Proceedings of the 1990 EPA/A&WMA
International Symposium
Measurement of Toxic and
Related Air Pollutants
Jointly sponsored by the
U.S. Environmental Protection Agency's
Atmospheric Research and Exposure Assessment Laboratory
and the
Air & Waste Management Association
Technical Program Chairmen
R.K.M Jayanty, Research Triangle Institute
Bruce W. Gay, Jr., U.S. Environmental Protection Agency
Raleigh, North Carolina
May, 1990
-------
Air & Waste Management Assoication Publication VIP-17
U.S. EPA Report Number EPA/600/9-90A)26
Measurement of Toxic and
Related Air Pollutants
Notice
Any policy issues discussed do not necessarily reflect the
views of the U.S. EPA and no official endorsement should
be inferred. The mention of trade names or commercial
products does not constitute endorsement or recommenda-
tion for use.
-------
Measurement of Toxic and
Related Air Pollutants
-------
-------
Table of Contents
Preface
xxi
Keynote Address - Air Toxics
David E. Price
Session I - Radon
Chairman: TJ. Maslany, U.S. EPA, Region m
Future Needs in Radon Monitoring
Prediction of Long-term Average Radon
Concentrations in Houses
Based on Short-term Measurements
An Analysis of the Parameters Influencing
Radon Variations in a House
Thomas J. Maslany
William E, Belanger
Albert Montague
Application of Building Diagnostic Techniques
to Mitigate Very High Radon Levels in a
Commercial Building on a Superfund Site D. Bruce Harris
An Automated, Semi-continuous System for
Measuring Indoor Radon Progeny Activity-
weighted Size Distributions, D : 0.5-500 nm
Chih-Shan Li
13
19
25
31
Session n - Atmospheric Chemistry and Fate of Toxic Pollutants
Chairman: B.W. Gay, Jr., U.S. EPA
The Atmospheric Stability of Polybrominated
Dibenzo-p-dioxins and Dibenzofurans
Occurrence and Vapor Particle Partitioning
of Heavy Organic Compounds in Brazzaville,
Congo
Heterogeneous Reaction of Nitrogen Oxides
on Sea Salt and Mineral Particles - A Single
Particle Approach
Reaction Products of Alternative
Chlorofluorcarbons - The
Hydrochlorofluorocarbons
Christopher C. Lutes 38
Barnabe Ngabe
Y. Mamane
Bruce W. Gay, Jr.
45
51
57
-------
Session in - Supercritical Fluid Extraction
Chairman: S.B. Hawthorne, University of North Dakota
Quantitative Supercritical Fluid
Extraction Coupled On-line to Capillary
Gas Chromatography for Environmental
Applications
The Role of SFE in Methods for the
Analysis of Toxic Compounds
Supercritical Fluid Extraction of
Polyurethane Foam Sorbents
Quantitative Extraction and Analysis of
Environmental Solids Using Supercritical
Fluid Extraction (SFE) and SFE-GC
Supercritical Fluid Extraction -
Applications in the AG Industry
Isolation of Polychlorinated Dibenzo-p-
dioxins and Dibenzofurans from Fly Ash
Samples Using Supercritical Fluid
Extraction
Extraction and Separation of Polar
Molecules Using Supercritical Fluids
with Subsequent Identification by
FT-IR Spectrometry
Quantitative Supercritical Fluid Extraction/
Supercritical Fluid Chromatography
from Aqueous Media
The Use of Supercritical Fluid Extraction
(SFE) in the Analysis of Environmental
Contaminants Isolated from Complex
Matrices
Indirect Supercritical Fluid Extraction
of Polychlorinated Dibenzo-p-dioxins
from Rainwater and Other Aqueous
Matrices
J.M. Levy
James H. Raymer
Mark S. Krieger
M.E. McNalfy
NickAlexandrou
Peter H Griffiths
J.L. Hedrick
Jennifer V. Smith
62
69
76
Steven B. Hawthorne 82
88
94
102
110
116
Michael J. Lawrence
123
VI
-------
Session IV - Acid Deposition - Mt Mitchell
Chairman: J. Vickery, U.S. EPA
Monitoring of Nitrogen Pollutants at
Mt. Mitchell, North Carolina
Parameterization of In-cloud Scavenging
of Sulfates and Nitrates in Acidic
Deposition Models
Measurements of Atmospheric Hydrogen
Peroxide in the Gas-phase and in
Cloudwater at Mt. Mitchell State Park, NC
Modeling Orographic Cloud Water Deposition
at Mt. Mitchell, NC: The Effect of Local
Topography
Remote Site Cloud Condensation Nuclei
"Fingerprints": Indicators of Air-mass
Transport and Mixing
Anuradha Murthy
N.-H. Lin
134
140
Candis S. Claiborn
Steven R. Chiswell
T.P. DeFelice
146
152
156
Session V - Determination of Polar Organic Compounds in Ambient Air
Chairman: J. D. Pleil, U.S. EPA
Vice Chairman: Don Adams, Consultant
Polar Volatile Organics: Overview of
Projects of the Monitoring Methods
Research Section - MRB/MRDD/AREAL,
U.S. EPA
Fourier Transform Infrared (FTIR)
Spectroscopy for Monitoring Polar and
Labile Airborne Gases and Vapors
Optimization of Analytical Procedure for
Characterizing Ambient Polar Volatile
Organic Compounds
Analysis of Polar Semivolatile Organic
Compounds by Liquid Chromatography/
Mass Spectrometry
Analysis of Volatile Organics in Air
Via Water Methods
Joachim D. Pleil
S.P. Levine
Sydney M. Gordon
Roberts. Whiton
J.H.M. Stephenson
167
176
183
189
194
vu
-------
Session VI - Delaware Superfond Innovative Technology
Evaluation (SITE) Study
Chairman: W.A. Mcdenny, U.S. EPA
Vice Chairman: E.N. Koglin, U.S. EPA
The 1989 Delaware Site Study William A. McClenny 200
Multi-adsorbent Preconcentration and
Gas Cinematographic Analysis of Air
Toxics with an Automated Collection/
Analytical System Albert J. Pollack 209
Passive Sampling Devices and Canisters:
Their Comparison in Measuring Air Toxics
During a Field Study J.L. Vams 219
Sector Sampling for Volatile Organics
Contributions to Ambient Air from
Industrial Sources Joachim D. Pleil 221
Remote FTIR Measurement of
Chemical Emissions Robert H. Kagann 244
A Comparison of FTIR Open Path
Ambient Data with Method TO-14
Canister Data George M. Russwurm 248
Short-term Sequential Canister-based
Sampling near Superfund Sites Karen D. Oliver 254
Spatial and Temporal Correlation of
Wind Direction and Speed over a
Small Region George M. Russwurm 260
Session Vn - Mobile Sources Emissions Characterization
Chairman: K.T. Knapp, U.S. EPA
Speciation of Organic Components of
Mobile Source Emissions Kenneth Knapp 266
Comparison of Data from a Fourier Transform
Infrared Automotive Emissions Sampling and
Analysis System with those Obtained from
Conventional Automotive Emissions Analytical
Instrumentation Alexander O. McArver 274
vm
-------
Remote Sensing Measurements of Carbon
Monoxide Emissions from On-road Vehicles
The Effect of Oxygenated Fuels on
Automobile Emission Reactivity
FTIR: Fundamentals and Applications
in the Analysis of Dilute Vehicle Exhaust
Emissions from a Flexible-fueled Vehicle
Robert D. Stephens 285
Charles D. Burton 291
CA Gierczak 300
Peter A. Gabele 314
Session VHI - Superfund Site Air Monitoring
Chairman: R.V. Crume, Midwest Research Institute
A Retrospective Analysis of a Baseline
Air Pathway Assessment at a Pre-remedial
Superfund Site
Ambient Air Monitoring of a SARA Title III
Facility Using the TAGA 6000E MS/MS
Development of a "Statement of Work (SOW)
for the Analysis of Air Toxics at Superfund
Sites" as Part of the Contract Laboratory
Program (CLP)
A Program to Validate Air Monitoring
Methods for the Superfund Program
Reducing Uncertainties in Inhalation
Risk Assessments
Design and Implementation of an Air
Monitoring Program at a Superfund Site
Evaluation of Methods for Detecting
Dimethyl Mercury in Ambient Air at
a Superfund Site
Hazardous Waste TSDF PM10 Emissions
A Methodology for Identifying and
Ranking Toxic Air Pollutants at
Superfund Sites
Richard W. Tripp
David B. Mickunas
Russell McCallister
W.J. Mitchell
Sally A. Campbell
Chris G. Harrod
Thomas H. Pritchett
C. Cowherd, Jr.
Emile I. Boulos
320
328
336
349
353
359
371
378
385
IX
-------
Air Toxics Considerations at an Active
Firefighter Training Facility
Michael]. Barboza
395
Traverse-monitoring for Superfund Sites
The U.S. and California Clean Air Acts:
Implications for the Future
Superfund Community Relations Plans:
Taking Advantage of the Requirement
A.J. Cimorelli 401
John T. Ronan III 407
Deborah CZ Hirsch 412
Session IX - Exposure Assessment
Chairman: KD. Pellizzari, Research Triangle Institute
The Evaluation of Breath VOCs Resulting
from Human Exposure to Microenvironments
Automated Specific Analysis of Breath
with a Portable Monitor
Routes of Chloroform Exposure from
Showering with Chlorinated Water
Biomarkers and Pharmacokinetics
Review of the Particle TEAM 9 Home
Field Study
Particle Total Exposure Assessment
Methodology (PTEAM): Statistical
Analysis of Spatial and Temporal Patterns
Measurements of Ozone Exposures
A Personal Exposure Monitor for CO
and Ozone
A Health & Welfare Canada Program
to Develop Personal Exposure Monitors
for Airborne Organics at UG/M3
New Dermal Exposure Sampling Technique
James H. Raymer
William R Penrose
Clifford P. Weisel
Jerry N. Blancato
RW, Werner
C. Andrew Clayton
P. Koutralds
William R Penrose
418
425
436
442
452
461
468
475
Rein Otson
J.P. Hsu
483
489
-------
A Low Cost Sampler for Monitoring
Worker Exposure to Herbicide
Residues in Forest Fire Smoke
Development of a Test Atmosphere
Generation Facility for Particle
Bound Organic Compounds
Charles K McMahon 498
Philip Fellin
Session X - Chemometrics and Environmental Data Analysis
Chairman: WJ. Dunn HI, University of Illinois
Vice Chairman: D.R. Scott, U.S. EPA
506
Spectroscopic Identification of Organic
Compounds by Library Searching:
Methods, Potentialities, Limitations
The Use of Principal Component Analysis
to Display PAH Concentration Patterns in
Indoor Air
Statistical Modeling of Ambient Air Toxics
Impacts during Remedial Investigations at
a Landfill Site
Comparison of the Source Locations and
Their Seasonal Patterns for Sulfur Species
in Precipitation and Ambient Particles in
Ontario, Canada
Monitoring Toxic VOCs in Urban Air
in Illinois
J.T. Clerc
Sotnenath Mitra
Steven C. Mauch
Yousheng Zeng
Clyde W. Sweet
512
518
523
529
536
Session XI - Nicotine in Environmental Tobacco Smoke
Chairman: D J. Eatough, Brigham Young University
Cabin Air Quality: Cotinine as a
Biomarker of Environmental Tobacco
Smoke in Commercial Aircraft
Problems with the Use of Nicotine
as a Predictive Environmental
Tobacco Smoke Marker
DelbertJ. Eatough 542
P.R. Nelson 550
XI
-------
Polycyclic Aromatic Compound
Concentrations in Residential Air
Associated with Cigarette Smoking
and Gas or Electric Heating and
Cooking Systems Jane C. Chuang 556
Comparison of Area and Personal Sampling
Methods for Determining Nicotine in
Environmental Tobacco Smoke William E. Grouse 562
The Impact of Cigarette Smoking on
Indoor Aerosol Mass and Elemental
Concentrations B.P. Leaderer 567
Session XII - Source Monitoring
Chairman: TJ. Logan, U.S. EPA
Chromium Sampling Method Frank R. Clay 516
Development of a Source Test
Method for Hexavalent Chromium J.E. Knoll 579
Development and Field Validation of
a Sampling and Analytical Method for
Airborne Hexavalent Chromium P. Sheehan 595
Determination of Average Ambient
PCDDs/PCDFs Concentrations in the
Vicinity of Pre-operational Resource
Recovery Facilities in Connecticut Bruce E. Maisel 602
Development and Evaluation of Methods
to Determine Pathogens in Air Emissions
from Medical Waste Incinerators RE. Segall 611
Priority Pollutant Metals in Respirable
and Inhalable (PM10) Particles James E. Houck 617
Session Xm - Effects of Air Toxics on Plants and Ecosystems
Chairman: W.W. Heck, North Carolina State University
Vice Chairman: T J. Moser, NSI Technology Services Corporation
Atmospheric Transport of Toxic Chemicals
and Their Potential Impacts on Terrestrial
Vegetation: An Overview Thomas J. Moser 623
xu
-------
Fluoride Phytotoxicity: Past, Present
and Future
Biogeochemistry of Trace Metals at the
Hubbard Brook Experimental Forest, NH
Use of the Phytotox Database to Estimate
the Influence of Herbicide Drift on
Natural Habitats in Agroecosystems
Detecting Effects of Air Toxics
Using Wildlife
Effects of Air Pollutants on Cold-desert
Cyanobacterial-lichen Crusts and Rock
Lichens: Chlorophyll Degradation,
Electrolyte Leakage and Nitrogenase
Activity
David C. MacLean
Charles T. Driscoll
James E. Nellessen
R. Kent Schreiber
630
637
649
655
Jayne Belnap
661
Session XIV - Measurement of Volatile Organic Pollutants
Ambient Air Long Path Spectroscopy and Other Methods
Chairman: RJC Stevens, U.S. EPA
Evaluation of a Differential Optical
Absorption Spectrometer as an Air
Quality Monitor
Environmental Control Using Long
Path Measurements
Comparison of Long Path FT-IR Data
to Whole Air Canister Data from a
Controlled Upwind Point Source
Analysis of Volatile Organic Compounds
with an Ion Trap Mass Spectrometer
Evaluation of a Non Cryogenic Automated
Multitube Thermal Desorption System
for the Analysis of Air Toxics
Stability/Instability of Gas Mixtures
Containing 1,3-Butadiene in Treated
Aluminum Gas Cylinders
R.K. Stevens
Ronald Karlsson
ML. Spartz
David W. Berberich
William F. Boehler
668
675
685
693
699
George C. Rhoderick 709
xm
-------
Toxic Organic Gas Standards in
High Pressure Cylinders
Humidified Canister Stability of
Selected VOCs
Evaluation of a Continuous Sampling
and Analysis System for Volatile
Organic Compounds
Canister Based Sampling Systems -
A Performance Evaluation
Mobile Ambient Air Sampling and
Analysis Experience of the Texas
Air Control Board
The Establishment and Operation of
an NMOC and Aldehyde Monitoring
Program - Experience of a State Agency
Discrepancies in Ambient Non-methane
Hydrocarbon Measurements among
Various Methods
Measurement of Acid Deposition
Components in Southern Commercial Forests
PM10 Hi-vol Collection and Quantitation
of Semi-volatile Methoxylated Phenols as
Tracers of Wood Smoke Pollution in Urban Air
Stephen B. Miller
Rita M. Harrell
James M. Hazlett
Dave-Paul Dayton
James L. Lindgren
Julian D. Chazin
Joel Craig
Robert L, Sutton
Steven B. Hawthorne
Session XV - General
Chairman' P.K. Hopke, Claikson University
Solid Sorbent Air Sampler for the
Characterization of Contaminants
in Spacecraft Atmospheres
Sulfur on Surfaces of Atmospheric
Minerals and Spores
Development of Atmosphere Generation,
Monitoring, and Clean-up Systems for
Acid Aerosol Animal Exposures
Thomas F. Limero
Yaacov Mamane
MA Higuchi
718
726
731
740
747
753
761
767
774
780
789
796
xiv
-------
The Atmospheric Transformation of
Automobile Emissions and Measurement
of the Formation of Mutagenic Products
Validation of Methodology for Direct
Bioassay of Atmospheric Volatile
Organic Compounds
Isolation of Mutagens in Ambient Air
Paniculate Extracts Using
Bioassay-directed Fractionation
Applicability of GC/MS Instrumentation
for the Analysis of Undried Air Toxic Samples
Laboratory Evaluation of Microsensor
Technology Gas Analyzer
A Real-time Monitor for Chlorinated
Organics in Water
Performance Optimization of
Photovac 10S70 Portable Gas
Chromatograph
Computer Software for Gas
Chromtography in the Field
Variability in Elution Time Data from
Microchip Gas Chromatographs:
Ramifications for Sample
Component Identification
T.E. Kleindienst
CM. Sparacino
Daniel J. Thompson
L.D. Ogle
Joseph S.C. Chang
Joseph R. Stetter
HE. Berkley
C.F. Steele
802
808
818
824
830
836
849
855
K.R. Carney
861
Session XVI - Air Pollution Dispersion Modeling
Chairman: SJ*. Arya, North Carolina State University
Vice Chairman: S.T. Rao, New York State Department
of Environmental Conservation
A Fractional (Fractal) Brownian Motion
Model of Atmospheric Diffusion
Numerical Simulations of the Mountain
Iron Tracer Data
Gifford
Tetsuji Yamada
868
875
xv
-------
The Inclusion of Pollutant Removal
Processes in Urban Air Quality Models K, Shankar Rao 881
The Influence of Measurement
Uncertainties on the Performance
of Air Pollution Dispersion Models Steven R. Hanna 887
Flow and Dispersion of Pollutants
within Two-dimensional Valleys William H. Snyder 893
Wind-tunnel Modeling of the Dispersion
of Odorants and Toxic Fumes about
Hospitals and Health Centers Robert N, Meroney 899
Estimating Exposures Downwind of
Isolated Buildings John S. Irwin 905
Concentration Fluctuations of a Toxic
Material Downwind of a Building William B. Peterson 911
Session XVQ - Measurement of Hazardous Waste Emissions
Chairman: T.T. Shea, New York State Department
of Environmental Conservation
Hazardous Waste Incinerator
Evaluation Program Shekar Viswanathan 917
Remote Optical Sensing of VOCs:
Application to Superfund Activities Timothy R. Minnich 928
Measurements of Emissions at a
Chemical Waste Water Treatment
Site with an Open Path Remote
Fourier Transform Interferometer Orman A. Simpson 937
Application of a Hene Laser to
Hydrocarbon Leak Detection
over an Oil Field Joel D, Cline 940
Evaluation of Community Exposure
to Airborne SARA Title III
Section 313 Chemicals Emitted
from Petroleum Refineries C. Hemdon Williams 948
xvi
-------
Development of a Technical Approach
for Determining the Volatilization Rate
of Hydrocarbons from a Landfill Using
Both Modeling and Direct Emission
Measurement Techniques C.E. Schmidt 955
Temperature-corrected Dispersion
Modeling of Volatile Emissions
from Hazardous Waste Sites Richard J. Hardy 961
Session XVHI - Measurement and Data Analysis
of Indoor Toxic Air Contaminants
Chairman: D.F. Nauglc, Research Triangle Institute
Vice Chairman: H. Oskaynuck, Harvard University
The Effect of Wood Finishing Products
on Indoor Air Quality Bruce A. Tichenor 968
Indoor Air Testing for Low-level
Volatile Organics: A Site-specific
Technical Approach Amy S. Johnson 974
Volatile Organic Compounds in
the Atmosphere of a Newly
Constructed Residence Barbara B. Kebbekus 981
Estimating the Cancer Risk from
Multi-route Exposure to Chloroform
from Chlorinated Water Wan K. Jo 988
Comparison of Indoor and Outdoor
Aldehyde Concentrations during
the 1ACP Roanoke, Virginia
Residential Study Roy Zweidinger 994
Risk Characterization of Noncancer
Health Effects Associated with
Indoor Air Pollutants Robert G, Hetes 1000
Assessment of the Health Risks Associated
with Indoor Benzene Vapor Emitted from
Building Foundation Soil: A Case Study Tibor T. Sarlos 1008
xvu
-------
Session XIX - Environmental Quality Assurance
Chairman: F. Smith, Research Triangle Institute
A Review of the Accuracy and Precision
of the Toxic Air Contaminant
Monitoring Program of the
California Air Resources Board Catherine Dunwoody 1015
Uncertainty Estimates for the
NAPAP Material Exposure
Monitoring Network Data P. Michael Barlow 1027
Quality Assurance for Contract Source
Testing in the South Coast Air Quality
Management District (California) Gary Dixon 1033
Lab Certification Versus In-home
Emissions Performance of Advanced
Technology Woodstoves Stockton G. Bamett 1039
Analysis of Air Pollutant Concentrations
below the Detection Limit 5. Trivikrama Rao 1044
SessionXX - General
Chairman: W.F. Gutknecht, Research Triangle Institute
Use of the Surface Isolation Flux
Chamber to Assess Fugitive
Emissions from a Fixed-roof on
an Oil-water Separator Facility C.E, Schmidt 1050
An Attempt to Measure the Air Toxics
Impacts of the Greater Detroit
Resource Recovery Facility James C. Seme 1055
Emission of Ozone and Dust from
Laserprinters. Presentation of a
New Emission Source Test Method Torben Eggert 1064
Determination of Light and Heavy
Hydrocarbons and Non-methane
Organic Compounds (NMOC) in
Ambient Air Using a Combination
of Method TO-12 and Method TO-14 John K Hawkins 1073
XVlll
-------
The Prediction of Atmospheric Stability
and the Dispersion of Emissions from
Superfund Sites C.C.Allen 1078
Subject Index 1084
Author Index 1094
xix
-------
Conference Committees
General Conference Chairmen
Gary Foley, U.S. EPA
Martin Rivers/ Air & Waste Management Association
Technical Program Chairmen
Bruce Gay, U.S. EPA
R.K.M. jayanty, Research Triangle Institute
Research Triangle Park Chapter
William R. Barnard, Acting Chairman
F. Vandiver Bradow, Secretary
M.B. Shards, Treasurer
South Atlantic States Section
Michael F. Tanchuk, Chairman
Robert Norton, Vice Chairman
Susan Wieman, Secretary
Fabiola Sepulveda, Treasurer
Kenneth Weiss, Membership
Ambient Monitoring Committee (EM-3)
Douglas A. Lane, Chairman
Thompson G. Pace, First Vice Chairman
R.K.M. Jayanty, Second Vice Chairman
Fred B. Dowling, Secretary
Source Monitoring Committee (EM-4)
Mark S. Siegler, Chairman
James Jahnke, Vice Chairman
Ron Jernigan, Secretary
Toxic Air Pollutants Intercommittee Task Force (ITF-2)
David Patrick, Chairman
Pat Bartholomew, Vice Chairman
Jitendra Shah, Secretary
Headquarters Support
Steve Stasko, Technical Program
Sharon DeAndrea, Meeting Coordination
Len Mafrica, Exhibition
James Morton, Production
Linda Despot, Production
Dan Denne, Registration
-------
Preface
The 1990 EPA/AWMA International Symposium Measurement of Toxic and Related
Air Pollutants sponsored by the Atmospheric Research and Exposure Assessment
Laboratory of the U. S. Environmental Protection Agency at Research Triangle
Park, North Carolina and co-sponsored by the Air and Waste Management Association
of Pittsburgh, Pennsylvania was held May 1-4, 1990, at the Mission Valley Inn
at Raleigh, North Carolina. The technical program consisted of 178 presentations
held in 20 separate sessions. The sessions focused on recent advances in the
measurement and monitoring of toxic and related pollutants found in the ambient,
indoor, and source emissions atmospheres. New sessions were added this year to
the symposium included: Radon, Atmospheric Chemistry and Fate of Toxic
Pollutants, Super-critical Fluid Extraction, Determination of Polar Organic
Compounds in Ambient Air, Mobile Sources Emissions Characterization, Effects of
Air Toxics on Plants, and Air Pollution Dispersion Modelling. Session on
Measurements of Volatile Organic Pollutants - Ambient Air focused on Long Path
Spectroscopy and other methods of monitoring toxic pollutants. During the
symposium over 70 exhibitors had booths displaying a wide range of pollution
monitoring instrumentation and consulting services. Over 850 attendees from the
U.S. and other countries attended the symposium. Contained in this volume are
the papers presented during the symposium. The keynote address of Congressman
David Price, Fourth District of North Carolina, presented by the Congressman's
Administrative Assistant, Mr. Gene Conti, is also included.
Air pollution measurement and monitoring research are designed to support
regulatory actions by developing an in*depth understanding of the nature and
processes that impact health and the ecology, to provide innovative means of
monitoring compliance with regulations and to evaluate efforts through the
monitoring of long-term trends, and to anticipate potential environmental
problems. The U.S. Environmental Protection Agency's Atmospheric Research and
Exposure Assessment Laboratory at Research Triangle Park, North Carolina is
responsible for research and development of new methods, techniques, and systems
for detection, identification, and characterization of pollutants in emission
sources and in indoor and ambient environments; for the implementation of a
national quality assurance program for air pollution measurement systems; and
supplying of technical support to Agency regulatory programs on local, regional,
and global scale.
This was the 10th consecutive year the symposium was held and the 5th year
of its co-sponsorship with AWMA. The objective of the symposium is to provide
a. forum for the exchange of ideas on recent advances for the reliable and
s.ccurate measurement and monitoring of toxic and related air pollutants in
indoor, ambient, and source atmospheres. The ever growing number of
presentations and attendees to this symposium represents advancements in the
current measurement and monitoring capabilities.
Bruce W. Gay, Jr., (EPA)
R. K. M. Jayanty (RTI)
Technical Program Chairmen
XXI
-------
-------
ENVIRONMENTAL PROTECTION AGENCY INTERNATIONAL SYMPOSIUM
KEYNOTE ADDRESS - AIR TOXICS
REPRESENTATIVE DAVID E. PRICE
MAY 1, 1990
I'm honored to be able to present these remarks as the keynote
to your conference today. I'd like to commend Drs. jayanty,
Gay, and Foley, and Mr. Kueser, as well as the Environmental
Protection Agency and the Air and Waste Management Association
for convening this annual symposium on toxic and related air
pollutants.
We are proud to have EPA's air research facility headquartered
in North Carolina and proud of our EPA scientists and the
private sector companies which support EPA's efforts here, many
of whom are constituents of this district. In my relatively
short time in Congress, I have had the pleasure of working with
EPA on several funding projects, including appropriations for a
new EPA affiliated research lab in Chapel Hill and the hoped
for acquisition of a supercomputer -- and, longer term, on a
major new research facility in the Research Triangle Park. I
value the close working relationship that we've developed and
anticipate its continuation.
In my brief remarks this morning, I would like to provide a
broad context for the environmental issues you will be
discussing this week and to give you an overview of some
specific legislative initiatives that will have an impact on
your work in the years to come.
Around the globe, we are confronting serious environmental
problems. Solving these problems will require a sustained
research effort that can provide policy makers with reliable
information. Over the last twenty years, EPA's research budget
has declined more than 20% in real terms and more than 500
scientists have been cut from the agency's payroll. Yet during
the same period, EPA's regulatory responsibilities have vastly
increased through the enactment of major new statutes.
-------
This pattern of neglect must be broken. You may be aware that
the Congress is currently working on legislation to elevate the
status of the Environmental Protection Agency. On March 28 of
this year, the House passed legislation to redesignate the EPA
as a cabinet level department. Similar legislation is awaiting
floor action in the Senate. We believe a cabinet-level EPA
will help serve notice that we have made the environment an
important national priority -- consistent with the status that
environmental protection has reached in many countries around
the world.
I was pleased to see President Bush submit a budget to Congress
this year that includes an increase in federal air quality
programs. The fiscal year 1991 budget request for air quality
programs is $390 million -- a $100 million or a 35% increase
over last year's funding level. Most of the increase would be
targeted to State assistance activities to implement the
provisions of a reauthorized and stronger Clean Air Act, but
$17 million will support more aggressive air research programs,
including a $9 million increase to address the specific issues
of air toxics and non-attainment of National Ambient Air
Quality Standards.
Funding for EPA, and virtually every other program within the
federal budget, has been constrained in recent years, but there
is still a tremendous reservoir of confidence and support in
Congress for the work of EPA. This is not to say EPA can or
should escape the scrutiny or the discipline that our current
budgetary crisis requires. But a solid case can be made for
research expenditures as a critical national investment, one
that pays rich dividends in the health and well-being and
security of our people and in the vitality of our research and
educational institutions.
As we stand on the brink of a new year and a new decade, we
look ahead to assess the critical issues of the 1990's. There
are few issues as important -- to us and to our children -- as
the protection of our global resources. A recent survey
indicated that 90% of Americans say that "taking stronger
actions to clean up the nation's air and water is a top
priority for government and business leaders."
Even more dramatic evidence comes from data analyzing the
willingness of the American public to pay for more
environmental protection. In 1981, the public split about
evenly on this question. By 1989, the public indicated a
willingness to pay by a margin of about 4 to 1. There will be,
of course, continuing discussion about how much people will
have to pay but there is no mistaking the strong public
commitment to a better environment.
-------
The "greening of America" is taking place in Washington as
well. Congress is paying a great deal of attention to
environmental issues now, and — perhaps more importantly --
we're working with an administration that recognizes the
problem as well. We still have our differences with the White
House -- witness several key votes on oil spill liability --
but the prospects for cooperation are far better than they were
during the Reagan years.
With this broad background in mind, let me review some of the
legislative action in the current session of the Congress.
Clean Air legislation is the number one environmental issue
before the Congress this year. Clean air legislation was put
off by President Reagan and did not get addressed by committees
of the Congress for the last decade because of an ongoing
deadlock between those who wanted stricter pollution controls
and those who wanted more lenient ones. However, the expiration
of compliance deadlines under the 1977 amendments and
heightened public concern about health and the environment
almost certainly will result, in the enactment of a strengthened
version of the Clean Air Act this year. Earlier this spring,
just prior to the Easter recess, clean air legislation took
giant strides toward enactment, clearing the House Energy and
Commerce Committee on a 42 to 1 vote and passing the Senate by
a wide margin, 89 to 11. I expect floor action on the House
version later this month.
In the meantime, the House Science, Space and Technology
Committee, of which I am a member, has approved legislation to
expand air pollution research under the Clean Air Act.
Provisions in this legislation include expansion of research
and development programs on the health effects of air
pollution, as well as improving monitoring and control
methods. This legislation will be offered as a floor amendment
to the House version of the Clean Air Act. Such legislation
should mean additional R&D dollars for North Carolina --
particularly the EPA lab in the Triangle. As you know, much of
the nation's cutting-edge environmental research is being done
here in the Triangle, and I will be pushing to commit adequate
resources to these efforts. To do a better job of implementing
environmental controls and environmental health policy, we need
a solid base of scientific knowledge.
We are also seeing an increased emphasis on control of air
toxics included in the House and Senate versions of the Clean
Air Act. This effort will require stepped up research on your
part if we are to rationally control these compounds.
It is extremely important to establish tougher standards for
toxic air emissions. One major flaw of the 1970 Clean Air Act
was the system established to regulate hazardous air
-------
pollutants. In the nearly two decades since the Clean Air Act
was enacted, EPA has only regulated seven pollutants. This
pace is too slow — uncontrolled air toxics pose substantial
health risks and we must improve this regulatory process. I
think we were all alarmed by the announcement EPA made in April
of 1989 showing that manufacturing industries released 2.7
billion pounds of toxic air pollutants in 1987. Most
observers, including EPA officials, expressed surprise at the
volume of toxic releases and called them unacceptably high.
The essential questions that must be answered in this debate on
air toxics revolve around two distinct issues. The first
involves the distinction between point sources of pollution --
i.e., specific, large sources of toxics — and area sources,
which include the exhaust from our cars, the discharges from
small businesses, and similar sources. The second involves the
type of remedy we seek from these sources — i.e., whether we
impose specific technology-based solutions or whether we impose
health-risk based solutions.
The President's clean air proposal required EPA to regulate
only 50% of the major sources of hazardous air pollutants.
During House committee consideration, an amendment was adopted
that would require EPA to regulate all categories of major
sources. The only exception to this rule, in the case of
sources with carcinogenic emissions, is that EPA may exempt a
source category from regulation if it determines that no source
in the category presents a lifetime risk of cancer greater that
1 in 1 million to any person.
Data presented to Congress indicate that toxic emissions from
area sources are collectively responsible for as many cancer
cases as are emissions from point sources. Under the
President's bill, regulation of these area sources was entirely
discretionary. The amended House bill now requires EPA to
regulate 90% of the area source emissions of each hazardous air
pollutant. EPA may elect to establish controls based on
"general available control technology" in lieu of the more
stringent controls based on "maximum achievable control
technology" that would apply to major sources.
The technology based standards required for major and area
sources under the amendment may not eliminate all health risks
from toxic emissions. To address the "residual risks" that may
remain, the amendment requires EPA to report to Congress on
legislative alternatives within ten years. If none are enacted
at that point in time, the law would revert to the more
stringent but thus far unenforced health-based standards of the
1977 Act. Although the Senate bill includes a second-stage
health-based standard, the likely final product will be closer
to the House bill language.
-------
Another area of concern to members of Congress is the issue of
indoor air pollution. People spend about 90% of their time
indoors, and air contaminants commonly found in the home and
workplace are known to cause a wide range of health effects.
Public concern about the threat of indoor air pollution to
human health and worker productivity has increased dramatically
in the past two decades. Media reports about the "sick
building" and "sick home" phenomena are now common.
Scientific information about the sources and effects of indoor
air pollution is increasing steadily. At the same time,
definitive conlusions about the health effects of indoor air
pollution remain controversial. It is difficult to associate
definitively a given response with a single air pollutant, in
or out of doors, because many factors in the human environment
can contribute to particular health conditions.
These issues were pursued during hearings before the Science,
Space and Technology Subcommittee on Natural Resources,
Agricultural Research and the Environment, last year. In
response to these hearings, our subcommittee recently marked-up
and reported H.R. 1530, the Indoor Air Quality Act. This
legislation would strengthen EPA's indoor air research program;
would promote development of health advisories to indicate
levels at which identified contaminants would have no adverse
health effect; would focus the use of existing regulatory
authority to implement a national response plan based on the
health advisories; would provide grants to the states to assist
them in developing basic management strategies and assessments;
would establish an EPA office for indoor air activities and a
coordinating interagency Council on Indoor Air Quality; and
would give EPA the lead responsibility for developing a federal
response to indoor air quality.
A companion bill has been introduced in the Senate by Majority
Leader George Mitchell and is pending consideration at the full
committee level. I am hopeful we can enact this necessary
legislation this year.
A large part of the indoor air pollution debate focuses on
radon. EPA estimates that radon is responsible for 5,000 to
20,000 case of lung cancer each year. If these figures are
accurate, radon may be the nation's leading radiation problem
and the second largest cause of lung cancer. Recent studies,
however, have challenged these health estimates and provide a
stronger basis for continued research on this issue.
Federal statutory authority for dealing with the radon problem
is not as clearly defined in existing law as it should be. The
Superfund Reauthorization Act authorized an EPA radon gas and
indoor air quality research program, including a national
-------
assessment of radon gas and a demonstration program to test
methods of reducing or eliminating radon. To support its Radon
Action Program, EPA received $12.4 million in fiscal year
1990. The President's budget requests $13.7 million for
FY1991, an increase of 11% from FY1990.
To address the increasing concern about the potential effects
of radon, in the first session of the 101st Congress, 11 pieces
of legislation related to radon issues were introduced. None
have been enacted thus far. H.R. 1530, the Indoor Air Quality
Act would address radon through a comprehensive indoor air
quality approach. Several major policy issues remain to be
addressed: which federal agency should deal with radon
regulations?; is the EPA-recommended radon action level
protective enough of human health?; and does scientific
information support public concerns and the need for EPA
guidelines? I would like to see these issues addressed by the
Congress in the near future and encourage those of you working
in this area to move forward aggressively with your research.
As you can see, our country is demonstrating increased
awareness and concern about environmental quality; the public
response to Earth Day leaves no doubt about that. And this
concern is being translated into policy initiatives, shaking
loose once-stalled items like oil spill liability and clean air
and stimulating discussion and debate on the environmental
agenda of the future — matters ranging from rain forest
destruction to ozone depletion to global warming to alternative
energy sources. No one is more aware than you how critical a
part research will play in determining the extent of the
dangers and what solutions we can feasibly pursue. We're also
aware that research -- an honest facing of what our best
scientists tell us — will not always or even usually be
comforting. The need for research must not be a
rationalization for inaction. On the contrary, as you are in a
better position than most to know, for this country and the
world community to make good on the lofty pronouncements of
Earth Day, it is going to require strong leadership,
considerable sacrifice, major commitments of resources, and
much determination in the years ahead. What is going on here
at EPA and other research facilities gives me great hope that
we can rise to the challenge, a challenge that should transcend
everyday politics.
Through these research efforts, we're investing in our future,
in the healthy future of our children and the well-being of
generations to come. I want to commend each of you for the
part you are already playing in that and to thank you once
again for the opportunity to participate in this symposium.
-------
FUTURE NEEDS IN RADON MONITORING
Introductory Remarks on Radon
Thomas J. Maslany
U. S. Environmental Protection Agency
Region III
841 Chestnut Street
Philadelphia, PA 19107
Prepared by W.Belanger, L. Felleisen, T. Maslany
Indoor radon monitoring is currently dominated by charcoal
adsorption devices for short term monitoring and alpha track
devices for long term monitoring. Several other devices are also
available. The simplest is an electrical device which allows radon
to be read with a special voltmeter. The most complicated are
delicate instruments requiring a skilled operator. This paper
discusses the radon measurements which will be needed in the future
and develops the measurement device characteristics necessary to
fulfill these future needs. The available radon monitoring methods
are evaluated with respect to their ability to satisfy this set of
future needs.
-------
Introduction
The Environmental Protection Agency believes that a large
fraction of the U. S. population is exposed to unacceptable risk
from radon and has advised everyone with significant exposure
potential to test their home. Unfortunately, only a small
fraction of homes have been tested, and many people believe the
complicated nature of the radon testing process is partially to
blame for this poor response. EPA suggests a two step testing
process; first a screening test is used to decide whether a problem
might exist in the house, then a long-term follow up test is used
to measure the radon exposure of the residents. Few people seem
willing to go through this testing process, which usually takes
more than a year. This testing scheme also is not well suited to
radon testing during real estate transactions. A simpler and
shorter testing scheme will probably be needed for residential
testing. Some means will be needed to relate this short term
testing to the average exposure of the residents. The main change
in radon monitoring foreseen by this author is an increased
reliance on short term tests.
Schools and offices pose special testing problems because they
are unoccupied a large fraction of the time. Heating and air
conditioning systems are frequently shut down or cut back during
this unoccupied time, which can produce substantially different
radon concentrations. This must be accounted for if radon
exposures in these buildings are to be accurately assessed. It has
also been observed that the working level ratio is significantly
lower in large buildings than in houses. Where 4 pCi/1 yields
about 0.02 WL in houses, it may yield only .004 WL in large
buildings.
Most radon testing today is performed using charcoal
adsorption or alpha track devices, though many other monitoring
methods are available. The charcoal devices are limited to short
monitoring periods and are well suited to the screening
measurements currently recommended by EPA. Alpha track devices
are well suited to the long term follow up measurements. The
remainder of this paper is devoted to a discussion of future
monitoring needs. The suitability of today's monitoring methods
to these needs will be discussed.
The "Ideal Radon Monitor"
The ideal radon monitor may not be possible with today's
technology. It is also possible that there may be conflicting
requirements for different applications. No one radon monitor will
necessarily be ideal for all purposes. The attributes of the
"ideal" radon monitor are given below.
Accuracy
The radon monitor should accurately measure the radon.
Accuracy limits of plus or minus 25 percent have been used by EPA
-------
for the Radon Monitoring Proficiency (RMP) program. This level of
accuracy does not reflect the limits of radon monitoring
technology. It was chosen to reflect the degree of certainty of
the radon risk estimates. It also accounts for the fact that radon
itself is variable and any measurement can only reflect the radon
at the time the measurement is made. While most methods can yield
better accuracies with careful procedures, a few
laboratories are have difficulty in meeting the 25 percent error
limit in double blind tests near 4 pCi/1.
Minimum Measurable Concentration
EPA has set an action level of 4 pCi/1 as an indoor average
radon concentration. The Indoor Radon Abatement Act sets a
national goal that indoor radon be as low as outdoors, and
requires EPA to review its action level. In response, the action
level may be revised in the near future. No decision has been made
on future level at the time of this writing. It is possible that
a radon concentration different from 4 pCi/1 could be chosen. It
Is also possible that it would be based on something other than an
annual average concentration in the living areas of the house.
Different action levels may be chosen for different applications
such as schools. It may be necessary to accurately measure radon
concentrations below 4 pCi/1 to reach this goal.
Sampling Time
While it is desirable to be able to measure radon over a
period of a year, it appears that most radon mitigation decisions
are being made based on much shorter tests. A full year of testing
is prohibitive in a real estate transaction. Most people who
monitor out of concern for their health base their actions on short
term tests. An analysis of radon variability has shown that the
minimum sampling time for a reliable test is one day, and that
errors decrease as sampling time is increased to two days. There
is a small additional improvement for sampling times of a week.
Beyond this time, little additional accuracy is gained until
sampling times become impractical to serve the purpose of the test.
An ideal sampling time of two days to a week is selected for use
here. The radon monitor should be able to measure within 25
percent at 4 pCi/1 over a 2 to 7 day period. Sample time should be
in multiples of one day to avoid diurnal variations in radon.
True Average
The radon monitor should provide a true average radon
concentration over the sampling time. If, for example, the device
is heavily biased toward the last day of the sampling period, then
this is effectively a one-day sample and is subject to the
variability associated with one-day measurements.
-------
Storage Capability
The device should not need immediate analysis. If a large
number of samplers are returned to a laboratory for analysis at
one time, the lab must be able to process them within the time that
a useful measurement can be made. Devices which store radon for
future counting are subject to radon decay. Half the radon
collected is lost every 3.8 days. A delay of a month reduces the
radon available for counting by a factor of 250.
Sensitivity to Environment
The measurement method should not be sensitive to normal
variations in temperature or humidity or other common
environmental parameters. Ideally, there should be quantitative
limits for this sensitivity, but for purposes of this paper, we
will require that common environmental parameters not cause errors
beyond the allowable accuracy limits at the minimum
measurable concentration.
Calibration of Active Devices
The device should remain in calibration for at least six
months and preferably as long as a year. Recalibration should be
a relatively easy. It should be possible to make a performance
check without transporting the device to a radon chamber.
Operator Skill
The required level of operator skill should be kept to a
minimum. College level mathematics should not be required in order
to operate the device.
Cost
Cost should be kept to a minimum. Current costs for
charcoal and alpha track tests ranges from about $10 to $30. The
cost per test should not be greater than the high end of this
range.
Proper Pollutant
There are two basic measurements, radon and radon decay
product. The decay products will not occur in a constant ratio to
the radon concentration. The ratio will depending on the age of
the air being sampled, the amount of air movement and the presence
of any filtration. In addition, some decay products will be
attached to dust particles, while others will be
unattached. While decay products are responsible for the bulk of
radon's health effects, the variables in decay product
measurement make it difficult to do a health assessment based on
a typical short term decay product measurement. This is because
of the variability of decay product concentrations. On the other
10
-------
hand, radon is the pollutant entering the building, so if one
wishes to know the effectiveness of a radon mitigation, radon is
the appropriate parameter to measure. In choosing whether to
measure radon or its decay products, it is important to have a
clear understanding of the purpose of the measurement and to
control those factors which could influence the measurement.
Intermittent Sampling
Buildings with intermittent occupancy, such as schools and
office buildings, may best be sampled only during hours of
occupancy, especially if ventilating systems are also operated
intermittently. This could result in significantly different radon
concentrations when the building is not occupied. The measurement
should be able to differentiate between these periods in some .;ay,
or its usefulness may be limited in this context.
Anti-tampering
As radon becomes a factor in real estate transfers, it might
be expected that radon testing will become more common in this
situation. In this circumstance, the party selling the house (and
usually occupying it at the time of the measurement) has an
economic stake in the outcome of the test. There is an incentive
to cheat. Radon tests need to be structured to prevent
tampering, either by the design of the test device or through some
external controls.
Conclusions
Many radon and decay product monitors are available which give
a wide range of costs and capabilities. No one monitor can do
every job, and each is suited to a different application. In the
future, the author expects an increased reliance on short term
tests, so methods which are good only for long time periods will
probably enjoy less success. This is apparently being anticipated
by the equipment manufacturers because the methods are being
adapted to shorter sampling times. The short term alpha track
detector is a good example of this.
The shortening of the sampling period will also allow
greater use of the more equipment intensive methods. Where it
would net be cost effective to leave a continuous radon or
working level monitor in place for several months to a year,
placing it for a day or two will probably be practical. Thus the
use of continuous monitors is likely to increase. These devices
are alsc particularly well suited to schools and offices where
occupancy is intermittent and where ventilation systems are often
cut back for weekends and evenings. These devices are really the
only way to accurately assess what is happening. Their increased
cost may also be better afforded by the institutions involved, but
this is heavily dependent on how many rooms must be sampled.
Perhaps the best strategy will be continuous monitors in one or
two locations and inexpensive samplers elsewhere. The
11
-------
differences in working level ratios in large buildings also suggest
that working levels as well as radon be monitored in these
buildings.
For monitoring of homes, the driving force for testing
increasingly seems to be real estate transactions. The trend
appears to be toward samplers placed by professionals, at least
for real estate purposes. Professionals have a great deal of
flexibility in their choice of sampling method and are able to use
skilled technicians. This segment of the testing market might be
expected to grow in the future. The professionals can be expec-
ted to purchase the instruments that work for them and that give
then a competitive advantage. It is likely that the larger
operators would have several methods at their disposal and use
whichever is best for the job at hand, or in combination as the
need arises.
Many homeowners testing for their own information still seem
to prefer to be able to buy a sampler at a local store. The local
mail order market might be expected to remain stable, and seems to
be the province of the charcoal and alpha track devices.
12
-------
Prediction of Long-Term Average
Radon Concentrations in Houses
Based on Short-Term Measurements
William E. Belanger, P.E.
U. S. Environmental Protection Agency
Region III
841 Chestnut Street
Philadelphia, PA 19107
Previous studies have examined the relationship between
short term screening tests and annual average radon
concentrations, but have generally been limited by short term
measurements conducted at only one location in the house or
during only one season. This study examines the usefulness of
short term measurements to estimate long-term averages when the
short term measurements are made in living areas of the house in
and in the basement. The effect of season is also examined.
This analysis is particularly useful in situations where there is
not time to do an annual measurement, but multiple rooms can be
easily tested. We used the raw data from five different
researchers to cover the full annual cycle and multiple sampling
locations. The result is a matrix of factors which can be used
to estimate the annual average radon based on a test at any time
of year. Separate factors are calculated for basement and first
floor monitoring locations.
13
-------
Introduction
The Environmental protection Agency Published the Citizen's
Guide to Radon in August of 1986. The Citizen's Guide outlines a
two-step testing process which calls for an initial screening
measurement followed by a longer term follow-up measurement. If
radon in excess of four picoCuries per liter is detected by the
screening test, additional tests of up to a year's duration are
recommended. The whole measurement process including the time
needed for the test, purchase of detectors, sample analysis,
mailing, and procrastination can take well in excess of a year.
Under some circumstances, a full year of monitoring is not
practical or is not desired by the person conducting the test. In
this circumstance it would be desirable to be able to compare the
short term test with the concentration that would have been
measured if a full year test had been conducted. Five existing
data bases have been collected and analyzed. These data bases
were selected because there were both short term and long term
tests performed in the same house. In addition, houses where
continuous monitoring has been conducted have been used to
investigate the variations in testing accuracy which result from
different sampling durations. The results suggest corrections
can be made to short term data to provide an estimate of long
term averages in houses. This paper uses radon measurements made
in several thousand houses to derive factors which can be used to
normalize short term radon data to whole house annual averages.
Use of a short term sample to estimate the annual average
can be expected to introduce additional error into the radon
measurement process. If the errors that are introduced are
smaller than those inherent in the measurements, the additional
errors would be unimportant. If, however, the errors introduced
by short term sampling are larger than the measurement errors,
then there will be the need for a policy decision on whether or
not these errors are acceptable. That decision is beyond the
scope of this paper, but we will attempt to quantify the errors.
Optimization should be an inherent part of this process.
The decision on the acceptability of errors should not be based
purely on the errors observed from simple comparison of short
term measurements with annual averages. These short term samples
will include variance due to seasonal differences, location
differences and differences due to house conditions as well as
"random" variations. Many of these sources of variance can be
reduced by the prudent choice of sampling strategy and
application of correction factors. When the estimate is as good
as can be made from the available data, then the decision on
whether or not the errors are acceptable should be made.
Experimental Methods
Five independent datasets were used in this analysis. These
are the first year data from the EPA State surveys, Data
14
-------
collected by George Mason University for a survey in Northern
Virginia, data collected by Pennsylvania in the Reading Prong,
continuous radon monitoring at one location in each of four
houses in Butte, Montana and continuous radon monitoring at
several locations in a house in Media, Pennsylvania conducted
under the auspices of Rutgers University. The original data was
obtained to allow a consistent analysis to be performed.
No new radon measurements were made for this investigation.
Because existing radon data was used, the author was limited to
whatever analytical methods and quality assurance procedures were
used by the person who collected the data. These ranged from
carefully calibrated quasi-continuous radon monitors to radon
test kits purchased on the open market. All radon measurements
were assumed to be accurate, i.e there was no systematic bias.
Precision was accounted for by eliminating measurements known to
be of low precision. For example, alpha track detectors bought
in the marketplace were assumed to have usable precision only
above a 1-year average of one pCi/1. This is well above the
limit of detection for these detectors, and typically yields a
coefficient of variation of about 20 percent (equivalent to 4
pCi/1 for 3 months). Charcoal detectors of many types were also
included in this analysis. These were again assumed to be
unbiased. This assumption is supported by the analysis of data
collected by Mose. Short term samples using three different
charcoal detectors were shown to be unbiased estimators of alpha
track averages at the same location when the lower radon
concentrations were eliminated. The EPA/State survey data was of
high quality because of careful quality assurance procedures.
The parameter of interest in this investigation is the ratio
of a short term measurement to a long-term average. This
requires that the short-term to long-term ratio is not a function
of radon concentration. This independence is shown by Perritt.
There is a noticeable decrease in precision at low
concentrations, which may account for a dependence being seen by
others. The long term average chosen here as a baseline was the
average of radon concentrations in all floors of the house. This
is not the same as the EPA follow up protocols which exclude
space that is not actually lived in. This was done because the
"lived in" status of floors of a house was generally not reported
in the data, and because the utilization of rooms in a house is
not a property of the house. A basement that is not lived in
this year may have a couch and a television in it next year. All
"livable" spaces were therefore assumed to be occupied.
Results
The EPA/State survey provides seasonal short-term to whole
house annual ratios for winter and spring. The George Mason
University data contains useful seasonal information for summer
and fall but lacks sufficient measurements to calculate a whole
house average. The adjustment of the George Mason University
data was done by normalizing to the EPA/State data. The spring
15
-------
season is contained in both datasets and so the George Mason
University seasonal factors were adjusted to make the spring
ratios the same as those from the EPA/State dataset. This
adjustment is intended to yield "whole house averages" from other
seasons of the George Mason data.
Both the Pennsylvania and New York data allow direct
comparison of winter and annual data. Floor-to-floor differences
can be examined in the New York and Pennsylvania data, but there
is insufficient data to directly calculate whole house averages
because of the lack of second floor measurements. It is not
clear whether the houses had second floors that were not measured
or if second floors did not exist. This leads to the dilemma
whether to compensate for the lack of second floor data in
calculating a household average. We have elected to make this
small correction as if second floors existed in all houses.
From the table below it can be seen that summer first floor
radon measurements should yield a result close to the whole house
average, while similar winter first floor measurements should be
higher than the whole house average by a factor of 1.6. Summer
basement measurements should yield results about 1.8 times the
whole house annual average while basement winter measurements
should be 2.4 times the whole house annual average. Note that
this data is unbalanced. The overestimation in winter is not
balanced by an underestimation in summer. The basement
overestimation is not accompanied by a first floor
underestimation. The author believes this is due to the inherent
bias in the screening locations and in the closed house
conditions used for the short term tests. The annual averages
include periods of low radon concentration which were "designed
out" of the short term tests by choice of sampling location and
closed house conditions. Note the similarity among the datasets.
winter spring summer fall
Basement Seasonal (closed house) to Whole House Annual
EPA/State 2.9 2.1
Geo. Mason 2.3 2.1 1.8 2.3
Pennsylvania 2.0
Average 2.4 2.1 1.8 2.3
First Floor Seasonal (closed house) to Whole House Annual
EPA/State 2.2 1.3
Geo. Mason 1.6 1.3 1.1 1.5
Pennsylvania 1.5
New York 1.2
Average 1.6 1.3 1.1 1.5
Another source of variance is the inability of a short term
sample to predict a long term concentration. We investigated
this effect by breaking continuous monitoring data in five houses
into simulated short term samples. With uncontrolled house
conditions the 95 percent confidence limits yielded a range from
1.4 times the average to 0.1 times the average. This variability
16
-------
can be considerably reduced by avoiding summer sampling, or
presumably by sampling under closed house conditions. When
summer samples are eliminated the range is from 1.4 to 0.7 times
the annual average. This reduction in variance is accompanied by
the introduction of a systematic overestimation in the radon
concentration by about 10 percent. If one wished the most
accurate prediction, it would be prudent to remove this bias. On
the other hand it may be decided that such a bias is desirable in
that it reduced the chance of under-estimating the radon. There
appears to be a small improvement in predictive ability as sample
time is increased from one day to about one week, with little
additional gain in accuracy until large fractions of a year are
sampled.
Conclusions
We have shown that short term sampling can provide estimates
of long term radon concentrations at the same location with
little bias. However there is considerable variance associated
with the ability of individual samples to predict a whole house
average. The variance is partly due to the differences between
houses, partly due to differences between floors in a house,
partly due to the time when the short term sample was taken and
also due to other random effects. If short term measurements are
to be used to predict a long term whole house average, then this
variance should be reduced to a minimum. Any procedure which can
be incorporated into a sampling protocol which reduces the
variance is desirable.
The variance can be reduced by a number of techniques. For
example there is a systematic difference between the radon
measured in the basements and first floors of houses. It is
possible to compensate for the systematic difference by applying
a correction based on our experience with large numbers of
houses. Further reduction in variance is possible if sampling is
done simultaneously on more than one floor. By sampling in this
way, it is no longer necessary to estimate the floor to floor
differences based on a large number of houses. The differences
become known, at least at the time of the sample. This should
reduces the variance in the resulting prediction. In addition
the number of floors is not the same in all houses. A house with
a second floor has twice the chance for non-basement radon
exposures as a one story house. One would calculate a different
whole house average for a two-story house than for a ranch house.
The number of floors in a house are easily counted, so there is
no reason to treat all houses as if they were of one design.
The author therefore recommends sampling on each floor of the
house as a means of minimizing the errors in the calculated
average.
It may also be possible to use the simultaneous basement and
first floor measurements as a confirmatory measurement. If floor
to floor differences fall outside an expected range, then the
measurements may be suspect. Basement to first floor ratios
17
-------
appear to fall mainly in a range from 1 to 4 from the New York
and Pennsylvania data. Thus if one were to make short term
measurements in a basement and on the first floor and the ratio
were between one and four, the data would seem credible. In the
data examined this ratio appears to fall outside these limits
about 25 percent of the time. A ratio of .75 to 5 is satisfied
about 85 percent of the time, so less "rejections' of tests can
be obtained by using wider limits. The ratio between basement
and first floor tests be used as a quality assurance tool with
limits of 1 to 4 or .75 to 5 depending on the level of confidence
desired.
By prudent choice of sample duration and house conditions,
it is possible to minimize the variance due to sample time. It
has been shown that closed house samples of about a week duration
can generally predict average radon levels at the same location
within about -25 to +40 percent (the bias due to closed house
conditions.) This suggests that any sampling protocol should
include closed house conditions and that the duration of short
term samples be more than a day and preferably as long as a week.
Seasonal variations were also obvious in the data. These
seasonal variations are a source of variance if a short term
measurement is to be used to predict an annual average. Unlike
floor to floor differences, seasonal variations in a particular
house cannot be determined except by actually measuring the radon
in more than one season. It will therefore be necessary to
utilize seasonal trends from large numbers of houses to reduce
this source of variance. It cannot be eliminated because of the
differences between houses in their seasonal radon patterns, but
at least the seasonal trends can be removed.
References
1. D. Mose, G. Mushrush and C. Chronsiak, "Reliability of
Inexpensive Charcoal and Alpha Track Radon Monitors,"
Natural Hazards, 1990
2. C. Granlund and M. Kaufman, "Comparison of Three Month
Screening Measurements with Yearlong Measurements Using
Track Etch Detectors in the reading Prong," Pennsylvania
Department of Environmental Resources (1987)
3. R. Perritt, T. Hartwell, L. Sheldon, B. Cox, C. Clayton, S.
Jones, M. Smith and J. Rizzuto, "Radon 222 Levels in New
York State Homes," Health Physics 58 no 2: 147. (1990).
4. "Radon Measurement Comparison Study," U. S. Environmental
Protection Agency, EPA 520/1/89/034. (1990)
5. "Seasonal Variations of Radon and Radon Decay Product
Concentrations in Single Family Homes," U. S. Environmental
Protection Agency, EPA 520/1/86/015. (1986)
18
-------
AN ANALYSIS OF THE PARAMETERS INFLUENCING RADON VARIATIONS IN A HOUSE
Albert Montague, P.E., EPA Region III, Philadelphia PA.
William E. Beianger, P,E., EPA Region III, Philadelphia PA.
Francis J. Haughey, Ph.D. Rutgers University, New Brunswick, N.J.
The purpose of this study was to devise a statistical meang of eliminating interfering
variables from data of specific interest. The primary data being sought were the comparative
differences in radon concentration in a residential dwelling under two operating modes.
Radon levels were measured in the basement, the first floor, and the second floor of the
house. A meteorological station was constructed that continuously measured the variable
parameters such as wind speed, wind direction, outside temperature, relative humidity, and
barometric pressure. The operating modes were changed every two weeks and the data were
examined in two-week cycles.
An attempt was made to remove the effects of the unwanted variables by using the standard
stepwiae linear regression. Since -this approach was not adequate without compromising the
quality of the data of primary interest, a novel time-scale method was developed to achieve
the desired discrimination. This utilized a computer routine to simulate the effect of an
electronic RC filter.
When the averages were taken for comparison of the effect of change on radon concentration
in two operating modes, they were seen to be separated by 15 standard deviations. Thus, this
new method of variance reduction, by removing the unwanted time scales, yields results that
are clearly significant.
19
-------
A residential dwelling in a. rural area near Media, Pennsylvania, was the
site for a recent radon study that encompassed a five-month period of
continuous meteorological and radon data collection. The home, a five-year
old, two story, colonial wood frame structure, with standard insulation and
a full basement, was modified with an apparatus to observe the effect of the
radon levels in the dwelling when the modification was switched on and off.
Previous observations, though limited in scope, showed radon levels in the
basement to range between 15-20 pCi/1.
A rigorous statistical analysis was performed on each of the data sets.
Initial efforts focused on assessing whether the raw data showed a normal
distribution by applying the Chi-square test, if the data were not normally
distributed, various numerical transformations were applied as appropriate,
to try and obtain a normal distribution, e.g., a 24-hour moving average
(smoothing function), log transformation, or other numerical corrections
associated with an observed phenomenon, for which there is a known or logical
explanation.
Since the data generated are time-based, i.e., they represent a time series
which may have cyclical characteristics. When such a trend exists —
described by the correlation of radon (dependent variable) with barometric
pressure, temperature, and other associated factors — the trend may be
removed by subtracting it (slope of the line generated by performing a linear
regression) from the original data set. Otherwise, the opportunity to detect
any difference in the mean values of the radon concentrations may be masked
by unwanted variation, especially if the difference is numerically very small
between the mean radon values representing the two operating modes.
When time series data are analyzed, one must account for the tendency of
each measured value of a parameter to depend in some way on the previous
values of that parameter. This dependence can occur in two ways. For
physical systems which cannot change instantaneously, the physical state of
the system at time t can best be represented by its state at time t - ,&t,
where At, is an increment of time. As at approaches zero, the value of the
parameter, F(t), approaches its value an instant ago, F(t-At). Auto-
correlation of the time series provides a measure of this effect, with At
used as a displacement in time. The decrease in autocorrelation as At
increases provides a measure of how fast the parameter can change. Some time
series also exhibit periodic or cyclical variations. These variations are
usually superimposed on random variations in the system (and the
measurement). In this case the autocorrelation will initially decrease as
At is increased, and then level off somewhat or even increase as A t
approaches the period of the cycle.
When a long time series, such as several months of weather data are
analyzed, there will be a number of components to the time series. Ambient
temperature provides a good example of this. First, there will be a tendency
of each temperature measurement to depend on the previous temperature. There
will also be a cyclical variation with a period of one day reflecting the
periodic nature of the driving force, sunlight. This will be superimposed
on a variation with a time scale of a few days as weather systems move across
the measurement device. Finally, there will be an annual cycle with a period
of one year. These variations can be represented in the frequency domain
using a Fourier transformation, and the result is a power spectrum.
If one desires to determine the effect of one variable on another, for
example, the effect of temperature on radon, variance can be reduced by
choosing the time scale of interest and excluding variations which occur on
greatly different time scales. For example, when changes happening on a two-
week scale are being examined, variations on an annual or a daily time scale
are of little interest. These variations contribute to the variance in the
data without adding useful information. This can best be illustrated by
graphic examination of the data. It can be seen that both radon and
temperature exhibit diurnal effects. This does not imply cause and effect
because household activities unrelated to outdoor temperature can affect
radon. At this point it must be decided whether these 24-hour variations are
central to the analysis of cause and effect, or if they are more likely to
introduce artifacts. In this case diurnal variations are more likely to
20
-------
introduce artifacts, and so they would be removed, one easy way to remove
these diurnal variations is to apply a transformation. The simplest way to
do this is to take a moving 24-hour average of the data. This removes
variations not of interest which reduces the likelihood of artifacts in the
analysis. This also reduces the variance in the data thereby improving our
ability to identify more subtle effects.
After diurnal variations are removed, it can be seen that in the short term
radon seems to increase as temperature increases. On the other hand, the
annual cycle, part of which can be seen in the data, shows radon going down
in the summer as temperature rises. These two opposite effects tend to
cancel when the time series are correlated. If the true relation between the
variables is to be investigated, it is necessary to analyze each time scale
independently. There is a need to transform the data in some way to isolate
time scales of interest in much the same way that diurnal variations were
removed.
It is possible to construct a mathematical filter to select the time scale
of interest. This may be accomplished relatively easily by making an analogy
to an electrical circuit. A resistor and capacitor provide a simple filter
to select a frequency range of interest. The electrical circuit diagram
appears below.
INPUT
.c
OUTPUT
The response of this filter can best be represented by a plot of the output
level versus frequency. This is done on log-log coordinates and is called
a Bode plot. The circuit above yields close to 100% response (output equals
input) for very low frequencies. For high frequencies, output decreases with
increasing frequency. As frequency is doubled, response is halved. There
is a critical frequency where the response transitions from flat to frequency
dependent. This critical frequency, in radians per second, equals 1/RC.
Angular frequency in radians per second is cycles per second -times two pi.
A Bode plot of the simple RC filter above is shown below.
log
output
0
-45°
-90C
amplitude
log frequency
phase
Note that a phase change is introduced by the RC filter. For frequencies
above 1/RC, there is a 90 degree phase shift. At 1/RC the shift is 45
degrees. Only at frequencies well below 1/RC is the output in phase with the
input. This is important in selecting the filter constants since this phase
change may appear as a time shift in the data near 1/RC. In order to utilize
this filter in analysis of data it is important to consider both the
amplitude and phase of the response. Properly used, such a filter will allow
variations with angular frequencies slower than 1/RC to pass while
attenuating those above. This low-pass filter can be very useful in
observing long term trends and variations in the data. It may also be used
to remove these long term trends by simply isolating them and subtracting
them from the input. This effectively produces a high-pass filter.
21
-------
The above paragraphs describe the desired response of the filter. A
practical realization of this transfer function may be formulated by taking
advantage of the fact that the response of an electrical circuit to a step
input in the time domain uniquely defines the frequency response of the
filter. If one desires to construct a function which responds in the
frequency domain as an RC filter, it is only necessary to construct a filter
with the same response to a step input. This is what is done below.
In response to a unit step input, i.e., an input which suddenly changes from
zero to one at time t a simple low pass RC filter produces a response of the
form l-e""1 where T is a time constant equal to the product of the resistance
and capacitance, RC. This is the same RC used above for the critical
frequency in the Bode plot. This function is graphed below and is a simple
exponential approach to a final value.
A useful property of this function is that the slope of the output function
at any point on the curve is tangent to a line which intercepts the final
value at t+T. This property arises from the fact that
d/dx(e*) =e*
By simply producing a formula that does this, the desired filter can be
created.
Let F(t) be the input function's present value and G(t) be the output
function's present value. At time t this function must be changing at a rate
which would cause G(t) to intercept input function F(t) at time t+T.
The slope of G(t) must therefore be (F(t)-G(t))/T. in discrete mathematics,
G(t+l) may be calculated by simply adding the slope to G(t>. Hence the
output at any given time is G(t+l)««(t)+(F(t)-G{t)>/T. This function is used
directly in the Lotus spreadsheet. Choice of T allows the filter function
to be tailored to the desired response, with the critical angular frequency
being 1/T.
Consistent with the analytical plan, observed radon values (raw data spikes
were adjusted, i.e., attenuated to compensate for concurrent rain events that
clearly affect radon levels. Other factors that also appeared to contribute
to abnormal basement radon values, e.g., windows or doors remaining open for
extended periods of time, were analyzed using daily log information, and to
the extent necessary, appropriate adjustments were made on the initial radon
values, by performing a linear interpolation.
The raw temperature data were plotted against time, for the entire
monitoring period. Examination of the graphical representation showed a mild
concave-up trend line for the period, a trend that was expected, due to
seasonal changes, but a trend that could hamper the analysis and opportunity
to see any difference between the two study modes. This long-term trend, was
successfully removed by employing the electrical analog filtering technique,
previously described.
At this point, each of the data sets namely, outside temperature, relative
humidity, barometric pressure, and wind were subjected to a Chi-square
analysis, to determine if each of the full period (November through March)
meteorological data sets, and, in particular, the now modified (rain-
corrected) basement radon data, satisfy a normal frequency distribution.
Basement radon values were found to be normally distributed after
transforming the data, i.e., doing a 24-hour moving average, and then
performing a log transformation. In light of the observed benefit, the
remaining data sets were also modified using the moving average technique.
However, the remaining meteorological parameters did not satisfy a normal
distribution, even though various transformations were applied to the data.
Of the four parameters analyzed, only the barometric pressure data came
22
-------
sufficiently close to being characterized as normally distributed, after
performing a 24-hour moving average data transformation. In light of the
above findings, it became evident that further statistical analysis of the
data and associated results should be viewed with a certain degree of
caution. A truly valid comparative statistical analysis, requires that each
data set has similar frequency distributions, e.g., normal frequency
distribution.
After all the data sets were transformed (smoothed), to remove any diurnal
variations - daily and seasonal variations were not the center of interest -
a series of statistical correlations (regressions) was performed on each of
the meteorological parameters, relative to the log-transformed, normally
distributed, basement radon data. This was done to prepare the data for the
correction (removal) of trends generally associated with time series data,
trends that could mask the possibility of observing any difference--
particularly if the difference is very small—in radon levels between the
two study modes. Each of the regression analysis-generated slope and y-
intercept values, with the values for m and b respectively, in the linear
equation, y= mx+b, for the line that best fits the data being studied, and
the coefficient of correlation.
To illustrate:
let Y - Radon Initial Value, at Time T,
X = Barometric Pressure Initial Value, at Time T,
Y, - New Radon Value, at Time T,
where, m = Slope of Line (Trend)
and, b - y - Intercept
then, Y, = Y - ( mX + b ).
These corrections were applied even if the correlation (slope of the trend
line) was low, and no plausible explanation could be offered or established,
to explain the relationship between the independent (meteorological) and
dependent (radon) variables. Two iterations were required to perform this
phase of the analysis.
The first iteration generated the following correlation (r value), and in
descending order of priority:
* Barometric Pressure 0.44
* Relative Humidity O.37
* Wind ( B-W Vector) 0.28
* Temperature 0.14
A second correlation analysis was done after the data sets were shifted in
time, specifically in three-hour increments, forward and backward from their
initial and corresponding (time dependent) values, until a maximum
correlation (r value) could be found. It yielded the following correlation
values:
* Barometric Pressure 0.54
* Relative Humidity 0.39
* Temperature — ——- 0.36
* Wind ( B-W Vector) 0.28
Consistent with the above findings and planned method of analysis, the trend
(relationship) between barometric pressure and the parameter having the
highest correlation value with radon was removed. The revised basement radon
values were again subjected to a third correlation analysis, with the
remaining meteorological data sets, namely, relative humidity, temperature
and wind.
Three more basement radon corrections were executed to remove the trends,
correlations—observed above, i.e., temperature, relative humidity and wind.
In each case, the last revised basement radon value was used in the
adjustment process. Revised basement radon values and corresponding
meteorological parameters were then subjected to a second regression
•analysis, after each trend correction was made to confirm that the adjusted
slope of the trend line (correlation) was indeed zero.
23
-------
Basement radon values were then graphically plotted versus time. Although
considerable effort was expended to carefully filter out—to the extent
possible—the unwanted (trends) variations, diurnal and seasonal bias, within
the observed basement radon variation—again presumed to be due to
meteorological influences—a concave-down (semi-cyclical) wave was observed
which presented yet another unknown influencing factor. This finding called
for the reuse of the trend-filtering technique, previously described.
At this point the adjusted basement radon data listed in a continuous
time series format, were split into two numerical groupings. One group,
representing basement radon readings for the bi-weekly periods, when the
dwelling operated under the first study mode. The second group represented
basement radon values, when the dwelling operated under the second study
mode. The data, were then subjected to an analysis of variance to see if
there was a statistical difference in basement radon values between the two
operating modes. It showed that the differences between the means was over
15 times the standard deviation of the means and therefore the differences
were very significant.
24
-------
Application of Building Diagnostic Techniques to Mitigate Very High Radon
Levels in a Commercial Building on a Superfund Site
by
D. Bruce Harris
Air and Energy Engineering Research Laboratory
U. S. Environmental Protection Agency
Research Triangle Park, NC
Thomas Staley
Radon Screening Services, Inc.
Englewood, CO
and
Philip C. Nyberg
Region 8
U. S. Environmental Protection Agency
Denver, CO
ABSTRACT
A small commercial facility built upon a minerals extraction
plant process waste fill was found to have average radon daughter
concentrations of O.4 working levels (WL) with peak concentrations
exceeding 2.0 WL. At the request of EPA's Region 8 Emergency
Response Branch, the Office of Research and Development's Air amd
Energy Engineering Research Laboratory/Radon Mitigation Branch
undertook and completed the mitigation of this building within 1
week after receipt of the request for assistance. The structure had
been built in four stages which meant that two interior footings
probably existed. Soil gas concentrations varied between 20,000 and
250,000 picocuries per liter (pCi/L)*. Pressure field extension
testing, developed in residential and school mitigation, indicated
that at least two suction points would be needed to successfully
apply sub-slab depressurization to the building. Since radon levels
were so high, each of the three slab sections was penetrated and
despressurized with two sharing one suction fan. This system has
mciintained, for at least 38 days, average concentrations of 0.0075
WL which is well below the 40CFR192 (1) standard of 0.020 WL.
(*) 1 pCi/L = 37 Bq/cu m
25
-------
INTRODUCTION
During the summer of 1989 a remedial investigation was
conducted at a major contaminated property which is part of the
Denver Radium Superfund Site, one of the original entries on the
National Priorities List of uncontrolled hazardous waste sites.
The property is an inactive chemical processing plant for the
extraction and purification of reagent-grade metals. Many years
before, in the 1917-1924 era, the plant had been the site of a
small radium extraction facility where ore containing radium and
other naturally occurring radioactive materials was brought by
train from western Colorado and subsequently purified for various
uses. The waste products from this operation remained largely
forgotten on the property, mixed by the activities of wind, water,
and humans, until 1980 when they were discovered as a part of a
major radioactive waste investigation.
In the course of the remedial investigation, properties
adjacent to the existing chemical processing facility were checked
to determine whether any radioactive contamination had migrated
beyond the present-day property boundaries. Elevated gamma
radiation exposure rates were identified in and around a small,
one-story office building due east and across a street from the
processing facility. With the cooperation of the building's owner,
substantially elevated gamma radiation levels and radon (222-Rn)
decay product concentrations were measured within the office
building. Since the investigation had been conducted by a
contractor for the Colorado Department of Health (CDH), and since
CDH lacked the resources to deal with an emergency situation of
this type, EPA was contacted and assistance was requested.
Confirmation measurements were quickly made, and additional
ventilation was instituted to bring the radon decay product
concentration to less critical levels. Initial grab samples showed
up to 4 working levels (WL, a special unit of radon decay product
concentration) in the front office, which may be compared to the
average allowable concentration in a uranium mine of 0.3 WL, set
by the Mine Safety and Health Administration (MSHA) of the
Department of Labor. MSHA also allows a maximum of only 1 WL before
closing a mine or requiring worker respiratory protection. The
initial ventilation measures reduced the average value to about 0.3
WL.
Having temporarily decreased the radon levels to a more
reasonable if not satisfactory range, EPA examined the options for
providing a more positive control mechanism. While there were
several administrative possibilities to deal with this situation,
it was decided to use the emergency response authority given to
EPA under CERCLA/SARA (Comprehensive Environmental Response,
Compensation and Liability Act/Superfund Amendments and
Reauthorization Act) to install a radon mitigation system in the
building. In this way it was possible to use some existing
26
-------
contractual arrangements to proceed rapidly with the system design
and installation. The EPA Region 8 Emergency Response Branch in
Denver contacted the ORD's Radon Mitigation Branch in North
Carolina. Through an existing contract with a Denver radon
mitigation firm, ORD was able to bring the necessary equipment,
supplies, expertise, and manpower to the scene within 1 week.
Diagnostic tests were conducted and a complete radon control system
was installed within a few days with minimum disruption to the
building's occupants.
PREMITIGATION TESTING
Testing on this building began with charcoal canisters placed
in the major office areas for 1 day followed by a repeat lasting
2 days. A gamma scan was performed at the same time. The data are
shown on the building schematic in figure 1. Grab samples of radon
and radon progeny taken at the same time agreed with the canister
data. The radon progeny levels were far above the accepted 0.020
WL, with the range between 0.13 and 4.0 WL and the equilibrium
ratio averaging 25%. These elevated levels triggered a more
intensive investigation.
Continuous working level and radon gas monitors were installed
in one of the offices. Even with the windows left open to reduce
the exposure levels, the average progeny concentration was 0.25 WL
(100 pCi/L @25% equilibrium) with diurnal variations between 0.012
and 1.0 WL. These data confirmed that a serious situation existed
and it was decided to mitigate the building under the emergency
response provisions of SARA.
BUILDING DIAGNOSTIC TESTING
Building diagnostics are performed to provide the necessary
data to design a mitigation system. The building was evidently
constructed in four stages. The slab was divided into a front slab
which included the bathroom and secretary's area, a middle section
containing the original office and the first added office, and a
rear slab with an office and storage area (figure 1) . The
diagnostics attempt to determine the flow characteristics and the
source strength of soil gas under the slab. Small holes were
drilled in the floor in each room and a continuous sample of
filtered soil gas was pumped into a Pylon AB-5 radon monitor which
measures scintillation activity of the alpha decay events within
the detection cell. Counts were printed out every 30 seconds and
the levels noted when a steady reading was reached in 3-4 minutes.
The concentrations of radon in soil gas under the floor slab were
estimated to range from 33,000 to 250,000 pCi/L which are
significantly elevated over typical background values and are quite
consistent with the elevated levels of airborne radon found in the
building. One soil sample obtained from beneath the slab showed a
radium (226-Ra) concentration of about 3700 pCi/g, which may be
27
-------
compared to 1-2 pCi/g found in normal soil, or to 15 pCi/g which
is the maximum allowable concentration remaining at reclaimed
uranium tailings sites.
With the extremely high levels found under the slab, extra
safety measures were necessary. All building diagnostic and
mitigation work was performed after normal business hours to limit
possible exposure to the building's occupants. Full face masks
using certified breathing air from pressurized tanks were used
whenever penetrations were open in the slab. Protective overalls
were available to reduce exposure to progeny deposited on skin and
clothes.
The gas flow under the slab is investigated to determine if
sub-slab depressurization can be applied to the building and the
distance from which a single suction hole can pull soil gas. This
test is called a pressure field extension measurement and uses a
vacuum cleaner to apply a suction to a hole drilled at the site of
a potential mitigation suction point and a number of small holes
at varying distances from the suction point to determine how far
a negative pressure can be detected. Potential suction points are
selected after examining available blueprints and visual building
inspection to determine if sub-slab barriers such as interior
footings or sewer pipes exist. This building had no prints
available, but the visual inspection indicated three or four
different construction activities. A test suction point was located
in each of three sections. The data indicated small pressure
communication (less than 0.003 in. water or 0.75 Pa between the
center and rear office sections but none with the front reception
area.
Because of the extremely high radon levels observed in this
building, a conservative mitigation strategy was adopted. It was
decided to locate a suction point.in each section of the building,
since the degree of sub-slab soil gas interconnection was not
strong. The two suction points in the center and rear sections were
connected to a common suction fan, while the one in the front
section was connected to a separate fan. Following common
mitigation practice, a hemispherical pit approximately l-foot (30-
cm) in radius was excavated under each 5-inch (12.5-cm) suction
point hole cut through the concrete floor slab. This procedure
minimizes the high pressure drop experienced as gas velocity
through the soil increases near the hole and extends the distance
the pressure field reaches. The fans were mounted outside the
building so if any leaks developed in the pipe the suction in the
pipe would pull room air in rather then inject soil gas back into
the building. A vertical mounting was selected for the fans to
allow any condensation to drain back under the slab and not collect
in the piping or fan and freeze in the winter.
28
-------
RESULTS
The effect of the mitigation system can be seen dramatically
In figure 2. During the week prior to mitigation activities, the
concentration ranged from 0.006 to 1 WL with the average about 0.2
WL. Mitigation activities caused a momentary increase as high as
2 WL, but when the fans commenced operation, the levels quickly
dropped to less than 0.01 WL and were very stable. The average
level over the next 38 days was 0.0075 WL, well below the accepted
EPA guideline of 0.02 WL. Subseguent testing with passive radon
monitors over the winter showed only minor seasonal increases in
that level, indicating that the radon concentrations within the
building have been successfully reduced and controlled. The gamma
radiation exposure rates in the building remain elevated, however,
and this will be dealt with in future remedial activities at this
property. The building continues to be used by the occupants on a
regular basis.
REFERENCES
1. Code of Federal Regulations. 40CFR192.
29
-------
4"{1O cm) PVC Mitigation Systen
90jiR
870 |83O| 115JJR
930
= 6/24-26 •.*.:;:.
1989 . /V.'*V •
FIGURE 1.- Radon concentrations (pCi/L) and gamma readings (pR/h)
u 1.5
MITIGATION FAN ON
0.5
0 B 16 0 8 16 0 B 16 0 8 16
23 24 AUG. 25 26
TIME, hours
FIGURE 2. Radon progeny levels during and after mitigation
30
-------
AN AUTOMATED, SEm-CONTJNUOUS SYSTEM
FOR MEASURING INDOOR RADON PROGENY
ACTIVITY-WEIGHTED SIZE DISTRIBUTIONS,
Dp: 0.5-500 NM
Chih-Shan Li, Philip K. Hopke,
and Mukund Ramamurthi
Department of Chemistry
Clarkson University,
Potsdam, NY 13699-5810
A system for the detection and measurement of indoor radon progeny activity-weighted size
distributions (particle size, d > 0.5 nm) and concentration levels has been developed. The system
is microcomputer-controlled and involves a combination of multiple wire screen (Graded Screen
Array) sampler-detector units operated in parallel. The radioactivity sampled in these units
permits the estimation of the radon progeny activity-weighted size distributions and concentration
levels on a semi-continuous basis. This paper presents details of the system and describes various
stages in the development of the system. Results of field measurements in a residential
environment are presented to illustrate the resolution, sensitivity and capabilities of the
measurement system.
31
-------
INTRODUCTION
A substantial interest in the properties and occurrence of radon and its progeny has arisen
in recent years from the recognition that the infiltration of radon into the indoor environment may
constitute a significant human health hazard. The inhalation and subsequent lung deposition of
the short-lived radon progeny, 218Po, 2UPb, and 214Biy214Po, have thus warranted the study of their
diffusivity and association with molecular cluster aerosols in the ultrafine cluster size range (0.5-5
nm) and to larger mode aerosols.
Traditionally, the ultrafine cluster and accumulation modes of the activity size distribution
have been termed as the "unattached" and "attached" fractions, respectively, in view of the
significant difference in their diffusivities. "Unattached" 218Po has been assumed to have a single,
constant diffusion coefficient, typically 0.054 cm2/sec (1), and samplers have been developed that
provide operationally defined estimates of these two modes of activity (9). Experimental
measurements of activity-weighted size distributions in recent years have shown that the
"unattached" fraction is in reality an ultrafine cluster fraction in the 0.5-5 nm range, whose
diffusivity and characteristics are dependent upon the nature of the indoor atmospheric
environment (4,10,11).
In view of the important contribution of the ultrafine cluster mode to the estimated lung
alpha dose (e.g., 5), efforts have been made to develop size distribution measurement techniques
that overcome the lack of sensitivity of conventional methods for particle size, d < 5 nm. The
alternative techniques developed have been adaptations of the wire screen diffusion battery
concept, and have been called Graded Screen Arrays (GSA) (4,8). GSA systems consist of varying
mesh number, single/multiple wire screen stages operated either in series or in parallel, with a
choice of a wide range of wire screen parameters and sampling flow rates, coupled with a
technique for the determination of the radioactivity associated with the particle size distribution.
GSA systems can be classified into grab sampling (4,8) and simultaneous sampling/analysis
categories (11,13). The conceptual design of the automated, semi-continuous GSA system
described in this paper thus evolved from an attempt to utilize the most advantageous features of
both of these systems (6). The primary objective of the system was to provide semi-continuous
estimates of 218Po, 214Pb, and 214Bi activity size distributions and concentrations. The system was
also required to be capable of adequate size resolution and sensitivity at low indoor radioactivity
levels, and be experimentally characterized for ultrafine cluster diffusional deposition (plateout)
losses. The development of the system is described in detail (8). The salient features of these
developmental stages and results of recent initial field measurements are presented in this paper.
MEASUREMENT SYSTEM
The measurement system involves the use of 6 compact sampler-detector units (Figure 1)
operated in parallel. Each sampler-detector unit couples wire screen penetration, filter collection
and activity detection in a way as to minimize depositional losses while being sufficiently rugged
for field operations. The system samples air simultaneously in all of the units through the sampler
slit between the detector and filter sections in each unit (Figure 1). The filter section consists of
a filter holder assembly that is inserted from the base of the lower aluminum block and allows easy
filter replacement. However, in the initial design and field experiments, the filter holder assembly
was sealed into the lower block. The sampled air is drawn through a 25 mm Millipore (0.8 \im
Type AA) filter that is supported by a stainless steel support screen. One of the sampler-detector
units is operated with an uncovered sampler slit, thus providing information on the total ambient
radon progeny concentrations. The sampler slits on the remaining units are covered with single
or multiple wire screens of differing wire mesh number.
32
-------
Rjip, Rl*tichl> ltd*I TLID
Fl I tan. Ml 11 Ipors. 0.8 un
Multichannel Bjffer. Ortae Model Bian
Hultlplwr. Or too Modal 470-8
fepim«r>. Ortao Modal BBS
Rlpna DatKtor, Ortao DlflD 460
Figure 1. Sampler-detector unit
in the measurement system.
Figure 2. Schematic representation of the
various components of the measurement
system.
The upper section of each sampler-detector unit incorporates an ORTEC Model DIAD
II ruggedized, 450 mm2 surface barrier alpha detector sealed into an aluminum block. The
detector is positioned in the aluminum block, concentric with the filter in the lower aluminum
block and detects the alpha particles emitted by the 218Po and 214Po atoms collected or formed on
the filter. The signals from the alpha detectors are connected through amplifiers into an 8-
segment multiplexer and routed to an personal computer-based multichannel analyzer.
Figure 2 is a schematic diagram of the various components of the measurement system.
A dedicated microcomputer controls acquisition of the alpha spectra, operation of the sampling
pump, sample time sequencing, and data analysis. A typical sequence utilized in sampling air with
1-20 Pci/1 of radon involves a 15 min sampling interval during which the first alpha spectrum is
acquired followed by a 20 min delay period and a 40 min second alpha counting period prior to
data analysis. The next sample is then begun following a delay period of between 15 min to 100
min to permit further decay of the 214Pb and 2I4Bi. The delay period duration is chosen based on
sensitivity constraints imposed by the residual alpha counts remaining from the previous sample(s)
since the filters are not changed between samples. This sequence of sampling, counting, and
analysis permits automated, semi-continuous operation of the system with a frequency of between
1.5 to 3 hours.
The alpha counts from 218Po and 214Po detected by each alpha detector in the two counting
Intervals are used to calculate of the radon decay product concentrations penetrating into each
unit (14). The observed concentrations of 218Po, 214Pb and 214Bi are used to reconstruct the
corresponding activity-weighted size distributions using the Expectation-Maximization (7) or
Twomey (15) algorithms. The penetration characteristics of the 5 stages with screens are
calculated using the Cheng-Yeh penetration theory since the wire screen parameters used in these
samplers are identical to those of Yeh et al. (16). This theory has been verified experimentally
in the size range dp > 4 nm (2,3,12). The theory was also recently assessed to be accurate in the
cluster size range for 30 and 145 in"1 mesh screens using a 218
cm2/sec (10).
!Po cluster aerosol, Davg=0.078±0.003
The determination of optimum sampler-detector design and operating parameters (sampler
33
-------
diameter, slit width, detector-filter distance and sampling flow rate) were determined by
experimental testing of a prototype sampler-detector unit in a 2.43 m3 radon-aerosol chamber and
theoretical studies that are described by Ramamurthi (8) and Ramamurthi et al (10). Table I
presents details of 6 optimized sampler-detector units.
Table I. Design and operating parameters for the optimized measurement system.
Unit Sampler Sampler
Slit Width Diameter
(cm) (cm)
1 0.5
2 0.5
3 0.5
4 0.5
5 1.0
6 1.0
5.3
5.3
5.3
5.3
12.5
12.5
Wire Screen dp(50%)
Mesh x Turns (0.5-500 nm range)
(nm)
_
145
145x3
400 x 12
635x7
635 x 20
_
1.0
3.5
13.5
40.0
98.0
Sampling flow rate = 15 1pm (each unit)
Detector-Filter separation - 0.8 cm (all units)
The number of stages (six) and stage progression shown in Table I was based on the
conclusions of the simulation study. The study also yielded concepts regarding the optimum size
resolution that could be obtained from the measurement system. The number and width of the
size intervals used in the reconstruction process was dictated by considerations of size distribution
accuracy and stability. An optimum number six inferred size intervals in geometric progression
within the 0.5-500 nm size interval were thus selected. This progression of size intervals maximizes
the differences in penetrability through the various stages insuring solution accuracy and stability
while yielding sufficient size resolution in the inferred activity distribution.
FIELD MEASUREMENTS
Measurements were made in a one-story residence with living room, dining room,
kitchen, two bedrooms, a study room, two bathrooms, and basement in the Princeton, N.J. area.
Activity size distributions were measured in the living room and the bedroom over two week
period (1/16-1/31/90). A total of about 10 measurement were made in the living room and more
than one hundred measurements in the bedroom with different types of particle generation.
Aerosols were generated from candle burning, cigarette smoking, vacuuming (electric motor),
cooking, and opening door from normal activities in the domestic environments. The particle
concentrations were measured by using a Gardner manual condensation nucleus counter. The
concentration and size distribution of radon progeny were determined by a semi-continuous graded
screen array system. A sequence (0-15, 0-15, 15-35, 35-75) with 75 minute sampling was chosen
because the radon concentration was in the range of 5 pCi/1 - 50 pCi/1.
The influence of cigarette smoking (20 minutes) on the radon progeny size distributions
in a closed bedroom are shown in Figure 3. The measurements were made 5 min after lighting
the cigarette (5-20 min), 80 min later (80-95 min), and 155 min later (155-170 min). The fraction
of 2J8Po in 0.9 nm size range changed from 60% to 8%. The fraction of 2UPb and 214Bi in 0.9 nm
size range was about 10% and becomes essentially zero. The fraction of three distributions in 1.5-
15 nm size range stayed the same. There is a large increase (from 40% to 80%) of 218Po in the
"attached" mode (50-500 nm size range) with insignificant changes (from 35% to 40%) in 2HPb
and 214Bi fractions in this mode.
34
-------
1,00
O.BO
0.60
0.4O
0.20
Rn -132: ll.WpCI/1
CN «nt: 3,aoo/an3
CD Po-218 : 6.94
i—A Pb-2i4:iBO
*—* 8-214:2.65,
i PAEC : M.3Q
0.5 1X1
10.0 1OO.O
Portido Diameter (nun)
Q.BO
0.60
0.40
0.20
tLflO.
Rn -222: 10pCI/l
CM oonc. : 1 NXOOO/cmJ
C3 Po-218 n2.52pCI/l (2.1 *)
«. 4 ftl-214 : «. 23 fa/In A *)
* — * Bf -214 : 2.69 pC1/l ».» «.
D D PAEC : 54 Jttnfa. (0.5 *)
^
/
O.S 1X1
10.0
Particle Diameter (ran)
100.0
A: Background Size Distributions.
B: During the active smoking period.
1.00
o.ao
0.80
0.40
0.20
0.00
Rn -222: 1S.25pC!/l
CN cone, . 22.000/071}
a Pa-£lB:11.65 pCI/l (2.2*)
A - A Pb-21* • 0.4O pCf/l M.+ X1
.
-214; 7.34 fCl/\l2.0
mWL (0.3
n PAEC : 87 .5 m
1.00
0.90
0.60
0.40
0.20
Rn -222 : 1S.5Oj>CI/!
CN cone.: 10,pod/cm3
2-1 K)
jr-
Pb-214 : 7.66
H-2Hi7.BB
PAEC i 7BJ mWL (0.3 »
0.6 1JD
10.0 100.0
Portid* DTameter (nm)
10.0 tOO.O
ParUde Mamettr (nm)
C: 80 minutes later.
D: 155 minutes later.
Figure 3. Activity size distributions in a bedroom before, during, and after smoking a cigarette.
The influence of cooking on the radon progeny size distributions with an open bedroom
door is shown in Figure 4. A steak was pan fried for 20 minutes (0-20 min) using a gas stove
burner in the kitchen. The measurements were made 5 min later (5-20 min), 80 min later (80-95
min), and 155 min later (155-170 min). The fraction of 218Po in 0.9 nm size range changed from
60% to 15%. The fraction of 214Pb and 214Bi in 0.9 nm size range changed from 15% to 10%.
There is a very low fraction of activity in 1.5-15 nm size range for background and it increases to
10% because of cooking. A large increase (from 35% to 70%) of 218Po is observed in the
"attached" mode peaked in 50-500 nm size range with only small changes (from 40% to 50%) in
214Pb and 214Bi distributions.
Because of the large amount of particle generated by normal activities in the domestic
environment, the working level will increase for a period of time while the "unattached" fraction
will decrease. The particles generated from cigarette smoke and cooking dramatically shifted
almost all of radon progeny to "attached" fraction and remained for a long period of time. The
particles produced from candle burning and vacuuming were much smaller with the average
attachment diameter around 15 nm. The candle and vacuuming particles did decrease the
"unattached" fraction, but returned to the original background distributions about 150 minute later.
35
-------
0
It
V
1
$
O.BO
0.60
0.40
0 20-
n nn.
n
AJ\_
^
Rn -222 : 15.97 pCi/l
CN cone. : 10,000/cm3
CD Po-218 : 13.95 pCl/l (2.0 X)
A; — A Pb-214 : 6.92 pa/m.7 *}
* * Bl -214 : 5.50 pCi/l (2-7 *)
a a PAEC : 70,0 mWL (0.4 *)
_A
A^^^-J^*~~*| "^* '— MJ
^fc— -ff^***5^
— '^=-^' ^ 1 |
0.5 1.0 10.0 100.0
Porttclo Diameter (nm)
A: Background Size Distributions.
1.00
0.80
0.60
0.40
0,20
0.00
Rn -222 : 2550 pCl/l
CN cone. ! 4QjOOO/cm3
0.5 1.0 10.0
Particle Diameter (nm)
B: 5 minutes later.
100.0
1,00
0.80
0.60
I" 0.40 ^
0.20
0.00
Rn -222 : 33.71 pCi/l
CH cone. : 23,OOO/cm3
Cn Po-218: 30.80 pCI/l (1.5 «)
A—A Pb-214 : 25.98 pCi/l (1.1 X)
» A Bl -214 : 17.98 pCl/l (1.5 V)
a a PAEC : 230.6 mVrt. (0,2 Si)
1.00
0.5 1.0 10.0 100.0
Particle Diameter (nm)
0.80
0.80-
0.40-
0.20-
0.00
Rn -222 : 39.05 pCi/l
CN cone. : 1O.OOO/cni3
CD Po-21 B : 32.67 pCl/l (1.5 X)
* * Pb-214 : 28.53 pCl/l (1.1 X]
»—A Bi -214 : 25.53 pCiVI (1.3 %\
a a PAEC : 273.5 mWL (0.2 X)
0.5 1.0 10.0 10O.O
Particle Diameter (nm)
Figure 4.
kitchen.
C: 80 minutes later. D: 155 minutes later.
Activity size distributions measured in a bedroom during and after cooking in the
SUMMARY
A measurement system has been developed for the purpose of characterizing indoor radon
decay product radioactivity on an automated, semi-continuous basis. The system was designed and
calibrated for sampling both the ultrafine cluster and attached modes of radioactivity present in
indoor air, with the optimized design based on the results of both experimental and numerical
simulation studies. The measurement system is capable of monitoring changes in activity
concentrations and size distributions as a function of time and indoor events. Further
measurements with the system should permit an improved estimation of the health hazards from
radon decay products in indoor air.
ACKNOWLEDGMENTS
This work was supported by the U.S. Department of Energy under contract DE-FG02-
89ER60876, and the New Jersey Department of Environmental Protection under contract J89-62.
We would also like to gratefully acknowledge the assistance of Rian Strydom and Chih-Shan Li
in the experimental work, as well as the collaboration of K. Gadsby, A Cavallo, and R. Socolow
of Princeton University in the field experiments.
REFERENCES
1. AC Chamberlain, E.D. Dyson, "The dose to the trachea and bronchii from the decay
products of radon and thoron," Brit. J. Radiol. 29:317-325 (1956).
36
-------
2. Y.S. Cheng, H.C. Yeh, Theory of screen type diffusion battery," J. Aerosol Sci. 11:313-319
(1980).
3. Y.S. Cheng, J.A. Keating, and G.M. Kanapilly, Theory and calibration of a screen-type
diffusion battery," J. Aerosol Sci. 11:549-546 (1980).
4. R.F. Holub, E.O. Knutson, In: Radon and its Decay Products: Occurrence. Properties and
Health Effects. Hopke, P.K., Ed., Washington D.C.: American Chemical Society, pp. 341-
356 (1987).
5. A.C. James, In: Radon and its Decay Products in Indoor Air. Nazaroff, W.W. and Nero,
A.V., Eds., New York: Wiley-Interscience, 259-309 (1988).
6. L.M. Kulju, M. Ramamurthi, P.K, Hopke, The detection and measurement of the activity
size distribution of ultrafine particles," Paper No. 86-40.6, Air Pollution Control
Association, Pittsburgh, PA (1986).
7. E.F. Maher, N.M. Laird, "EM algorithm reconstruction of particle size distribution from
diffusion battery data," J. Aerosol Sci. 16:557-570 (1985).
8. M. Ramamurthi, The detection and measurement of the activity size distributions (d >0.5
nm associated with radon decay products in indoor air. Ph.D. Thesis, University of Illinois,
Urbana-Champaign (1989).
9. M. Ramamurthi, P.K. Hopke, "On improving the validity of wire screen "unattached"
fraction radon daughter measurements," Health Phys. 15:189-194 (1989).
10. M. Ramamurthi, R. Strydom, P.K. Hopke, "Verification of wire and tube penetration
theories using a 218PoOx cluster aerosol," J. Aerosol Sci. (in press, 1990).
11. A. Reineking, J. Porstendorfer, "High-volume screen diffusion batteries and «-
spectroscopy for measurement of the radon daughter activity size distributions in the
environment," J. Aerosol Sci. 17:873-879 (1986).
12. H.G. Scheibel, J. Porstendorfer, "Penetration measurements in the ultrafine particle size
range," J. Aerosol Sci. 15:549-556 (1984).
13. J.C. Strong, "The size of attached and unattached radon daughters in room air," J. Aerosol
ScL 19:1327-1330 (1988).
14. R.J. Tremblay, A. Leclerc, C. Matthieu, R. Pepin, M.G. Townsend, "Measurement of
radon progeny concentration in air by a-particle spectrometric counting during and after
air sampling," Health Phvs. 36:401-411 (1979).
15. S. Twomey, "Comparison of constrained linear inversion and an iterative nonlinear
algorithm applied to the indirect estimation of the particle size distribution," J. Comput.
Phvs. 18:188-200 (1975).
16. H.C. Yeh, Y.S. Cheng, M.M. Orman, "Evaluation of various types of wire screens as
diffusion battery cells," J. Coll. Int. Sci. 86:12 (1982).
37
-------
THE ATMOSPHERIC STABILITY OF POLYBROMINATED
DIBENZO-p-DIOXINS AND DIBENZOFURANS
Christopher C. Lutes, M. Judith Charles and Richard M. Kamens
Department of Environmental Sciences and Engineering, School of Public
Health, University of North Carolina at Chapel Hill,Chapel Hill, North
Carolina, 27599-7400.
ABSTRACT
Polybrominated dibenzo-p-dioxins (PBDDs) and polybrominated
dibenzofurans are generated during the combustion of brominated organics
such as polybrominated diphenyl ethers (PBDPEs) which are used as flame
retardants in plastics, carpets and other materials (Buser 1986). Due to
increasing use of PBDPEs, atmospheric emissions of PBDDs and PBDFs
are likely to increase in the future. Regulating emissions of these
compounds requires information on their atmospheric stability because
biological exposure and uptake occurs via atmospheric transport and
depositional processes. Laboratory experiments indicate that PBDDs and
PBDFs rapidly photodegrade in solution and also degrade when sorbed on
quartz surfaces. Unknown are the rates of PBDD and PBDF
photodegradation under realistic outdoor conditions. In this study, we
introduced emissions from the combustion of polyurethane foam containing
PBDPEs into 25 m3 outdoor Teflon film chambers. Concentrations of
tetra- and pentabrominated dibenzo-p-dioxins and dibenzofurans were
monitored over time by collecting and analyzing filter and adsorbent
samples. The results show that PBDDs and PBDFs are stable on soot
particles over periods of hours and suggest that photochemical production
of PBDFs and PBDDs from residual PBDPEs may occur.
INTRODUCTION
As incineration is becoming a more important form of disposal of
wastes an assessment of risks associated with toxic organic emissions
becomes vital. Polychlorinated dibenzo-p-dioxins have been a major focus
of scientific investigation and public concern due to the toxicity of
tetrachlorinated dibenzo-p-dioxin (TCDD) and tetrachlorinated
dibenzofuran (TCDF) in guinea pigs (LD50 = 0.6ug/kg and 5-10ug/kg,
respectively, Poland and Glover, 1977). Polybrominated dibenzo-p-dioxins
and dibenzofurans (PBDDs and PBDFs) are structurally similar compounds
which are potentially either as or more toxic than their chlorinated analogs
(Mason, 1987).
The combustion or photodegradation of polybrominated diphenyl
ethers (PBDPEs) produces PBDDs and PBDFs. Yields of 19% were
38
-------
observed for the thermolysis of pentabrominated diphenyl ether sorbed on
quartz surfaces at 630C and 0.5-1% at 530C (Buser, 1986).
Photodegradation of decabrominated diphenyl ether in hexane has been
shown to produce PBDFs with a 10-20% yield by UV or sunlight
irradiation (Watanabe and Tatsukawa, 1987).
In 1988, polybrominated diphenyl ethers were produced at a rate of 75
million pounds with production expected to increase at a rate of ten to
fifteen percent per year (Mazek, 1988). Due to increasing production and
use of incineration, emissions of PBDDs and PBDFs are likely to increase
in the future. The atmospheric behavior of these compounds affects
biological exposure through transport and deposition processes. Thus in
order to understand the potential impact of these emissions, information on
the atmospheric stability of PBDDs and PBDFs is needed.
A paucity of data exists on the decay of PBDDs and PBDFs once
formed under realistic atmospheric conditions. In laboratory experiments,
PBDDs and PBDFs readily degrade by denomination with half lives on
the order of minutes in solution and hours on surfaces (Buser, 1988). Work
has previously been done to evaluate the atmospheric stability of polycyclic
aromatic hydrocarbons on realistic soot particle surfaces (Kamens, 1987).
This work helps form the methodological basis for this research discussed
whose overall objective is to study the stability of PBDDs and PBDFs on
incinerator soot particles under realistic outdoor conditions.
METHODOCOGY
Polybrominated dibenzo-p-dioxins and furans were produced in a
small scale incinerator from known precursors (industrial grade
pentabrominated diphenyl ether as 8.5% w/w component of polyurethane
foam), introduced to a 25 m3 outdoor teflon chamber and allowed to age
within this captured air parcel (see Figure 1). During the aging period,
samples were collected on 47 mm T60 420 teflon impregnated glass fiber
filters followed by a 4 x 1.5" polyurethane foam trap (see Figure 1).
Additional atmospheric variables monitored include ozone concentrations,
nitrogen oxide concentrations, total solar radiation, temperature and
particle size. These results are summarized in Table 1. In the laboratory,
C12 labeled PBDDs and PBDFs were added as internal standards and the
samples were soxhlet extracted in toluene. The extracts were then passed
through a series of three gravity columns; acidic silica gel, florisil and
carbon/celite to separate PBDDs and PBDFs from other components in
the extract. The resulting extracts were analyzed by GC/MS by selective-
ion-monitoring at a resolving power = 10,000. The results for the tetra-and
penta- brominated dibenzo-p-dioxins (TBDDs and PeBDDs) as well as
tetra, penta and hexa dibenzofuran (TBDFs, PeBDFs and HxBDFs) are
reported on a congener basis.
39
-------
RESULTS AND DISCUSSION
The results of initial samples for two experiments conducted on
Dec. 20, 1989 and March 14, 1990 are reported. Comparison of these
results provides a means of assessing the starting points for the chamber
aging experiments. Concentrations varied by orders of magnitude for the
congener classes monitored following the order TBDF > TBDD , PeBDF
> PBDD > HxBDF (Figure 2). This is expected since the formation of
PBDFs from PBDPEs is an intramolecular process while the formation of
PBDDs from PBDPEs is an intermolecular process. The preponderance of
lower brominated species can be explained by the primary abundance of
pentabrominated diphenyl ether in the flame retardant material and
simultaneous debromination reactions occurring during thermolysis (Buser,
1986). Comparison of the initial vapor and paniculate samples shows that
in each case the paniculate concentration exceeds the vapor phase
concentration by at least two orders of magnitude (Figure 3).
Samples were obtained over the course of a two hour period in each
of these experiments (Figure 4). Concentrations of PeBDD remained
steady at approximately 0.6 ng/mg on Dec 20. PeBDD concentrations on
March 14 began at 1.8 ng/mg and remained steady or perhaps exhibited a
slight decline over time. Similarly PeBDF held steady at approximately 13
ng/mg during the Dec 20, 1989 experiment and held steady or slightly
decreased from a level of 68 ng/mg on March 14. Similarly the March 14
experiment showed HxBDF to be stable or slowly decaying from a
concentration of 540 pg/mg. These results suggest degradation half lives
on the order of many hours. This implies that Buser's quartz surface
experiments are a better predictor of rates on realistic particles then his
solution experiments.
Results from the other congener classes were more surprising.
TBDF concentrations (3-6 ug/mg) appeared to slightly increase midway
through the sampling period and then decline. This cannot be explained by
the debromination of PeBDF due to the differences in concentrations
between TBDF (ug/mg) vs. PeBDF (ng/mg) but may be evidence for
photolysis of PBDPEs that survived the combustion process in a reaction
similar to that observed by Watanabe and Tatsukawa (1987). TBDD
seemed to be steady or decay from a concentration of 140 ng/mg on Dec
20 and increase on March 14 from 9.2 to 12.9 ng/mg. Further experiments
are needed to determine if these trends are due to analytical variability or
photolytic production. Some rational exists for why photolytic production
may be occurring in the tetra congener classes and not in others.
Watanabe and Tatsukawa (1987) found the reaction to occur with
simultaneous debromination and our starting material was the pentabromo
diphenyl ether primarily. On March 14 the initial concentration of TBDD
was relatively low perhaps allowing formation processes to outweigh
40
-------
degradation processes for a time. It is important to note that Watanabe
and Tatsukawa (1987) observed photolytic formation for PBDF and not
PBDD. However, it is possible that the availability of oxygen donating
species needed for the formation of dioxins is less in hexane solution then
on soot particle surfaces under atmospheric conditions.
Aerosol size distributions monitored from the December 20
experiment revealed that though the total number of particles in the
chamber decreased over time (probably due to air infiltration, removal of
sample and diffusion to the walls) the overall size distribution remained
fairly constant (Figure 5).
CONCLUSIONS
Additional work is needed to deconvolute rates for the processes that
seem to be occurring; degradation through debromination of PBDDs and
PBDFs and formation of TBDDs and TBDFs from PBDPEs. However
several conclusions can be reached from this initial work. At 650C-675C
substantial quantities of PBDD and PBDF are produced from the
combustion of small amounts (0.59 to 0.7 g) of polyurethane foam
containing polybrominated diphenyl ethers. Paniculate phase
concentrations of PBDDs and PBDFs seem to remain relatively steady over
periods of hours under realistic atmospheric conditions. The data suggest
that a photolytic production of TBDDs and TBDFs from PBDPEs may be
occurring. In addition, procedures and protocols developed can be applied
to study the atmospheric stability of other toxic organics emitted during
incineration.
ACKNOWLEDGEMENTS
We want to thank Randall Goodman for help in designing and
constructing the ignition vessel, G. Dean Marbury for assistance in mass
spectrometry analyses, Jay Odum for technical assistance and The
Brominated Flame Retardant Industry Panel for providing funds for
chemical standards. Primary funding for the project was received from The
EPA Research Center for Waste Minimization and Management, N.C.
State University.
REFERENCES CITED
Buser, H.R. (1986) Polybrominated Dibenzofurans and Dibenzo-p-dioxins:
Thermal Reaction Products of Polybrominated Diphenyl Ether Flame
Retardants. Environmental Science and Technology, 20, 404-8.
Buser H.R. (1988) Rapid Photolytic Decomposition of Brominated and
Brominated/Chlorinated Dibenzodioxins and Dibenzofurans. Chemosphere,
17:889-903.
41
-------
Kamens R.M., Guo, Z., Fulcher, J. N. and Bell, D. A. (1988) Influence of
Humidity, Sunlight and Temperature on the Daytime Decay of
Polyaromatic Hydrocarbons on Atmospheric Soot Particles. Environmental
Science and Technology, 22(1): 103-8.
Mason, G. M, Denomme, A., Safe, L. and Safe, S. (1987) Polybrominated
and Chlorinated Dibenzo-p-Dioxins: Synthesis Biologic and Toxic Effects
and Structure-Activity Relationships. Chemosphere, 16:8/9, 1729-31.
Mazek, C. 1988. Great Lakes Chemical Corporation, personal
communication.
Poland A, and Glover E. (1977) Chlorinated biphenyl induction of aryl
hydrocarbon hydroxylase activity: A study of the structure- activity
relationship. Mol Pharamacol 13, 924-38.
Watanabe, I. and Tatsukawa, R. (1987) Formation of Brominated
Dibenzofurans from the Photolysis of Flame Retardant Decabromobipheny
Ether in Hexane Solution by UV and Sunlight. Butt. Environ. Contam.
Toxicol 39:953-9.
42
-------
EMtQMDbMT
(26 rf)
Flgun l:S
ConqMri
CHAMBER TEMPERATURE
DBWPOINT (t»
[NO] (ppa)
[NOJ (HMD)
tOJ (ppm)
INCINERATOR
TOTAL SOLAR
RADIATION
(Ctl cn^min'1)
Ow. 20
1968
510-2
-6.0
0.040
0.036
0.019
060
0.7
to 0.06
for
ao
u
0.1
0.1
2
o.
to
rimenial append*.
'IBDD PsBDD TBDF PeBDF HxBDF
s
TBDD
TBDF
PeBDF
DEC 50. 1989
MARCH 14,1990
VAPQH PHASE
PARTICLE PHASE
2: CompvUon erf (he fint filter Moople
Flgura 3: Coopariwa of T«par nl partoto pb
u tor the initial Mmpk tat ttw
Dae. 80, 1089 axperinmit.
Oin?/ce)
'•ja TIME (pm)
5:43
6;42
O.M22 0.075 0.133 0.237 D.4?2 0.7C
Particle Diameter (Jim)
FlgmC: Yob
In niintrihar during ttieoourae of tto Dec. 80.1080 opaimeni.
43
-------
160
,2, 80
t
8 40
a „
2
s5» 1 -
a 0.5 -
fl o
1 TBDD -\ _-****
~^,. ^ —
CH 14. 1«W (1(W
I _^*^ "~ DECEMBER 30. I960
0123
Chamber Aging Time (hrs.)
PeBDD *-. MARCH 14, 1090
DECEMBER 20 1989
0123
Chamber Aging Time (hrs.)
f> 8OO
-| 600 -
* 4OO -
| 200 -
n
HxBDF
_3» 0
-gj 60-
j» 6O-
>S 40 -
§ 20 -
5
U o
"Tgrjp ^ — ____MARCH 14,1990
-"" DECEMBER 20 1»W ^^° ^~°
0 1 2 S
Chamber Aging Time (hrs.)
PeBDF MARCH 14.1990 ^^^-^^
DECEMBER 20 1989
0122
Chamber Aging Time (hrs.)
\__™o
0123
Chamber Aging Time (hrs.)
Figure 4: Concentrations of PBDDs and PBDFs during
chamber aging period.
-------
OCCURRENCE AND VAPOR PARTICLE PARTITIONING OF HEAVY ORGANIC
COMPOUNDS IN BRAZZAVILLE, CONGO
Barnabe Ngabe, Terry F. Bidleman
Department of Chemistry
University of South Carolina
Columbia, South Carolina 29208
INTRODUCTION
Organochlorine (OC) insecticides are used in several tropical and
subtropical countries for malaria control and for commercial agriculture.
Whereas DDT has been banned in North America and Europe, Third World
countries still use DDT in the war against malaria and sleeping sickness
and in agriculture (1-3). In India and Central America re-emergence of
malaria has accompanied the rapid spread of mosquito resistance caused by
elevated application of DDT during the late 1960s and 1970s (4). Dispersal
of pesticides through the atmosphere is responsible for environmental
contamination on a regional and global scale (5-8) . Reports of high
levels of DDT in the Indian troposphere have shown that India has become a
point source (9-10). DDT consumption in Africa in the period between
1980-1989 was 1294 tonnes (11). In Zimbabwe, wildlife has been
contaminated with the DDT used in the irradication of the tse-tse fly (3).
High levels of DDT in plants from the west coast of Africa is indicative
of local use (12). DDT volatilization from tropical soils has been
reported (13). DDT residence time in soils under tropical conditions is
relatively short because of volatilization which is considered to be the
most important pathway of dissipation in atmosphere (14-15). Is Africa
another point source of DDT and other OC insecticides? To the best of our
knowledge no information exists on atmospheric transport of pesticides in
Africa. This work was carried out to determine the types and levels of
airborne pesticides, polychlorobiphenyls (PCBs) and polycyclic aromatic
hydrocarbons (PAHs) in Brazzaville, Congo, and to assess the distribution
of these compounds between the particle and gas phases in atmosphere.
EXPERIMENTAL
Air samples (331-680 m ) were collected during August-September, 1989
at the top of the American Cultural Center in Brazzaville, Congo (4.14 S,
15.14 E) using a high volume sampler containing two glass fiber filters
and two polyurethane foam plugs. The_average sampling temperature was
25 C. The average TSP was 56 //g m . Samples were shipped to the
University of South Carolina where they were extracted and analysed by
capillary GC with electron capture detection (OCs) and GC-MS with selected
ion monitoring {PAHs). Collection and analytical methods are described
elsewhere (16-17).
RESULTS AND DISCUSSION
a) Organochlorine Pesticides
The mean concentrations (ng m~ ) of DDT-related compounds,
hexachlorocyclohexanes(HCHs), and chlordanes are shown in Table 1 and are
compared with those found from various locations of the world in Table 2.
Total DDT in Brazzaville (3.0 ng m ) ranged between levels reported in
Porto Novo and New Delhi (South-India) where DDT is still used, and was
30-100 times higher than the levels found in European and North American
countries where DDT is not used. In Brazzaville, p,p'-DDT was the
prevalent isomer followed by p,p'-DDD, o,p'-DDT and p,p'-DDE.
Africa consumed 5213 tonnes of HCH products from 1980 to 1989 (11).
45
-------
The levels of total HCH (0.39 ng m ) in Brazzaville were comparable to
those reported from European and North-American cities and 40 to 3000
tiroes smaller than those reported from India (Table 2). With a mean
concentration of 0.35 ng m , -^HCH was the predominent isomer. HCH
products are used in many countries for antimalarial and agricultural
purposes. In central Europe where pure lindane (99.5% r-HCH) is used, Y-
HCH concentrations exceed those of o-HCH (18-20). In India technical HCH
(70% ct-HCH, 14% Y-HCH, 9% S-HCH and 7% 8-HCH) is largely used and in most
cases a-HCH > v-HCH (9,10). In Brazzaville, ratios of r-HCH to a-HCH
ranged between 6.8 to 12.7 with an average of 9.3. This high r/a ratio
can only be due to the use of y-HCH and not technical HCH.
Typical Southern Hemisphere levels of chlordane (ng nf ) have been
reported (8,21). Eastern Indian Ocean: 0.02, western Australia: 0.027,
Reunion: 0.013-0.027. In Brazzaville the mean concentration of chlordane
(cis- + trans-chlordane + trans-nonachlor - 0.04 ng m ) was comparable to
the levels in the atmosphere of the open Southern Hemisphere oceans, and
15 - 43 times smaller than in Columbia, SC and Hyogo, Japan. Major
chlordane use in the U.S.A and Japan has been for termite control and has
led to fairly high levels in ambient air. The fact that chlordane
concentrations in Brazzaville are similar to open-ocean concentrations
suggests that little or no local use occurs.
b) PCBs and PAHs
-3
The mean PCB concentration was 0.60 ng m . This concentration is
slightly lower than observed in U.S. cities and closer to that reported
from Hyogo, Japan. Figure 1 shows the distribution of PCB congeners in
the air of Brazzaville. The mean concentrations of PAHs are also shown in
Table 1. As with PCBs, PAHs with higher vapor pressures predominated.
For example, phenanthrene represented more than 45% of the total PAH. PAH
levels in Brazzaville were almost six times lower than those reported from
Osaka, Japan and comparable to those found in U.S. cities (Table 2).
Table 1. Oxganochlorine Pesticides, iFCB and PAH in Brazzaville Air
Pesticides and IPCB
Cojnpound
p/P'-DDT
o,p'-DDT
p,p'-DDD
p,p'-DDE
rHCH
a-HCH
trans-chlordane
cis-chlordane
trans-nonaclor
LPCB
ng/m3
1.26
0.59
0.73
0.45
0.35
0.04
0.01
0.02
0.01
0.60
EAH
Compound
PH
MePH
AN
FLA
PY
BaA
CHRY
BbF
BkF
BeP
BaP
BghiP
ng/m3
21.60
5.32
1.94
5.36
5.84
0.48
1.10
0.76
0.63
0.44
0.25
0.85
PH = Phenanthrene
MePH = Methylphenanthrene
AN = Anthracene
FLA = Fluoranthene
PY = Pyrene
BaA = Benz(a)anthracene
46
CHRY = Chrysene
B(b)F = Benzo(b)Fluoranthene
B(k)F = BenzodOFluoranthene
B(e)P = Benzo(e)pyrene
B(a)P = Benzo(a)pyrene
B(ghi)P = Benzo(ghi)perylene
-------
Table 2. Concentrations Pesticides, PCBs and PAHs from Different Locations, ng/m3
Locations and Survey Year
Brazzaville, Congo
1989
Delhi, India
1980-1982
Porto-Novo
1987-1989
Columbia, SC (USA)
1977-1980
Denver, CO (USA)
1980
1985
Portland, OR (USA)
1984-85
Delft, Netherlands
1980
Paris, France
1986-87
Ulm, West Germany
1986
Southern Sweden
1983-85
Hyogo, Japan
1988
Osaka, Japan
1982
Reunion, Southern Indian
Ocean 1986
EDDT
3.03a
46-73a
0.28a
0.14b
0.02C
0.03a
QJOtf
0.12a
<0.05-0.22C
0.025°
0.007
0.03C
1HCH
0.39
160-930
15
1.10
0.30
0.34
0.61
1.6
3.0
0.49
0.41
Chlordane
0.04
0.60
0.07
0.06
0.01
1.7
0.015
LPCB
0.60
30
2.2
2.0
0.%
5-44
0.29
0.45
0.03f
PH+AN+FLA+PY
35
54
79
46
<8.5-11.5
185
BeP+BaP
0.70
0.508
1.78
2.4
<0.3-6
11.2
References
This work
9
10
17,27
17
26
29,22
19
20
18
17
28
30
8
a = pp'-DDT+op'-DDT+p^'-DDD+pp'-DDE
g = only BaP
b = p,p'-DDT+p,p'-DDE
f = congener (28-180)
c = p,p'-DDE
-------
00
co oo r- i*-*-1
c) Vapor-Particle Relationships.
To assess the removal mechanisms, reactivity and health effects due
to inhalation of organic compounds in the atmosphere, the study of the
vapor-to-particle partitioning is required (22). Many researchers have
shown that the relationship between the apparent vapor-to-particle ratio
and the liquid vapor pressure of the organic compounds (P .) at a given
temperature is:
Log (A/F)TSP - mLogP°L + b (1).
TSP is the particle concentration (//gin~ ). A and F are, adsorbent and
filter-retained concentrations of organic compounds (ng m~ ). These were
calculated from quantities found on the two filters and PUF plugs using
relationships given by Ligocki and Pankow (22). Average percentages of
particulate OCs and PAHs were: p,p'-DDT » 5, o,p'-DDT + p,p'-DDD « 3,
p,p'-DDE • 1, BeP+ BbF+BkF > 80, chrysene - 20, fluoranthene - 1.2,
phenanthrene - 0.40. Plots ol (1) for OCs and PAHs are shown in Figures 2
and 3. Values of P were from the literature (23,24). Slopes,
intercepts, and R values were: OCs: m - 0.74, b - 5.76, R - 0.97. PAHs:
m - 0.85, b - 5.29, R^ - 0.96.
Figure 4 shows a comparison between the fraction of particle-
associated organic compounds as described by Junge-Pankow (J-P) model and
by Brazzaville data. The J-P model is:
Where c » 17.3 pa-cm (22); S_~ Particle surface area per volume of air
(cm /cm ). The curves for Brazzaville OCs and PAHs were calculated from:
4> = 1/[1+(A/F)] (3).
The field data for PAHs were very close to J-P curve. The differences
between PAHs and OCs suggest different strengths of adsorption. This was
seen in a comparison of PAH data from Tokyo and OC data from Columbia and
Stockholm (25) but not seen for PAHs and OCs in Denver (26). However the
fractions of OCs on particles in Brazzaville were very small and the
liquid vapor pressure range of OCs examined was much less than for the
PAHs.
ACKNOWLEDGEMENTS
Support was provided by the International Agency for International
Development and the National Science Foundation. We are grateful to the
authorities of the American Cultural Center in Brazzaville for allowing us
to use their building.
48
-------
FIGURE 2
CO
E-^
-------
KEFEKENCBO
1) Filial, M. K., Samuel, T. (1989). Environ. Pollut.,57, 63- 77.
2) Ultra., J., Raghu, K. (1989). Environ. Pollut., 61, 157-
170. ~~
3) Matthielson, P. (1985). Environ. Pollut. Ser. B., 10, 189-
211. ~~
4) Chapin, G., Wasserstrom, R. (1981). Nature, 293, 181-185.
5) Atlas, E. L., Giam, C. S. (1988). Water, Air, Soil Pollut.,
38, 19-36.
6) Pacyna, J. M., Oehme, M. (1988). Atroos. Environ., 22, 243-
57. *~
7) Patton, G. W., Hinckley, D. A., Walla, M. D., Bidleman, T.
F. (1989). Tellus, 41B, 243-255.
8) Wittlinger, R., Ballschmlter, K. (1990). Fresenius Z. Anal.
Chem., 336, 193-200.
9) Kaushik"7~ C. P., Pillai M. K. K., Raman, A., Agarwal, H. C.
(1987). Water, Air, Soil Pollut., 32, 63-76.
10) Ramesh, A., Tanabe,S., TatsuTcawa, R., Subramanian, A. N.,
Palanicharay, S., Mohan, D., Venugopalan, v. K. (1989). Environ.
Pollut. 62, 213-222.
11) Ottar, B., Semb, A. (1989). Proceedings of the 8th International
Conference of the Comite Arctique International (CAI). Oslo, Norway,
18-22 September.
12) Bacci, E., Calamari, D.f Gaggi, C., Biney, C., Focardi,
Morosini, M. (1988). Chemosphere, 17, 693-702.
13) Yeadon, R., Perfect, T. J. 7/1981). Environ. Pollut., B, 2,
275-94.
14) Sleicher, C. A., Hopcraft, J. (1984). Environ. Sci., 22, 1-
15. ~
15) Samuel, T., Agarwal, H. C., Pillai, M. K. K. (1988). Pestic.
Sci., 21, 1-15.
16) Patton, G. W., Bidleman, T. F. J. Geophys. Res. (submitted).
17) Billings, W. N., Bidleman, T. F. (1983). Atmos. Environ., 17, 383-
91. ~~
18) Wittlinger, R., Ballshmiter, K. (1987). Chemosphere, 16, 2497-513.
19) Guicherit, R., Schulting, F. L. (1985). Sci. Tot. Environ., 43,
193-219.
20) Chevreuil, M., Chesterikoff, A., Letolle, R., Granier, L. (1989).
Water, Air, Soil Pollut., <43, 73-83.
21) Kawano, M., Tanabe., S., Inoue T., Tatsukawa, R. (1985).
Transactions of the Tokyo University of Fisheries, 6, 59-66.
22) Ligocki, M. P., Pankow, J. F. (1989). Environ. Sci. Technol., 23,
75-83. ~~
23) Hinckley, D. A., Bidleman, T. F., Foreman, W. T. (1990). J. Chem.
Eng. Data, (in press).
24) Yamasaki, H., Kuwata, K., Kuge, Y. Nippon Kagaku Kaishi (1984),
1324-29, Chem. Abst.'101, 156747p.
25) Bidleman, T. F., Billings, W. N., Foreman, W. T. (1986). Environ.
Sci. Technol., 20, 1038-1042.
26) Foreman, W. T., Bidleman, T. F. (1990). Atmos. Environ, (in
press).
27) Keller, C. D., Bidleman, T. F. (1984). Atmos. Environ., 18, 887-
845. ~
28) Nakano, T., Tsuji, M., Okuno, T. (1989). Atmos. Environ, (in
press).
29) Ligocki, M. P., Leuenberger, C., Pankow, J. F. (1985). Atmos.
Environ., 19, 1619-26.
30) Yamasaki, H., Kuwata, K., Miyamoto, H. (1982). Environ. Sci.
Technol., 16, 189-94.
50
-------
Heterogeneous Reaction of Nitrogen Oxides
on Sea Salt and Mineral Particles -
A Single Particle Approach
Y. Mamane and J. Gottlieb,
Environmental Engineering, Technion
Haifa 32000, Israel
Heterogeneous reactions of NO. and HNO3 at sub ppm levels with sea salt
and mineral particles were investigated. Experiments were conducted in a
static reaction chamber made of Teflon, where particles deposited on filters
and on electron microscope grids were exposed to NO_ or HNO- under
controlled conditions. Nitrates formed on the particle loaded filters were
determined by bulk analysis. In parallel, sea salt and mineral particles
were observed in an electron microscope to detect the presence of nitrate
on the surfaces of each individual particle. Microspot techniques were
applied for the latter.
Under the present experimental conditions the formation of nitrates on
sea salt particles was in the range of 0.1 to 3.3 mg NO,~/g NaCl. Slightly
higher values were obtained for mineral particles: 0.2 to 4.7 mg NO,~/g
aerosol ([NO,,]=0.18 ppm and 0.54 ppm; [HNO3J=0.04 ppm; exposure time 1 to
8 days; relative humidity =70%).
Application of electron microscopy and specific microspot techniques
provided direct evidence for the heterogeneous reaction of sea salt and
mineral particles with NO, and HNO- to form a layer of nitrate on the
particle surfaces. Forty to 50% of the soil and almost all sea salt
particles that were exposed to NO_ and HNO- form mixed nitrate particles.
51
-------
Introduction
The presence of ammonium nitrate in the fine particle mode and nitrates
associated with sodium or minerals in the coarse mode has been documented
by several groups . The fine particles are formed through homogeneous
gas phase, while the coarse mode is formed through the reaction of sea salt
and mineral particles with nitrogen oxides. No direct evidence was
provided. Other heterogeneous processes for nitrate formation involve
reaction of gaseous nitric acid with sea salt particles .
It is the purpose of this research to use electron microscopy techniques
to study the reactions between sea salt or mineral particles and nitrogen
oxides or nitric acid, and to provide direct evidence for the formation of
nitrates on the surfaces of ambient particles.
Experimenta1
Laboratory studies of heterogeneous reactions between atmospheric
particles and nitrogen oxides were carried out in a static Teflon chamber,
under controlled conditions and no illumination. NaCl, CaCO_, sand, soil,
and sea salt particles were deposited on electron microscope grids, and on
47 mm Nuclepore and Teflon filters. The latter were used for bulk
analysis. The particles were exposed to known concentrations of NO. {0.18
and 0.54 ppm) or HNO3 (0.04 ppm) at 70% relative humidity (RH), using
permeation tubes. Exposure varied from 1 to 8 days. Any formation of
nitrate on loaded filters were determined by standard wet chemistry methods.
Electron microscope grids were investigated using microspot and electron
microscopy techniques . Bulk analysis allowed to quantify nitrate
formation in terms of mg NO-~/g aerosol, while microscopy provides direct
evidence of such formation.
Results and Discussion
Bulk Analysis
In this study changes in relative humidity (40% to 85%) were found to
have a negligible effect on nitrate formation. A 70% RH was therefore
chosen for all experimental runs. Figure 1 is a plot of the nitrate
formation for different aerosols exposed to 0.54 ppm and 0.18 ppm NO- from
one to 7 days. The Figure indicates that exposure of particles to higher
NO- concentrations results in higher nitrate capacities. Capacities (mg
NO_/g aerosol) are defined as the maximum quantity of nitrates in mg formed
on one gram of aerosol. Figure 2 shows nitrate formation versus time for
exposure to 0.04 ppm HNO, at 70% RH. Nitrate capacities for sea salt
(including NaCl) and minerals were around 3 and 3 to 5 mg NO.~/g aerosol
respectively.
Individual Particle Analysis
Particles collected on Nuclepore filters were observed in the microscope
before and after they were exposed to NO_ and HNO,. No significant changes
52
-------
in particle morphology were detected. The particles collected on electron
microscope grids, after exposed to nitrogen oxides, were coated with a thin
layer of nitron in order to detect the presence of nitrate formation on
the particle surfaces. Forty five to fifty percent of the mineral particles
reacted with either NO- or HNO- to form mixed nitrate-mineral particles.
Figure 3 is a photomicrograph of a mineral particle that reacted to form
nitrate on its surfaces. The mineral particle is surrounded by
characteristic fibers, the result of nitrate reaction with the nitron film ;
thus this micrograph provides direct evidence for the nitrate formation on
the particle surface. Not all minerals had reacted to form nitrates.
Particles that formed nitrate seem to be identical to those without any
nitrate formation. Only that the later group had a larger size range.
Almost all sea salt and NaCl particles reacted with the above gases as
expected, following the well known equation :
HNO
>3(g) + NaCl (s) = NaNO3(s) + HCl(g) .
In this case the reaction was not always on the surface but rather a
reaction with the whole particle mass.
Acknowledgements
This research was supported by grant No. 013-107 from the National
Council for Research and Development, Jerusalem, Israel, and GSF, Munich,
Germany. The assistance of Mrs. E. Melamed is gratefully acknowledged.
References
1. Yoshizumi, K. and A. Hoshi (1985), "Size distribution of ammonium
nitrate and sodium nitrate in atmospheric aerosol". Environ. Sci. Technpl.
19, 258-261.
2. Wall, S.M., W. John, and J.L. Ondo (1988), "Measurement of aerosol size
distributions for nitrate and major ionic species", Atmos. Environ. 22.
1649-1656.
3. Savoie D.L., and J.M. Prospero (1982), "Particle size distribution of
nitrate and sulfate in the marine atmosphere", Geophvs. Res. Letters 9.
1207-1210.
4. Harrison, R.M. and C.A. Pio (1983), "Size differentiated composition of
inorganic atmospheric aerosols of both marine and polluted continental
origin", Atmos. Environ. 17. 1733-1738.
5. Wolff, G.T. (1984), "On the nature of nitrate in coarse continental
aerosols", Atmos. Environ. 18. 977-981.
6. Mamane, Y. and M. Mehler (1987), "On the nature of nitrate particles in
a coastal urban area", Atmos. Environ. 21. 1989-1994.
7. Mamane, Y. and R.F. Pueschel (1980), "A method for the detection of
individual nitrate particles", Atmc-a. Environ. 14, 629-639.
53
-------
8
2
.0) 3
?J
O
D)
2-
O
'•§
I 1
i
0
0.18 ppm 0.54ppm--
3 5
Exposure times in days
CaCO3
Soil
Sea salt
Figure 1. Formation of nitrate, in terms of mg nitrate per g of aerosol,
on carbonate, soil and sea salt aerosol as a function of exposure time. NO_
concentrations were 0.18 and 0.54 ppm.
54
-------
5
o
(0
CO
O
en
g
is
,0
I
4.5-
4-
3,5-
3
2.5-
2-
1.5-
1-
0.5
3 . 5
Exposure times in days
-»~ NaCI
-H- Soil
-«— CaCO3 -»i
-x- Sea salt
*- Sand
Figure 2. Formation of nitrate, in terms of mg nitrate per g of aerosol,
on NaCI, carbonate, sand, soil and sea salt aerosol as a function of
exposure time. HNO3 concentration was 0.04 ppm.
55
-------
Figure 3. Typical reaction spot of mineral particle coated with a thin
layer of nitron. The reaction spot that surrounds the mineral particle,
appears as fibers of nitron-nitrate, is indicative of the presence of
nitrate on the mineral surface.
56
-------
REACTION PRODUCTS OF ALTERNATIVE CHLOROFLUOROCARBONS - THE
HYDROCHLOROFLUOROCARBONS
Bruce W. Gay, Jr. and Edward 0. Edney
U.S. Environmental Protection Agency
Gas Kinetics and Photochemistry Research Branch
Chemical Processes and Characterization Division
Atmospheric Research and Exposure Assessment Laboratory
Research Triangle Park, North Carolina 27711
Abstract
The reaction of HCFC-22 (CHC1F2) , HCFC-123 (CHC12CF3) , and HCFC-
141b (CHjCCljF) were studied to determine their oxidation products.
The reactions were carried out in the laboratory at low ppm reactant
concentrations using long path Fourier transform infrared
spectroscopy to monitor reactants and products. The laboratory
chemical initiation took place by photolyzing chlorine gas in the
presence of the HCFC. The chlorine atoms formed in the photolysis of
molecular chlorine abstract H atoms from the HCFC. Addition of
oxygen to the newly formed radical produces an HCFC peroxy radical
that undergoes further reaction. Major carbon-containing products
for the reaction of the HCFCs were: F2C(0) from HCFC-22, CF3C1C(0)
from HCFC-123, and CO and FC1C(0) from HCFC-141b. The products
observed are expected to be similar to those generated under ambient
conditions with OH radicals.
Introduction
Stratospheric ozone depletion by chlorofluorocarbons (CFCs) led to
the 1987 signing by many countries of the Montreal Protocol. The
protocol called for curtailing the production and eventual banning of
CFCs over the next decade. The phaseout of CFCs has prompted
development of hydrochlorofluorocarbons (HCFCs) that can undergo
atmospheric reaction. Replacement HCFCs include HCFC-22 (CHClF2) for
home air conditioners and foam food containers, HCFC-123 (CHCljFjC)
and HCFC-141b (CH3CC12F) as substitutes for CFC-11 (CC13F) in the
plastic foam industry. Alternative HCFCs are more reactive than CFCs
57
-------
because the hydrogen atom in their molecular structure can be
abstracted by tropospheric hydroxyl (OH) radicals. The subsequent
oxidation of the HCFC lead to new compounds having unknown
atmospheric fate and effects on stratospheric ozone.
The reactions of HCFC-22, HCFC-123, and HCFC-141b were studied in
the laboratory using long path Fourier transform infrared
spectroscopy. The initial abstractions of hydrogen atom from HCFC
by OH radicals in the atmosphere are simulated in the laboratory
using chlorine atoms. Chlorine atoms are produced by the
photodissociation of chlorine gas. In the absence of NOX and other
chemical species usually found in the atmosphere all of the oxidation
products generated by hydroxyl radical reactions will not be observed
in the laboratory with the use of chlorine atoms. Only those
products generated under ambient conditions resistant to further
reaction are observed in the laboratory system due to the highly
reactive nature of chlorine atoms. Therefore only the more stable
oxidation products are observed, however, these products are among
the products expected to be produced under normal atmospheric
conditions.
Experimental
The experiments were carried out in a 670 liter photochemical
reaction cylindrical chamber constructed of borosilicate glass 9 m in
length and 0.3 m diameter. To illuminate the reactants 72 black
light and 24 sun fluorescent lamps were evenly spaced around the
length of the chamber. The chamber is evacuable and contains an
eight mirror optical system set at 144 m for insitu infrared
absorption analyses. The chamber was optically coupled to a Digilab
Model 80 Fourier transform infrared spectrometer and used a Nernst
glower as an IR source and a mercury-cadmium-telluride LNX cooled
detector to cover the spectral range of 650 to 4000 cm"1. Standard
spectral absorption techniques were used to obtain data.
The HCFCs of high purity were obtained from E. I. DuPont DeNemours
and Company and used without further purification. The HCFCs were
added to 650 torr of zero grade tank air in the reaction chamber by
injecting jul liquid samples of HCFC-123 or HCFC-141b into a manifold
connected to the chamber inlets. Experiments with HCFC-22 a gas at
room temperature used a gas-tight syringe to add a pre-determined
amount to the manifold. A gas-tight syringe was also used to inject
chlorine (>99.9% pure J. T. Baker Chemical Company) into the
manifold. The HCFCs and C12 injected into the manifold were swept
into the cell with prepurified tank nitrogen. The initial conditions
of experiments were 13 ppm HCFC-220 and 12 ppm C12, 14 ppm HCFC-123
and 12 ppm C12, and 15 ppm HCFC-141b and 12 ppm C12. Experiments were
carried out at 700 torr total pressure and 24 ± 1°C. Before
irradiation the chamber contents were monitored by obtaining the
infrared spectrum of the reactants. After 10 minute periods of
irradiation, the chamber contents were again monitored via its
infrared spectrum. Substraction of unreacted HCFC from the
irradiated spectrum results in an IR absorption spectrum of the
products. Authentic samples of F2CO and CF3C(0)C1 were obtained from
PCR, Inc. and used to identify HCFC reaction products. All reaction
products were also identified from literature absorption band values.
58
-------
Results and Discussion
In the chlorine initiated oxidation of HCFC-22, the only products
observed in the infrared spectra after removal of unreacted HCFC-22
were HC1 and F2CO. The reaction sequence and formation can be
explained as follows:
C12 + ht -» 2 Cl (1)
Cl + HCC1F2 -> HC1 + CC1F2 (2)
CC1F2 + 02 -" C1F2C02 (3)
2(C1F2C02) -» 2(C1F2CO) + 02 (4)
C1F2CO -* Cl + F2C(0) (5)
Chlorine atoms formed in reaction 1 abstract a hydrogen from HCC1F2
(reaction 2) to form HC1 an observed IR product. In reaction 3 the
CC1F2 radical quickly adds 02 forming a peroxy radical. Because the
system does not contain NO or other species which the peroxy radical
can react, it reacts with itself as shown in reaction 4 forming the
haloalkoxy radical. The product F2CO observed and identified by its
IR absorption bands and comparison with an authentic F2CO sample
forms via the elimination of chlorine atom (reaction 5). Chlorine
atom elimination has been reported for other halogenated alkoxy
radicals and is a viable mechanism.
The oxidation of HCFC-123 via chlorine initiated reaction produced
HC1 and CF3C(0)C1. Both compounds were identified by IR bands from
the literature and the trifluoroacetyl chloride as compared with an
authentic sample. The formation of products can be explained by the
following reactions. The chlorine atoms formed in reaction 1
abstracts a hydrogen forming HC1 and the fluorochloro alkyl radical
(reaction 6).
Cl + CF3CHC12 -»• HC1 + CF3CC12 (6)
CF3CC12 + 02 -" CF3CC1202 (7)
2(CF3CC1202) -* 2CF2CC120 + 02 (8)
CF3CC120 -* CF3C(0)C1 + Cl (9)
Reaction 7 addition of 02 to form the peroxyradical, reaction 8 to
form the alkoxy radical and the elimination of chlorine atom in
reaction 9 are similar to reactions 3, 4, and 5. Thus the main
carbon-containing product observed in the oxidation of HCFC-123 is
trifluoroacetyl chloride. This compound is stable and is resistant
to further attack by chlorine atoms.
The infrared spectra of irradiated HCFC-141b/Cl2 mixtures showed
the formation of HC1, CO, and FC1C(0) as main products. The HCl can
be accounted for as in previous HCFC reaction 2 and 6. The resulting
HCFC peroxy radical in recombination with itself forms the alkoxy
radical (reactions 10, 11, and 12).
59
-------
Cl + CFC12CH3 •+ CFC12CH2 + HC1 (10)
CFC12CH2 + 02 -*• CFC12CH202 (11)
2(CFC12CH202) -» 2 CFC12CH20 + 02 (12)
The alkoxy radical can react with 02 to form an aldehyde or cleave
the C-C bond to form H2C(0) and CFClj.
CFC12CH20 + 02 -* H02 + CFC12CH(0) (13)
CFC12CH20 -» H2C(0) + CFC12 (14)
The CFC12 radical reacts with 02 forming a peroxy radical.
CFC12 + 02 -* C12FC02 (15)
Reaction of two peroxy radicals to form the haloalkoxy radical
followed by chlorine atom eliminations results in the formation of
fluorochlorocarbonyl FClC(O).
2(C12FC02) - 2 C12FCO + 02 (16)
C12FCO -* Cl + C1FC(0) (17)
The aldehyde CFC12CH(0) formed in reaction 13 is reactive toward
chlorine atoms in the system and loses its aldehydic hydrogen.
Cl + CFC12CH(0) -<• HC1 + CFC12CO (18)
The resulting radical adds 02 forming a peroxyacyl radical which re-
combines with itself to form a alkoxy radical,
CFC12CO + 02 -" CFC12C(0)02 (19)
2(CFC12C(0)02) -» 2CFC12C(0)0 + 02 (20)
The alkoxy radical dissociates to C02 and a CFC12 radical.
CFC12C(0)0 - C02 + CFC12 (21)
Chlorofluorocarbonyl is formed from the CFCl2 radical via reactions
15, 16, 17.
Formaldehyde formed in reaction 14 is quickly oxidized in this
highly reactive chlorine system to carbon monoxide an observed
product by the following reaction scheme.
H2C(0) + Cl -> HC1 + CH(0) (22)
CH(0) + 02 •* H02 + CO (23)
As in the case of F2C(0) and CF3C(0)C1 heterogeneous reactions
will be of importance in knowing the atmospheric lifetime and fate of
FC1C(0).
60
-------
Conclusion
The reactions of HCFCs were studied in the laboratory using
chlorine atoms to abstract hydrogen atoms as a way of simulating
tropospheric OH radical reaction. Due to the very reactive nature of
chlorine atoms, the absence in the laboratory experiments of NO and
other reactive species usually found in the atmosphere, only those
products resistant to further reaction were observed. Chlorine atom
initiated oxidation of HCFC-22, HCFC-123, and HCFC-141b resulted in
the formation of unique halogen containing carbonyl compounds. The
photolysis of chlorine in air and HCFC-22 resulted in the formation
of F2C(0). The chlorine atom initiated oxidation of HCFC-123 led to
the formation of CF3C(0)C1 as the only carbon containing product.
Reaction of HCFC-141b resulted in the formation of CO, C02, and
FC1C(0). More than one reaction sequence may have been responsible
for the formation of FC1C(0). The formation single carbon atom
containing molecules as products indicated C-C bond cleavage in the
reaction of HCFC-141b. Future work is needed, especially with
heterogeneous reactions, in order to determine the final fate of the
oxidation products arising from the degradation of HCFCs.
Disclaimer
Although the research described in this article has been supported
by the U.S. Environmental Protection Agency, it has not been
subjected to Agency review and therefore does not necessarily reflect
the views of the Agency and no official endorsements should be
inferred.
61
-------
QUANTITATIVE SUPERCRITICAL FLUID EXTRACTION COUPLED ON-LINE
TO CAPILLARY GAS CHROMATOGRAPHY FOR ENVIRONMENTAL APPLICATIONS
J.M. Levy*, A.C. Rosselli. D.S. Boyer, and K. Cross,
Suprex Corporation
SFC Research Center
125 William Pitt Way
Pittsburgh, PA 15238
The usefulness and ease of utilizing supercritical fluid extraction (SFE) directly coupled to
capillary gas chromatography (GC) as quantitative or qualitative analytical problem-solving tools
will be demonstrated. As an alternative to conventional liquid solvent extractions, SFE presents
Itself as a means to achieve high extraction efficiencies of different compounds In complex solid
matrices In very rapid time frames. Moreover, SFE has an additional advantage of being able to
achieve distinct extraction selectivltles as a function of the solubilizing power of the supercritical
fluid extracting phase. For on-line SFE/GC, the extraction effluent is directly transferred to the
analytical chromatograph. On-line SFE/GC Involves the decompression of pressurized extraction
effluent directly Into the heated, unmodified split capillary split Injection port of the GC. In this
respect, SFE introduction Into GC can be used as an alternative means of GC injection, compara-
ble to such modes of injection as pyrolysis and thermal desorptlon.
62
-------
PfTROpUCTIQN
Supercritical fluids have been used successfully for years for different Industrial applica-
tions (1). A large scale application of supercritical fluid extraction (SFE), for example. Is to Increase
crude oil recoveries from porous rocks In oil fields by pumping in gases such as carbon dioxide and
nitrogen. In this environment, the pressures and temperatures are high enough that supercritical
conditions exist and contribute to enhanced recoveries. Extractions using supercritical fluids are
attractive when compared to conventional liquid extractions for a number of reasons. While super-
critical fluids have solvent strengths that approach those of liquid solvents, they have lower
viscosities (10-4 N-sec/m2 versus 10-3 N-sec/m2) and higher solute diffusivities (10-4 c/m^/sec
versus 10-2 cm2/sec). These properties improve mass transfers from solid or liquid matrices and
thus significantly decrease the overall time needed for supercritical fluid extractions. By increasing
the density, the solvent strength of a supercritical fluid Increases. Therefore, conditions can be
optimized for the extraction of a specific solute or class of solutes from a complex matrix by
changing the extraction pressure or temperature. Close to the critical point of the supercritical
fluid, temperature or pressure changes dan change solute solubilities by a factor of 100 or even
1000. By using different supercritical fluids for extractions, such as carbon dioxide, nitrous oxide,
and sulfur hexafluoride, preferential extraction can be achieved for different solutes. Moreover, the
use of fluids that have low critical temperatures (I.e. CO% and ^O) allow extractions under
thermally mild conditions, thereby protecting thermally labile components. Since supercritical
fluids, such as CO^, ^O and SFg are gases at room temperature, off-line component collection or
concentration is greatly simplified. Because supercritical fluids undergo expansive (Joule-
Thompson) cooling upon decompression, even volatile components can be quantitatively and
efficiently collected Into solvents off-line after extractions. It Is also possible to directly Interface
supercritical fluid extraction with analytical chromatography, such as capillary gas chromato-
graphy (GC) and, supercritical fluid chromatography (SFC). Recent reports have demonstrated the
potential of using SFE as an alternative to time consuming, less efficient and less quantitative
conventional liquid solvent extraction techniques. Specific solutes ranging from environmental
priority pollutants to spices and fragrance components have been qualitatively and quantitatively
extracted using supercritical fluids from a variety of liquid and solid sample matrices (2-10). Direct
Interfaces of SFE to capillary GC and SFC (7-20) have been also demonstrated.
The benefits of directly coupling SFE to GC are that no sample handling Is required
between the extraction step and the GC separation step and that extraction effluents can be
quantitatively and reproduclbly transferred for on-the-fly analyses. When employing flame ioniza-
tlon detectors, no detector responses (I.e. solvent peaks) appear for reasonably pure supercritical
fluid grade CO2 or ^O. This permits the determination of volatile solutes which are often masked
by liquid solvents when using conventional extraction techniques. Moreover, when modifiers such
as methanol or propylene carbonate, are used to augment the solubllizing power of primary super-
critical fluids, they elute as distinct peaks In respective GC or SFC separations. The limitations of
coupling SFE to GC are defined by the volatility constraints of higher molecular weight solutes In
complex matrices that may not necessarily completely elute from GC columns.
This paper will demonstrate the applicability of SFE/GC techniques towards the quantita-
tive and qualitative characterization of some environmental matrices.
EXPERIMENTAL
On-line SFE/GC was performed on a Suprex Model SFE/50 stand-alone extractor
equipped with an electronic Valco four-port high pressure selector valve and a Hewlett-Packard
Model 5890 gas chromatograph equipped with a spllt/splitless capillary Injection port and flame
lonization detector.
63
-------
Figure 1 shows a schematic diagram of the SFE/GC Interface. The Suprex SFE/50 extractor
consists of a 250ml syringe pump with pressure limits up to 500 atm.
Healed Transfer Line
Extractor
Via
ilnjuctor
Pump
Control SFE Oven
GCOvun
Figure 1. On-Llne SFE/GC Schematic Diagram
The oven of the extractor was large enough to accommodate multiple extraction vessels or extrac-
tion vessels up to 50 ml In volume. The electronically actuated Valco four-port selector valve was
used to perform the static and dynamic extractions and to divert the extractor effluent flow Into the
Injection port of the GC. The controlling software of the Suprex SFE/50 permitted the automatic
operation of the four-port selector valve and automatically Initiated the run on the GC after
dynamic transfer of the extractor effluent. Both 1/32 Inch O.D. X 0.007 inch I.D. stainless steel
and 15 or 25 micron I.D. fused silica tubing have been used as transfer lines between the SFE/50
and the 5890 gas chromatograph. When stainless steel tubing was used. It was necessary to
restrict the flow by crimping to allow a flow of 40-80 ml/minute of expanded decompressed gas at
the specified extraction pressure. The transfer line tubing was Inserted 35-40 mm directly into the
spllt/splltless capillary Injection port which was kept at 225°C to minimize the Joule-Thompson
cooling which occurred when the supercritical fluid phase decompressed. For purposes of solute
focusing. It was also necessary to cryogenically cool the gas chromatographlc oven. The oven was
kept cool long enough to allow the dynamic transfer of the respective vaporized solutes onto the
head of the capillary gas chromatographlc column. The level of cooling depended on the volatility
of the solutes of Interest. Generally, the GC oven was never cooled below -50°C which would cause
freezing of the decompressed carbon dioxide.
RESULTS AND DISCUSSION
The use of SFE on a quantitative analytical scale presents a number of distinct advantages
when compared to conventional solvent extractions. Depending on the sample matrix, the nature
of supercritical fluids allows for rapid extractions in usually less than one hour with high extrac-
tion efficiencies. Moreover, the ability to transfer the SFE effluent to a GC or SFC In an automated
fashion permits sensitive quantitative or qualitative determinations of solutes In different solid or
liquid matrices.
The quantitative reproducibllity of on-line SFE/GC was Investigated by performing
comparative triplicate analyses using SFE with split GC and flame ionizatlon detection and
conventional syringe split GC Injections of methylene chloride extracts of the spiked clay shown In
Figure 2.
64
-------
1 =
*• • •* I
5 i" I3 I h
-i\v.r..»rW^^^MJ
*r 5
=<
7 *
10
S
LJ
13
14
15
C-
V3
In:)-ct
1. 2— Chloro(>h«nol
3. 4-chIoro-»«thyl ph«nol
4. J-chloro n*phlhala»«
!>. 9r4 ill n I Li ul uluaita
U.
V. f lll •
11. • lili In
13.
14.
IS.
Figure 2. SFE/GC-FID Analysis of Priority Pollutants In Clay. GC temperature program: O°C (20
minutes) programmed to 300°C at 7°C /minutes.
The operating conditions for SFE Included 400 atm pressure of supercritical CC>2 at GO^C for 20
minutes using 650mg quantities of. clay in a 500 mlcrollter SFE vessel. A 50 meter X 0.2 mm I.D.
methyl slllcone (PONA) capillary GC column was used to provide the separation. The SFE effluent
was transferred directly to the capillary GC injection port using a fused silica 15 micron I.D.
transfer line. All peak Identities were confirmed using a mass spectrometer. Table I lists the peak
area reproduclbillty results for selected priority pollutants in the clay.
Table 1. Comparison of Peak Area Reproducibillty for Priority Pollutants In Spiked Clay with On-
Une SFE/GC and Conventional GC Split Injections.
Priority Pollutant
2-chlorophenol
Naphthalene
1 -chloronaphthalene
Hexachlorobenzene
Phenanthrene
Pyrene
Benzo(a)pyrene
%RSD*
SFE/GC
1.8
2.1
5.6
5.8
4.0
4.2
5.5
%RSD*
Split GC
2.0
4.6
8.1
7.8
3.8
5.6
6.4
Concentration
(ng/ull
50
200
60
50
300
200
20
•Based upon raw peak areas resulting from an average of three replicates.
65
-------
As can be seen, the SFE/GC results compared favorably with those obtained by conventional
syringe GC Injections. Moreover, the percent relative standard deviations for the SFE/GC-F1D
results Include contributions from sample Inhomogenelty. weighing, and technique errors as
opposed to only Injection and integration errors for the methylene chloride extract Injections. It
was also very important to thoroughly grind the clay sample before loading the SFE vessel to obtain
consistent results. Certain matrices, such as some clays, have sufficient density to trap certain
solutes for longer periods of time thereby disrupting the efficiency of the extraction process. Figure
3 shows a SFE/GC-FID chromalogram of another environmentally important sample matrix
namely, marine sediment.
AROCLORS
>Li Jl
Inject
5.
101
15 20
TIME (minutes)
25
Figure 3. SFE/GC-FID Analysis of aroclors in marine sediment at low ppm levels. GC temperature
program: -15°C (5 minutes) programmed to 300°C at 15°C/mlnute.
Approximately 1 gram of this sediment was extracted In a 5 milliter vessel at 300 aim using super-
critical CO2 at 60°C for 40 minutes. The same 50 meter PONA column was used to provide the GC
separation. As can be seen, the sediment was contaminated with a mixture of aroclors at 5 to 10
ppm levels (as determined by external standard calibration standards and retention times). If an
electron capture detector would have been used, significantly more sensitivity and selectivity could
have been provided for the aroclors. Since this particular sample contained significant amounts of
water L30%), approximately 1 gram of sodium sulfate was added to the sediment In the extraction
vessel as an adsorbent. In general, on-line SFE with a split GC injector is more capable of
handling wet samples without restrlctor plugging as opposed to on-line SFE with an on-column GC
Injector (18). The conventional sample preparation procedure for this marine sediment generally
Involves 6-8 hours of mulli-solvent extractions and 2 hours of concentration before injection into a
GC-MS as opposed to a total sample preparation and analysis time of 80 minutes for the SFE/GC
technique. Another example of using on-line SFE/GC for quantitative analyses Is shown in
66
-------
Figure 4 with the determination of aromatlcs and chlorinated aromatlcs in contaminated soil which
was taken from a spill site.
Extraction
- 1.2.4-TnnwMiylMnMiu
..
nCjo
20
TIMEtMINOTESI
Figure 4. SFE/GC-FID Analysis of pollutants In soil, GC temperature program: 30°C (7 minutes)
programmed to 310°C at 7°C/minute.
Approximately 170 mg of the soil was extracted In a 0.5 ml vessel at 375 atm using supercritical
CO2 at 50oC for 30 minutes. A 30 meter x 0.25 mm I.D. DB-Wax capillary column was used to
provide the GC separation. Hexachlorobenzene was used as an Internal standard which was
spiked directly Into the soil before extraction. Table II lists the quantitative results for replicate
analyses of the soil.
TABLE II. Replicate SFE/GC-FID Determinations of Aromatlcs and Chlorinated Aromatlcs In
Contaminated Soil
COMPOUND
CONCENTRATION (PPM)«
2 3.
ethylbenzene
cumene
2-chloronaphthalene
1,2,4 trimethylbenzene
44
30
51
25
40
30
50
26
42
32
51
29
43
34
48
24
•Based upon Internal standard calculations
67
-------
CONCLUSIONS
The use of directly coupled SFE/GC as an analytical techniques has shown excellent potential for
the quantitative and qualitative characterization of different solutes in different matrices of
environmental significance Using on-line SFE/GC, an entire analysis which includes the extrac-
tion, concentration, clean-up, and analytical separation steps, can be accomplished in usually less
than one hour. Selective extractions can also be performed by varying parameters such as pres-
sure, temperature, and type of supercritical fluid extracting fluid. Moreover, the analytical
versatility and flexibility of the technique can be further enhanced by the utilization of such
chromatographlc detectors as mass spectrometry, electron capture, nitrogen-phosphorus, and
sulfur-specific.
REFERENCES
1. M.A. McHugh and V.J. Krukonls. Supercritical Fluid Extraction. Principles and Practice.
Butterworths, Stoneham. MA (1986).
2. B.W. Wright, C.W. Wright, J.S. Fruchter. Energy and Fuels 3: 474-480 (1989).
3. J.W. King, J.H. Johnson and J.P. Friedrich. J. Apic. Food Chem. 37: 851-954 (1989).
4. F.I. Onueka and KJV. Terry. J. High Resolut. Chromatogr. 12: 357-361 (1989).
5. J.R Wheeler and M.E. McNally. J. Chromatogr. Sci. 27: 534-539 (1989).
6. S.B. Hawthorne and D.J. Miller. Ana}. Chem. 59: 1705-1708 (1987).
7. S.B. Hawthorne and D.J. Miller. J. Chromatogr. Sci. 24: 258-264 (1986).
8. S.B. Hawthorne, M.S. Krieger, and D.J. Miller. Anal. Chem. 61: 736-740 (1989).
9. S.B. Hawthorne, M.S. Krieger, and D.J. Miller. Anal. Chem. 60: 472-477 (1988).
10, J.M. Levy and A.C. Rosselli. Chromatographta. 28: issue 11/12 (1989)
11. B.W. Wright, S.R. Frye, D.G. McMinn, and RD. Smith. Anal. Chem. 59: 640-644 (1987)
12. J.M. Levy, J.P. Guzowski and W.E. Huhak. J. High Resolut. Chromatogr. Chromatogr.
Commun. 10: 337-347 (1987).
13, J.M. Levy, J.P. Guzowski. Fresenius Z. Anal. Chem. 330: 207-210 (1988)
14. J.M. Levy, R.A. Cavalier, T.N. Bosch. A.F. Rynaski, and W.E. Huhak. J. Chromatogr. Set. 27:
341-346 (1989).
15. S.B. Hawthorne and D.J. Miller. J. Chromatogr. 403: 63-76 (1987).
16. S.B. Hawthorne, D.J. Miller, and M.S. Krieger. J. Chromatogr. Sci. 27: 347-354 (1989).
17. M.W.F. Nielen, J.T. Sanderson, R.W. Frel. and U.A.T. Brinkman. J. Chromatogr. 474: 388-395
(1989).
18. S.B. Hawthorne, D.J. Miller, and J.J. Langenfeld. J. Chromatogr. Sci. 28: 2-8 (1990)
19. K. Sugiyama, M. Saito, T. Hondo, and M. Senda. J. of Chromatogr. 332: 107-116 (1985).
20. S.B. Hawthorne. Anal. Chem.. in press
68
-------
THE ROLE OF SFE IN METHODS FOR THE
ANALYSIS OF TOXIC COMPOUNDS
James H. Raymer and George Velez
Research Triangle Institute
Analytical and Chemical Sciences
3040 Cornwallis Road
Research Triangle Park, North Carolina
Abstract
The recoveries of ^C-labelled 7~BHC, hexachlorobiphenyl, ethyl
parathion, and anthracene from Tenax-GC and four polyimide sorbent materials
were studied using off-line supercritical carbon dioxide extraction (SFE)
and thermal desorption methods. SFE was superior to thermal desorptlon.
On-line SFE/GC analysis both with and without an intermediate Tenax-GC
adsorption step was studied using a mixture of nonradiolabelled pesticides.
The addition of the Tenax-GC step allowed for larger extraction volumes than
were possible using direct SFE/GC with analyte transfer into the
chromatographic column. The intermediate trapping step also Improved the
chromatographic efficiency relative to direct SFE/GC. Replicate analyes
indicated variabilities less than 3% relative standard deviation.
69
-------
Introduction
As more becomes known about the toxic effects of organic compounds
present in ambient air, the identification and quantification of these
compounds, often present in trace quantities, are more important than ever
to understand the extent and magnitude of the exposure problem. Organic
compounds in air range from those that are very volatile (WOC, e.g., 1,1,1-
trichloroethane) to volatile (VOC, e.g., chlorobenzene) to semi-volatile
(SOC, e.g., pesticides and phthalates). Over the years, many analytical
methods tailored to specific volatility categories have been developed.
Collection of air into evacuated canisters with subsequent gas
chromatographic analysis of the contents using mass spectrometric detection
(GC/MS) has been applied to WOCs. As the volatilities of the compounds
decrease, they are present in the atmosphere at lower levels and the
analysis of a small volume of air does not provide the necessary
sensitivity. As a result, methods were developed that rely on the selective
adsorption of the organic compounds as large volumes of air are drawn
through an adsorbant-filled tube. Tenax-GC is widely used as a sorbent to
collect VOCs with the analysis performed on-line with a GC/MS through the
use of thermal desorption of the retained chemicals . For SOCs such as
PCBs, pesticides, and phthalates, polyurethane foam (PUF) or XAD resin is
used for collection. The recovery of these semi-volatile analytes requires
solvent extraction methods. GC/MS analysis is performed on the extract.
Although such preconcentration techniques are very powerful, they are
not without limitations. One problem is found in the selectivity of Tenax-
GC; this material retains nonpolar compounds to a much greater extent than
polar compounds. If quantification of polar compounds is desired, the
volume of air sampled needs to be reduced to avoid "breakthrough" of these
polar compounds and this can limit the sensitivity of the method. The
utilization of sorbents more polar than Tenax-GC could help circumvent this
problem; four polyimide sorbents synthesized at RTI and found suitable for
use with thermal desorption/GC have been reported . The competitive
effects of adsorption for polar and nonpolar analyte vapors on the
polylmides were studied as a means of improving the sampling methods for
polar analytes .
A second problem is found when the thermal desorption of strongly
adsorbed compounds is attempted. Thermal desorption can fail because a
temperature sufficient to desorb the analyte might also destroy the sorbent,
the analyte, or both. In addition, if a compound is thermally unstable,
thermal desorption can Invalidate quantification and introduce artifacts
even if the analyte is only weakly adsorbed.
Ongoing research in our laboratory has been directed at the utility of
supercritical carbon dioxide extraction for environmental analysis. The
high solvating capability of the fluid can facilitate the recovery of
strongly adsorbed analytes from sorbents for subsequent collection and
analysis. The extract can either be expanded into a solvent to provide a
solution that can be analyzed later (off-line) or or it can be introduced
directly into a chromatographic system for on-line analysis. On-line SFE
methods can preserve the favorable concentration factors of thermal
desorption GC and provide for methods with very low limits of detection over
an extended range of volatilities. Because the critical temperature of C02
Is approximately 31*C, high temperatures can be avoided and thermally-
induced decomposition of analytes is not a concern. In addition, SFE can
result in methods devoid of large volumes of organic solvents. For example,
methods that utilize Soxhlet extraction for the recovery of analytes (e.g.,
PUF) require large volumes of organic solvent most of which is sent to waste
70
-------
or into the atmosphere during volume reduction steps.
Work presented in this paper demonstrates the utility of SFE in the
recovery of low levels of semivolatile compounds of environmental concern
::rom Tenax-GC and four polyimide sorbents. Additional preliminary work will
also show how such sorbents can be used to permit the on-line extraction and
GC analysis of SOCs where large extraction volumes might be needed.
Experimental Methods
Solutions of 7-[14C(U)]BHC (1,2,3,4,5,6-hexachlorocyclohexane), [9-
•L4C] anthracene, [ring-2,6-14C]parathion and [1AC(U)]-2,3,5,2',3',5'-
hexachlorobiphenyl were prepared individually in dichloromethane to a
concentration of between 11 and 40 ng per microliter. One microliter
aliquots were introduced into a GC injector and swept with helium onto a
cartridge (4.8 cm long x 4.6 mm ID x 0.25 in OD; 0.8 mL total volume) packed
with 40-60 mesh Tenax-GC or one of the polyimides as previously
described '. The cartridge was then extracted with CC>2 at 3000 psi and
40*C (density of approximately 0.85 g/cm ) and the effluent was expanded
into a scintillation vial containing 15 mL scintillation cocktail.
Alternatively, the cartridge was left in the GC, heated to 250*C, and purged
with helium at 1.5 mL/min for 8.5 min or 6.4 mL/min (Tenax-GC) or 12 mL/min
(polyimides) for 10 min. The 1.5 mL/min flow for 8.5 min purged the sorbent
with the same number of column volumes as is used in routine application of
Tenax-GC thermal desorption where the sorbent cartridge is ten times larger.
The effluent was directed through a heated transfer line into a vial
containing scintillation cocktail; cocktail was used to rinsed any residual
compound from the transfer line into the vial. In each case, recoveries
were determined from radioactivity recovered from a known mass of compound
deposited into the thermal or SFE desorption system in the absence of a
sorbent.
The work with SFE/GC and SFE/SFE/GC was conducted as follows. A Valco
HPLC injector fitted with a 500 nL rotor was placed in-line and before the
0.41 mL SFE cell which was filled with sea sand and held at 50*C in a
modified Lee Scientific Model 501 SFE/SFC system. The valve allowed for the
reproducible introduction of a methanol solution of pesticides
(approximately 100 ng each) into the cell so that SFE conditions could be
mimicked. The pesticides used were molinate, propoxur, atrazine, 7-BHC,
triallate, terbutryin, ethyl parathion, "y-chlordane, and phosmet. The
outlet of the SFE cell was directed through a multiport swithing valve to a
fused silica restrictor (13 cm x 25 flm ID). In the direct SFE/GC
configuration, the restrictor was placed into the first few cm of a DB-5
capillary column (30 m x 0.32 mm ID) at ambient temperature. In the
S7E/SFE/GC configuration, the effluent of the SFE cell was expanded onto the
head of a column comprised of a steel tube (6 cm x 4 mm i.d.) with fritted
column end-fittings and packed with 0.14 g Tenax-GC. The Tenax-GC cell was
also held at 50*C. After this extraction and deposition, supercritical C0£
was used to extract the Tenax-GC in the direction opposite to that of
analyte deposition. This effluent was expanded through another restrictor
(13 cm x 25 ftm ID) into the first few cm of the GC column, again at ambient
temperature, as in the SFE/GC experiment. After the SFE step(s), the column
was purged with helium and the temperature was programmed to effect
separation. Flame lonization detection (FID) was used and data were
collected by either an HP 3390A integrator or a Nelson Analytical Data
system. All extractions were conducted at 400 atm resulting in C02
extraction densities of 0.928 g/cm .
71
-------
Results
Polymer Desorption
SFE Desorption An initial study of the total radioactivity recovered
as a function of C02 volume was performed for each compound on each sorbent.
From these data, we chose extraction volumes for each analyte/sorbent
combination for use in the experiments to determine recoveries. The results
of the recovery experiments are shown in Table I. The analytes were
recovered to a higher extent using a smaller extraction volume from Tenax-GC
than from the polyimides. This is consistent with earlier work where lower
GC retention volumes were measured on columns packed with Tenax-GC than on
those packed with any of the polyimides.
Thermal Desorption The recoveries measured using thermal desorption
are shown in Table II. For these semivolatile test compounds, thermal
desorptions using 50 column volumes of helium provided very poor recoveries
from all of the sorbents. Increasing the desorption volume and time led to
improved recoveries of BHC and anthracene from Tenax-GC while the recoveries
for hexachlorobiphenyl and parathlon were not changed. The results for the
polyimides are indicative of the tenacity of these sorbents for the test
compounds. For the low volume desorptions, none of the compounds was
recovered well. Using the more rigorous desorption conditions, the
recoveries of the analytes improved; only BHC was well-recovered from
polyimides 119, 149, and 115.
SFE/GC
Direct SFE/GC experiments using extraction times of 7 and 10 minutes (8
and 11 column volumes, respectively) revealed that better recoveries and
peak shapes were obtained at the shorter times suggesting that the analytes
were deposited and then lost at longer extraction times. The flow rate of
C(>2 during extraction was 81 mL/min at ambient conditions. A lower trapping
temperature might improve this situation but was not tried because the goal
here was to determine if the presence of the secondary Tenax-GC trapping
step could minimize the problems associated with the longer time of the
first extraction step. Replicate SFE/GC analyses provided percent relative
standard deviations (%RSD) of less than 3% using an extraction time of 7
minutes. Figure 1 shows the chromatogram obtained after SFE/SFE/GC of the
pesticide mixture. In this case, the time of extraction from the Tenax-GC
was 7 minutes and the time of extraction from the sand was 15 minutes. The
use of sand extraction times up to 30 minutes did not affect the recoveries
or chromatographic efficiencies as long as the Tenax-GC extraction time was
maintained at 7 minutes. This allows for a great deal of flexibility in the
first extraction time as long as the analytes of interest are well retained
by the Tenax-GC. For compounds of relatively low volatility, this should
not be a problem. A gas chromatogram of the test pesticides obtained using
conventional splitless/split injection is shown in Figure 2 for comparitive
purposes. Although some broadening of the chromatographic peaks is seen in
the SFE/SFE/GC case of Figure 1, relative to Figure 2, the efficiencies of
the separations shown in Figure 1 are certainly adequate for most purposes
and could probably be Improved upon through a careful optimization of flow
rate and column temperature during extraction of the Tenax-GC cartridge.
Replicate analyses using SFE/SFE/GC provided %RSDs of less than 3%.
Table III shows a comparison of the chromatographic peak area to height
ratios for each of the pesticides for both SFE/GC and SFE/SFE/GC. This
ratio is indicative of chromatographic efficiency, with a low ratio
reflecting a higher efficiency. The extraction times for both SFE/GC and
72
-------
SFE/SFE/GC were 7 minutes at the same extraction flow rate, and thus
differences reflect the effect of the Tenax-GC in the system. It can be
seen that, in all cases, the presence of the Tenax-GC resulted in sharper
peaks. This is presumably due to the accumulation of the analytes at the
head of the Tenax-GC cartridge. When the Tenax-GC is back-extracted, the
analytes are Introduced into the column as a tight band. In SFE/GC, the
analytes were introduced onto the column over a longer time because of the
band spreading that occurred during the migration through the sand. Such an
effect would be expected to be more pronounced in a real extraction
situation where desorption of the analytes from the matrix would take place
over a longer time with the result that they would be spread out even more.
The use of a secondary sorbent can minimize this effect.
Conclusions
The use of polymeric sorbents in conjunction with SFE can allow for the
recovery of less volatile and more polar analytes than is possible using
thermal desorption methods. In addition, the use of the sorbent Tenax-GC in
an on-line SFE/GC analysis scheme (SFE/SFE/GC) can provide greater
flexibility in the extraction of the sample than can direct SFE/GC analysis
of semivolatile organic compounds. The use of on-line SFE/GC methods has
the potential to allow for the collection of smaller samples (smaller air
sampling volumes) and, because the entire extract is analyzed, to lower the
limits of detection of the analytical method.
Acknowledgements
This project was supported by grant R010H01218 from the National
Institute for Occupational Safety and Health of the Centers for Disease
Control with additional funds for the SGE/GC work provided by RTI.
References
1. K. J. Krost, E. D. Pellizzarl, S. G. Walburn, S. A. Hubbard,
"Collection and analysis of hazardous organic emmfsslons," Anal. Ghent. $4(4):810 (1982).
2. R. G. Lewis, A. R. Brown, M. D. Jackson, "Evaluation of polyurethane foam for sampling
of pesticides, polychlorlnated blphenyls, and polychlorlnated naphthalenes in ambient
air," Anal. Chem. 49(12^.1668 (1977).
3. L. S. Sheldon, D. Hhltaker, J. Sickles, E. Pellizzarl, D. Hesterdahl, R. Wiener in
Indoor Air Pollution-. Its Causes. Its Measurement and Possible Solutions. J. Miller,
editor, Lewis Publishing Co, in press 1990.
4. S. D. Cooper, E. D. Pellizzarl, "Characterization of polylmide sorbents by using tracer
pulse chromatography," J. Chroiiiatoor. 498:41 (1990).
5. J. H. Raymer, E. D. Pelilzzarl, "Toxic organic compound recoveries from 2,6-dfphenyl-p-
phenylene oxide porous polymer using supercritical carbon dioxide and thermal desorption
methods," Anal. Chem. 59:1043 (1987).
6. J. H. Raymer, E. D. Pellizzarl, S. 0. Cooper, "Desorption characteristics of four
polyfmlde sorbent materials using supercritical carbon dioxide and thermal methods,"
Anal. Chem. 59:2069 (1987).
7. £. D. Pellizzarl, B. Demlan, A. Schlndler, K. Lam, W, Jeans, Preparation and Evaluation
of New Sorbents for Environmental Monitoring. Volume 1. final report on EPA contract 68-
02-3440 (1982).
73
-------
TABLE I. PERCENT RADIOACTIVITY RECOVERED BY
SUPERCRITICAL C02 DESORPTION
Sorbent
Compound
7-BHC
hexachlorobiphenyl
parathion
anthracene
Tenax-GC
100(14)a
95(1A)
92(14)
81(14)
PI-109
100(45)
72(60)
58(101)
b
PI-119
86(30)
79(30)
74(45)
b
PI-149
92(30)
77(30)
76(45)
b
PI-115
99(30)
85(30)
77(45)
D
a Approximate number of column volumes of C02 used at 3000 ps± and 40*C.
Recovery not determined.
TABLE II. PERCENT RADIOACTIVITY RECOVERED
USING THERMAL DESORPTION AT 250*C
Tenax-GC
Compound 50
260
PI-109
50
500
PI-119
50
500
PI-149
50
500
PI-115
50
500
7-BHC
hexachlorobiphenyl
parathion
anthracene
27
13
12
10
88
13
12
99
1
1
8
8
50
14
3
3
6
16
3
8
88
21
9
6
9
17
8
11
83
57
25
26
22
32
12
14
94
59
36
18
a These values indicate the number of column volumes of helium used for
the desorption.
TABLE III. CHROMATOGRAPHIC EFFICENCY OF PESTICIDE SEPARATION
AS REFLECTED BY AREA/HEIGHT RATIOS
Compound
Molinate
Propoxur
Atrazlne
7-BHC
Triallate
Terbutrin
Ethyl parathion
7-Chlordane
Phosmet
SFE/GCa (%RSD)
7.
7.
8.
8.
7.
8.
8.
8.
12.
4
6
2
4
4
5
3
9
8
(3.
(4.
(1.
(3.
(5.
(A.
(5.
(4.
(3.
9)
7)
2)
8)
6)
2)
7)
3)
7)
SFE/SFE/GCb
5.
5.
5.
6.
5.
5.
6.
6.
8.
4
4
8
2
7
8
0
3
8
(A.
(7.
(6.
(7.
(6.
(1.
(5.
(6.
(1.
(% RSD)
9)
5)
5)
4)
3)
0)
8)
0)
3)
Extraction time of 7 minutes; triplicate analysis.
Extraction of sand for 15 minutes followed by 7 minute extraction of
Tenax-GC; triplicate analysis.
74
-------
t
200
250
—t—
300
20
30
300
—t—
Figure 1. SFE/SFE/GC chromatogram of pesticide test mixture.
Compound identifications are (1) molinate,
(2) propoxur, (3) atrazine, (4) Y-BHC,
(5) triallate, (6) terbutyrin, (7) ethyl parathion,
(8) r-chlordane, (9) phosmet. Conditions as
described in the text.
•ro :
*M* (
i
i
u . .
2
3
m> tass
t
4
t
<
7
DO MO
i * . m 9b « w •
Figure 2. GC chromatogram of the pesticide test mixture
after splitless/split injection. The large
peak before molinate is either an impurity or
thermal decomposition product of propoxur.
75
-------
SUPERCRITICAL FLUID EXTRACTION OF POLYURETHANE FOAM SORB-
ENTS
Mark S. Krieger and Ronald A. Kites
Department of Chemistry and School of Public and Environ-
mental Affairs, Indiana University, Bloomington, IN 47405
Steven B. Hawthorne and David J. Miller
University of North Dakota Energy and Mineral Research
Center, Campus Box 8213, Grand Forks, ND 58202
Polyurethane foam (PUF) sorbents have become widely
used for air sampling because of their low resistance to
air flow and ease of handling. However, the usefulness of
PUF is limited by the time consuming liquid solvent ex-
traction required. Supercritical fluid extraction (SFE)
of PUF is a rapid alternative to conventional liquid
solvent extraction. SFE can achieve quantitative extrac-
tion of polychlorinated biphenyls (PCBs), polycyclic
aromatic hydrocarbons (PAH) ranging from naphthalene to
perylene, heteroatom-containing PAH, n-alkanes ranging
from C-J2 to C24, and organochlorine pesticides from PUF in
10-20 minutes. In addition, the direct coupling of SFE
with gas chromatography (SFE-GC) is possible. SFE-GC
quantitatively transfers all compounds collected on the
sorbent plug to the gas chromatographic column for analy-
sis. Thus, SFE-GC can reduce the mass of sample needed by
two orders of magnitude, and it can also decrease sampling
time by two orders of magnitude. The application of SFE
and SFE-GC of PUF for the rapid extraction and analysis of
variety of semivolatile organic compounds will be demon-
strated.
76
-------
Introduction
Polyurethane foam (PUF) has become a widely used sorb-
ent for the collection of semivolatile organic compounds.
Polyurethane foam has a low resistance to flow, is inex-
pensive and easy to use, and is effective at collecting a
broad range of semivolatile organic compounds. However,
traditional liquid extraction techniques that are used to
extract PUF, such as Soxhlet extraction and sonication,are
very time consuming, error prone, and require large vol-
umes of high purity solvents.
Supercritical fluids have many advantages over liquid
solvents for extraction of PUF. Supercritical fluids have
lower viscosities and higher solute diffusivities than
liquid solvents, resulting in improved mass transfer, and
therefore a shorter extraction time. The solvent strength
of a supercritical fluid can be controlled by changing the
pressure (and therefore, it's density), in order to gain
selectivity in the extraction process. Supercritical
fluids such as CO, and N20 have low critical temperatures
(32 °C), making them ideal for extraction of thermally
labile solutes or sorbents. Fluids such as CO2 and N20
are also gases at ambient temperature and pressure, which
simplifies solvent removal and sample concentration.
Also, the direct coupling of supercritical fluid extrac-
tion to gas chromatography (SFE-GC) is possible. Coupled
SFE-GC has the potential to reduce the volume of air that
one needs to sample for collection of semivolatile organ-
ics by two orders of magnitude, thus decreasing sampling
time by two orders of magnitude.
EXPERIMENTAL METHODS
Extractions were performed using either a SFT Inc.
model 250-TMP syringe pump (Salt Lake City, UT) or an ISCO
model jxLC-500 syringe pump (Lincoln, NE). Supercritical
fluid grade CO2 and N2O in aluminum cylinders was obtained
from Scott Specialty Gases. Extraction cells were con-
structed from Parker brand stainless steel fittings1.
Supercritical pressures were maintained inside the extrac-
tion cell by using 15 cm lengths of 20-30 /im i.d. X 150 /im
o.dk fused silica tubing (Polymicro Technologies, Phoenix,
AZ). Extraction temperature was maintained by inserting
the extraction cell into a thermostatted tube heater.
Extracted compounds were collected by inserting the outlet
of the fused silica restrictor into a vial containing ca.
4 ml of hexane. Approximately 400 ng of PCB congener 204
(2, 2 ', 3,4, 4',5,6,6'-octachlorobiphenyl) was spiked into
each vial before the extraction as an internal standard.
Coupled SFE-GC experiments were performed by inserting the
fused silica restrictor directly into the capillary gas
chromatographic column through the on-column injection
port1. Extracted analytes were cryogenically trapped at 5
77
-------
°C in the gas chromatographic oven. After completion of
the extraction, the restrictor was withdrawn from the
injection port and gas chromatographic analysis proceeded
in a normal manner. All GC/MS and SFE/GC/MS analysis were
performed on a Hewlett-Packard 5985B GC/MS system using
electron capture negative ionization (ECNI). Ion source
temperature was 100°C and the reagent gas was methane at a
source pressure of 0.45 torr. The GC/MS was equipped with
a 30 m DB-5 (250 jiin i.d, 0.25 /im film thickness) capillary
column (J&W Scientific, Folsom, CA).
RESULTS
The ability of supercritical COj to recover a variety
of organochlorine pesticides from PUF is shown in Table 1.
Approximately 100 ng of each compound was spiked into the
center of the PUF plug. After allowing the solvent to
evaporate, the plug was extracted for 10 minutes with CO2
at 300 atm and 50 °C. These results show that supercriti-
cal CO2 is able to rapidly and quantitatively remove these
compounds from the PUF matrix.
Off-line extractions can be performed using a relative-
ly high flow rate (ca. 1 ml/min with a 30 /urn) restrictor.
However, the extraction flow rate is limited during a
coupled SFE/GC/MS by the efficiency of the cryotrapping
step and the pumping speed of the mass spectrometer. A
practical limit for on-line SFE/GC/MS is approximately
200 /il/min of liquid flow. Therefore, it takes roughly 5
times longer to pump an equivalent volume of supercritical
fluid through the extraction cell during a coupled
SFE/GC/MS experiment. In order to minimize the time
required for the extraction step during future on-line
analyses, we examined short extraction time intervals to
determine what percentage of each analyte is extracted
over a given time period. Figure I shows that from 80-95%
of each of the test compounds was extracted in the first
two minutes with both supercritical N20 and CO2. The
remaining fraction was collected in the next two minute
interval, leaving less than 1% of analyte for the final
six minutes of the extraction. These results imply that
these compounds could be quantitatively extracted from PUF
in ca. 15-20 minutes during a coupled SFE/GC/MS experi-
ment.
SFE/GC/ECNI/MS analysis of a PCB standard (one congener
from each level of chlorination from tetrachloro through
decachlorobiphenyl, ca. 250 pg/congener) spiked onto PUF
showed mixed results. Comparison of the amount of each
individual congener detected in two consecutive 20 min
extractions of the same sample indicated that approximate-
ly 90-95% of each congener had been extracted in the first
20 minutes. However, comparison with an external standard
78
-------
associated with individual measurements. The QA/QC results further
emphasized the need to consider all the data from the study in making
decisions.
Many of the testing results were "not detected," and the measured
concentrations were all less than 0.5 ppmv. The data analysis focused on
comparing the study area results to those for the control area and on
comparing indoor and outdoor data sets. These comparisons were critical,
especially in presenting data to the community, since many of the results
were not zero. The outdoor and control area data sets provided perspective
with which to evaluate the study area indoor results. As shown in the
example in Table I, when whole data sets were examined, the study area
sample concentrations were very similar to those measured in the control
area samples and in the samples from the outdoor locations. If a
subsurface source were contributing significantly to the indoor
concentrations, the study area indoor levels would be expected to be higher
than those in the control area and higher than the outdoor levels. A
nonparametric statistical method, along with a standard t-test, was used to
compare the study area and control area indoor and outdoor data sets for
individual target compounds to determine whether there were any significant
differences between these sets that might indicate a subsurface impact.
Both types of statistical analyses produced the same results: the study
area indoor levels were not significantly higher than the study area
outdoor levels for any target compounds and were not significantly higher
than the control area levels for most compounds. The test also indicated
that any differences between the study area and control area indoor air
were less than the outdoor differences between the two areas. The
nontarget compounds detected in the study area indoor samples were similar
in type and level to those detected in the control area and were not
similar to the compounds that had been measured in ground water on site.
The results suggested that when concentrations of target compounds
were measured in the study area structures, their presence was likely
associated with household product use, outdoor air, or possibly natural
gas, rather than with a subsurface source. On an individual structure
basis, indoor concentrations were very similar to the corresponding outdoor
concentration in most cases, and the houses where the indoor levels were
higher than the outdoor levels were scattered throughout the testing area
and were not clustered in one location.
Conclusions and Recommendations
Results of two exposure assessments conducted at apparently similar
sites demonstrated the importance of developing a site-specific technical
approach. For Site 1, the available data and observations suggested that a
subsurface impact to residential indoor air was possible, or even likely;
and for certain structures, the testing results indicated an impact. The
technical approach emphasized full characterization of the subsurface
contamination, investigation of indoor point sources, and speciation
analysis to recognize the "fingerprint" species and ratios. The priorities
of the facility and the regulatory agencies to move rapidly toward any
necessary mitigation were addressed in developing this technical approach,
which provided information key to the design and scoping of mitigation
measures in individual structures and recovery of the liquid hydrocarbon.
The available information for Site 2 indicated that a measurable
impact was not likely, and that the target compounds were very common
solvents that could be expected to occur in the outdoor air and in
household chemical products. The technical approach therefore emphasized
simultaneous outdoor testing, so that any measured indoor levels could be
978
-------
evaluated with respect to the corresponding outdoor levels; testing in a
control area to provide data representing typical levels of the target
compounds in a residential setting; and inventories of household chemical
products in the tested structures. Since all measured concentrations were
expected to be very low, the study was designed to collect representative
sets of data that could be compared to determine any impact. Results of
the testing did not indicate any subsurface influence on indoor levels, as
the levels measured in the study area indoor and outdoor samples, and in
the control area samples, were all similar. Detection limit studies and
thorough QA/QC sampling and analysis were conducted to ensure that the
potential for false positive and false negative results, and the degree of
variability in the data, were fully assessed. The outdoor air and control
area data were critical in providing perspective, mainly to prevent the
incorrect interpretation of any measured concentration as indicative of an
impact. The number of structures tested was more a function of assurance
for the residents than a scientifically-based, representative sampling
scheme.
Neither of these approaches would have been well-suited to the other
site, and considering these examples emphasizes that no single, or
"generic" technical approach will be appropriate for exposure assessment at
a given site. In these examples, point source sampling and
"fingerprinting" chemical species would not have been effective measures
for Site 2 due to the very low levels and mixtures of subsurface compounds.
Conducting such sampling and analysis would have been an unnecessary
expense. Similarly, for Site 1, a study design focused on comparison of
whole data sets - for instance comparing all of the study area data to all
of the control area data - would not have allowed prioritization of
individual structures and areas for further investigation and mitigation.
In general, it is recommended that the technical approach for each
residential indoor air investigation be tailored for the specific site by
considering such important factors as:
• subsurface conditions,
• types, toxicities, and levels of subsurface contamination,
• other potential sources of similar compounds,
characteristics of the residential structures,
outdoor ambient air conditions,
initial indoor observations, and
priorities of the regulatory agencies and the community.
Making full use of available information in the categories listed here will
ensure that resources can be focused on the most critical elements of the
investigation to produce scientifically defensible and cost-effective
assessment of the exposure potential for the site.
979
-------
Table I
Average subsurface and indoor/outdoor air concentrations for
representative target compounds in two studies designed to
assess subsurface impacts to residential indoor air.
Average Concentrations in Parts Per Million
•SITE 1"
Indoor Air
Level 1 Structures PS
Rm Arab.
Level 2 Structures PS
Rm Amb .
Other Structures PS
Rm Amb .
Control Structure Rm Amb.
Outdoor Air
Shallow Soil Vapor
Vapor Above Shallow Ground Water
TNMHC
9.
3.
23
2.
24
0.
1.
0.
8,
29
1
3
2
68
2
16
100
,000
Benzene
0.
0.
0
0
.76
.048
.55
.033
NA
MA
0,
0
.0038
.00097
180
732
Isopentane
13
0.
2.
0.
0.
0.
0.
0.
33
5
23
038
018
13
005
800
2,
800
2-Methyl
14
0.
3.
0,
0.
0.
0 .
0.
-2-butene
43
0
23
0
0054
0018
003
700
3,
200
Ratio*
1.
1.
1.
0.
0.
0
0
1
1
1
.1
.1
.2
.87
.0
.58
.015
,3
,1
.2
wsxyEtm
Indoor Air
Study Area
Control Area
Outdoor Air
Study Area
Control Area
Shallow Soil Vapor
Study Area
Control Area
Shallow Ground Water
Acetone
0.029
0.047
0.030
0.011
0.016
0.006
<0.05
Average Concentrations in Parts Per Million
Cyclohexane Benzene Trichloroethene
0.0068 0.0010 0.00031
0.0064 0.0014 0.00033
0.0086 0.00085 0.00034
0.0020 0.00064 0.00030
0.0004 0.0010 HA
0.009 0.0023 HA
<0.05 <0.05 <0.05
TNMHC - Total non-methane hydrocarbons .
PS - Point source measurement.
Rm Amb - Room ambient measurement
NA - Not available
Average - One-half the detection limit was used for "not detected" results in calculation.
* - Ratio refers to the ratio of 2-methyl-2-butene concentration to isopentane concentration in
individual samples.
980
-------
VOLATILE ORGANIC COMPOUNDS IN THE
ATMOSPHERE OF A NEWLY CONSTRUCTED
RESIDENCE
Barbara B. Kebbekus, Han Chou and Gesheng Dai
Chemistry, Chemical Engineering and Environmental
Science Department of New Jersey Institute of
Technology, Newark NJ 07102
A newly constructed house located in Princeton NJ. was chosen to determine
the initial concentrations of some selected volatile organic compounds (VOC) and to-
tal hydrocarbons and to follow the decay of these levels as construction materials out-
gassed. A Tenax adsorbent trap sampler was used to collect samples in and out side of
the house simultaneously during the first six months after the house was occupied.
Thermal desorptipn with capillary GC and GC/MS were used for quantitative and
qualitative analysis. The concentration distribution of more than 12 compounds both
inside and outside the house was observed. The results show that the concentration of
the some compounds such as toluene, xylene which are solvents in paint and other fin-
ishes was much higher inside the house than outside, just after construction was com-
pleted, but decreased rapidly to near the outside level. The total hydrocarbon con-
centrations behaved similarly. The concentration of chlorinated compounds inside the
house were always higher than those outside, but the source of these has not been
identified. A concentration distribution model of VOC's based on diffusion has been
developed and the parameters of the model are given, based on the determined data.
981
-------
Introduction
More and more attention has been paid to indoor air pollution recently because the
concentration of many pollutants is often higher indoors than outdoors. Identifying the
source of the indoor air pollutants, finding the relations between indoor and outdoor air
contaminant concentrations and developing models for indoor air pollutant concentration
distribution are active subjects of study.
A EPA report[l] indicated that more than 50 volatile organic chemicals, including
aromatic hydrocarbons, halogenated hydrocarbons, alcohols, esters, aliphatics and
aldehydes are normally found in indoor air and their concentration level is always higher
than that outdoors, especially in new buildings. Jarke, et. al.[2] tried to identify the organic
contaminants in indoor air and find their relation to outdoor contaminants. They selected
about 36 homes in the Chicago metropolitan area, using Chromosorb 102 as adsorbent for
GC and GC/MS analysis. 118 compounds were found and identified in indoor air but only
29 compounds were identified in outdoor samples. They concluded that the indoor con-
taminants probably arise for the most part, from the carpeting, clothing, furniture and the
residents' activity in the home. The location of the home relative to industrial operations in
the community had a slight effect on the number of contaminants found in the home. Gam-
mage and Matthews[3] gave some examples of occupant activities, and consumer and con-
struction products inside residences that cause transient or persistent emissions of volatile
organic compounds(VOC) and found that emissions arise from window cleaning, carpet
laying, smoking and automobile operations. Montgomery and Kalman [4] used air bag and
charcoal tube sampling methods for measurement of some volatile organic compounds at
17 residences in Ruston, Washington and made a comparison of indoor and outdoor air.
They showed that indoor air environments were more polluted than the immediate outdoor
environments. Johansson [5] investigated the air of two schoolrooms and his results showed
that the qualitative composition of indoor and outdoor air is about the same.
This study focuses on a newly constructed and furnished residence. Measurements
of both indoor and outdoor air for 12 selected volatile organic compounds were made at
random intervals over a 6 month period to observe the concentration distribution begin-
ning as soon as the house was occupied. A model based on molecular diffusion describes
the concentration distribution of indoor pollutants.
Experimental
The house chosen for the study is located in a newly constructed townhouse com-
plex near Princeton, NJ. The three story house is heated by gas fueled hot air and has cen-
tral air conditioning. Gas is also used for cooking. While there is a garage in the first level,
it was not used for automobile storage before or during the test period. Most of the floors
are carpeted with nylon carpeting and urethane foam padding. The walls are painted with
latex paint. Some woodwork has a stain and polyurethane finish.
Samples were taken in the kitchen/dining area on the middle floor of the house.
Since the house has a fairly open plan, and forced air heat, this central location sould be
impacted by pollutants emitted in any area. Outdoor samples were taken simultaneously on
a balcony on the second floor, at the front of the house. Since the rest of the complex was
mostly incomplete and unoccupied, there was little automobile traffic in the area during
the study.
The sampling period was entirely in the cooler months, when ventilation through
open windows and doors was minimal. A wood burning fireplace was used 5 times during
the period, but not on the same days that sampling was being done. The air conditioning
was not used during this time.
Samples were taken by pumping air through Tenax adsorbent at 8 - 10 ml/min for
24 hrs. The traps were analyzed within 2 days. Both indoor and outdoor samples were
982
-------
taken simultaneously. Sampling began in October 1989 ended in April 1990, and samples
were taken randomly with an average of 7 days interval. Samples were taken more fre-
quently in the early stages of the project, when concentrations were changing more rapidly.
The samples were analyzed using a Tekmar 5000 desorption system coupled with a
Varian 3700 GC. The sample tubes were desorbed at 250°C for 12 min., and injected into
the GC using a -150°C secondary cold trap for focussing the peaks. The column is a 50
meter, 0.3mm I.D. methyl silicone with a 5 um film thickness (Hewlett Packard). The ef-
fluent is split between FID and ECD detectors. Compounds are identified by retention
times, compared with a standard gas mixture containing the target compounds (Alphagaz).
Additional samples were run by GC/MS, using a similar column, to verify compound
identifies. A blank was done with each day's sample and the minimum detection level was
0.01 ppb for each compound.
Results and Discussion
Figure 1 shows the GC/MS chromatpgram of a sample, with the major peaks
identified. The total hydrocarbon concentration was obtained by summing all the FID peak
areas and using the hexane calibration factor. Figure 2 shows the time distribution of the
total hydrocarbon concentration. Initially, there is a very high concentration in indoor air as
the construction and finishing materials putgassed. During most of the construction period,
the house was well ventilated by open windows, but these were closed for about 2 weeks
before sampling began. Within 10 to 15 days, the concentration of indoor hydrocarbons
dropped rapidly. After 30 days the concentration became stable and remained fairly con-
stant for the next five months. The total hydrocarbon concentration in the outdoor air
remained between 20 and 50 ppb during the six months. This was generally lower than the
indoor level. The rapid decrease in the indoor air concentration at beginning indicates that
the contaminants are emitted from surface finishes and carpets and are eventually ex-
hausted by diffusion or ventilation, or are adsorbed on interior surfaces. Because of the
dead volume inside the house, it is considered that diffusion is the main mass transfer
phenomena. Finally, the difference in concentration Between indoors and outdoors be-
comes smaller. The rate of emissions from the various sources becomes equal to the rate of
exhaust to the outdoors and a steady state arises.
The concentrations of compounds such as toluene and xylenes, used in paints,
showed comparable changes with time. Toluene concentration variation is shown in Figure
3, A similar pattern of concentration is seen in the total hydrocarbons. Benzene and
hexane, more volatile compounds, were also like each other, but were different from the
heavier compounds. These concentrations decreased more quickly in the indoor air and
finally approached that of outdoor air (Figure 4). This may indicate that the lighter com-
pounds are more easily exhausted from emission sources and also are more readily cleared
from the indoor air through diffusion. Several CIO hydrocarbon isomers are prominent in
the analysis and showed a decline over the course of the project. The source of these was
not determined.
The concentration distribution of total chlorinated compounds (i.e.total ECD active
compounds) as well as a typical chlorocarbon, 1,1,1-trichloroethane, are shown in Figures 5
and 6. The concentration of these in indoor air was always higher than outdoor, and did not
drop during the 6 months. This indicates that the emission source was fairly constant, and
did not become depleted. Drycleaned clothing brought into the house may be a con-
tributor, but the actual source has not been identified. Several chlorofluorocarbons of un-
known origin were also present at constant levels. Chlorofluorocarbon 113 was the most
prominent, with a concentration of about 10 ppb.
The simplest model to describe indoor air concentration distribution based on diffu-
sion and ventilation may be written as:
983
-------
_ v = - Ri + FCi + DiA( C-L - Coi ) (1)
dt
V : total house volume
Cj : indoor concentration of compound i
t : time, hours
Rj : emission rate of indoor source of compound i
F : total ventilating flowrate
Dj : diffusion coefficient of compound i
A : total diffusion area from indoor to outdoor
Coi: outdoor concentration of compound i
During a certain period, if the source and ventilation are considered to make slight
contributions, then equation (1) can be rewritten as :
dCi
_ v = D.jA( ^ - Coi ) (2)
dt
Ci - CQi DiA
or In = - t (3)
C°i- Coi V
Where C°- is concentration of compounds i at initial time.
If the emission is significant, and is considered as a diffusion from the source,
equation (1) can be rewritten as :
d^
_ v = - DeiAe( Cei- ^ ) 4- D-jAf Ci - CQi ) (4)
dt
where D -, Ae, Cei are the diffusion coefficient, area, and concentration from emit-
ting materials of compound i respectively. Equation (4) is a linear first order differential
equation and its solution based on initial conditions t = 0, Cj' = COi is:
Ci = a + ( C°i - a ) e~bt (5)
Ci - a
or In - - bt where (6)
r°-- a
c 1 a.
DeiAecei + DiACOi DeiAe + DiA
a = and b = (7)
DeiAe + DiA v
From equation (6) it can be seen that, after a period of time, the concentration
of indoor contaminants tends to become constant. The concentration is a function of the
emission concentration, the diffusion coefficient of the contaminant and the diffusion area
of both emission and exhaust. Actually, when the concentration of contaminants in indoor
984
-------
air is low enough, and the ventilation has little effect on the concentration of contaminants,
diffusion can be considered as the main mass transfer path for exhausting contaminants.
The plots of the concentration term in equation (3) vs. time for the toluene, xylenes and to-
tal hydrocarbon data show the predicted linear behavior early in the sampling, and in later
samples, show a different, near zero, slope (Figure 7), indicating early clearance of pol-
lutants by diffusion and later equilibration with surfaces in the residence.
References
[1] L. R. Ember, "Survey Finds High Indoor Levels of Volatile Organic Chemicals",
Chemical and Engineering News, December 5, 1988
[2] F. H. Jarke, A. Dravnieks and S. M. Gordon,"Organic Contaminants in Indoor Air
and Their Relation to Outdoor Contaminants" ASHRAE Trans., 87(Part 1): 153-
166, (1981)
[3] R. B. Gammage and T. G. Matthews, "Volatile Organic Compounds in Indoor Air:
Types, Sources, and Characteristics", Environmental Progress, Vol.7, No.4, 279,
(1988)
[4] D. D. Montgomery and D.AKalman "Indoor/Outdoor Air Quality : Reference
Pollutant Concentrations in Complaint-Free Residences", Appl. Ind. Hyg., vol.4,
no.l, 7,(1989)
[5] J. Johansson, "Determination of Organic Compounds in Indoor Air with Potential
Reference to Air Quality", Atomspheric Environment, vol.12, 1371, (1978)
985
-------
o
o
qdd
qdd
O
O
TJ
4—
O
o
c
t/1
o
C7>
3/H
Lew
331-1'1'I
a
986
-------
CO
00
2.0-r
Fig, 4 Benzene
Fig 6. Total Chloracarbons
0.0
0 15 30 45 60 75 90 105 120 135 150 1 S5 180
Day
60 180
2.0
Fig 5. 1.1 ,1 —Trichloroethane
t— 1 1 H 1
0 20 40 60 80 100 120 140 160 180
Day
o
o
I
o"
o
o
I
a~
o
-4--
-6--
Fig. 7 Diffusion model
A Xylene
O Toluene
f Total Hydrocarbons
—\—
10
-+-
-+-
20 30
Day
.
40
50
60
-------
showed that only 45-55% of the total sample was detected.
It has previously been demonstrated that PCBs can be
extracted quantitatively from PUF . Therefore, something
must be interfering with analysis of the extracted ana-
lytes. There are two possible explanations to this prob-
lem. One possibility is that analytes are being are not
being efficiently collected during the cryogenic collec-
tion process. The other possibility is that co-eluting
interferants (contaminants from the CO,) are suppressing
the ECNI process in the ion source of tne mass spectrome-
ter. Many co-eluting impurities are observed in the total
ion chromatogram, although we have as of yet been unable
to identify these compounds. Levels of freons in SFC
grade CO2 high enough to preclude the use of an BCD for
dynamic extractions have been reported by other workers in
the field3. Since ECNI responds well to similar classes
of compounds as the BCD, it is likely that this is a
significant effect. Therefore, it may be necessary to
obtain a higher purity grade supercritical fluid for
SFE/GC/ECNI/MS experiments, or to spike the PUF with
isotopically labeled standards to account for variations
in detector response.
CONCLUSIONS
SFE is a rapid alternative to liquid solvent extraction
for the removal of semivolatile organic compounds from
PUF. SFE can reduce the amount of time needed to perform
the extraction step from days to minutes. Coupled SFE-GC
analysis should allow sample collection volumes for ambi-
ent air to be reduced by two orders of magnitude, and thus
decrease sampling time by the same factor. A decrease in
sample volume requirements makes small personal sampling
pumps useful for collecting samples in locations that
otherwise would be inaccessible with a high-volume air
sampler (e.g., balloons, remote areas, indoor sites).
REFERENCES
1. S.B. Hawthorne, D.J. Miller, J. Chromatoqr. 403: 63-76
(1987).
2. S.B. Hawthorne, M.S. Krieger, D.J. Miller, Anal. Chem._
61: 736-740 (1989).
3. F.I. Onuska, K.A. Terry, JJ. High Res. Chrom. 12; 357-
361 (1989).
79
-------
ESTIMATING THE CANCER RISK FROM MULTI-ROUTE EXPOSURE TO CHLOROFORM FROM
CHLORINATED WATER
Wan K. Jo*' Clifford P. Weisel and Paul J. Lioy
Joint Graduate Teaching Program in Human Exposure,
Rutgers: The State University and UMDNJ - Robert Wood Johnson
Medical School and The Environmental and Occupational
Health Sciences Institute, Piscataway, NJ 08854
*current address: InJae University
Dept. of Environmental Science
A-Bang Dong, KimHae
Seoul, Korea
Abstract
Showers are currently suspected to be as an important source of exposure
to chloroform organic as water ingestion. To better estimate the internal
chloroform dose from dermal and inhalation exposure from showering,
chloroform breath concentration before and five minutes after exposure to
chlorinated water was measured. An increase greater than an order of
magnitude above the corresponding background breath concentrations was found.
The chloroform breath concentration was also determined to decrease
exponentially with time following the cessation of the exposure, reaching
background levels within two hours. Approximately 30% of the internal dose
was expired as chloroform. The calculated internal dose from showering
(inhalation plus dermal) was comparable to estimates of the dose from daily
water ingestion. The risk associated with a single, ten minute shower was
estimated to be 1.2 x 10"4, while the estimated risk from daily ingestion of
tap water ranged from 0.13 x 10"4 to 1.8 x ICf4 for 0.15 and 2.0 liters,
respectively. Since the estimates of chloroform risk from domestic water use
for the three exposure routes, ingestion, inhalation and dermal are similar,
all routes must be used to calculate the total risk when making policy
decisions regarding the quality of the municipal water supply.
Introduction
Drinking water regulations for chemical contaminants have been based
on the assumption that a daily ingestion of two liters represents the
principal source of exposure to chloroform (Interim Primary Drinking Water
Regulations). However, showering also exposes individuals to volatile organic
compounds (VOC) contained in the water. The exposure from showering is via
two routes: inhalation of VOC after transference to air (1,2) and dermal
absorption of VOC after water contacts the body (3,4). Chloroform can enter
the body via these exposure routes during showering thereby increasing its
body burden. Models based on Fisk's law, diffusion considerations and
transfer estimates through the skin have indicated that exposure in a shower
could yield a similar VOC dose as ingestion (5,6). Since many people take
at least one shower each day throughout their lifetime, the potential health
risk could be significant. Estimates of chloroform health risks have been
extrapolated from high dose animal studies. The partitioning of chloroform
among the amount expired, metabolized and absorbed by tissue need to be
similar for the extrapolation to be valid. Fry et al. (7) demonstrated that
for a 0.5 g dose of chloroform, between 18% and 67% was expired, with amount
expired inversely related to body weight. They also found that the majority
of the unexpired chloroform was metabolized to carbon dioxide, with little
excreted. The objectives of the present study were to: 1) examine the
relationship between the chloroform concentration of shower water and breath
2) estimate the percentage of the chloroform internal dose expired at
988
-------
environmentally relevant levels and 3) estimate chloroform dose and cancer
risk from a shower and compare it to that from water ingestion.
Methods
Chloroform exposure and internal dose from showering was estimated from
thirteen showers taken by six subjects (5 males and 1 female) using a defined
set of parameters. These parameters were either standardized or measured
(4). Breath samples were collected from the subject prior to and after each
exposure. A water sample was collected along with each shower. The percent
of chloroform dose expired was estimated from five series of breath
measurements. For three, inhalation only exposure was monitored, and for
two, a normal shower was taken. A model shower chamber, constructed of
stainless steel, was used to examine inhalation only exposure for durations
of 5, 10 and 15 minutes. Breath samples were collected immediately following
the exposure, at 30 minutes, one hour and two hours. Breath samples were
collected at more frequent intervals following a 10 minute shower for the
combine dermal, inhalation exposure. A description of the sampling and
analysis for the dose estimates is available in Jo et al (4). The time
series samples were collected using a system model after Pellizzari et al (8)
and analyzed by thermal desorption/GC/MS.
Estimates of the Expiration of Chloroform
A mathematical model describing the post exposure breath concentration
with time was determined by fitting the decay curve to an exponential
equation. The equation was integrated to estimate the amount of chloroform
expired following the exposure. It was assumed that chloroform was also
expired during showering. This was estimated assuming a linear relationship
with time.
Estimates of Chloroform Dose from a Shower
The total chloroform dose from a shower was estimated from the sum of the
inhalation exposure and dermal exposure. Thfi chloroform dose from inhalation
exposure was calculated using the following equation:
Di = Er x Ca x R x T/Wt (1)
where Di = chloroform dose from an inhalation exposure (ng/inhalation
exposure-kg); Er = respiratory absorption efficiency of chloroform (0.77)
(9); Ca = shower air concentration Ug/m3); R = breathing rate(0.014
m3/min) (10); T = Shower duration (10 min); and Wt = body weight of a
reference person (70 kg)
Jo et al. (4) estimated the relative chloroform body burdens for dermal
exposure and inhalation exposure from chloroform breath concentrations, while
controlling for exposure intensity. The ra~io of the body burden resulting
from dermal exposure to that from inhalation exposure was 0.93. The
chloroform dose from dermal exposure was then estimated from that ratio as
follows:
Dd = Di x F (2)
where Dd = chloroform dose from a dermal exposure^g/dermal exposure-kg); and
F = ratio of the body burden from dermal to inhalation exposure (0.93)
Estimates of Chloroform Exposure from Water Inqestion
The chloroform dose from water ingestion was estimated using the
following equation:
989
-------
Dig = Ei x Cw x Aw/wt (3)
where Dig = Dose from water ingestion (^g/ingestion exposure-kg); Ei
gastrointestinal tract absorption efficiency (100%); Cw = mean tap water
concentration Ug/L); and Aw = water amount ingested per day(0.5 and 2
1iters/day)
Estimates of Cancer Risk from a Shower and Water Ingestion
The chloroform risk associated with a shower and water ingestion for a
reference person was calculated from the estimated doses. A linearized model
was used to estimate the cancer potency of the chloroform exposure (11). The
model extrapolates animal data at high experimental doses to low
environmental exposure levels in order to estimate cancer risk for humans.
The cancer risk from a shower was estimated by extending the model developed
for ingestion exposure to inhalation and dermal routes of exposure.
Pd = q x D x 10'3 (4)
where Pd = lifetime risk; q = cancer risk potency slope (.26 mg/kg-day)"1
(11); and D = chloroform dose^g/kg-day)
Results and Discussion
Breath analyses confirmed that a chloroform internal dose resulted from
both dermal and inhalation exposure during showering. The post-exposure
exhaled breath concentrations ranged from 6.0 to 21 i^g/m3. Chloroform was
not detected in any breath samples collected prior to a shower. The minimum
detection limit (MDL) for the packed GC method for chloroform was 0.86 ^g/m3.
Using an estimate of the pre-shower breath concentration as one half the MDL,
the chloroform body burden from a shower was 14 to 49 times higher than the
background chloroform body burden.
Chloroform was measured in the exhaled breath as a function of time after
exposure. The initial study showed a linear increase in breath concentration
with exposure duration in a sample collected immediately after an inhalation
exposure (figure 1). The pre-exposure breath concentration of three subjects
was again less than 0.86 ng/m , while the post-exposure, exhaled breath
concentrations were 15, 21 and 26 jtg/m3 for 5, 10 and 15 minute exposure
durations, respectively. The air concentration in the model shower chamber
was estimated to be 160 *jg/m3. It is hypothesized that there is a net
absorption of chloroform by the body during showering since the exchange of
chloroform between alveolar air and blood across the lung/capillary interface
is based on an equilibrium process and the maximum breath concentration was
below that in the air being breathed. However, since an equilibrium process
exists, some of the previously absorbed chloroform could be expired during
showering. The amount expired during shower should be estimatable using the
linear relationship shown in figure 1, provided the body is not saturated
with chloroform. Once the exposure ceased, the breath concentration
declined. The concentrations after one hour were 1.8, 1.5 and 2.4 ^g/m3, for
the 5, 10 and 15 minute exposures, respectively, and at or below the
detection limit after two hours.
To better define the elimination rate of chloroform, a series of breath
samples were collected at frequent intervals following an exposure to
chlorinated water. The chloroform breath concentrations measured after
exposure to shower water were elevated above the pre-exposure concentrations
of <0.4 Mg/m3. The breath concentration was observed to decrease
exponentially with time (figure 2). The chloroform breath concentrations
990
-------
reached background levels within two hours of the end of exposure. An
estimate of the amount of chloroform exhaled was made by determining the best
fit for the data, assuming that the breath concentration decay with time
could be mathematically modelled using an exponential equation. The
following equation was derived:
C = 2.2 e-°-023T (5)
where C is the concentration Ug/m3) and T is the post exposure time
(minutes). Equation 5 was integrated between time equal zero and 120
minutes, the time required for the exhaled breath concentration to the return
to the background concentration. The time integrated concentration, 89 ng-
min/m3' was multiplied by the average breathing rate of 0.014 m/min to
determine that a total of 0.99 ng of chloroform expired subsequent to
exposure. To estimate the amount expired during the shower, the background
chloroform breath concentration, 0.4 /ug/m3, and the highest breath
concentration measured, 3.9 ^g/m3' were used as the two endpoints for the ten
minute shower exposure. The upper concentration is an underestimation, since
the first post-exposure breath concentration was measured more than 2 minutes
following the shower and the breath concentration is expected to have
decreased during that time interval. The integrated concentration
determined, 22 ^g-min/m3, was multiplied by the average breathing rate of
0.014 m3/min, to estimate that 0.30 ng was exhaled during showering. Thus,
an estimated 1.3 /*g of the chloroform internal dose was expired as
chloroform.
The internal dose from showering results from both inhalation and dermal
exposures. The inhalation exposure can be estimated from equation 1 using
the chloroform air concentration measured during this experiment of 20 ng/m* .
Thus the internal dose from inhalation exposure for the individual, who
weighed 75 kg, was calculated to be 2.2 ^g. The internal chloroform dose
from dermal exposure was calculated using equation 2 to obtain a value of 2.0
-------
The chloroform risk estimates are summarized by exposure type in Table
1. The risk associated with exposure from a single daily ten minute shower
was estimated, from equation 4, to be 1.2x10 (6.2xlO"5 for inhalation
exposure and 6.0xlO~5 for dermal exposure), which is comparable to the risk
from a daily 2-L water ingestion (l.SxlO"4). The risk from a shower was
approximately a factor of 10 larger than from ingestion daily of 0.15-L water
ingestion (0.13xlO~4). If one million people were exposed for a lifetime,
the excess cancer risk would be 122 from a single daily 10 minute shower, 180
from a daily 2-L water ingestion, and 13 from a daily 0.15-L water ingestion.
Conclusion
Individuals are exposed to chloroform from daily showers when using
chlorine-treated municipal tap water. The chloroform body burden from a
shower was estimated to be 14 to 49 times the background chloroform body
burden, depending on the shower tap water concentration. The breath
concentration rapidly returns to background values within two hours, with
approximately 30% of the internal dose expired. Chloroform dose and cancer
risk from a single, ten minute shower was equal to or greater than that from
daily water ingestion. Hence, in situations where individuals are told not
to drink water because it has been contaminated with VOC they should also be
told not to shower or bath with the water. Furthermore,, the chloroform dose
received from showers and from other uses of chlorinated tap water must be
considered when agencies and health officials evaluate the quality of a
chlorinated water supply.
Acknowledgements
This research was funded by the New Jersey Department of Environmental
Protection, Division of Science and Research, Dr. Jo received Fellowship
support from the Environmental and Occupational Health Sciences Institute.
References
1. Andelman, J.B. (1985) Inhalation exposure in the home to volatile
organic contaminants of drinking water, Sci. Total Environ. 47:443-460.
2. McKone, T.E. (1987) Human exposure to volatile organic compounds in
household tap water: The indoor inhalation pathway. Env. Sci. & Tech.
21:1194-1201.
3. Wester, R.C. and Maiback H.I. (1989) Human skin binding and absorption
of contaminants from ground and surface water during swimming and bathing.
J. of the Amer. Col. of Tox. 8:853-860.
4. Jo, W.K., Weisel, C.P.; Lioy, P.J. (1990) Routes of chloroform exposure
and body burden from showering with chlorinated tap water. Risk Analysis, in
press.
5. Brown, H.S.; Hattis, D. (1989) The role of skin absorption as a route
of exposure to volatile organic compounds in household tap water: A simulated
kinetic approach. Journal of the American College of Toxicology, 8:839-851.
6. McKone, T.E. (1989) Household exposure models. Toxicol. Let. 49:321-
339.
7. Fry, B.J.; Taylor, T.; Hathway, D.E. (1972) Pulmonary elimination of
chloroform and its metabolite in man. Arch. int. Pharmacodyn. 196:98-111.
8. Pellizzari, E.; Thomas, K.; Raymer, J; Smith, D. and Cooper, S. (1990)
Measurements of exhaled breath using a new portable sampling method. U.S.
EPA Final Report from Research Triangle Park RTI/140/03.
9. Ambient Water Quality Criteria for Chloroform (1980); U.S. Environmental
Protection Agency, EPA 440/5-80-033.
10. W. S. Syneder, M. J. Cook, E. S. Nasset, L. R. Karhausen, G. P. Howells,
I. H. Tipton, "Report of the Task Group on Reference Man," International
Commission on Radiological Protection Papers, No. 23, Pergamon Press, New
York, p. 360 (1984).
992
-------
Table 1 - Chloroform dose and the corresponding risk estimates
for the chloroform water concentration of 24.5
Exposure Type
Dose
Ug/kg-day)
Risk
(per mill ion)
Normal Shower
Inhalation Exposure 0.24
Dermal Exposure 0.23
Total 0.47
Water Ingestion
2-1 Ingestion 0.7
0.15-L Ingestion 0.05
62
60
122
180
13
the dose was estimated based on one shower oer dav
CHLOROFORM BREATH CONCENTRATION
AFTER INHALATION ONLY EXPOSURE
B
E 3°
A
T 25
H
20
C
O 19
N
C 10
U 5
G
i
1
- ' • :!
i i
'
_
RIO 10 20 3O 40 60 BO 70 80 9O
3 TIME (minutes)
Figure 1
CHLOROFORM BREATH CONCENTRATION
B AFTER 10 MINUTE SHOWER
R
t . ..
A 5r
T
H *
C 3-
0
N 2
C
1 -
u
* i
*
*
0 '
" 1
i
/ 0 10 20 30 40 50 60 70 80 90
m TIME (minutes)
3
• 1303A * 13038 * 1503/v " 1503B
Figure 2
993
-------
COMPARISON OF INDOOR AND OUTDOOR ALDEHYDE CONCENTRATIONS
DURING THE IACP ROANOKE, VIRGINIA RESIDENTIAL STUDY
Roy Zweidinger and Alan Hoffman
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Leslie Gage
NSI Technology Services Corp.
Research Triangle Park, NC 27709
The U.S. EPA's Integrated Air Cancer Project (IACP) conducted a field
study in Roanoke, Virginia during November 1988-February 1989. As part of
this study, samples were collected in ten pairs of homes, each pair
consisting of one home heated by oil and one heated by electricity or gas.
Paired homes were located near each other and concurrent sampling was
conducted inside each home and outside the one not heated by oil. Average
concentrations of the major aldehydes inside homes with and without oil heat
were similar and significantly greater than outside levels. A few homes
exhibited a diurnal variation of indoor concentrations, but no consistent
trends were observed among the homes on a daytime/nighttime, weekend/weekday
or day of week basis.
Introduction
Between November 1988 and February 1989, the Integrated Air Cancer
Project (IACP) conducted a field study in Roanoke, Virginia. This study was
conducted in similar fashion to our previous study in Boise, Idaho during
the period November 1986 to February 1987. In the Boise study the focus was
on residential wood combustion and mobile sources while the Roanoke study
includes residential oil combustion as an additional source. An overview of
the Boise study and preliminary results were presented at a previous
symposium^.
The residential phase of the Roanoke study involved ten pairs of homes
wherein one home had an oil heating system (IN:WITH) while the other was
heated by gas or electricity (IN:W/0). Each week samples were concurrently
collected from inside one pair of homes, outside the home without oil
heating (OUTSIDE), and at three primary ambient sites. The primary sites
included a mobile source site located adjacent to an interstate highway next
to Roanoke's Civic Center (CIV), a general residential site located at
994
-------
Morningside Park (MSP) and a background situ located outside the city at
Carvins Cove (COV). Samples were collected 7am-7pm and 7pm-7am every day at
the primary sites and on Saturday through Tuesday at the residential sites
to observe daytime vs. nighttime and weekend vs. weekday variations.
Experimental Methods
Aldehydes were collected on dinitrophenylhydrazine (DNPH) coated silica
gel cartridges^ in duplicate (collocated). All sample flow rates were
maintained using mass flow controllers, Tho DNPH derivatives of carbonyl
species were eluted from the sampling cartridges with 5 ml of acetonitrile
and analyzed by high performance liquid chromatography (HPLC). Two C-18
columns (25 cm x 4.6 mm) in series were employed using an acetonitrile/water
gradient and detection at 360 nm. Aldehyde identities were determined by
retention time comparison with standards.
Results
The collocated samples were analyzed for about one half of the sampling
periods of the residential study. The percent difference between ppb
concentrations for the collocated samples averaged 2.85% for formaldehyde
and 10.39% for acetaldehyde (indicating a contamination problem with
acetaldehyde). Review of all field blanks ^evealed consistently low
formaldehyde levels, while acetaldehyde levels varied in a random fashion
with no correlation relative to sampling site or storage time between
cartridge preparation and analysis. Comparison of blank cartridges stored
at RTP, NC with those stored in Roanoke (and not taken out to sampling
sites) showed elevated levels of acetaldehyde for the Roanoke samples. This
indicated some of the contamination observed in the blanks may have occurred
in Roanoke during storage or during transportation to and from RTP.
Problems with acetaldehyde blanks or contamination has not been evident in
previous IACP field studies or numerous other studies where we have employed
the cartridge technique. The source of contamination is still under
investigation. Because the contamination is of a random nature and blanks
do not exist for every sampling period, the data presented here have not
been corrected for the blank background. Based on a typical sample, the
average acetaldehyde contamination is equivalent to 1.73 ppb (v/v). Acetone
contamination was also evident, but this is typical due its ubiquitous
nature.
Table I gives the average aldehyde concentrations observed during the
residential study. Formaldehyde, acetaldehyde, and acetone account for
about 90% of all the carbonyls measured. Except for a trivial difference
for o-tolualdehyde, the average indoor concentrations are always greater
than outside and indoor concentrations show little sensitivity to the type
of heating employed. The problem with contamination from acetaldehyde and
acetone is readily apparent from the measurements made outside the homes,
i.e. the acetone concentration is 4 times greater than formaldehyde and the
formaldehyde/acetaldehyde ratio is 1.1. Formaldehyde/acetaldehyde ratios of
2 are more typical which suggests the actual outdoor concentration of
acetaldehyde may be about 1.1 ppb. In our previous IACP field study
conducted in Boise, Idaho, the outside formaldehyde/acetaldehyde average
ratio was 1.85 based on individual samples and average concentrations of
formaldehyde and acetaldehyde were 4.40 and 2.38 ppb respectively3. Note
that the average ambient formaldehyde concentrations observed in Boise are
about twice those of Roanoke. Average formaldehyde concentrations inside
Boise homes without wood stoves (16.3 ppb) ^ere similar to the homes in
Roanoke without oil heating (16.9 ppb). Average formaldehyde concentrations
inside homes with wood stoves in Boise (22.2 ppb) however, were higher than
995
-------
those inside Roanoke homes either with or without oil heating.
Figure 1 shows the average formaldehyde concentrations observed for the
residential samples in Roanoke relative to daytime and nighttime while
Figure 2 shows average concentrations (daytime + nighttime) for each day of
the week. In all cases, no significant trends were observed. Figure 3
shows the average concentrations of formaldehyde, acetaldehyde and acetone
for all ambient sites sampled in Roanoke for the Saturday-Tuesday period.
Concentrations are similar for the residential outside (RES) and residential
park (MSP) samples while the mobile source site (CIV) is slightly elevated.
The background site (COV) shows reduced formaldehyde but is similar to the
residential samples with respect to acetaldehyde and acetone, again
indicating the contamination problem with these species.
The formaldehyde levels observed for each sampling period inside and
outside the homes without oil heating are shown in Figures 4 and 5,
respectively. Mean concentrations are shown by the solid line while the
dotted and dashed lines represent plus and minus one or two standard
deviations, respectively. The first block in each group represents
Saturday-daytime followed by Saturday-nighttime etc., the last block being
Tuesday-nighttime. While an occasional day>night pattern is evident, e.g.
week 9, there is no consistent pattern at the various homes studied. There
is no correlation between cooking and or frying in the homes and the
concentration of formaldehyde (and other aldehydes such as hexanaldehyde and
acrolein) found. This may relate to the fairly long 12-hour sampling time
relative to these activities and the effects being averaged over adjacent
sampling periods. On the other hand, these activities may only be minor
contributors to the overall concentrations observed.
In the indoor and outdoor data plots, one statistically significant
outlier is evident, i.e. week 4 for the outside samples and week 7 for the
indoor samples. Outdoor samples were collected via a glass manifold located
inside a trailer parked adjacent to the home without oil heating. Outside
air was continuously pulled through the manifold and a portion of this air
pulled through a smaller manifold where the outdoor cartridge samples were
collected. A small leak in any of the connections between the glass
manifold and the cartridges could result in contamination by air inside the
trailer. The indoor samples had the cartridges located inside the home with
all sampling pumps, etc. located in the trailer downstream of the
cartridges. Any leaks in this system might result in reduced flow through
the cartridges, but would not be a source of contamination. We have
observed indoor concentrations similar to week 7 in previous studies and
have no valid reason to reject these samples. Outdoor concentrations and
air exchange rates appear to have little influence on observed indoor
formaldehyde concentrations.4 Personal activities, home construction and
furnishings appear to be the main variables which influence indoor
concentrations. The outdoor concentrations observed for week 4 appear
unrealistic and are harder to explain. (Mean values reported in Table I do
not include these samples.) While the observed levels could be a result of
some local point source, there is some indication that these outdoor samples
may have been contaminated with indoor air due to a leak in the sampling
system. For week 4, the outdoor samples have formaldehyde concentrations
slightly greater than their corresponding indoor levels. Hexanaldehyde
concentrations were also elevated relative to the indoor samples. In the
Boise study, we did have a few elevated outdoor formaldehyde samples, but
did not also see corresponding higher levels of minor species such as
hexanaldehyde. We currently consider the week 4 samples atypical and will
compare the aldehyde data to other gas phase samples, e.g. organic
hydrocarbons, when that data becomes available.
996
-------
Conclusions
Average concentrations of most carbonyls were higher inside homes than
outside with formaldehyde concentrations being 7 times higher indoors than
outdoors. The indoor formaldehyde concentrations were similar in all the
homes with one exception and no correlatior with heating systems, i.e., oil
burners or gas and electric was found. In general, aldehyde concentrations
indoors did not exhibit consistent diurnal or day of week trends.
Activities of individuals, furnishings, etc. are likely the major factors
affecting carbonyl concentrations in homes. Correlation of cooking
activities with observed indoor aldehyde levels was not consistent.
Contamination of field samples with acetalciehyde was observed, the causes of
which are still being investigated.
Disclaimer
The research described in this paper has been reviewed by the
Atmospheric Research and Exposure Assessment Laboratory, US EPA and approved
for publication. Approval does not signify that the contents necessarily
reflect the views and policies of the Agency nor does mention of trade names
or commercial products constitute endorsement or recommendation for use.
References
1. Session on Integrated Air Cancer Prograir Study (13 papers presented),
1988 EPA/APCA Symposium on Measurement of Toxic and Related Air
Pollutants, RTP, NC, 1988, pp 799-895.
2. S. Tejada, "Evaluation of silica gel cartridges coated in situ with
acidified 2,4-dinitophenylhydrazine for sampling aldehydes and ketones in
air," Intern. sL. Environ. Anal. Chem.. 26: 167 (1986).
3. R. Zweidinger, S. Tejada, R. Highsmith, H. Westburg and L. Gage,
"Distribution of volatile organic hydrocarbons and aldehydes during
the IACP Boise, Idaho residential study," 1988 EPA/APCA Symposium on
Measurement of Toxic and Related Air Pollutants, RTP, NC, 1988, pp 814-
820.
4. C. Lewis, R. Zweidinger, and R. Stevens, "Apportionment of residential
indoor aerosol, VOC and aldehyde species to indoor and outdoor sources,"
1990 Symposium on Indoor Air, Toronto, Canada.
997
-------
Table 1
Recovery of organochlorine pesticides from PUF
using supercritical CO2
COMPOUND MW % RECOVERY9 SD
HCB 282 96.5 3.2
Heptachlor 370 113.1 5.0
Heptachlor Epoxide 386 99.8 2.3
Oxychlordane 420 99.6 3.2
gamroa-Chlordane 406 111.6 6.8
alpha-Chlordane 406 106.8 3.2
trans-Nonachlor 440 106.5 9.0
E,P_'-DDE 316 108.9 7.5
Dieldrin 378 105.5 5.0
Endrin 378 97.9 8.2
aAverage of three extractions. The PUF plug was ca. 1 cm in
diameter and 2.5 cm long
80
-------
Table I. IACP Roanoke residential study average aldehyde
concentrations, ppb(v/v) .
COMPOUND OUTSIDE3
Formaldehyde 2.26
Acetone 9.51
Acetaldehyde 1.91
Unknowns 0.63
Hexanaldehyde 0.03
Propanaldehyde 0.20
Butyraldehyde + MEKb 0.27
Valeraldehyde 0.02
Acrolein 0.11
Benzaldehyde 0.05
p-Tolualdehyde 0.01
Iso-valeraldehyde 0.00
Crotonaldehyde 0.02
m-Tolualdehyde 0.00
o-Tolualdehyde 0.01
Total w/o acetone 5.72
Total carbonyls 15.03
a. Does not include data from
b. MEK = methylethylketone.
IN: WITH
15.68
12.86
9.16
1.31
1.21
1.02
0.45
0.48
0.50
0.38
0.11
0.08
0.09
0.03
0.00
30.51
43.38
week 4 .
IN: W/O
16.87
12.57
9.60
1.45
1.09
0.85
0.64
0.43
0.38
0.38
0.11
0.07
0.06
0.02
0.00
31.95
44.52
OUTSIDE
IN:WITH
CO IN:W/O
DAYTIME NIGHTTIME
Figure 1. Average formaldehyde concentations observed at
residential sites during daytime and nighttime.
B SATURDAY
SUNDAY
cz MONDAY
TUESDAY
16
14
12
in'
.
8
C
1
4
a
_J?
m*
/
y
y
/
y
y
/
/
1
*-,
/
/
y
/
/
y
y
/
/
— FORMAL
K3 ACETAL
m ACETONE
1
/
/
y
/
y
y
/
1
f
/
/
/
/
y
y
y
/
/;
OUTSIDE IN:WrTH IN:W/O
Figure 2. Average indoor and outdoor
concentrations of formaldehyde
observed by day of the week.
cov
MSP
RES
Figure 3. Average ambient concentrations
of major carbonyls observed at
residential and primary sites.
998
-------
au -
£ 40-
Q.
z
O lf\ _
g
2 20-
UJ
U
Z <
8 10-
n -
MEAN = 16.8 SIGMA = 9.42 J\
L
•^
r ^\
JUl J •' T • • ^rf' 1
^^TLP J _ lj^ • ^^
"^ ^« 1 A fu*^
JL iW - JVV ¥
1 23456789 10
WEEK
Figure 4. Formaldehyde concentrations observed for each sampling period
at homes without oil heating.
12-
-«j
D. 10 -
0.
6-
UJ
u
MEAN = 3.02 SIGMA = 2.63
IlIlflllllllTIIilITIllITIIP I11IITFI1IT11
123456789 10
WEEK
Figure 5. Formaldehyde concentrations observed for each sampling period
outside homes without oil heating.
999
-------
RISK CHARACTERIZATION OF NONCANCER HEALTH EFFECTS
ASSOCIATED WITH INDOOR AIR POLLUTANTS
Robert G. Hetes
Terrence K. Pierson, Ph.D.
Center for Environmental Risk and Geosciences
Research Triangle Institute
Research Triangle Park, North Carolina
To date, risk characterization efforts associated with indoor air poll-
utants have primarily focused on carcinogenic risks. While cancer risks
are an important concern, experience to date has shown that, related to
indoor air pollutants, noncancer health effects may be more widespread and
significantly affect the work environment, worker health and performance.
Sick Building Syndrome (SBS) has become a common problem attributed to
volatile organic compounds (VOC), biological contaminants and other pollu-
tants. SBS symptoms are noncancer in nature, emphasizing the importance of
noncancer health effects from indoor air pollution. Very little work has
been carried out in characterizing noncancer risks. Most effort has focus-
ed on defining acceptable daily intakes or Reference Doses (RfD) rather
than estimating incidence and severity of the wide range of effects within
an exposed population. There are several significant differences between
risk characterization for cancer and noncancer health effects. These diff-
erences have significant bearing on the manner in which indoor air pollu-
tants should be measured and exposures characterized.
INTRODUCTION
The objectives of this paper are to review the nature of characterizing
noncancer health effects, identify implications of this risk character-
ization on indoor air pollutant data and measurements, attempt a risk char-
acterization for an example complex mixture of indoor air pollutants,
assess the quality of available data on Indoor air pollutants, and identify
future research needs.
NATURE OF RISK CHARACTERIZATION OF NONCANCER HEALTH EFFECTS
Noncancer effects are assumed to have thresholds, and vary widely,
typically having multiple endpoints, multiple target organs and multiple
symptoms of varying intensity and severity. Coupled with the thresholds
concept is the issue as to what constitutes an "adverse health effect".
1000
-------
The severity of effects vary greatly from a physiological response with no
apparent effects in an individual as a whole, to a clinical response or
disease, and finally to disability and/or death. "Adverse" effects are
defined as any effects which result in functional impairment and/or patho-
logical lesions which may affect the performance of the whole organism, or
which reduce an organism's ability to respond to an additional challenge.
At low exposure concentrations, a homeoutatlc state occurs in which the
organism has the ability to adapt to a mino:: insult with no change in over-
all physiology, no physical effect is observed, though psychological or
subjective changes may be perceived by an individual. As concentration or
dose increases, a compensatory state ia reached where functional impairment
and subtle effects which are not adverse may occur without loss of overall
integrity of the organism through some defense mechanism such as metabolic
detoxification. As exposure or dose furthe:: increases, small portions of
the populations exhibit adverse frank effects, where the dose has exceeded
the body's compensatory or defense limits and organ and system function is
impaired, disease may be apparent. Wh«n do!>e is increased the incidence of
adverse effects will Increase, as will the .severity of the effect observed.
At some point, the entire population would exhibit an adverse response,
though there would be a distribution oi: severity of effects. The goal of
noncancer risk characterization is to define the distribution of effects
and severity of these effects within the population for the organ systems
affected.
The complexity of characterizing noncanoer health effects is
exemplified by a condition commonly associated with Indoor air exposures,
Sick Building Syndrome (SBS), a common problem consisting of acute,
nonspecific sensory irritation and oth«sr sensory effects. The World Health
Organization (1983) defined SBS symptoms to Include (!) eye, nose, and
throat irritation, (2) sensation of dry mucous membranes, (3) erythema
(skin irritation, redness), (4) mental fstique and headaches, (5) high
frequency of airway Infections and cough, (<>) hoarseness and wheezing, (7)
itching and unspecific hypersensitlvity, and (8) nausea and dizziness. The
variety of endpoints Is apparent in this condition, affecting multiple
organs (eye, respiratory etc...) with varying severity (eye irritation to
unspecific hypersensitlvity).
IMPLICATIONS ON RISK CHARACTERIZATION
This multi-dimensional nature of noncancer health effects has Impli-
cations on all components of the risk characterization process. As stated
above, the goal Is to define the distribution of effects within a popu-
lation for several organ systems for various levels of severity. Each of
the organ systems would have a threshold for each of the adverse effects.
Thresholds are both time and concentration dependant, as concentration
increases response time is decreased. Appearance of noncancer effects may
be immediate or delayed, and either permanent or transient. To
characterize effects, most efforts to date have focused on the use of
thresholds for effects from long-term exposures and to define protective
exposure levels (i.e., RfD) for such exposures rather than specific dose-
response functions. In theory, thresholds could be determined for a range
of effects for specific exposure patterns such as acute, short-term and
long-term exposures, not just for long duration exposures. However, even
the simplified approach of using thresholds presents problems for risk
characterization. Average lifetime exposures are Inadequate for comparison
to acute and short-term thresholds. Greater detail is required in defining
dose, exposure, arid pollutant concentration, Peak and cumulative values,
in addition to an overall average, are required for comparison to the
1001
-------
appropriate threshold levels. This is especially important considering
most indoor pollutant exposures are transient with varying concentration.
Individual exposures in the indoor environment are heavily influenced
by the time individuals spend in particular microenvironments and the
activities they are engaged in while in those microenvironments. The
timing of an exposure is important when concentrations are variable over
time, as in the decay profile of offgassing from new building materials.
Likewise, activities are important in that it may result in very high
transient exposures (e.g., the use of consumer cleaning products) or affect
dosimetric factors (e.g., increased breathing rate associated with physical
activity) which may increase the effective dose.
Each microenvironment has its own concentration profile, peak, and
average concentration. Actual overall exposures are the aggregation of all
individual microenvironment exposures. An overall average exposure level,
aggregated over all microenvironments, does not allow for comparison of
peak exposures, from individual microenvironments, to relevant thresholds.
Therefore greater detail is needed in characterizing pollutant concentra-
tion and time activity patterns for several common microenvironments.
APPLICATION TO AN EXAMPLE COMPLEX MIXTURE
Table 1 presents an example complex mixture of VOCs identified as major
pollutants in a problem building. Employees experienced increased health
complaints following building renovations which included painting and in-
stallation of carpets and office partitions. Sampling of the work environ-
ment indicated a typical indoor air quality problem, multiple compounds at
low concentrations. Actual ambient sampling data were limited and incon-
sistent, for both constituents and their concentrations.
The data in Table 1 are from direct sampling and model estimates. The
first seven compounds were Identified as major constituents in carpet
offgas. The values are from samples taken from new carpet of the same
batch stored in a warehouse. As offgassing emissions from new products
tend to decay with time, these measurements are assumed to be maximum
exposure levels. The remaining three compounds have been associated with
office partitions and other materials using particleboard. The expected
ambient concentration estimates of the three compounds attributed to office
partitions as a source were calculated based on chamber studies to
characterize emissions rates, and a model which considers environmental and
ventilation conditions. The values in the table are assumed to be worst-
case (at 0.1 air changes per hour, ACH). These 10 compounds and their
respective concentrations will be used to attempt a risk characterization
for this complex mixture.
IDENTIFICATION OF HEALTH EFFECTS
Initial efforts focused on identifying the possible health effects
associated with the compounds identified in Table 1. A search was made in
the College of Medicine's database, TOXNET, for each of the compounds in
the following fields: human toxicity excerpts; populations at special risk;
absorption, distribution, and excretion; metabolism/metabolites; biological
half-life; mechanisms of action; interactions; and threshold limit values
(TLV). Only noncancer health effects which have been reported for humans
were recorded and summarized for the following organ/systems, eyes, nose,
respiratory, central nervous system, blood, skin, as well as a miscell-
aneous category. As an example, the summary matrix for the central nervous
system is presented in Table 2. The designation of an X within a cell
1002
-------
TABLE 1. VOLATILE ORGANICS FOUND IN EXAMPLE COMPLEX MIXTURE
Compound
Toluene
Ethylbenzene
Xylene
Styrene
Cumene
4-Phenylcyclohexene
DIchlorobenzene
Formaldehyde
Acetaldehyde
Acetone
Concentration (ppb)*
13.0 - 22.0
3.7 - 4.6
4.6 - 8.6
31.0 - 33.0
4.1 - 6.9
70.7**
18.0 - 68.0
61.0***
15.0***
55.0***
Method
GC/MS, GC/FID
GC/MS, GC/FID
GC/MS, GC/FID
GC/MS, GC/FID
GC/MS, GC/FID
GC/MS, GC/FID
GC/MS, GC/FID
Chamber, Model
Chamber, Model
Chamber, Model
Value
Peak
Peak
Peak
Peak
Peak
Peak
Peak
Worst-case
Worst-case
Worst-case
* Range of concentrations represent the values from GC/FID and GC/MS.
Measurements made on new stored materials and are assumed to represent
the maximum exposure levels.
** 4-PCH concentration from GC/FID, presence confirmed but no
concentration quantified on GC/MS.
*** Predicted concentrations at 0.1 ACH due to other building materials
(office partitions) using a model developed by the EPA, Air and Energy
Engineering Research Laboratory (AEERL).
Source: *USEPA, Internal memo from Burchett^a and Singhvi to T. Fields,
dated 5/22/88.
***USEPA, Internal memo from Tichenor to D. Weitzman dated 8/25/88
indicates that effect has been reported in humans for a given compound. An
asterick by the effect listing indicates that that particular effect was
reported in the problem building. The information in the Table is not
exhaustive, but illustrative of the range of possible effects due to a
particular compound. The absense of an X does not imply the effect is not
possible, only that it has not been reported in TOXNET. This method is
qualitative and limited by the availability of toxicological or epidemio-
logical data for specific compounds. It should also be noted that the
exposures for which effects were reported in TOXNET were typically occupa-
tional, at exposures significantly higher than that found in the indoor
environment. While limited, this method is useful in identifying the range
of probable effects and in identifying partlcluar compounds of concern.
To introduce a quantitative element to the risk characterization, the
observed concentrations were compared to two common thresholds, Threshold
Limit Values (TLV) and odor thresholds. Th« TLV is intended for use in
occupational settings and is not intended for application to the general
public. However, it is an easily recognized standard that is widely used
to establish Indoor and outdoor exposure Units and does provide a gross
appraisal of possible health effects. Odor thresholds may play an impor-
tant role in Indoor air quality problems. Jiome hypothesize that odors may
be either bringing attention to otherwise ignored symptoms, association of
the source of odors as the source of the symptoms, or leading to symptoms
directly. TLVs and odor thresholds have b«:en found for all compounds
except 4-phenylcyclohexene (4-PC) for which only an odor threshold has been
defined. These are presented in Table 3. Observed concentrations were
several orders of magnitude lower than their respective TLVs, and odor
100:5
-------
Table 2: SUMMARY OF REPORTED NONCANCER HEALTH EFFECTS DUE TO INDOOR AIR
POLLUTANTS COMPRISING THE MIXTURE
EFFECT
CENTRAL NERVOUS SYSTEM
Tinnitis
Headache
* Dizzines
* Depression
* Fatigue
* Confusion
Drowsiness
Vertigo
Slowed Reaction Time
Intoxication: Euphoria
Exhileration
Boastful., talkative
Incoordination (ataxia)
* Anasthesia
Edema
Weakness
COMPOUND
AcA ACE COM DCB EtB FOR STY TOL XYL
x
x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
* Effects Reprted Associated With The Example Complex Mixture
thresholds were exceeded for three compounds, acetaldehyde, formaldehyde,
and 4-PC. This approach can be expanded to any effect which has an
identified threshold.
The efforts described above have focused on assessing the toxlclty of
individual compounds. Indoor air exposures are typically mixtures, as in
the example complex mixture, so it is desirable to develop some method to
assess the toxicity of a mixture rather than just for the individual
Table 3: COMPARISON OF OBSERVED CONCENTRATIONS TO TWO EXAMPLE THRESHOLDS
Acetaldehyde
Acetone
Cumene
Dichlorobenzene
Ethylbenzene
Formaldehyde
Styrene
Toluene
Xylene
4-Phenylcyclohexene
Maximum
Obs. Cone, (ppm)
0.015
0.055
0.0069
0.068
0.0046
0.061
0.033
0.022
0.0086
0.072
TLV (ppm) Odor Threshold*(pptn)
100 0.00011 - 2.3
750 19.675 - 668
50 0.008 - 1.3
75 2.0 - 30
100 2.0 - 200
1 0.05 - 49
500 0.047 - 200
100 2.14 - 70
100 0.08 - 40
0.001 - 0.01**
Source: Ruth, 1986.
** Source: Unpublished reports
1004
-------
components. Based on a review of suggested EPA approaches and others
proposed in the literature, an approach of using mixture index values has
'aeen applied. Several options for these mixture index values have been
identified and include: (1) Hazard Index, (2) Margin of Exposure, (3)
Additivity (with Relative Potency), (4) Response Addition, (5) Comparative
Potency and Toxicity Equivalent Factors, (6) Total Organics (or by Chemical
Class), (7) Indicator Compound Concentrations, (8) Interactions, and (9)
Tiered Approach.
The hazard index, margin of exposure, and additivity assume additivity
of effects, and involve the summation of health effects for individual com-
pounds. Comparative potency is different in that the toxicity of the mix-
ture is assessed directly without attention to individual components. Com-
parative potency is based on the assumption that animal bioassays are appli-
cable to human health prediction and mixtures are compared based on bioassay
results. The total chemical class approach assumes little difference in
relative potencies between compounds with the same chemical characteristics.
The indicator compound approach assumes that a single compound is indicative
or responsible for a large fraction of total health effects. Interactions
is a formal approach for addressing the physiological effects individual
compounds have on one another, either synergistically or antagonistically.
The tiered approach integrates elements of previous approaches. It would
have a ceiling threshold to protect against a severe (clinical) effect, and
a dose-response component to estimate the distribution of less severe
effects at concentrations below the ceiling threshold.
Two mixture index value approaches have been applied to the example
complex mixture, the hazard index (HI) and margin of exposure (MOE). These
Indexes and their application to the complex mixture (for those constituents
with EPA verified inhalation RfDs) are described in Table 4. MOE^s have
been calculated for these compounds and are all within an order of magnitude
of each other, and are several orders of magnitude lower than their
individual no observed adverse effects level (NOAEL). For a mixture, the
MOE is sum of the MOE^ for the individual constituents for a given target
organ. Individual MOE^ should only be summed for the same target organ
system. A MOE^ for the central nervous system (MOEcns) was calculated
summing the MOE^s for the 3 compounds having a NOAEL for the central nervous
system. The MOEcns was calculated as 0.00443. This is interpreted as the
exposures are about 0.4% of the "threshold value for the mixture".
The hazard index approach is the most commonly applied method for
noncancer effects from mixtures. Values exceeding 1 indicate that the RfD
has been exceeded. The HI^ values calculated for those 5 compounds having
verified inhalation RfDs vary from 0.032 for toluene to 3.056 for cumene.
This is interpreted as toluene exposure is at about 31 of the RfD, while
cumene exposures are more than 3 times the RED level. Cumene exposures are
predicted to exceed the RfD. However, before any conclusions are made it
should be noted that cumene has a high uncertainty factor (10,000) and a low
confidence designation. Therefore in reviewing the mixture and relative
importance of each constituent and in the overall index value these factors
should be considered.
An overall mixture Hazard Index was calculated for the central nervous
system (HIcns). The value of 3.184 indicates that the exposure is more than
3 times the estimated "mixture RfD". Howeve;:, it should be noted that the
cumene which, as described above, has the highest uncertainty and lowest
confidence designation, contributes about 96% to the overall index value.
Therefore this uncertainty should be included in any interpretation of this
index value. While there is significant: uncertainty in the RfD, this is the
1005
-------
only method which predicts or substantiates some health (CNS) effects for
the example mixture.
CONCLUSIONS
In reviewing the application to the example complex mixture it can be
concluded that the data are preliminary and limited, with compounds and
effects identified. The data have been shown to be adequate in defining
peak pollutant concentration levels and for subsequent comparison to some
threshold values. However, existing data are inadequate to define actual
exposures and dose and subsequent use of dose-response relationships.
Existing data was also inadequate in that only 5 of the 10 compounds had
verified inhalation RfDs which limited the use of the mixture index
approaches. Concentration profiles were also nonexlsting prohibiting the
calculation of actual exposures and dose. It was not known if all compounds
were present at their respective maximum concentrations simultaneously. It
is also not known whether the sampling taken from the stored virgin material
actually represents the peak concentration, actual peak may be higher to
increased emissions due to increased surfaces area exposed. To apply the
mixture index values, concentration profiles for each chemical in the
mixture should be used for the same time period. Exposure to peak
concentrations different chemicals at the same time versus different times
will give different index values.
ACKNOWLEDGEMENTS
This work was sponsored under cooperative agreement with the U.S. EPA
Environmental Criteria and Assessment Office, Research Triangle Park, NC., EPA
Contract Number CR 815509-01-0.
REFERENCES
1. DeRosa, C.T., Dourson, M.L., and Osborne, R., (1989), Risk Assessment
Initiatives for Noncancer Endpoints: Implications for Chemical Mixtures,
Toxicology and Industrial Health, Vol. 5, No. 5, pp. 805-824.
2. World Health Organization, (1983), Indoor Air Pollutants: Exposure and
Health Effects, Euro Rep. Stud., 78. Copenhagen, HO, 42 pp.
3. Tichenor, B. A., (1987), Organic Emmislon Measurements Via Small Chamber
Testing, Indoor Air, Proceedings of the 4th International Conference on
Indoor Air Quality and Climate, Berlin.
4. Kolega, H. S., (1987), Environmental Annoyance, in Environmental
Annoyance: Characterization, Measurement and Control, H.S. Kolega, ed.,
Elsevier Science Publishers, 1987.
5. Ruth, J. H., (1986), Odor Thresholds and Irritation Levels of Several
Chemical Substances: A Review, Am. Ind. Hyg. Assoc. J. Vol. 47, March
1986, A-142 to A-151.
6. U.S. Environmental Protection Agency, (1987), Guidelines for the Health
Risk Assessment of Chemical Mixtures, Federal Register, 51FR 34014-44025.
7. Schoeny, R.S. and Margosches, E., (1989), Evaluating Comparative
Potencies: Developing Approaches to Risk Assessment to Risk Assessment of
Chemical Mixtures, Toxicology and Industrial Health, Vol. 5, No. 5, pp.
825-837.
1006
-------
TABLE 4. INDEX APPROACH EXAMPLES
o
o
ObsT (ppb)
Cenpouoi (i) CCnc. NC&EL
Acetaldehyde 15.0 66,000
(LQAEL)
Curene 5.5 18,700
(LQAEL)
1,4-Dichlorbbenzene 43.0 11,210
Toluene 17.0 40,000
Xylene 6.6 6,200
1Ccnf idence : Study /Eata Base/RfD
H = High
M = Medium
L = Low
Study
Species Organ/Effect (M3E)i
Rats Histopathological 0.00227
changes in resp. tract
Rats CNS, nasal irritaticn 0.00294
Eats Urinary protein output 0.00384
incr. liver, kidney wts.
Hunan CNS {Dizz. headache) 0.000425
eye, nose irrit.
Human CNS, irritation 0.00106
Margin of Exposure QCE)
MEi = Obs. Conc./NQAEL
e.g. M^CNS = E MDEj
MEcNs = M>ECLM + M^ETCL + M^EXYL
= (0.00294) + (0.000425) + (0.00106)
(M3E)CNS * 0.00443
-> 0.4Z of observed threshold level
RED
UP ng/ra? ppb Conf.l
3,000 0.04 22 M/L/L
10,000 0.009 1.8 L/L/L
100 0.7 115 H/M/M
100 2.0 533 M/M/M
100 0.3 69 M/M/M
Hazard Index (HI)
HIi
0.682
3.056
0.374
0.032
0.096
HIi = Obs. Conc./RfD
(Uiere RS) = NQAEL/UF)
e.g. HlcNs = £ HIi
HCNS = HIOM + HlTQL +
= 3.056 + 0.032 +
HlcHS - 3.1B4 »1
HIXYL
0.096
-------
Figure 1
HCB
Heptochlor
Heptochlor Epoxide
Oxychlordane
gammo-Chlordone
olpha-Chlordone
trans-Nonochlor
p.p'-DDE
Dieldrin
Endrin
Recovery vs Time
300 atm C02
50 C
xx\\\\\\ma
"^ 0-2 min
40 60
% in Fraction
100
Recovery vs Time
300 atm N20
50 C
HCB
Heptochlor
Heptochlor Epoxide
Oxychlordane
gommo-Chlordane
alpha—Chlordone
trans-Nonachlor
p.p'-DDE
Dieldrin
Endrin
^XXXXXXXX^XX<^XXN\XXXX\^
XX^XXXXXXXXXXXXXXXXXXCV^ S
xx\\\x\x\\\\\\\\\x\x\\\\x^8a
\\\\\\\\\\\\\\\\\^
0—2 min
2-4 min
4-6 min
20
40 60
% in Fraction
80
100
Percent recovery vs time for several organochlorine
pesticides. The 6-8 minute and 8-10 minute fractions
showed no detectable traces of any of the analytes.
81
-------
ASSESSMENT OF THE HEALTH RISKS ASSOCIATED WITH
INDOOR BENZENE VAPOR EMITTED FROM BUILDING
FOUNDATION SOIL: A CASE STUDY
Tibor T. Sarlos
Bruce E. Fishman, Ph.D.
Dennis J. Paustenbach, Ph.D., D.A.B.T.
ChemRisk
A Division of McLaren
1135 Atlantic Ave.
Alameda, CA 94501
Abstract
Benzene is often the main chemical of concern associated with gasoline
residues in soil. Occupants of buildings constructed on gasoline-
contaminated soil may be exposed to significant indoor air concentrations
of benzene vapor emitted from soils underlying building foundations. This
paper describes a screening-level approach for evaluating the potential
hazards associated with inhalation of indoor benzene vapor emitted from
building foundation soil. A combination of two predictive models is used
to derive an upper-bound estimate of the airborne benzene concentration
inside a structure built on gasoline-contaminated soil. First, the benzene
vapor flux from the soil is estimated using the known or estimated soil
benzene concentration and Farmer's steady-state model. The benzene
concentration inside the building is then estimated using a simple
diffusion-ventilation or "stirred-tank" model. The variables used in this
model are the benzene generation rate in indoor air and the building's
ventilation rate. The benzene generation rate is the product of the
predicted benzene flux and the total area of openings in the foundation that
allow entry of gaseous benzene into the building. The minimum ventilation
rate is estimated using building code requirements. Uptake of benzene by
building occupants and the attendant theoretical excess cancer risk is
estimated using standard factors for exposure and risk prediction.
Predictive modeling approaches such as those used in this assessment are
useful tools for conducting screening level health risk assessments when
cost or time constraints preclude the measurement of vapor flux and/or
airborne chemical concentrations.
1008
-------
Introduction
The leakage of volatile liquids from underground storage tanks is a
significant environmental problem. Because many substances stored in these
tanks contain chemicals considered to be hazardous to human health, leaks
draw concern from the public and environmental regulatory agencies.
However, the occurrence of a chemical spill or leak alone is not enough to
pose a health threat to the public. An accurate evaluation of the potential
for an underground tank leak to actually pose a significant health threat
to the public involves an assessment of the magnitude and extent of the
contamination, the toxicological properties of the agents, and the potential
degree of human exposure. Environmental regulatory agencies often require
that a risk assessment be conducted in order to ensure that no threat to
public health exists or will exist as ;i consequence of environmental
contamination, such as that produced by an underground tank leak or spill.
In this paper, an underground gasoline spill that occurred in Southern
California was evaluated. A developer wanted to build a "fast food"
restaurant on the site having gasoline contaminated soil. There was a
concern regarding the possibility for the chemicals to migrate through the
soil and the building foundation, contaminating the air inside the
restaurant. Inhalation of vapors is the only pathway of exposure considered
since no human contact with the soil or groundwater is expected to occur.
For the sake of simplicity, benzene is the only chemical considered in this
example, however, the screening risk assessment methodology presented herein
can be applied to other chemicals, with the appropriate consideration of
their volatilities (diffusion coefficients) and potencies. The approach
employed here can also be applied to similar spills at other sites, with the
careful consideration of differing site-specific information.
Due to time constraints and the relatively low levels of volatile organic
compounds (VOCs) detected in the soil, a screening-level assessment of the
potential health risks associated with inhalation of benzene vapor is
deemed to be sufficient for initial evaluation. The screening methodology
involved using the combination of two predictive models: one for estimating
the benzene vapor flux from the contaminated soil, and one for estimating
the attendant airborne benzene concentrations inside the restaurant. The
latter prediction involves the use of several assumptions regarding the
potential for chemical vapor to penetrate building foundations and the
ventilation standards for commercial structures. Finally, estimates of
carcinogenic health risks associated with exposure to the estimated indoor
airborne benzene concentrations are calculated using standard factors for
exposure and risk prediction,
Experimental Methods
Modeling Benzene Vapor Emissions
Although direct measurement of vapor flux from area sources contaminated
with volatile organic compounds (VOCs) I;; possible they are not always
feasible or cost-effective. However, predictive models are available for
estimating gas emission rates from areas contaminated by losses of VOCs from
spills, leaky underground storage tanks, pipelines or surface impoundments.
1009
-------
Such predictive models are useful when project cost or time constraints
preclude direct measurement of vapor flux, or when rough "upper-bound"
estimates of flux are sufficient.
Farmer's model is used to predict the benzene vapor flux from the
gasoline-contaminated area. This mathematical model is based on vapor-phase
diffusion, as described by Pick's First Law of molecular diffusion. Vapor
phase diffusion is the dominant mechanism for the transport of volatile, low
solubility chemicals through the vadose (unsaturated) zone. The model was
developed for estimating chemical emission rates from covered landfills
without internal gas generation. It was validated In a laboratory using
industrial landfill soil contaminated with hexachlorobenzene, The model is
used to predict the movement and steady-state vapor loss rates of chemicals
from landfills. The predicted non-decreasing emission rate Is generally an
overestimate because It disregards decreases in soil contaminant
concentrations due to volatilization, leaching, and/or biodegradation.
Farmer's equation Is as follows (Farmer, 1980):
where:
F = Vapor flux from the soil surface (mg/sec-cm2) ,
D = Vapor diffusion coefficient In air at 25°C (cm2/sec),
P. = Volumetric air content of soil (cm3,- /cm3,,),
n -r o VT SOI I
PTOT = Total soil porosity (cm3al> + yater /cm3^),
1 - (b/p)
where:
b = Dry soil bulk density (gdry Soil/cm3sojl) ; and
p = Soil particle density (assumed to be 2.65 g/cm ) .
CG = Concentration of chemical in soil vapor below cover
(mg/cm3ai-r),
CQ = Concentration of chemical at soil surface. (mg/cm .r).
Assumed to be zero,
L = Thickness of clean soil cover (cm).
Estimating Benzene Concentration in Soil Vapor
In order to employ Farmer's model in estimating the benzene vapor flux
from soil, the concentration of benzene in the soil vapor must be
determined. An approach to estimating the chemical concentration in the
soil vapor given its total concentration In soil Is suggested by Jury et.
al. (Jury et. al, 1983), and Is utilized In this report. Jury's approach
Is based on the fact that the chemical resides in a combination of the
vapor, liquid, and solid, or adsorbed, phase, and it focuses on
determining how a given quantity of chemical will partition between these
three phases. Jury uses Henry's law for liquid-vapor partitioning and a
linear equilibrium sorption isotherm for solid-liquid partitioning. The
following equation relates the total chemical concentration In the soil to
its vapor phase component:
CG = CT / RG
where:
CG = Chemical concentration in vapor phase (mgchem /cm ajr)
CT = Total chemical concentration In soil (mgchem /cm sojt)
1010
-------
RQ = Gas partitioning coefficient that gives ratio of
chemical concentration in vapor phase to total
concentration (cm . /cm . ()
" bVKH + PL/KH + PA
where :
b = Dry soil bulk density
where:
F • = Fraction of organic carbon in the soil
K = Chemical's organic carbon distribution
coefficient (ml/g)
KH = Chemical's dimensionless Henry's Law constant
3 liquid /Cm3air>
P. = Volumetric moisture content of soil
, ,
PA = Volumetric air content of soil (cm a-r /cm S01-^)
Predicting Airborne Benzene Concentration in Indoor Air
The indoor benzene concentration under transient conditions may be
approximated by the following mass balance relationship :
the rate of change of the mass = generation rate - removal rate
of benzene in indoor air
This relationship Is represented mathematically by the following
differential equation:
d(Vc) = G - Qc.
dt
where :
V = Volume of building (m3)
c = Indoor air concentration of contaminant (mg/m3)
t = Time (hours)
G = Generation rate of chemical in Indoor air (mg/sec)
F x A
where :
F = Benzene vapor flux predicted by Farmer's model (mg/sec-
cm2)
A = Total area of openings in building foundation (cm2)
Q = Ventilation rate in building (ms/sec)
The above differential equation can be solved for c(t) to obtain:
c(t) = (G - Ge"Qt/v) / Q, which, at steady state, reduces to c = G/Q .
To estimate the generation rate of benzene in Indoor air, it was
c.ssumed that the amount of benzene vapor which will diffuse through the
building foundation concrete is negligible in comparison with the benzene
vapor which will enter the building through openings in the foundation.
This assumption is "based on a study by Zapalac (1983) in which he showed
that this was true for radon diffusion through concrete. The benzene
generation rate in indoor air, then, is simply the product of the benzene
vapor flux predicted using Farmer's model and the aggregate area of all
openings in the building foundation (e.g. cracks, floor-wall joints, loose
1011
-------
fitting pipes, weeping tiles). The average California home has 2-10 cm
of openings per m2 of floor area, and the values in this range are greater
than the estimated leakage area for commercial buildings (Grimsrud et al.,
1983). Therefore, this assessment uses the lower limit of the range for
homes in California, 2 cm per m of floor space.
The ventilation rate for the proposed restaurant was assumed to be 7
ft3/min per person inhabiting the building. This rate is suggested as a
minimum ventilation rate by the American Society of Heating,
Refrigerating, and Air-Conditioning Engineers (ASHRAE) for fast food
restaurants in which smoking is not permitted. If smoking is allowed, a
5-fold increase in the ventilation rate is required. These ASHRAE
standards are used in the building codes of 45 states in the U.S. (ASHRAE,
1981). The proposed restaurant will have an occupancy of 14 persons,
resulting in a minimum ventilation rate of 98 ft^/min or 0.046 rn^/sec.
Estimating Carcinogenic Risk Associated with Inhalation of Benzene
The incremental lifetime cancer risk associated with exposure to
airborne benzene vapor inside the proposed restaurant is estimated using
the following equation:
RISK - Indoor Air Concentration (Jig/m3) x Unit Risk Value (Mg/m3)"1
The Unit Risk Value (URV) for inhalation Is a chemical-specific value
derived from the chemical's inhalation potency factor. The URV Is defined
as the risk posed by inhalation of air containing 1 Mg/m3 of the chemical.
It assumes a breathing rate of 20 m3/day, a body weight of 70 kg, and a
lifetime of constant exposure. Since no one is expected to spend their
entire lifetime in a restaurant, this approach adds another layer of
health-conservatism to this screening-level assessment.
Results
Based on soil characteristics measured at the site and chemical-
specific parameters for benzene, the flux of benzene vapor from soil was
predicted to be 2.9 x 10"8 mg/sec-cm2, using Farmer's model. Based on
this upper-bound flux estimate, and the fact that the floor area of the
restaurant will be 74 m2' the generation rate of benzene vapor In the
indoor air of the restaurant Is calculated as follows:
c = FA/Q = 2.9 x 10"8 mg/sec-cm2 x 74 m2 x 2 cm2/m2 / 0.046 m3/sec
9.3 x 10'5 mg/m3 = 0.093 /ig/m3
The inhalation URV for benzene is 8.3 x 10'6 (/Llg/m3)"1 (SHEAS, 1989),
which results in a incremental lifetime cancer risk of 7.7 x 10"7 or 7.7
in 10 million.
Conclusions
In this screening analysis, an upper-bound for the potential cancer
risk posed by indoor benzene vapor emitted from gasoline-contaminated soil
underlying a proposed restaurant was evaluated. Because the building In
this case study had not been built yet at the time the risk assessment was
1012
-------
required by a regulatory agency, a predictive method was used to estimate
the worst-case indoor airborne benzene concentration. The indoor benzene
concentration was based on the benzene emission rate from the soil
underlying the building foundation, which, due to time and cost
constraints, was also estimated using a predictive tool. The estimated
vapor flux from soil was also an upper-bound prediction.
Predictive models such as the ones used in this study are very useful
tools for the risk assessor, because cost and/or time constraints often
preclude the measurement of chemical concentrations in environmental
media. Also, some chemicals may pose a health threat at levels which are
not detectable using currently available monitoring procedures, so
predictive modeling may be the most feasible way to evaluate the potential
for environmental contamination to threaten public health.
The incremental lifetime cancer risk associated with exposure to the
benzene vapor inside the restaurant was calculated using standard
inhalation exposure assumptions and a risk prediction factor for benzene
suggested by the EPA. The calculated risk is certainly an upper bound for
the potential incremental cancer risk posed by benzene emissions from the
gasoline-contaminated soil. The carcinogenic risk estimate was 7.7
increased cancer risks for every 10 million people exposed to the benzene
vapor. Any risk lower than 1 in 1 million is traditionally considered to
be an insignificant risk by environmental regulatory agencies (Young,
1987).
This assessment is a screening analysis that is likely to greatly
overestimate the actual levels of human exposure. Its purpose is to
determine whether the site warrants additional study to more accurately
estimate the true risks. Since the calculated risk level was so low, it
is clear that a more detailed analysis is not necessary. In summary, this
analysis shows that modeling can be a useful tool in screening assessments
of contaminated soil.
1013
-------
References
ASHRAE, 1981. The American Society of Heating, Refrigerating, and Air-
Conditlonlng Engineers, Inc. ASHRAE Standard. "Ventilation for
Acceptable Indoor Air Quality."
Farmer, W.J., M.S. Yang, J. Letey, and W.F. Spencer (1980). Land Disposal
of Hexachlorobenzene Wastes: Controlling Vapor Movement in Soil. EPA-
600/2-80-119, U.S. Environmental Protection Agency, Office of Research and
Development, Municipal Environmental Research Laboratory, Cincinatti, OH.
69 pp.
Jury, W.A., Spencer, W.F., and Farmer, tf.J. (1983) Behavior assessment
model for trace organics in soil: I. Model description. Jour. Environ.
Qual. 12: 558-564.
SHEAS, 1989. Superfund Health Effects Assessments Summary Tables and
User's Guide (SHEAS). Office of Emergency and Remedial Response,
Washington, D.C.
Grimsrud, D.T., M.H. Sonderegger, and R.C. Sherman (1983). A Framework of
a Construction Quality Survey for Air Leakage in Residential Buildings, In
Proceedings of Thermal Performance of External Envelopes of Buildings.
Page 422-452, published by ASHRAE.
Grimsrud, D., Univ. of Minnesota, Personal Communication, August 17, 1989.
Young, F.A. 1987. Risk Assessment: The Convergence of Science and Law.
Regul. Toxicol. Pharmacol. 7:179-184
Zapalac, G.H. (1983) A time-dependent method for characterizing the
diffusion of 222Rn in concrete. Health Physics 45 (2):377-383.
1014
-------
A REVIEW OF THE ACCURACY AND PRECISION OF THE TOXIC AIR
CONTAMINANT MONITORING PROGRAM OF THE CALIFORNIA AIR RESOURCES
BOARD
Catherine Dunwoody and Robert Effa
Monitoring and Laboratory Division
California Air Resources Board
Sacramento , CA
The California Air Resources Board (CARS) operates a 20
site air monitoring network to measjre ambient levels of toxic
air contaminants (TAC). The network has been in operation
since 1985. Gaseous samples are collected over a 24-hour
period using a low volume gas sampler designed by CARB. The
samples are transported to the laboratory for analysis.
Several routine checks are used to assess the accuracy and
precision of this system, and to ensure that data of the
highest possible quality are produced. The results of several
quantitative measures used to assess the quality of the data
are presented in this paper.
In 1988, the CARB staff initiated a performance audit
program which is designed to test the accuracy and precision
of the overall sampling and analysis system for gaseous TACs.
The audit procedure involves introducing an audit sample
directly into the intake probe of tie sampler. Results for
the most recent year of these "throjgh-the-probe" audits show
that, on average, the monitoring system is unbiased for 8 of 9
compounds. Several individual audit results are in error by
up to 70%, although the majority (70%) of the results are
within ± 20%. The average precision for all compounds is
± 40%. The individual estimates of precision for 6 of 7
compounds compare well with the precision estimates derived
from ambient collocated data.
1015
-------
Introduction
The California Air Resources Board's (CARS) monitoring
network for gaseous toxic air contaminants (TAC) consists of
20 sites where 24-hour samples are collected twice each
month . Samples are collected using a low volume sampler
designed by CARB. From 1985 through 1989, Tedlar bags were
used as the collection media. Beginning in 1990, 6 liter
Summa polished canisters are used to collect samples. Samples
are transported to the laboratory for analysis of 10
compounds. In 1988, a "through-the-probe" audit system was
implemented to test the accuracy and precision of the entire
gaseous TAC sampling, transportation and analysis procedure.
The audits are conducted annually at each of the sites in the
TAC monitoring network. The results of through-the-probe
audits yield information regarding the quality of the ambient
TAC data produced.
Configuration and Operation of the Audit System
The through-the-probe audit system (Figure 1) consists of
a pure air source (Aadco 737 pure air generator plus a methane
reactor), a gas cylinder containing Research Triangle
Institute (RTI) certified concentrations of TACs, and a
dilution system (a modified Dasibi 1009 MC). The dilution air
generated by the pure air source has been analyzed by GC/MS
and GC/ECD-PID and is well within CARB's laboratory criteria
for clean air acceptability. The Dasibi 1009 MC was modified
by replacing all sample lines and fittings with cleaned
stainless steel parts, removing the large mixing chambers and
enclosing the mass flow controllers in an insulated chamber
with a heater to minimize potential wall effects. The mass
flow controllers are certified every three months against a
National Institute of Standards and Technology (NIST)
traceable primary flow standard. The development of the
through-the-probe audit method is described in more detail
elsewhere .
The output of the dilution system is directed to the
inlet of the low volum'e toxics sampler (Xbntech Model 910) for
a 24-hour period. The sample is collected and transported to
the laboratory for analysis in the same manner as an ambient
sample . The audits are performed on routine TAC sampling
dates, so the laboratory is not aware that the sample is an
audit until after the analysis is completed.
Most of the through-the-probe audit data available to
date is from audit samples collected in Tedlar bags. CARB is
currently in the process of switching to 6 liter canisters as
the collection media for TAC samples. Several audits have
been conducted at sites which have initiated canister
sampling, but currently there are insufficient data to
characterize canister samples. Results for audits using
canisters are not included in this paper.
Nine of the ten compounds monitored are included in the
audit cylinder. They are methylene chloride (DCM), chloroform
(CHC13), carbon tetrachloride (CC14), ethylene dichloride
(EDC), 1,1,1-trichloroethane (TCEA), ethylehe dibromide (EDB),
1016
-------
trich loroethene (TCE), perch 1 oroethylene (Perc) and benzene.
Although ambient TAG samples are analyzed for 1,3-butadiene,
this compound is not included in the TAG cylinder currently
used for through-the-probe audits.
Limitations and Assumptions of the Audit Procedure
The gas cylinder used for the through-the-probe audits
can be diluted to a wide range of air to gas ratios. However,
the relative concentration of the TACs is the same for each
audit. Additionally, the audit cylinder is a pure mixture of
the nine compounds listed above. Therefore the audits do not
account for matrix effects which may affect ambient data.
When the audit results are used to represent the quality of
ambient data, it is assumed that the ambient matrix does not
substantially affect the TAG results.
Another assumption essential to an analysis of the audit
results is that each site is indistinguishable from any other
site regarding the monitoring equipment and the way it is
operated. Since the goal of the program is to audit each site
annually, only one or two audits are currently available for
each site. The assumption makes it possible to group the
individual results into the same population. It is a
reasonable assumption since precautions have been taken to
ensure that common equipment is used and a common protocol is
followed at each site.
Summary of Results From 1989 - 1990 Audits
Although GARB has used the through-the-probe audit
procedure since 1988, only results from 1989 and 1990 audits
are presented here. Beginning in 1989, the laboratory has
used ambient concentration standards produced by NIST. All of
the results presented in this paper were obtained using these
high quality laboratory standards.
Results for 12 to 15 individual audits were averaged to
obtain mean bias for each compound. These data are presented
in Table I. The sample mean bias is statistically
significantly different than zero for one compound, methylene
chloride (DCM), at the 95% level of significance. However, a
regression analysis of DCM observed audit values versus
expected values shows that the slope of the linear
relationship is not significantly different than 1.0, and the
y-intercept is not significantly different than zero. This
suggests that the bias, although statistically significant,
is small compared to the variability of individual data
points.
The 95% confidence intervals for the true mean bias for
each compound are shown graphically in Figure 2. Only one
compound, DCM, has a mean bias that is significantly different
than zero. The 95% prediction intervals for individual
results are also shown. While the prediction intervals show
that individual results may be in error by up to 75% (eg. EDC
and TCEA), on average there is no bias in the results for 8 of
the 9 compounds. This suggests that imprecision rather than
bias is responsible for the inaccuracy of individual results.
1017
-------
QUANTITATIVE EXTRACTION AND ANALYSIS OF
ENVIRONMENTAL SOLIDS USING SUPERCRITICAL
FLUID EXTRACTION (SFE) AND SFE-GC
Steven B. Hawthorne, David J. Miller,
and John J. Langenfeld
Energy and Environmental Research Center
University of North Dakota
Grand Forks, ND, 58202
Supercritical fluid extraction (SFE) reduces the time needed for the
extraction and recovery of environmental pollutants from sorbent resins, air
particulates, and soils and sediments to 5-30 minutes compared to several hours
required by conventional liquid solvent extraction methods. SFE essentially
eliminates the generation of waste solvents, as well as the need for concentration
steps. When SFE is directly coupled to capillary GC (SFE-GC), sample collection,
extraction, analyte concentration, and gas chromatographic separation can be
completed in a total time of < 1 hour. The use of SFE and SFE-GC for the rapid
and quantitative determination of organic air, water, and soil pollutants including
PAHs, heteroatom-contalning PAHs, PCBs, wood smoke phenolics, fuel
components, and ionic surfactants will be discussed. Quantitative claims for SFE
and SFE-GC are supported by the analysis of NIST certified standard reference
materials (air and diesel particulates and marine sediment), multiple extractions,
and spike recoveries.
82
-------
Determination of Trends and Outliers
Figures 3 through 11 show individual audit results in
chronological order. The two lines above and below zero bias
are data screening lines. These lines are determined by
calculating two relative standard deviations (2RSD) using
valid audit results which represent normal operating
conditions. The lines are not straight because they represent
a moving 2RSD, ie. poijits on the data screening lines are
recalculated after each audit result.
The chronological plots can be used to identify trends
and unusual events. If an audit result exceeds the data
screening lines, or if a trend of increasing or decreasing
audit results occurs, the sampling and analysis procedures are
investigated. If no problems are detected, the result is
included in the calculation of the data screening lines and
used to determine if future data points are unusual. If a
problem is detected, corrective action is taken. Although the
audit result is valid, it is not used in calculating the data
screening lines because it is not representative of normal
conditions. In some cases, the problem is in the audit
procedure itself, and the audit results are considered
i nvalid.
Several data points or groups of data which exceed the
data screening lines can be seen in Figures 3 through 11. In
most cases when these audit results were investigated, no
problems were found. Therefore, in most cases the results
were included in the calculation of subsequent points on the
data screening lines. There were, however, two specific
examples of unusual results which led to the discovery of an
unsuspected problem.
When the TCEA result for audit 89-17 (Figure 7) was
investigated, the laboratory discovered an instrument
malfunction which affected TCEA only. The malfunction caused
negative biases in the three subsequent audits as well,
therefore these results were not included in the calculation
of the data screening lines. As a result of this discovery,
all routine sample analyses were shifted to a second
instrument until the problem was solved. Ambient TCEA data
for samples analyzed with the malfunctioning instrument were
deleted from the database.
EDB results for audits 89-10 through 89-13 (Figure 8)
show a trend of increasing bias. Upon investigation, it was
determined that the EDB concentration in the laboratory
standard was decreasing. Therefore these three audit results
were not included in the calculation of the data screening
lines. As of November 1989, a new standard was purchased and
assigned for EDB based on the NI5T primary standard.
Calculation of Method •Variability
If a sufficient number (10 to 15) of through-the-probe
audits have been performed, the results can be used to
estimate total system method variability. This estimate of
total system method variability can be compared to the
1018
-------
estimate obtained using collocated sampling data for each
compound where both data sets are available. Method
variability (defined as ± 2RSD) for each compound is presented
in Table II.
Three of the 20 sites in CARB's TAG network have
collocated (duplicate) gaseous samplers. Method variability
can be estimated from collocated data using the following
equat ion.
Method variability = ( 2 * s / avg. cone.) * 100
4
where s is estimated by
s = [ (sum of differences}2/ 2 * k ]1/2
where k = number of data pair:;
Data from all three sites were combined to obtain
estimates of method variability. It is assumed that there are
no important differences between sites. It is also assumed
that these three sites are representative of the other 17
sites in the network. Because each site has similar
monitoring equipment and is operated according to the same
protocol, these are reasonable assumptions.
Total system variability estimates using the two
estimation methods are comparable for 6 of 7 compounds. The
collocated method variability estimate for TCE is
significantly higher than the through-the-probe estimate.
This may be due to the lower average concentration of TCE
(0.17 ppb) detected in ambient air as compared to the average
audit concentration (1.0 ppb). Two compounds, EDC and EDB,
are consistently lower than the detection limit in ambient
air. Therefore no col.located method variability estimate is
available. Since 1,3-butadiene is not contained in the audit
cylinder, a through-the-probe estimate of method variability
is not available.
Future Plans for the Through-the-Probe Audit Procedure
The current through-the-probe audit system is designed to
dilute a single audit cylinder. Although the cylinder can be
diluted to a wide range of ratios, the relative concentrations
of the compounds are always the same for that audit cylinder.
CARB is currently developing a modified dilution system which
will dilute up to three cylinders to different ratios. The
result will be more flexibility in the relative concentrations
of the audit gas mixture. Additionally, this system will
allow CARB to perform audits using NIST standards which are
similar to the laboratory NIST standards but at higher
concentrations. Due to gas stability issues, the NIST
certified compounds are contained in four separate cylinders.
The use of these standards for through-the-probe audits has
been limited because the current system can dilute only one
cylinder at a time; therefore only a few compounds could be
audited at one time.
1019
-------
Conclusions
CARB's through-the-probe audit system has been in
operation for two years. The most recent data, collected in
1989 and early 1990, show that CARB's TAC monitoring network
is unbiased for 8 of 9 compounds. The exception, methylene
chloride, is biased by an average of + 124. The variability
of the audit results is representative of the variability of
routine ambient data for 6 of 7 compounds. The average
variability for all compounds is ± 40 i at the audit
concentrations of TACs. These results suggest that the
potential inaccuracy of any individual data point is due
primarily to variability in the sampling and analysis system,
rather than bias.
References
1, D. Crowe, "Defining Toxics Problems at the State Level-
The State of California's Monitoring Program", Mon i tor i ng
Methods for Toxic-s in the Atmosphere. ASTM 5TP 1052. W.L.
Zielinski, Jr. and W. D. Dorko, Eds., American Society
for Testing and Materials, Philadelphia, 1990, pp. 25-33.
2. R. Effa and P. Vanicek, "A Procedure For Measuring Total
System Error In Ambient Organics Toxics Monitoring",
Proceedings of the 1989 EPA/APCA International Symposium
on Measurement of Toxic and Related Air Pollutants. (May
1989).
3. "Standard Operating Procedure for the Determination of
Volatile Organics in Ambient Air Using Tenax Trap
Preconcentration Gas Chromatography and Tandem
Photoionization/Electron Capture Detectors", Method 002,
California Air Resources Board, Monitoring and Laboratory
Division, October 1986.
4. J. Taylor, Quality Assurance of Chemical Measurements.
Lewis Publishers, Inc., Michigan. (1987)
Table I.
Mean Bias and Standard Deviation for Through-the-Probe Audits
January 1989 - February 1990
Compound n Mean Bias SD Bias p value Significant
(56) (%) at alpha = 0.05?
DCM 15 11.9 15.9 0.01 yes
CHC13 15 3.4 20.6 0.54 no
CC14 15 1.3 12.5 0.70 no
EDC 14 0.2 34.7 0.99 no
TCEA 12 -5.2 32.3 0.56 no
EDB 13 -4.3 15.6 0.34 no
TCE 15 -2.9 10.9 0.32 no
Perc 15 6.4 14.7 0.11 no
Benzene 15 0.0 15.4 1.00 no
1020
-------
Table II.
Comparison of Total System Method Variability (2RSD)
Through-the-Probe Audit Estimate and Collocated Estimate
Compound
DCM
CHC13
CCL4
EDC
TCEA
EDB
TCE
Perc
Benzene
Butadiene
Avg. TTP
Audit Cone
(ppb)
2.4
0.35
0.24
1.1
1.1
0.24
1 .0
0.32
2.4
MA
Avg. Detected
Ambient Cone.
(ppb)
1
0
0
<0
0
<0
0
0
2
0
4
03
12
2
9
01
17
27
4
34
TTP
Est imate
(*)
31.7
41.2
25.0
69.5
64.6
31.3
21.7
29.4
30.7
NA
Col located
Est imate
(*)
44. 1
37.1
15.3
NA
76.7
NA
125.6
28. 1
24.5
23.8
* Average detected ambient concentration is calculated using
ambient data above the limit of detection ( > LOD ) only.
F i gure 1.
Dfigrim of thi "Through-Thi-Prob*' Audit Syitca
To Out** Vwt
1021
-------
Toil* Air ContealMMt Accuracy «nd ProcUlo*
(••t«d on 191* - It** Thr*«|h-Th«-rr
X
CHOJ|n-1S) OCL4(n-1S)
R-14) TCEA(n-12)
TOE(««15)
("•IS)
961 Confidence Int«rv«l for Hun Bttt V ' 9B1 Prediction Intervel for Indtvtdutl Viluei
X
F I gure 3.
SOO -r
r B
c I
e t
rt ft
I
Through-The-Probe Audttt 19*9 - 1990
Olchloroaethine (DCM)
Av«r«g» Detected Aabtent Concentration • 1.4 ppb
Avenge TIP Audit CencentritIon • 2.4 ppb
-300 --
-400
-SOO
ZftSO
ZRSO
8B-i m-i eo-3 »a-4 m-s m-t m-i w-io w-iz w-u M-IS W-K M-IT *»-ie n-i
1022
-------
Thr«ufk-TM-Pr*M Audit* !••> - 1*«0
Cbl»rtf*r* (CHCIJ)
Avtrif* Dtt«ctM A«H«»t Concentration . 0.01 ••»
Avortg* m Audit Concontrttto* • fl.JS ppb
2RSO
-2RSD
M-7 M-10 W-12 »11 W-K W-1* W-17 «*-1l fO-1 W-2
s,
Throuoh-Th»-Prot>« Audtti 1)19 - 1*90
Ctrbon T«tr tcMor Id* (CCL4)
A»«rtg« 0*t*ct*4 A»bt»nt ConctntritIon • 0.12 ppb
Avcrig* TIP Audit Conetntrit Ion • 0.24 ppb
2RSO
-ZRSO
IB-3 M-4 «0-S t»4 W-7 W-10 W-12 W-1J **-!$ «•>!• W-17 M-K
90-2
1023
-------
•00 -T-
400 -
200 --
0.0
•400
••00 --
••0.0 -L
' ltt«
Ethyl*** B1cMortd« (CDC)
Avtrig* D«t«ct*4 Avbltnt C«nc*ntrltIon - <0.2 ppb
Avcrif* TTP Audit Concintntlon • 1.1 ppb
'-JRSO
«*-S
W-10 H-12 M-11 W.1S «».!« ».,7 „.
F1gur» 7.
Througri-Th«-Prob» Audits 1919 - 1990*
1,1,1-Trlchloroithtnt (ICEA)
Av*r«gt D«t»ct«d Avblint ConcintrttIon - 0.9 ppb
Av«rtg« TTP Audit Cone»ntr«tton - I.I ppb
2RSO
2RSO
te-7 M-10 *B-12
W-15 »1« M-17 W-11
00-2
Audtti «9-17 through 90-Z not Included In ditt jcrttnlng (ZRSD) lines.
1024
-------
M|vrt I.
Throvfft-TM-rrtM *M*1t» 11** - 1W*
Ethyl*** Ol»r*»)«« (CO!)
Av«ng« D*ttct«4 Aabttnt C«*ctntntten • 2RSO
10 --
TIT
H ^
No CM*
-10 4
-20
~-2RSO
M-5 104 W-7 M-10
W-1S W-1C W-17 Wl
«0-1
1025
-------
II.
» IB +
I
a
• 0
•10
-30 --
AytfUl 1919 - 1*10
(Mre)
Aa*1«nt C»«Ci«tr«t 1«i» • O.ZT ppb
Avtrtf* TTP Ayrflt Canecntrttlon • 0.32 ppb
*2RSO
^-H
-2RSD
M-IO M-I
-i* W-IT
F igurt 11.
30 --
20 --
10 --
Through-Th«-Prob» Audttt 1919 - 1990
B«iutn*
Aving* Dttttttd A»b1tnt Conctntrtt 1on - Z.4 ppb
A»»r«g* TTP Audit Concentration • 2.4 ppb
JZL
•2RSD
•10 --
•X --
•90 --
TI
-2RSO
W-1 1*4
»»10
-13
•*• W-IT M-U 90-1 90-2
1026
-------
Uncertainty Estimates for the NAPAP
Material Exposure Monitoring Network Data
P. Michael Barlow
Applied Technology Division
Computer Sciences Corporation
Research Triangle Park, NC
Richard C. Shores
Center for Environmental
Quality Assurance
Research Triangle Institute
Research Triangle Park, NC
John If. Spence and Fred H. Haynie
Atmospheric Research and
Exposure Assessment Laboratory
Research Triangle Park, NC
Abstract
This materials effects study was conducted to provide research data for the
National Acid Precipitation Assessment Progran (NAPAP), to determine the
effects of acid precipitation on materials, and to aid in the development of
models describing damage to air quality I concentrations of SOg, 03, NOg, NO,
NOX) and certain meteorological variables {tenperature, humidity/dewpoint, and
wind speed),
To determine the accuracy of the models, lit was necessary to estimate the
uncertainty of the aerometric variables using quality assurance (QA) data.
Performance and system audits were conducted on the Materials Exposure
Monitoring sites by Research Triangle Institute, under contract to the
Environmental Protection Agency. A total of five QA audits were conducted over
a 4-year period, from 1985 through 1988.
Two methods were used in making the uncertainty calculations. In the first
method, the data were fit to simple linear rejjr ess ions so that the uncertain-
ties could be estimated using the confidence limits on the regression at speci-
fic concentrations. In the second method, uncertainties were estimated from
statistics calculated on the differences between the site and audit con-
centration measurements. The two methods produced consistent results.
Introduction
A field exposure study investigating the effects of acid deposition on
materials damage was performed by Task Group VII (Effects on Materials and
Cultural Resources) of the National Acid Precipitation Assessment Program
(NAPAP). Samples of various materials were exposed to the environment at five
field sites in the eastern United States, Ai:~ quality (including particulate
concentrations), meteorology, and rain chemistry [provided by Bureau of Mines
(BOM)] were measured continuously to allow accurate assessment of the quanti-
tative effects of acid deposition on materials.
The purposes of this study were to (1) determine the extent of damage due
to acid deposition on materials, and (2) aid in the development of models
describing the materials effects damage. The models relate materials damage to
air quality parameters (such as concentrations of SC^, 03, NOg, and NOX) and
meteorological parameters (such as temperature, wind speed, and humidity/
dewpoint).
1027
-------
Introduction
The extraction of organic pollutants from environmental solids including
sorbent resins, air particulates, sediments, and soils, is often the slowest and most
error-prone step of an entire analytical scheme. It is ironic that, while
chromatographic methods can separate, identify, and quantitate hundreds of
individual species per hour, the most frequently used sample extraction method
(liquid solvent extraction in a Soxhlet apparatus) was in common use when Tswett
first reported chromatography in 1906. Besides requiring several hours to perform,
liquid solvent extractions yield samples that often need to be concentrated, result
in large volumes of waste solvent (and exposure of laboratory personnel to
potentially harmful solvents), and can introduce considerable error into the final
analytical result.
Supercritical fluids have several characteristics that make them useful for
analytical-scale extractions of organic pollutants from solid matrices:
1) Supercritical fluids have solvent strengths that approach those of liquids
but, in contrast to liquid solvents, the solvent strength of a supercritical fluid can
easily be controlled by the pressure (and temperature) of the extraction. Extraction
at lower pressures (e.g., 80 atm) favors less polar analytes while higher pressure
extractions (e.g., 400 atm) favors more polar analytes.
2) Mass transfer in supercritical fluids is ca. two orders of magnitude faster
than in liquids. Thus, SFE can be completed much faster than conventional liquid
solvent extractions.
3) Supercritical fluids such as C02, N20, and SF6 are gases at ambient
conditions, which simplifies sample concentration steps and makes the direct
coupling of SFE with capillary GC simple to perform.
Experimental Methods
Supercritical fluid extractions were performed using syringe-type
supercritical fluid pumps (Suprex and ISCO) and either C02, N2O, or C02 with
added methanol modifier Supercritical pressures were maintained inside the
extraction cells (0.1 to 10 ml_ depending on sample size) with 15 to 30 um i.d. X
150 um o.d. fused silica capillary tubing for outlet restrictors. Temperature was
maintained during extraction by inserting the cell into a thermostatted tube heater.
For non-coupled SFE, the extracted species were collected by inserting the outlet
restrictor into a vial containing 2 ml methylene chloride. GC/FID and GC/MS
analyses of these extracts were performed in a normal manner. The direct
coupling of the supercritical fluid extraction step with gas chromatography (SFE-
GC) was achieved by inserting the SFE outlet restrictor directly into the capillary
gas chromatographic column through the on-column injection port (on-co!umn SFE-
GC) or by inserting the restrictor into a split/splitless injection port through an SGE
septumless injector (split SFE-GC). Extracted species were cryogenically trapped
in the gas chromatographic column which was held at - 30 to 5°C. After the
extraction was completed, the restrictor capillary was withdrawn from the injector
and gas chromatographic analysis was performed in a normal manner.
83
-------
The Atmospheric Research Effects and Assessment Laboratory (AREAL) of the
U.S. Environmental Protection Agency (EPA) supported this effort by the cre-
ation and maintenance of the Materials Aerometric Data base (MAD). AREAL has a
contract with Computer Sciences Corporation to maintain, analyze, and validate
the data constituting the MAD. The MAD contains air quality and meteorological
data, measured at the five exposure sites. On-site system and performance
audits were conducted by Research Triangle Institute (RTI) at each of the five
sites.
Materials Effects Study
Continuous measurements of the air quality, meteorological, and
particulate data were made at five field exposure sites, including the
following:
Sulfur dioxide (S02)
Ozone (03)
Oxides of nitrogen (NOX)
Nitric oxide (NO)
Nitrogen dioxide
Wind speed average
Wind direction vector
Temperature
Dew point
Relative humidity
Precipitation
Solar radiation
The deposition of these species of pollutants in the presence of water or water
vapor leads to the corrosion of exposed materials.
The locations of the material effects sites were located in Newcomb, NY;
Chester, NJ; Steubenville, OH; Washington, DC; and RTP, NC. A description of
the specific sites and monitoring instrumentation has been previously reported
[1,2]. The formats and detailed summaries of the data are described elsewhere
[3,4]. A summary of the S02 concentrations for the sites for 1985-1987 is
given in Table I,
Materials Exposure Network Audits
Audits were conducted over a 24-hour period, initially just once a year
for each site but, increasing to twice a year during the latter years of the
study. On-site audits were conducted over a four year period from 1985 to 1988
for a total of five audits during the study [5].
The purpose of the audit was two-fold: (1) to furnish a means of rapidly
evaluating the specific operations atmospheric measurement and data recording
devices at the sites, and (2) to provide a continuing index of the validity of
site environmental data.
Prior to the audit, all auditing systems were verified at RTI and/or EPA
using traceable and/or known standards. The gaseous analyzer auditing system
was verified using analyzers calibrated with the National Institute of
Standards and Technology (NIST) Standard Reference Materials (SRMs) and/or
primary standards. The flow rate auditing devices were verified (calibrated)
using primary volumetric standards. The meteorological auditing systems were
verified (calibrated) by either the manufacturer or by using primary standards.
On-Site Audit Procedures
Gaseous Analyzers: The on-site audit procedure for the gas analyzers
consisted of challenging the monitors with pollutant concentrations over two
1028
-------
ranges:
(1) 0 to 25 ppb and (2) 0 to 500 ppb.
The lower range is representative of actual ambient air conditions, while the
higher range is comparable to other QC/QA activities conducted at the site.
Test levels of 03 were generated using an ultraviolet ozone generator as
described in the Code of Fj-jeral Regu1ations, Appendix D of Part 50. Test
levels of SOg were generated using dynamic dilution. NO and NOg test levels
were generated using cylinder dilution and gas phase tiltration, respectively,
Meteorological Instrumentation: The audit procedure for meteorological
sensors was to collocate audit sensors. A data logger was used to record not
only the auditing sensor signal, but also the site sensor's signal to ensure
compatibility in data reduction procedures.
Particle Collection Devices: A high volume orifice was used to audit the
flow rate of the high volume samplers. Two calibrated orifices were used to
audit the total, fine, and coarse flow rates of the dichotomous samplers.
System Audit and Data Acquisition: A systems audit was conducted accord-
ing to guidelines given in the Quality Assurance Handbook for Air Pollution
Measurement Systems, Vol. II, [6]. Important points which were considered
during the on-site evaluation included (1) evaluation of site according to EPA
siting criteria, (2) set up of continuous monitors and meteorological
instruments, (3) availability and suitability of the calibration system, and
(4) check of the site operator's recordkeeping and documentation including
traceability of calibration standards.
Collocation of ambient air data and meteorological data collection were
conducted with a Fluke Model 2280a data acquisition system. This system was
programmed to scan every 10 seconds and calculate hourly averages. Using these
data (volts) and the calibration relationship as provided by the site operator,
a comparison was made between hourly data reported by the site and hourly data
acquired during the audit. Differences between the hourly averages and
standard deviations of the differences were calculated and reported [5].
Audit Results: After each site audit, the findings were discussed between
the audit team and the site operators so that corrective action could be taken
immediately. Upon completion of the audit, these findings were presented to
EPA management to insure that corrective action had been taken at each site.
Linear regression coefficients were calculated, assuming linearity of the site
analyzers. The criteria for evaluation of the gaseous analyzers were based on
the calculated slope and intercept, as follows:
Satisfactory = 0.85 _< slope _> 1-15, -3% full scale _< intercept <+3%
full scale
Unsatisfactory = slope <0.85 or >1.15, and <-3% full scale or > +3%
full scale.
The criteria for evaluating the meteorological instrumentation were based upon
the recommended tolerance limits as given in the Quality Assurance Handbook
Vol. IV [7]. The performance of the particle samplers was based on the average
percent difference in flow rates:
1029
-------
Satisfactory = * difference < ± 10%
Unsatisfactory = % difference > _+ 10%
Uncertainty Estimation
The audits described above provided data which were used to estimate the
measurement uncertainties. In the following discussion we will limit ourselves
to 862 uncertainties.
A preliminary method of data analysis is to plot the site measurement
versus the audit concentration levels to look for trends. For an accurately
calibrated instrument, the plot would be a straight line with a slope of 1,
with the site measurement exactly correlated to the audit concentration.
Differences between the site measurement and the audit value were calcu-
lated for each pair of data within the concentration ranges. In this case, the
differences themselves represent the uncertainty of the site measurements.
Di = Yj - X|
Yj is the site measurement, and
Xj is the audit concentration or measurement
The average of the difference can be attributed to the calibration bias of
the site instrument. The standard deviation of the differences represents
actual measurement variability of the site instrument. The average difference
(D) and the standard deviation of the difference (o^) were calculated for each
concentration range using:
D = (E Di)/n
-------
Results
Table II contains the uncertainty estimates for the indicated analyzer
concentrations using the two methods described above.
The QDEV displays lower values at lower concentrations, giving a better
approximation to the spread in the real data, The average DEVO did not provide
consistent results between sites, but did provide estimates representing the
spread of the real data.
For the air quality variables, in this case S02, the QDEV and the DEVO
displayed increasing spread in the data for higher values. Both QDEV and DEVO
measure the uncertainty in slightly different ways and give similar results.
Since both statistics are conservative, the best estimate of the uncertainty is
the minimum of the pair.
Conclusion
Uncertainties for S02 means are given in Table III. These uncertainties
were calculated using the QDEV estimates, averaged over the indicated range of
ambient air concentrations. The uncertainties in the lowest range are on the
order of 5 ppb, which is comparable with the instrument capabilities. The
calculations indicated that uncertainty estimates should be used on a site-by-
site basis.
1. D.R. Flinn, S.D. Cramer, J.P. Carter, and J.W. Spence, "Field Exposure
Study for Determining the Effects of Acid Deposition on the Corrosion and
Deterioration of materials — Description of Program Preliminary Results,"
Durability of Building Materials, 3:147-175 (1985).
2. R-T Tang, P.M. Barlow, and J.W. Spence, "Monitoring and Operations at
Materials Effects Sites," U.S. Environmental Protection Agency Internal
Report (1988).
3. R-T Tang, P.M. Barlow and J.W. Spence, "Materials Aerometric Database for
use in Developing Materials Damage Functions," Environmental Protection
Agency Project Summary (1989).
4. NSI particulate data report,
5. R.C. Shores and R.W, Murdoch, "Performance Audit of the Material Exposure
Air Monitoring Network," Five RTI Technical Reports, Research Triangle
Institute (1985-1988).
6. Quality Assurance Handbook for Air Pollution Measurement Systems: Volume
II, Ambient Air Specific Methods, U.S. Environmental Protection Agency.
7. Quality Assurance Handbook for Air Pollution Measurement Systems: Volume
IV, Meteorological Measurements, U.S. Environmental Protection Agency.
1031
-------
Table I. MAD S02 concentrations summary for 1985-86-87*
Mean
Min
Max
DC
11
0
143
NC
3
0
80
NJ
6
0
188
NY
2
0
29
OH
22
0
500
* Values in ppb.
Table II. Uncertainty of S02 data, Washington, D.C. exposure site*
Cone .
0
5
10
15
20
QDEV
6
7
7
9
8
DEVO
9
9
9
11
10
Cone.
0
50
100
200
300
500
QDEV
7
7
11
17
27
42
DEVO
10
8
13
21
32
51
* Values in ppb.
QDEV = Quadrature of deviation.
DEVO = Square root of the average squared deviation from zero.
Table III. MAD S02 uncertainties by concentration (in ppb)
Concentration
DC
NC
NJ
NY
OH
Range
0
5
20
100
200
5
- 20
- 100
- 200
- 300
> 300
6
7
9
14
22*
34*
7
11
12
16*
20*
24*
4
6
14
15
16*
22*
2
6
18*
37*
46*
68*
8
8
10
22
41
61
Uncertainty estimates above maximum reported ambient air concentration.
1032
-------
QUALITY ASSURANCE FOR CONTRACT SOURCE TESTING
IN THE SOUTH COAST AIR QUALITY MANAGEMENT1 DISTRICT
(CALIFORNIA)
Gary DIxon and Ramiro Gonzalez, Jr.
South Coast Air Quality Management District
Quality Assurance Branch
Technical Services Division
9150 Flair Drive
El Monte, CA. 91731
The South Coast Air Quality Management District (SCAQMD) conducts
about 500 stationary source tests each year, To augment and expand
SCAQMD's testing capabilities to 750 source tests, required additional
tests are performed through contract laboratories. This paper discusses
the various activities in Quality Assurance (QA) for the Source Test
Contracts Program in the SCAQMD.
Developments leading to the establishment of the QA program are
enumerated and discussed including a rerview of QA that applies to
fundamental principles and methods for the attainment of accurate and
reliable results. The discussion includes j;ome of the specific factors
and elements of the QA Plan such as pre:test preparation, sample collection
and analysis, data reporting and validation.., calibration, concurrent
testing and audits. Also entailed are the QA reporting process and
practices, development of audit findings, evaluations, recommendations,
implementation of corrective actions, and follow-up resolutions.
As an integral part of the QA program, procedural guidelines and
specific criteria for system and performance: audits have been established
and performed. Also discussed are guidelines and criteria addressing the
evaluation of acceptability, validity of data and adequacy of documentation
in contract source test reports.
1033
-------
Introduction
In January 1987, the South Coast Air Quality Management District
implemented a program after nearly two years of development that would allow
contractors to perform a portion of the stack emissions tests conducted
by the District to determine source compliance with air pollution rules
and regulations. This program utilized pre-qualified, established Engi-
neering firms as contractors selected through an open competitive bid
process. Contracts were awarded based upon fixed-price bids for nine
separate groups or zones of sources (nominally 25 sources per zone). The
zones contained sources grouped geographically that included glass melting
furnaces, asphalt batch plants, aluminum furnaces, cement plants, foundries,
gas turbines, boilers and various other industrial processes or equipment.
Testing was limited to measurements for particulate matter concen-
tration, carbon monoxide, oxides of nitrogen (NO ), sulfur dioxide (and other
oxidized sulfur compounds), lead concentration, oxygen, water content and
mass emission rate. The South Coast District test methods were specified
for all tests except those in which New Source Performance Standards (NSPS)
were applicable. The NSPS tests were performed according to the appropriate
Federal (EPA) test methods as found in the Code of Federal Regulations,
Title 40, Part 60.
The Quality Assurance Program was designed to ensure that all testing
performed under the contracts program met the high standards required for
compliance determinations. This program consisted of: A quality assurance
plan relating quality controls such as standardized testing methods or
report formats to acceptability requirements; performance audits to
challenge the accuracy and precision of contractor test data; field audits
to enhance contractor performance; final report evaluations to validate
and approve contractor results; and source facilities follow-up to resolve
any safety or testability concerns encountered at the designated contract
sources.
Discussion
The Quality Assurance Plan was prepared to provide the contractors and
District staff with an organized description of the quality information and
activities required by the Source Test Contracts Program. The plan included
a brief project description, program staff organization (District),
distribution list and an outline of the Quality Assurance program. Included
in the Quality Assurance program were: Pretest preparation; sample collection
and analysis; data reporting and validation (with detailed chain-of-custody
procedures); calibration procedures; audits and concurrent testing; and,
other considerations (scope of quality assurance activies and an amendment
to the test method for lead (Pb) analysis to clarify contractor requirements).
The plan also included guidance for the determination of compliance with
District Rule 217 -.Provision for Sampling and Testing Facilities. This
information was provided to promote universal understanding and consistency
in evaluations of the testing facilities for sources to be tested by the
contractors.
1034
-------
The Quality Assurance Program was conducted in an intensely proactive
manner in order to maximize the benefits of Quality Assurance information
to contractor personnel and to minimize the number of rejected source tests.
Contractors were not paid for any source tests performed until the final
source test report was evaluated and approved by the District Quality
Assurance Manager. This rather strict control was necessary to ensure that
compliance information generated through the Source Test Contracts Program
was reliable and equivalent in quality to that produced by the District
in-house source test group. Proactive field auditing allowed contractor
personnel to receive real-time input from qu.ality assurance staff regarding
test methods, District policies, engineering practices or format require-
ments. Application of this input by the contractors to the in-progress
tests significantly reduced the number of rejected final reports.
Field Audits by District quality assurance staff provided validation
for the subsequent contractor source test reports; verification of contrac-
tor skills and capabilities; and technical support to the contractors.
The field audits utilized extensive checklists specific to the various
test methods being used for the source t.est. Pertinent observations,
comments and interactions were included on the audit form as well as
recommendations for enhanced performance. The field auditors performed
another very important function, particularly during the implementation
year of the program. Source operators or plcint managers reluctant to
a.ccept contractor testing for District Rules compliance determinations
could interact directly with District auditors who would, in turn, act as
"buffers" to facilitate the completion of the: source tests.
Pretest preparation included site-specific evaluations to assess the
safety and technical adequacy of the sampling facilities (can the test be
performed in a safe and technically reliable manner?) and contractor
planning to ensure that appropriate personnel and equipment were allocated
for the test. Source samples collected for the determination of particulate
matter require isokinetic (collection of particulate matter at the same *
velocity as that of the stack gases) sampling. This implies that the stack
gas velocity profile is steady without excessive turbulence or cyclonic
flow. The selection of appropriate sampling locations is a necessary pre-
requisite to the reliable measurement of particulate matter from emission
sources. Safety practices followed by the District may have been more
extensive than those followed by contract firms in the course of their
normal field testing. Guardrails, caged ladders, adequate platforms and
appropriate sampling ports were installed in order to provide safe access
and sampling conditions. District test methods require specific procedures
to be followed so that follow-up actions will be enforceable and legally
valid. Due to the rigidity of District testing protocols, contractors had
to place considerable emphasis upon pretest evaluation and planning. The
contractors used the information gained to compensate for the loss of
flexibility (in the field).to modify methods, equipment or techniques to
adjust for unforseen conditions.
1035
-------
Performance audits were conducted to assess the analytical performance
of the contractors. These audits consisted of "blind" (preanalyzed samples
submitted without indicating concentration values) audit materials and user-
friendly audit forms were developed specifically for this program. The
audit samples simulated:
1) SO. and SO. collected in an impinger train
2) NO collected in a glass bulb/absorbing solution
3) Preweighed particulate filters
4) NO , SO,,, CO, CO and 0^ in compressed gas cylinders
to simulate extractive stack analysis by continuous
monitoring instruments (Mobile Source Test Vehicles or
MSTV's)
Comparative data was also collected by field auditors to evaluate
temperature, stack gas velocity and physical dimension measurements through
the use of a traceable temperature reference, standard pitot tube and
measuring tape. A limited number of parallel source tests between District
and contractor test teams provided additional verification of the overall
contractor reliability.
Report evaluation and approval were the most significant and effective
broad-based quality controls in the program. By linking the payments to
contractors for work performed with District _approval of final test reports,
the quality of contractor testing could be very effectively managed. Final
source test reports submitted by contractors were thoroughly reviewed by
District Quality Assurance Branch staff to confirm that the appropriate test
methods, equipment, plant process conditions, and technical accuracy
requirements were followed. Reports found to be substandard were either
returned to the contractor for revisions or they were rejected. In some
cases, however, reports from tests conducted under less than optimal
conditions were approved for information purposes only. These reports were
evaluated on a case-by-case basis with the decision being based upon
District need for the information, the scope of the test deficiencies and
the extent of the efforts made by the contractor to meet all 'specified test
requirements. The test report approval consisted of an approval as to form
by the contracts coordinator and the final approval/release for payment by
the Quality Assurance Manager.
Formalized contractor source test report audits as well as annual
quality assurance reports on the Source Test Contracts Program provided the
opportunity to evaluate the contractors over a broad range of activities in
a very detailed manner. In the report audits, twenty contractor source test
reports were selected at randon and thoroughly reevaluated according to
previously established criteria. Significant findings were reviewed to
determine technical validity as well as impacts on test quality or reliability
Annual quality assurance reports summarized the progress of the program for
District management and administrators. This information was used to
support the reallocation and funding of the program as part of the
budgetary process.
Summary
The quality assurance program developed by the South Coast Air Quality
Management District for source emissions testing by contractors was
designed to validate the source test data and reports generated by the
program.
1036
-------
Concern over the technical adequacy and le§;al enforceability of contractor
produced compliance information resulted ir. extensive, comprehensive
quality assurance requirements for the contractors, their resulting source
test activities and final test reports.
The program consisted of a quality assurance plan summarizing quality
requirements for contractor performance, various types of audits covering
all of the technical aspects of the program and evaluations of contractor
performance and reports as well as source facilities. Follow-up actions
were taken, as needed, to ensure that testing and reports maintained a high
level of quality and reliability and that E.ny observed weak areas were
resolved and not recurrent.
The Source Test Contracts Program has been proven to be an effective
adjunct to the South Coast District's source emissions compliance testing
program due, in part, to a comprehensive, proactive quality assurance
program.
Acknowledgement
Many thanks to Dipak Bishnu and John Higuchi for their work in the
development and administration of the SCAQMD Source Test Contracts Program.
References
1) Code of^ Federal Regulations, Title 40, Part 60,
Standards of Performance for New Stationary Sources,
Government Printing Office, 1988, pp. 195-969.
2) Dixon, G., "Quality Assurance Plan for the Source
Test Contracts Program", South Coast Air Quality
Management District, March 19, 1
-------
Results
Supercritical fluid extraction (SFE) reduces the time needed for the
extraction of organic pollutants from sorbent resins and air particulates to 5-30
minutes compared to several hours required for conventional liquid solvent
extractions. When supercritical fluids are used that are gases at ambient
conditions (e.g., C02 and N20), SFE essentially eliminates the generation of waste
solvents, as well as the need for sample concentration steps. The direct coupling
of SFE with capillary GC (SFE-GC) allows an entire analysis including sample
collection, extraction, analyte concentration, and gas chromatographic separation
to be completed in a total time of less than 1 hour, as demonstrated by the
analysis of roofing tar organics that were collected on polyurethane foam (PDF)
sorbent resin near the face of the tar vat operator (Figure 1). On-column SFE-GC
also yields maximum sensitivity since 100% of the collected analytes can be
transferred to the GC column for analysis.
Spike recoveries, multiple extractions, and the analysis of certified standard
reference materials have demonstrated the ability of SFE and SFE-GC to
quantitatively extract a variety of organic pollutants including fuel hydrocarbons,
polycyclic aromatic hydrocarbons (PAHs), heteroatom-containing aromatics, and
polychlorinated biphenyls (PCBs) from Tenax and PUF sorbent resins; and air
particulates released from vehicle exhaust, cigarette smoke and wood smoke1"6.
Figure 2 shows the results of two sequential SFE-GC/MS analyses of cigarette
smoke volatiles collected on PUF. As shown by the lack of significant peaks in the
second 10-minute SFE extraction, the first 10-minute SFE extraction yielded
essentially quantitative recovery of the cigarette smoke organics.
The most convincing demonstration of the ability of SFE and SFE-GC to yield
quantitative results is shown in Table I by a comparison of SFE and SFE results for
PAHs with the concentrations certified by the National Institute of Standards and
Technology (NIST). Note that while the conventional liquid solvent extractions
used by NIST required 16 to 48 hours to perform, SFE and SFE-GC extractions
required only 10 to 30 minutes per sample, yet quantitative agreement was
excellent. With coupled SFE-GC analysis, sample size can also be dramatically
reduced since all of the extracted analytes can be transferred to the
chromatographic system. For example, SFE-GC/MS analysis of the urban dust
sample required only 2 mg, compared to 1 gram required for the NIST method
using liquid solvent extraction.
Conclusions
While considerable research is needed to understand and optimize SFE and
SFE-GC methods for the rapid and quantitative extraction and analysis of organic
pollutants from sorbent resins and environmental solids, the results reported to
date1"6 clearly demonstrate the potential of SFE and coupled SFE-GC to reduce
extraction times to <30 minutes and to eliminate the production of waste
solvents. The ability of coupled SFE-GC to quantitatively transfer all of the
extracted analytes into the GC column is especially attractive when the analysis
of trace organics from small samples is desired.
84
-------
Table 1
RATING CRITERIA for
SOURCE TEST REPORT AUDIT
Good (G) : Valid source test results; adequate
documentation to support
conformance with District methods &
procedures; all data and
information are reported as
required for source test reports.
Acceptable (A) : Validity of test results is
acceptable; however, essential
documentation and/or data are
missing in the report, which are
necessary to supplement conformance
with District methods i procedures.
Unsatisfactory (U) : Test results are considered
invalid; test was not conducted
according to District methods &
procedures.
Table 2
CONTRACT SOURCE TEST and AUDITS PERFORMED
FY1987-88 FY1988-89
o SOURCE TESTS 193 185
O FIELD AUDITS 5 15
o PERFORMANCE AUDITS 18 9
O SOURCE TEST REPORT AUDITS 19 39
O PARALLEL TESTING 1 2
1038
-------
LAB CERTIFICATION VERSUS IN-HOME EMISSIONS PERFORMANCE OF ADVANCED
TECHNOLOGY WOODSTOVES
Stockton G. Barnett and Robert Roholt
OMNI Environmental Services
10950 SW 5th Street
Beaverton, OR 97005
Participate emissions trends were evaluated for thre; models of catalytic and two models of non-
catalytic woodstoves under "in-home" burning conditions during the 1988-89 heating season in the Glens
Falls, N.Y. area. The results (averaging 9.4 g/h and 9.4 g/kg) showed about a 55% reduction in emissions
compared to conventional woodstoves and demonstrate that the emissions performance of new woodstove
technologies has improved compared to that of stoves in earlif.r field studies. Emissions for the non-catalytic
stoves were about 50-55% and for the best performing catalytic stove about 80% lower than conventional
woodstoves. Two of the catalytic stove models displayed elevated emissions; in one case a significant
degradation trend developed, hi the other emissions were elevated throughout the test period, Leaky bypass
systems appear to be a major cause as well as catalyst deterioration resulting from lack of flame shielding
and inadequate air/fuel mixing.
Field emissions significantly exceed certification values as they have in past studies. Differences in
wood loading patterns and stack draft have been identified as possible causes as well as emissions
deterioration over time.
1039
-------
INTRODUCTION
During the 1970s, as a response to the Energy Crisis, there was a dramatic increase in the sale and
use of residential woodstoves. By the 1980s residential wood combustion had become a major source of
particulate pollution, especially in the many woodburning areas where wintertime temperature inversions
were common. A marked improvement in combustion efficiency was needed to reduce these air quality
problems.
With the introduction of the catalytic combustor for use in woodstoves by Corning in 1980, the
production of clean-burning woodstoves appeared feasible. Research and regulatory activity aimed at
producing low emissions appliances increased rapidly for both catalytic and non-catalytic designs.
However, because woodstove development and evaluation activities were generally conducted under
laboratory conditions, and operating conditions in homes are significantly different from those in the lab,
the effectiveness of lab-certified clean-burning technologies under real world "in-home" conditions was
unknown. Recent research efforts have been directed toward evaluating "in-home" performance.
A field study, conducted from 1979-1982 in New York and Ohio1, using both conventional and
catalytic stoves, demonstrated that clean-burning catalytic technology can operate effectively under "in-
home" conditions and its effects on emissions and efficiency can be dramatically positive. Due to the small
sample size, the question of adaptability of this technology to widespread use was not answered.
Additionally, monitoring of the catalytic stoves lasted only one season; therefore long-term woodstove
component durability was not addressed.
The results of two more recent studies, the first Northeast Cooperative Woodstove Study2
conducted in New York and Vermont from 1985-1987, and the Whitehorse Study3, conducted in
northwestern Canada in 1986-1987 suggested that while newer technologies did reduce particulate emissions,
reduction was less than 50% and emissions values were higher than certification values. Stove durability
problems were recognized, but attempts were not made to identify cause-and-effect relationships between
failure of specific components and emissions degradation.
The 1988-89 Northeast Cooperative Woodstove Study (NCWS)4, summarized herein, was designed
to evaluate the latest technologies in comparison with their certification values and evaluate the woodstove
durability issue in depth.
OBJECTIVES
The 1988-89 NCWS study attempted to 1) evaluate the effectiveness of certification values as a
predictor of the "in-home" emissions levels, 2) attempt to explain certification - field differences, if they
exist, 3) identify new woodstove technologies that can significantly reduce particulate emissions (by
approximately 80%) under "in-home" conditions, 4) document emissions degradation, should it occur and
3) determine the cause-and-effect relationships of poor emissions performance.
The "in-home" performance of advanced-technology woodstoves was evaluated for woodstoves that
met the following conditions:
1) The stoves were Oregon DEQ 1988 certified.
2) The stoves passed a laboratory stress test which screened for durability.
3) The stove installations had adequate draft, and the stoves were sized correctly for the
home.
1040
-------
4) The homeowners were experienced woodbi rners. They were given only a minimum of
specialized training about their new stoves,
METHODOLOGY
"In-home" sampling of 25 advanced-technology stoves., installed in homes in Glens Falls, N.Y. (7500
degree day winter climate), was conducted for five week-long periods from January through March, 1989.
All stoves were new and had been installed in December, 1988. AWES (Automated Woodstove Emissions
Sampling) and OMNI Data LOG'r systems5 were used to measure particulate stack emissions, thermal
efficiency, stack and catalyst temperatures, stack oxygen, stack draft, burn rate and wood use patterns.
Five stove models, three catalytic and two non-calalytic, were selected for the study from a
candidate group of ten stoves based on their performance in a laboratory stress test designed especially for
this project. Five stoves of each of the five stove models were evaluated in the field.
RESULTS
The overall average burn rate of the new technology stoves was 1.09 dry kg/h. Average wood
moisture was 27% dry basis (21% wet basis). And average draft was -0.074 in. water column and average
stack oxygen was 15.4%,
The overall average particulate emissions was 9.4 g/h (9.4 g/kg). The lowest emissions woodstove
model, the Country Flame BBF-6, a catalytic stove, had average emissions of 4.6 g/h (4.1 g/kg). The
emissions of the other four stove models appear statistically to constitute a single population, with emissions
about twice as high as the Country Flame's.
A relatively high average net thermal efficiency of 67% was attained by the Country Flame BBF-6
stoves. Efficiencies of non-catalytic stoves were about 50-55% similar to conventional stoves. These lower
efficiencies were due primarily to the lower combustion efficiency, relatively high levels of excess air and
high stack temperatures characteristic of non-catalytic stoves.
Two of the three catalytic stove models did not perform as well as had been expected. One, the
Blaze King Royal Heir, experienced a significant trend in performance degradation which began after one
month of stove operation. The other, the Oregon Woodslove, displayed generally elevated emissions
performance throughout the test period.
Examination of the stoves during and after the emissions testing period identified the failure of a
small number of identifiable components as the cause of emissions degradation, and identified areas of
potential premature component failure,
1) Blaze King Royal Heir 2200 (catalytic): Warping of some bypass support areas developed,
causing leaks around the catalyst. Probable partial catalyst failure occurred in some
stoves, induced, at least in part, by lack of fl.ime impingement shielding and high internal
catalyst temperatures.
2) Oregon Woodstove (catalytic): Bypasses generally fitted loosely, causing leaks around the
catalyst. Bypass control mechanisms also would not close the bypass consistently.
3) Regency R3/R9 (non-catalytic): Some baffles oxidized and/or warped.
4) Country Comfort CC150 (non-calalytic): Oxidation and warping of bypass support areas
developed.
1041
-------
CONCLUSIONS
The performance of new technology woodstoves has improved markedly compared to the results
reported for stoves in the earlier Northeast Cooperative Woodstove and Whitehorse studies, which were
conducted in northeastern U.S. and western Canada. The overall average paniculate emissions of 9.4 g/h
(9.4 g/kg) for the woodstoves evaluated in the current study represents a 55-60% reduction in emissions
compared to the EPA emissions factor of 21.3 g/h6 for conventional stoves, and demonstrates that current
technology is capable of significantly reducing participate emissions from woodstoves.
The lowest emissions woodstove model, the Country Flame BBF-6, a catalytic stove, with average
emissions of 4.6 g/h (4.1 g/kg), approaches the 1990 EPA certification limit of 4.0 g/h. These emissions are
about 80% lower than those of conventional woodstoves.
The emissions of the two brands of non-catalytic stoves, the Regency R3/R9 and the Country
Comfort CC150, which average 9.3 g/h (8.2 g/kg) and 11.3 g/h (11.2 g/kg) respectively, are about 50-55%
lower than those for conventional stoves, and are higher than the 7.5 g/h 1990 EPA certification limit for
these stoves.
"In-home" performance of four of the five stove brands did not agree closely with the certification
emissions values, exceeding these values by up to four times as much. On the other hand, the EPA-weighted
certification emissions values and the results for the "in-home" tests on the remaining stove brand, the
Country Flame BBF-6, were nearly identical.
The field performance of one of the catalytic stove models degraded with time magnifying the
discrepancy with certification values. Signs of physical degradation in three of the other models suggest that
emissions deterioration may develop with them in subsequent years.
It is concluded that certification emissions values probably continue to understate field performance.
There are two identified reasons: 1) certification test conditions are different from those in the field
(especially wood loading geometry and stack draft) and 2) the emissions performance of an as yet
undetermined number of stove models degrade over tune.
Because certain conditions in the field differ from those in the lab, the common practice of using
certification testing as a stove design development tool appears to be having a negative impact on stove
design. For example, "in-home" burning conditions produce stack gases that are about twice as dilute as
those produced by the Douglas fir cribs used hi certification tests, and "in-home" drafts are almost twice
as high. To perform satisfactorily under certification testing conditions, manufacturers therefore tend to
design stoves with more secondary air than is needed in most home burning situations. This excess air
creates higher than necessary catalyst temperatures and may shorten the life of catalysts. This situation also
encourages the use of smaller diameter catalyst eel! sizes. In homes these cells are more prone to clogging
and can develop locally higher catalyst temperatures than larger cells do.
ACKNOWLEDGEMENTS
This research was sponsored by the U.S. Environmental Protection Agency, the New York State
Energy Research and Development Authority, the Coalition of Northeast Governors Policy Research Center,
the Canadian Combustion Research Laboratory and the Wood Heating Alliance. An advisory committee
of sponsors and industry representatives guided the research.
1042
-------
REFERENCES
1. S.G. Barnett, "The Effects of Stove Design and Control Mode on Condensable Participate
Emissions, Flue Pipe Creosote Accumulation and the Efficiency of Woodstoves in Homes," in
Energy from Biomass and Wastes Symposium, Institute of Gas Technology, Chicago, 1982, pp 283-
318.
2. OMNI Environmental Services, Inc., Performance Monitoring of Catalyst Stoves. Add-Qns, and
High Efficiency Stoves, Field Testing for Fuel Savings, Creosote Build-Up and Emissions. Prepared
for the Coalition of Northeastern Governors, New York State Energy Research and Development
Authority and the U.S. Environmental Protection Agency, 1987.
3. C.A. Simons, P.D. Christiansen, L.C. Pritchett, G.A. Beyerman, Whitehorse Efficient Woodheat
Demonstration. Prepared for the City of Whitehors; and Energy, Mines and Resources Canada,
1987.
4. S.G. Barnett, Field Performance of Advanced Technology Woodstoves in Glens Falls, N.Y..
Prepared for the Coalition of Northeastern Governors, New York State Energy Research and
Development Authority and the U.S. Environmental Protection Agency, 1990.
5. J.E. Houck, C,A. Simons, P.G. Burnet, "A System to Obtain Time Integrated Woodstove Emission
Samples," in Proceedings of the 1986 EPA/APCA Symposium on Measurement of Toxic Air
Pollutants. VIP-7, Air Pollution Control Association. 1986, pp 724-735.
6. Office of Air Quality Standards, USEPA. In-situ emission factors for residential wood combustion
units. EPA-450-3-88-013, U.S. Environmenta! Protection Agency, Research Triangle Park, 1987.
1043
-------
ANALYSIS OF AIR POLLUTANT CONCENTRATIONS
BELOW THE DETECTION LIMIT
S. Trivikrama Rao and Jia-Yeong Ku
Division of Air Resources
NY State Department of Environmental Conservation
Albany, New York
and
K. Shankar Rao
Atmospheric Turbulence and Diffusion Division
Air Resources Laboratory, NOAA
Oak Ridge, Tennessee
Air quality data often contain several observations reported only as below the
analytical limit of detection (LOD), resulting in censored data sets. Such censored
and/or truncated data sets tend to complicate statistical analysis. We discuss several
procedures for estimating the mean concentration and its 95% confidence interval for
air contaminant data that contain values below the LOD. A quantitative approach
for assessing the uncertainty inherent in the estimated mean concentration due to the
presence of values below the LOD in the data set, as well as the natural variability of
atmospheric concentration data, is described. The utility of this approach in the analysis
and interpretation of air pollutant concentration data is demonstrated for a singly-
censored hypothetical data set drawn from a normally distributed population, and for
a multiply-censored, multi-pollutant observed concentration data set.
The methodologies discussed here should be particularly useful in estimating the
risks associated with long-term exposure of populations to toxic air contaminants, and in
assessing the uncertainty associated with these estimates. This information is valuable
to policymakers for making informed decisions regarding the environmental risk.
1044
-------
Introduction
Recent advances in ambient air quality monitoring technology are enabling us
to measure very low levels of toxic air contaminant concentrations. However, because
of inherent limitations of the chemical/analytical methodologies, air quality data sets
often contain several observations reported us below the analytical limit of detection
(LOD). The presence of values below the LCD results in censored data sets, and
complicates all related statistical analyses.
Many of the methods in the statistical literature1 for analyzing censored
data are sensitive to assumptions about the underlying distribution. For a left-
censored concentration data set, if LOD values are used in place of actual "net" (as if
measured) concentrations, the mean H; will be biased high (i.e., on the average af will
be larger than /j. , the true mean) and the variance s2 will be biased low (i.e., s will
be smaller than a , the true standard deviation). Biases will also arise if the values
below the LOD are ignored, or if they are replaced by "zeros" before computing the
mean and the variance. Another method suggests replacing concentrations below the
LOD by the mid-point between zero and the LOLi value. In principle, we must regard
each of these approaches as unacceptable for one or more reasons if our objective is to
obtain unbiased estimates of fj, and a.
The long-term (e.g., the annual mean) concentration of a toxic contaminant
must be determined to quantify the risk associated with chronic human exposure.
Often, samples of certain pollutants collected over a period of time may be censored
at different levels as changes in analytical technology alter the LOD of the selected
method. Among examples of this are the ambient measurements for the total 2,3,7.8-
TCDD (tetrachlorodibenzo-p-dioxin, commonly referred to as '"dioxin") concentration.
Analysis of such multiply-censored, multi-pollutant data is more complex than that of
singly-censored data for a single pollutant.
In this paper, we examine selected techniques for the analysis of air quality
data which contain values below the LOD. Several methodologies for quantifying the
uncertainty associated with the presence of values below the LOD in the data set, and
the naiural variability of atmospheric concentration data, are examined. The utility
of these methods is examined by applying them to a hypothetical data set drawn
from a normal distribution, and to an observed data set of multiply-censored, multi-
pollutant (dioxin) concentrations.
Quantification of Uncertainty
LIncertainties associated with ambient air quality data (assuming the mea-
surements to be error-free) stem from the fact that (a) the "net" concentrations
are always unknown when the data are reported as below the LOD, and (b) any
given ambient measurement reflects a single event of a population. Therefore, a 95%
confidence interval which describes the limitations of the methodology chosen for
treating the data below the LOD, as well as the natural variability in air pollutant
concentrations, needs to be estimated for communicating the inherent uncertainty in
the estimated mean concentration to a decision-maker.
We utilized the following approaches to "fill-in" for the data below the LOD in
estimating the statistical parameter of interest, ar.d the uncertainty associated with
this estimate:
1045
-------
(i) The reported value below the LOD is replaced by a value chosen randomly from
a uniform distribution which assumes that any value in the range from 0 to
the reported LOD value is equally likely to occur. The region between 0 and
the LOD is divided into 10 equal segments, and each segment's mid-point is
taken as the concentration value representing that segment. A segment number
is randomly selected and the concentration associated with that segment is
taken as the net concentration replacing the below-LOD value. This procedure
is repeated for all concentrations reported as below the LOD, and the mean
concentration and the standard deviation for the total sample are computed.
The above procedure is repeated several (say 100) times, for assessing the
variability in the estimated mean concentration of the pollutant. Using the
results of 100 Monte Carlo simulations, cumulative distributions for the
computed means and the standard deviations are developed, and their medians
are then determined. These median values may be considered to represent robust
estimates of the statistical parameters of interest for each pollutant.
Once the data set for the year for each pollutant (isomer) is assembled in this
manner, the annual mean concentration of a multi-pollutant (e.g., dioxin), ~x , is
then computed using the weighted sum of averages as:
x = ^(wi xt) (1)
k
where ~x is the weighted sum of k different isomers, the Wi are the weights, and
the x; are the means of samples of size n, for each isomer. The problem of the
distribution of a weighted sum was discussed by Satterthwaite2 , and Gaylor and
Hopper3. Assuming independent samples, the 95% confidence interval (CI) for
the mean can be determined from4:
CI = x ± i(a/2, n cr(c) (2)
<72(c) = (w? <72/nt) (3)
where
-------
a uniform distribution), and combining them with the detected concentrations to
reconstruct the original sample. This reconstructed sample is then subjected to
100 bootstrap replications (random sampling with replacement), and the median
of the 100 means and the median of the 100 standard deviations are determined,
along with the 95% confidence bounds for the mean concentration based on the
percentile method.
Results and Discussion
The limitations associated with the treatment of the values below LOD are
assessed by invoking either the uniform or normal distribution approach to fill-in
the region below the LOD, followed by Monte Carle simulation. These procedures
are applied to a hypothetical data set with a sample size of 60 values which is drawn
randomly from a normal population N(//,
-------
Credit
The support of the U.S. EPA, Office of Exploratory Research is gratefully
acknowledged. Instrument loans from Suprex and ISCO are also appreciated.
References
1. Hawthorne, S.B.; Miller, D.J. J. Chromatogr. Sci. 1986, 24. 258.
2. Hawthorne, S.B.; Miller, D.J.; Krieger, M.S. J. Chromatoor. Sci. 1989, 27.
347.
3. Hawthorne, S.B.; Miller, D.J. Anal. Chem. 1987, 59. 1705.
4. Hawthorne, S.B.; Miller, D.J. J. Chromatoar. 1987, 403. 63.
5, Hawthorne, S.B.; Krieger, M.S.; Miller, D.J. Anal. Chem. 1989, 61. 736.
6. Hawthorne, S.B.; Miller, D.J.; Langenfeld, J.J. J. Chromatogr. Sci. 1990,
28_, 2.
Table I
Concentrations of PAHs in NIST Standard Reference Materials
Determined Using SFE and SFE-GC
Concentration: certified va!ue/SFE value (uq/g)a
fluoranthene
benz[a]anthracene
benzo[a]pyrene
indeno[1,2,3-cd]-
pyrene
marine sed.
(SRM 1941)
1.22/1.45
0.55/0.60
not reported
0.57/0.56
urban dust
(SRM 1649)
7.1/7.3
2.6/2.6
2.9/2.8
3.3/3.0
diesel part.
(SRM 1650)
51/53
48/47
1.2/1.4
not reported
The certified value is given first followed by the concentration determined using
SFE (for SRM 1650) or SFE-GC/MS (for SRMs 1941 and 1649). SFE extractions
for SRMs 1941, 1649, and 1650 were (respectively) 10 minutes with N20, 15
minutes with N20, and 30 minutes with C02. See references 1,2,4, and 6 for
details.
85
-------
bounds estimated from the Monte Carlo uniform fill-in approach compare favorably
to these estimates from the more computer-intensive approach.
Conclusions
We presented several methodologies to quantify the inherent uncertainty due to
the presence of values below the LOD in the data set and the natural variability of
the atmospheric concentration data. The analysis methods discussed here are able
to estimate the mean concentration of the pollutant reasonably well, but the results
indicate that the 95% confidence limits for the mean are sensitive to the assumptions
invoked regarding the nature of the underlying data. Although the bootstrap method,
as applied here, does not require the determination of the true distribution of the
data, the results may be sensitive to the procedure used for reconstructing the
original sample.
The upper 95% confidence bound for the annual mean may be more appropriate
in exposure assessment and risk analysis, because this higher limit allows us to make
conservative estimates of the risk associated with the human exposure to toxic air
contaminants. If the distribution of the original data is not known a priori, one
may have no choice except to assume that all values between zero and the reported
LOD are equally likely to occur, or to apply distribution-free techniques. Under
these conditions, the results from this study suggest that the assumption of uniform
distribution to fill-in for the values below the LOD, followed by the Monte Carlo
simulation procedure, is a reasonable methodology for estimating the upper 95%
confidence bound for the mean concentration.
Acknowledgements
The authors would like to express their gratitude to Drs. Cynthia Hirtzel,
Ray Hosker, Steven Porter, Gopal Sistla, and Ram Uppuluri for their many helpful
suggestions and comments on this work. The encouragement of Thomas Allen and
Bruce Hicks during this study is gratefully acknowledged.
References
1. R. 0. Gilbert, Statistical Methods for Environmental Pollution Monitoring, Van
Nostrand Reinhold Company, New York, 1987.
2. F. E. Satterthwaite, "An approximate distribution of estimates of variance
components," Biometrics Bulletin 2: 110 (1946).
3. D. W. Gaylor, F. N. Hopper, "Estimating the degree of freedom for linear
combinations of mean squares by Satterthwaite formula," Technometrics 11: 691
(1969).
4. P. S. Porter, personal communication, 1989.
5. B. Efron, The Jackknife, the Bootstrap, and Other Resampling Plans, CBMMS-
NSF-38, SIAM, Philadelphia, PA, 1982.
6. S. T. Rao, G. Sistla, V. Pagnotti, W. B. Petersen, J. S. Irwin, D. B. Turner,
"Resampling and extreme value statistics in air quality model performance
evaluation," Atmos, Environ. 19: 1503 (1985).
7. S. T. Rao, J. Y. Ku, K. S. Rao, "Analysis of toxic air contaminant data
conatining concentrations below the limit of detection," Manuscript submitted
to J. A & WMA (1990).
1048
-------
TABLE I. Comparison of the medians of the means and standard deviations, and
the 95% confidence intervals for the means, for different "fill-in" methods for the
hypothetical data set of sample size 60 drawn from N(20,5).
Scenario
Original
Censoring
-------
USE OF THE SURFACE ISOLATION FLUX CHAMBER TO ASSESS FUGITIVE
EMISSIONS FROM A FIXED-ROOF ON AN OIL-WATER SEPARATOR FACILITY
Dr. C.E. Schmidt
Independent Consultant
1479 Salmon Falls Road
Folsom, CA 95630
John Clark
Radian Corporation
10395 Old Placerville Road
Sacramento, CA 95827
The release of fugitive volatile organic compound (VOC)
emissions from an oil-water separator facility at a refinery was
studied. The first phase of the testing was a screening of all
sources of hydrocarbon emissions on the fixed roof of the
separator. These sources included seams in the cement roof and
covered observation ports. All seams were tested and
categorized into a characteristic range of VOC source. It was
then possible to express the fugitive VOC emissions from the
seams in terms of number of lineal feet of roof seam per class
of seam (e.g., 125 feet of seam, range 0 to 10 ppmv VOC etc.)
Likewise, all ports were surveyed and categorized in a similar
fashion.
After all sources of VOCs were surveyed and categorized,
representative sources from each category (seams and observation
ports) were studied using an emission isolation flux chamber to
determine the emission rate of VOCs from the facility. This
approach was cost-effective since a minimum number of emission
measurements were performed. In this way, a correlation between
the onsite TNMHC and hydrocarbon speciation data was made.
Total TNMHC emissions were calculated by summing the
emissions from all categories of seams and all categories of
ports and expressing the fugitive emissions as TNMHC emissions
from the roof on the facility.
1050
-------
Introduction
Radian Corporation performed testing of fugitive total
non-methane hydrocarbon emissions from a fixed-roof oil/water
separator designed by the American Petroleum Institute (API) at
a refinery in California. The refinery was required to
demonstrate that the control measures placed on the separator
were effective at controlling fugitive emissions to the
acceptance of the local air quality management district agency.
The EPA recommended surface emissions isolation flux chamber was
used to measure the emissions from the API separator. Although
not specifically intended for tnis purpose, the direct
measurement approach using the flux chamber provided to be an
effective measurement approach for this fugitive source. These
data were then used to demonstrate effective control of VOC
emissions by the fixed-roof. As such, this paper will focus on
the application of this emission assessment technology to this
unique VOC emission source.
Facility Description
The fixed-roof separator involved in the program was an
experimental model that was constructed and tested to satisfy a
local air district control requirement. The design of the
separator was unique in that the fixed-roof construction was
intended to contain all fugitive emissions from the oil waste.
The features of concern for the fugitive emissions testing
program were the seams in the roof and the observation ports.
The separator consisted of two cells each having an
outside dimension of about 22 feet by 174 feet. The roof was
constructed of metal I-beams and concrete slabs. The unique
feature of the roof design was a compression gasket placed
between the metal beams and the concrete slabs. Each cell of
the separator had a total of thirty, 4 feet by 21 feet slabs
that covered the main area of each cell. These slabs were
fastened to the metal beam structure but the slabs were not
united in any way. They were, however, sealed with a caulk
material. In fact, the cell that was tested had four types of
caulk material that was being evaluated for effectiveness (hard
rubber, soft rubber, metal tape, and cement).
In addition to the slab seams, there were a total of six
observation ports having a dimension of 1 foot by 3 foot. Each
port had a lid and a rubber gasket to seal the lid to the roof
port. These lids were not air-tight and later proved to be a
primary source of fugitive emissions, as did those seams that
had weathered or incomplete caulk. There was no active control
technology on these units and the waste streams were not treated
in any way upstream of the separators. The concept of this
design was to seal the separator with a fixed-roof and prevent
fugitive emissions by using a "lecik-tight" roof. No attempt was
made to limit the head space air volume and there was no concern
of explosion hazard.
1051
-------
Summary of the VOC Testing Program
There are several ways in which to assess the fugitive
emissions from a facility such as the separator. One class of
technologies that could be used are indirect emission
measurement technologies. These technologies involve the
collection of ambient air samples downwind of the facility (or
source) and rely on dispersion modeling to estimate a source
term that could have generated those measured downwind
concentrations. Normally, indirect assessment technologies are
preferred for situations like this one where the emissions are
from many fugitive sources that collectively act as a
heterogeneous area source. However, the disadvantages of the
indirect approach outweighs the advantages. Indirect
technologies are subject to upwind interferences which can be
significant at a refinery, especially when the same or similar
compounds can be found upwind as well as downwind of the source
being tested. Another significant disadvantage is that the
downwind concentrations will be low if the source has a low
emission rate, which is the case here. Lastly, indirect
technologies are governed by the dispersion of air contaminants
at the facility. The efficiency of collecting data and the
quality of the data are controlled by parameters that the
scientists has no control over. For these reasons, the testing
approach selected was from the class of technologies known as
direct emission assessment technologies.
The testing approach for assessing the VOC emissions
from the separator employed field screening and direct emission
measurements using the Surface Isolation Emission Flux Chamber.1
This direct emissions measurement technology was developed for
EPA by Radian Corporation and is now a recommended technology
for measuring emission rates from fugitive area sources. The
primary advantage of using this technology is that the emission
rate is calculated from measured parameters, all of which the
scientist has control over. The technology is free from upwind
interferences, and is applicable to a variety of surfaces
including emissions from cracks, vents, and fugitive emissions
from observation ports. The disadvantage is that the entire
source must be assessed and that means that the source must be
understood and then representatively tested.
The emissions testing program consisted of screening to
select representative sampling locations and emission
measurements that could be summed in order to calculate an
emission rate of VOC's from the facility. Preliminary screening
indicated that the majority of fugitive emissions were from the
caulked seams in the cement slab construction and from the
observation ports throughout the roof. The screening activities
using real-time analyzers indicated that there were 20
categories of seams that required testing: 4 types of caulk
material; and 5 ranges of VOC concentration above the caulked
seams. The seam length per each type of seam caulk was measured
and representative emission rate measurements were made using
the isolation emission flux chamber where needed to satisfy the
screening data needs. Not all types of seam caulk had VOC
concentration measurements in each of the five concentration
ranges identified and flux chamber measurements were not
1052
-------
required for all 20 categories.
The results of the seam testing by category were used to
calculate the total non-methane hydrocarbon (TNMHC) fugitive
emissions by multiplying the unit emission rate data per
category of seam by the lineal feet of each category of seam and
summing these data to obtain the TNMHC emissions for caulked
seams on the separator cell (Table L) . It was estimated that
the TNMHC fugitive emissions from the seams was 0.56 pounds per
day calibrated as hexane. In addition, testing was conducted on
ports with the lids open and closed, and with lids with and
without over gaskets of rubber. It was found that a metal lid
controls emissions about 44 percent and 80 percent with an
additional gasket.
The same approach was used for the observation ports;
screening with real-time analyzers and the flux chamber
measurements over the covered ports to estimate the fugitive
emissions from this source (Table 2). In the same way, an
estimate of TNMHC fugitive emissions of 0.34 pounds per day
calibrated as hexane was made for all. six observation ports per
cell.
Considering that the total fugitive emissions for TNMHC
were believed to come from the caulked seams and the observation
ports, the total emissions estimate for the separator cell was
0.90 pounds per day calibrated as hexane.
The effectiveness of the fixed-roof control measure was
assessed by summing the fugitive emission sources and dividing
by the estimated, uncontrolled emission rate estimate derived by
the Litchfield equation. The control efficiency of the fixed-
roof was estimated to be 98 percent.
Summary and Conclusions
The primary conclusions from this testing and research
effort are as follows:
o The fugitive TNMHC emissions from a fixed-roof API
oil/water separator on the day of testing were
measured and found to be 0.90 pounds per day as
hexane.
o The overall control efficiency of the fixed-roof on
the separator without waste pre-treatment or an
additional control technology was estimated at 98
percent.
o Additional controls consisting of over gaskets on
ports and flexible continuous caulking in seams can
improve the control efficiency of fugitive emissions
from fixed-roofs on separators.
o If used with a screening technology, the direct
measurement technologies can be effective in
assessing fugitive emissions from these types of
fugitive emission sources.
1053
-------
TABLE 1. RESULTS OF DIRECT EMISSIONS TESTING OF SEAMS
SEAM EMISSIONS: EMISSION RATE PER CATEGORY OF OVA CONCENTRATION (ppmv) as Lbs hexane/foot-day
TYPE < 10 ppmv OVA 10 - 50 ppmv OVA SO - 200 ppmv OVA 200 - 1,000 ppmv OVA >1,000 PPMV OVA
HARD RUBBER
SOFT RUBBER
590' 3 1.6E
-5
23' a 9.0E
-4
18- a 6.9E
236' a 1.0E
-5
3' a 9.0E
-4
1.3 a 1.9E
-2
5- a 5.5E
NA
-2
h- NA - NOT APPLICABLE
O
TABLE 2. RESULTS OF DIRECT EMISSIONS TESTING OF PORTS
NA
NA
METAL TAPE
CEMENT
TOTAL EMISSIONS FROM SEAMS:
233' a 6.5E ' 5' a 9.0E H
62' a 1.4E"5 NA
0.56 Ibs/day as hexarve.
3' 3 1.5E 1.3 a 5.5E ' NA
NA NA NA
PORT EMISSIONS: EMISSION RATE PER CATEGORY OF OVA CONCENTRATION (ppmv) as Ibs. hexane/port-day
TYPE
NUMBER
OVA SCREEN (ppmv)
1,000 - 2,000
6,000 - 7,000
NOT TESTED
EMISSION RATE (Ibs. hexane/port-day)
r-2
INLET PORT
OUTLET PORT
PUMP PORT
6.8E
4.9E
-2
4.9E
-3
TOTAL EMISSIONS FROM PORTS: 0.34 Ibs./day as hexan*
-------
AN ATTEMPT TO MEASURE THE AIR TOXICS IMPACTS OF THE GREATER
DETROIT RESOURCE RECOVERY FACILITY
By: JAMES C. SERNE, P.E. AND JOSEPH M. MARTINI
INTRODUCTION
An intensive ambient air monitoring program was initiated in
January 1988 in the vicinity of the Greater Detroit Resource
Recovery Authority's (GDRRA) resource recovery facility located
in Detroit, Michigan. In July 1989 the facility finished
initial compliance testing and began commercial operation. The
facility is designed to receive and process a maximum of 4000
tons/day of municipal solid waste (MSW). Typically, the
processing equipment will shred, screen and magnetically
separate 3300 tons of solid waste, to produce refuse derived fuel
(RDF) five days a week. Each day, seven days a week, up to 2400
tons of RDF can be used as boiler fuel for production of steam
and electricity.
The ambient air monitoring program being conducted by Roy F.
Weston, Inc. (WESTON) for GDRRA began in January 1988. The
objectives of the air monitoring program are to:
• Document the background levels of the pollutants of
concern that will be emitted from the waste-to-energy
facility.
• Gather sufficient data to determine the actual ambient
impact of the GDRRA facility emissions.
WESTON established and operates ambient air monitoring equipment
at two sites in the vicinity of the GDRRA facility. One of the
monitoring stations, the "Playground Site", is located near the
predicted location of the maximum annual ground-level concentra-
tion and annual maximum dry pollutant deposition. The second
site used by WESTON, the "Wayne County Site", was selected as
representative of the regional background air quality beyond any
significant influence of the resource recovery facility. In
addition to the two sites operated by WESTON, Environment Canada
operates two monitoring stations; in the Windsor, Ontario area
that provide data of potential use in WESTON's air quality
impact analysis.
The pollutants of concern include trace metals, organic com-
pounds (which could result from incomplete combustion of the
MSW) , acid gases and inhalable particulate matter. The organic
compounds being sampled and analyzed include dioxins, furans,
PCB's, polynuclear aromatic hydrocarbons, chlorobenzenes and
chlorophenols. The metals being analyzed include lead, copper,
zinc, mercury, nickel, manganese, arsenic, chromium, vanadium,
selenium, cadmium and antimony.
The first phase of the ambient air monitoring program was
designed to gather background air qua.lity data for at least one
year prior to the start-up of the GDRRA facility. The monitor-
ing program will continue for at least one year after the GDRRA
facility begins operation to determine if measurable air quality
impacts occur.
1055
-------
The design of the ambient air toxics monitoring program was
coordinated with an ad-hoc interagency committee of the federal,
state, provincial and local air pollution control agencies in
the Detroit-Windsor area.
SAMPLING SITES
Ambient air quality data is collected by WESTON for the GDRRA at
two monitoring stations in Detroit. The "Playground Site" is
located at the Grandy Street/Medbury Street Playground, which is
predicted by air quality modeling to be in the area of the maxi-
mum air quality impacts from the GDRRA facility. The Play-
ground Site is located approximately 0.65 miles northeast of the
GDRRA facility.
The second monitoring station operated by WESTON is called the
"Wayne County Site". This site, which is located near the
Southeastern High School on Goethe Street, was selected as
representative of the regional background level concentrations
for the pollutants of concern. The Wayne County monitoring
station is located approximately 3.8 miles east of the GDRRA
facility. The Wayne County Site is being considered a back-
ground site in the sense that it is anticipated to be much less
influenced by the emissions from the GDRRA facility relative to
the Playground Site.
In addition to the two monitoring stations operated by WESTON
for GDRRA, Environment Canada operates two air quality monitor-
ing stations in areas potentially impacted by the GDRRA facili-
ty, but to a much lesser degree than the Detroit Playground
Site. The two Environment Canada sites are referred to as the
Windsor Site and the Walpole Island Site. The Windsor Site is
located approximately four miles south of the GDRRA facility in
Windsor, Ontario. Environment Canada began sampling at the
Windsor Site in July 1987. In January 1988, they began sampling
at the Walpole Island site which is located approximately 35
miles to the northeast of the GDRRA facility. The Walpole
Island Site is considered representative of a remote background
site.
SUMMARY OF SAMPLING AND ANALYTICAL METHODS
The pollutants of concern and the sampling methods and analyti-
cal procedures are described below. This is the first time an
attempt has been made to identify and quantify most of these
pollutants in Detroit.
Pollutants of Concern
While there are many pollutants of concern in ambient air, this
study focuses on those pollutants that are specifically thought
to be associated with the combustion of municipal solid waste.
Table 1 provides a list of pollutants that are included in the
monitoring study. Table 2 provides a list of the semi-volatile
organics analyzed for in the PAH and PCB, chlorobenzene,
chlorophenol samples. Many of these pollutants may be emitted
from resource recovery facilities. They have been classified as
pollutants of concern in previous health risk studies. The
pollutants of concern are compounds or elements, which poten-
tially contribute to long-term, chronic health effects. While
1056
-------
some of these compounds also have short-term, acute health
effects, there are no conceivable plant operating conditions
that could lead to an exceedance of accepted short-term stan-
dards or criteria nor any long-term or chronic exposure crite-
ria. Although none of the many studies that have been conducted
concerning waste-to-energy facilities have identified any
potential health impacts, this study is being conducted to
confirm such results.
Sampling and Analytical Methods
The sampling and analytical methoc.s employed by WESTON are
identified in Table 3. The selected sampling and analytical
methods are published EPA and NIOSH inethods. In general, these
are the same methods used by Environment Canada in their air
quality monitoring program at the Windsor and Walpole Island
Sites. This was an important point in method selection. The
use of the same or very similar sampling and analytical methods
will enable the monitoring results ::rom both the Michigan and
Ontario sides of the Detroit River to be merged into a single
database to determine the relative air quality impact of the
GDRRA facility.
The detection limits for the pollutants of concern are also
noted in Table 3. The detection limits for dioxins and furans
are in the 0.05 to 0.5 picograms per cubic meter range (1CT12
grams/m3) depending upon the species of dioxin or furan being
measured. The detection limit for POB's and chlorobenzenes and
chlorophenols are approximately 2 to 20 pg/m3. The detection
limits for the other organics and trace metals are in the nano-
grams per cubic meter range (10~9 gra:i\s/m3) .
Sampling Schedule and Period
PAH, metals and acids samples are collected at both WESTON sites
every six days. PCB's and chloroben2enes and chlorophenols are
sampled every twelve days. Dioxins and furans are sampled every
24 days. To collect a sufficient a:_r sample volume to obtain
the desired detection limits for dioxin/furans, PAH's, PCB's,
CLB and CLP, WESTON samplers are operated for 48 hours. The
samples for the other pollutants of concern are collected over
24 hour periods by WESTON. The 43-hour samples are collected
from noon to noon.
Methods For Identifying Ambient Air Quality Impacts
Several methods are planned to identify and characterize the air
toxics impact using the pre- and pos:- operation ambient data.
A preliminary evaluation has been performed each quarter since
the beginning of the sampling activities. Each calendar quarter
the air sampling, meteorological arid boiler operations data are
tabulated and entered into a database. Quarterly data reports
are prepared which present tabulated and graphical (bar chart)
summaries of the mean values, standard deviations, maximum
values and number of values above the detection limits. The
quarterly summaries have been compared visually or "by-eye"
against previous quarterly summaries to flag trends or obvious
impacts. To date no significant air toxics impacts have been
1057
-------
p
o
I
1st Extraction
2nd Extraction
n-A!kanes
m/z-57
C16
C17
C15
C14
C13
C18
C19
XJbiJll-Ji , .1
o Branched Alkyl Benzenes
131
162
190
Dibenzothiophenes
184 212
190
226
Retention Time (min)
^ 10 ' ' ' ' 15 ' ' ' ' 20
Retention Time (mini
Figure 1. SFE-GC/MS analysis of roofing tar volatiles collected near the tar vat
operator's face. The entire analysis including sample collection (10 minutes),
supercritical fluid extraction (10 minutes), and GC/MS analysis (except data
reduction) required < 1 hour. The numbers above the selected ion plots indicate
the mass of the selected ion plot. See reference 5 for details.
86
-------
observed and the facility does not appear to be increasing
ambient air toxic concentrations above existing pre-operational
levels.
Only preliminary analysis and interpretation of the data have
been accomplished to date and the collection of post operation
ambient data is ongoing. During the first half of 1989 the
facility was undergoing start-up activities and hence the
combustion of RDF was intermittent and quite variable. The
summer season (July through September) of 1989 was the first
quarter in which the facility operated in a routine manner,
although RDF combustion at the design rate was seldom achieved.
Since completing start-up the facility has typically operated
with only one boiler and only about 1000 TPD of RDF has been
combusted.
The summer season of 1989 has been compared to the summer of
1988. A series of barcharts provides a side-by-side comparison
of the seasonal data from the two years. These figures are
included as an appendix to this paper. This simplistic type of
comparison identifies no air toxics impact from the facility.
It appears that the emissions from the facility have little or
no measurable ambient air toxics impact. For most of the
pollutants included in the study, the ambient concentrations
actually decreased in 1989. Meteorology must be considered and
a larger database with more seasons of sampling are needed
before any conclusions can be made.
Other methods will be used to investigate potential impacts.
Dispersion modeling will be performed to study specific sampling
days to compare the predicted air quality impacts from the air
toxic emissions from the facility to the actual measured ambient
concentrations. Meteorological data are continuously gathered
at the Playground monitoring station and records are kept of the
rate of RDF combustion in each boiler and per day. Emission
test results are also available from the initial compliance test
program to relate pollutant emissions to the quantity of RDF
combusted. Days with prevailing winds that would carry the
plume towards one of the monitoring stations can be studied and
comparisons of predicted air toxic impacts to measured ambient
concentrations can be performed.
When approximately one year of post-operation data are avail-
able, a comprehensive statistical analysis of the database is
planned. Multi-variate analysis of variance will be used to
determine if measurable air toxic impacts have occurred. Each
pollutant emitted from the facility can be evaluated and these
results for different pollutants compared. The large number of
variables, in particular, meteorological factors (such as wind
speed, direction, atmospheric stability, etc.) which may vary
significantly during most sampling days and from one sample day
to the next are expected to make this statistical analysis quite
difficult. Further complicating this analysis is the vari-
ability in pollutant emission rates which vary from day to day
(and hour to hour) depending on boiler operation and RDF
combustion rates .
1058
-------
TABLE 1
LIST OP POLLUTANTS THAT ARE INCLUDED
IN THE ODRRA MONITORING STUDY
TABLE 2
O
Oi
CD
EMISSION
Pioxjns and furana
Polychlorinated
dibanto-p-dloxina and
Polychlorinated
dibenzofuran*
Seal-Volatile Organic*.
Polynuclear Aromatic
Hydrocarbon*
Polychlorinated Biphenyla
Chlorobenzene* and
Chlorophenol*
Trace Metal*
Antiaony
Araenic
Cadniua
Chroaiua
Copper
Lead
Mercury
Selenivn
Vanadium
Zinc
Inhalable Partlculate Matt
Inhalable Particulate*
Acid Gaae*
Hydrogen Chloride
and Inorganic Acid*
a Generally recognized
facilities
b Federally-regulated:
c Federally-regulated:
d Ej»i»aion of concern
SYMBOL REASON FOR SAMPLING
PCDD,PCDF
a, d
PAH a, d
PCB a, d
C1B, C1P
Sb d
A* d
Cd d
Cr d
Cu d
Pb b, a, d
Mn
Hg c, d
Ni d
8* d
V d
Zn d
tr
PM10 a, b, d
HC1 a
pollutant emission from resource recovery
"criteria pollutant."
"PSD pollutant".
fro» a standpoint of high toxicity, and/ or
SEMI-VOLATILE
POLYNUCLEAR AROMATIC HYDRO-
Acenaphthylene
Acenaphthene
Fluorene
Phenanthrene
Anthracene
Fluoranthene
Pyrene
Chryaene
POLYCHLORINATED BIPHENYLS
10 I«omer Croup*
CHLOROBEHZEVES (CLB)
Tr i ch 1 orobe n z en«
Tetrachlorobenzene
Pentachlorobenz en*
Hexachlorobenzene
CRLOROPHZNOLS (CLP)
Dichlorobenzene
Trichlorophenol
Tetrachlorphenol
Panta Chlorophenol
ORGANIC MEASUREMENTS
CARBONS (PAH*)
Benzo ( a ) anthracene
Benzo (b, k) f luoranthene
Benzo ( a ) pyren*
Indeno(l,2, 3-cd) pyrene
Dlbeni (a , h) anthracene
Benzo ( g , h , 1 ) pery len*
(PCB«)
high inhalable particle concentration.
-------
TABLE 3
AMBIENT HOHITORINS METHODS AND DETECTION LIMITS
o
o
COMPOUND
DioxirWFuran
PCS
Chlorobenzenes
Ctilorophenols
PAH
Particulat.es (PIHO)
Metals
Inorqanic fields
VOC
SAMPLING
HE T HOD
Hi-Val/PUF
Hi-Vol/PUF
Hi-Vol/PUF
Hi-'v'ol/PUF
Hi-Vol/PUF-KAD
Ned-Flo*
Med-Flow-TFE
NIQSH 7903
EPA TO-14
ANftLfTICAL
METHOD
GC-MS (82801
GC-MS
GC-MS
6C-HS
GC-MS-SIM
Gravimetric
XRF
1C Scan
c-c-ns
SAMPLE
VOLUME
800 ffl3
a oo »3
800 «3
800 iJ
700 «3
160 a3
160 §3
0.5 §3
16 L
TftPGET DETECTION
LEVEL
0.05 - 0.5 pg/«3
2-20 pq/«3
10 pq/s3
10 pq/*3
0.02 - 0.2 ng/s3
—
0.09 - 9 nq/«3
2-10 uq/«3
0.1 ug/u-3
SfiMPLlNB EQUIPMENT
General fSetd Works PS-1
Beneral Metal Works PS-1
General Metal Works PS-i
General Metal Works PS-1
General Metal Works PS-1
Sierra Anderson Medium Flon
Sierra Anderson Mediun Flow
Silica gel tube
Evacuated Canister
-------
EMISSION OF OZONE AND DUST
FROM LASERPRINTERS.
PRESENTATION OF A NEW
EMISSION SOURCE TEST METHOD.
Torben Eggert, Arne Grove
and Iver Drabaek.
Danish Technological Institute,
Department of Environmental
Technology,
DK-2630 Taastrup, Denmark.
Tel. +45 42 99 66 11.
Abstract
A new method has been developed to measure emissions from
office equipment to provide figures for the contribution
to local indoor air pollution from different sources e.g.
laser printers.
The tests are conducted in a climatized room. The total
air emissions from the equipment, e.g. a printer is col-
lected in a funnel, the equipment is tested for leaks,
the total air flow is measured, the concentration of
ozone and particles in the outlet air is determined, and
the results are given as micrograms pr. minute of ozone
and dust. As the results are given as a single number for
each tested item the method also gives the possibility
for ranking between printers for the tested component.
The average emission of ozone from printers without fil-
ter is 440 /Lig/min for printers. With built-in filters the
average emission is 100 /^g/min. It is shown that these
emissions will give significant contributions to exposure
concentrations in an office.
1061
-------
Introduction
Installation of personal office equipment in poor venti-
lated small offices is often seen to cause unacceptable
indoor air pollution. The nuissance from office equipment
is often complex and can be described as a combination of
draw, elevated temperatures, dust from printers, and
ozone generated from corona discharges in laser printers
and photocopying machines.
Sources to indoor ozone exposure
Ambient air is the common source of indoor ozone. The
concentration is varying in time and by place. The indoor
concentration level from this source is determined by the
actual ambient levels, reduction in the air conditioning
system, and air exchange rate. It is also known from the
literature that electrostatic filters, photocopying ma-
chines, laser printers and other equipment with electric
discharging generate and emit ozon /!/.
Printer technology
The principles of the laser printer technology are shown
in Figure 1. The formation of ozone takes place in an
electric field (AC or DC) between the corona wire (DC,
positive or negative) and a photosensitive surface on the
"image drum".
The generation of ozone from atmospheric oxygen in this
field depends on the field type and the field strength,
which is related to the distance from the corona wire to
the photosensitive surface. The generated ozone will
decompose by catalytic reactions at surfaces. In a number
of printers this decomposition is enhanched by high tem-
perature, a long residence time and a built-in filter.
Equipment and ventilation of the modern office
During the last 5-10 years we have seen the introduc-
tion of personal computers, computer networks and local
laser printers into the modern office. This development
has introduced new relative large sources of ozone and
heat into the office. The contribution from ambient ozone
to indoor ozone levels is much lower in Denmark than in
e.g. metropolitan areas in the United States.
Many Danish office buildings have ventilation systems
which are unable to meet the requirements needed to se-
cure acceptable low ozone concentrations and temperature.
1062
-------
Definitions and principles
The emission of ozone/dust is defined as the total amount
of ozone/dust leaving the printer pr. minute. This defi-
nition allows for catalytic decomposition of ozone and
deposition of dust inside the printer. The definition
means that the emission includes point sources and dif-
fuse sources.
To collect all air streams from the printer it is equip-
ped with a funnel. Diffuse source's and leaks are sealed
with film and tape (Figure 2) and checked by leak test-
ing for ozone. The airflow rate through the funnel and
the concentration of ozone arid dust are determined.
To make comparable results the printer is set to produce
standard test prints at maximum printing rate on standard
80 g/m2 paper, and the tests are performed in a test room
with controlled temperature, humidity and ozone (20°C, 50%
relative humidity, < 0.001 ppm respectively).
Every test includes two independent tests with replace-
ment of the funnel between the tests.
The emission of particulate matter from the printer is of
primary relevance for the deposition inside the printer
or on the built-in filter, where deposition of particula-
te matter increases the pressure loss over the filter and
will contribute to catalytic destruction of ozone.
Experimental
Air volume flow rate: The air velocity v (m/s) is deter-
mined with a calibrated spinwheel anemometer (Lambrecht)
as the mean of 3 repeated 1 minute average values for
each independent test. The area of the funnel opening is
A (m2) , and the air volume flow rate Q (m3/h) is calcula-
ted as:
Q(m3/h) = v(m/s) * A(m2) * 3600(s/h)
Air volume flow rates are reported at standard conditions
(20°C and 1 atra. ) .
1063
-------
Determination of ozone concentration: A conventional
ozone monitor (AID 560) based on the chemiluminicence
principle is used for determination of the ozone concen-
tration C (ppm) in the funnel opening cross section. To
avoid errors from non-uniform distribution over the cross
section the teflon sample probe is traversed. Before each
test the ozone monitor is calibrated using the AID 565
ozone generator, which is calibrated at regulary inter-
vals against a standard iodometric method /2/.
The range for the actual instrument is 0.001 - 10 ppm
ozone.
The ozone emission concentration (E) is determined as the
average of 3 repeated 1 minute average values of the 2
independent tests. The results are read from the instru-
ment display.
The emission of ozone E(microgram/min) at 20° C is cal-
culated as:
E(/ig ozone/min) = C(ppm) * 1995 |Ltg/m3/ppm * Q(m3/h)/60
(min/h)
Determination of particle emission: From the funnel a
representative sample of air is taken isokinetical
through a 37 mm pre-weighed membrane filter (Millipore
AA, 0.8 Mm). The sample time is 60 min. The air wolume is
determined by a calibrated gas meter.
The amount of dust sampled on the filter is determined
gravimetrically with a microbalance (Cahn 25) at con-
trolled air conditions (20°C and 50 % relative humidity).
Detection limit < 0.01 mg/m3.
As a standard routine the composition of the dust is
characterised by optical microscopy.
Results
During the last 5 years our Institute has tested approx.
40 different printers. All printers have been tested
without a built-in ozone reduction filter, and 50% of
these have been tested with an internal ozone reduction
filter. In addition to this our Institute has performed
tests to describe the ozone reduction for new and aged
filters. The tests have been performed for producers and
suppliers to the Danish market, and results for individu-
al printers and external filters can be obtained from the
relevant supplier or producer of the printer.
1064
-------
Test results are shown on the Figures 3-6 for ozone and
Figure 7 for particle emissions.
Table I gives a summary of the obtained results.
Ozone emission from printers without built-in filter
It is possible technically to reduce the ozone produc-
tion rate without the use of filters, e.g. by improved
principles for the electrostatic proces.
Ozone emission reduction by built-in filters
The reduction efficiency is defined as:
R = (l- Efwith fliter)/E(without filter)) * 100 %
E(with filter) and E(without filter) measured at identi-
cal test conditions, paper qualit.y and printer settings
etc.
For the tested printers the ozone reduction varies from
10 to 90%. The sources of variation in the ozone reduc-
tion are:
type of filter (corrugated paper or PUR-foam impreg-
nated with activated charcoal or an activated charcoal
grid)
- leakages caused by incorrect or inexpedient mounting of
filter
increased catalytic reduction of ozone caused by longer
air residence time inside the printer
We have not performed tests to quantify sources of varia-
tions in ozone reduction efficiency. The obtained maximum
reduction efficiency of 90% can be obtained when the de-
velopment, production, installation and operation of
printers are performed according to accepted rules for
quality control. This reduction efficiency can be in-
creased by further development of the internal filters in
the different printers. No apparent or statistical signi-
ficant connection between any of the factors emission of
particulates, the ozone emission or the ozone emission
reduction can be seen.
Ozone emission reduction by external filters
For the different types of external filters on the Danish
market the ozone reduction efficiency has been tested in
1065
-------
connection with different printers from different sup-
pliers. For new filters, correctly installed etc. the
ozone emission concentration is at or below the detection
limit of the instrument, which gives an emission reduc-
tion efficiency beyond 99.9 %.
Discussion
The test method described is shown to give results that
provide us with information on the ozone emission from
laser printers and the methods used to reduce the ozone
emission from these printers. The information on emission
levels can ranking of printers according to their emis-
sion and for predicting exposure levels for different
categories of printers.
Emission reduction:
It is important to notice that with the current used
filters we have experienced a rapid decrease in the re-
duction efficiency of these filters due to particular
deposition on the filters. Higher average values will
then develop with time.
The emission source test method has been used actively in
Denmark for the development of active external filters
based on the principle of a prefilter for particulants
followed by granulated active carbon. This technology
effectively removes dust and ozone, and the ozone reduc-
tion efficiency does not decrease as rapidly as for other
filter types.
Exposure levels for ozone in offices:
In Denmark regulations of ozone emission and exposure are
limited to 100 ppb for industrial work-exposure. For in-
door exposure in offices there is no exposure limit.
Based on the test results of ozone production from the
laser printers and the common model for calculation of
average concentrations in rooms with known ventilation
rate, volume, inlet concentration of ozone, rate con-
stant for surface catalyzed decomposition average concen-
trations can be calculated.
The results are shown in Table 2 for one example and for
different printer categories and printing rates. These
results are also shown on Figure 8.
The equilibrium concentrations obtained by this simple
box-model will give average values for the concentration
of ozone in the room or in a part of a room confined by
thermal movements in the room.
1066
-------
Although equilibrium concentrations are often used for
regulatory purposes, problems are experienced at concen-
trations well below the treshold limit value. In situa-
tions where a confined plume from the printer points
towards a person, the concentration of ozone in the
breathing zone will often be higher depending on the
initial concentration and th€>. dilution by distance and
thermal movements. Consequently such a person might be
effected by the ozone.
References
1. B. Andersen, T.B. Hansen, "Czone and other Air Pol-
lutants from Photocopying Machines", Am. Ind. Hyg.
Assoc. J. 47: 659 (1986).
2. "Ozone in Air method", NIOSH, P&CAM. 154.
1067
-------
c
OJ
0
C
0
ro
4-<
o
C16 Acid
C16 Acid (=)
C15 Acid
C14 Acid\
^ '-'WA.
Pi ,v''« j >1'
C18 Acid (=)
wl/
C18 Acid
•V^^Ji'Ul
l 1 1 1
2nd extraction
f—i—>—r-^i—i—i—i—r"-i—i—i—T^T—i—i "r-i"—r
5 10 15 20
Retention Time (min)
25
Figure 2. SFE-GC/MS analysis of air collected onto a PDF sorbent plug in the
office of a cigarette smoker. Sample collection was for 5 minutes, then the PDF
plug was analyzed using split SFE-GC/MS with a 10 minute extraction (400 atm
CO,).
87
-------
Table I.
Summary of test results
Without filter With filter
Avg.+/-St.dev. Avg.+/-St.dev.
Ozone ng/min. 438+/-27V 102+/-98
Ozone cone., ppb 504+/-290
Particulates Mg/min. — 61+/-74*
* Note: Particulate emission data not normal distributed
Table II.
Ozone equilibrium cone, (in ppb) at different printing
rates.
Printer category Printing rate
100 % 50 % 10 %
Worst case 380 190 39
Average 135 68 14
Low 30 15 3
Minimum background/other sources
Note: ventilation rate q (1 pr. hour), volume v (25 m3) ,
inlet concentration of ozone Co, rate constant for sur-
face catalyzed decomposition k (0,05 min."1) average con-
centrations can be calculated.
1068
-------
Toner Cartridge
(Development)
Print
Cartridge
(Cleaning)
(Charge)
Transfer
Oetack
Figure 1: The xerographic proces
BuilHn filter
Printer
IL
Printer outlet
F:unnel
50cm
Funnel outlet
Measuring crossection
Figure 2; Test principle.
1069
-------
MEASURED EMISSIONS OF OZONE
With and without build-in filters
m
9
/
m
i
n
4 7 10 13 16 19 22 25 28 31 34 37
Printer number
NO Fitter
BUILD-IN Filter
Figure 3: Ozone emissions
PRINTERS WITHOUT
FILTER
OZONE micrograms/min
Figure 4: Ozone emissions from printers without
built-in filter. Histogram.
1070
-------
PRINTERS UITH
BUILD-IN FILTER
396 49>
OZONE microgr/min
Figure 5: Ozone emissions from printers
with built-in filter. Histogram.
PRINTERS WITHOUT
FILTERS OR POOR FILTER
OZONE CONG, microgr./min
Figure 6: Ozone emission concen-
trations. Histogram.
1071
-------
PARTICULATES
FROM PRINTERS
31
U
C
01
01
L
IL
PARTICULATES microgr,/min.
Figure 7: Particulate emissions.
Histogram.
OZONE CONTRIBUTIONS FROM PRINTERS
equilibrium concentrations
MINIMUM LOW AVERAGE WORST
PRINTER CATEGORY
Max./10
Figure 8: Calculated equilibrium
concentrations.
1072
-------
DETERMINATION OF LIGHT AND HEAVY HYDROCARBONS
AND NON-METHANE ORGANIC COMPOUNDS (NMOC) IN
AMBIENT AIR USING A COMBINATION OF METHOD TO-12
AND METHOD TO-14
John V. Hawkins, B. Rubert, S. Wilhite and K. Holloway
Pioneer Laboratory, Inc.
11 East Olive Road
Pensacola, FL 32514
John L. Deuble, Jr.
ERC Environmental and Energy Services Co.
5510 Plorehouse Drive
San Diego, CA 92121
Michael Stroupe
Nutech Corporation
2806 Cheek Road
Durham, NC 277O4
The relationship between ambient concentrations of
precusor organic compounds and subsequent downwind
concentration of ozone has been described by a variety of
photochemical dispersion models. The most important
application of such models is to determine the degree of
control necessary in an urban area to achieve compliance to
air quality standards for ozone?. Elaborate theoretical models
require detailed organic species data obtained by
multi-component gas chromatography. The combination of
techniques found in T012 and T014 allow for successful
identification and speciation of photochemical reactive
organic gases (ROG). Haw the merging of the two methods was
applied to the analysis of ROG samples and what commercially
available equipment was utiliz£?d will be discussed.
1073
-------
INTRODUCTION
The analytical strategy for method TO-14 involves using
a high-resolution gas chromatograph (GC) coupled to one or
more appropriate GC detectors. Method TO-12 utilized the
same type of cryogenic concentration technique used in Method
TO-1 and Method TO-14 for high sensitivity with the simple
flame ionization detector (FID) of the GC for total MNOC
measurements without the GC columns and complex procedure
necessary for species separation.
Approaching the analytical system as an in vivo ambient
sample is the best way to see how the two methods were merged.
Once the canister containing the collected sample is attached
to the analytical system, a gas stream flow is established.
As in Method TO-14 a Nafion dryer is used to remove water
vapor to avoid its formation in the analytical system.
Method TO-12 did not utilize the dryer concept. The sample
is directed to the cryogenic trap and the ROGs are collected
in the cryogenically cooled trap. After a set time the cryogen
is then removed and the temperature of the trap is them raised.
The ROGs originally collected in the trap are revolatilized and
transferred into the gas chromatographic system.
EXPERIMENTAL
As ambient methods have been adopted, modification and
adaptations have made use of technical advancements and
improvements. Compendium Method TO-1 used cryogenic cooling of
the GC inlet to trap the desired analytic. Compendium Method
TO-2 used an efficient trap off-line that trapped and
transferred compounds into the gas cromatograph. Compendium
Method TO-3 made use of the Nafion dryer to reduce moisture, an
offline cryogen cooled trap and multidetector analytical system.
Analytical methods employing trapping concepts are identified in
Table 1.
For the ROG analysis discussed in this TO-12/TO-14 Method a
commercially available system is utilized." A Nutech Model 8533
universal sample concentrator is interfaced to the gas
chromatograph. The model 8533 sample concentrator consists of
two sample concentrator devices, a standard Tenax/charcoal trap
for traditional purge-and-trap operations, and a Compendium
Method TO-14 liquid cryogen trap. This concentration technique
is utilized for many instrumental-analytical methods (i.e., Gas
Chromatography: 501, 601, 602, 8020, 8010, 603 and 8030; and
Mass Spectrometry: 501.3, 524, 624, 8240, 5040 and 5030.)
When in use for ambient air methods the Model 8533 valve
oven is kept at a constant elevated 150 C. The only valve
utilized is the cryo-valve to switch the off-line trap in line
with the gas chromatograph analytical column. Compendium
Method TO-12 collects data valid for the Empirical Kinetic
Modeling Approach (EKMA) which does not require speciation.
1074
-------
Since speciation is important for many projects, the non-
methane organic compound (NMOC) identification and separation
is accomplished by transferring a "discrete bullet" of
collected NMOC onto a fused silica high resolution capillary
column. The column, which is cryogenically cooled and slowly
ramped to a higher separation temperature, allows for optimum
separation.
Compendium Method TO-12 takes the cryo-focused sampled
directly to a flame ionization detector (FID). The organic
compounds previously collected in the trap revolatise due to
the increased temperature and they are carried into the FID,
resulting a response peak or peaks from the FID. The area
of the peak or peaks is integrated, cind the integrated value
is translated to concentration units by using a previously
obtained calibration curve relating integrated peak areas with
known concentrations of propane.
In developing this method for NMOC speciation, a Hewlett
Packard 5890 Series II Gas Chromatograph was utilized along
with a Carle Gas Chromatograph for destermining methane and
carbon monoxide concentrations. The HP 5890 is considered to
have the most stable and reliable, heating and cooling control
which is quite important for speoiation and fingerprinting.
Dual detectors are in-place for the siame reasons as stated
in Compendium Method TO-14. Identification errors can be
reduced by: (a) employing simultaneous detection by different
detectors, and (b) verification of key components to detect
shifts in eluting order or retention time. Interferences on
the non-specific detectors can still cause errors in identifying
a complex sample. The non-specific detector system (i.e., GC-
PID-FID, however, is successfully use^d for quantification of
relatively clean samples. The non-specific detector system can
provide a "snapshot" or "fingerprint" of the constituents in
the sample, allowing determination of: (1) extent of misidentif-
ication due to overlapping or coeluting peaks; (2 position of
the ROGs within or not within the concentration range desired;
and (3) retention time shift patterns.
The O.I. Analytical PID/FID system was chosen for its
design and reliability characteristics that made dual detector
analysis successful. Sensitivity and linearity are well
established with this system.
Compendium Method TO-12 uses electronic integrators for
ssample quantification. For the ROG procedure a Waters Maxima
820 personal computer based data collection system is utilized
to identify and quantify organic analytes. The diagram shown
below denotes how the detector sends an analog signal to the
chromatographic interface (WD24 board). The magnitude of this
signal (measured in microvolts) corresponds to the amount of
sample present in the detector at a given time. The
chromatographic interface converts the analog signal to
digital information.
1075
-------
CONCLUSION
Compendium method TO-12 uses a NIST traceable standard,
propane, at the level of 1-100 ppm (3-300 ppmC) for its
calibration standard. The importance of the Standard
Reference Material (SRM) or a NIST U.S. EPA approved Certified
Reference Material (CRM) cannot be overemphasized. This
modification or merging of Compendium Methods TO-12 and TO-14
lead to the selection of multiple standards for the ROG
procedure. Since speciation is the goal of the ROG analytical
scheme, a Hydrocarbon Library of over 200 compounds containing
either neat liguids or gas mixtures was established from
commercial vendors. The combination of SRM reference standards
and these individual compounds yields accurate and adeguate
identification and quanification. Merging the TO-12/TO-14
Method allows for successful analysis for reactive organic gases
and hydrocarbon speciation.
REFERENCES
Jayanty, R.K.M., et al. (1982), "Laboratory Evaluation of
Non-Methane Organic Carbon Determination in Ambient
Air by Cryogenic Preconcentration and Flame lonization
Detection," EPA-600/54-82-019, USEPA/EMSL, Research
Triangle Park, NC. July 1982.
McClenny, W.A. and J.D. Pleil, J.W. Holdren, and R.N. Smith
(1984), "Automated Cryogenic Preconcentration and Gas
Determination of Volatile Organic Compounds," Anal.
Chem. 56:2947.
McElroy, F.F., V.L. Thompson and H.G. Richter (1985), "A
Cryogenic Preconcentraion-Direct FID (PDFID) Method
for Measurement of NMOC in the Ambient Aire,"
EPA-600/4-85-063, USEPA/EMSL Research Traingle Park, NC.
Rasmussen, R.A. and Khalil, M.A.K. (I960), "Atmospheric
Halocarbons: Measurements and Analyses of Selected
Tract Gases," Proc. NATO ASI on Atmospheric Ozone,
209-231.
1076
-------
Table 1
ANALYTICAL METHODS UTILIZING TRAPPING CONCEPTS
Compen-
dium
Method Description
Target
Compounds
Pre treat-
ment of
Sample
Detector
TO-1
TO-2
TO-3
TO-12
TO-14
TO-12/14
Tenax GC
adsorption and
GC/MS analysis
Carbon
Molecular Sieve
adsorption and
GC/MS analysis
Cryogenic
trapping and
GC/FIDor
GC/ECD
analysis
Canister
collection,
cyrogenic
trapping and
GC/FH) analysis
Canister
collection,
cyrogenic
trapping and
GC/MS or
GC/FTD-PID-
ECD analysis
Canister
collection,
cyrogenic
trapping and
GC/PED-FID
analysis
Volatile, nonpolar organic None GC/MS
(i.e., aromatic hydrocarbons,
chlorinated hydrocarbons)
having boiling points in the
range of 80° to 200°C.
Highly volatile,, ncnpolar None GC/MS
organics (i.e., vinyl chloride,
vinylidene chloride, benzene,
toluene) having boiling
points in the range of -15° to
+ 120°C.
Volatile nonpolar organics
having boiling points in the
rangeof-10°to20D°C.
Organic compounc s
specifically hydrocarbons of
aromatic, aliphatic and
olefinic nature treated as total
NMOC.
Organic compounc.s; a
detailed list of halogenated
and aromatic compounds
have been verified down to
the ppbV level.
Nafion®
dryer to
remove
moisture
None
Nafion®
dryer to
remove
moisture
GC/FID
GC/ECD
GC/FID
GC/MS
GC/FID-
ECD-P1D
Organic connpouncs and Nafion® GC/PID-
reactive gases; methane, dryer to FTO
carbon monoxide, "thane remove
ethylene, acetylene and a moisture
detailed list of parafins
olefms, aromarics and
terpenes.
1077
-------
SUPERCRITICAL FLUID EXTRACTION -
APPLICATIONS IN THE AG INDUSTRY
M. E. McNally
E. I. Du Pont De Nemours and Co.
Agricultural Products Department
Experimental Station, P. O. Box 80402
Wilmington, DE 19880-0402
In the agricultural industry, the identification and
quantification of residual herbicides and their metabolites is
crucial to the registration process. Conventional chemical
extraction procedures for the parent compound and its
metabolites, sometimes unknown, are often difficult, time
consuming and risk interference or transformation of the
species of interest. Once identified, quantitative recoveries of
the residues from the parent compound and its metabolites
88
-------
The Prediction of Atmospheric Stability and the Dispersion of
Emissions from Superfund Sites
C. C. Alien
Center for Environmental Systems
Research Triangle Institute
Research Triangle Park, North Carolina
27709
At Superfund sites, remediation procedures may be needed that result in
emissions of dense vapors, cooling tower emissions, fumes from elevated vents,
and other types of emissions. Because the risk from the remediation activities
can occur during specific atmospheric conditions, it is especially important to
have access to dispersion models that are both site-specific and weather-
specific in order to evaluate the potential effect of (1) remediation activities
before they are carried out or (2) accidents after they occur.
This paper discusses of one approach to model mathematically the meteorological
conditions that are present in the planetary boundary layer and then to use
these calculated conditions to predict the dispersion characteristics of the
atmosphere. These theoretical methods can be used for improving the prediction
of the dispersion of emissions, especially when the meteorological conditions
change during the dispersion, or if there are density effects.
The numerical model dispersion predictions for stable and unstable conditions
are similar to the ISC Gaussian dispersion model using the Pasquill-Gifford
stability classification system. The numerical model requires a surface heat
flux and a geostrophic velocity (wind velocity at an elevated height, 3 km) as
input parameters. The numerical model can predict more stable dispersion than
the most stable class or more unstable dispersion than the most unstable class,
depending on the parameters used.
Basic turbulent dispersion phenomena are discussed and examples are provided for
using the computer model for calculating three-dimensional emission
concentrations near the site boundary or near a specific source. Dispersion
under stable conditions is predicted to be influenced by wind rotation near the
top of the planetary boundary layer. Both static dispersion and dynamic
dispersion (the wind characteristics change with time) are discussed.
1078
-------
The dispersion of air emissions in our environment is complicated by the
wide diversity of source types and the complex nature of air turbulence
that disperses these air emissions downwind. Measurements of air
concentrations can vary significantly according to the wind speed, the
vertical wind shear, the wind direction, and other factors. The
dispersion characteristics of vapors in the planetary boundary layer are
important because the ultimate fate arid environmental impact of vapor
releases depend on the path of the vapors and reaction conditions in the
wind (such as temperatures and concentrations of trace pollutants).
Reaction mechanisms are different at upper elevations in the atmosphere
due to differences in radiation and atmospheric conditions.
Conventional approaches to modeling the structure of the atmosphere rely
on turbulence parameters and correlations. The applicability of the
correlations are limited, so the overall sclution must be obtained by
integrating a series of different correlations into a unified approach.
Danard (1989) presents a good example of this approach.1
The dispersion of air emissions at specific sites is also Important
because of potential health risk to workers and the surrounding
populations. The theoretical approach presented in this paper provides a
method for calculating the wind conditions and for accounting for the
effects of vapor density of the dispersed vapors.
The proposed method of calculating the stability characteristics of the
atmosphere is based upon a direct calculation using geostrophic
meteorological conditions. This numerical method 1s currently under
development, and shows promise because of the power and flexibility of the
method.
TURBULENCE
Turbulence 1s the flow of air 1n swirls or eddies. The flow of vapors,
mists, and smoke follows the flow of the air in the eddies. The flow of
air in eddies can be observed 1n the motion of visible particles in the
eddies. At the edge of the eddy swirls, the smoke density is lowered as
the eddies mix with surrounding air. This reduction in density is eddy
dispersion and it accounts for almost all dispersion of air emissions due
to wind.
The region of the atmosphere where the effects of turbulence are most
apparent is called the planetary boundary layer (100 m to 3000 m over the
Earth's surface). Turbulence can also be important above the planetary
boundary layer if there are temperature or density effects. The model of
the wind described in this paper uses the same equations for the wind
above the planetary boundary layer as are used for the wind below the
planetary boundary layer.
The effect of atmospheric structure on dispersion
The atmosphere generally is being heated or cooled by contact with the
surface of the planet. Neutral stability occurs near the surface when the
surface is the same temperature as the wind over the surface. A situation
similar to neutral stability may occur during certain times during the
day, with high wind velocities near the surface, or when a heavy cloud
cover isolates the surface from solar heating. Turbulent forces shift
eddies with a motion that adds a highly variable velocity component to the
average wind velocity. At high elevations, the shifting forces due to
1079
-------
turbulent surface interaction are reduced. When the forces due to a
density increase in the atmosphere are greater than the forces due to
turbulence, the potential for vertical eddy translations is reduced and
the atmosphere is more stable. This stability is different at each point
in the atmosphere.
Conventional Gaussian dispersion processes assume only one atmospheric
stability factor, but the numerical method can be used to continuously
update the calculated stability of the atmosphere as a function of both
position and time. The ability to change the dispersion characteristics
with time permits the calculation of long term dispersion effects.
The flow of air over thermally neutral surfaces is well characterized,
even if not well understood. From momentum transport, heat transport, and
mass transport correlations developed for flow of fluids in closed
conduits, we can predict the turbulent characteristics of air away from
surfaces.2 The numerical model 1s based upon the concept of wind flow in
discrete eddies, with a shift of position of the eddies from one position
to a higher or lower position. The frequencies of these eddy shifts
depend on shear forces. The density differences in eddies also contribute
to forces acting on the eddy shifts. The density forces are added to the
shear forces. If the eddies in the upper layers are more dense, the total
forces will be greater than in the neutral case and the eddy shift
frequencies will be higher than in the neutral case (this causes unstable
conditions). If the eddies 1n the upper layers are less dense, the total
forces will be less than in the neutral case and the eddy shift
frequencies will be less than in the neutral case (this causes stable
conditions).
The conventional method of estimating the dispersion of vapors from stacks
and other sources depends on empirical correlations such as those used
with the Industrial Source Complex (ISC) models for air dispersion.3
There are short-term and long-term ISC models which differ in how the
results of the dispersion equations are integrated. These models are
generally considered as a standard method of predicting the dispersion of
vapors from sources. The proposed method of calculating the dispersion
characteristics using wind simulation provides a convenient method of
accounting for many factors that these empirical methods do not consider,
such as the effects of height, changes in conditions, changes in wind
direction with time, and plume density.
Model parameters needed for atmospheric simulations
The computer model is designed to simulate the planetary boundary layer
from the global perspective. The meteorological parameters needed are
those that can be measured or predicted remotely from the location of
interest. The only site-specific information that is needed are the
surface characteristics. These surface characteristics are expected to be
constant, with perhaps seasonal variations in some cases.
The geostrophic wind speed and direction at some distant elevation over
the site of interest is an input. Variation of the geostrophic wind with
time during the simulation is permitted.
The heat flux from the surface as a function of time is a significant
input. The heat flux can be calculated from the solar flux and the
surface characteristics, or 1t can be measured on a site specific basis by
interpreting temperature gradients and velocity gradients at the surface.
The temperatures of the atmosphere as a function of height are needed for
1080
-------
the simulation. The model can be used to estimate the temperature
distributions from prior meteorological data if limited information about
the temperatures is available.
Conventional information about the source that is emitting vapors is
needed: area, height, and emission rate,
NUMERICAL MODEL CALCULATIONS
The numerical simulation proceeds with the following simplified flow
structure:
• The vertical temperature profile is calculated from the available
input data. Linear interpolation between data points is used and
the resulting profile is numerically smoothed.
• The wind velocity vectors in both tie x and y horizontal directions
are calculated as a function of the vertical position.
• The eddy diffusivities are calculated as a function of the vertical
position. The eddy dlffuslvitles (!< value) determine the rate of
pollutant dispersion in the wind.
• The surface heat flux is calculated from linear interpolation of
specified flux data.
• The vertical temperature profile is recalculated by solving
Fourier's law. The eddy diffusivities are used to calculate the
thermal conductivities of the wind. The resulting equation that
describes the changes in temperature with time, is a non-linear
second order partial differential eeiuation that is solved by means
of Thomas's algorithm.
• The dispersion of an emitted component is carried out with a three-
dimensional simulation 1n a zone defined independently of the wind
simulation. Concentrations, the location of the center of the
plume, and the horizontal and vertical standard deviations of
concentration about the center of the plume are calculated.
• The above steps are repeated with the updated temperature
distributions.
A comparison of the model results to conventional dispersion correlations.
When the model results are compared to conventional Gaussian dispersion
correlations, it is important to note that the model results depend on
wind velocities, surface heat flux, and the prior history of the wind.
The model results are continuous and do not follow discrete classes. For
the purpose of a comparison between the model results and Gaussian
correlations, the standard deviations of the dispersion distributions were
obtained by a numerical integration of the results of simulation elements
in horizontal and vertical lines intercepting at the point of the maximum
plume concentration.
Table 1 compares of one set of unstable atmospheric dispersion predictions
of the model to conventional Gaussian predictions (stability class A, the
most unstable class). The numerical simulation generated the model of an
unstable atmosphere by heating a neutral boundary layer for 2 h at a
typical mid-afternoon rate of heating (0.005 cal/cm^-s). The dispersion
from a height of 2 m was used for comparison. The horizontal and vertical
1081
-------
dispersion are characterized by ay and az, respectively. The simulation
results are comparable to the results obtained with Gaussian methods,
although the simulation results are non-Gaussian (not normally
distributed).
Table 2 compares of one set of stable atmospheric dispersion predictions
of the numerical model to conventional Gaussian predictions (stability
class F, the most stable class). The numerical simulation generated the
model of the stable atmosphere by passing a neutral atmosphere over a
cooling surface (-0.000075 cal/cm2-s) for a few hours. The model
predictions of the horizontal dispersion were substantially greater than
those of the vertical dispersion, due mainly to the wind turning effect
(Coriolis effect). The standard deviation of the vertical dispersion
predicted by the numerical model at 400 m was less than the Gaussian
correlation.
Figure 1 presents a computer simulation of a cross-section of the plume
under stable atmospheric conditions 42 m downwind of the source. The three
different zones in the figure represent concentrations within a factor of
10, of 100, and of 1000 of the plume maximum concentration. The
distortion of the plume 1s due to predicted wind directional differences
at different heights. Figure 2 presents the same simulation as Figure 1,
but with 10 g/s release of butadiene, which is heavier than air. The
density of the butadiene caused the simulation to predict increased
fumigation.
Differences may exist between the dispersion predictions of the two models
due to factors not considered in the current version of the numerical
simulation, such as terrain effects and wind fluctuations. Experimental
data are needed to confirm the use of the numerical simulation, especially
where the numerical simulation and the Gaussian predictions are different.
Since the Gaussian methods are based on empirical data, the Gaussian
methods are considered more reliable for situations that are very similar
to the dispersion test conditions. For conditions that differ from the
basis of the Gaussian correlations, there may be some advantages with the
numerical simulation. The numerical method is not limited by some of the
constraints of the Gaussian method.
Figure 3 Illustrates fumigation of a stable plume of pollutants as
simulated by the model. The height of the release (84 m) is above the
planetary boundary layer, but as the sun heats the earth's surface, this
boundary layer increases 1n width until the top reaches the plume's
height. When the planetary boundary layer reaches the plume, the plume is
rapidly dispersed and the pollutant quickly reaches the earth's surface.
This numerical prediction illustrates a capability that is beyond the
capability of conventional Gaussian predictions. Both the stable
dispersion factors and the unstable dispersion factors of the Gaussian
correlations would predict lower surface concentrations after fumigation
than are predicted by the numerical method.
REFERENCES
1. M. Danard, "A Prognostic Model for the Surface Temperature, Height of
the Atmospheric Boundary Layer, and Surface Wind," Monthly Weather Review,
117:67 (1989)
2. J. G. Knudsen and D. L. Katz, Fluid Dynamics and Heat Transfer.
McGraw-Hill Book Co., New York, 1958, pp. 155-170.
3. Industrial Source Complex (ISC) Dispersion Model User's Guide-- Second
Edition. Vol I, EPA-450/4-86-005a, (June 1986)
1082
-------
OJ
E
CL
"o
£
gi
X
15
10
I
I
I
I
I
I
-20 -15 -10 -5 0 5 10 15
Lateral Distance from Source (m)
20
Figure 1. The dispersion of a neutral density
plume 40 meters downwind of a point
source.
o>
|
E
'o
£
D)
'01
15
10
I
I
-20 -15 -10 -5 0 5 10 15
Lateral Distance from Source (m)
Figure 2. The dispersion of a dense plume
40 meters downwind of a point
source.
Tablet
A Comparison of Unstable Atmosphere Model
Predictions to Gaussian Predictions3
Distance from
source (m)
50
103
203
303
403
Model prediction"
cz
6.8
13
26
38
47
ay
12
23
43
62
65
Gaussian prediction*1
oz
7.2
13.9
29
47
71
ay
10.3
20.6
40.7
60
79
aPascuill-Gifford Stability class A
oy is calculated at the maximum width of the plume This occurs at a height that is
greater than the point of maximum concentration, ay al the point of maximum
concentration is somewhat less than az.
Table 2.
A Comparison of Stable Atmosphere Model
Predictions to Gaussian Predictions3
Distance from
source (m)
!50
100
200
300
400
Model prediction
az
2.0
2.7
3.4
3.7
4.2
ay
2.3
3.3
5.7
8.4
11.7
Gaussian prediction
oz ay
1.3
2.3
3.1
4.6
7.0
1.7
3.3
6.8
9.9
13
Q>
JD
a.
125
E 100 -
Planetary Boundary Layer
123456
Distance (km)
Figure 3. The fumigation of a stable plume
as the planetary boundary layer
increases with surface heating.
aPa&}uill-Gitford Stability class F
1083
-------
SUBJECT INDEX
Absorbance Spectra 688
Absorbents 69-75,76-81,82,83
94,97,99,100,124-133,200,
202,209-218,339,343,349
350,361,372,483,747-752,
780-788,948
Multi-absorbent Trap
209-218,695-708
Acceptable Ambient Levels 389
Accuracy 8,9,457,1015-1026
Acetoacetone 432-434
Acid Deposition
SESSION 4: 134-166
767-773,1027-1032
Acid Rain 135-139,156,408,409
529-535,655-660,661-668
Acidic Precipitation in
Ontario Study 529-535
Activated Charcoal 129
Activation Energy 98
Aerosols 32,34,41,452-460
462-467,506-511,567-575,
796-801,888,892
Elemental Concentration
567-575
Generation 796-801
Jets 888,892
Mass 567-575
Size Distributions 41
Agriculture Industry 88-93,625
649-654
Agroecosystems 649-654
Aircraft Cabins 542-549,553
Air/Fuel Ratio 292
Air-Mass Transport 156-166
Air Modeling 353-358,378-384
Air Sampling Program 395-400
Air Toxics
SESSION 13: 623-668
SESSION 18:968-1014
336-348,349-352,353,385-394
395-400,407,409,523-528,
536-541,686,695-708,719-725
824-829,849,911-916,1015-
1026,1044-1049,1055-1060
Alcohol Monitors 425-435
Blood 426
Breath 425,426
Aldehydes 268,484-486
753-760,994-999
Alkanes 76,228,229,267
302,310,484,862-867
Alpha Track Devices 7,8
Aluminum 489-497,640,647,648
Foil 489-497
Amberlite XAD Resin 124,129
Ambient Air 183-188,194-199
227-243,371-377,523-528,
536-541,693-698,747-752,
761-766
Ames Test 802,808,809
Ammonia 280,799,800,948-952
Amperometric Electro-
chemical Detection 476
Anaerobic Degradation 372
Analytical Methods 62-87,90
94,95,111,112,117,118,
126-128, 135,136,141,147,
183-199,210,267-284,300-
313,336-352,371-377,389,
419-435,437,543,557,563,
568-570,581-601,603,612,
613,618,677,695-708,719-
725,747-766,781-788,790,
803,804,810,819,824-835,
837,894,899-901,921,1063
1073-1077
Animals 628,630-636
655-660,796-801
Acid Aerosol Exposure 796-801
Anion Exchange Resin 168,170
Annular Denuder System 134,135
789,790,796,798
Aperture Area 305,306,313
Aroclors 66,67,106
Aromatic Amines 483-488
Aromatic Hydrocarbons 194-199
219-226,231,251,291-299,
330-332,483,484,849,919,
948,1055-1060
Atmospheric Chemistry
SESSION 2: 38-61
Atmospheric Diffusion 868-874
Atmospheric Minerals 789-795
Atmospheric Photo-
chemical Processes 267
Atmospheric Simulations 1080
Atomic Emission
Detection 169,173
Audit Procedure 1017-1038
Automated Gas
Chromatographs 200,202
209-218
Automated Systems 695-708
Automobile Emissions 802-807
B
Bacteria 611,802,804,899
Baseline Air Pathway 320-327
Beer's Law
Benzene
Bicycle Transport
Bioassay
301,677
1008-1014
248-250
808-823
1084
-------
Biogeochemistry 637-648
Biological Detoxifi-
cation 371-377
Biological Effects 626,627
634,655-660,803,804
Biomarkers 442-451,542-550
Biomonitoring 655-660
Bode Plot 21
Bootstrap Method 1046,1047
Breath Analysis 418-435
Briefcase Automated
Sampling
Brominated Organics
Brownian Motion
957
38-44
968-973,
Buildings
899-916
Commercial
Diagnostic
Public
Radon
Wakes
Butadiene
543,562
38,39
868-874
8,25-30,395-400
Testing
709
25-30
27
395-400
8
911-916
-717,749
851,857
Calibration Library
861,862
Calibration System 270,278
California Air
Resources Board 1015-1026
Canada Health and
Welfare Program 483-488
Canisters 194-200,219-243
248-259,328-335,339-352,418
536,686-692,726-730,740-746
753-760,833,850-854,948,
1016,1074,1077
Canonical Correlation
Analysis 523-528
Capillary GC, see
Chromatography
Carbonates 54,55
Carbon Dioxide 62-66,69-75
77,78,83,84,94-99,102-107,
110-112,124-129,279-284,
286-290,301,303
Carbon-14 Labelling
Carbon Molecular
sieve Tubes
Carbon Monoxide
507,581,680
70,89
749
279-281
285-290,314,316,425-435
475-482,543,581,
Carcinogens
536,557
818-823
969
168,342,385
579,595,710,809,
988-999,1008-1014
Integrated Air
Proj ect
Catalytic Combustor
Cations
Cancer
994-999
1039-1043
770,771
Chamber Studies
506-511 ,955
1050-1054
Surface Isolation
Flux 955,957,1050-1054
Charcoal Absorption 7,8
Chemical Emissions 244-247
328-335,948-954
Chemical lonization, see
Chromatography
Chemical Kinetic
Mechanism 753-760
Chemical Manufacturing
Facility 975-980
Chemical Mass Balance 537-541
Chemical Processing
Plant, Waste 26,918
Chemical Storage
Warehouse 320-327
Chemical Structures 512-517
Chemical Transfor-
mation 881-886
Chemical Waste Sites 359-372
Chemiluminescence
Analyzer 134,135
Chemometrics
SESSION 10: 512-541
Children 489-497
Chlordane, see
Organochlorine Insecticide
Chlorinated Hydro-
carbons 65-67,70-76
94-1O1 , 123-133,194-199,
204,231-235,244-249,256,
322,324,331,483,484,719,
720,727,836-854,919,948-
954,981,988-993
Chlorinated Water 436-441
988-993
Chlorofluorocarbons 57-61
Chloroform 436-441,988-993
Chlorophyll Degra-
dation 661-668
Cholic Acid Derivatives 107
Chromatography,see also
Gas Chromatography/
Mass Spectrometry
SESSION 3: 62-139
169-172,183-185,190,209,
213-216,254,321,359-361,
372,374,484,491,557,580,
595-601,686-692,711,721-
725,731-739,748,753-767,
830-835,849-860,1016,
1074-1077
Capillary 62-68,830-835
855-867,976,981,995,
1073-1077
Chemical lonization 183-188
580, 595-601
1085
-------
Electrolytic Conductivity
Detector 731-739,1016
Electron Capture
Detection 128,169,209
216,491,557,721-725
Flame lonization
Detection 71,112,117-121
169,171,209,213,215,254,
321,484,686-692,711 ,721-
729,731-739,1074-1077
High Performance
Liquid 103,105,112,190
Ion Trap Detector 169,172
183-188,484,748,767
Matrix Isolation
Microchip
Photoionization
Detection
372,374,1016
Portable Equipment
849-860
Pre-construction
Direct FID
Chromium Sampling
Hexavalent
Trivalent
Citizen Involvement
Clay
Clean Air Act
710,747-752
Federal
State
California
Texas
Clean Air Standards
321
861-867
359-361
830-835
753-766
576-601
579-601
579
414-417
65
407-411
407-409,710
409-411
747-752
285-290
Clean-up Systems 94-101,796-801
Clothing 553,554
Cloud Condensation
Nuclei 156-166
Cloud Water Sampling 136
140-152,156-166
Collisionally Induced
Dissociation 693-698
Combustible Gas 395-400
Combustion Sources 579,580
Community Exposure 948-954
Community Relations
Plans 412-417
Handbook/Guidelines 413
Comparator Controller
Circuit 125,126
Compendium Methods 1073-1077
Complex Matrix 116-120
Comprehensive Environ-
mental Response,
Compensation and Liability
Act of 1980 26,336,337
386,413
Computer Application 31-37
211,254-259,269,280,302,
390-394,427,445,577,649-
654,669,678-685,731-739,
831,851 ,855-868,875-880,
1078-1083
Concentration Pattern 518-522
Congo 45-50
Continuous Sampling 731-739
Contour Maps 896-898
Control Strategies 285-290
Control Techniques 382,383
Cooking Effects 34,35,518,519
556-561,575
Correlation Coeffi-
cients 519,523
Cotinine 542-547
Coulometry 425-435
Cryogenics 110,111,183-188
196,209-218,221,229,255,
321,339-342,537,695-708,
711,727,754-760,780,825-
829,1074-1077
Cyanobacteria 661-668
Cylinders 709-727,780
D
Data Analysis
SESSION 10: 512-541
SESSION 18: 968-1014
Principal Component 518-522
Database,TOXNET 1002
Datalogger 475-482
Data Quality Objec-
tives 928-936
Data Reduction
Techniques 230,231
DDT 45,46
Decay Rates 550-555
Delaware Superfund
Innovative Technology
Evaluation Study
SESSION 6: 200-265
Density 103,110
Deposition, see Acid
Deposition
Dermal Exposure 436-441
489-497,988-993
Desorption 127
Detection Limits 693-698
1044-1049
Dichotomous Sampler 790
Differential Optical
Absorption Spectrometer
669-685
Dimethyl Mercury 371-377
Dioxins 38-44,94-101,123-133
602-610,922-927,1046,1047,
1055,1059,1060
1086
-------
Dispersion 876,893-898
1078-1083
Dispersion Modeling,
see also Models
SESSION 16: 868-916
940,942,961-967
Diurnal Variables 20,21,136
146,148,292,322,462-467,
731,739,961-967,994,998
Dose 445-451
Drainage Water 637,638
Dry Deposition 881-886
Dryers 824
Dust 85,116-119,1061-1072
Dynamic Coulometry 425-435
Economic Effects 408-411
Ecosystems
SESSION 13: 623-668
Electrolyte Leakage 661-668
Electron Capture
Detection, see Chroma-
tography
Negative lonization, see
Mass Spectrometry
Electron Microscopy 51-56
Electroplating Industry 576-579
Elemental Concen-
trations 571-575
Elution Time Data 861-867
Emergency Response 930-936
Emission Measurement 955-960
Emission Sources 518-522
968-973,1008-1014
Environmental Protection
Agency 8,14,25,167-175
195-199,228,248-253,328-
352,360,372,383,384,386,
453,611,612,754,762,917,
1073-1077
AP-42 Emission
Factors 383,384
Method TO-12 1073-1077
Method TO-14 167-175,195-199
228,248-253, 1073-1077
Resource Conservation and
Recovery Act 917
Environmental Tobacco
Smoke, see Tobacco Smoke
Environment Chambers 550-555
Error Analysis 14,15
Margins 679
Ethers 38,39
Evaporative Emissions 317,319
Exhaust 285-319
Exposure Assessment
SESSION 9: 418-511
543-550,1046,1066
Exposure Chambers 796-805
1087
Federal Reference
Methods 669-675
Field Testing 34,35
Filter Sampling 948
Finite Line Source 906
Firefighter Training
facility 395-400
Fireplaces, Wood-
burning 518-522,774-779
Fixed Roof Studies 1050-1054
Flame lonization Detection
see Chromatography
Flexible-fueled
Vehicles 314-319
Floor Wax, see Wood
finishing Products
Flow 893-898
Flue Gas 680
Fluorescent Pigment
Tracers 876
Fluoride Phytotoxicity 630-636
Fluorometric Analyzer 147
Fly Ash 94-101
Forest Fires,see
Smoke, Forest Fires
Forests 135,140,147,148
153,156,630-648,767-773
Southern Commercial 767-773
Formaldehyde 168,170,282,284
301,311,314-319,426,670,
994-999
Fourier Transform Infrared,
see Spectroscopy
Fractals 868-874
Fractionation Process 818-823
Fuel Composition 291-299
Fuels 266,267,271
279,280,291-299,314-319
Fugitive Particulate
Emissions 378-384,617-622
1050-1054,1061-1072
Furans 38-44,94-101
602-610,922-926,1055,1059,
1060
Gamma Radiation 26
Gas Analyzer 830-835
Gas Chromatography,see
Chromatography;GC/MS
Gas Chromatography/
Hass Spectrometry 167-174
183-188, 194-199,229,244-
248,267-269,420,491,558,
603,700,774-780,788,824-
829,922,977,981,986,1016,
1077
Gas Metrology 709-717
Gas Wells 395-400
-------
from the appropriate soil, plant or animal matrix are
performed.
We have demonstrated supercritical fluid extraction as a
technique that has eliminated some of the tedious steps of
current liquid-liquid and solid-liquid extraction procedures.
Extraction procedures can be developed in series, or in
parallel, to selectively remove several components, either
individual compounds, their metabolites or known
interferences, from complex matrices.
The extraction of several components of interest, has
been conducted on a quantitative basis for common agricultural
compounds.
Introduction
SFE has shown great potential in offering shorter
extraction times with higher recoveries(1-6). Sample handling
is minimized and loss due to several time consuming
extraction steps is eliminated. Cleaner sample extracts and
more accurate results can be obtained.
The SFE studies presented focus on optimizing recoveries
for methylphenylureas and sulfonylureas, agricultural
compounds applied to a variety of crops. The environmental
fate of these compounds and their metabolites in a variety of
soil matrices is determined for identification and
quantification purposes. Carbon-14 radiolabeled species of
interest were extracted from both Day 0 and aged soils. The
resulting radiolabeled extraction effluent was collected,
counted via liquid scintillation counting, and overall
efficiencies were quantitatively evaluated.
Extraction conditions have been explored in terms of
mobile phase modifier, extraction phase additives, pressure or
89
-------
Genetic Influences 634
Geographic Analysis 529-535
Global Concentrations,
Pesticides 47
Glow Discharge Source 168,171
Gold Film Sensor 371-377
Graded Screen Arrays 31,32
Graphite 100
Gravitational Settling 881-886
Greenhouse Gas
Reduction Plan 410,411
Groundwater Contami-
nation 395-400
Guaiacols 774-779
Guinea Pigs 38
H
Half-life 41.8-424
Hand Heel Press 489-497
Hazard Index 1005-1007
Hazardous Substances
Priority Lists 385-394
Hazardous Wastes 841
Emi ssions
SESSION 17: 917-967
Hazardous Waste Sites 353-358
378-384,961-967
Hazard Ranking System 930,936
Health Effects 372,385-394
401,929-933,1000-1014
Heating Effects 518,519
556-561,568,994-999,1039-
1043
Henry's Law Constant 137,149
Herbicides 88-93,498-505
649-654
Heterogeneous Reactions 51-56
High Performance Liquid
Chromatography,see
Chromatography
Hi-vol Collectors 774-779
Homes 322,418,452-467
518-522,553,556-561,567-
575,968-987,994-999,1039-
1043
New Construction 981-987
Test House 968-973
Hospitals/Health
Centers 899-904
Human Exposure 388,418-424
483,974-980,988-993,1000-
1008,1045
Humidity Effects 52,824-829
Hydrocarbons 266-284,291-299
314,319,483,719-725,727,
753-760,802-807,919,948-
960,975,981,1073-1077
Hydrochlorofluoro-
carbons 57-61
Hydrogen Chloride - 918
Hydrogen
Hydrogen
Peroxide
Sulfide
146-151
426,432
_
Impinger Train
612-616
Incineration
579-594,611-
Medical Waste
Sewage Sludge
580-601
38,39,95
615,917
611-616
581,586 587
Incinerator Evaruabiorr917-927
Indoor Aerosol Sources 34-37
Indoor Air
SESSION 18: 968-1014
453,461-467,476,483,518-
522,556-575,1061-1072
Industrialized Urban
Areas 536-541
Industrial Sites 201-208
220,227-243,595-601,624,
632,974-980
Infrared Transmission
Monitors 200,286-290
Ingestion 988-993
Inhalation Risk
Assessment 353-358
436-441,988-993
Innovative Technology,
see Delaware Superfund
Innovative Evaluation
Study
Inorganic Anions 770,771
Insecticides, see
Organochlorine
Insecticides
Interlaboratory Compa-
risons 759,763-766
Ion Trap Detector, see
Chromatography
Irradiation 802-805
J_
Jet
Separators
K
Ketones
Kilauea, Hawaii
Krypton-85
824-829
268
142,144
868-874
Laboratory Studies 1039-1043
Landfill 220,224,244,379
381,395-400,523-528,948-
950
Laser 940-947,1061-1072
HeNe 940-947
Printers 1061-1072
Lateral Dispersion 906,907
Lead 639,647,1034
1088
-------
Leak Detection System 940
Legislation 656
Library Search 512-517
Lichens 661-668
Liquid Chromato-
graphy 189-193
Long Path Spectrometry
SESSION 14: 669-779
M
Malaria Control 45
Margin of Exposure 1005-1007
Marine Sediment 66,83,85
117,121
Markers 556-561,611-616
Masking 303,311
Mass, see Aerosol Mass
Mass Balances 637-648
Mass Selective
Detectors 116,117
Mass Spectrometry, see
also GC/MS 78,79,104
171-173,183-193,328-335,
693-698
Materials Aerometric
Data Base 1028-1032
Matrix Isolation,see
Chromatography
McMurdo, Antarctica 142-145
Measurement Methods
SUBJECT OF CONFERENCE
SESSION 14: 669-779
SESSION 17: 917-967
SESSION 18: 968-1014
Measurement Uncer-
tainties 887-892
Medical Wastes 611-616
Membrane Technology 836-838
876
Mercury 371-377
Metals 339,344,350,351
360,361,617-622,637-648,
948,951,1055,1059,1060
Priority Pollutant 617-622
Trace 626,637-648
Meteorological Data 19-24,160
889,890,930-934,940-954,
962,1027-1032,1078-1083
Methane Detection 940-947
Methanol 279,314-319,487
Methoxylated Phenols 774-779
Microenvironments 418-424
452-467,550-555,1002
Microorganisms 611-616
Microsensor 830-835
Microspot Techniques 51
Migration Pathways 336,354
396-398,928,930,962,964-
980,1009
Mineral Particles 51-56
Mobile Laboratory 748
Mobile Sampling 747-752
830-835
Mobile Source Emissions
SESSION 7: 266-319
407,409,418,802-807,
818-823
Evaporative 266,291,314-319
Tailpipe 266,291,314-319
Models
SESSION 16: 868-916
140-145,152-155,157-161,
378-384,401,402,421,444,
445,498,523-528,649-654,
733,955-973,981-993,
1002-1014,1027-1032,
1078-1083
Air Pollution Dispersion 402
CHEMDAT 955-960
Diffusion 894,1010
Farmer's 1010,1011
Finite Line Source 905-910
Fluid 900
Fractional(FRACTAL)
Brownian Motion 8680874
Saussian Plume 881-886
895,911-916,1080
Gradient-Transfer 881
Hazard Ranking System 930
HOTMAC 875-880
HP-BASIC 733
Indoor Air Quality 968-973
Jury 1010
MGCP PLUVIUS 152,153
Pharmacokinetic 444,445
PHYTOTOX 649-654
PUFF 911-916
RAPTAD 875-880
Molecular Spectra 512-517
Monitoring
SESSION 8: 320-417
SESSION 12: 576-622
7-12,27,134-139,167-175,
200-208,442-451,498-505,
519,536-541,669-675,680,
731-739,753-760,763,767-
773,780-788,797-801,836-
848,928-936,948-954,1015-
1032,1055
Monitoring Methods
Research Section 167-175
Mo:ite Carlo Simulation 104
1046,1049
Mosquitoes 45
Motor Vehicles 285-299
Mountain Cloud Chemistry
Project 152,156,157
Mountain Iron Tracer 875-880
1089
-------
Mt. Mitchell
SESSION 4: 134-166
Municipal Waste Site 247
Municipal Water 988-993
Mutagens 168,557,802-823
N
National Acid Precipitation
Assessment Program 1027-1032
National Institute of
Standards and Technology
709-717
National Weather Service
164,165
Nicotine Levels
SESSION 11: 542-575
see also Tobacco Smoke
Nitrogenase Activity 661-668
Nitrogen Oxides 51-56,77,83,84
94-99,124,127,128,134-139,
140-145,147,279-294,300,
303,304,313,314,316,426,
468-470,475,543,631,669-
675,680,802-807,1027,
1028,1034
Nonattainment Areas 408,409
Nonmethane Organic
Compounds 753-766,948-950
955-960,1050-1054,1073-
1077
0
Ocean Emissions 529-532
Odorants 899-904
Odors 168,1003-1007
Office Equipment 1061-1072
Oil Fields 63,940-947
Oil-water Separator 1050-1054
Olympic Mountains 159
On-line Applications 62-75
Ontario 529-535
Organic Compounds,
see also VOC,SVOC,etc. 45-50
65-81,316,506-517
Particle Bound 506-511
Organochlorine Insec-
ticides 45-47
76-81,103
Orographic Cloud 152-155
Outdoor Air 464,467,519,521
556-561,567, 575,974-980,
994-999
Overwater Tracer
Experiment 888
Oxygenated Fuels 291-299
Oxygenates 279-284,291-299
Ozone 135,137,267,280
291-199,408,468-482,631,
669-675,753-760,803,804,
1027,1028,1061-1073
P
Paints 981-987
Paper Samples 104,107
Particle Analyzer 158
Particle Beam 189
Particulates 359-370
378-384,452-467,498-511,
532,543,552-555,557,567-
575,595,617-622,661-668,
774-779,789-795,805,818-
823,917,1034,1039-1043,
1055,1059,1060
Passive Samplers 219-226
483-488
Passive Smoking 550-557
Pathogens 611-616
Performance Optimi-
zation 849-854
Personal Samplers 202
452-505,562-566
Pesticides 45,46,69-78
489-497,624
Petroleum Products,
Waste 955-960,1008
Petroleum Refinery 948-954
975-980
pH 770
Pharmacokinetic
Analysis 418-424,442-451
Phenol Recovery 110-115,170
Photochemical
Oxidants 135,146-151
802-807,1073-1077
Photoionization
Detector, see Chromato-
graphy
Pine Seedlings 767-773
Piseco Lake, NY 142,144
Planetary Boundary
Layer 141
Plants
SESSION 13: 623-668
Plume Generator 686
Plumes 401,402,901,911
930,940,942,967
Polar Molecules 102-109
Polar Volatile Organic
Compounds
SESSION 5: 167-199
Pollutants 38-61,65,82-87
Polychlorinated Biphenyls
76-78,82,102-106
131,132,339,342,360,361,
922,1055,1059,1060
Polycyclic Aromatic
Hydrocarbons 45,46,65,76
82-85, 189-193,342,483,
485,487,506-511,518-522,
556-561
Polyimide Sorbents 69-75
1090
-------
Polynuclear Aromatic
Hydrocarbons 118-122,922
Polyurethane, see
Wood Finishing Products
Polyurethane Foam 70,76-81
489-505,774-779
Portable Air Sampling
System 543,562-566
Portable Chromatograph
see Chromatography
Potassium 571
Potential Source Contri-
bution Function 529-535
Power Plant Emissions 661-668
Prairie Grass
Experiment 888,892
Precipitation,see
also Acid Rain 123-133
769,770
Precision 217,457,598
601,701,705,1015-1026
Preconcentration 209
Pre-remedial Super-
fund Site 320-327,386
Pressure Effect 305
Principal Component
Analysis 518-522
Pseudoephedrine 112,115
Public Input 414-417
Pulp 129-131
Pulsed-wire Anemometer 895
Purge and Trap
System 195,342
Q
Quality Assurance
SESSION 19: 1015-1049
320,321,437,603,701,749,
753,756,759,977
Quality Control 197,921
922,977,1034
Questionnaire 455-460
72,74,899
R
Radioactivity
Radon
SESSION 1: 7-37
Radon Monitoring
Proficiency Program 9
Radon Progeny 32-34
Rainfall 140,159,323,327
Rate Constants 579-594
Ratios 552,553
Reactive Organic
Compounds
Reactivity
Real-time Aerosol
Monitor 360,361
Real-time Emissions
* ' ' 279-284
956
291-299
Analysis
Reference Standards 389
Reid Vapor Pressure 292-299
315
Remedial Action 415-417
930-936
Remote Sensing 177,179,244
285-290,686-692,928-939
Removal Processes 881-886
Renormalized Range
Statistic 868-874
Residences, see Homes
Residential Testing,
Radon 8,13-24,31-37
Resource Recovery
Facility 602-610,1055-1060
Restaurants 562-566
Retention Index 862-867
Risk Assessment 353-358
655,1000-1007,1045-1049
Road Dust 379-384,617-622
Roofing Tar Volatiles 86
Salmonella 808,809
Sample Collection 210,268
330,360,437,489
Samplers,see also
Personal Samplers 31-37
79,200-208,219-243,267,
268,303,321,395-397,437,
452-460,506-511,536,543,
556,557,562-566,568,
576-601,617-622,695,
731-752,780-788,790 1015,
1035,1036,1056-1060
Dilution 617-622
Isokinetic 1035
Mobile 747-752,1036
Solid Sorbent 780-788
Sampling System
Certification 741-744
918-921,948-954
Sand 55
Sand, Gravel and Stone
Operation 401-406
Scanning Electron
Microscopy 789-794
Scavenging, In-cloud 140-145
Sea Salt 51-56
Seasonal Patterns 529-535
604,607
Sector Sampling 227-243
Selected Ion Moni-
toring 96
Semi-volatile Organic
Compounds 70-81,339
349-352,360,361,506,774-779
Sensor 836-848
Showers 436-441,988-993
Shuttle Atmosphere 780-788
1091
-------
Sick Building
Syndrome 1000-1007
Silica 124,131
Site Description 158,164
Site-Specific Studies 974-980
Size Distributions,
Radon Progeny 31-37
Sleeping Sickness 45
Smog 408,409
Smoke 498-505
774-779,818-823
Smoke Visualization
Techniques 901-904
Software 177,832-835,855-867
Soils 55,67,82,83,103-106
116,117,359,371-377,595-
601,647-648,661-668,962,
976,1008-1014
Building Foundation 1008-1014
Crusts 661-668
Solar Radiation 137
Solid matrix 103-107,129
Solvents 95
Sorbents, Solid, see also
Absorbents 780-788
Souce Monitoring
SESSION 12: 576-622
623-629,676-685
Source Testing 1033-1038
Soviet-American Work
Program 893-898
Soxhlet Method 94-98,117,122
Spacecraft Atmosphere 780-788
Spatial Patterns 461-467
Speciation Studies 753-766,798
Spectrometry 157-163
167,170-174,183-188,189-
199,229,340,345,346,617-
622,693-698,790
Energy Dispersive
X-ray 790
X-ray Fluorescence 617-622
Spectrophotometrie
Procedure 580,767-773
Atomic Absorption 767-773
Diphenyl Carbazide 580
Spectroscopic Identi-
fication 512-517
Spectroscopy,Fourier
Transform Infrared 57-61
102,104,171,176-182,202,
244-253,279-284,300-313,
321,686-692
Spectroscopy,Long Path
SESSION 14: 669-779
937-939
Spherocarb Sorbent 372,374
Spores 612-615,789-795
Stability 709-730,1078-1083
Stack Plumes 889
Standards 25,407,469,709-717
Standard Vapor IR
Spectra 177-182
Statement of Work,
Development 336-348,386
State Regulations 388
Statistical Analysis 20-24
461-467,518-528,1044-1049
Storage Studies 702,703
Subsurface Sources 974-980
Sulfur Enrichment 789-795
Sulfur Oxides 134,140-145
147,581-594,631,669-675,
680,789-795,1027-1032,1034
Sulfur Species 529-535
SUMMA, see Canisters
Supercritical Fluid
Extraction
SESSION 3: 62-133
Superfund Amendments and
Reauthorization Act
SARA Title III 328-337
386,928-936,948-954
Superfund Site
SESSION 8: 320-417
25-30,200-208,220,224,244-
247,930,936,1078-1083
Surface Isolation Emission
Flux Chamber, see
Chamber Studies
Target Compound List 385-394
Tedlar Bags 808,809,1016
Telescope, Newtonian 940-947
Temperature Effects 137
140-145,303,304,312,832,
835,852,961-967
Temporal Profile
Analysis 255,461-467
Tenax-GC Adsorption 69-72
94,97,99,123,128-131,183-
188,219,361,371-377,419,
780-788,808-817,969,981-
987
Terrain Amplification 894-898
Test Atmosphere
Generation Chamber 506-511
Thermal Conductivity
Detector 831
Thermal Desorption 70-74
695-708,749,750,780,810,
813,815,981
Thermal Gradient
Diffusion 157
Thermal Pump 123-126
Third World Countries 45-50
Threshold Limit Values 322
1003,1004
Time Series Data 20
1092
-------
Tobacco Smoke 34,87,459
465,518-522,542-575
Total Exposure Assessment
Methodology 419,452-467
Toxic Compounds, see also
Air Toxics
SESSION 2: 38-61
SESSION 13: 623-668
69-75,209-243,328-360,396
483,718-725,837,911-916
EPA Toxic Equivalency
Factor 602-610
Toxic Release Inventory 624
Toxic Fumes 899-904
Toxicology 401-406
Toxic Organic Mixtures,
Group V 718-725
Tracers 774-779,875-880
894-896,922,924,927,938,
939,941,969
Trajectories 152-155,165
Transport, Atmospheric 355,356
623-629,637-654,876
Traps 211-218
2.68,557,562,711,732,733,
755,824-829
Traverse Monitoring 401-406
Treatment, Storage and
Disposal Facility 378,379
Triprolidine 112,115
Troposphere, OH Radical 61
Turbulent Diffusivity 153
868-880,900
Turbulent Dispersion 1078-1083
U
373
Ultraviolet Instrument
Ultraviolet/Photometric
Ozone Analyzer 468-472
Uncertainty Estimates 1027
-1032,1045-1049
Urine Samples 542-549
V
Valleys, Two
Dimensional 893-898
Valveless GC/MS Inlet 174
Vapor Flux 1008-1014,1050-1054
Vapor Generator 188
Vapor Particle
Partitioning 45-50
Variance Studies 14-20
Velocity Fields 893
Ventilation 26
Vibrating Orifice 507,508
Video Imaging 901,905-916
Volatile Organic
Compounds (VOC),
see also Polar VOCs
SESSION 14: 669-779
70,209,320-327,336-348,
349-352,359-370,395-401,
418-424,436,453,484-486,
506-511,526,536-541,780-
788,808-817,830-854,861-
867,928-936,962,968-993,
1000-1014,1017-1026,1050
-1054,1073-1077
Volatilization Rate 955-967
Volcanoes 632
W
Waste Water Treatment 937-939
waiter 824-829,836-848,1034
Ws.ter Analysis 127
WE.ter Chemical
lonization 183-188
Watershed 637-648
We.ter Vapor 183-188
194-199,301,876
Weather Data 320-327,361
Wetlands 372
Whitetop Mountain, VA 148-150
Wind 260-265,320,327,361
401,402,524,525,536-541,
686-692,739,875-879,889,
890,893,894,899-916
Direction 260-265,320,327
361,402,524,686-692,739,
875-879,889,890,905-910
Trajectory Analysis 536-541
Tunnels 893,894,899-916
Velocity 260-265,320-327
402,525,875-879,889,890
Wood-burning Stoves
1039-1043
Wood Finishing
Products 968-973,981-987
Workplaces 418,710
XYZ
XAD Resin 70
X-ray Fluorescence 790-795
ZnSe Window 104
1093
-------
AUTHOR INDEX
Nick Alexandrou
C.C. Allen
Eric R. Allen
Frank Allen
Mark Allen
J.M. Andino
Viney P. Aneja
J. Arello
B
Robert E. Bailey
James Balders
Michael J. Barboza
Jerry R. Barker
P. Michael Barlow
Stockton G. Barnett
William E. Belanger
Douglas A. Bell
J.B. Bell
Jayne Belnap
David W. Berberich
Richard E. Berkley 254
Terry F. Bidleman
Jerry N. Blancato
G.C. Blanschan
Catherine Bobenhausen
William F. Boehler
Anthony S. Bonanno
Emile I. Boulos
D-S. Boyer
Ronald L. Bradow
Scott Braithwaite
Brian Brass
S.L.K. Briggs
T.H. Brixon
Lance Brooks
G. Brorby
D.A. Brymer
Susan S. Bunker
Charles D. Burton
P.B. Bush
Steven Businger
J.W. Butler
Steven H. Cadle
Fern M. Caka
David E. Camann
Zhuang Cao
Sally A. Campbell
Bruce S. Carhart
K.R. Carney
R.E. Carter
A.C. Carver
T.A. Casey
N.P. Castillo
94
1078
767
194
753
300
134,146
686
731
761
395
623
1027
1039
7,13,19
818
669
661
693
,830,849
45
442
611
395
699
102
385
62
152
542
371
567
200
818
595
824
875
291
498
152
300
336
134
740,
266
285
542
489
836
353
385
855,861
686
579
401
808
Kenneth J. Caviston
Joseph S.C. Chang
M. Judith Charles
Julian D. Chazin
Steven R. Chiswell
Han Chou
Judith C. Chow
Jane C. Chuang
A.J. Cimorelli
Candis S. Claiborn
John Clark
L.D. Claxton
Frank R. Clay
C. Andrew Clayton
J.T. Clerc
Joel D. Cline
Jan Clover
R. Murray Colquhoun
Stephen D. Cooper
C. Cowherd, Jr.
Joel Craig
John Crawford
William Crews
H.L. Crist
K. Cross
William E. Grouse
Walter L. Crow
Michael N. Crunk
L.T. Cupitt
F. Curtis
D
Gesheng Dai
D.W. Davies
D-P. Dayton
Donald L. Decker
T.P. DeFelice
Robert B- Denyszyn
Ralph DeSimone
John L. Deuble, Jr.
W.G. DeWees
Gary Dixon
Iver Drabaek
R.J. Drago
Charles T. Driscoll
David Dropkin
Bruce E. Dumdei
Catherine Dunwoody
Thomas G. Dzubay
Delbert J. Eatough
Edward 0. Edney
Robert Effa
Torben Eggert
W.L. Elmore
Darell L. Ernst
830
266
726
,887
38
753
152
981
617
556
401
146
1050
802
576
461
512
940
761
123
418
378
761
542
,314
349
62
562
948
940
802
611
981
796
740
385
156
718
244
1073
579,611
1033
1061
669
637
266
359
1015
789
542
57
1015
1061
378
506
1094
-------
F
B.J. Fairless
J.H. Fateley
W.G. Fateley
L. Felleisen
Philip Fellin
Melvin W. Findlay
Bruce E. Fishman
S. Flack
John S. Fletcher
Linda Forehand
686
686
686
7
506
425,475
1008
595
649
336
J.P. Hsu
Alan H. Huber
E.E. Hudgens
G.F. Hudson
Jody Hudson
T.J. Hughes
Ronald L. Huttie
John S. Irwin
905
489
911
802
796
320,686
808
695
905
Peter A. Gabele 314
Leslie Gage 994
Theodore J. Galen 780
Bruce W, Gay, Jr. 57
C.A. Gierczak 300
F.A. Gifford 868
James M. Godowitch 881
Ramiro Gonzalez, Jr. 1033
Larry W. Goodwin 940
Sidney M. Gordon 183,830
J. Gottlieb 51
Peter R. Griffiths 102
Arne Grove 1061
Zhishi Guo 968
D.F. Gurka 686
H
R.M. Hammaker 686
Steven R. Hanna 887
Lee D. Hansen 542
Rita M. Harrell 726
D. Bruce Harris 25
Chris G. Harrod 359
Francis J. Haughey 19
John V. Hawkins 1073
Steven B. Hawthorne 76,82,774
Fred H. Haynie 1027
James M. Hazlett 731
D.L. Heavner 550
J.L. Hedrick 110
K.M. Hendry 611
William F. Herget 244
.Robert G. Hetes 1000
Jeff B. Hicks 974
M.A. Higuchi 796
Kenneth M. Hill 699
John Hillary 753
Deborah C.Z. Hirsch 412
Ronald A. Kites 76
Xiao H-k 176
Alan Hoffman 994
Michael W. Holdren 209,227
K. Holloway 1073
T. Holloway 686
Philip K. Hopke 31,529
James E. Houck 617
Merrill D. Jackson 968
John T. James 780
Gilbert R. Jersey 940
Wan K. Jo 436,988
Amy S. Johnson 974
Chris E. Johnson 637
R.F. Jongleux 740
K
Larry Kaelin
Robert H. Kagann
Richard M. Kamens
Vinod Kansal
Ronald Karlsson
Barbara B. Kebbekus
Ebrahim Khalili
Leon H. Khurshudyan
Harry Kimball
T.H. Kleindienst
J.J. Kliment
Kenneth Knapp
Charles R. Knipe
J.E. Knoll
T.J. Korniski
P. Koutrakis
Mark S. Krieger
Keith G. Kronmiller
Ji?i Yeong Ku
Michael R. Kuhlman
Michele Kyle
244
328
371
937
38
371
676
981
359
893
320
802
200
266
116
579
300
468,567
76,774
468
475,849
1044
556
489
L
D.D. Lane 686
Andrew J. Lange 102
John J. Langenfeld 82,774
Michael J. Lawrence 94,123
Robert E. Lawson, Jr. 893
B.F. Leaderer 567
K. Leese 611
S.P. Levine 176
J.K. Levy 62
Paul S. Lewandowski 948
Edwin A. Lewis 542
Robert G. Lewis 489
Chi-Shan Li 31
Gene E. Likens 637
1095
-------
Thomas F. Limero
N.-H. Lin
James L. Lindgren
Paul J. Lioy
Christopher C. Lutes
M
Gregory A. Mack
G. Jordan Maclay
David C. MacLean
Bruce E. Maisel
Yaacov Mamane
G.A. Marotz
Joseph M. Martini
Thomas J. Maslany
Steven C. Mauch
Russell McAllister
Alexander 0. McArver
William A. McClenny
200,227
Jim McElroy
Willie T. McLeod
Charles K. McMahon
M.E. McNally
Richard E. Means
Robert N. Meroney
David B. Mickunas
M.R. Midgett
780
140
747
436,988
38
Wayne S
Louis M
David J
Michael
Stephen
Timothy
William
Somenath
Miles
Militana
Miller
Miller
B. Miller
R. Minnich
J. Mitchell
Mitra
Albert Montague
Thomas J. Moser
James D. Mulik
Anauradha Murthy
N
Igor V. Nekrasov
James E. Nellessen
P.R. Nelson
James R. Newman
Barnabe Ngabe
Marcia G. Nishioka
Kelly L. Norton
Philip C. Nyberg
0
L.D. Ogle
Guy B. Oldaker III
Karen D. Oliver
Robert G. Orth
Rein Otson
E.B. Overton
H. Ozkaynak
556
425,475
630
602
51,669,789
686
1055
7
523
336
279
167,183
248,254
761
260,669
498
88
726
899
328
579
116
523
76,82,774
183
718
928
349,726
518
19
623
,475
134
893
649
550
655
45
818
102
25
824
550,562
227,254,849
693
483,506
855,861
452
219,468
P
D. Pahl 452
Li Pan 425,475
Dennis Paustenbach 595,1008
Janusz Pawliszyn 94,123
Edo D. Pellizzari 418,452
William R. Penrose 425,475
E. Perry 802
William B. Petersen 905,911
Terrence K. Pierson 1000
C. Pietarinen 808
William S. Pipkin 116
Joachim D. Pleil 167,183,227
Albert J. Pollack 209,227
David E. Price 1
Lyle C. Pritchett 617
Thomas H. Pritchett 176,328
371,928
Q
Shraddha Quarderer 489
R
Mukund Ramamurthi 31
Andrea J. Randall 385
K. Shankar Rao 881,1044
S. Trivikrama Rao 1044
Susan A. Rasor 968
C. Rawn 808
James H. Raymer 69,418
George C. Rhoderick 709
R. Ricks 595
J. Rizzuto 567
Robert Roholt 1039
John T. Ronan III 407
D.L. Ross 796
A.C. Rosselli 62
B. Rubert 1073
George M. Russwurm 248,260
Tabor T. Sarlos
Thomas E. Sassaman
H. Sauren
V.K. Saxena
Herbert Schattenberg
C.E- Schmidt
R. Kent Schreiber
Robert L. Scotto
R.R. Segall
James C. Serne
P. Sheehan
M.C. Shepherd
R.T. Shigehara
Richard C. Shores
Thomas G. Siccama
John E. Sigsby, Jr.
Orman A. Simpson 244
Mark S Sirinides
III
955
1008
718
669
140
489
1050
655
928
611
1055
595
824
611
1027
637
266,279
937,940
718
1096
-------
Paula Siudak
Tim Slagle
J.L. Slater
D.F. Smith
Jennifer V. Smith
L.J. Smith
R.L. Smith
Richard F. Snow
William H. Snyder
C.M. Sparacino
Leslie E. Sparks
M.L. Spartz
John W. Spence
J. Spengler
Thomas Staley
C.F. Steele
S.C. Steinsberger
Robert D. Stephens
J.H.M. Stephenson
Joseph R. Stetter
R.K. Stevens
J.J. Stout
David G. Strimaitis
Michael Stroupe
Fred D. Stump
David Suder
J.C. Suggs
Robert L. Sutton
Clyde W. Sweet
Thomas 2. Tan
L.T. Taylor
Kent W. Thomas
M. Thomas
Daniel J. Thompson
Roger S. Thompson
Bruce A. Tichenor
David T. Tingey
Richard W. Tripp
R.D. Turpin
Robert J. Tyson
279.
425
314
194
468
802
116
349
401
802
893
808
968
686
1027
452
25
855,861
579
285
194
,475,836
669
855
887
1073
266
955
349
767
536
899
110
418
686
818
893
968
623
320
176
718
V
J.L. Yarns
George Velez
S.J. Vermetti
Kevin Villalobos
Shekar Viswanathan
W
L.. Wallace
Rachel Ward
J, Warner
John G. Watson
Clifford P. Weisel
jEiy M. Wendling
Bert Wheeler
R.B. White
Robert S. Whiton
D. Whittaker
R.W. Wiener
S. Wilhite
C. Herndon Williams
D.D. Williams
Ren Williams
Nancy K. Williams 189
William T. Winberry
Eric D. Winegar
M.R. Witkowski
J.M. Wolfson
R.L. Wong
XYZ
H-k Xiao
Tetsuj i Yamada
Jyisy Yang
Kenneth L. Zankel
Yousheng Zeng
Roy Zweidinger
219
69
536
489
917
452
669,789
808
617
436,988
693
489
824
189
452
452
1073
948
468
818
556
336
974
686
468
861
176
875
102
353
529
994
219
518
1097
-------
density, temperature, equilibration time and extraction mode
i.e. either static or dynamic.
In order to examine the extraction effluents, SFE has
been coupled to SFC and LC. However, more rapid method
development is obtained with off-line examination of the
extraction effluents. Typical residue studies require detection
limits in the parts per billion range. Large injection volumes
along with column switching techniques have been used to
obtain the desired detection limits.
Experimental
1. Sample Preparation: Samples are prepared for
extraction by filling a hollow stainless steel tube (5 cm x 4.6
mm) with the matrix to be extracted. Alternatively, a
cartridge holder for a liquid chromatography guard column can
be used. Little sample pretreatment is required. Soil samples
are generally sieved and the larger pieces of debris and stones
removed. Plant materials are chopped or ground to obtain
smaller particle sizes, often the matrix has been freeze-dried
before this treatment. Stainless steel frits, 0.045 u.m, are
used at both ends of the extraction vessel to prevent the loss
of the matrix during the extraction. Known amounts of reagent
grade sand were also packed at both ends to prevent clogging
of the frits by fine particles in the matrix.
Day 0 samples were prepared in'the laboratory. Known
amounts of the species of interest were spiked onto the matrix
of interest from solutions dissolved in methylene chloride.
The excess methylene chloride was evaporated under a nitrogen
stream. Aged samples were from actual field studies of the
compounds of interest. These samples were subjected to the
natural aging processes due to weather and compound
degradation for the time periods described.
90
-------
-------
2. Instrumentation: The extraction experiments were
conducted on three instruments with equivalent results
obtained although not over equivalent time periods. These
instruments were: a Hewlett Packard 1082 liquid
chromatograph that had been modified for supercritical fluid
delivery (4), a Lee Scientific Model 501 supercritical fluid
chromatograph that was modified with a Rheodyne No 5704
tandem switching valve, and a Suprex Model 200A dual oven
SFE/SFC.
The liquid chromatographic experiments were conducted
using a Hewlett Packard 1090 LC. The liquid chromatograph
was equipped with a photodiode array detector and a column
switching valve to expedite sample clean-up.
3. Extraction mode: Static and dynamic extraction
conditions were used in these experiments. In static
extraction, the mobile phase was allowed to remain in contact
with the sample matrix for a predetermined time period. In
the dynamic mode, the extraction fluid was pumped
continuously through the extraction vessel.
Results
Table 1 shows the results of experimentation comparing
classical methodology with SFE under a variety of conditions.
The addition of extraction phase modifiers demonstrated the
need for these additives to obtain the desired high recoveries.
Without them, acceptable recoveries were not obtained. The
volume of the extraction phase additive was found to be
optimum for a given sample size, but excess additive did not
reduce acceptable recoveries. In certain cases, an optimum
modifier has been found, but in general methanol seems to be
the most universal modifier to achieve high extraction
efficiencies.
91
-------
Static extraction was found to enhance recoveries during
shorter time periods. This is especially useful when the
extraction is coupled with on-line chromatography or analysis.
In this way, larger amounts of solute of interest can be
transfered in a shorter time period.
Table 1
Compound Time
Modifier Method
Volume
Recovery
Diuron 35 min 0.0 ml_ static-dynamic SFE 0%
(CO2 mobile phase)
Diuron 35 min 200 u.L static-dynamic SFE 99%
methanol (CO2 mobile phase)
Diuron 50 min 200 u.L dynamic SFE 86%
methanol (CO2 mobile phase)
Diuron
Diuron
Diuron
105 min 50.0 mL dynamic SFE 99%
methanol (CO2 modified mobile
phase)
3 days > 1 Liter Classical extraction 96%
of mixed methodology with
solvents separatory funnels
35 min 0.0 mL
dynamic SFE
(CO2 mobile phase)
0%
Increases in temperature were found to enhance the
extraction efficiencies of Diuron and Linuron,
phenylmethylurea compounds. The volatility of these
compounds (boiling points less than 100°C) could account for
this effect. In actuality, the reverse trend would be expected
92
-------
since lower temperatures would yield higher densities thus
enhancing the recovery. In the cases examined, the
phenylmethylureas, sulfonylureas, and their metabolites,
increases in density or pressure have increased the extraction
efficiencies obtained.
Conclusions
Reproducible, quantitative recoveries have been obtained
for the extraction of these moderately polar compounds from
soil matrices. Optimization of the experimental method is
rapid. Overall SFE offers the versatility that is needed to
become a dependable analytical sample preparation technique.
References
1. H. T. Kalinoski, H.R. Udseth, B.W. Wright, and R.D. Smith, Anal.
Chem. 58:2421 -25(1986).
2. P. Capriel, A. Haisch, and S. U. Khan, J. Agric. Food Chem.
34:70-73(1986).
3. M. M. Schantz and S. N. Chesler, J. Chromatogr. 363:397-
401(1986).
4. M. E. McNally and J. R. Wheeler, J. Chromatogr. 435:63-
71(1988).
5. S. B. Hawthorne, M. S. Krieger and D. J. Miller, Anal. Chem.
60:472-77(1988).
6. M. E. McNally and J. R. Wheeler, J. Chromatogr. 447:53-
63(1988).
93
-------
ISOLATION OF POLYCHLORINATED DIBENZO-P-DIOXINS AND DIBENZOFURANS
FROM FLY ASH SAMPLES USING SUPERCRITICAL FLUID EXTRACTION
Nick Alexandrou
Michael J. Lawrence
Janusz Pawliszyn
Department of Chemistry
University of Waterloo
Waterloo, Ontario, Canada
N2L 3G1
The use of supercritical fluid extraction yields rapid and quantitative extraction
of polychlorinated dibenzo-p-dioxins and dibenzofurans from municipal incinerator
fiyash. One hour of extraction with nitrous oxide (N2O) at 400 atm and 40°C removes
over 90% of the tetrach!orodibenzo-p-dioxins from fiyash compared to 20 h of
extraction with the Soxhlet method. Pure carbon dioxide (CO2), which does not
remove dioxins at 400 atm, can be used in a clean-up procedure to remove weakly
adsorbed organics. However, CO2 at 400 atm and extraction temperature of 60°C
does isolate dioxins when 10% benzene is used as a modifier, or when the fiyash
matrix is etched with hydrochloric acid. The extraction rate is controlled by the
kinetics of desorption and not by solubilities. The presence of carbon in fiyash inhibits
the isolation of dioxins. Also, particle size and or porosity of the material significantly
effects the extraction process.
Preliminary results indicate that some sorbents may be used, in combination
with supercritical fluids, in the ciean-up procedure of complex matrices such as fiyash.
This process is similar to that used in solid phase extraction. In this procedure,
however, the supercritical fluid replaces the organic solvents. Polychlorinated
biphenyls were partially fractionated from polychlorinated dibenzo-p-dioxins using
TENAX-GC. Similar results were obtained with graphite at 650 atm and 40°C using
benzene modified N20.
94
-------
Introduction
The public's concern with the presence of a number of toxic chemicals in food
and environmental samples has created a demand for improved methods of analysis
of these compounds. The problem associated with municipal refuse disposal
increases as the population of North America increases. Incinerators are now used
to eliminate the problems associated with dumping the waste in landfill sites.
Thousands of tons of municipal incinerator flyash are produced annually with 1-2%
escaping into the atmosphere through stack emission. However this industry has
come under criticism with the discovery of polychlorinated dibenzo-p-dioxins (PCDDs)
and dibenzofurans (PCDFs) in municipal incinerator fly ash.
The first step in environmental analysis of solids such as municipal incinerator
fly ash involves separating the organic compounds of interest from the matrix. This
process is presently achieved by using liquid extractions. This method is time
consuming and very expensive since it requires high purity organic solvents. It also
generates a significant amount of toxic waste since the solvent recycling process is
very difficult to implement.
However, after extraction a complex organic mixture remains which, often
requires application of clean-up procedures to remove interferences for proper
quantitation of trace compounds. At the present time the only methods used require
large amounts of organic solvents, are labour intensive and difficult to automate. For
example, current clean-up procedures for the determination of PCDDs/PCDFs present
in complex samples involve several chromatographic column separations and require
days to complete.
An attractive alternative is supercritical fluid extraction (SFE). This approach
has been explored for some time by chemical engineers1"3 and has recently attracted
the attention of analytical chemists4"10. Supercritical fluid extraction exploits the
properties of this medium at temperatures and pressures near the critical point. This
method combines both distillation and extraction in a single process since both vapour
pressure and phase separation are involved. The major features of supercritical fluid
extraction include: low toxicity of the supercritical carbon dioxide with its low cost and
high chemical inertia; low temperature extractions of non-volatile compounds;
selectivity of the process through density programming (by varying the fluid pressure);
rapid extractions due to low viscosity of fluids and high diffusivities compared to
liquids; easy separation of the solute from the mixture by lowering the pressure; and
low energy consumption due to low temperatures of adsorption and desorption
compared to the high thermal energy required for distillation. Organic solvents can
be replaced by supercritical fluids and sorbents in the clean-up procedures of complex
organic mixtures (such as flyash) produced after SFE of complex environmental
samples.
Experimental
Distilled in Glass Grade solvents (benzene, methanol, hexane, and
dichloromethane) used in this study were purchased from BDH Chemicals, Toronto,
95
-------
Ontario. The liquified, bone-dry carbon dioxide, nitrous oxide, helium, and nitrogen
were acquired from Inter City Welding, Kitchener, Ontario. The flyash was supplied
by the Ontario Ministry of Environment from an electrostatic precipitator of a municipal
incinerator in Toronto, Ontario.
In order to obtain reproducible results, the flyash to be extracted was sieved
to 150 urn. Approximately 10 g samples were Soxhlet extracted with benzene (300
m!_) for 40, 20, 6 and 2 h. Coarse porosity glass fritted extraction thimbles were used
in the extractions. The solutions were reduced in volume by rotary evaporation,
transferred to a 25 Ml pear flask for further concentration and then concentrated, in
a vial, to 500 \iL by a gentle stream of nitrogen.
A high pressure vessel11 was used to supply supercritical fluids for extractions.
The extraction vessel, with a capacity for 0.8 g of flyash, was constructed of Swagelok
316 stainless steel (SS) fittings and 5 cm x 1/4" SS tubing. A fused silica capillary,
20 cm x 20 urn, was connected to the extraction vessel with a graphite-vespel ferrule.
The capillary was directed into a 1.5 mL vial with septum and vent containing
approximately 1.0 mL of hexane. A measured flow rate of approximately 120 mL/min
of the depressurized gas was obtained at 400 atm. The samples were concentrated
to 100 u.L by directing a gentle stream of nitrogen into the vial.
The extraction of PCDDs/PCDFs was accomplished in three different ways.
Benzene (10% by volume) was added to the CO2 in the high-pressure vessel. A
pressure of 400 atm was used to extract the PCDDs/PCDFs at 60°C for 2 h.
Secondly, nitrous oxide (with no modifier) was used as the supercritical fluid. The
temperature of the extraction was 40°C for 2 h at an extraction pressure of 400 atm.
The flyash was treated with 1 N HCI to break down the matrix and facilitate the ease
of the extraction of the PCDDs/PCDFs. The flyash was then stirred for 2 h in 1N
HCI, centrifuged, washed with copious amounts of distilled water, and air-dried. CO2
was used to extract it at 400 atm for 2 h and 40°C. This flyash was also extracted
with nitrous oxide at 400 atm for 2 h and 40°C. In the above SFE procedures,
approximately 0.5 g of fly ash was extracted.
A Varian 3500 capillary gas chromatograph, equipped with an on-column
capillary injector, an electron capture detector (ECD), flame ionization detector (FID)
and 30 m x 0.25 ^im fused silica capillary column (DB-5)(J & W Scientific, California)
was used for chromatographic analysis.
Quantitation of the extracted PCDDs/ PCDFs was achieved with a Hewlett-
Packard HP5890 GC-MSD system. An ionization voltage of 70eV and ion source
temperature of 300°C were used. Before each set of analyses, the instrument was
tuned with the compound perfluorotributylamine (PFTBA). This compound was used
since it is stable and produces fragments throughout the entire mass range. Three
such peaks at m/z 264.00, 413.95, 501.95 were used since they are close to the ion
masses of the PCDDs and PCDFs. Since the mass of the PCDDs and PCDFs are
known, selected ion monitoring (SIM) was used. SIM of the (M-COCI)+, M+, and
(M+2)+ ions for the T4CDD/T4CDF through the H7CDD/H7CDF were used for
identification. The (M-COCI)+, (M+2)+, and (M+4)+ ions were monitored for the
0BCDD/O8CDF. M denotes the parent molecule. To ensure correct identification and
quantitation of PCDDs and PCDFs, a standard mixture of these compounds was
injected before a new set of unknowns was analyzed. A DB-5 column was also used
96
-------
for GC-MSD analysis. The temperature program consisted of an initial oven
temperature of 65°C for 1 min, programmed to 230°C at 15°C/min, held there for 1
min, programmed to 300°C at 3°C/min, and held there for 30 min.
One percent and ten percent (by weight) carbon was added to approximately
10 g of flyash and mixed. Approximately 0.4 g of this mixture was extracted at 400
atm and 40°C for 2 h with N2O. These samples were concentrated to 100 pi with
nitrogen and analyzed with the Varian 3500 GC.
The sorbents (except for TENAX-GC which was cleaned at 400atm for 2 h with
N20) were extracted with the Soxhlet method for 20 h with 250 ml benzene to
remove interfering compounds. Coarse porosity glass fritted extraction thimbles were
also used in these extractions. The sorbents were air dried and approximately 0.2 g
were used in supercritical extractions. A solution consisting of tetrachlorobenzene
(T4CB), hexachlorobenzene (H6CB), trichlorobiphenyl (P3CB ), pentachlorobiphenyl
(P5CB), 1,2,3,4-tetrachlorodibenzo-p-dioxin (T4CDD), octachlorodibenzo-p-dioxin
(O8CDD) was spiked (100 u.L) onto the sorbents. The sorbents were extracted at 400
atm and 40°C with CO2 and N2O. 100 ^iL methoxychlor was used as the internal
standard. Fractions were collected after 5, 10, 15, 25, 45 and 90 minutes. These
samples were concentrated to 100 ^iL with nitrogen and analyzed with the Varian
3500 GC.
Results and Discussion
The method validation procedure was complicated by the lack of certified
standards of flyash. Therefore, in order to evaluate the accuracy, the SFE method
was directly compared with the standard Soxhlet extraction procedure. Analyses were
reproduced at least seven time to ensure good estimation of the accuracy and
precision of the analytical results. Also, different batches of material were analyzed
to ensure good efficiencies of SFE for a range of contamination levels12.
Initial results obtained with SFE were very discouraging. No native pollutants
were removed from flyash using pure C02 at 6000 psi. However, 13C labelled
standard spikes of 2,3,7,8-tetrachlorodibenzo-p-dioxin was successfully recovered from
the flyash matrix. This result clearly indicates that native dioxins and furans are
strongly bound to the matrix; they are most likely chemisorbed onto the original active
sites where they were formed (Figure 1). In this situation the extraction process is
kinetically limited. The large energy barrier assosicated with dissociation of the
analyte-matrix complex prevents the effective extraction. This is in contrast to the
case when analyte molecules are physically adsorbed onto the matrix when the
equilibrium described by the corresponding partition coefficient K is reached rapidly
limited only by the diffusion coefficients. In order to facilitate extraction with CO2 at
400 atm it was decided to etch the matrix with hydrochloric acid in order to eliminate
chemisorption sites and the disorption energy barrier associated with it. This resulted
in complete removal of native pollutants (Table I), but this procedure required an
additional step. The precision of data was also poor due to highly inhomogeneous
material produced after treatment of the matrix with acid. Therefore, in the next step
small amounts of benzene were added to supercritical carbon dioxide. This solvent
97
-------
was chosen because it has been proven to be effective in Soxhlet procedures.
Indeed, quantitative removal of native dioxins was achieved with this approach.
However, this procedure was less desirable as it used toxic solvents and replacement
of this solvent with methanol resulted in lower extraction efficiencies. In the final
procedure nitrous oxide, a slightly polar gas which has been proven to extract
polyaromatic molecules more rapidly, was used at 400 atm. Using this procedure the
native dioxins and furans were quantitatively removed from the untreated fly ash
matrix. The extraction time is about 1 h versus 20 h for Soxhlet extraction.
j
•Bt/BT
BU3HXK4D
Figure 1 Activation energy barrier
GROUP OF COMPOUNDS
POLYCHLORINATED
DIBENZODIOXINS TOTAL
T4CDD
CO2+10%
benzene
400 atm
60°C
2h
92.4±9.8
117±12
CO2add
treated fly-
ash, 400 atm
40°C
2h
86.5±11.4
9€±25
N-O
400
aim
40°C
2h
B4±4
98±7
TOTAL POLYCHLORINATED
DIBENZOFURANS 91 ±13.6 9€.4±10.4 88±4
Table I Percent extraction efficincies, as compared to 20 h Soxhlet method, of
polychlorinated dibenzo-p-dioxins and dibenzofurans from flyash with supercritical fluids.
98
-------
The dramatic difference between SFE results obtained from CO2 and N2O,
structurally similar molecules, is somewhat unexpected. We believe that investigations
prompted by this result will lead to a better understanding of the nature of supercritical
fluid extraction. We are presently studying the extraction rates of the native dioxins
and furans from flyash as a function of the SFE polarizability/polarity parameter by
varying the fluid density. The initial results indicate that the nitrous oxide fluid
molecules are involved in a specific interaction with the matrix which lowers the
activation energy barrier to the desorption of chemisorbed molecules.
The difference in extraction efficiencies between CO2 and N2O can be used
effectively to clean-up complex organic mixtures which are present in the flyash.
Weakly adsorbed interferences can be removed first with CO2 followed by extraction
of compounds of interest with N2O. In addition, our more recent results indicate that
pure CO2 is capable of removing native dioxins and furans from fly ash matrix when
the extraction pressure is 650 atm or higher.
The results outlined above clearly illustrate the procedures which can be used
to optimize conditions for supercritical fluid extraction. It appears that by increasing
the pressure, and therefore the density and polarizability of the fluid, the desorption
strength of the fluid increases. These observations emphasize the difference
between the engineering applications of the supercritical fluid extraction, where
relatively large amounts of the product are dissolved in the fluid and transferred to the
collection vessel, and the trace analysis applications where minute quantities of
specifically adsorbed material must be removed quantitatively from the matrix. In
order to achieve the engineering goal, it is, in most cases, sufficient to ensure that the
material of interest is soluble in the supercritical fluid, while to ensure quantitative
removal of pollutants, it is necessary to use a strong solvent that will remove organic
components from the matrix.
Our experience with the flyash matrix indicates that the extraction efficiencies
obtained for a spiked standard can be drastically different than those for native
pollutants. It is difficult or, in some cases, impossible to prepare spiked standards
which are equivalent to native contaminants, due to aging effects associated with
phenomena such as chemisorption and porosity effects13"14. Therefore, estimation of
accuracy of analytical methods by procedural spikes need to be treated with caution.
Preliminary results indicate that even different size flyash particles yield significant
extraction profiles.
There is some concern that carbon levels in flyash affects the SFE recoveries
of PCDDs/PCDFs. This is critical to the general applicability of the SFE method as
the type of refuse, the incineration temperature and the fuel used may affect carbon
levels in the ash. It is known that carbon adsorbs PCDDs/PCDFs, as the final step
in the clean-up procedure of complex samples15 consists of a carbon column. Initial
results involving 1 % and 10 % carbon revealed that carbon does indeed prevent the
extraction of PCDDs/PCDFs from flyash. 51.9% of O8CDD was isolated from the
flyash matrix when 1 % carbon was added. The extraction of this compound dropped
to 18.3% when 10% carbon was added to flyash. Both carbon and PCDDs have a
planar structure which allows for a greater adsorption of these compounds by carbon.
Various sorbents are used in model studies for the extraction of environmental
-------
pollutants. A number of these sorbents (TENAX-GC, CM, C18 and graphite) were used
to investigate the possible clean-up of organics from environmental matrices. This
fractionation scheme resembles the process of solid phase extraction except, the
organic solvents are replaced by supercritical fluids.
Ideally, a sorbent will retain all compounds initially, but release progressively
less soluble compounds as the pressure is increased. Initial results indicate that C18
and CN produced full recoveries of T4CB, H6CB, P3CB, P5CB, T4CDD, O8CDD after
5 minutes of extraction with both CO2 and N2O.
Two possible sorbents that may be used to clean-up complex organic mixtures
are graphite and TENAX-GC. Extractions of graphite at 650 atm with N2O removed
T4CB, H6CB, P3CB, and P5CB. Addition of benzene, as a modifier, recovered the
planar T4CDD. N2O extraction of Tenax produced quantitative recoveries in 5 minutes.
However, CO2 was able to fractionate some of the compounds. The T4CB and H6CB
were extracted in the first 5 minutes while P3CB was isolated in the first 10 minutes.
The O8CDD was removed in the last hour. Therefore, both Tenax and graphite may
be used to clean-up complex mixtures produced after SFE of samples such as flyash.
Conclusions
The time necessary to isolate dioxtns from municipal incinerator flyash was
reduced from 20 h with the Soxhlet method to 2 h with supercritical fluids.
Substantially smaller quantities of flyash are needed in SFE than with the Soxhlet
method. Furthermore, the limiting step in the extraction rate appears to be the
kinetics of desorption rather than the solubilities of these species in CO2 or N2O. The
variance and standard deviation associated with supercritical fluids was lower than
that of the Soxhlet method.
The presence of carbon in flyash reduces the efficiency of the isolation of
PCDDs/PCDFs from the matrix.
Some sorbents may be utilized in the same manner as in solid phase extraction
for the clean-up of complex organic samples. Clean-up procedures can be based on
changing the supercritical fluid, addition of a modifier or varying the pressure.
References
1. G., Wilke,"Extraction with supercritical gases-a foreword",
Anoew. Chem. Int. Ed. Enal.. 17, 701, (1978).
2. D.F., Williams, "Extraction with supercritical gases",
Chem. Eng. ScL 36(11). 1769, (1981).
3. M.A., McHugh, V.J., Krukonis, Supercritical Fluid Extractions:
Principles and Practice. Butterworths, Toronto, 1986.
100
-------
4. S.B., Hawthorne, D.J., Miller, "Extraction and recovery of organic pollutants
from environmental solids and TENAX-GC using supercritical CO2M, Anal.
Chem.. 57, 1705, (1987).
5. B.W., Wright, S.F., Frye, D.G., McMinn, R.D., Smith, "On-line supercritical fluid
extraction-capillary gas chromatography", Anal. Chem.. 59, 640, (1987).
6. S.B., Hawthorne, D.J., Miller, J.J., Langefeld, "Quantitative analysis using
directly coupled supercritical fluid extraction/capillary gas chromatography (SFE-
GC) with a conventional split/splitless injection port", J. Chromatoar.
ScL 28, 2, (1990).
7. K., Sugiyama, "Directly coupled laboratory-scale supercritical fluid extraction-
supercritical fluid chromatography, monitored with a multiwavelength ultraviolet
detector", J. Chromatoar.. 332, 107, (1985).
8. F.I., Onuska, K.A., Terry, "Supercritical fluid extraction of 2,3,7,8-
tetrachlorodibenzo-p-dioxin from sediment samples", J.. High Resol.
Chromatoar.. 12, 357, (1989).
9. B.W., Wright, C.W., Wright, R.W., Gale, R.D., Smith, "Analytical supercritical
fluid extraction of adsorbent materials", Anal. Chem.. 59, 38, (1987).
10. S.B., Hawthorne, M.S., Krieger, D.J., Miller, "Supercritical carbon dioxide
extraction of polychlorinated biphenyls, polychlorinated aromatic hydrocarbons,
heteroatom-containing polycyclic aromatic hydrocarbons, and n-alkanes from
polyurethane foam sorbents", Anal. Chem.. 61, 736, (1989).
11. J., Pawliszyn, "Inexpensive fluid delivery system for supercritical fluid
extraction", J. High Resol. Chromatoar.. 13, 199, (1990).
12. N., Alexandrou, J., Pawliszyn, "Supercritical fluid extraction for the rapid
determination of polychlorinated dibenzo-p-dioxins and dibenzofurans in
municipal incinerator fly ash", Anal. Chem.. 61, 2770, (1989).
13. D., Di Toro, L, Horzempa, Environ. Sci. Technol.. 16, 594, 1981.
14. 2., Gerst, Y., Chen, V., Mingelgrin, B., Yaron, eds., Toxic organic chemicals in
porous media. Springer-Verlag, New York, NY, 1989.
15. Paprican clean-up for pulp extracts. Pulp and Paper Research Institute of
Canada, Personal Communications.
101
-------
EXTRACTION AND SEPARATION OF POLAR MOLECULES USING
SUPERCRITICAL FLUIDS WITH SUBSEQUENT IDENTIFICATION
BY FT-IR SPECTROMETRY
Peter R. Griffiths.
Kelly L. Norton,
Andrew J. Lange,
Jyisy Yang, and
Anthony S. Bonanno
Department of Chemistry
University of Idaho
Moscow, Idaho 83843
Organic compounds of medium polarity can be rapidly and
efficiently extracted from most solid substrates with neat
supercritical carbon dioxide. The difficulty of extracting
organics using supercritical fluids increases with the polarity of
both the analyte and the substrate. Extraction efficiency can
usually be increased under isobaric conditions by increasing the
temperature, even though the density of the fluid may be
decreased. Extraction efficiency is also increased by the
addition of polar modifiers to C02. Extracts are condensed on a
short length of an appropriate chromatographic column mounted
immediately after the restrictor, and separated as soon as the
extraction is believed to be complete. A generalized method of
trapping chromatography eluites as small spots on a moving ZnSe
window has allowed Fourier transform infrared (FT-IR) spectra of
low nanogram and subnanogram quantities of organic analytes to be
measured in real time. This chromatography/FT-IR interface
provides a complementary technique to mass spectrometry through
which the substitutional isomers of aromatic compounds can be
readily distinguished. The performance of the supercritical fluid
extraction/supercritical fluid chromatography/FT-IR spectrometry
interface is illustrated by the extraction and identification of
polychlorinated biphenyls from soils.
102
-------
The ease of extracting organic molecules from solid matrices
using supercritical carbon dioxide decreases both with the
polarity of the molecules being extracted and that of the matrix.
The closer is the polarity of any molecule to that of the
extractant the greater is its solubility. Since CO is a nonpolai
(albeit highly polarizable) molecule, compounds of low and medium
polarity have the highest solubility. It is often assumed that
the density of supercritical fluids controls their solvent
strength. Certainly for extractions carried out isothermally, the
efficiency increases with pressure, as not only does the density
increase but also the mass flow rate through a given restrictor
increases with the pressure differential. For fluids at a
constant density, the solubility of organic molecules increases
with temperature. Keeping the density constant while varying the
temperature often necesitates a disproportionately large variatior
in pressure. We have found, that extraction efficiency often
increased with temperature at a given pressure (i.e. with a set
flow rate), even though the density of the extracting fluid may
drop by more than a factor of two.
The ease of extracting organic molecules from solid
substrates also depends on the strength of the interaction betweer
the analytes and the substrate surface. The greater the number of
hudrogen-bond donors and acceptors contained by the analyte and
the matrix, the poorer will be the extraction efficiency, since
CO- does not interact strongly with hydrophilic substrates or very
polar analytes. In this case, the extraction efficiency should be
dramatically increased by the addition of a polar modifier such as
methanol or water to the CO., as the modifier molecules can from
hydrogen-bonded complexes with polar analytes and will interact
strongly with OH groups on the surface of the substrate,
effectively blocking each active site from interacting with polar
analytes.
Polychlorinated biphenyls (PCBs) are analytes of medium
polarity that do not interact strongly with siliceous surfaces.
It is relatively easy, therefore, to extract PCBs from soils using
supercritical CO,. The very sharp change in extraction
efficiency occurring around 1200 psi is very surprising in light
of the relatively slow change in density with pressure. Extraction
efficiency was measured by accumulating the extracted PCBs on a
short length of a capillary supercritical fluid chromatography
(SFC) column for a time of 20 minutes, after which time they were
separated by capillary SFC, see Figure 1A. This technique
incorporates many of the principles that were developed by Miller
and Hawthorne for their supercritical fluid extraction (SFE)/gas
chromatography (GC) interface [1]. The total quantity of the
extract was esitmated by measuring the integrated area under the
chromatogram and comparing the result with the integrated area
under the chromatogram obtained after a conventional injection of
a solution of the same PCB at a known concentration, see
Figure IB. It is equally feasible to trap supercritical fluid
extracts on wider bore tubing for subsequent separation by gas
chromatography or on short lengths of packed columns for
separation by packed column SFC or high performance liquid
chromatography (HPLC).
For extractions of standard samples of PCBs in
soil (Environmental Resource Associates) at 50°c, the extraction
efficiency is strongly dependent on the pressure, as shown in
Figure 2. Also shown on this plot is the variation of the density
103
-------
of CO with pressure at 50°C.
The more polar are the organic analytes, the more slowly they
are extracted from polar matrices, presumably because of the
existence of strong hydrogen bonding between the analytes and the
matrix. This effect is illustrated by the SFE of indigeneous
organic molecules from paper using neat CO-. Supercritical fluid
chromatograms of four consecutive extractions of a paper sample
are shown in Figure 3. The earlier peaks in the chromatogram are
significantly reduced in intensity in Figure 3E, indicating that
these components are being extracted at fairly high efficiency.
Conversely, the later-eluting peaks do not diminish appreciably in
intensity from Figure 3A to 3E indicating that they are more
tightly bound to the cellulose matrix and possibly even form a
saturated solution under the conditions of the extraction.
The higher the polarity of the analytes, the more difficult
they are to solvate and to separate by capillary SFC using neat
CO as the mobile phase. Polar analytes exhibit greater peak
widths and a greater degree of tailing than nonpolar compounds
separated using the same chromatographic parameters. In this
case, the addition of a polar modifier to the C02 mobile phase
will alleviate the problem of tailing, although neither the
retention time nor the peak width are changed to a great extent,
see Figure 4. For such molecules, reverse-phase HPLC appears to
be the optimal separation technique.
The identification of large molecules often requires the
combination of several spectroscopic techniques. Mass
spectrometry (MS) is the technique of choice for identifying GC
eluites, but even for GC the additional measurement of the
infrared spectrum reduces the ambiguity of the identification [2].
The higher the molecular weight of the analyte, the more ambiguous
is its mass spectrum and the greater is the need for additional
spectroscopic data. In view of the complementary nature of
infrared and mass spectra, infrared spectrometry would appear to
be the technique of choice in this regard. However, interfaces
between chromatographs of all types and FT-IR spectrometers are
commonly perceived to be of much lower sensitivity than the
corresponding chromatography/MS system.
In the past few years, we have shown that "direct deposition"
GC/FT-IR [3,4] and SFC/FT-IR [5,6] interfaces significantly
reduce the minimum injected quantity required for the measurement
of an identifiable spectrum. In these interfaces, each eluite is
deposited on a moving ZnSe window as a spot of very low area
(<0.02 mm ), so that even a 1-ng deposit has a thickness that is
sufficient to yield an identifiable spectrum in one second using
an FT-IR spectrometer equipped with microscope-style optics. To
illustrate the power of SFC/FT-IR measurements based on this
principle, the spectra of the six peaks indicated in Figure 1 are
shown in Figures 5 and 6, along with the identification of five of
these components made solely on the basis of their infrared
spectra. The sixth peak appears to be a mixture of at least two
unresolved components.
The SFC/FT-IR spectra shown in Figures 5 and 6 were measured
by first depositing each component on a ZnSe window at ambient
temperature held immediately below the restrictor. The spectra
were then measured subsequently using a Perkin-Elmer Model 1800
FT-IR spectrometer equipped with a Spectra-Tech (Stamford, CT) IR
Plan microscope, with the aperture set a 100 /*m. A GC/FT-IR
interface based on the principle of direct deposition is now
104
-------
available commercially (Digilab Division of Bio-Rad, Cambridge,
MA). We are presently modifying this interface for on-line
3FC/FT-IR and there is every indication that it will yield
equivalent spectra to these shown in Figures 5 and 6. We have
already tested this interface with water-modified CO , and there
is no trace of water in the measured spectra, even wnen the windo1
is cooled to -20°C.
It was noted above that for very polar analytes, reverse-
phase (RP) HPLC will probably yield superior separations than SFC
because of the greatly increased polarity of the mobile phase. W<
are developing a direct deposition RP-HPLC/FT-IR interface which
is based on the same principles as the GC/FT-IR and SFC/FT-IR
interfaces described above. Aqueous solvents can be eliminated b]
a concentric flow nebulizer. With flow rates of water up to
!50 /il/min, the area of deposits can be reduced below 0.02 mm .
This has not yet been applied to true HPLC separations, and it is
presently being tested using injections of solutions of a single
analyte in a "flow injection analysis" ("FIA") mode. A
preliminary spectrum, of 60 ng of methyl violet 2B deposited from
100% water at a flow rate of 50 pd/min measured in this manner is
shown in Figure 7. The high signal-to-noise ratio of this
spectrum indicates that subnanogram detection limits will be
achievable for HPLC/FT-IR in the very near future.
Acknowledgments
The authors gratefully acknowledge current support by the
U. S. Environmental Protection Agency under Grant No. R-814441-0
and previous support by EPA Environmental Monitoring Systems
Laboratory/Las Vegas uncer Cooperative Agreement No. CR-812258-03.
We also wish to acknowledge the invaluable contribution of Dr.
Roger Fuoco of the National Research Council, Pisa, Italy, who
developed the methodology for the SFE/SFC/FT-IR studies of PCBs or
soil.
References
1. S.F. Hawthorne and D.J. Miller, "Directly coupled
supercritical fluid extraction - gas chromatography analysis
of polycyclic aromatic hydrocarbons and polychlorinated
biphenyls from environmental solids:, J. Cromatogr. 403: 63.
(1987) .
2. C.L. Wilkins, "Linked gas chromatography, infrared and mass
spectrometry", Anal. Chero^ 59; 571A. (1987).
3. P.R. Griffiths and D.E. Henry, "Coupled gas chromatography anc
Fourier transform infrared spectrometry", Progr. Analyt.
Spectrosc. 9: 455. (1986).
4. A.M. Haefner, K.L. Norton, P.R. Griffiths, S. Bourne and R.
Curbelo, "Interfaced gas chromatography and Fourier transform
infrared spectrometry by eluite trapping at 77K", Anal. Chero.
60.: 2441. (1988).
5. S.L. Pentoney, K.H. Shafer and P.R. Griffiths, "A solvent
elimination interface for capillary supercritical fluid
chromatography/Fourier transform infrared spectrometry using
an infrared microscope", J. Chromatoar. Sci. 24; 230. (1986).
6. P.R. Griffiths, S.L. Pentoney, G.L. Pariente and K.L. Norton,
"A unified approach to the chromatography/FT-IR interface",
Mikrochim. Acta fWienl. Ill: 47. (1988).
105
-------
B
SO
SS 60
TIME (MINUTES)
Figure l. (A) SFC of standard solution of Aroehlor 1260 (960 ng
injected) after removal of solvent, (B) SFE/SFC of a
certified soil sample containing Arochlor 1260. If all
the PCBs were removed quantitatively, 910 ng would have
been extracted.
100
CO
X 90
S 80
-2
o
N
60
50
40
3°
20
10
0.8
0.7
0.6
0.5
CuO
0.4 OT
0.3 o
0.2
600 1000 1400 1800 2200 2600 3000
Extraction pressure (psi)
Figure 2. (n) Extraction efficiency as a function of pressure for
same certified soil sample used for Figure 1; (°)
Variation of density of C02 as a function of pressure.
106
-------
5 10 15 20
Time (min)
Figure 3. Chromatograms of successive supercritical fluid
extractions of a paper sample.
CO
e
o
CL,
VI
V
o -\
-20 -
3X
B
o
2000
2O 4O
Time (min)
27OO 34OO
Pressure (psi)
6O
4000
Figure 4. Chromatogram of a mixture of cholic acid derivatives
(lithocholic acid, methyl cholate, chenodeoxycholic
acid, deoxycholic acid, and cholic acid) separated on a
biphenyl capillary column using (A) carbon dioxide, and
(B) CO containing 0.1% water.
107
-------
.06-
.055-
.05-
,045-
.04-
« .035-
o
c
o
€ .03-
o
V)
XI
< .025 -j
.02-
.015-
.01
.005
0
Peak #3: 2,2',3,3',5',6-Hexachloro
Peak #2: 2,2',3,5',6-Pentachloro
Peak #1 2,2',4,5,5'^Pentachloro
u
bAJ
4000 3500
3000
2500
2000
1500
toco
Wavenumbers (cm-1)
Figure 5. SFC/FT-IR spectra of the first three major peaks in the
chromatogram shown in Figure 1; tentative
identifications are indicated above each spectrum.
108
-------
.11
.09
.08
.07
v
% -06
.
0
5 .05
.04
.03-
.02-
.01-
0-
Peak #6: 2,2',3,3',4,4',5-Heptachlorc
Peak #5: Un-
resolved mixture
Peak #4: 2,2',3)4,4',5'-Hexachloro
4000 3500 3000 2500 2000 1500
1000
Wavenumbers (cm-1)
Figure 6,
Figure 7
SFC/FT-IR spectra of the last three major peaks in the
chromatogram shown in Figure 1.
A
B
S
o
R
B
A
N
C
E
0.2
0.1
4000
1000
3000 2000
WAVENUMBERS
"FIA"/FT-IR spectrum of 60 ng ot metnyivioier <>.& arcer
elimination of mobile phase (100% water) at a flow
rate of 50 /il/min.
109
-------
QUANTITATIVE SUPERCRITICAL FLUID EXTRACTION/SUPERCRITICAL FLUID
CHROMATOGRAPHY FROM AQUEOUS MEDIA
J. L. Hedrick and L. T. Taylor
Virginia Polytechnic Institute and State University
Department of Chemistry
Blacksburg, VA 24061-0212
Supercritical fluid extraction using CO2 has been used to
extract various analytes from aqueous media. Both static and
dynamic extraction modes are reported. The rate of extraction is
shown to be a function of not only the density of the
supercritical fluid but also the ionic strength of the aqueous
media. A new scheme for trapping of analyte onto a solid phase
by both cryogenic and adsorption mechanisms is described. The
new scheme shows marked improvement over simple trapping into a
liquid solvent. The recovery of phenol from water at the 20 ppm
level is reported as 80% with an RSD of 9%.
110
-------
Introduction
To date most of the work which has been done with
supercritical fluid extraction (SFE) has centered on the
extraction of analytes from solid, insoluble or slightly soluble
matrices1"4. Experimentation, however, is needed on systems
where undesired soluble matrix components are present. King2 has
noted that it may be possible to isolate low levels of analytes
from interfering matrix components by precisely controlling the
extraction pressure. For example, he has shown that most of the
DDT can be removed from lipids if the extraction is preformed at
approximately 100 atm. A wet matrix will obviously be a problem
in CO2 dynamic extractions even though water is soluble in
supercritical CC>2 to only approximately 0.1 %. Over time
extracted water can cause restrictor plugging, and modification
of the deposition and/or chromatographic solid phase. If the
accumulator is a nonpolar liquid phase, the presence of water in
the extract can result in a two-phase system.
Previous work in the area of extracting material from wet or
liquid media has involved essentially conventional supercritical
fluid extraction techniques employed on dried (freeze dried or
heated) or trapped material (by solid phase extraction). Methods
of this type are useful in cases where the analyte is (a) weakly
retained, (b) not extremely volatile, and (c) soluble in a SF.
Extraction of analytes from the raw sample using SFs, however,
might prove useful in cases where the aqueous phase is hostile to
direct injection onto a chromatographic column or to clean up by
solid phase extraction. In addition, thermally labile materials,
irreversibly retained materials, as well as polar materials with
high water solubility which consequently do not solid phase
separate well, may be better extracted from the aqueous matrix.
Hawthorne et al.5 have reported that split SFE-GC (cryotrap = 5°
C) allows the analysis of wet samples, in contrast to on-column
SFE-GC where ice formation is a problem. In the split mode,
clepressurization of the SFE effluent occurred in the heated
injection port. SFE-GC-MS analysis of wet fuel contaminated
sediment (20 % water by weight) has been reported. The
application of this SFE-GC technique was limited to the n-octane
and higher boiling point species when wet samples were analyzed.
The extraction of aqueous samples with SF CO2 is currently
being studied in our laboratory . Preliminary findings have been
recently communicated. We describe here additional results which
employ both dynamic and equilibrium extraction modes. Concurrent
with our research efforts have been the reports of a similar
system for the extraction of cholesterol from water as well as a
segmented-flow system for on-line liquid/SFE of aqueous
matrices8.
Experimental
The extraction vessel currently in use was acquired from
Suprex Corp. (Pittsburgh, PA) and was 1 cm I.D.. x 10 cm in
111
-------
length (8 ml volume). The supercritical CO2 (Scott Specialty
Gas, Plumsteadville, PA) was delivered to the extractor and any
subsequent chromatography was done with a Suprex SFC 200 (Suprex
Corp.,Pittsburgh, PA). The equilibrium extraction system has
been previously reported6.
SFE of the aqueous matrix followed by deposition of the
analyte onto a solid phase extraction tube as well as directly
into solvent was also done. A flame ionization detector heating
block from a Suprex SFE 50 was modified so that a standard solid
phase extraction (SPE) tube (Analytichem International, Harbor
City, NJ) could be mounted on what would normally have been the
flame jet (Figure 1). Nitrogen was plumbed in through the
hydrogen inlet to provide a stream of heated gas to ensure that
the pulled restrictor tip would not ice-up and possibly foul.
The outlet restrictor was pulled so that at 150 atm a flow
of 2 ml/min of SF was sent through the extraction vessel. Unless
otherwise noted the extractions were performed with a total flow
of 30 ml of supercritical CC>2. Once the SFE was complete the
analyte was eluted from the SPE tube with 4 ml of 50/50
MeOH/Water. For collection into a solvent a slower flow was
used, 0.6 ml/min at 150 atm, and the restriction was fed directly
into a sample vial containing 5 ml of 50/50 MeOH/water.
Liquid chromatography was performed with a LC/9560 ternary
pump (Nicolet, Madison, WI) equipped with a TriDet UV 254
detector (Perkin-Elmer).
Results and Discussion
Aqueous solutions of triprolidine and pseudoephedrine were
analyzed by equilibration SFE/SFC. Both triprolidine and
pseudoephedrine were originally in the form of hydrochloride
salts, and as salts they were expected and found to be insoluble
in supercritical 003. In hopes of first forming and then
extracting the free bases, a molar excess of tetrabutylammoniura
hydroxide was added to 3 ml of 1 mg/ml solutions of both
compounds. SFE/SFC traces for both extracted compounds are shown
in Figures 2a and 2b. The chromatographic traces show that the
free bases were extractable from water with CC-2 as well as
chromatographable under supercritical conditions.
While equilibrium SFE/SFC should perform well for nonpolar
to moderately polar analytes, our primary goal was to analyze for
highly polar materials. In this regard, several problems arose
from poor elution behavior of the extracted analytes using 100 %
C02.
Trapping systems which would allow for the isolation of
analytes were therefore developed. Dynamic extraction was
performed by decompression of the SF directly into a conventional
solvent as well as being deposited onto a solid silica support as
described in the Experimental section.
112
-------
Figure 3 shows the effect of SF density upon the extraction
behavior of phenol from water. At low density phenol is seen to
be extracted quite slowly from water. After 30 ml of CC>2 were
passed through the cell only 17% of the phenol was extracted.
Much greater quantities of phenol could be extracted at higher
density using the same amount of SF (60% extraction at 150 atm
and 50° C). These numbers were obtained using deposition directly
into a conventional solvent.
The major problems associated with both forms of collection
(liquid and silica) arise from the large volume of gas which is
produced on the low pressure end of the restrictor. A compressed
flow of 1 ml/min corresponds to 100-500 ml/min of decompressed
gas, depending upon the density and nature of the SF used for the
extraction. Collection of the analyte into a liquid solvent
consequently was especially difficult. Violent bubbling of the
liquid solvent by the gas caused sporadic loss of analyte. This
was evident as the extraction efficiency for phenol out of water
(150 atm, 50° C with 30 ml of CC>2 used) was found to be 60% with
relative standard deviation of the extraction at 15% at the 400
ppm level. Even faster flow rates are desirable as they allow
for faster extractions to be run, because a greater volume of
fluid can be put through the extraction vessel in a given amount
of time.
Deposition of the analyte onto a solid support was intended
to remove the problem of loss of the analyte due to violent
bubbling solvent. Before an extraction was carried out, the
effect of two different size restrictors upon the resulting
temperature of the solid support trap were investigated. By use
of a thermocouple installed in the dead volume of the SPE tube,
it was determined that a 25 jum capillary, providing a condensed
flow of 0.5 ml/min at 350 atm, cooled the dead space to -10° to -
15° C. A 50jiro capillary provided 2 ml/min and cooled the dead
space to -30° to -35° C. The 50 /im capillary obviously allows
for a faster extraction and the increased flow also cools the
head of the trap more efficiently. It was therefore hoped that
the added cooling would allow the SPE tube to also act as a
cryogenic trap, thus providing two mechanisms for the trapping of
analytes.
The solid trapping system indeed provided a significant
improvement over collection of analyte into a solvent. At 20
ppm, phenol was found to be extractable to 80% (16 jig) with 30 ml
of supercritical CC>2 (150 atm and 50° C) . RSD of the entire
method, including the chromatography was determined to be 9%.
The extractable amount of phenol was confirmed by observing the
amount of phenol left in the raffinate (4 nq with an RSD of 20%).
The SFE/SFC of phenol from 6 M sulfuric acid with collection in
water was carried out. It represents the case of an analyte in
an aqueous environment which is hostile to many forms of clean up
cis well as analysis. A representative trace for the separation
of the extracted phenol is shown in Figure 4.
113
-------
CONCLUSION
Clearly much additional work is needed to perfect
liquid/liquid extraction. Extension to a wider range of analytes
including highly basic compounds is desirable. Reproducible and
quantitative extraction/analysis at low ppb levels should be
demonstrated. The feasibility for on-line SFE/SFC for trace
analysis should be established. However, based on research to
date the potential of the liquid/fluid extraction is encouraging.
ACKNOWLEDGEMENT
Financial support form the U.S. Environmental Protection
Agency and B.P. America is greatly appreciated.
REFERENCES
1 Hawthorne, S.B.; Kreiger, M.S.; Miller, D.J. Anal.
Chem. 60, 472-477 (1988).
2 King, J.W.; Eissler, R.L. and Friedlich, J.P. ACS
SVP. Ser. 366, 63-68 (1988).
3 Krukonis, V.J. ACS Symposium Series 366,
26-43 (1987).
4 McNally, M.E.P.; Wheeler, J.R.; J. Chromatoar.
11, 38-42 (1988).
5 Hawthorne, S.B.; Miller, D.J.; Langenfeld, J.J.,
J Chrom. Sci. 28, 2-8 (1990).
6 Taylor, L.T.; Hedrick, J.L. Anal. Chem. 61, 1986-
1989 (1989).
7 Ong, C.P.; Ong, H.M.; Li, S.F.Y.:; Lee, H.K.,
J. Microcol. Sep. 2, 69-73 (1990).
8 Thiebaut, D. J. Chromatogr. 477, 151 (1989)
FIGURES
Figure 1
Modified flame ionization heating block for the
deposition of analyte onto a solid support.
Figure 2a
Triprolidine chromatographic trace.
Figure 2b
Pseudoephedrine chromatographic trace.
Figure 3
Extraction of 95 ppm phenol as a function of the amount
of SF used for the extraction.
Figure 4
Chromatographic trace of phenol extracted from 6 M
Sulfuric acid.
114
-------
Figure 1
Ml In III U»
Figure 2
T-^_r
0-0
«; SO"C
PT.IIUI.: 340 Mr
Loop; 20|iL
CMumm OH.T4BQMO™ CH
iriUmtn CO2 Const
150 ta 490 uUmm M 3.8
<» Ul/'mht. H«M
B«l*eSo«: aw nm
Figure 2b
Figure 3
Hiirtmmirtm HCI
urti 50OO
t 29 III
OMIlbond CN
ftn9«f«liirii to 0
Pt«*M*t 340 **r
Finn > nl/nki IV«IHI M>
190 uUnlit
D J!4 no
Figure 4
0.2i
0.15
0,1
035
Wt Phenol vs Volume SF
rag Ph«ne>! cztraeind
10
20 30
mL SF
40
100 aim (I7*> -+- 128 aim (62%) -•- 150 aim
0.2B1 mff Kunol Ida.
«2 BOm plltno) In 8M H3SO*
3m« infraction
t03 •'"• 19 G
10 inlmil«t
50/9O
l nH/n
UV Ii
219 « 4IHMI
115
-------
THE USE OF SUPERCRITICAL FLUID EXTRACTION (SFE) IN THE
ANALYSIS OF ENVIRONMENTAL CONTAMINANTS ISOLATED
FROM COMPLEX MATRICES
Jennifer V. Smith, Charles R, Knipe, William S. Pipkin, and Wayne S. Miles
Hewlett-Packard Company
P.O. Box 900
Avondale, PA 19311
In this study the HP 7680 Supercritical Fluid Extractor (SFE) was used to extract urban dust
and soils under a variety of conditions. The extracted analytes were analyzed using the HP 5890A
Gas Chromatograph with a variety of detectors, including mass selective detectors (MSD), and the
results compared with those from previously used extraction techniques.
Extraction efficiencies are greatly affected by sample matrix. The interactions of environ-
mental contaminants and sample matrices, such as density, modifier type, and modifier amount, were
studied to optimize conditions for maximum recoveries.
116
-------
Introduction
Analyzing complex matrices for the presence of environmental contaminants is the primary
concern of many industrial- and environmental-testing laboratories. Current methods are often com-
plicated and time consuming, involving numerous steps. The need for alternate sample preparation
techniques eliminate the use of chlorinated solvents and reduce extended extraction times may be
met with supercritical fluid extraction.
SFE is a rapid, clean, reproducible sample preparation technique that replaces the many labor-
ijatensive steps of current sample preparation techniques and greatly reduces the potential for human
error. Advantages of SFE include decreased sample preparation time, short extraction time, reduced
smalyte decomposition, and ease of use. In addition, the supercritical extraction parameters may be
fine-tuned to maximize extraction efficiencies.
Experimental Methods
The conditions for the extraction of the the various samples are given in table I.
Table I. Supercritical Fluid Extraction Parameters
Instrument HP 7680
Sample
ID
Urban Dust fSRM 1649)
LID1
II D2
LID3
LID4
Marine Sediment fSRM 19411
MSl+Hexane
MS2 + H20
Spiked Soil
S1
C02 Density
(g/ml)
0.25
0.45
0.60
0.90
0.25
0,90
0.85
Flow
(ml/min)
1.0
1.0
1.0
1.0
1.0
1.0
4.0
Temp
(°C)
40
40
40
40
40
40
40
Time
(min)
15
15
15
15
15
15
40
For comparison, the spiked soil sample was extracted with a Soxhlet apparatus. The solvent
used was methylene chloride, and the sample was extracted by Soxhlet for eight hours.
The gas chromatographic conditions are given in table II.
Table II. Gas Chromatographic Conditions
GC: HP 5890A Series tl
Flame lonization Detector
Detector Temperature 350 °C
Irjection Port On-column, Oven Track on
Constant Pressure Mode Initial Pressure, 4.6 psi
Column 25 m, 530 p., HP-5
Oven Program 40°C (2min),10°C/min to 350°C (5min)
Injection Volume 1 jal
Data System HP 3365
IV ass Selective Detector
W ass Range 10to800amu
Transfer Line 280 °C
Data System HP 9836C (Rev 3.1.1.)
Column 50 m, 320 u, HP-5
Injection Port On-column, Oven Track on
Constant Pressure Mode Initial Pressure, 17 psi
Oven Program 70°C (0.7min), 30°C/min to 160°C (0.75), 5°C/min to 280°C (3.75)
Injection Volume 1 nl
117
-------
Results
Using several densities, the extraction results of urban dust (SRM 1649) are shown in figures 1
to 3. These analyses were performed on the GC/MS, allowing the monitoring of the polynuclear aro-
matic ions. The polynuclear aromatic compounds can be extracted at various densities, depending on
their molecular weight, as shown in figure 4, As the densities increase, greater amounts of hydrocar-
bons are also extracted from the urban dust, as illustrated in figure 5. This is a composite of the re-
sults of the analysis of supercritical extracts by GC/FID. These data indicate the importance of
density in efficient extractions.
Modifiers may be added directly to the thimble containing the CO2 or to the sample. The
effects of modifier addition may be to decrease the density at which a particular analyte is extracted.
This effect is shown in experiments with marine sediment and modified CO2. In these experiments
CO2 modified with hexane, 5% by weight, was used as the extraction fluid. Experiments with CO2
alone resulted in limited extraction of hydrocarbons or polynuclear aromatics even at high densities.
As illustrated in figure 6, densities of 0.25 g/ml with modifier had increased extraction efficiency. The
addition of water, which acts as a modifier, was added to the sample such that the sample was satu-
rated. Figure 7 shows the result, which was to inhibit the extraction of an analyte from marine
sediment.
The comparison of Soxhlet extraction and supercritical extraction of polynuclear aromatic
compounds from spiked soil is illustrated in figures 8 and 9.
Conclusion
This study indicates the feasibility of supercritical extraction in the preparation of samples.
Hydrocarbons and polynuclear aromatic compounds were extracted from urban dust, marine sedi-
ment, and spiked soil. These experiments show the importance of density, modifier type, and modi-
fier amount in the extraction of a particular analyte from a complex matrix. Further studies are in
progress to optimize the extraction conditions.
Supercritical Extraction of Urban Dust
Density 0.25 g/ml
41 minutes
Figure 1. Total ion chromatogram for sample UD1.
118
-------
Supercritical Extraction of Urban Dust
Density 0.45 g/ml
41 minutes
Figure 2. Total ion chromatogram for sample UD2.
Supercritical Extraction of Urban Dust
Density 0.90 g/ml
41 minutes
Figure 3. Total ion chromatogram for sample UD4.
119
-------
Urban Dust (N1STSRM1 649)
Density vs Amount
MW178 NIW202
Normalized Area Counts
MW 202
MW252
MW252
0.4
0.3
0.2
0.1
0.0
25
35 4 45 5
Density (g/ml)
55
Figure 4. Comparison of the amounts of five polynuclear aromatic compounds vs extraction
density.
Supercritical Extraction of Urban Dust
0.6 g/mi
44 minutes
Figure 5. Merged FID chromatograms for samples UD1, TJD2, UD3, and UD4.
120
-------
Supercritical Extraction of Marine Sediment
5% Hexane Modifier in CO?
36 minutes
Figure 6. FID chromatogram for sample MS2 plus hexane.
Supercritical Extraction of Marine Sediment
Density 0.90 g/mi
Sample Saturated with Water
41 minutes
Figure 7. FID chromatogram for sample MS2 plus H2O.
121
-------
H
o
E
H*
§
I
1
00
i
en
I
PHENflNTHRENE
FLUORRNTHENE
PYRENE
BENZQ(e)PYREHC
-BEN20(»)PYRENE
BENZO(k)HNTHRHCENE
uSnB^^—^*~«-
CHRYSENE
BENZOCbJfLUORHNTHENE
BENZO t h) FUORflNTHENE
INDENO
-------
Indirect Supercritical Fluid Extraction of Polychlorinated Dibenzo-p-Dioxins from
Rainwater and Other Aqueous Matrices
Michael J. Lawrence
R. Murray Colquhoun
Janusz B. Pawliszyn
Department of Chemistry
University of Waterloo
Waterloo, Ontario, Canada
Modified solid phase extraction (SPE) procedure for isolation of organic
contaminants present in aqueous samples is described. The organic solvents
commonly used in the desorption step of SPE methods are replaced by the use of
supercritical fluids. The water analysis procedure uses a 10 port valve to direct the
flow of aqueous sample and supercritical fluid through the adsorbent. Simple
automation of the device can be achieved through an electronic actuator and
microprocessor based controller. Water samples spiked with various chlorinated
aromatic compounds were analyzed by supercritical fluid desorption from various
adsorbents, including Tenax-GC, Octadecyl C-18, paper pulp, activated carbon,
Florisil, and Ambertite XAD-2 resin. The extraction efficiencies of the adsorbed
species from the sorbents using nitrous oxide and carbon dioxide was investigated.
Class fractionation of complex samples is possible using density programming, and/or
the stepwise use of different supercritical fluids and/or modifiers. The application of
supercritical fluids to ultra-trace analysis is possible, since all the extracted compounds
can be deposited at the front of a chromatographic column1. A novel method of
supercritical fluid delivery based on a thermal pump is described. The design is
simple and inexpensive to assemble. Its pressure stability is dependant on the type
of pressure controller used. An inexpensive comparator system gave stability of ± 2%
at 6000 PSI.
123
-------
Introduction:
Environmental contaminants such as chlorinated aromatic hydrocarbons are
accumulated and transported in the atmosphere, and then released into the
environment by precipitation. It is necessary to develop rapid, cost-effective and
accurate methods for monitoring rainwater pollutants to protect the environment and
the population.
Current methods used to analyze organic contaminants in aqueous samples
involve liquid-liquid extractions or adsorption onto sorbents followed by solvent
desorption. These methods are generally time consuming and labour intensive. Also,
they often involve the use of high purity and expensive solvents such as benzene and
dichloromethane. With growing concern over occupational health and safety, an
alternative to the use of such organic solvents is desirable. A further disadvantage
of these traditional techniques is the difficulty in automating the process.
To improve this situation, solid phase extraction (SPE) procedures have been
developed. The instrumentation associated with these techniques is based on
cartridges filled with material similar to column packings used in solid chromatographic
columns, and uses chemically modified silica2'3. The primary advantage of SPE is
the reduced consumption of high purity solvents, thereby reducing analysis cost and
the need for solvent disposal. The time required to isolate the analyte of interest is
also greatly reduced when compared to classical extraction methods. However, SPE
still uses a small amount of solvent.
In this paper we present and discuss a sample isolation approach based on
supercritical fluid extraction using carbon dioxide and nitrous oxide. The supercritical
fluid is delivered from a new thermal pump4. The extraction process has been
developed by chemical engineers5'6 and has recently attracted deserved attention by
analytical chemists1'7*10. This separation technique exploits the properties of the
compressed gas, in most cases inexpensive and non-toxic carbon dioxide, at
temperature and pressure above its critical point. The method combines both
distillation and extraction in a single process since vapour pressure and phase
separation are involved. The major features of a supercritical fluid extraction include:
rapid extraction due to low viscosity and high diffusivity compared to liquids; selectivity
of the process through density programming; and low temperature extractions of non-
volatile compounds. Supercritical fluid extraction systems can be easily automated
since the transfer of the analyte can be accomplished by flowing the mixtures in the
system and analyte separation from the fluid can be achieved by lowering the
pressure.
Analytical supercritical fluid extractions have been applied to isolate organic
contaminants, in most cases polycyclic aromatic hydrocarbons from solid matrices
such as flyash, dust, or sorbents such as Amberiite XAD resin8 or Tenax GC1. This
paper describes inexpensive instrumentation and procedures which enable rapid
isolation of pollutants from aqueous matrices.
124
-------
Experimental Methods:
Supercritical Fluid Delivery: Supercritical fluid was delivered to these experiments
by a thermal pump, based on the thermal expansion of a compressed gas4. The
pumping system consists of a supply tank for the SF gas connected to a high
pressure vessel (thermal pump) via a high pressure valve. The vessel has
independent heating and cooling systems, has a pressure transducer which monitors
internal pressure, and has a thermistor monitoring temperature. A rupture disk is an
integral part of the safety system, and will vent the system at 7000 PSI, well below the
design rating of the vessel. The extraction vessel is thermostatically controlled by an
independent heater circuit. The extraction vessel is constructed of a suitable length
of stainless steel tubing with Swagelok fittings at either end. Another such tube filled
with activated carbon serves as a trap to remove impurities from the SF stream.
Operation of the pumping system involves cooling the pump to -15° C by
flowing liquid nitrogen or carbon dioxide through a copper tube coiled around its
exterior. The supply of gas is allowed to fill the vessel and condense, increasing in
density as the temperature decreases. After the system has equilibrated at this
temperature, it is closed to the gas supply tank and heated to increase its internal
pressure. The amount of heat required to reach operating pressure is dependant on
the initial temperature of the system. Figure 1 shows the response of the pump at
various initial temperatures.
6.7S •
6.00
5.25
t. 50
3.75
3. DO
2,35
\ ,50
-32.0 -24.0 -ts'.O -8.00 .000 8.00 16.0 24.0 32.0 40.0
>EO
TEWERATURE ( DEC C I
Figure 1 Thermal pump response at different initial temperatures.
As the SF is delivered, the amount of heat applied must be increased to
maintain operating pressure. This has been accomplished in two ways. Initially, a
simple op-amp comparator circuit was constructed (Figure 2) which switched the
power to the heater on or off in response to the signal from the pressure transducer;
on if the pressure dropped below the working pressure, off if it went above. There are
also thermal shutdown and manual reset features, as well as variable heater duty
cycle. This controller produced pressure stability of + 200 PSi @ 6000 PSI. To
improve the stability of the system, a commercial 3 mode controller (Proportional
Integral Derivative - PID) (Omega Engineering) was used. This controller is infinitely
tunable and has the advantage that it can be programmed to pressure ramp,
analogous to temperature programming of a GC oven. This controller increased the
125
-------
stability to + 35 PSI @ 6000 PSI, or a deviation of less than 1%. There are two
limiting factors on the stability of the thermal pump. The dynamic conditions in the
pump during its operating cycle indicate that it can only be tuned for one set of
conditions, thus the controller is only optimally tuned for a short period of time, and
the tune is compromised for the remainder of the cycle. This problem, combined with
the large thermal mass of the 25 Ib stainless vessel cause the small oscillations in
pressure.
400 W HEATER
Figure 2 Simple comparator controller circuit.
In order to address the stability problem, as well as to increase the safety of
the design, a low volume twin tube system was constructed from two lengths of 1/2"
stainless steel pipe, fitted with valves at each end. Figure 3 schematically shows the
system. With this system, which operates on the same principle as the larger pumps,
a continuous supply of SF can be delivered. One stage of the pump is cooled, filled,
and then heated to bring it up to pressure. As it is being operated, stage two is
cooled, filled and pressurized to just below operating pressure. When stage one is
nearly empty, stage two is brought up to operating pressure, and stage one is quickly
cooled. The valves can be either check valves or active solenoid valves, operated at
the appropriate time by differential pressure or an electronic signal from a central
processor. This system is thus also easily automated.
GAS
SUPPLY
TO
CHROMATOQflAPHY
SYSTEM
Figure 3 Twin tube low volume thermal pump.
126
-------
Water Analysis System: The system to remove organic contaminants from aqueous
samples was composed of a ten port valve (Valco Instruments, Houston, Texas)
connected to the extraction vessel, an inlet for the water sample, a SF inlet, an exit
to a waste bottle and an exit for analyte collection.
MY II
UATII SMTLE
CO,
BATU
VASTI
TO CC
Figure 4 Water analysis apparatus.
In position A, water was forced under pressure from a dry nitrogen cylinder
through the adsorbent in the extraction vessel and into the waste bottle. In this step,
the organics from the water are adsorbed onto the adsorbent in the extraction vessel.
After the water has all passed through the adsorbent, the adsorbent is dried by
continuing to pass the dry nitrogen through the adsorbent while heating the extraction
vessel. Once the adsorbent has been dried, the valve is switched to position B. The
SF flows through the adsorbent in the reverse direction and desorbs the organics
adsorbed in the first step. The analyte is recovered by bubbling the exiting SF into
a via! containing about 1 mL hexane. The contents of the vial are then blown dry with
a gentle stream of nitrogen, then brought up to 100 uL volume with hexane and
analyzed by GC-ECD.
Desorption Step Recoveries: Desorption of an OCDD standard from various
adsorbents using supercritical carbon dioxide and nitrous oxide was investigated.
Cartridges made from 1/4 inch stainless steel tubing and Swagelok nuts and ferrules,
were filled with about 80 mg of the adsorbents. A 5.0 microlitre portion of an OCDD
standard (3.114 ng/uL) was injected into the centre of the adsorbent bed. The
standard compound was then extracted using the supercritical fluid at 400 atm for 1h
with a gaseous flow rate of 120 mL/min. The extracted materials were trapped by
bubbling the supercritical fluid into a vial containing about 1 mL of hexane. After
completion of the extraction, the hexane was blown to dryness using a gentle stream
127
-------
of dry nitrogen, then brought up to final volume in hexane. The sample was then
analyzed using GC-ECD to determine OCDD concentration and extraction efficiency.
Leaching Studies: A 10 ng spike of 1,2,3,4-TCDD in hexane was injected into the
centre of the adsorbent bed. A 50 ml quantity of water was passed through the
adsorbent at 3 mL/min., the adsorbent was dried and then extracted using N20 or
CO2 at 6000 psi and 50'C for 1 h. The extract was then analyzed using GC-ECD to
determine whether the water had leached any of the TCDD from the adsorbent.
In-Flow Injection: A T-union was connected to the entrance of the extractor, the
water inlet was connected to one port of the T-union and a septum was placed in the
other port of the T-union. While water was being forced through the extractor, a
syringe was used to pierce the septum and inject a 10 ul spike of the 1,2,3,4-TCDD
standard (prepared as 1 ng/uL in MeOH) into the flow. The adsorbent filled extractor
was dried and extracted to determine extraction efficiency against an external
standard using GC-ECD.
Analysis of Spiked Water: The 50 mL water sample was spiked with 10 ul of a 1
ng/uL solution of 1,2,3,4-TCDD standard in MeOH. The water was then placed in an
ultrasonic bath for 10 minutes to facilitate dissolution. The concentration of the dioxin
in the water was 0.2 parts per billion (ppb) which is 1/3 of the solubility reported by
Schler et. al10. The water was then forced through the adsorbent-filled extractor at
a flow rate of approximately 3 mLVminute, then the extractor was dried and extracted.
The extraction efficiency was determined using GC-ECD and an external standard.
GC Analysis: All of the analyses were done on a Varian 3400 gas chromatograph
equipped with an electron capture detector (ECD), a septum programmable injector
(SPI), and a 30m X 25um DB-5 capillary column ( J&W Scientific, California). The
analyses were carried out using an external standard method. Each time a spike of
an adsorbent or water was done, an identical aliquot of the standard was injected into
a vial, evaporated to dryness and diluted to 100 uL A 0.5 uL injection of this standard
is then analyzed , using the peak area of the compound in the standard as 100%
recovery.
Fractionation Studies: The Tenax filled extractor was spiked with a mixture
containing tetrachlorobenzene, hexachlorobenzene, tricloro-PCB, pentachloro-PCB,
1,2,3,4-TCDD and OCDD. Extractions were attempted with CO2 and N2O at various
pressures to determine whether class-selective fractionations could be done.
Discussion and Results:
The development of a method for extraction of polychlorinated dibenzo-p-
dioxins from water proceeded in a series of steps. Initially, the ability of supercritical
N2O and CO2 to desorb dioxins from a number of different adsorbents was
128
-------
investigated. An OCDD standard was chosen as the standard to be spiked onto the
adsorbents because it is expected to be the most difficult to extract due to its low
solubility. It was assumed that if this particular dioxin isomer could be desorbed then
all of the others should also be able to be desorbed. The extractions were performed
at 400 atm and 50*C and the extract was collected for 1 h. Table I illustrates the
results achieved.
These results indicate that N2O is a more efficient supercritical fluid for the
extraction of dioxins from these adsorbents. A strong matrix effect is also evident,
since the OCDD was desorbed well from Tenax GC, the octadecyl phase and pulp but
very poorly desorbed from activated charcoal, florisil and Amberiite XAD-2, Obviously,
the OCDD is much more strongly bound to the latter group of adsorbents.
Table 1 Extraction of OCDD
Adsorbent
Tenax GC
Octadecyl (Cie)
Activated Charcoal
Florisil
Amberlite-XAD2
Pulp
Carbon Dioxide
% Recovery^ 10%
35
92
ND
ND
ND
80
Nitrous Oxide
% Recovery +. 10%
95
100
ND
10
ND
92
ND-none detected
-extractions done for Ihr.
at SOOOpsi. and 50*C
Based on these results, continuing experiments were limited to Tenax-GC and
octadecyl extracted with N2O and pulp extracted with CO2. Pulp is not a common
adsorbent, but in other experiments conducted in the laboratory, it was noted that
dioxins were easily desorbed from pulp. Therefore, pulp was considered for use as
an adsorbent.
The second step of the research was to determine which adsorbents would
adsorb dioxins from water. In these experiments, a 1,2,3,4-TCDD standard was used
instead of the OCDD. The TCDD standard is much more water soluble than OCDD
(more than 1000 times more soluble)10 and therefore it was much easier to extract
a detectable quantity. Initially, water leaching of the TCDD from the adsorbents was
examined. A direct spike on the adsorbent was followed by passing water through the
adsorbent. The adsorbent was then extracted to determine if any leaching of the
TCDD had occurred. Three replicate experiments were done and deviations were
calculated as shown in Table II.
129
-------
Table H Extraction of 1,2,3,4 TCDD
Adsorbent
Tenax GC (N2O)
Octadecyl (N2O)
Pulp (C02)
% Recovery
100+4%
100±6%
87+ 3%
*extractions done at 6000 psi and 50*C for 1 h.
The excellent recoveries from Tenax-GC and octadecyl indicate that leaching is not
occurring. Less than 100% recovery from the pulp is likely due to incomplete
desorption rather than leaching.
In the second stage, the TCDD was spiked in the flow of water as it passed into
the adsorbent bed. In this way, better simulation of the dioxin actually being dissolved
in solution was hoped to be achieved. Three replicate runs of each adsorbent were
done.
Table HI Extraction of 1,2,3,4 TCDD
Adsorbent
Tenax(N2O)
Octadecyl(N2O)
Pulp(C02)
% Recovery
80+5%
85+ 10%
76+ 3%
extractions done at 6000 psi and 50 C for 1 h.
The results in Table III show lower recoveries than obtained with direct spiking onto
the adsorbents but this is a result of the dioxins never actually contacting the
adsorbent. Due to its low solubility in water, the TCDD likely precipitates out, to some
extent, at the point where it is injected into the flow of water and therefore is never
able to be adsorbed and subsequently extracted. Dioxins that are dissolved in water
should not encounter this problem. The final stage of this step was to actually
dissolve the 1,2,3,4-TCDD in water and put this spiked water through the extraction
procedure. The water was spiked at 0.2 ppb which is about 1/3 its reported solubility
level10. This was done to ensure that the TCDD would not precipitate out as in the
in-flow injection of the standard. Again three replicate experiments were done with
each adsorbent, and deviations were calculated.
130
-------
Table IV Extraction of 1,2,3,4-TCDD
Adsorbent
Tenax(N2O)
Octadecyl(N2O)
Pulp(CO2)
% Recovery
67+ 20%
68± 30%
55+ 20%
* extractions done at 6000 psi and 50TT
for 1 h.
The results shown in Table IV illustrate significantly lower recoveries than had
been achieved in other experiments. This lowering of extraction recoveries indicates
a loss of analyte at some point in the system. The possibility of break-through of the
TCDD in the water was investigated by re-running the same water through the system.
No more TCDD was recovered so break-through is not likely. It is likely though that
the TCDD is adsorbed and immobilized onto the glassware used as a water reservoir
and thus was never able to reach the adsorbent. Three runs were done with each
adsorbent to determine the deviation in recoveries. The water reservoir bottle was
rinsed with MeOH and hexane between runs to remove any TCDD adsorbed on the
glass and prevent cross-contamination. One method for compensating for the loss
of analyte adsorbed onto the glassware would be the use of a 13C labelled internal
standard. It could be assumed that the 13C labelled TCDD would be adsorbed to the
same extent as the unlabelled TCDD. Using GC/MS one could quantitatively
compensate for this adsorption effect.
A larger deviation (30%) was observed in the runs using octadecyl bonded to
nnicroporous silica as the adsorbent. Excellent recoveries were achieved with the first
extraction on fresh octadecyl but the recoveries dropped as the same adsorbent was
used repetitively. Stripping of the octadecyl phase by the supercritical fluid is likely
occurring, leaving behind a bare silica surface which binds strongly to the dioxin and
causes poorer recoveries.
The ability of supercritical fluids to do class-selective extractions or fractionate
samples was investigated. A mixture of chlorinated benzenes, polychiorinated
biphenyls (PCB's), and dioxins was injected directly onto the Tenax GC adsorbent in
an extractor. Extractions were done with N2O and CO2 at various pressures and the
extract was taken at various time intervals to try to separate these three classes of
compounds. It was found that N20 desorbs all of the compounds within 20 minutes
and very tittle separation of the compounds could be achieved. Since CO2 was found
to desorb dioxins poorly from Tenax, it was tested to see if it could be used to
fractionate the mixture (Table V). At 5000 psi and 50'C, 74% of the pentachloro-PCB
and all of the trichloro-PCB and chlorinated benzenes were removed within 20 min.
with loss of 23% of the TCDD and no loss of OCDD. At SOOOpsi and 50*C, 60% of
the pentachloro-PCB and all of the trichloro-PCB's and chlorinated benzenes were
removed after 40 min. with only a 5% loss of the TCDD and no loss of the OCDD.
131
-------
Table V Class fractionation study.
Extract.
Time
20 min.
40 min.
Extract.
Pressure
3000 psi
5000 psi
3CI-PCB
Removal
100%
100%
5CI-PCB
Removal
60%
74%
TCDD
Loss
5%
23%
OCDD
Loss
0%
0%
It appears that better separation is achieved at lower pressure but the analysis time
is greatly increased. Once the interferences are removed, N2O can be used to desorb
the dioxins.
Conclusions:
Solid phase extraction using supercritical fluid desorption is a good alternative
for isolation of organic contaminants such as polychlorinated dibenzo-p-dioxins from
aqueous samples. This process can be controlled by a single six port valve, though
we used a ten port device, and can thus be easily automated. The combination of this
method with direct on-column deposition of analytes by decompression of the
supercritical fluid9 will facilitate ultra-trace analysis. In this procedure, all extracted
material is injected into the column, rather than only a small fraction as is common
when performing solvent extractions. Class fractionation of complex samplos is
possible using density programming and step-wise use of different supercritical fluids
and/or modifiers. More work is needed to optimize this process. Tenax-GC requires
the use of nitrous oxide as the supercritical fluid to obtain good recoveries.
Quantitative extractions of dioxtn can, however, be ensured with carbon dioxide if
chemically modified silica or paper pulp are used as the adsorbent.
132
-------
References:
1. S. Hawthorn, D. Miller, "Extraction and recovery of organic pollutants from
Environmental solids and Tenax-GC using supercritical CO2", J. Chrom. Sci.. 24:258
(1986).
2. M. Zief, R. Kiser, "Solid phase extraction analysis of biological samples
", American Laboratory. JAN: 70 (1990).
3. G. A. Junk, J. Richard, "Organics in water: solid phase extraction on a small
scale ", Anal. Chem.. 60: 451 (1988).
4. Janusz Pawliszyn, "Inexpensive fluid delivery system for supercritical fluid
extraction." Journal of High Resolution Chromatoaraphv. 13:199 (1990).
5. G. Wilke, "Extraction with supercritical gasses", Anaew. Chem.. Int. Ed. Enal..
17:701 (1978).
6. M. McHugh, V. Krukonis, Supercritical fluid extractions: Principles and practice.
Butterworths, Stoneham MA, (1986).
7. K. Sugiyama, M. Saito, T. Hondo, M. Senda, " Directly coupled labscale
supercritical fluid extraction-supercritical fluid chromatography with a multiwavelength
ultraviolet detector", J.Chromat.. 332:107-116 (1985).
8. B. Wright, C. Wright, R.Gale, R. Smith, "Analytical supercritical fluid extraction
of adsorbent materials", Anal. Chem..59: 38 (1987).
9. S. Hawthorn, D. Miller, "Directly coupled supercritical fluid extraction gas
chromatography of polycyclic aromatic hydrocarbons from environmental solids", .L
Chromat.. 403: 63 (1987).
10. Schler, Doucett, Gobas, Andren, Mackay, Environ. Sci. Tech.. 22: 651-656
(1982).
133
-------
MONITORING OF NITROGEN POLLUTANTS
AT MT. MITCHELL, NORTH CAROLINA.
Anuradha Murthy, Viney P. Aneja, and R. Bradow
Department of Marine, Earth and Atmospheric Sciences,
North Carolina State University,
Raleigh, North Carolina 27695-8208.
ABSTRACT
Gaseous nitrogen compounds (nitric acid, nitrous acid, ammonia, and nitrogen
dioxide), and paniculate nitrogen compounds (nitrate and ammonium) were measured
in ambient air using Annular Denuder System (ADS) and chemilumincescence
nitrogen oxides analyzer at ML Mitchell State Park (~2006m MSL), North Carolina.
Measurements were made during May through August of 1988 and 1989. Further,
measurements of nitrate and ammonium in cloud water samples were also made during
the same period Concentrations of gaseous nitric acid using ADS were found to be in
the range 0.13-5.62 M-g/m3 with a mean of 1.07 M-g/m3 during 1988, and in the range
0.55-2.60 jig/m3 with a mean of 1.39 ng/m3 during 1989. Nitric acid levels were
found to be higher during the daytime compared to the nighttime levels, which are
consistent with the photochemical formation of gaseous nitric acid. Gaseous ammonia
levels were found to be in the range of 0.01-4.98 |ig/m3. Paniculate nitrate,
ammonium and sulfate concentrations were found to be have range of 0.02-
0.21iig/m3,0.01-4.72 yg/m3 and 0.21-18.13 ng/m3 for 1988, and 0.1-0.78 ng/m3,
0.24-2.32 jig/m3 and 0.5-11.8 Jig/ai3 for 1989 respectively. The fine aerosol fraction
was dominated by sulfate particles followed by ammonium particles. Mean nitrate and
ammonium ion levels in cloud water samples for 1988-1989 ranged from 142-192
(jmol/1 and 153-185 (imol/1 respectively. The incorporation of the pre-existing
gaseous nitric acid/nitrate into the cloud is not efficient to account for the measured
nitrate concentrations in die cloud water.
134
-------
INTRODUCTION
Air pollution is thought to be one of the contributing factors for the high
elevation forest decline in the eastern United States (Cowling, 1985; Woodman and
Cowling, 1987). High amounts of acidic deposition and elevated levels of gaseous
atmospheric photochemical oxidant pollutants (e.g., ozone, hydrogen peroxide) have
been observed at Mt Mitchell State Park, (Mt Mitchell, elevation -2038m MSL),
North Carolina. Mt Mitchell is the highest peak east of the Mississippi river. Ozone,
a photochemically formed secondary pollutant has been observed at levels exceeding
the National Ambient Air Quality Standards (NAAQS) (Ancja et al., 1990); and has
frequently been >50 ppbv during the growing season (May through September) which
could cause damage to plants (Mohnen and Cowling, 1988). Significant quantities
of acidic substances are also found to deposit on the canopy which may contribute to
the observed decline in the ecosystem.
Greater than normal atmospheric deposition of nitrogen compounds is thought
to be one of the likely mechanisms by which airborne pollutant chemicals could cause
injury to spruce-fir forests. According to this hypothesis (Nihlgard, 1985), nitrogen
containing pollutants deposited on foliage and soils could cause leaf damage; and also
may extend the period of growth later into the fall, thus inhibiting the "hardening off1
processes by which needles protect themselves from frost injury during the severe
winter weather (Friedland 1984).
Since this mechanism involves the mineral nutrition of growing plants, and,
specifically, the availability of nitrogen at a specific time in the growing season, it is
important to establish the extent to which nitrogen is supplied by atmospheric
exposure throughout the growing season. Thus, it is not clear whether general
overfertUization or time-specific fertilization is important, or that there is, in fact, a
seasonal factor in the exposure/deposition of nitrogenous air pollutants. Hence,
characterization of the atmosphere for nitrogen containing species and their chemistry
becomes important in exploring the forest decline phenomenon.
EXPERIMENTAL TECHNIQUES
Gas- phase measurements: Ambient gaseous nitric acid, gaseous nitrous acid,
gaseous ammonia, and paniculate nitrate and ammonium measurements were made
during the summers (May through August) of 1988 and 1989 utilizing the Annular
Denuder System (ADS) technique (Possanzini et al., 1983). Continuous gas phase
measurements of nitrogen oxides (NO and NO2) were made using a
chemiluminescence nitrogen oxides analyzer. The annular denuder system used at Mt
Mitchell consisted of an impactor preseparator to remove coarse aerosol fraction (>2.5
|im), three annular denuder tubes in series to collect the gaseous species and a filter
pack to collect the aerosol fraction. The first denuder was coated with 0.1 % solution
of NaCl to collect gaseous nitric acid to remove gaseous nitric acid from the other
acidic species. The second denuder was coated with a solution of 1% glycerine and
1% NazCOs solution in a 1% mixture of methanol and distilled water to capture
nitrous acid and sulfur-dioxide. The third denuder was coated with 1% citric acid in
methanol for the collection of ammonia. The filter pack consisted of a teflon filter for
the collection of fine particulates and a nylon filter to capture any volatalized nitric acid
during the dissociation of ammonium nitrate aerosol. The denuder system was
operated in a laminar mode at a constant flow rate of 16.5 1/m. Though the
135
-------
measurements were primarily made for 24 hours duration, short time measurements
(12 and 4 hours) were also made to study the diurnal variability and trends in the
nitrogen compounds.
Cloud water measurements: Cloud water samples were collected on an hourly
basis using an ASRC passive cloud collector (Kadlaeck et al., 1983). The collected
samples were analyzed for pH immediately upon collection at the site. The samples
were later analyzed for pH, conductivity, major anions and cations. Liquid water
content measurements based on gravimetric technique (Valente, 1988) were also made
simultaneously during cloud collection.
RESULTS
Ambient nitrogen pollutant concentrations: The observed nitric acid
concentrations during the 1988 monitoring season ranged from 0.13-5.62 |ig/m3 and
the mean concentration was 1.07 jig/m3. The mean concentration of nitric acid for
1989 sampling season was 1.39 Hg/m3 while the range of values observed was
between 0.55-2.60 p.g/m3. The concentrations of nitric acid observed at the Mt
Mitchell site are comparable to the measurements at other remote sites (Cadle et al.,
1985; Shaw et al., 1982; Harrison and Allen, 1990). The gaseous ammonia varied
from 0.01-4.98 ug/m3 for 1988, and 0.2-3.23 ng/m3 for 1989, while the mean values
were 0.54 ng/m^and 1.47 Hg/m3 respectively. Gaseous nitric acid concentrations
were in general higher in 1989. A diurnal variation was noted in the concentrations of
nitric acid from the short time interval denuder measurements. Eight measurements
each of 4 hours of duration were made during one day. The observed concentrations
were found to peak during the afternoon hours (1200-1600). The daytime
concentrations were significantly higher than the nighttime levels (2.5 |ig/m3
compared to 1.3 M-g/m3 in 1989; and 1.21 compared to 0.8 jig/ni3 in 1988
respectively). The observations are consistent with the photochemical pathway for the
production of nitric acid during the daytime.
The precision associated with the ADS measurements was observed to be
good. Collocated sampling runs produced similar concentrations (1.27 & 1. 16jig/m3
for HNO3). The laboratory and field blanks suggested no source of contamination.
The average concentrations of nitrogen dioxide were 2.5 ppb and 4.0 ppb for
the 1988 and 1989 sampling seasons respectively. The levels of nitric oxide (NO)
were always very low (usually less than the detection limit of the instrument, which is
2 ppb). The concentrations of nitrogen di-oxide were also consistently higher in 1989
than the 1988 values. x
The mean concentrations of paniculate ammonium were 1.4 Ug/m3 and 0.92
p.g/m3 for the 1988 and 1989 sampling periods respectively. Mean sulfate paniculate
concentrations were 4.35 ng/m3 and 4.33 M£/m3 for the 1988 and 1989 monitoring
periods respectively. Nitrate concentrations (0. 11 and 0.22 (ig/m3 for 1988 and 1989
respectively) were much lower compared to sulfate and ammonium The aerosol
fraction was thus dominated by sulfate particles followed by ammonium particulates,
Cloud Water Concentrations: The mean concentrations of nitrate in the cloud
water samples were 195 (iequ/1 in 1988, and 142 fiequ/l in 1989. The ammonium ion
concentrations were 185 and 153 p.equ/1 for the 1988 and 1989 field seasons
respectively. The ionic concentrations in the cloud water were lower in 1989
compared to the previous years. Nitrate and ammonium contributed ~13% and -15%
respectively to the total ionic concentration in the cloud water.
136
-------
DISCUSSION
The concentrations of nitric acid measured during the course of one day in
1988 and 1989 were compared with the corresponding meteorological and ozone data
to determine the generation and removal mechanisms for gaseous nitric acid. A
multiple regression of HNOs concentrations on temperature, relative humidity, solar
radiation and ozone of the 1988 data suggested that HNOs concentrations observed
were related to the above variables with an r2 of 0.9. Similar analysis for the 1989
data showed a good correlation of r^=0.97. Solar radiation alone explained "75% of
the variation in the nitric acid production in 1988 and ~50% variability in 1989.
Relative humidity and temperature independently explained 74% and 65% of the
variation in the nitric acid concentrations respectively for the 1988 measurements. A
weak positive relationship between HNO3 and ozone (r2 = 0.27) was also observed.
The correlation between the solar radiation and nitric acid suggests that the
major pathway for the formation of nitric acid at Mt, Mitchell is oxidation of nitrogen
di-oxide via OH radicals. Considerable levels of gaseous nitric acid were observed
even during nights, though the concentrations reached a maximum during midday.
The production of nitric acid during the daytime is believed to proceed through the
reaction.
NO2+ OH -» HNO3.
The oxidation of NO2 by OH radicals does not occur during the night (no production
of OH radicals). To account for the levels observed during night, an oxidation
mechanism by 63 may be present at Mt Mitchell, since ozone levels are higher during
the night than during the day at Mt Mitchell.
NOz + O3->NO3 + O2.
NOs can combine with the existing NO2 to produce ^05. Under high relative
humidity conditions, N2Os combines with water vapor in the atmosphere to generate
nitric acid.
The Henry's law constant for gaseous nitric acid is 2*105 moles I'1 atnr1 at
25®c (Schwartz et al., 1981). Using an average ambient concentration of 0.5 ppbv for
gaseous nitric acid, the nitrate concentration in cloud water was analytically obtained
to be -130 jimolesAiter. The measured average nitrate concentration in the cloud water
samples were 195 (imoles/I and 141 ^moles/1 during 1988 and 1989 which are
higher than the Henry's law derived value. This seems to suggest that the
incorporation of the pre-existing gaseous nitric acid/nitrate into the cloud accounts for
most of measured nitrate concentrations in the cloud water. Though the pathways for
the excess nitrate in the cloud water are not clear, Oxidation of the dissolved nitrogen
dioxide by dissolved hydrogen peroxide or ozone seems a possible mechanism at Mt.
Mitchell. Sharp decreases in pH and ionic concentrations during the leading edge of
an orographic cloud event may now be attributed primarily to initial droplet scavenging
of gaseous nitric acid and deliquescence on pre-existing aerosol. Water vapor
condensation and evaporation of the droplets have been suggested as important
mechanisms in the leading and trailing edges respectively of an orographic cloud
event Thus, this role of HNO? may explain the concave trend (decreasing at the
beginning and rising toward the end) observed in the concentration of the ions
measured during an event (Kim & Aneja, 1990).
137
-------
The heterogeneous removal of HNOa appears to have three possibilities:
1. the incorporation of nitric acid on to aerosols
2. the incorporation of nitric acid into precipitation/cloud droplets
3. dry deposition of nitric acid on surfaces.
According to our data, the incorporation of nitric acid on to aerosols does not seem to
be present because of the following observations: (a) The concentrations of gaseous
nitric acid were larger than the paniculate nitrate levels by an order of magnitude; (b)
The correlation between paniculate nitrate and gaseous nitric acid were very low; (c)
A diurnal variation, similar to that for gaseous HNOj was not observed for nitrate
participates. A rapid removal of nitrate particles must be occurring if the conversion of
HNO3 to nitrate particulates was indeed present Such a short lifetime of an aerosol
species seems unreasonable, because the deposition velocities are very small compared
to the deposition velocity of the gas (1*1Q-3 cm/s for particles compared to 2.5 cm/s
for gaseous nitric acid), and the scavenging of nitrate particles in the
cloud/precipitation should be less than that for gaseous HNOs scavenging under
similar conditions. Having a high deposition velocity (2-4 cm/s), dry deposition of
nitric acid at the surfaces seems to be a likely removal mechanism in the absence of
cloud/precipitation pathway.
Acknowledgement
This research has been funded through a cooperative agreement with the U.S.
Environmental Protection Agency (8113934-91-2) as part of the Mountain Cloud
Chemistry Program.
Disclaimer
The contents of this document do not necessarily reflect the view and policies of
the Environmental Protection Agency, not the views of all members of the Mountain
Cloud Chemistry consortia, nor does mention of trade names or commercial or non-
commerical products constitute endorsement or recommendation for use.
138
-------
REFERENCES
E. B. Cowling, "Critical review discussion papers - effects of air pollution on
forests", J. Air Pollut. Control Association. (36): 916 (1985).
S. E. Schwartz, and W. H. White, "solubility equilibria of the nitrogen oxides and
oxyacids in dilute aqueous solution". Adv. Environ. Sci. Eng., (4), 1 (1981).
R. M. Harrison, and A. G. Allen, "Measurements of atmospheric HNO3, HC1, and
associated species on a small network in Eastern England, Atmos. Environment
(24A), 2, 369, (1990).
V. P. Aneja, C. S. Claiborn, Z. Li, and A. Murthy, "Exceedances of the National
Ambient Air Quality Standard for ozone occurring at a pristine area site", J. Air and
Waste Management Association. (40): 217, (1990).
D. Kim, and V. P. Aneja, "chemical composition of clouds at Mt. Mitchell, North
Carolina", for presentation at the 18th Australasian Chemical Engineering Conference,
Auckland, New Zealand, August 27-30, (1990).
J. N. Woodman and E. B. Cowling, "Airborne chemicals and forest health", Env.
Sci. & Tech. (21): 120(1987).
B. Nihlgard, "The ammonium hypothesis-an additional explanation to the forest
dieback in Europe", AMBJQ, (14): 2 (1985).
Friedland. Canadian Journal of Forest Research. (14V 963 (1984).
J. Kadlecek, S. McLaren, N. Camarota, V. A. Mohnen, and J. Wilson, "Cloud water
chemistry at Whitcface Mountain" In: PnaHpitaHflflfl y^ygngi"g dry deposition and
resuspension. H. R. Pruppacher et al., eds. Newyork, 103 (1983).
R. J. Valente, "Development of field implementation of a new instrument for
gravimetric measurement of cloud Liquid Water Content," J. Atmos. Oceanic Tech.
(1988).
M. Possanzini, A. Febo, and A. Liberti, "New design of a high performance denuder
for the sampling of atmospheric pollutants", Atmos Environ. (17): 2605 (1983).
R. Shaw, R. K. Stevens, and J. Bowennaster, "Measurements of atmospheric nitrate
and nitric acid: the denuder difference experiment," Atmos. Environ. (16): 845
(1982).
S. H. Cadle, "Seasonal variations in nitric acid, nitrate, strong aerosol acidity and
ammonia in an urban area," Atmos. Environ. (19): 181 (1985).
139
-------
PARAMETERIZATION OF IN-CLOUD SCAVENGING OF SULFATES AND
NITRATES IN ACIDIC DEPOSITION MODELS
N.-H. Lin and V. K. Saxena
Department of Marine, Earth and Atmospheric Sciences
North Carolina State University
Raleigh, NC 27695-8208, U.S.A.
ABSTRACT
Scavenging of sulfates and nitrates — two most common ions leading the cloudwater
acidity - was investigated during field studies atop a site in Mt. Mitchell (35°44'05"N,
82°1T15"W) State Park where the highest peak (2,038 m MSL) of the eastern U.S. is
located. Experiments were conducted during the growing seasons (May 15 - September 30)
of 1986 and 1987 using an instrumented meteorological tower (16.5 m tall) and a passive
cloudwater collector. Clouds were frequently observed in which the Fraser fir and red spruce
stands stayed immersed 28 % and 41% of the time during the 1986 and 1987 seasons
respectively. Rate of cloudwater deposition on the forest canopy was determined using an
inferential cloud deposition model. It was found by analyzing 9 short duration (lasting 8 h or
less) and 16 long duration cloud events that the ionic concentration (SC>42- and NCV) is
inversely proportional to the rate (Ic) of cloudwater deposition (in mm h"1) and can be
expressed by the following relationship: [SC>42-] = a lc~b or [NCV] = a ICA The b values for
predicting SO42' concentration were found in the range of 0.14 - 1.24 for short duration and
0.062 - 0.63 for long duration cloud events respectively. The corresponding b values for
predicting NCV concentrations were 0.19 - 1.16 and 0.072 - 0.59 respectively. When b
parameter was between 0.2 - 0.6, the correlation coefficients between measured and predicted
ionic concentrations were found to exceed 0.7. The ratio of a parameter for SC>42- to NCV
varied between 1.75 - 6.95 indicating that the SC>42' contributes to the total ionic concentration
substantially more than the NOs'. The above parameterization is similar to the one that is
frequently used to relate ionic concentration in precipitation to the rainfall rate. Assuming that
the SC>42' in cloudwater is the result of in-cloud scavenging of SO42' aerosols in the cloud
forming airmass, a functional dependence on the cloudbase temperature (Tct») was found:
[SCU2'] = C exp[TCb/273] where C is a constant. Existing data from Piseco Lake (NY),
Kilauea (HI), South Pole and McMurdo (Antarctica) have already shown conformity to the
above relationship.
140
-------
1. INTRODUCTION
Upon formation, clouds are very efficient scavengers of air pollutants if they lead to
precipitation subsequently, while upon dissipation, in the absence of precipitation, they
efficiently transform and redistribute the air pollutants in the planetary boundary layer. With
regard to the interaction between pollutants and clouds, Scott1 developed a mathematical
model for which the SO42' concentration in precipitation is predicted to be directly
proportional to the SC>42' concentration of air ingested into cloud and inversely proportional to
the liquid water content of cloud. Based on Scott's hypothesis, Hogan2 interpreted the
precipitated concentration of SO42~ ion as a function of cloudbase temperature with a natural
logarithmic relationship. Furthermore, de Pena et al? investigated 10 precipitation events and
showed the dependency of [SO42-] upon the precipitation intensity.
A study of cloud chemistry at Mt. Mitchell (35°44'05"N, 82°17'15"W; 2,038 m MSL),
North Carolina, began in May, 1986. The current investigation are primarily focused on
characterizing the chemical features of mountain clouds4 and assessing the relative
contribution to acidic deposition through wet, dry and direct cloud capture mechanisms5.
Based on the database obtained in 1986 and 1987, we present in this paper the representative
cases of 9 short (lasting < 8 h) and 16 long duration cloud episodes by analyzing the
micrometeorology, cloudwater chemistry and deposition flux over the cloud evolution period.
Furthur, we present a scheme of parameterization of in-cloud scavenging of SO^- and NQj-,
based on the earlier studies1-2-3 and an evidence for relating the SC>42- concentration in
precipitation (direct or occult) to the cloudbase temperature is investigated.
2. EXPERIMENTAL
At the Mt. Mitchell site, a 16.5 m tall tower was fully instrumented with meteorological
sensors. The cloud water was hourly collected during the cloud episodes with an ASRC
(designed by Atmospheric Science Research Center, State University of New York at Albany)
collector placed atop the tower. The cloud episode was signaled when a stationary object at a
distance of 1 km from the observation point became obscured by clouds and stayed
consistently off the view for more than 15 min. The meteorological and micrometeorological
parameters were measured. All the details of the experimental setup have been given
elsewhere5.
The field observations began each year in May and ended in the middle of October, thus
covering the duration of the growing season at the site. The observed cloud episodes are
categorized into two classes: long ones exceeding 8 h, and short ones with duration less than 8
h. The former are found generally attributed to meso-scale or synoptic-scale disturbances and
the latter are primarily the result of orographic lifting mechanisms. The cloud immersion was
found4-5 to be 28% and 41% for 1986 and 1987 field seasons, respectively.
During the individual cloud episode, the rates of cloud water deposition were calculated for
each hour with an inferential cloud deposition model which was first proposed by Lovett^and
further modified by Mueller and Weatherford7. This model is able to simulate the uptake of
cloud water by the tree components (boles, branches and needles) within the forest canopy.
3. A PROPOSED METHOD FOR PREDICTING CONCENTRATIONS OF
PRINCIPAL ANIONS: PARAMETERIZATION SCHEME FOR ACIDIC
DEPOSITION MODELS
The SC»42- concentration in precipitation is a consequence of several cumulative processes
occurring within and below clouds. Nucleation scavenging , Brownian motion, phoretic
attachment, and inertial impaction primarily play a role in removing the SO42* aerosols from
air and attach them to the cloud and precipitation elements. In addition, SO42' may be
generated within the cloud and precipitation water through the oxidation of gaseous SO2.
Scott1 developed a mathematical model for calculation of SO42' aerosol removal by
precipitation. Scott1 developed a mathematical model for calculation of SO42* aerosol removal
141
-------
by precipitation. The SC>42- concentration in precipitation is predicted to be directly
proportional to the SC>42' concentration of air ingested into cloud and inversely proportional to
the cloud water concentration. The airborne SC>42- scavenged by precipitation is strongly
dependent upon precipitation formation mechanism in a certain portion of clouds. This can be
mathematically expressed1 as
14000Afs(0) 750(1-4.41x10-V°-88)
r _o.88 (1.56+0.44 InR)
5°« , (1)
where/is a washout rate defined as the ratio of the SC>42- concentration in precipitation to that
in air, 50 is the SO42' concentration of air just below the cloud base, My(0) is the total SCU2"
concentration of air contained in all hydrometers falling past a fixed level at the top of the
riming zone, and ^? is the precipitation intensity. The model assumes that the in-cloud
oxidation of SOx is negligible and the SO42- deposited upon the ground represents the
scavenging of pre-existing aerosols during the lifetimes of individual cloud elements.
Based on the Scott's hypothesis, Hogan2 derived a formula to relate the SCU2' concentration
in precipitation with the cloudbase temperature as expressed by
1 , (2)
where Tcb is cloudbase temperature and C is a constant. He used this formula to correlate the
earlier observations obtained in different cloud types (warm or cold clouds) and in different
locations (Piseco Lake, New York State; South Pole, Antarctica; Kilauea, Hawaii), as shown
in Fig. 1 with the dash-boxes. The SC>42' concentrations were found to be reasonably
predictable by the cloudbase temperatures. Here, we use our database and the observations
of Saxena et a/.8 at McMurdo, Antarctica, which are listed in Table 1 to fit Eq. (2). The
former were based on the ground measurements and the latter were obtained during airborne
investigations. The results are shown in Fig. 1 with the solid-boxes. The temperature ranges
of the data of McMurdo and Mt. Mitchell are similar to those of Piseco and Kilauea,
respectively. The minimum values are found to be in good agreement with Hogan's
prediction, but the averages for the results of both McMurdo and Mt. Mitchell are about three
times that of Hogan's prediction. As shown in Fig. 1, the slope of our predicted semi-
logarithmic line is approximately equal to that of Hogan's predicted line. The above method
could be used to parameterize acidic deposition models and deserves further careful
investigation under varying conditions of cloud forming airmasses, especially for the warm
clouds such as the ones observed at Mt. Mitchell.
Eq. (1) for predicting the SO42" concentration in precipitation can be further simplified as
[S04-2] = a I/, (3)
where L, represents the precipitation intensity (same as R) in mm h"1 which can be directly
obtained from the precipitation collectors, and a and b are constants highly dependent on the
cloud type3. Nitrate can be predicted with the same expression as Eq. (3) as well. It was
found by de Pena et al? that b ranged from 0.34 to 1.71 (mean=0.47) and from 0.03 to 1.19
(mean=0.68) for SO42~ and NOa', respectively, for 10 precipitation events.
We found that the cloud droplet capture mechanism essentially behaves somewhat similar to
the precipitation scavenging. Thus, Eq. (3) can be implemented to predict the SO42'
concentration in cloud water by replacing the precipitation intensity with the rate of cloud
water deposition (Ic) as has been computed for each cloud event with the cloud deposition
model6-7. Eq. (3) then becomes
142
-------
alc- , (4)
where Ic is the rate of cloud water deposition.
Selecting 9 short and 16 long cloud events with complete measurements for related
parameters to fit Eq. (4), the results are listed in Table 2 in which r is the correlation
coefficient between the measured and predicted SCU2' concentrations. Nitrate is also
predicted with the same equation. The results show excellent correlation for most short cloud
events with r greater than 0.75, but somewhat inconsistent in some cases with very low r for
long cloud events. The b values for predicting SC>42 are in the range of 0.14 - 1.24 and
0.062 - 0.63 for short and long cloud events, respectively, whereas for NCV, these are 0.19 -
1.16 and 0.072 - 0.59, When b falls between 0.2 - 0.6, the correlation coefficients between
measured and predicted concentrations are mostly above 0.7 level. The average of b is well
compared with the value of 0.3 suggested by Scott1 and de Pena et a/.3.
4. CONCLUSIONS
By analyzing 25 cloud events, our results support the previous findings for estimating SC>42-
concentration in precipitation with the knowledge of related meteorological parameters.
According to Scott's1 hypothesis and Hogan's2 inferred relationship, SO42- concentration in
the cloud water samples can be related to the average cloudbase temperature with a natural
logarithm relationship, but our results are higher by a factor of 3 when compared to the
Hogan's predictions. Scott1 proposed a scheme for the parameterization of in-cloud
scavenging of the SC»42' aerosol for which the concentration can be related to R~a, where R is
the precipitation intensity in mm tr1 (similar to Ic) and a is a constant dependent on cloud
types and the precipitation formation mechanism, and is suggested to be 0.3. Sulfate and
NOs' concentrations in cloud water were found to be predictable by the inverse relation with
the rate of cloud water deposition (Ic) calculated from the cloud deposition model during
individual cloud episodes. According to our database, the predicted SCU2' and NOa' ion
concentrations were in good agreement with the measured values having correlation
coefficients above 0.7, when a ranged from 0.2 to 0.6. These findings will find usefulness
in parameterization schemes of acidic deposition models.
ACKNOWLEDGMENTS
This study has been funded through cooperative agreements with the United States
Environmental Protection Agency (agreement No. ESRL-CA-01 and contracts CRS 812444-
01-0, -02-0, and -03-0 with the North Carolina State University). The contents of this paper
do not necessarily reflect the views and the policies of the EPA, nor the views of all members
of the Mountain Cloud Chemistry Project (MCCP) consortia, nor does the mention of trade
names or commercial or non-commercial products constitute endorsement or recommendation
for use. Professor Volker Mohnen is the Principal Investigator on the MCCP and the EPA
project officer is Dr. Ralph Baumgardner. In this study, Dr. V. P. Aneja, in-charge of the
field studies, provided the logistical support for the 1987 field season. Dr. W. Robarge was
responsible for the chemical analysis of the cloud water samples. The cloud deposition model
codes were provided by Dr. S. Mueller.
REFERENCES
1. B. C. Scott, "Parameterization of sulfate removal by precipitation," J. Appl. Meteor. 17:
1375-1389, (1978).
2. A. W, Hogan, "Estimation of sulfate deposition," J. Appi. Meteor. 21: 1933-1936,
(1982).
3. R. G. de Pena, T. N. Carlson, J. F. Takacs and J. O. Holian, "Analysis of
precipitation collected on a sequential basis," Atmos. Environ. 18: 2665-2670, (1984).
143
-------
4. V.K-Saxena, R. E. Stogncr, A. H. Hendlcr, T. P. DcFelicc, R. J.-Y. Ych andN.-H.
Lin, "Monitoring the chemical climate of the ML Mitchell State Park for evaluating its
impact on forest decline," Tellus 41B: 92-109, (1989).
5. V. K.Saxena andN.-H. Lin, Cloud chemistry measurements and estimates of acidic
deposition on an above cloudbase coniferous forest Atmos.Environ. 24:329-352,
(1990).
6. G. M. Lovett, "Rates and mechanisms of cloud water deposition to a subalpine balsam fir
forest," Atmos.Enviio. 18: 361-371, (1984).
7. S. F. Mueller and F. P. Weatherford, "Chemical deposition to a high elevation red spruce
forest," Water. Air and Soil Pollut 38:345-363, (1988).
8. V. K. Saxena, F. P. Parungo and F. H. Ruggjero, "Airborne measurements of the
Antarctic cloud water acidity," Antarctic J. U.S. 19:201-203, (1985).
o
o
o
UJ
.1-:
.01
KILAUEA
McMURDO
PISECO,.
vj.
^ A
XI
SPOLEx
.
iJTITT Ml MITCH ELL
v
-40 -20 0 20
CLOUD BASE TEMP (° C)
Figure 1. The sulfate concentrations as a function of cloud-base temperature. The ranges of
cloud-base temperature and sulfate concentration are enclosed by labelled box. The solid-
boxes are based on the data of Mt Mitchell and McMurdo. The dash-boxes are duplicated
from the figure of Hogan (1982). The mean values for the former are the squares with the
horizontal bars and for the latter are open dimond, square, triangle and circle. For dash-box,
the solid inverted triangles represent the values obtained in a single warm midalttitude storm
and the open circle, triangle and square represent the values measured in a winter snowstorm
with a similar wind structure.
144
-------
Table 1 Sulfate chemistry of cloud water sampled at Mt. Mitchell, NC, and McMurdo,
Antarctica.
Samples
]a LWC
(ueqH) (gm-3)
Temp.
(ppm)
McMurdo 6 range
mean
ML Mitchell 86
(20 cloud events)
31-330
116
39-790
370
0.01 - 0.18
0.1
0.05 - 0.51
0.23
-14 --5
-9.5
1.3 - 14
10.8
0.28 - 2.97
1.04
0.35-7.11
3.33
a. Sulfate ion concentration measured in cloud water samples.
b. Sulfate concentration in air, converted from sulfate ion concentration measured in cloud
water samples.
Table 2 Sulfate and nitrate concentrations as a function of the rate of cloud water deposition.
N is the number of cloudwater samples, r is the correlation coefficient between the observed
and predicted values, and Ic is the cloud deposition rate estimated from the cloud deposition
model.
[804*] = a V6
Event
24/06/86
29/06/86
01/07/86
02/07/86
05/07/86
11/07/86
19/08/86
14/05/87
21/05/87
03/06/86
29/06/86
09/07/86
10/07/86
12/07/86
29/07/86
12/08/86
13/08/86
09/09/86a
09/09/86b
10/09/86
11/09/86
22/05/87
04/06/87
18/06/87
12/10/87
Type
S
S
S
S
S
S
S
S
S
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
L
N
4
3
3
5
3
4
4
3
6
11
10
7
11
5
8
17
7
6
9
18
7
8
8
7
12
a
729.669
146.551
593.496
70.776
565.619
68.420
4.138
88.682
334.204
316.670
140.957
883.455
399.895
97.085
1010.19
14.251
39.856
67.619
282.171
135.871
41.047
164.602
371.376
236.422
49.307
b
0.341
0.602
0.144
0.681
0.138
0.679
1.239
0.270
0.262
0.062
0.303
0.326
0.235
0.341
0.094
0.567
0.134
0.629
0.193
0.065
0.400
0.306
0.219
0.240
0.173
r
0.901
0.984
0.885
0.778
0.996
0.880
0.784
0.658
0.811
0.075
0.669
0.882
0.239
0.957
0.586
0.716
0.178
0.616
0.499
0.130
0.396
0.851
0.808
0.478
0.386
a
298.522
31.568
85.810
18.274
187.526
35.792
1.263
41.209
94.211
85.654
43.76
231.396
82.001
45.829
226.421
8.153
16.395
139.645
55.585
23.540
19.719
61.412
72.812
85.577
26.223
b
0.229
0.609
0.294
0.576
0.193
0.772
1.158
0.305
0.292
0.072
0.290
0.156
0.319
0.402
0.163
0.590
0.194
0.133
0.308
0.314
0.299
0.236
0.501
0.202
0.111
r
0.536
0.983
0.996
0.768
0.979
0.897
0.847
0.727
0.876
0.084
0.665
0.431
0.417
0.970
0.723
0.882
0.316
0.423
0.737
0.563
0.269
0.873
0.929
0.424
0.202
S = Short cloud event (duration < 8 hours)
L = Long cloud event (duration > 8 hours)
145
-------
MEASUREMENTS OF ATMOSPHERIC
HYDROGEN PEROXIDE IN THE GAS-
PHASE AND IN CLOUDWATER AT MT.
MITCHELL STATE PARK, N.C.
Candis S. Claiborn* and
Vincy P. Aneja
Department of Marine, Earth and Atmospheric Sciences
North Carolina State University
Raleigh, North Carolina 27695-8208
Abstract
Measurements of atmospheric hydrogen peroxide in the gat phase and in
cloudwater were made at the Mt. Mitchell State Park, North Carolina, during
the growing teaion (May through September) of 1988. Cloudwater hydrogen
p«roxide was also measured during the late summer and fall of 1987 (August
and October). LcTtls were found to be similar to those reported for another
high elevation location in the southeastern United States. Cloudwater sam-
ples collected during these periods showed a wide range of lereli (~ 0 to 219
/onoles/liter) and average values of SI /moles/liter aad 44 jaaoks/liter for the
entire sampling sessions of 1988 and 1987, respectively. ftigninVint seasonal
variation was noted both in 1987 and 19*8, with doudwater levels of hydrogen
peroxide much higher in the summer than in the Jkll.
Gas-phase hydrogen peroxide levels ranged from nearly sero to above 3
ppbv. Gas-phase hydrogen peroxide demonstrated a nighttime maximum in
the summer but not in the fall. This reverse diurnal pattern ii unlike the typ-
ical pattern of a daytime maTi'miim, as observed in most locations.
'Department of Chemical Engineering
146
-------
Introduction
Ambient hydrogen peroxide plays an important role in the aqueous phase chem-
istry of precipitation acidification, and may be the primary oxidant involved in the
production of aqueous phase sulfuric acid in cloudwater, when the pH is less than
4.5(lp3). Like ozone, hydrogen peroxide is a strong oxidizing agent. At least one
study, conducted on Norway spruce, indicated that hydrogen peroxide-containing
mist can cause damage to trees^. Since hydrogen peroxide is formed from the
recombination of peroxyl radicals when atmospheric levels of NO. are low enough
such that NO does cot compete for consumption of H0j§, this may be significant to
the rural eastern United States, where add deposition and air pollution have been
thought to be the cause of, or at least contribute to, the forest decline in the Ap-
palachians. For these reasons gas phase and cloudwater levels of hydrogen peroxide
were measured, in conjunction with other chemical, physical, and meteorological
parameters, during the growing seasons of 19S7 and 1988 at Mt. Gibbs, a mountain
adjacent to Mt. Mitchell in western North Carolina and where significant forest
decline is reported^4).
Experimental
Ambient, gat-phase hydrogen peroxide was measured using the continuous fluoro-
metric analyzer based on the horseradish peroxidase method*1) periodically during
the latter portion of the growing season (July through September) of 1988 at the
high elevation site (Site 1) at Mt. Gibbs, N.C. The dual channel fluorometric ana-
lyser measures total peroxides on one channel, and by specific enzymatic destruction
of hydrogen peroxide, organic peroxides only on the second channel. Hydrogen per-
oxide is obtained by difference. Gas phase total and organic peroxide data were
recorded on a chart recorder and extracted manually as 12-minute averages. These
data were then consolidated into hourly averages. The hydrogen peroxide analyser
was calibrated at least once daily, and calibration solutions were checked weekly.
Baseline checks were performed automatically, usually several times per day.
>•
Cloudwater samples were collected hourly at both sites 1 and 2, "fixed" for hy-
drogen peroxide according to the derivatisation technique^ to minimise losses due
to decomposition, and immediately refrigerated. These samples were later sent to
the Tennessee Valley Authority lab in Muscle Shoals, Alabama, where they were
analyzed by the fluorometric technique adapted for precipitation samples^7) for hy-
drogen peroxide content. A limited number of samples were collected in 1987, during
August and October. A larger data set was collected in 1988.
The Mt. Gibbs research station, the meteorological and climatological param-
eters measured there, and the quality assurance protocols, have been described
previously^7'8). Unless otherwise noted, statistical significance is assumed to be at
the 99 % confidence level.
147
-------
Results and Discussion
Gas-phase Hydrogen Peroxide
During the field season of 1988, 273 hourly hydrogen peroxide measurements were
recorded. Gas-phase hydrogen peroxide at Mt. Mitchell ranged from ~0 to 3.3
ppbv. In general, atmospheric hydrogen peroxide levels at the Mt. Mitchell State
Park are comparable, or higher than, values reported in the literature. At nearby
Whitetop Mtn, VA('), a maximum of 2.6 ppbv was reported in the summer of 1986,
and a maximum of 0.57 ppbv in the fall. Values over 4 ppbv have been observed
aloft, over the eastern United States^10).
Statistically significant seasonal variation in the ambient hydrogen peroxide level
during 1988 was observed, with summertime levels of hydrogen peroxide (mean
0.62 ppbv) sigraficantly greater than those observed in the fall (mean 0.19 ppbv).
Other researchers have noted similar seasonal variations in ambient hydrogen per-
oxide. A strong seasonal dependency was also observed at Whitetop Mountain^
with highest levels in both the gas and liquid phases occurring during the summer,
and lowest values in the spring and fall. In 1988, a seasonal variation was noted
in the ambient ozone levels as well, with summertime levels were higher than fall
levels at Mt. Mitchell^11).
Nighttime hydrogen peroxide levels measured during the summer of 1988 (mean
0.77 ppbv) were significantly higher than daytime levels (mean 0.46 ppbv). This
result was not expected, based on our current understanding of the photochemistry
of hydrogen peroxide formation. Atmospheric hydrogen peroxide is formed in the
absence of NO from the combination of hydroperoxyl radicals, which are formed
from the photooxidation reactions of reactive hydrocarbons or of carbon monoxide.
The nighttime maximum in ambient hydrogen peroxide at Mt. Mitchell is very
different from the typical diurnal pattern reported in the literature. In southern
California'11), a daytime maximum in hydrogen peroxide was observed in the early
afternoon, corresponding to a minimum in the NO. and a maximum in the Oi, 1
to 3 hours after the daily peak of solar radiation. Daytime values exceeded the
nighttime values by 26% at Whitetop Mountain, VAO. The reversed diurnal trend
observed at Mt. Mitchell was not noted during the fall, where the nighttime levels
(mean 0.20 ppbv) were not found to be significantly higher than the daytime levels
(mean 0.17 ppbv).
The reason for the nighttime maximum in ambient hydrogen peroxide at Mt. Mitchell
is still unknown, but because of the elevation of Site 1 (2006 m MSL), it is possible
that, during the night, the mountaintop is above the nocturnal boundary layer,
which might allow the mountaintop to be exposed to the free tropospheric, HjOj-
rich air above the mixed layer; a vertical gradient in ambient hydrogen peroxide,
increasing with height, has been reported^10'. Within the mixed layer itself, losses
of ambient hydrogen peroxide would be high due to deposition to the forest and
decomposition on dirt or water surfaces. Above the mixed layer, these losses would
be minimized.
148
-------
Cloudwater Hydrogen Peroxide
Cloudwater hydrogen peroxide levels at Site 1 ranged from 0 to 219 /iM during
1988. The mean hydrogen peroxide level was 38 /iM at Site 1 and 49 /iM at Site 2
(however, this difference was not statistically significant at the 95 % level). Cloud-
water levels at Mt. Mitchell have also been found to be comparable or higher than
most cloudwater level* reported elsewhere. The only higher values reported in the
literature were for Whitetop Mountain, VA^'\ where a maximum of 247 /iM was
observed in the summer.
Fall hydrogen peroxide levels were much smaller at Mt. Mitchell (maximum 55
/iM). In this work, seasons are roughly defined (at Site 1) as May and June be-
ing spring, July and August being summer, and September being fall. Based on
this definition of the seasons, the levels for spring, 1988, (mean 57 /iM) were not
significantly higher than those of summer (mean 47 /iM). However, the summer con-
centration was significantly higher than that of fall (mean 17 /iM) (99 % confidence).
Cloudwater samples were analyzed according to whether the cloud event was "long"
(8 hours or longer) or "short" (less than 8 hourc)(7). This classification was intended
to give an indication of the type of cloud; short cloud events are usually orographic
in nature, while the long event* generally tend to be frontal in origin. Based on
this categorization, short events have been reported to exhibit significantly higher
concentrations of mil ionic species and lower liquid water content than long events
exhibit^7'*). The ratio of concentration in the long event cloudwater to the con-
centration in the short event cloudwater was 0.52, 0.52, 0.50, and 0.58 for S0j~,
NO, , NHJ, and H+, respectively. The ratio of liquid water content for the inort
event to the long event wu 1.8 (the inverie of which is 0.56). Therefore, dilution
may be responsible for the difference in ionic concentrations between long and abort
events. In contrast, although the hydrogen peroxide content of the short event u
significantly higher than that of the long event (95% confidence), the ratio of the
concentration in long events compared to short events is 0.71, which is significantly
higher than the ratio for other cloudwater species, indicating that, although there
seems t&be an inverse relationship between cloud liquid water content and hydro-
gen peroxide content, dilution is not the only factor determining the cloudwater
content of hydrogen peroxide. This inverse relationship may be more a result from
equilibration processes rather than dilution processes.
The cloudwater hydrogen peroxide content was compared to that predicted by the
Henry's Law constant for those periods in which there were simultaneous mea-
surements of cloudwater and gas phase HjOj and temperature. The Henry's law
constant is given
H = tXP(j;-B) (1)
where H, the Henry's law constant, has units of M atm'1, T is in degrees Kelvin,
A = 6621, and B = 11.
149
-------
Good agreement was found between the liquid and gas phase concentrationi of hy-
drogen peroxide, with the slope of the line = 0.8, and the correlation coefficient r3
= 0.84. Usually, the aqueous phase concentration has not quite reached equilibrium
with the gas phase level. Exceptions occur mostly at very low concentration (gas
phase hydrogen peroxide less than about C.2 ppbv), which may be due to the ap-
proaching of the level of detection for the ambient air instrument.
Conclusion!
Hydrogen peroxide levels at Mt. Mitchell were comparable, or higher, than those
levels previously reported in the literature. Gas phase hydrogen peroxide levels
reported previously in the literature for other sites which were higher than Mt.
Mitchell levels were found aloft. Cloudwater levels at Whitctop Mtn, VA, were the
only cloudwater levels found to be higher than those at Mt. Mitchell. Due to the
southern location of the Mt. Mitchell site and the observation of the latitudinal
gradient of hydrogen peroxide, with levels increasing as we move south, this is con-
sistent with other observations reported in the literature.
Contrary to the typical diurnal variation observed at most monitoring sites, with
midday maxima, gas phase hydrogen peroxide at Mt. Mitchell exhibits a significant
"reverse diurnal variation", in the summer, but not necessarily in the fall. This
may be explained by the lowering of the nocturnal boundary layer to below the
moun taint op at night, so that the site is exposed to free tropospheric air, which
could contain higher levels of hydrogen peroxide. A vertical gradient of hydrogen
peroxide, increasing with increasing elevation, has previously been reported in the
literature.
Consistent with our current knowledge of the behavior of hydrogen peroxide at other
locations, there is significant seasonal variation in the fas phase hydrogen peroxide
concentration at Mt. Mitchell, with levels dropping in the late summer and fall.
Cloudwater hydrogen peroxide content depends on chemical parameters and cannot
be explained on the basis of dilution alone, like the ionic species. It does, however,
appear that the liquid water content of the cloud does have an effect on the level
of hydrogen peroxide in the cloudwater, although this effect is not as significant as
the effect on ionic species.
Acknowledgement
This research has been funded through a cooperative agreement with the U.S. Envi-
ronmental Protection Agency (813934-01-2) as part of the Mountain Cloud Chem-
istry Program. We express sincere appreciation to Dr. Ken Olszyna, Tennessee
Valley Authority, for analysis of the cloudwater hydrogen peroxide samples.
150
-------
References
1. S.A.. Penletl, B.M.R. Jonei, S.A. Brice, A.E.J, Eggleton, "The importance of atmospheric oionc
and hydrogen peroxide in oxidising mlphu: dioxide in cloud and rainwater," Atmoi. Environ. IS
(123-137, (1979).
2. L.R. Martin, D.E, Darnichen, "Aqueoui oxidation of lulfur dioxide by hydrogen peroxide at low
pH," At mo.. Environ. 16:9 (1C15-1821). (1981).
3. E.J. Mallant, J. Slaaina, G. Maaueh, A. Kettrup, "Eiperimenti OB fljOj-eontaining fog expo-
lurei of young treei," Aerosols: Research, Riik Assessment, aad Control Strategies. (1988).
4. R.I. Bruck, W.P. Robarge, "Change in forwt structure in tke bore*] montane ecosystem of Mount
MitcheU, North Carolina » Eai. J. For. Path. II: (3S7-3«e). (1918).
6. A.L. Lairui, G.L, Kok, J.A. Lind, S.N. Gitlin, B.G, leikes, R.E. Shetter, "Automated fiuoro-
metric method for HjO, in lii," Anal. Cfom- £8:694-S9?.(198B).
6. G.L. Kok, K. Thompton, A.L. LMriu, S.E. McLaren, *Derintii*tioii tecanique for the determi-
nation of peroxidei in precipitation,* Anal. Chem. 68 (1192-1194). (1986).
7, A.L. Lurut, G.L. Kok, S.N. GiiUa, J.A. Lbd, S. McLaren, "Automated fiuorometric method for
hydrogen peroxide in atmotpherie precipitation,1' Anal. Cbem 17 (917-922). (1985).
8. V.K. Saiena, R.E. Stofner, A.H. Headier, T.P. deFelice, R.J.-Y. Yea, N. -H. Ian, -Monitoring
the chemical climate of the Mt. Mitchell State Park for tnlaatioi of iti impact on fbiwt decline,"
TeUu. 41B (9M08). {1980).
6. Aneja., V.P., C.S. Claiborn, A. Moithy, D.-S. Kin, and Z. Li (1990). CharacUriiation of the
chemical and physical diinatolofj at Mt. Mitchell, N.C. for erahwtioji of the role of air pollution
in forejt decline, in preparation.
10. K.J. Oliiyna, J.P. Meagher, E.M. Bailey, 'Gat-phate, clond aad rainwater measurement* of
hydrogen peroxide at a high-elevation nte,* Atmc«. Raviroa. 13:1 (1W9-17M). (1988).
11. B.F. leikes, G.L. Kok, J.G. Waltga, A.L. Laarmi, 'H»Oj aad SO* MMiaitmtaU IB tilt lover
troposphere over the emstexn United States during fall,' J. Q*ophy». *M. B2:D1 (B1S-931). (1917).
12. V.P. Aneja, S. Boaster, Z. Li, C. CUibom, A. Marthy *O»ea« efiaiatolofj at high elevatiou
in the Southern AppakeUau," J. Qeophyi. EM. (ItIO),
13. E. Sakugawa, I.R. Kaplan *B|O] and O« is the atmoaphere of Lot Angela and iti ricinity:
factori controlling their formation and their rokt as oxidanti of SO,," J. Geophyi. Ret. 94:D10
(12957-12173). (1989).
14. J.A. Lind, G.L. Kok, 'Henry'§ law determinations for aqneoos tolation* of hydrogen peroxide,
raethylhydioperoxide, and peroxy&cetk acid,* J. Geophyt. Re>. 91:D7 (7889-7895). (1686).
151
-------
MODELING OROGRAPHIC CLOUD WATER DEPOSITION AT MT.
MITCHELL, NC: THE EFFECT OF LOCAL TOPOGRAPHY
Steven R. Chiswell and Steven Businger
Department of Marine, Earth and Atmospheric Sciences
North Carolina State University, Raleigh, NC 27695-8208, U.S.A.
Ronald L. Bradow
U.S. Environmental Protection Agency
ASRL/MD-59, Research Triangle Park, NC 27711
A two-dimensional orographic cloud model MCCP PLUVIUS is applied to the Mt.
Mitchell Mountain Cloud Chemistry Project (MCCP) site in order examine possible sources of
variability within sampling site measurements due to topographic forcing. Model simulations
are used to determine the effect of local topography on observed pollutant deposition and
cloud water concentrations. The dissolved pollutant concentration at the mountain summit was
found to be highly dependent on the distance which over which deposition occurred. The
equilibrium balance between pollutant loss through deposition and replacement by turbulent
diffusion is shown to vary with cloud size. Orographic clouds generated by trajectories
crossing the summit from west to east are found to produce more acidic clouds at the Mt.
Mitchell MCCP site for small to moderate sized clouds, while the clouds passing from east to
west are more acidic for exceptionally moist flows. The use of sampling site measurements to
classify atmospheric conditions is shown to be inconclusive.
152
-------
Introduction
The summer of 1988 was among the driest ever recorded for the southeast. Despite
these conditions, Mt. Mitchell, NC (elev 2037 m) was within a cloud 960 hours out of a total
of 3260 hours sampled, representing 29% of the total time. The extreme presence of clouds
present at this elevation, even during a period of regional drought, as well as their ability to
scavenge available pollutants from the atmosphere has lead to the hypothesis that clouds are
one of the leading mechanisms for the transfer of atmospheric pollutants to the above cloud
base forests. The Mountain Cloud Chemistry Project (MCCP) was initiated in order to
investigate the role of direct cloud water interception as a mechanism in the decline of the high
altitude forests of the Appalachians1. As part of the project, a two-dimensional orographic
cloud model MCCP PLUVIUS was developed to aid in the evaluation of the data collected at
field sites2. PLUVIUS simulates the chemical and physical processes associated with
orographic cloud formation as a result of air flow over a mountain, including aerosol and
gaseous scavenging, in-cloud chemical transformations, and cloud water deposition.
In 1988, over two-thirds of the time that clouds were present at the summit sampling
site at Mt. Mitchell (Site 1), the local wind direction was observed to be within 20° of the
normal to the ridge axis. Observations show that this is the result of channeling the wind field
between adjacent ridge lines. Thus it is possible for similar wind conditions at Site 1 to result
from radically different trajectories. In general, the easterly facing slope of the Mt. Mitchell is
steeper than the westerly facing slope, and the prevailing wind crosses the ridgeline from west
to east approximately 60% of the time. In order to evaluate the performance of PLUVIUS
based on measurements made at Mt. Mitchell, we must therefore first be able to assess the
influence that local topography will have on the resultant model prediction.
Evaluation of Topographic Influences
The local topography used within PLUVIUS determines the relative distance from the
sampling site at which a cloud will form along a streamline, and consequently, how far the
cloud must travel until reaching the site. Two parameters that effect the dissolved pollutant
concentration and cloud water deposition within the model are turbulent diffusion and
deposition through impaction. Turbulent diffusion within the model continuously acts to
redistribute areas of high and low concentrations of a quantity in order to produce a more
uniform distribution. Deposition within the model, on the other hand, only takes place when a
the column of air being advected over the mountain becomes saturated in the layer adjacent to
the forest canopy.
Deposition within PLUVIUS is parameterized by a relationship provided by Lovett3
which is assumed applicable to the Mt. Mitchell forest. Since Deposition through impaction is
generally much larger that that for settling given the typical wind speeds found in a mountain
cap cloud, deposition through settling is neglected. Turbulent diffusivity is based upon mixing
length considerations using a value of 10 m2 cnr1 suggested by Luecken et a/.1 for a
moderately stable environment.
The effect of local topography on the above parameterizations was tested using a base
case chemical atmosphere, and three soundings each representing different magnitudes of
orographic cloud formation. Various topographic profiles were implemented to explore the
general effects of trajectory slope on cloud development and deposition. The initial height of a
streamline was constant for each profile tested, as was the summit height. Topographic
profiles derived from the terrain surrounding Mt. Mitchell were then used as input to the model
to examine the effect which local topography may have on measurements obtained from the
Mt. Mitchell MCCP site.
153
-------
Results and Discussion
Pollutant concentrations at the cloud base were found to drop quickly once deposition
was initiated (Figure 1). This condition is the result of the relative inefficiency of diffusion to
replace pollutants lost through deposition when the mass flux is large. After the initial drop,
the pollutant concentration reaches a point at which the mass flux out of the cloud approaches
an equilibrium with diffusion of pollutants from the surrounding air into the cloud base
producing a much more gradual decline. From Figure 1, the pollutant value sampled at a
sampling site can be seen to be a function of the distance over which deposition occurred.
Trajectories that rise steeply once saturation is reached in the column base travel a shorter
horizontal distance above cloud base, thereby reducing the distance over which deposition
occurs; thus, higher spatial pollutant concentrations result.
Pollutant mass lost through deposition within the model is replaced only through
diffusion; alternatively, liquid water lost through deposition is replaced not only by diffusion,
but by additional condensation if lifting continues. Therefore the pollutant mass within a cloud
is affected to a higher degree by deposition than is the liquid water content (LWC). In
opposition to deposition which is only invoked once a cloud forms, diffusion is always acting
to redistribute areas of high and low concentration. In this manner, diffusion acts to dissipate
saturation within the cloud over the entire length of the trajectory. Steeply sloped trajectories
resulted in higher LWC's than did those with more gradual slopes. This result lead to lower
dissolved pollutant concentrations for steep trajectories. As a result, steep trajectories which
favored higher pollutant masses through less deposition were often coincident with conditions
favorable for greater dilution of the pollutants remaining in the cloud.
When local topography was introduced into the model, small to moderate clouds (<
=0.6 g nr3) resulting from trajectories passing over the Mt. Mitchell ridge from west to east
were found to be more acidic than those originating from the east This is in general agreement
with experimental results reported by Saxena et a/.4 The amount of variability decreased with
increased cloud size, ranging from 3.0 - 3.5 pH for an LWC of 0.1 g nr3 to 4.6 -4.8 pH for
an LWC of 0.6 g nr3.In the case of small clouds, the distance over which deposition occurred
is very small in comparison to the distance over which diffusion occurs. In these cases, the
steeper trajectory (originating from the east) produced the greatest LWC, and therefore the
greatest pollutant dilution. For very moist clouds (> 0.6 g nr3) in which deposition occurred
over a long distance, clouds passing from east to west were found to be more acidic. For this
case, the distance over which deposition occurs becomes of an order comparable to that of
diffusion. Depletion of pollutants thus limits the acidity of clouds when diffusion cannot
balance the mass flux out of the column, and since steeper trajectories decrease the distance
over which deposition occurs, these trajectories may produce more acidic clouds.
Conclusions
Local topography was found to greatly influence the pollutant concentration within the
model cloud and the pollutant deposition to the mountain. When actual topography was
introduced, trajectories approaching Site 1 from the west side of the ridgeline were found to
generally produce more acidic clouds than those approaching from the east. In large clouds
where depletion of pollutants is important, the steep eastern facing side of the Mt. Mitchell
ridge produced the most acidic clouds. The model parameterizations for deposition and
diffusivity strongly influence the equilibrium concentration of pollutants within a cloud, and
should be taken into account when comparing model predictions to observed conditions.
Topographic forcing of the flow field is therefore an important contributor to the values
measured at the Mt. Mitchell MCCP sites. Clouds resulting from similar air masses and
observed wind conditions on the mountain can display significantly different characteristics in
pollutant concentrations. Alternatively, similar cloud acidity does not necessarily imply a
similar pollutant environment. It may therefore be misleading to classify an airmass or cloud
environment based solely on field measurements.
154
-------
References
1. V. K. Saxena, "Mountain Cloud Chemistry Project at Mt. Mitchell, North Carolina:
Strategies and Highlights," Trans. Amer. Geophvs. Union (EOS) 6&: 270. (1987)
2. D. J. Luecken, C. D. Whiteman, E, G. Chapman, G. L. Andrews, and D. C. Bader,
Description of the Mountain Cloud Chemistry Program Version of the PLUVIUS
MOD 5.0 Reactive Storm Simulation Model. PNL-6242, Pacific Northwest
Laboratory, Richland Washington. 1987, pp 5 - 42.
3. G. M. Lovett, /'Rates and Mechanisms of Cloud Water Deposition to a Subalpine
Balsam Fir Forest." Atmos. Environ. 18: 361-371. (1984)
4. V. K. Saxena, R. E. Stogner, A. H. Hendler, T. D. DeFelice, R. J-Y. Yeh, and N-H.
Lin, "Monitoring the chemical climate at the Mt. Mitchell State Park for evaluation of
its impact on forest decline," Tellus 41B: 92-109. (1989)
Acknowledgements
This research has been funded under the US EPA grant number 814612-01-0. The
contents of this paper do not necessarily reflect the policies or opinions of the EPA nor those
of the MGCP consortia. The authors wish to express their gratitude to C. D. Whiteman and D.
J. Luecken for their willingness to discuss the use of MGCP PLUVIUS and offer their
opinions on its development
C-4
50
45
40
35
30
25
20
15
10
-o—— Sounding * 1
-•—— Sounding *2
-M— Sounding *3
-2000 -1500 -1000 -500 0 500 1000 1500 2000
Distance from Summit, meters
Figure 1. Hydrogen ion (shown as a representative field) flux to the model mountain surface
for three sounding cases using a gaussian mountain. Positive X direction is upwind of
the mountain.
155
-------
Remote Site Cloud Condensation Nuclei 'Fingerprints':
Indicators of air-mass transport and mixing
T. P. DeFelice
National Research Council/ NASA
Post-Doctoral Resident Research Associate,
NASA-Ames Research Center, SGP: 245-4, Moffett Field Ca. 94035
The cloud condensation nuclei, CCN, concentration was measured in Mt. Mitchell State
Park, NC during the summer of 1988, in conjunction with the Mountain Cloud Chemistry
Project, to help further the understanding of the physico-chemical characteristics of the air
which delivers wet acidic deposition to forests. The CCN measurements in this study were
made using a CCN-Spectrometer, and they ranged between <100 and 3,600 cnf ^. The
synoptic conditions, the local meteorological and the microphysical parameters (i.e., total
droplet concentration) were used to establish whether air-mass transport or mixing yielded
the measured CCN concentrations. Large variations in CCN concentration (>200 cm"^ h~l)
were found to be related to the sampling of a non-homogeneously mixed air parcel or near the
boundary of different air-masses. Large temporal CCN variations that occur near the end of
an event are likely due to evaporation. More suttle changes in the CCN concentration (<200
'-* h~l) were associated with air-mass transport.
156
-------
1. Introduction
The cloud condensation nuclei, CCN, concentration was measured in Mt. Mitchell State
Park, NC during the summer of 1988, in conjunction with the on going measurements of the
Mountain Cloud Chemistry Project, to help further the understanding of the physico-chemical
characteristics of the air delivering the wet acidic deposition to the local forests. Such
measurements are the first in this area, although they have been made elsewhere l~&. The
relation of CCN measurements to their source^ (e.g. maritime, continental, evaporation),
and to the dynamics^ involved in producing a critical supersaturation with respect to water
vapor and hence the microstructure of clouds is known. The intent is to enhance the present
knowledge by uniquely applying this data to the question of whether air-mass transport, or
mixing, yielded the measured CCN concentrations. My measurements suggest the existence
of a threshold magnitude of the time rate of change in the CCN concentration (>200 cm'3 h~
1) that is associated with the non-homogeneous mixing of different air parcels or air-masses
(e.g. near their boundaries, evaporation). Temporal changes in CCN concentration less than
the threshold are associated with synoptic scale transport. Knowing how the CCN, in the air
that flows across the site, are generated would be a beneficial input to acid deposition models
since this knowledge would make such models more indicative of the present dynamical
processes, the source and the physico-chemical characteristics of these aerosols. The
accuracy and operationality of these models will also improve. CCN measurements would
also prove valuable in investigations related to the effect that they might have on climate, e.g.
dimethylsulfide and climate 10, or even on cloud formation in other planetary atmospheres.
2. Background
Measurement of CCN
The instrument used to measure the CCN^, termed the CCN-Spectrometer (Figure 1),
yields a continuous measure of the number of aerosols activated over the entire range of
atmospheric supersaturations in any instant. The ability of tackling the problems associated
with the role of the physical-chemical characteristics of the aerosols on the formation of
clouds was allowed, for the first time, on the order of the lifetime of a cloud. By 1988 a
similar spectrometer^ had evolved. It consists of 5 to 8 basic thermal gradient diffusion
chambers in series with a ROYCO particle counter capable of simultaneously counting five
different particle size ranges at the exhaust end of the instrument from which a
supersaturation profile is inferred. The CCN-Spectrometer produces a supersaturation
profile, equivalent to that yielded by a number of basic thermal gradient diffusion chambers
in parallel, which is derived from the variation of the vapor density profile across the width
of the chamber due to the difference between the top and bottom plate temperature profiles
across the width of the chamber. The desired supersaturation range is obtained by adjusting
157
-------
the temperature difference between the two plates. The resulting activated particles are
counted by a CLIMET 208 particle analyzer as its inlet sweeps the entire width of the
chamber within fifteen seconds. See Fukuta and Saxena^ 1 for details.
Cloud condensation nuclei (CCN) spectra were measured before, during, and after each
event The data was fitted to the standard functional relationship between supersaturation, S,
and cloud condensation nuclei, CCN, namely
N = C 8 (1)
where N is the number concentration of CCN per unit volume with critical supersaturation
less than S (in percent); C is the concentration parameter of the distribution and k is its slope
parameter. The CCN-Spectrometer measures N at any given range of S instantaneously.
The value of k is related to the dynamics involved in producing the S°. A k value very much
less than one, i.e., <0.6, implies the cloud droplet concentration is essentially determined by
the composition of the CCN concentration active at different supersaturations. A k greater
than unity suggests that the cloud droplet concentration is more related to the rate of the
ascent, which is proportional to the rate of the supersaturation generation (dS/dt), rather than
to the aerosol content. Such k values may also be the result of existing CCN source regions,
since recently generated particles are likely to be small and require high activation
supersaturations yielding higher k values^.
Site Description
The site in the Mt. Mitchell State Park, NC is located on Mt. Gibbs (=3.2 km
southwest of Mt. Mitchell), which is located in the westernmost portion of the state, and
consists of a 17.1 m walk up tower equipped with temperature, pressure, wind speed, wind
direction, and humidity instruments near its top. A carriage, positioned on the tower's
northern face, carries the Atmospheric Science Research Center (ASRC) passive teflon string
cloudwater collector, and the Particle Measuring Systems Forward Scattering Spectrometer
Probe (FSSP) from the ground to as high as a couple of meters above the top of the tower.
An instrument shed at the base of the tower houses the gaseous instruments which are
hooked into a sampling manifold that extends to the top of the tree canopy (approximately
12.1 m below the top of the tower). The CCN-Spectrometer was operated from a tool shed
located about 8.0 m to the northeast of the tower. Its inlet is an inverted funnel (0.1 m
diameter) that extends 0.5 m out from and whose opening is 2 m above the top of the shed.
Further details of the site are given in Saxena et al.^3.
158
-------
3. Results and discussion
_T
The CCN measurements at our site ranged between <100 and 3,600 cm0 and are
consistent with other studies^. The <100 cur 3 CCN concentrations were also measured in
the Olympic Mountains^ and have been observed to follow widespread rain". CCN
concentrations <200 cnr^ are typical of maritime air-masses^ 5, Unpolluted continental
air-mass concentrations^^ are usually between 200 and =2,000 cm"3, while polluted
continental air-masses have concentrations above 2,000 crrr-*. Table I shows the typical total
CCN concentrations (based on 5=0.92%) and the respective k parameter for the given time
periods before and after four cloud events during the 1988 field season. The events in this
table are associated with air-masses that range from maritime (JULY 22) through 'aged1 -
continental (JUNE 30). Two of the events in Table I (namely, JUNE 30 and SEPTEMBER
24) are accompanied with no change in air-mass, while the remaining two events (namely,
JUNE 24 and JULY 22) are. The values in this table are based on averages of 3 h or longer.
The JUNE 24 and JUNE 30 events occurred under different meteorologically dynamic
processes1 ^. Observations suggest that the pre-JULY 22 CCN concentrations of <100 cm"3
are due to the 24 h a-priori period of widespread precipitation. It is possible that the low
CCN measurements may be due to operational shortcomings of the instrument during this
time. Regardless of the reason, a misinterpretation of the measurement is inevitable. The
CCN measurement of 3,600 cm~3 was taken near the end of the JUNE 24 cloud event. If
this measurement is encountered, after the fact by a modeler, it might be interpreted as a
change to a polluted air-mass. This would imply that a different composition and source of
the activated aerosol reached that site. For example, such a measurement could be due to
evaporation in which case the sampled aerosols are likely to have the same general
characteristics as those prior to the evaporation. An illustrative case, i.e., June 24-25, 1988,
is presented.
Case Study
.Tune 24-25. 1988. The June 24-25, 1988 event began at 1630 EST with
thunderstorms that delayed sampling until 1926 EST and continued for another 9.5 h as a
surface high, which extended itself into the Great Lakes, the mid-west and the mid-Atlantic
states 12 h prior to this event (Figure 2a), became oriented southwest to northeast by 0700
EST June 25 due to an eastward moving southern Canadian frontal system (Figure 2b). The
850 mb level, the closest standard vertical level to our site (=810 mb), shows a large high
pressure system centered south southeast, SSE, of the site that extended north to off the
South Carolina coast 12 h prior to this episode (Figure 3a). The high strengthened as its
center moved northwestward, NW, and south of the site (Figure 3b) by 1900 EST June 24.
A low moving eastward, E, across southern Canada strengthened the high pressure (now
located over mid-southern Tennessee) by 0700 EST June 25 (Figure 3c). The winds at 850
159
-------
mb went from moderate northwest, NW, 12 h prior to the event to light and variable at 1900
EST June 24 to moderate-strong west northwest, WNW, by 0700 EST June 25. Figure 3c
also shows that the 850 mb 24-48 h back trajectories (ending 0700 June 24 (2) and 0700
EST June 25 (1)) shifted from WNW to NW flow during the course of this event. The 500
mb level, i.e., the level which steers synoptic systems, is typical for summer in this area,
namely, placid. The shift in the windfield, particlularly at 850 mb and at our site,
corresponds to the passage of an upper level disturbance.
Figure 4a shows the temporal variation of the total droplet concentration, N(j, and the
cloud condensation nuclei concentration, CCN. The N^ varied between 75 and 500 cm~3
while the CCN concentration ranged between 60 and 3570 cirr^. A period of light drizzle
was observed between 0103-0203 EST and explains the latter peak in the droplet
concentration. The slightly earlier peak (i.e., 0003-0103 EST) corresponds to a bimodal
cloud droplet spectra. The total droplet and CCN concentration values show a significant
change after the 2 h period ending 0203 EST compared to before it, excluding the likelihood
of evaporation during the last part of the event. This change is also evident in the concurrent
variations of the cloudwater ions. The site wind direction is also given at the top of this
figure. The wind direction begins to shift westward toward north around 0100 EST. The
0.3h averaged pressure dropped by 1.0 mb (P) between 0145 and 0200 EST. The segment
of the CCN curve between the open boxes (0218-0309 EST) represents a time rate of change
in the magnitude of the CCN concentration of >200 cnr^ h'* (208 crrf3 h'1). The open
oval, in a similar vain to the open boxes, denotes the onset of evaporation. The 0.3 mb
averaged pressure rose by 1.0 mb (circle P) between 0645 and 0700 EST. Figure 4b shows
the primary ions, excluding H+, present in the collected cloudwater. The early event
maximum, especially in the case of sulfate, may be the result of additional sulfate production
caused by the scavenging of the 100 ppb ozone concentrations present at the beginning of
this event!'. The minimum in the concentration of these ions is likely due to the presence of
precipitation, and the rebound could be indicative of a new source of CCN. Evaporation may
also have enhanced this rebound. Note that all of the cloudwater ions had the same temporal
pattern. The measured CCN spectra showed two, occasionally three, different C and k
parameter sets to exist (for example, at 2118, 0437 and 0456 EST), and may be due to
measuring polluted air. Saxena (1980) also attributed such to the sampling of non-steady
state circumstances. While the nonconformity to Eq. 1 at 2118 EST may possibly be due to
inhomogeneous mixing between maritime and continental air-masses^-15, the latter two
occasions (i.e., 0437 and 0456 EST) are due to the non-steady state occurrence of
evaporation since they were measured at the end of this event. The change in the total droplet
concentration, the CCN concentration, the local meteorology and in the ionic composition of
the cloudwater is in phase with that from the synoptic scale meteorology (Figure 2-3).
160
-------
4. Concluding remarks and recommendation
The real-time measurement of CCN spectra (i.e., CCN concentration versus
supersaturation), or the CCN "fingerprint", was found to be indicative of air-mass transport
and mixing. The use of these fingerprints in conjunction with microphysical, meteorological
and chemical measurements could be quite beneficial in making acid deposition models more
indicative of the present dynamical processes, the source and the physico-chemical
characteristics of these aerosols. The consequential result would be an increase in the
accuracy and in the operationality of these models. Studies relating to the effect that CCN
might have on climate change and those relating the role CCN might play in other planetary
atmospheres would also benefit.
The results of the case studies are summarized;
• The magnitude of the temporal change in CCN concentrations of <200 cnr 3 h~ 1 are
associated with the transport of an air-mass associated with a frontal system, an upper level
disturbance, or the rising of local valley fog. However, in contrast, a >200 cm'3 change in
the magnitude of the CCN concentration, for example, from 100 cnr^ to 300 cnr3, implies
that sampling was first conducted in a maritime air-mass that switched to a clean continental
air-mass. Such a change that is <200 cnr^ is not interpretted as a change in air-mass.
• The magnitude of the change in CCN concentrations of >200 cm~3 h~ 1 is due to
sampling near the boundary of a frontal system, upper level disturbance, two different air-
masses, or within a non-homogeneously mixed air parcel. At the end of an event such a
change in CCN concentration is most likely result from evaporation.
Further research using CCN measurements in conjunction with microphysical,
chemical and meteorological processes is highly recommended.
5. Acknowledgements
This study has been funded through cooperative agreements with the United States
Environmental Protection Agency (agreement ESRL CA-01 and contracts CRS 812444-01-0,
-02-0, and -03-0). The contents of this note do not necessarily reflect the views and the
policies of the EPA, nor the views of all members of the Mountain Cloud Chemistry Project
(MCCP) consortia, nor does the mention of trade names or commercial or non-commercial
products constitute endorsement or recommendation for use. Professor Volker Mohnen is
the Principal Investigator on the MCCP and the EPA project officer is Mr. Ralph
Baumgardner. Dr. V. P. Aneja handled the logistics of making the measurements during the
161
-------
1988 field season. The author wishes to thank Professor Dr. V. K. Saxena for his most
enlightening and inspirational discussions during this work.
6. References
1. Garmy, M., R. Serpolay, "An isothermal diffusion tube for getting low-supersaturation
spectra of CCN by coupling with a POLYTEC HC-15 spectrogranulometer," J.
Aerosol Sci. 17: 401-405 (1986).
2. Hobbs, P. V., D. A. Bowdle, L. F. Radke, "Particles in the lower troposphere over the
High Plains of the United States. Part II: Cloud condensation nuclei," J. dim. Appl.
Meteorol. 24: 1344-1356(1985).
3. Hudson, J. G., "Airborne CCN spectral measurements," Lecture Notes in Physics,
309, Atmospheric Aerosols and Nucleation, Proc. 12th Intl. Conf. on Atmos.
Aerosols and Nucleation. Vienna Austria, 22-27 August, P. E. Wagner, G. Vali,
eds., Springer-Verlag, New York. 1988, pp. 575-578.
4. Hudson, J. G., P. Squires, "An improved continuous flow diffusion cloud chamber,"
J. Appl. Meteorol. 15: 776-782 (1976).
5. Juisto, J. E., "Aerosol and cloud physics measurements in Hawaii," Tellus 19: 359-
367 (1967).
6. Kaye, A. D., W. I Megaw, "Measurement of CCN with the thermal diffusion tube,"
Lecture Notes in Physics, 309, Atmospheric Aerosols and Nucleation, Proc. 12th
Intl. Conf. on Atmos. Aerosols and Nucleation. Vienna Austria, 22-27 August, P.
E. Wagner, G. Vali, eds., Springer-Verlag, New York. 1988, pp. 571-574.
7. Shaw, G. E., "Cloud condensation nuclei associated with arctic haze," Atmos.
Environ. 20: 1453-1456(1986).
8. Twomey, S., "The nuclei of natural cloud formation. Part I. The chemical diffusion
method and the application to atmospheric nuclei," Geofisica Pura e Applicata 43:
227-242 (1959).
9. Charlson, R. J., J. E. Lovelock, M. O. Andreae, G. S. Warren, "Oceanic
phytoplankton, atmospheric sulfur, cloud albedo and climate," Nature 326: 655-661
(1987).
10. Fukuta, N., V. K. Saxena, "A horizontal thermal gradient cloud condensation nucleus
spectrometer," J. Appl. Meteorol. 18: 1352-1362 (1979).
11. Twomey, S., Atmospheric Aerosols.Elsevier. New York. 1977, pp. 98-182.
12. Saxena, V. K., "Some wintertime cloud aerosol interactions over Lake Michigan," L.
Rech. Atmos. 14: 255-265 (1980).
13. Saxena, V. K., R. E. Stogner, A. H. Hendler, T. P. DeFelice, R.J.-Y. Yeh, N.-H.
Lin, "Monitoring the chemical climate of the Mt. Mitchell State Park for evaluating its
impact on forest decline," Tellus 41B: 92-109 (1989).
162
-------
14. Radke, L. F., P. V, Hobbs, "Measurements of cloud condensation nuclei, light
scattering coefficient, sodium containing particles, and aitken nuclei in the Olympic
Mountains of Washington," J. Atmos. Sci. 26: 281-288 (1969).
15. Twomey, S., T. A. Wojciechowski, "Observations of the geographical variation of
cloud nuclei," J. Atmos. Sci. 26: 684-688 (1969).
16. DeFelice, T. P., "Characterization of extreme deposition of air pollutants in Mt. Mitchell
State Park: Potential for forest decline and opportunity for cloud deacidification,"
PhD thesis. NCSIL MEAS. Raleigh, NC 1989, 199pp.
17. DeFelice, T. P., V. K. Saxena, "Temporal and spatial distribution of ionic composition
and acidity in clouds: Comparison between modelling results and observations," L
Atmos. Sci. 47: 000-000 (1990).
Table I Representative total CCN concentrations of measured CCN spectra (based on
S=0.92%) and the respective k parameters for the given time periods (parentheses)
before and after the sampling of four cloud events in 1988. Also shown is the
duration of the respective cloud events
EVENT CCN CONCENTRATION (cm'3) EVENT
BEFORE AFTER DURATION
JUNE 24 344,k=1.8 (18.0 h) 672, k=0.6 (8.0 h) 12.3 h
JUNE 30 539,k=2.0 (10.0 h) 527, k= 1.0 (3.0 h) 3.7 h
JULY 22 94,k=5.5 (24.0 h) 234, k=6.3 (10.5 h) 3.5 h
SEPTEMBER 24 254, k=2.0 (0.5 h) 291,k=0.9 (0.5 h) 0.5 h
163
-------
Moisture Supply Profile
(Upper Plate)
Inlet
Climet Particle
Analyzer
Exit Chamber
Sample tube into sensor
Figure 1 The CCN-Spectrometer.
Figure 2 National Weather Service surface charts for 0700 EST June 24 (a) and 0700 EST
June 25 (b). The star indicates the location of the Mt. Mitchell State Park, NC
sampling site.
164
-------
(b)
(cj
Figure 3 The National Weather Service 850 mb maps for 0700 EST June 24 (a), 1900 EST
June 24 (b), for 0700 EST June 25 (c), and the 24-48 h 850 mb back trajectories
ending 0700 EST June 24,(2), and 0700 EST,(1), June 25 are shown by the
arrows.
165
-------
o
8
B
O
U
800
600
400
200
SULFATE
DRIZZLE
(b)
AMMONIA
2200 CCN/cc/h
1600
2100 0000 0200 0400
SW SSW SW Wj WNW (a)
1000
U
U
E
41
U
E
O
U
500
CN
1700 1900 2200
0200 0400 0700
1100
Time (EST)
Figure 4 The temporal variations of cloud droplet concentration, N4=, NC>3' and NH4+, (b) during the June 24-
25, 1988 event. The open boxes indicate the period during which the magnitude
of the time rate of change of the CCN concentration is >200 cm~3 h~l. The oval
represents the same as the open squares, except were evaporation is suspected.
The upper case P denotes a 1.0 mb fall in the 0.25 h average atmospheric
pressure. The circled upper case P denotes a 1.0 mb rise in the 0.25 h average
atmospheric pressure.
166
-------
POLAR VOLATILE ORGANICS: OVERVIEW OF
PROJECTS OF THE MONITORING METHODS
RESEARCH SECTION - MRB/MRDD/AREAL, U.S. EPA
Joachim D. Pleil and William A. McClenny
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
The diversity of physical and chemical properties of the
various classes of polar volatile organic compounds (PVOCs)
requires the application of different monitoring and analytical
strategies. Though monitoring methodology for PVOCs at part per
million concentrations exists for industrial hygiene applications,
these methods are not adequate at ambient concentration levels.
The Monitoring Methods Research Section (MMRS) is currently
performing or sponsoring research to develop specific methods for
the determination of trace levels of PVOCs in an air matrix. These
projects include the extension and modification of nonpolar VOCs
methodology (Method TO-14) for ketones, epoxides and acrylates;
development of novel adsorbents for phenolic and other acidic
compounds; real-time methods for formaldehyde; application of new
analytical systems such as a glow discharge ion source interfaced
to a tandem mass spectrometer, an ion trap system, and a gas
chromatograph with atomic emission detector; development of
sampling and sample introduction hardware; and development of
calibration methods. A brief overview and current status of these
projects for PVOCs is presented.
167
-------
Introduction
The Monitoring Methods Research Section (MMRS) of the EPA
Atmospheric Research and Exposure Assessment Laboratory developed
EPA's Method TO-14 for the determination of trace levels of
specific nonpolar volatile organic compounds (VOCs) in air (1) .
This methodology is not, however, suitable for what are loosely
referred to as the "polar organics" or PVOCs. This diverse
grouping of compounds is best characterized by their relative
polarity with respect to the hydrocarbons and halogenated hydro-
carbons for which TO-14 was developed, or alternatively, the PVOCs
are those volatile organics containing atoms such as oxygen, sulfur
and nitrogen. These compounds are difficult to deal with at the
trace level for a variety of reasons, depending on the specific
sub-class. Some tend to be chemically active on metal or other
surfaces (in sample tubing, canisters, cryotraps, etc.), and others
may be affected by the drying step necessary to remove the bulk of
the cocollected water vapor.
The polar fraction of the airborne organic compounds (PVOCs)
has been related to high mutagenic and carcinogenic activity, par-
ticularly for the semi-volatile and nonvolatile compounds (2-4),
and for some specific PVOCs such as formaldehyde (5) and ethylene
oxide (6, 7). As such, their trace level determination in air is
of interest from a risk assessment viewpoint. Secondly, PVOCs are
often implicated as the cause of odor related nuisance complaints
from the public (8-13).
Though there exist a variety of industrial hygiene related air
methods for high levels of many of the PVOCs, these methods are
geared to occupational exposure levels (threshold limit values in
the parts per million by volume range) and lack the sensitivity for
ambient work, typically by 3 to 5 orders of magnitude. To address
the need for determination of PVOCs in ambient air, MMRS has
embarked upon a number of projects to develop both sampling and
analytical methods for various specific classes of compounds of
interest. These classes include, but are not limited to, al-
dehydes, amines, epoxides, thiols, alcohols, phenols, and
isocyanates.
This paper presents an overview of the major research efforts
of MMRS that are directed to determination of PVOCs:
• Application of TO-14 methodology to certain PVOCs
• Anion exchange resin as a sorbent for phenolic compounds
• Real-time trace level determination of formaldehyde
• Glow discharge source combined with MS/MS instrumentation
for in-situ atmospheric analysis
168
-------
• Chemical ionization with ion trap detector (ITD) using a
water reagent
• Atomic emission detection (AED) coupled to a capillary
column gas chromatograph
• Calibration methodology using dynamic and headspace methods
• Valveless GC/MS inlet and preconcentrator
Discussion
TO-14 Extension
The canister sampling methodology for nonpolar VOCs is
currently under investigation for extension to certain PVOCs.
In-house testing by contractor personnel (NSI Technology Services,
EPA Contract 68-02-4444) has shown that over a seven day period the
variations in analysis results can be summarized in terms of
percentage relative standard deviations (%RSD) of concentration
measurements. The listing below is divided into three groups,
depending of the range of values of the %RSDs.
Compounds RSD %
N=14
Ethyl acrylate, methyl methacrylate <3
Acetone, acrylonitrile, isopropanol <10
2-butanone
Butanol, acetonitrile, ethanol <23
Samples for these preliminary tests were prepared at the 1-
to 4-ppbv level with dynamic dilution of standards from Scott
Specialty Gases (Plumsteadville, PA) in a matrix of ambient air.
Analyses were performed using standard TO-14 analytical methodology
(1) modified to bypass the Nafion tube arrangement used for drying
the sample. Only parallel flame ionization and electron capture
detectors (FID and ECD) were used to avoid problems with co-
collected water vapor. Descriptions of the specific analytical
steps are available in TO-14 (1) and other references (14, 15).
This work is now being extended to a larger group of compounds
through a research project at Battelle Memorial Institute
(Columbus, OH) under EPA Contract 68-DO-0007. In addition to the
use of canisters, multisorbent tubes will be used for sampling, and
the ion trap detector (Finnegan MAT, San Jose, CA) will be tested
for PVOCs applications.
169
-------
Anion Exchange Resin for Phenolic Compounds
The phenolic compounds are classified as acidic PVOCs. As
such, they tend to react in the presence of metal surfaces and
dissolve easily in water. A research program was initiated at
Battelle Memorial Institute under EPA Contract 68-02-4127 to
develop a specific sorbent method for phenolic compounds based upon
their acidic properties.
Laboratory work has shown that a strongly basic anion exchange
resin AG MP-1 (BioRad Laboratories, Richmond, CA) can be used to
quantitatively remove phenols from air and subsequently release
analyte ions via an ion exchange mechanism (16, 17). This method
to be effective for a variety of phenolic classes. So far, the
following compounds have been tested and have exhibited greater
than 80% collection/recovery efficiences:
Phenol 2-N02-phenol
2-CH3 -phenol 3-NO2-phenol
4-CH3 -phenol 6-CH3-2-N02-phenol
2,3-diCH3-phenol
2-Cl-phenol 2-OH-benzaldehyde
4-Cl-phenol 2-OH-biphenyl
2,4-diCl-phenol
PentaCl-phenol
The ultimate goal is to develop a fieldable sampling method
based upon resin-filled sorbent tubes that sample at 1 liter/min.
Subsequent laboratory analysis would be based upon a negative ion
chemical ionization mass spectrometric (NICI-MS) method specific
for a'group of phenolic analytes. It is expected that this method
will allow measurement of a wide variety of phenolic compounds and
have sub-part per billion by volume (ppbv) sensitivity.
Real-Time Trace Level Formaldehyde
Various schemes for real-time formaldehyde (HCHO) determi-
nations have been under investigation both at Battelle Memorial
Laboratories (EPA Contract 68-02-4127) and as in-house projects by
NSI Technology Services (EPA Contract 68-02-4444). These efforts
have produced three distinct types of sensitive real-time HCHO
monitors; the prototypes are now undergoing rigorous in-house
laboratory testing, limited field testing and intermethod
comparisons.
The development effort at Battelle has resulted in two
different monitor prototypes and two formaldehyde calibration
methods. The first monitor relies on an aqueous scrubber for HCHO
followed by fluorescence detection of the cyclization product of
ammonia, beta-diketone and HCHO (excitation at 254 nm, emission
around 520 nm). This prototype has been shown to have a detection
170
-------
limit of about 0.2 ppbv and a selectivity of 10,000:1 relative to
acetaldehyde, and 2,000:1 relative to benzaldehyde (18). The
second prototype monitor utilizes a spectroscopic method adapted
from commercial S02 monitors using direct fluorescence detection
(bandpass of 380-550 nm) through UV excitation (bandpass of 280-
350 nm) (18). The detection limit of this instrument is currently
about 50 ppbv, due to excessive background fluorescence. Opti-
mization of signal to background is planned subject to the results
of interference testing. The calibration methods rely upon dynamic
dilution of a known HCHO concentration in a stable flow. This
known concentration is generated either by the "porous Teflon
source," a purged teflon tube immersed in a dilute HCHO solution,
or by the "trioxane source,11 which relies on the sublimation of
solid trioxane prior to catalytic conversion to formaldehyde (19).
The in-house effort by NSI personnel produced a prototype
portable HCHO monitor based upon the CEA Instruments Model TGM 555
air monitor (CEA Instruments, Emerson, NJ) equipped with the
formaldehyde analytical module. This instrument detects 570 nm
color development due to the pararosaniline reaction with aqueous
formaldehyde in the presence of sodium sulfite. Sample air is
drawn into the instrument, scrubbed from the air stream into the
pararosaniline-containing aqueous solution and mixed with the
reagents. Sulfite is added and color development occurs as the
solution passes to the colorimeter. Specific details of the system
and its modifications are given in the literature (20) . The
monitor has been laboratory tested for real-time (with 16 min.
delay) operation in the 10- to 100- ppbv range and has been field
tested in two indoor air studies.
Glow Discharge-Mass Spectrometer Source
A novel mass spectrometric ionization source is currently
under development at the Illinois Institute of Technology -
Research Institute (IITRI, Chicago, IL) under EPA Contract
68-D8-0002. This source is a modification of pioneering work on
subatmospheric glow discharge ion sources by Glish and HcLuckey,
et al.t (21, 22) at Oak Ridge National Laboratories. The purpose of
this work is to utilize the speed, sensitivity and specificity of
tandem mass spectrometry for specific application to the PVOCs.
The glow discharge, as envisioned, will provide an instantaneous
source of analyte ions for presentation to the mass spectrometer.
No surface-induced reactions or adsorption is expected to corrupt
the sample. The capability of the tandem mass spectrometer for ion
separation and filtering obviates the necessity of a separation
method such as gas chromatography. The source currently under
construction is to be physically compatible with the ITMS mass
spectrometer (Finnigan MAT, San Jose, CA) at IITRI. Upon comple-
tion, the system will be tested using a variety of PVOC analytes.
Future applications include field screening at Superfund sites and
171
-------
exposure assessment from exhaled human breath analysis. This
methodology promises to be generally applicable, but requires the
very expensive tandem mass spectrometry capability and will require
on-site analysis.
Chemical lonization Ion Trap Detection
Existing methodology for the nonpolar VOCs requires both
drying of the sample and preconcentration of the analytes to
achieve adequate sensitivity in conventional quadrupole mass
spectrometers (see reference 1, and references therein). The PVOCs
pose special problems in such systems; the drying step affects
their quantitative throughput for analysis, and a reduction in
overall sample, thereby reducing cocollected water vapor, also
reduces sensitivity. Secondly, the conventional electron impact
sources of quadrupole instruments provide a "hard" ionization that
can excessively fragment the light PVOCs, thus reducing the
characteristic fingerprint of the molecular ion. To combat both
the problem of cocollected water and "hard" ionization, a project
that uses chemical ionization (CI) with water as the reagent gas
was initiated. The Finnigan Ion Trap Detector (ITD 800, Finnigan
MAT, San Jose, CA) was chosen for this project for its high
sensitivity and for the relative ease of implementation of water
CI. The work was performed at IITRI under EPA Contract No.
68-D8-0002.
This project involves only the analytical portion of PVOCs
determination and it is assumed that analytes can be presented to
the inlet of the system, either in-situ or via some unspecified
sampling medium, without adversely affecting their integrity.
Samples were prepared in zero-grade humidified air with a micro-
syringe driver, continuous injection system; only laboratory
testing was performed. The calibration methodology and the
development/evaluation of the ITD-based system are presented in an
EPA report (23) and other publications (24,25).
Results indicate that the analytical methodology is
appropriate for the PVOCs. The CI capability and the ITD's high
sensitivity allow full scan mode spectra at trace levels for many
PVOCs. Comparisons of electron impact (El) spectra and CI spectra
show that water CI is much more likely to preserve the high mass
(molecular ion, M, or protonated molecular ion, M+l) fragments;
therefore, the distinction among similar species is much clearer
than for El. Compounds in this category that have been tested
include the following:
ethylene oxide heptanal
3-octanone benzaldehyde
acetophenone ethanethiol
benzophenone 2-methyl-l-propanethiol
acrolein
172
-------
Two problems were encountered with this project. Certain
compounds were lost in the cryotrapping hardware despite the
removal of the dryer, and difficulty with the chromatography in the
presence of excess water occurred. In particular, the alkylthiol
compounds exhibited low recoveries and some lighter ketones were
not well resolved chromatographically. These issues will be
addressed in future work through the use of smaller sample sizes
and by the addition of novel concentration/inject ion hardware being
developed specifically for the ITD.
Atomic Emission Detector
Though mass spectrometry is generally considered the primary
identification tool for gas chromatography, often the identifica-
tion of unknowns is complicated when similar functional groups or
chemical structures (causing similar ionic fragmentation patterns)
occur in closely eluting species. Also, mass spectrometric frag-
mentation can result in neutral group losses that completely
obscure the identity of compounds. The atomic emission detector
(AED) is a gas chromatographic detector capable of detecting the
elemental composition of compounds, regardless of the molecular
structure. It operates by energizing GC column effluents in a
microwave plasma, thereby reducing the analytes to excited atoms.
As the excited atoms relax back to their lower energy levels, they
emit light at characteristic frequencies; this radiated energy is
dispersed and detected by a photodiode array. Up to four distinct
elements can be detected simultaneously with this system. Detailed
descriptions of this new instrumentation are available in the
literature (26-28). The AED is a complementary detector to an MS,
providing confirmatory identification information on atomic
constituents not generally possible with mass spectrometry.
An AED system (HP-5921A, Hewlett-Packard, Avondale, PA) has
been procured for application to PVOCs analysis and is operated by
NSI personnel under EPA Contract No. 68-02-4444 as an in-house
project. To date, the instrument has been installed and tested for
operation. It is planned to perform trace level work with a
-variety of PVOCs.
Calibration Methodology for PVOCs
Due to the varied reactivities of the PVOCs, each of the
various sub-classes may require a differing calibration methodology
at the trace level. The conventional method of dynamic dilution
has been tested for a subset of the compounds listed in the section
of this document entitled "TO-14 Extension." Further calibration
work was performed in-house using modified TO-14 hardware to
determine approximate responses and chromatographic parameters for
a variety of PVOCs, including:
173
-------
ethylene oxide methanol
propylene oxide trimethyl silanol
2-chloroethanol acetone
1-butanol pyridine
acetonitrile hexamethyl cyclotrisiloxane
acrylonitrile octamethyl cyclotetrasiloxane
tert-butyl alcohol vinyl acetate
1-propanol benzyl alcohol
2-butanone (MEK) nitrobenzene
ethyl acetate 4-phenyl cyclohexene
tetrahydrofuran butylated hydroxytoluene (BHT)
This work was presented at the 1989 Pittsburgh Conference
(29) . Additional in-house work is in progress for calibrating
permeation tube devices coupled to dynamic, static, and exponential
dilution of samples. The other projects discussed above have their
calibration methods presented as part of the cited references.
Valveless GC/MS Inlet
A current, in-house project is to develop a preconcentration
interface wherein the analyte cannot contact any surface other than
the interior of the column. This is based upon a sampling method
reported by Stephens (30) and upon work briefly described at the
1989 ACS meeting by Mouradian, et ai.t (32) and Arnold, et &l., (32).
A prototype inlet system was designed and built to be used
with the existing TO-14 equipment (HP-5880 GC and HP-5970 MSD,
Hewlett Packard, Palo Alto, CA). The inlet was fully automated and
tested, and then applied to a variety of odorous headspace samples
including coffee, fruits, consumer products, and plants. Initial
tests were successful in detection of the expected PVOCs, when
compared to published chromatograms of liquid extracts, or samples
collected on other media. As discussed earlier, the sensitivity
of this technique using the TO-14 analytical equipment was
restricted to analyte amounts in the 100 to 1000 nanogram range
because of cocollected water vapor; however, the combination of the
valveless GC/MS with the CI-ITD promises to provide the requisite
sub-nanogram sensitivity for ambient work.
Conclusions
The projects in progress in the Monitoring Methods Research
Section cover a wide variety of topics related to the determination
of PVOCs. It is anticipated that a subset of these projects will
provide specific methods for general use by the regulatory
community, analogous to the acceptance and implementation of the
TO-14 methodology for nonpolar VOCs. The MMRS continues to pursue
174
-------
PVOCs methods development and to support and sponsor scientific
exchange through the EPA/AWMA meetings and other similar forums.
Disclaimer
The research described in this article does not neces-
sarily reflect the views of the Agency and no official
endorsement should be inferred. Mention of trade names
or commercial products does not constitute endorsement
or recommendation for use.
References
1. U. S. Environmental Protection Agency Compendium of Methods
for the Determination of Toxic Organic Compounds in Ambient Air,
Method TO-14. U.S. Environmental Protection Agency, EPA-600/4-
84-041, Research Triangle Park, NC (1989).
2. Shuetzle, D., J. Lewtas, "Bioassay-Directed Chemical Analysis
in Environmental Research," Anal, chem., vol. 58, No. 11,
(1060A-1075A) September, 1986.
3. Bayona, J. M., K. E. Markides, M. L. Lee, "Characterization
of Polar Polycyclic Aromatic Compounds in a Heavy Duty Diesel
Exhaust Particulate by Capillary Column Gas Chromatography and
High-Resolution Mass Spectrometry," Environ. sd. Technoi., vol. 22,
No. 12, (1440-1446) 1988.
4. Nishioka, M. G. , C. C. Howard, D. A. Contos, L. M. Ball,
J. Lewtas, "Detection of Hydroxylated Nitro Aromatic and Hydro-
xylated Nitro Polycyclic Aromatic Compounds in an Ambient Air
Particulate Extract Using Bioassay-directed Fractionation,"
Environ. Sci. Technoi.f Vol. 22, No. 12, (908-915) 1988.
5. "Formaldehyde and Other Aldehydes," Committee on Aldehydes,
Board of Toxicology and Environmental Hazards, National Research
Council, National Academy Press Washington, DC, 1981.
6. Shane, B. S., "Human Reproductive Hazards," Environ. Sci.
Technoi., Vol. 23, No. 10, 1989.
7. "Toxicology and Carcinogenesis of Ethylene Oxide in B6C3F1
Mice (Inhalation Studies)," National Toxicology Program, Research
Triangle Park, NC, National Institutes of Health, NIH Publication
No. 88-2582, November, 1987.
8. Moore, J. G. , L. D. Jessop, D. N. Osborne, "Gas Chromato-
graphic and Mass Spectrometric Analysis of the Odor of Human
Feces," Gastroenterology, Vol. 93, No. 6, (1321-1329) 1987.
175
-------
FOURIER TRANSFORM INFRARED (FTIR) SPECTROSCOPY FOR
MONITORING POLAR AND LABILE AIRBORNE GASES AND VAPORS
SP Levine and Xiao H-k
Industrial Hygiene Program
University of Michigan
Ann Arbor, Michigan 48109-2029
and
T Pritchett and RD Turpin
US EPA (ERT)
Raritan Depot
Edison, NJ 08837
ABSTRACT
The FTIR has been evaluated for monitoring polar and
labile airborne gases and vapors. The limit of detection
(LOD) for FTIR when using a 10 meter closed path gas cell, a
HgCdTe or InSb-MCT detector, 2 cm"1 resolution, and least
sguares fitting (LSF) software is as low as 9 ppb for bis
(chloroethyl) ether in air. The LODs for other species, and
for these vapors in mixtures, is higher by a factor of up to
100. Software has been developed to identify unknown
compounds in the spectra of mixtures of organic vapors.
176
-------
INTRODUCTION
Efforts in our laboratory have centered on the
investigation of the optimal FTIR hardware, software and
data base for the near real time qualitative and
quantitative analysis of trace levels of airborne gases and
vapors (1-7). Much of this work parallels earlier work by
Herget (8-10) who has collaborated with us in this present
effort, and by Hanst (11).
Air monitoring applications areas have also been
delineated by our group, especially in the hazardous waste
site monitoring area (12,13). Other groups working with
FTIR (14), and with non-FTIR instruments have also defined
the needs for instrument development and application for air
monitoring (15-17).
It is the purpose of this paper to summarize the
results of the efforts in the FTIR air monitoring field on-
going in our laboratory in 1989-90, especially with respect
to the limit of detection (LOD) of polar and labile organic
vapors.
EXPERIMENTAL SECTION
Samples:
All mixtures of known and unknown samples of vapor
mixtures in air were supplied T. Pritchett, U.S. EPA
Environmental Response Team.
Hardware:
A Nicolet 20 SXB Fourier transform infrared
spectrometer was used with a liquid nitrogen cooled HgCdTe
(MCTJ-InSb sandwich detector and 10 meter Hanst gas cell.
Interferograms were collected to yield spectra of from 0.5
to 8.0 cm'1 resolution through deresolution of the 0.5 cm"1
interferograms by the method described in previous papers
(2-7). Sixty reference standard vapor IR spectra were
acquired as a reference library using GC-certified gas
cylinders (Scott Specialty Gas Co.)-
In addition, a University of Michigan-Nicolet prototype
"Baby-ROSE" (Remote Optical Sensing) FTIR has been
constructed for applications requiring a transportable
system with a pathlength of up to 30 meters. This system has
a resolution of 2 cm"1, and operates using an 80386/387 PC
and Nicolet-PC software.
Software:
For the detailed theory of LSF software, see papers by
Haaland, et al (18,19). Detailed discussions of the
177
-------
application of. LSF software to quantitative analysis of
trace gases in ambient air have also been published (2-7).
In addition, to quantitative analysis with LSF,
qualitative analysis was also performed using modified
Nicolet LSF software. This software performs iterative LSF
(ILSF) necessary for qualitative analysis.
DISCUSSION
Examination of the calibration spectra of polar and
labile compounds was made after dilution to 10 ppro, 1 ppm,
100 ppb and 10 ppb using a precision dilution manifold after
the design of Herget (1) as modified by Strang (2,3).
Investigation of the optimal treatment of the data to obtain
the lowest possible LOD was made by Strang (3).
Strang determined that, except for hydride gases, a
resolution of 2.0 cm"1 was optimal. Not only is there no
need for higher resolution, but higher resolutions are
counter-productive in that they require more expensive and
larger instruments, as well as longer computational time and
data storage capacity. However, for the hydride gases,
resolution of as high as 0.5 cm"1 may be needed.
The spectra of some of the polar and labile compounds
investigated in this study are shown in the figure.
The question of LODs obtainable using FTIR for
compounds in mixtures of up to 12 components was also
addressed by Ying (5,6). In most cases, there is a serious
degradation in LOD for each component of the mixture. For
example, benzene, in a 12 component solvent test mixture,
has an LOD in the range of 1-2 ppm. However, certain
analytes are less affected or unaffected by the presence of
solvent mixtures. This includes some of the Freons as well
as the hydride gases (2,3).
The result of determination of LODs for a few polar and
labile compounds is given below. The library of spectra
compounds in our files approximately numbers 70, over half
of which are polar. All spectra have been acquired at 0.5
cm"1 resolution, and all are at certified concentrations, or
dilutions of those concentrations. All dilutions were
performed using an orbital welded, electropolished dilution
system with precision pressure/vaccuum control.
178
-------
Limit of Detection(a)
Compound (ppb)(b)(ug in beam)(c)
acetone
acetaldehyde
acrylonitrile
arsine(d)
benzene
bis (chloroethyl) ether
chlorobenzene
3-chloropropene
o-chlorotoluene
2-butanone
diborane(d)
o-dichlorobenzene
1,4-dioxane
2-ethoxyethanol
ethylene oxide
isopropanol
nitrogen trifluoride
phosphine(d)
propylene oxide
pyridine
80
39
70
8
28
9
60
50
95
23
25
140
16
32
37
20
20
60
100
830
0.7
0.4
0.4
0.06
0.2
0.1
0.7
0.4
1.2
0.2
0.07
2.1
0.1
0.3
0.2
0.1
0.1
0.2
0.6
6.7
(a) LOD for individual compound in ambient air; LOD for
multi-component mixtures must be determined for each
mixture, and are likely to be 2-100 times higher.
(b) LOD for the indicated concentration in ppb @ 10 meter
path length.
(c) Weight of analyte in beam, assuming a total of 2.5
liters actually in beam path in cell used for calibration.
(d) Requires an inSb/MCT detector. All other compounds
require only a MCT detector.
Since the LOD is actually proportional to the product
of pathlength X concentration, longer path, remote sensing
(ROSE) systems may have lower LODs. In addition, these LODs
may be improved through the use of improved LSF software
with calibration included for the 41 organics listed as
target trace atmospheric gases by the EPA-RTP. The use of
iterative least squares fitting (ILSF) methods allows both
qualitative and quantitative analysis to be accomplished
under certain circumstances.
A description of the utility of the FTIR for air
monitoring, especially when contrasted to commercially
available filter-based systems, has also been published.
179
-------
ACKNOWLEDGEMENTS
The authors thank Ying Li shi whose invaluable
contributions to the FTIR and use of LSF for this
application is central to the success of this work. In
addition, we thank Dan Sparks and Bill Herget at Nicolet
Analytical Instrument Company for valuable guidance, and
Mary Weed for art work.
The authors thank the U. S. EPA, Environmental Response
Team (EPA-ERT) (research contract 68-03-3255, and related
contracts) and the Centers for Disease control (CDC-NIOSH)
(research grant 1-R01-02404) for their generous support.
REFERENCES
1. Herget, W.F. and S.P. Levine: Fourier Transform Infrared
(FTIR) Spectroscopy for Monitoring Semiconductor Process Gas
Emissions. Appl. Indus. Hyg. 1: 110 (1986).
2. Strang, C.R., S.P. Levine and W.F. Herget: Evaluation of
the Fourier Transform Infrared (FTIR) Spectrometer as a
Quantitative Air Monitor for Semiconductor Manufacturing
Process Emissions. Amer. Ind. Hyg. Assoc. J., 50: 70-77
(1989).
3. Strang, C.R. and S.P. Levine: The Limits of Detection for
the Monitoring of Semiconductor Manufacturing Gas and Vapor
Emissions by Fourier Transform Infrared (FTIR) Spectroscopy.-
Amer. Ind. Hyg. Assoc. J., 50: 78-83 (1989).
4, Ying, L-S, S.P. Levine and C.R. Strang: Fourier Transform
Infrared (FTIR) Spectroscopy for Monitoring Airborne Gases
and Vapors of Industrial Hygiene Concern. Amer. Ind. Hyg.
Assoc. J., 50: 354-359 (1989).
5. Ying, L-S and S.P. Levine: Evaluation of the
Applicability of Fourier Transform Infrared (FTIR)
Spectroscopy for Quantitation of the Components of Airborne
Solvent Vapors in Air. Amer. Ind. Hyg. Assoc. J., 50: 360-
365 (1989).
6. Ying, L-S and S.P. Levine: Fourier Transform Infrared
Least-Sguares Methods for the Quantitative Analysis of
Multi-Component Mixtures of Airborne Vapors of Industrial
Hygiene Concern. Anal. Chem. 61: 677-683 (1989).
7. Xiao H.K., S.P. Levine and J.B. D'Arcy: Iterative Least
Squares Fit Procedures for the Identification of Organic
Vapors Mixtures by Fourier Transform Infrared (FTIR)
Spectrophotometry. Anal. Chem. 61: 2708-2714 (1989).
180
-------
8. Herget, W.F.: Analysis of Gaseous Air Pollutants Using a
Mobile FTIR System. Am. Labs. 72 (1982).
9. Herget, W.F. and J.D. Brasher: Remote Optical Sensing of
Emission. Appl. Opt. 18: 3404 (1979).
10. Herget, W.F., J. Staab, H. Klingenberg, and W.J. Riedel:
Progress in the Prototype Development of the New
Multicomponent Exhaust Gas Sampling and Analyzing System.
Soc. Automot. Engin. Conference, Feb., 1984, paper no.
840470.
11. Hanst, P.L. IR-Spectroscopy of the Atmosphere. Frez.
Z. Anal. Chem.. 324: 579-588 (1986).
12. Levine, S.P., Costello, R.J., Geraci, C.L. and Conlin,
K.A.: Air Monitoring at the Drum Bulking Process of a
Hazardous Waste Remedial Action Site. Amer. Ind. Hyg. Assoc.
J. 46: 192 (1985).
13. Portable Instruments for Monitoring Airborne Emissions
from Hazardous Waste Sites. (Draft Document) Organ. Int.
Metrol. Legale., Pilot Secretariat 17, Reporting Secretariat
5 "Measurement of Pollution from Hazardous Waste Sites",
S.P. Levine, Chairman, ACGIH, Cincinnati, Ohio (1988).
14. small, G.W., R.T. Kroutil, J.T. Ditillo, and W.R.
Loerop. Detection of Atmospheric Pollutants by Direct
Analysis of Passive Fourier Transform Infrared
Interferograms. Anal. Chem. 60:264-269 (1988).
15. Burggraaf, P. Hazardous Gas Safety and the Role of
Monitoring. Semi. Intl. 56-62, Nov., 1987.
16. DeCorpo, J.J., J.R. Wyatt and F.E. Saalfeld: Central
Atmospheric Monitor. Naval Engin. J. 42: 231 (1980).
17. Helmers, C.T., Jr. Industrial Central Atmospheric
Monitor. Sensors, 20-25, Sept., 1984.
18. Haaland, D.M. and R.G. Easterling: Improved Sensitivity
of Infrared Spectroscopy by the Applications of Least
Squares Methods. Appl. Spectrosc. 34: 59 (1980).
19. Haaland, D.M., R.G. Easterling and D.A. Vopicka:
Multivariate Least Squares Methods Applied to the
Quantitative Spectral Analysis of Multicomponent Samples.
Appl. Spectrosc. 39: 73 (1985).
20. Levine, S.P., L.s. Ying, C.R. Strang and H.K. Xiao:
Advantages and Disadvantages in the use of FTIR and Filter
Infrared Spectrometers for Monitoring Airborne Gases and
Vapors of Industrial Hygiene Concern. Appl. Ind. Hyj. 4: 180
(1989).
181
-------
0.006 -i
0-
0.018'
O
z
CD
o:
o
CO
0
0.004
w
I]/""
Y ^v' f
1300 1122
944
WAVENUMBERS
766
Figure. Spectra of Acetaldehyde (A), N-Propanol (B),
and Pyridine (C) at 5 ppm in Ambient Air.
182
-------
OPTIMIZATION OF ANALYTICAL PROCEDURE FOR CHARACTERIZING
AMBIENT POLAR VOLATILE ORGANIC COMPOUNDS
Sydney M. Gordon, Michael Miller
IIT Research Institute
Chicago, IL 60616-3799
Joachim D. Pleil, William A. McClenny
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
The analysis of ambient levels of volatile organic compounds (VOCs)
in air usually involves cryogenic trapping of the analytes, followed by
thermal desorption and low-temperature refocussing onto a column for
analysis by high-resolution gas chromatography and electron impact mass
spectrometry (GC/MS). This approach has been widely used for nonpolar
VOCs, but the ambient water vapor co-collected with the sample causes
problems which have prevented Its use for more polar species.
Preliminary experiments have indicated that chemical 1onizat1on (CI)
GC/MS with the Ion Trap Detector™ (ITD) using water, including the water
present 1n the air, as the CI reagent 1s suitable for the measurement of
polar VOCs in whole air samples. Chemical ionization of the water gives
intense pseudomolecular ions for various polar compounds and enhanced
sensitivities in comparison with electron impact data. These experiments
also showed, however, that delivery of the compounds to the GC/ITD for
analysis, as well as calibration of the system, are aspects which demand
attention. Recent work has accordingly focussed on developing and testing
such methods for the generation of experimental data to evaluate the
approach for several target polar compounds of Interest, Including
alcohols, ketones, and thiols.
183
-------
Introduction
The need to develop better measurement techniques for polar volatile
organic compounds of low molecular weight has been spurred by Increasing
public concern over these compounds. Certain polar VOCs (e.g., alkyl-
thiols, amines, aldehydes, etc) are often the subject of nuisance
complaints from the public, while others (e.g., ethylene oxide, propylene
oxide, acroleln, etc) are sometimes cited as potentially toxic compounds
from Industrial emissions. However, the analysis of trace levels of polar
VOCs in air poses many problems, and the nature and distribution of these
compounds in the atmosphere cannot be adequately characterized at present.
The measurement of nonpolar VOCs in air is well established using
polished stainless steel sampling canisters together with cryogenic
trapping and gas chromatography/mass spectrometry.1 Cryogenic trapping
preconcentrates the analytes by condensing them on the cold surface of the
trap, after which they are desorbed and flushed onto the GC column for
analysis. Co-collected water vapor 1s dealt with by passing the sample
through a Nafion membrane dryer before 1t enters the GC.2 This technique
is, however, not generally applicable to small polar molecules since they
permeate the walls of the dryer tube along with the water vapor, and are
consequently lost.3
Most methods used to measure polar VOCs 1n air stem from Industrial
hygiene technology, and rely mainly on liquid Impinger sampling. As a
result, they are generally restricted 1n sensitivity to ppmv levels. Most
efforts to sample polar VOCs at trace levels have been largely ineffective
due to their greater chemical reactivity, affinity for surfaces, and
tendency to undergo polymerization.3
In a recent report, we proposed the use of chemical ionlzatlon GC/MS
in the quadrupole ion trap as a promising method for measuring polar VOCs
in whole air samples.4 The approach makes use of the water vapor present
in the air as the CI reagent. We showed that water is an effective CI
reagent gas in the ion trap, and reacts with high sensitivity with a range
of polar compounds. This paper describes an extension of this work 1n
which we have explored conditions for the use of GC/water CI-ITD, in the
full scan mode, for the analysis of selected polar VOCs in whole air
samples.
Experimental Methods
The system used was a cryogenic sampling unit (Nutech 3538-02)
coupled to a Varian 3400 GC and Finnigan MAT 800 ITD (Figure 1). Analyses
were performed using a nonpolar 30 m x 0.53 mm ID HP-1 methyl siHcone
column (0.88 urn film thickness), the outlet of which was connected to the
ITD via a heated transfer line. The ion trap was controlled by a COMPAQ
DESKPRO™ 386/20 personal computer using Automatic Reaction Control™
software for CI supplied by the manufacturer. For the CI experiments,
water in a reservoir attached to the reagent gas Inlet port was used to
augment the water concentration 1n the ion trap and maintain a constant
reagent gas pressure.
The test atmospheres used in this study were supplied by a vapor
generator system and included ketones, alcohols, thiols, nitrobenzene, and
three nonpolar compounds. The vapor generator is shown in Figure 2. It
made use of a SAGE micro-syringe pump and was designed to produce constant
concentrations of the compounds 1n dry or humidified air. The air was
humidifed by first passing through a flask partially filled with water.
184
-------
For all of the experiments described here, the relative humidity was
adjusted to about 35X. The output of the syringe was attached to a heated
septum, and the vapor concentration was carefully controlled by the
appropriate choice of flow rates of the air and the liquid mixture from the
syringe.
In order to check the response of the analytical system, the vapor
generator was calibrated using a sorbent preconcentration technique. These
measurements were made by connecting a Tenax cartridge at the outlet from
the vapor generator to the cryogenic trap. The gas flow through the
cartridge was maintained at 20 cmVmin, the flow rate of air typically used
when collecting a sample in the cryogenic trap. At the end of the 10-
minute sampling period, the Tenax cartridge was removed and analyzed by a
standard thermal desorption GC/MS technique.5 The GC/MS was calibrated
using Tenax cartridges loaded with the test mixture from a flash
evaporator.
Results And Discussion
The calibration experiments showed that Tenax cartridges together
with thermal desorption GC/MS analysis provide an effective and independent
method for measuring the concentrations of the selected polar compounds at
the outlet of the vapor generator. The peak area measurements for the
polar compounds loaded directly onto Tenax cartridges by flash evaporation
show a level of precision (±10* relative standard deviation) comparable to
that obtained for nonpolar compounds by this technique.5 The same approach
was used to establish the peak area reproducibility at the outlet of the
vapor generator. Although the precision for the polar compounds 1s not as
high as before (±16X relative standard deviation), the values are neverthe-
less still quite acceptable and similar to those for the nonpolar compounds
in the test mixture.
In order to evaluate the trapping and recovery efficiency of the
cryogenic Interface, the vapor generator was attached to the interface and
samples were collected on Tenax cartridges at the cryotrap outlet. The
cartridges were again analyzed by thermal desorption GC/MS. The results
are presented in Table 1. The recoveries obtained are very good for all of
the compounds, except for the thiols and naphthalene. These results
indicate that the water vapor present does not have a deleterious effect on
the behavior of these polar compounds in the cryotrap. The hot surfaces to
which the compounds are exposed within the cryotrap during the transfer
operation to the cartridges may be responsible for the virtual disappear-
ance of the thiols. The low recovery measured for naphthalene is somewhat
surprising in view of its lack of polarity.
The final consideration in these tests was that the water vapor 1n
the humidified air stream might have an adverse effect on analytical
performance. To explore this possibility, the output from the cryogenic
interface was attached directly to the GC/CI-ITD system and the experiments
were repeated. The results are also Included in Table 1. An examination
of the data shows that, in contrast to the previous situation, all of the
measured recoveries are very low, except for the nonpolar compounds and
nitrobenzene. The alcohols in the test mixture gave "smeared-out" peaks
which could not be quantitated. These low recoveries Indicate a severe
humidity effect on chromatographic performance when the water vapor in the
sample is allowed to pass directly to the GC column.
One way of reducing the deleterious effect of water vapor on
chromatographic performance may be to decrease the sample size and take
185
-------
advantage of the high sensitivity of the ITD 1n the water CI mode.4 To
test this approach, the vapor generator was adjusted to give a test
atmosphere containing the VOCs at a constant concentration, and experiments
were performed 1n which the volume of air sampled in the cryogenic trap was
systematically decreased from 250 cm3 to 50 cm3. The results are presented
1n Figure 3 for the polar compounds nitrobenzene, acetone, and 1-butanol,
and the nonpolar p-d1chlorobenzene. Except for 1-butanol, which shows a
fairly rapid decrease 1n signal response with sample volume, the plots for
the remaining polar compounds suggest that 1t is feasible to use much
smaller samples with water CI-ITD as a means of reducing the adverse
effects of water vapor on system performance.
Conclusions
The ITD has several features that combine to make it an Ideal
detector for the analysis of polar VOCs 1n air. Among these, the most
important are its chemical ionlzation capability and Its high sensitivity
1n the full-scan mode. This study has shown that whole-air samples can be
effectively analyzed with the ITD by exploiting the water vapor present in
the air, and a number of polar compounds can be transferred through the
cryogenic preconcentrator without significant loss. Although the water
vapor seriously degrades chromatographlc performance, this problem can be
alleviated to a large extent by using smaller air samples and taking
advantage of the ion trap's sensitivity.
Disclaimer
Although the research described was funded by the U.S. Environmental
Protection Agency through Contract 68-D8-0002 to IIT Research Institute,
this document has not been subjected to Agency review and does not
necessarily reflect the views of the Agency. No official endorsement
should be inferred by its Inclusion in these proceedings.
References
1. W. A. McClenny, J. D. Pleii, M. W. Holdren and R. N. Smith, "Automated
cryogenic preconcentration and gas chromatographlc determination of
volatile organic compounds 1n air," Anal. Chem. 56: 2947 (1984).
2. J. D. Plell, K. D. Oliver and W. A, McClenny, "Enhanced performance of
Naflon dryers in removing water from air samples prior to gas
chromatographlc analysis," JAPCA, 37: 244 (1987).
3. J. B. Clements and R. G. Lewis, "Sampling for organic compounds," in
Principles of Environmental SamplingT L. H. Keith, Ed., American
Chemical Society, Washington, DC. 1987, pp. 287-296.
4. S. M. Gordon, M. Miller, J. D. Plell, W. A. McClenny and R. G. Lewis,
"Evaluation of chemical 1onization-1on trap detector for the analysis
of ambient polar volatile organic compounds," presented at the 82nd
Annual Meeting & Exhibition, Air & Waste Management Association,
Anaheim, California, June 25-30, 1989.
5. K. J. Krost, E. D. Pellizzari, S. G. Walburn and S. A. Hubbard,
"Collection and analysis of hazardous organic emissions," Anal. Chem.
54: 810 (1982).
186
-------
Table 1. Recovery of polar and nonpolar compounds In air from cryogenic
trap, measured using Tenax cartridges and by direct analysis with
GC/ITD
Percent Recovery*
Compound
Cryotrap-Tenax
Cryotrap-GC/ITD
Acetone
1-Propaneth1ol
2-Methyl-1-Propanethiol
2-Butanone
2-Propanol
Ethanol
1-Butaneth1ol
1-Butanol
Nitrobenzene
1,1, 1-Tr1chloroethane
p-D1chlorobenzene
Naphthalene
117
12
1
103
125
119
ndc
103
92
123
103
57
nqb
12
25
44
nq
nq
4
nq
99
59
114
105
» Average of three experiments.
b nq = Not quantltated; peak observed but "smeared-out"
c nd = Not detected
GC Column
Figure 1. Cryotrap and GC/CI-ITD system for analysis of VOCs
1n air.
187
-------
Syringe Pump
Needle
Valve
Heated Zone
To
Cryogenic
Trap and
Analytical
System
Flow
Controller
Humidifier
Vent
15
12--
in
o
5. 9-1-
o
6--
Figure 2. Schematic of vapor generator.
• Nitrobenzene
A Acetone
A 1-Sutanol
o p-Dichlorobenzene
100 200
Air Volume (ml)
300
Figure 3. Variation of peak area with sample volume for selected polar
and nonpolar VOCs.
188
-------
ANALYSIS OF POLAR SEMIVOLATTLE ORGANIC COMPOUNDS BY LIQUID
CHROMATOGRAPHY/MASS SPECTROMETRY
Robert S. Whiton
NSI Environmental Sciences
P. O. Box 12313
Research Triangle Park, NC 27709
Nancy K Wilson
Atmospheric Research and Exposure Assessment Laboratory
United States Environmental Protection Agency
Research Triangle Park, NC 27711
Nonpolar semivolatile organic compounds (SVOCs) are routinely analyzed by
gas chromatography/mass spectrometry, but the technique has severe limitations when
polar SVOCs and nonvolatile organic compounds are the analytical targets. Polar
compounds may present difficulties due to thermal instability, insufficient volatility,
and adsorption in the chromatographic system, creating a need to develop methods
based on liquid chromatography/mass spectrometry (LC/MS). Experiments involving
the analysis of polar derivatives of polycyclic aromatic hydrocarbons by particle beam
LC/MS demonstrate the ability of the instrument to generate useable electron impact
and electron capture negative ionization spectra for some compound classes and
illustrate the problems associated with analysis of hydroxy-aromatic compounds.
189
-------
Introduction
The need to analyze airborne paniculate matter for polar organic compounds
has been demonstrated by groups working with bioassay-directed fractionation, who
have found substantial mutagenic activity in the polar fractions of extracts.' The
difficulties inherent in gas chromatographic (GC) analysis of polar semivolatile
organic compounds (SVOCs; i.e., thermal instability, adsorption, and insufficient
volatility) have hindered efforts to characterize these polar fractions, spurring our
interest in liquid chromatography/mass spectrometry (LC/MS).
We are currently evaluating a commercial particle beam LC/MS system for
use in the analysis of polar SVOCs, especially substituted polycyclic aromatic
hydrocarbons (PAHs), in airborne particulate matter. The particle beam interface,
developed by Willoughby and Browner2 and improved by Winkler si al-,3 is based on
a two-stage momentum separator. The LC effluent is nebulized in a stream of
helium, and the solvent evaporates from the droplets in a desolvation chamber at
reduced pressure. The solvent vapor and analytes then pass through the momentum
separator, where most of the solvent is pumped away, and the analytes are transferred
to the ion source of the mass spectrometer as a stream of dry particles. The apparent
advantage of the particle beam interface is solvent removal without exposure of the
analytes to heated surfaces, which might cause decomposition of sensitive compounds.
Solvent removal allows the use of the standard modes of ionization, including electron
impact (El) and chemical ionization (CI) in an unmodified MS ion source.
This paper highlights the results of our preliminary studies with polar PAH
derivatives. We have examined optimization of the particle beam operating
parameters and investigated the potential for analysis of substituted PAHs by both El
and electron capture negative ionization (ECNI) LC/MS.
Experimental Methods
All experiments were performed on a Hewlett-Packard (HP; Palo Alto, CA)
5988A mass spectrometer equipped with a switchable EI/CI ion source, negative ion
detection, and an HP 59980A particle beam LC/MS interface. Negative ion
experiments were performed with either methane or carbon dioxide to moderate the
gas introduced through the particle beam interface. The high-performance liquid
chromatograph (HPLC) was an HP 1090 with ternary-gradient solvent delivery system
and a diode array detector, which was used in line with the particle beam interface.
The HPLC columns were Brownlee Labs (Santa Clara, CA) 10-cm x 2.1-mm Spheri-5
cyano columns with 5-pm particles. The mobile phase flow rate was 0.3 ml/min.
Results and Discussion
There are several operational parameters which can be adjusted to optimize
the performance of the particle beam LC/MS for a specific analysis. These are (1)
LC mobile phase composition, (2) interface variables (nebulizer gas flow, nebulizer
capillary position, and desolvation chamber temperature), and (3) ion source
temperature. The dependence of sensitivity on mobile phase composition was noted
by Willoughby and Browner.2 We have confirmed that, for PAHs, organic solvents
yield better sensitivity than water-organic blends and that hexane and methanol yield
slightly better sensitivity than acetonitrile. Hexane and acetonitrile produce slightly
less chromatographic peak tailing than does methanol. The optimum nebulizer
capillary position is highly dependent on both the analyte and the solvent
composition, whereas the nebulizer gas flow is less so, and both must be determined
190
-------
experimentally. The desolvation chamber temperature does not appear to be a
critical variable. The ion source temperature has strong effects on sensitivity,
chromatographic peak shape, and spectral profile for the PAH derivatives in both the
El and CI modes; temperatures of 250 *C to 300 °C are optimum for El operation
with compounds up to molecular weight 300. At lower temperatures, sensitivity is
reduced, and the less volatile compounds show chromatographic peak tailing.
We have obtained El and ECNI spectra of a variety of substituted PAHs,
including hydroxy-, nitro-, and amino-PAHs, and benz[a]acridine. The El responses
for hydroxy-PAHs are considerably weaker than for nitro- and amino-PAHs and
unsubstituted PAHs. At an El source temperature of 300 *C, there may be
considerably more fragmentation than found in GC/MS or standard library spectra
for some compounds. Under these conditions, the nitro-PAHs produce little or no
molecular ion signal (Figure 1).
The CO2-ECNI spectra of nitro-PAHs are more predictable, containing
primarily molecular ions and small [M-Oj" peaks (Figure 2). The limit of detection
for the nitro-PAHs is approximately 1-2 ng by selected ion monitoring and 10 ng full
scan, but the response under all conditions is nonlinear, severely curtailing the
potential for quantitative analysis.
The hydroxy-nitro-PAHs tested by ECNI were not detected at levels from 100
to 150 ng. That is consistent with the poor performance of the interface with other
hydroxy-aromatics. Hydroxy-PAHs, pentachlorophenol, and Dinoseb all show reduced
response relative to the nonhydroxylated analogs. For example, the El detection limit
for pentachlorophenol is approximately 2.5 /*g on column versus 20 ng for
pentachlorobenzene. In addition, the hydroxy-PAHs show significant tailing and
carry-over, indicative of adsorption in the particle beam interface, which could not be
alleviated by addition of acetic acid to the mobile phase.
Conclusions
The particle beam LC/MS can generate useful El and ECNI spectra of PAHs
and substituted PAHs, although the degree to which the El spectra match the
reference libraries is variable. The difficulty in analyzing hydroxy-aromatic
compounds indicates the need for further development of the particle beam interface.
References
1. D. Schuetzle, J. Lewtas, "Bioassay-directed chemical analysis in environmental
research," Anal. Chem. 5_8_: 1061A (1986).
2. R. C. Willoughby, R. F. Browner, "Monodisperse aerosol generation interface for
combining liquid chromatography with mass spectroscopy," Anal. Chem. 5_& 2626
(1984).
3. P. C. Winkler, D. D. Perkins, W. K. Williams, R. F. Browner, "Performance of an
improved monodisperse aerosol generation interface for liquid chromatogTaphy/mass
spectrometry," Anal. Chem. 6JJ: 489 (1988).
191
-------
A.
1C
B.
[M-NO] +
217
[M-CNO2]+
189
108
j
)0
[""I""!1"1!""
120 140
160 180
'
r
218
—
r- 100
E-90
f 80
j-70
j-60
E- 50
E-40
5-30
f20
=-10
MM llll|ll |
-------
A.
Time
B.
24(M-
231
248
249
100
90
80
70
60
50
40
30
20
10
0
180 190 200
210 220 230 240 250 260
Mass/Charge
Figure 2. A. ECNI-LC/MS TIC of repeated injections of 3-nitrofluoranthene.
B. Mass spectrum of 20 ng of 3-nitrofluoranthene.
193
-------
ANALYSIS OF VOLATILE ORGANICS IN AIR
VIA WATER METHODS
J. H. Myron Stephenson, Chief
Frank Allen, Tim Slagle
GC/MS Unit
Organic Chemistry Section
Laboratory services Branch
Environmental Services Division
U.S. EPA, Region IV
College Station Rd.
Athens, GA 30613
A modified compendium method TO14 for volatile organic compounds in
air is used in the EPA Region IV laboratory. This method is a combination
of EPA volatile organic water methods (i.e. 524.2, 8240) and method TO14
for the analysis of air canisters. A standard water purge and trap device
with a solid sorbent trap is used. Air standards containing 63 analytes
plus 3 internal standards and 3 surrogates are prepared in canisters using
the water method purging vessel. The analysis is by a gas
chromatograph/mass spectrometer (GC/MS) operating in full scan mode.
Cryofocusing of the gas chromatograph oven is necessary only for analysis
of the very volatile gases.
194
-------
Introduction
The EPA Region IV laboratory has been involved in the analysis of
volatile organics in water and soil/sediment samples for more than ten
years. These analyses are performed by standard EPA protocols for
programs such as NPDES permits, drinking water regulations, and OSWER
regulations. Recently the laboratory was requested to analyze this same
group of volatile analytes in air samples. All of our GC/MS systems were
devoted to the analysis of samples from other media; therefore, a protocol
was developed to utilize existing instrumentation to perform these
analyses with a minimum of disruption. It was decided that using 6 liter
air canisters and Compendium Method TO14 with minor modifications would
best meet Region IVs immediate needs. Four separate techniques were
developed for the entire analytical system. These involved cleaning
canisters for laboratory blank, standard, and audit material use;
preparing standards mixtures in canisters; sample canister preparation for
analysis; and developing the analytical system. (See Figure 1 and Figure 2
for an overall view of the analytical system.}
Experimental
Canister Cleaning Procedure
The most effective and time efficient canister cleaning technique
involves a two stage process. The following equipment was used:
1. A direct drive mechanical pump with a free air displacement of 9
m /h with molecular sieve trap on the inlet side of the pump.
2. A rheostat controlled hot plate.
3. An ultra high purity nitrogen cylinder with a stainless steel
diaphragm and a humidifier vessel plumbed in the line to the canister
with 1/16 in. stainless steel tubing.
4. A 0-1500 mm Hg air gauge calibrated to read Absolute pressure.
The canister is first evacuated to less than 25mm Hg Absolute. Then it is
pressurized with humidified nitrogen to 1500 mm Hg. The canister is then
placed on a warm hot plate and evacuated for 15 minutes. The canister is
pressurized again, then re-evacuated for another 15 minutes while still on
the hot plate. The canister is removed from the hot plate, pressurized to
1500 mm Hg, then re-evacuated until the canister is cool (approx 15 min).
The canister at this point should read less than 25 mm Hg and is held
under vacuum until used.
Calibrated Standard Canister Preparation
Utilizing knowledge gained from sparging volatile organic analytes
from water samples onto a purge and trap system, a sparging system was
developed to purge these same volatile organics from water into an air
canister. The following equipment was used for preparing analytical
standard mixtures of volatile organics in canisters:
1. Zero grade air cylinder with stainless steel diaphragm and 1/16 in.
stainless steel tubing.
2. A 5 ml needle sparging vessel, used in sparging water samples,
plumbed with 1/16 in. stainless steel tubing to a pressure flow
controller on the nitrogen cylinder used in canister cleaning.
3. Standard solution of volatile organics in methanol (currently a 63
analyte mix).
4. A methanol solution containing 3 internal standards and 3
surrogates.
5. 6 liter stainless steel air canister.
6. 5 ml gas tight syringe.
7. 25 ul gas tight syringe.
8. Heat wrap for sparge vessel.
A previously cleaned canister still under vacuum is pressurized to 760 mm
195
-------
Hg Absolute (1 atm) with zero grade air. The air gauge is removed from
the canister and the sparging vessel is connected to the canister with the
shortest length of 1/16 in. stainless steel tubing possible. (Extra
efforts were made to minimize possible areas of dead volume to maximize
transfer of analytes from the water to the canister.) Five milliliters of
water is spiked with the standard solution and the internal
standard/surrogate solution. This water is transferred into the sparge
vessel and the water is purged with nitrogen for 10 min at 100 cm /min
while being heated at 40 °C. At the end of 10 minutes the sparge vessel
is removed from the canister, the air gauge re-installed and the canister
is pressurized with pure nitrogen to 1500 mm Hg Absolute (approx 15 psi).
The canister is allowed to equilibrate overnight before use. (See Figure
3 for an overview of the canister preparation technique.}
Sample Canister Preparation
The sample collection protocol requires that the canisters return to
the lab at near but less than 760 mm Hg Absolute (approx 1 atm). At this
point the sample canister is prepared as the standard canister is,
starting after the pressurization to 760 mm Hg with air. The one
exception is that the 63 analyte standard mixture is not added to the
water; only the internal standard/surrogate solution is added. Samples
are allowed to equilibrate a minimum of 2 hours after pressurizing before
analysis. (See Figure 4 for results of equilibration study.)
Analytical System Description
The analytical system is a combination of EPA volatile organic water
methods (i.e. 524.2, 8240) and method TO14 for the analysis of air
canisters. A standard water purge and trap device with a solid sorbent
trap is used. The analysis is by a gas chromatograph/roass spectrometer
(GC/MS) utilizing a capillary column and operating in full scan mode.
Cryofocusing of the gas chromatographic oven with liquid nitrogen is
necessary if the analysis of the very volatile gases is required. The
following equipment is used:
1. Automated water and soil/sediment purge and trap equipment.
2. Solid sorbent trap consisting of two hydrophobic adsorbent beds
(200 mg Carbopack B and 50 rag Carbosieve S-III, both 60/80 mesh) which
minimize water vapor collected during the canister subsampling.
3. Heated transfer line (1/16 in. stainless steel tubing with 1/16 in.
stainless steel tee at canister end).
4. GC/MS system with 60 m megabore column, liquid nitrogen cooling,
and full scan mass spectrometer mode.
5. Small diaphragm type vacuum pump (approx 0.25 m /h displacement).
6. Mass flow controller calibrated to handle up to 100 cm /min.
A transfer line heated to approx 100 °C is connected to the purge outlet
on one of the ports of the automated purge apparatus. A 1/16 in.
stainless steel tube is connected on one end to the purge flow inlet of
the automated purge apparatus and on the other end to a cut-off valve at
the tee on the canister end of the heated transfer line. The tee is
connected directly to the canister with an adaptor fitting. The trap vent
outlet on the purge and trap system is connected to a mass flow controller
set at 100 cm /min. The small vacuum pump is connected downstream of the
mass flow controller to maintain a positive pressure differential across
the flow controller. (See Figure 2 for an overview of the analytical
system.)
Air Canister Analysis
The cut-off valve on the purge inlet line should be closed during the
purge mode but open during the desorb and bake modes. This flow flushes
the transfer line during the bake cycle. With the mass flow controller
196
-------
set to allow a flow of 100 cm3/min, the canister valve is opened and the
purge mode is started on the purge and trap system. The purge mode is
maintained for 10 min giving a total volume of 1000 cm3 (1 liter). The
trap temperature is maintained at less than 30°C during the purge mode.
The gas chromatographic oven is cooled to -50 C and the desorb mode is
started. The oven is programmed up to 180°C at 8°C/min. The mass
spectrometer is scanned from 35-300 amu at 1 amu/sec. A standard curve
is generated from canisters prepared from a concentration range of 5 ug/m
to 150 ug/m . A daily standard is analyzed at 50 ug/m . This daily
standard is compared to the original curve as a quality control measure.
Internal standard response is monitored throughout the day and surrogate
recoveries are monitored in every analysis. A laboratory blank canister
is analyzed daily to monitor laboratory contamination. The quality
assurance measures used in the Superfund Contract Laboratory Program
statement of work for multi-level multi-media are followed very closely.
Results
The precision of the method including canister preparation was
demonstrated to be equivalent to the EPA water methods based on seven
replicate analyses. The highest %RSD of the seven replicates of the 63
analytes was 7.3 for methyl ethyl ketone. These seven replicates came
from three separate canisters prepared at nominally 50 ug/m per analyte.
The curve data was equally good with only one analyte having a %RSD
greater than 20 (1,1,2-trichloroethane, 21 %RSD) in the curve. The %RSD
of the majority of analytes was less than 10. (See Figures 5 & 6.)
Laboratory blank analysis continues to be a problem area at the 5-20
ug/m range for a few compounds. Acetone and methyl ethyl ketone are
persistent contaminants in the reagent water used for sparging and
humidifying the canisters. Naphthalene tends to carry over from high
concentration samples (>50 ug/m3} into the next analysis. Extra efforts
must be made to eliminate higher boiling point analytes from the system.
Five audit canisters prepared by EPA,RTF were analyzed to determine
accuracy. (See Figures 7 & 8.) Of the 28 analytes present in the audits
only 3 could be considered outside of normal quality control ranges.
Vinyl chloride in two of the audits was in the 125-150 % recovery range.
The problem with this data was caused by an inaccurate concentration in
the methanol solution used to prepare the standard canister and not by the
air analysis. Acetone and methyl ethyl ketone were in the 150-300 %
recovery range. This is not surprising since the audits were in the 10
ug/m range and the laboratory blanks contain 10-20 ug/m of these
analytes. With a purer quality of water, these analyte results should
improve.
Conclusion
This method allows Region IV to utilize the same instrumentation for
air, water, and soil analyses. The target analyte list is the same for
the different media and includes several "polar" water soluble organic
compounds. The potential of switching analyses between media on the same
instrument with minimal difficulty is an obvious time as well as cost
advantage. Few interferences are noticed on the GC/MS system and the
precision and accuracy of this technique indicate equivalency to a
cryofocusing trap for air analysis as well as water analysis. A limited
evaluation of the standards preparation procedure gives encouraging
results for the utility of these techniques for an easy and inexpensive
alternative for standard preparation.
197
-------
USEPA REGION IV ANALYTICAL METHOD
FOR VOC'S IN AIR CANISTERS
• initial Prmun Chock*d
•Confitar Praaurti«4 i« -15 PSI
WcUr Purg«
• Intmal SM Addod
ChromatoBraphv/
• Cryofoomlng • -50 C
Tun. Wllh FC43
Verily W«WB A**lgnin«iil with p-BfB
InUrnol Std QuonlllrHon
Full Sain Mod* lor S3 Tang* Analytn
Uethonol Stomfard Solution Uud
UOL'S 1-10 PPftV
EPA REGION IV AIR CANISTER ANALYSIS
Figure 1
Figure 2
REGION !V AIR CANISTER PREPARATION
AUDIT CANISTER PREPARED 02-12-90 14:25
Equilibration Study
mie
VALUE
IliU
COHC
MMCklCTHAHE
\.i-otCHiatattHtnf.
•ETHVLfW MLHWC
r-1 J-04CHL0MCTWNC
1,1-tMCHUMtOCTtUkNC
ICTHfi. HMYL KETOHC
11.1
7.0
I J ^1-TtlCHLOMCTHUC
UJtKH TETUCHLWOH
CMZEM:
[RtCHLMOCTWHE
IJ EMCHUMWHOTMIE
IClUEME
TTRMHUHUCTHEHE
L KNIEK
i-KTU*
mrtoic
Figure 3
Figure 4
198
-------
AIR CURVE FOR CHLORINATED HYDROCARBONS
(6-160 UQ/M3)
USEM REGION IV
5 10 19 20
% RELATIVE STANDARD DEVIATION
AIR CURVE FOR AROMATICS & POLARS
(6-160 UO/M3)
U8EM REGION IV
2 4 0 a 10 11
* RELATIVE STANDARD DEVIATION
Figure 5
Figure 6
AIR AUDIT RESULTS
CHLORINATED HYDROCARBON ANALYTES
(8-60 U8/M3)
10 40 M 60 100 120 140 1M IK)
I UICHT )
I AUDIT I
I MIHT >
I Mint 4
AIR AUDIT RESULTS
AROMATIC t POLAR ANALYTES
(6-60 UO/US)
IMhf I >H>rl
AUDIT I
AUDIT 6
Figure 7
Figure 8
199
-------
THE 1989 DELAWARE SITE STUDY
William A. McClenny
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
J. J. Kliment and T. V. Brixen
Delaware Department of Natural Resources and Environmental Control
The Delaware SITE (Superfund Innovative Technology Evaluation)
Study was carried out by a personnel from the Atmospheric Research
and Exposure Assessment Laboratory (AREAL), U. S. EPA and AREAL
contractors Battelle Columbus Laboratory, NSI Environmental
Sciences and Tecan Remote, Incorporated. Personnel of the Delaware
Department of Natural Resources and Environmental Control (DNREC)
hosted the operation and obtained permission to use local sites for
monitoring. The objective of the study was to field test several
monitoring methods that have progressed through a feasibility
testing stage and appear ready for predemonstration testing at
Superfund sites. Monitoring occurred near four Superfund sites in
the vicinity of New Castle, Delaware and at the Delaware Reclama-
tion Plant located just north of the Delaware Memorial Bridge.
Several different types of new monitoring equipment were deployed
including: (1) automated gas chromatographs; (2) canister-based
sequential samplers; (3) personal sampling devices; (4) canister-
based sector samplers; (5) long path infrared transmission
monitors; and (6) solid sorbant-based phenolic compound samplers.
200
-------
Introduction
The Superfund Innovative Technology Evaluation (SITE) Progran
promotes the development, acceptance and use of promising innova-
tive technologies capable of meeting the objectives of the overall
Superfund Program. One objective of the program is to provide the
means for developers of technology to demonstrate innovative
technologies at Superfund sites as alternatives to the systems
currently in use. A second objective is to provide support to
stimulate the development of promising concepts and technologies
to the point that they are suitable for demonstration at Superfund
sites. In the case of the Delaware SITE program, a proposal was
made by AREAL to the SITE program coordinator to carry out
predemonstration testing of new monitoring techniques. After
acceptance, a schedule was established for the testing in the
summer of 1989.
The schedule of the SITE project consisted of the following
items:
• Selection of a field site
• Predemonstration plan preparation
a. Planning/coordination meeting with participants; and
b. Preparation of plans for predemonstration testing.
1. Documentation of experimental design; and
2. Preparation and clearance of a quality assurance
plan.
Preliminary field screening study
Preparation of a target compounds list for each test site
Performance of the field testing
Review of data and preparation of a highlight report
Presentation of preliminary data at the 1990 EPA/AWMA
Symposium on Measurement of Toxic and Related Air
Pollutants
• Development of information products for the SITE program
The site proposed by J. J. Kliment of the Delaware Department
of Natural Resources and Environmental Control (DNREC) was
selected. The site was near New Castle, Delaware and included four
Superfund sites and one site near the Delaware Reclamation Plant.
A map of all the sites is shown in Figure I and detailed maps of
aill four sites are shown in the Figures II-IV. Figure I shows the
Superfund sites: Army Creek, Delaware Sand and Gravel, Halby, and
Standard Chlorine along with the Pigeon Point site which is near
the Delaware Solid Waste Authority. The location of a monitoring
site in the residential area referred to as Llangollen Estates and
the P4 monitoring site maintained by the state of Delaware are also
shown. The Standard Chlorine plant site was part of a large
201
-------
industrial complex. Figure II shows some detail for the Army
Creek, Delaware Sand and Gravel and the Llangollen Estates site.
Figure III shows the industrial complex of which Standard Chlorine
(location 12) is a part. Two state maintained monitoring sites
marked S-20 and S-8 are shown on the map near Delaware City.
Figure IV shows the Halby and Pigeon Point sites.
Experimental Plan
The experimental plan called for five methods of monitoring
to be tested in the field. An automated gas chromatograph (GC)
using a solid sorbant to preconcentrate volatile organic compounds
(VOCs) from the ambient air was placed in the P4 station. The unit
cycled through an automated sample and analysis sequence every
hour. The objective for this monitoring method was to evaluate the
use of a solid sorbant as a VOC preconcentrator and further to
demonstrate the use of automated GCs as a means to establish the
variability of VOC concentrations in time.
The P4 site was also used for placement of sector samplers.
These units were placed at the P4 and S20 sites to characterize the
industrial complex near Standard Chlorine. Data was also taken
near the Army Creek and Delaware Sand and Gravel site using the
sector samplers.
Personal sampling devices (PSDs) were used as fenceline
dosimeters at several of the sites. These units sample by
diffusion, are small and convenient to use. Since the units are
less than two inches in diameter, they can be attached to any
convenient structure and used to obtain a time-integrated ambient
air sample. To establish the validity of the fenceline monitors
for screening of VOCs, they were used in side-by-side tests with
canisters.
Long path monitors based on selective absorption of infrared
radiation by target gases, were used at four of the sites. The
objective was to define the field capabilities of FTIR-based long
path monitors. These monitors use source and receiver at one end
of a monitoring path and a retro-reflector at the other end. The
pathlength is typically 300 m long, giving a total pathlength of
600 m.
New solid sorbents specifically chosen for retention of
phenolic compounds were also used at the Superfund sites. However,
the field study was intended as a scoping study in this case since
only laboratory studies had been done up to this point.
Screening Studies
A screening field study was performed at the Superfund sites
in preparation for the main field study. Duplicate canister
samples were taken at several locations simultaneously with the
202
-------
operation of a Photovac Model 10S70 gas chromatograph. One
canister was sent to each of two laboratories for analysis.
Comparisons of the results of analysis have been given previously
(1) . Table I gives the results of a field screening analysis taken
April 6, 1990, near a plant producing aromatic hydrocarbons. Other
similar screening analyses at other sites gave significantly lower
concentrations but were effective in characterizing the sites.
By using the results of screening analysis and the information
obtained from emission inventories maintained by the state and from
the SARA Title III disclosure statements, a target compound set was
established for each site. Table II shows an example of a site
specific target compound list. The field monitoring capabilities
anticipated during the field study for the target compounds are
indicated also.
Field Study
The main field study was carried out during the period 24 July
through 9 August 1990 at the Superfund sites as planned.
Individual experiments and experimental results were explained as
part of the session on "The Delaware SITE Study" and are included
in this Proceedings as the following six papers. One of the most
interesting conclusions is that the dominant concentrations noted
in monitoring close to "old" Superfund sites were due to local
(within 5 km) industrial emissions.
Acknowledgements
The authors wish to thank members of the DNREC and of the
AREAL management for their support in carrying out this project.
Disclaimer
The information in this document has been funded in part by
the U. S. Environmental Protection Agency. It has been subjected
to Agency review and approved for publication. Mention of trade
names and commercial products does not constitute endorsement or
recommendation for use.
References
Richard E. Berkley, Jerry L. Varns, William A. McClenny,
James Fulcher, "Field Evaluation of the Photovac 10S70
Portable Gas Chromatograph," Proceedings of the 1989
EPA/AWMA International Symposium on Measurement of Toxic
and Related Air Pollutants. Air and Waste Management
Association, Pittsburgh, PA, 1989.
203
-------
Table I. Site Specific Target Compound List
Site #1 Title III Screen Capability
Benzene
Chlorobenzene
p-Dichlorobenzene
o-Dichlorobenzene
m-Dichlorobenzene
Trichlorobenzene
Vinyl Chloride
Carbon Monoxide
Carbon Tetrachloride
x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Table II. SITE Screening Analysis*
Compound ppbv *
Benzene
Chlorobenzene
p-Dichlorobenzene
o-Dichlorobenzene
Trichlorobenzene
4.5
23.4
103.6
35.3
16.8
x
X
X
X
X
#
X
X
* Confirmed by Separate Analysis
# Photovac Confirmation
+ 6 April 1989, near plant producing aromatic hydrocarbons
204
-------
•».»/•«»•* \ .
• 4' •>>'• .'«' '
Jt / - •*
Figure I - Monitoring Sites for the 1989 Delaware SITE Study
205
-------
1. AIR PRODUCTS
2 AKZO CHEMICALS
3. CARDOX ICHEMETROM)
4. CHLORAMONE
5. DELMARVA POWER. DEL. CITY
6. FORMOSA PLASTICS
7. GEORGIA SULF
8. HOECHST CELANESE
9. KEYSOR
OCCIDENTAL CHEMICAL
ORIOLE CARRIERS
STANDARD CHLORINE
STAR ENTERPRISE (TEXACO)
LOC OF MET TOWER
S-20. AVAILABLE SITE FROM DEL. STATE
Industrialized
-------
.•.•_ 3 ...•.•xn
^V' •'!''--!''t'---1
, 1 ADVANCE FILM DIV JAMES RIVER
2. CEPT, Of NATURAL RESOURCES & ENV, CONTROL
3. LLANGOLLEN ESTATES
4 ARMY CREEK
A 5 DELAWARE SAND & GRAVEL
\)
V
CO/
Figure III - The Area Including a Residential Community,
Llangollen Estates(3) Near the Superfund Sites of
Army Creek(4) and Delaware Sand and Gravel.(5)
-------
1 AMERICAN MINERAL
2 B&G INDUSTRIES
3 BLACKTOP PRODUCTS
4 BRANDYWINE CHEMICALS
(FORMERLY HALBY CHEMICALS)
5. CHERRY ISLAND LANDFILL
6 DELAWARE TERMINAL
7 DOVER EQUIP HOT MIX PLANT
8 DUPONT CHAMBERS WORKS. NJ
9 DUPONT CHRISTINA LABS.
10 FORBES STEEL
11 GEORGIA STEEL (2 SITES)
12 H M. DAVIS St SONS
1 3. LAIDLAW Of THE EAST
14 SICO
15 WILMINGTON CHEMICAL CORP
16^ DEL SOLID WASTE AUTHORITY
(Pigeon Point)
17 ENERGY GEN. FACILITY
(Pigeon Point)
18. ICI ATLAS POINT
-
Industrialized Area Between the Christina River
the Delaware Memorial Bridge.
208
-------
Multi-Adsorbent Preconcentration and Gas Chromatographic Analysis
of Air Toxics with an Automated Collection/Analytical System
Albert J. Pollack and Michael W. Holdren
Battelle Memorial Institute
Abstract
In order to identify and quantify the trace ppb levels of volatile
organic compounds (VOCs) present in ambient air it is necessary to
preconcentrate samples prior to analysis. Laboratory tests indicate that a
collection trap containing several different adsorbents (multi-adsorbent trap)
operating at ambient temperature, gave comparable sample collection
capabilities as a glass bead trap that was cryogenically cooled. An automated
gas chromatograph (GC) equipped with a multi-adsorbent trap was constructed
and operated at a field site over a 12-day period to collect hourly samples of
ambient air. Performance of the automated GC system was demonstrated by
analyzing the data from daily calibration checks of the system using a
41-component VOC mixture at the 2 ppb level. Parallel flame ionization and
electron capture detectors were used to detect these components. Relative
standard deviation values of less than 10 percent were obtained for most of
the 41 compounds. No downtime was experienced with the instrument, and
operator interface was limited to performing daily calibration checks,
transferring data to floppy disks, and replacing gas supplies.
Introduction
Many of the volatile organic compounds (VOCs) found in ambient air are
present at very low parts per billion (ppb) and parts per trillion (ppt)
levels. In order to identify and quantify these species, researchers must
employ collection techniques that preconcentrate sufficient amounts of these
materials for analytical detection.
The use of cryogenic trapping to concentrate VOCs prior to analysis has
been established as a proven technique for VOC monitoring (1-5). This method
•involves collecting the sample on an inert material at subambient tempera-
tures. The temperature of the trap is below the condensation temperatures for
trace VOCs but above the condensation temperature for major constituents of
ambient air. After collection the trap is rapidly heated, and the VOCs are
desorbed and typically analyzed using a gas chromatographic (GC) system. For
optimal analytical resolution, the GC oven is also cryogenically cooled. This
allows the desorbed VOCs to be refocused onto the front of the column as a
tight band prior to temperature programming.
The use of cryogen has proven to be not only a costly consumable but also
a logistic problem when monitoring air toxics during field sampling studies.
An alternative to cryogenic trapping is the use of adsorbent materials, which
collect VOCs at ambient temperatures. Popular adsorbents include Tenax-TA,
silica gel, carbon molecular sieves, and activated charcoal. Desorption again
is accomplished by elevating the trap temperature prior to GC analysis. A
deficiency associated with mono-adsorbent traps is their inability to trap
and/or desorb VOCs possessing a wide range of molecular weights (6,7).
The purpose of this paper is to: (1) describe a multi-adsorbent trapping
system designed for the automated collection and analysis of ambient air
samples; (2) compare the efficiency of trapping a variety of VOCs using a
209
-------
multi-adsorbent trap versus a cryogenic assembly; and (3) discuss results from
a 12-day ambient air sampling field study using a multi-adsorbent system.
Experimental Section
Analytical Instrumentation
The gas chromatographic system that was assembled for the automated
collection and analysis of ambient air samples consists of a Hewlett
Packard 5890A GC equipped with a flame ionization detector (FID) and an
electron capture detector (ECD). Hewlett Packard 3396A and 3393A integrators
in conjunction with a 9122 disk drive receives detector output signals and
stores data. The disk drive also provides access to the BASIC program used to
automate sample collection and analysis.
Separations chemistry is accomplished using two 30-meter, HP-1 series,
capillary columns joined with a zero dead-volume, butt connector. Internal
diameter (ID) of the capillary is 0.53 mm with a 2.65 ^m film thickness. The
end of the capillary column is fitted with a zero dead-volume tee. Lengths of
0.32 mm ID, deactivated, fused-silica («25 cm for ECD, «40 cm for FID) are
used to direct about 30% of the column effluent to the ECD while w70% goes to
the FID. Optimal analytical separation is obtained using a temperature
program of -50°C to 200"C with an initial hold of 2 minutes and ramp of
8°C/minute. Zero-grade helium with a flow rate of 7.5 cm3/minute constitutes
the carrier gas. FID gases are ultra zero-air (300 cnfVminute) and zero
hydrogen (30 cmVminute). Zero nitrogen, (40 cmVminute) passes through an
Alltech Oxy-Trap (Cat. No. 4003), and serves as makeup gas for the ECD.
Sample Collector
A modified Nutech 320 sample preconcentration unit is used to collect
ambient air. The unit contains two subsystems: (1) an electronics console
that regulates various temperature zones and (2) the sample handling
subassembly containing a six-port valve and the trap assembly (Figure 1). The
console controls the temperatures of the valve body (120*C), sample transfer
lines (120°C), and the trap. The trap temperature during sample collection is
regulated by the console with the controlled release of liquid nitrogen via a
solenoid valve. Set-points are -150°C during collection for the glass bead
trap and 40°C for the multi-adsorbent trap. During thermal desorption, the
glass bead trap is heated by a 250 watt heater to 90°C. The same type of
heater is used to desorb the multi-adsorbent trap at 220°C. A Valco,
Model A602, air actuator drives the six-port valve from the sample collection
to sample injection positions. The entire sample handling subassembly is
mounted onto the HP5890A GC where the injectors would normally be located.
The connection between the valve and analytical column is made using a
stainless steel union with graphite vespel ferrules.
Trap Design and Sample Collection
The trap assembly is shown in Figure 2. The trap consists of 0.2-cm ID
stainless steel tubing coiled around the desorption heater. The cryogenic
trap is packed with 0.20 gram of glass beads (60/80 mesh). The
multi-adsorbent trap, which was purchased from Supelco, Inc.
(Cat. No. 2-0321), contains 0.20 gram of Carbopack B (60/80 mesh) and
0.05 gram of Carbosieve S-III (60/80 mesh). This type of trap is used to
perform U.S. EPA Methods 624 and 824 which address the collection and analysis
of VOCs from water by purge and trap. The Supelco trap is cut to length and
wound around a 1.8-cm o.d. stock prior to being coiled onto the heater.
210
-------
Sample flow through the trap is controlled using: (1) a Tylan readout
control unit, Model R032-B; (2) a Tylan 0 to 100 standard cubic centimeters
per minute mass flow controller, Model MFC-260; (3) a Thomas dual diaphragm
pump; and (4) a Perma Pure Dryer, Model MD-125-48F. The readout control unit
in conjunction with the mass flow controller regulates air flow rate through a
sampling manifold into the trap. Typically, the sample flow rate is
35 cnf/minute with a collection time of 10 minutes. The perma pure dryer with
a tubular hygroscopic ion-exchange membrane (Nafion) is used to remove water
vapor selectively from the sampled gas and is placed upstream of the trap.
The tube size is 30 cm x 0.1 cm ID, embedded within a shell of Teflon tubing
of 0.25 cm ID. A countercurrent flow of dry zero air (200 to 300 cm3/minute)
Is used to purge the shell. The use of this type of dryer has been shown to
have no affinity for the VOCs of interest in this study (4,5,8). The
multi-adsorbent trap is installed in a manner that allows sample flow
initially through the Carbopack B material and then onto the Carbosieve S-III.
During desorption, the trapped VOCs are backflushed off of the trap. The
sample flow scheme is shown in Figure 3.
Automated Sampling and Analysis
A BASIC program permits automated sampling and analysis procedures to be
carried out. A run is initiated by activating the program using the HP-3396A
integrator via the INET loop connecting the GC and the 9122 disk drive. When
the program starts, five prompts are presented that require an input from the
analyst. These include the number of samples to run, run name, trap
collection time, trap cool down time, and clock start run time. Details of
the BASK program are presented elsewhere (10).
Gas Standard Used In Evaluating GC System
The gas chromatographic system is calibrated by injecting measured
amounts of a diluted mixture from a calibration cylinder. A gas phase dynamic
dilution system is used to generate nominal concentrations of 2 to 4 ppb of
each of these target compounds. Zero air (Aadco, Inc.) serves as the diluent.
The target compounds were obtained as gases or neat liquids (>98 %
purity) from Matheson, Eastman Kodak, Baker Chemical, or Aldrich Chemical
Company. A compressed gas cylinder mixture of the compounds is prepared by
injecting predetermined amounts of each test compound into an aluminum
cylinder that has been flushed with high-purity nitrogen gas and evacuated.
After injection of the gas or neat liquid, the cylinder is pressurized to
1,200 psig with zero air. In order to verify component concentrations, an
accurately measured quantity of each of the compounds is injected into a
17.3 m3, Teflon-lined, aluminum chamber. After equilibration, a GC response
factor is determined for each compound by withdrawing a known volume of air
from the chamber and injecting it into the gas chromatographic system. This
same procedure is used to analyze the gas cylinder standard. Concentrations
are calculated using the derived response factors. For the above compounds,
the calculated cylinder concentrations are within ± 20% of the 200 ppbv
nominal targeted values.
Trap Evaluation
A glass bead trap was initially installed into the trap/valve
subassembly. A humidified 41-component standard mixture (2 to 4 ppb) was
directed into a clean 6-liter, Summa polished, stainless steel sampling
canister and the trapping/analysis system was challenged with this calibration
standard. Evaluation of instrument performance was made by comparing
211
-------
qualitative/quantitative results for the newly constructed unit with
analytical results obtained using a reference gas chromatographic system
equipped with parallel flame ionization and mass selective detectors. The
glass bead trap/full cryogenic reference GC system has a record of excellent
performance based on audit analysis on previous studies carried out at
Battelle (9). Once it was shown that comparable results are being obtained on
both systems, then the multi-adsorbent trap was challenged with the same
41-component mixture. Performance of the multi-adsorbent system was based
upon comparing target compound recoveries with those observed for the
cryogenic, glass bead system.
Field Test of Automated System
The automated GC system, with the multi-adsorbent trap as the whole air
concentrating device, was subjected to a 12-day field test. During this time
the system ran continuously for 10 days collecting and analyzing whole air
samples on an hourly basis. The only interruption in the sampling regime
occurred when a daily performance check was made using the 41-component
calibration mixture. The analytical conditions were the same as those already
mentioned in this paper.
Results and Discussion
Comparison of Cryogenic to Multi-Adsorbent Trapping
Tables I and II show the results comparing the relative performance of
the multi-adsorbent trap to the cryogenic trap. The FID results show
excellent agreement between the two traps for most of the compounds. The
species displaying lower than expected recoveries are in general compounds
with poor FID response (peaks 1 through 7 in Table I). Because of this poor
response, typically both the cryogenic and multi-adsorbent traps displayed
large relative standard deviation (RSD) values. These compounds did respond
well to the BCD and the RSD values reflect acceptable recoveries as shown in
Table II.
Six test runs were made that involved the side by side comparison of two
similar GC systems, one equipped with a cryogenic trap and the other with the
multi-adsorbent trap discussed in this paper. These analyses of ambient air
showed that most of the FID and ECD responding compounds for the two GC
systems were within ±25 percent of the mean concentration (10).
Field Study
Due to the favorable comparison of the multi-adsorbent trapping system to
a full cryogenic unit, the adsorbent system was utilized in a field study.
Field data results are reported elsewhere (11). However, several results that
further demonstrated trapping efficiencies and instrument performance are
reported here.
During the field study a single point calibration check was performed
daily. The results are shown in Table III. The RSDs were derived from the
raw peak areas observed over the 12-day period. The data indicate excellent
precision for most of the 41 components. Compounds that respond to both the
FID and ECD have both RSD values reported. Differences in these values
generally reflect individual response characteristics of those compounds to
the specific detector. The one compound that did show poor RSD values on both
detectors was dichlorodifluoromethane. This result agrees with earlier
laboratory data shown in Tables I and II.
212
-------
On the final day of the field sampling program, the room environment that
housed the GC system during the study was sampled. Immediately after that
analysis, a sample of zero air (Airco) was collected and processed through the
GC system. The room air was indicative of the types and concentrations of
compounds found during the study. Several of the predominant peaks are
identified. In reviewing the chromatogram of the zero air, it is apparent
that no carryover occurred from the previous sample collection. The only
apparent artifact peak from the multi-adsorbent trap occurred at 24.2 minutes
and is observed in both analyses. Based upon an equal carbon atom response by
the FID, this peak corresponds to a concentration of 0.2 ppb carbon.
Concurrent with the evaluation of the trap was the assessment of the
automated sample collection/analytical hardware. During the study the system
performed uninterrupted analyses with minimal operator interaction. The
hourly ambient air sampling was halted to perform the daily calibration run
and change any consumables such as chart paper or gases, but these were the
only times the instrument was not online monitoring the status of the air at
the site. No downtime repairs were required.
The field GC system employed liquid nitrogen to cool the analytical
column and thereby enhance peak resolution. Prior to the field study, efforts
were made to either eliminate or reduce the consumption of cryogen without
sacrificing peak resolution. Various types of capillary columns and a
prototype cryo refocusing device were evaluated, and these results are
reported elsewhere (10). We used the adsorbent trap instead of the cryogenic
cjlass bead trap yet still utilized liquid nitrogen to cool the analytical
column. In this configuration a 30% reduction of liquid nitrogen consumption
was realized. Another advantage in using the adsorbent trap is that extended
collection times do not result in greater cryogen consumption as would be the
case with glass bead trapping.
In the current configuration, the system is sampling the air for
10 minutes every hour. Potentially, this could result in the loss of
detecting a discrete plume moving across the site. The flexibility inherent
with this analytical system makes it possible to modify the BASIC program to
allow extended collection coverage. Up to 79% sampling time coverage of the
monitoring cycle is possible using a single system by overlapping sample
collection for the next run while the analysis of the previous collection is
being performed (4).
Trap/Valve Subasserably Modification
The evaluation of adsorbent traps brought about a modification to
previously constructed air sampling units. The trap/valve subassembly shown
in Figure 1 allows for the rapid installation and removal of different traps.
Earlier designs demanded several hours of effort to change traps (4,5). The
new configuration has converted the trap/valve into two separable parts: one
contains the valve/actuator, while the other contains the trap/heater. The
trap/heater attaches to the rest of the assembly with Swagelock and electrical
quick-connect fittings. This modification allows for the removal of the
existing trap assembly and installation of another complete trap/heater unit
within 20 minutes. As a result, it is possible to change traps rapidly to
meet new analytical needs or to replace a defective unit.
The use of a multi-adsorbent trap has resulted in a system that is more
versatile and less costly to operate than other units without sacrificing
analytical performance.
213
-------
References
1. R.A. Rasmussen, D.E. Harsch, P.M. Sweany, J.P. Krasnec, D.R. Cronn,
"Determination of atmospheric halocarbons by a temperature-programmed gas
chromatographic freezeout concentration method," JAPCA 27:579 (1977).
2. S.O. Farwell, S.J. Cluck, W.L. Bamesberger, T.M. Schuttle, T.F. Adams,
"Determination of sulfur containing gases by a deactivated cryogenic
enrichment and capillary gas chromatographic system," Anal. Chem. 51:609
(1979).
3. H. Westberg, W. Lonneman, and M. Holdren, "Analysis of individual
hydrocarbon species in ambient atmospheres: Techniques and data
validity," In: L. H. Keith (ed.). Identification and Analysis of Organic
Pollutants in Air, Butterworth Publishers (1984).
4. W.A. McClenny, J.D. Pleil, M.W. Holdren, R.N. Smith, "Automated cryogenic
preconcentration and gas chromatographic determination of volatile
organic compounds in air," Anal. Chem. 56:2947 (1984).
5. M. Holdren, S. Rust, R. Smith, J. Koetz, "Evaluation of cryogenic
trapping as a means for collecting organic compounds in ambient air,"
EPA-600/4-85-002, a final report on Contract 68-02-3487 from Battelle
Columbus Laboratories, U.S. Environmental Protection Agency Research
Triangle Park, NC (1985).
6. R.H. Brown, C.J. Purnell, "Collection and analysis of trace organic
vapour pollutants in ambient atmospheres," J of Chrom. 178:79 (1979).
7. E. Pellizzari, B. Demian, K. Krost, "Sampling of organic compounds in the
presence of reactive inorganic gases with Tenax GC," Anal. Chem. 56:793
(1984).
8. J.D. Pleil, K.D. Oliver, and W.A. McClenny, "Enhanced performance of
Nafion dryers in removing water from air samples prior to gas
chromatographic analysis," JAPCA 37:244 (1987).
9. D.L. Smith, "Method evaluation of TAMS network sampling," a final report
on Contract 68-02-4127 from Battelle to U.S. Environmental Protection
Agency, Research Triangle Park, NC (September, 1989).
10. M.W. Holdren, A.J. Pollack, "Development and evaluation of an automated
gas chromatograph equipped with a multi-adsorbent preconcentration
device," a final report on Contract 68-02-4127 from Battelle to U.S.
Environmental Protection Agency, Research Triangle Park, NC (September,
1989).
11. W.A. McClenny, G.M. Russwurm, M.W. Holdren, A.J. Pollack, J.D. Pleil,
J.L. Varns, J.D. Mulik, K.D. Oliver, R.E. Berkley, D. Williams, and
K.J. Krost, "The Delaware SITE study - 1989," Atmospheric Research and
Exposure Assessment Laboratory, Research Triangle Park, NC, In
preparation.
214
-------
Table I. Performance of the muHi-adsorbent trap compared with
the cryogenic trap (3-runs), using flame ionization detector
response.
Concentration (ppb)
Compound
1) dichlorodif luoroaethane
2) •ethyl chloride
3) l,2-dichloro-l,l,2,2-tetraf luoroethine
4) vinyl chloride
5) lathy 1 broiide
6) ethyl chloride
7) trichlorof luorowthane
8) 1,1-dichloroethene
9) dichloroiethane
10) 3-chloropropene
11) l.l^-trichloro-l^.Z-trif luoroethane
12) 1,1-dichloroethane
13) cis-l,2-dichloroethene
14) trichloroiethane
15) 1,2-dichioroethane
16) 1,1,1-tciehloroethane
17) benzene
18) carbon tetrachloride
19) 1,2-dichtoropropane
20) trichloroethene
21) cis-l,3-dichloropropene
22) trans-l,3-dichloropropene
23) 1,1,2-trichloroethane
24) toluene
26) 1,2-dibroaoethane
26) te trichloroethene
27) chlorobenzene
28) ethyl benzene
29) i+p-xylene
30} styrene
31) 1,1,2,2-tetrachloroethane
and o-xlyene
33) 4-ethyt toluene
34) 1 , 3, 5-triiethyl benzene
35) 1, 2, 4-triiethyl benzene
36) benzyl chloride
and i-dichlorobenzene
38) p-diehlorobenzene
39) o-di chlorobenzene
40) 1,2,4-trichlorobenzene
41) hexachlorobutadiene
Cryogenic
2.60
2.64
2. 68
4.61
3.06
2.68
2.74
3.22
4.11
3.22
2.72
3.07
3.48
3.45
3.34
2.66
2.83
2.96
2.69
2.93
2.93
2.93
2.83
2.46
3.06
2.61
2.E4
2.16
2.14
2.25
4.37
1.83
1.83
1.88
4.20
1.78
2.16
1.64
1.37
X
R.5.D.
8.18
15.17
*
17.83
7.49
6.30
84.20
6.26
7.19
3.34
5.25
12.39
2.38
7.29
2.52
3.46
0.97
6.27
2.13
2.36
1.46
3.24
2.18
2.45
6.76
7.61
6 42
5.31
5.86
6.51
3.52
5.97
4.90
6.19
8.66
4.37
6.68
10.08
9.70
Multi- % %
Adsorbent R.S.D. Recovery
0.91
3.25
**
3.92
2.18
2.23
1.46
3.12
3.96
3.13
2.62
2.78
3.52
3.48
3.33
2.58
2.98
3.16
2.S6
3.22
2.82
2.61
2.81
2.51
2.90
2.47
2.58
2.19
2.21
2.16
4.29
1.86
1.80
2.07
3.92
1.79
2.19
1.67
1.24
78.77
20.99
*t
1.60
9.81
3.49
47.84
3.82
6.42
2.21
6.09
5.41
3.58
9.90
1.67
6.39
1.50
3.03
1.46
3.82
5.09
1.72
1.31
3.13
4.71
4.90
2.36
1.46
4.21
1.07
1.07
1.87
2.12
10.47
3.29
1.76
2.94
12.33
8.57
35
123
**
85
71
86
63
97
96
97
96
91
101
101
100
97
105
107
99
110
96
89
99
102
9E
98
102
101
103
96
98
101
98
110
93
101
101
96
91
only one data point,
** = no data.
215
-------
Table II. Performance of the multi-adsorbent trap compared with
the cryogenic trap (3-runs), using the electron capture detector
response.
Compound
Concentration (ppb)
Cryogenic R.S.D.
Multi-
Adsorbent
R.S.D. Recovery
1) dichlorodifluoroiethane 2.60 4.09 2.90 1.03 112
2) l^-dichloro-l.l^^-tetrafluoroethane 2.88 5.73 2.81 0.24 106
3) lethyl broil da 3.06 11.35 2.68 1.44 83
4) trichlorofluoroiethane 2.74 1.33 2.79 0.27 102
5) l.l^-trichloro-l^^-trifluorethane 2.72 0.07 2.71 0.20 100
6) trichloroiethane 3.45 0.74 . 3.45 0.11 100
7) 1,1,1-trichloroethane 2.66 0.32 2.60 0.2S 98
8) carbon tetraehloride 2.96 0.49 2.92 0.12 99
9) trichloroethene 2.93 0.54 3.22 0.42 110
10) 1,2-dibroirethane 3.06 1.45 2.93 0.45 96
11) tetrachloroethene 2.51 0.92 2.53 0.03 101
12) 1,1,2,2-tetrachloroethane 2.23 3.01 1.99 0.38 89
13) i-dichlorobenzene 2.16 ** 2.30 3.24 107
14) p-dichlorobenzene 1.78 ** 1.63 2.89 108
15) o-dich!orobenzene 2.16 6.76 2.29 1.66 106
16) 1,2,4-trichlorobenzene 1.64 ** 1.82 6.10 111
17) hexachlorobutadiene 1.37 8.16 1.28 0.03 93
** = only single run used.
216
-------
Table III. Analytical precision of multi-adsorbent GC system
during field study (12 runs).
Coi pound
1) dichlorodif luoroiethane
2) (ethyl chloride
3} 1 , 2-di ch 1 oro-1, 1 , 2, 2-tetraf I uoroethine
4) vinyl chloride
5) (ethyl broiide
6) ethyl chloride
7) trichlorof luoroiethane
8) 1,1-dichloroethene
9) dichloroiethane
10) 3-chloropropene
11) 1 , 1, 2-trich 1 oro-1 , 2, 2-tr i f 1 uoroethane
12) 1,1-dichloroethane
13) cis-l,2-dichloroethene
14) trichloroaethane
15) 1,2-dichloroethane
16) 1,1,1-trichloroethane
17) benzene
18) carbon tetrachloride
19) 1,2-dichloropropane
20) trichloroethene
21) cis-l,3-dichloropropene
22) trans-l,3-dichloropropene
23) 1,1,2-trichloroethane
24) toluene
25) 1,2-dibroKoethane
26) tetrachloroethene
27) chtorobcnzene
28) ethyl benzene
29) i+p-xylene
30) styrene
31) 1,1,2,2-tetrachloroethane
and o-xlyene
33) 4-ethyl toluene
34) 1, 3, 5-triiethyl benzene
35) 1,2,4-triiethylbenzene
36) benzyl chloride
and •-dichlorobenzene
38) p-dichlorobenzene
39) o-dichlorobenzene
40) 1,2,4-trichlorobenzene
41) hexachlorobutadiene
Chal lenge
Concentration
(ppb)
2.60
2.64
2.68
4.61
3.06
2.58
2.74
3.22
4.11
3.22
2.72
3.07
3.48
3.46
3.34
2.66
2.83
2.96
2.69
2.93
2.93
2.93
2.83
2.4E
3.06
2.61
2.64
2.16
2.14
2.25
4.37
1.83
1.83
1.88
4.20
1.78
2.18
1.64
1.37
FID
X
R.S.D.
*
9.63
63.0
1.37
3.61
3.90
142
3.11
7.41
3.77
3.92
1.34
2.02
1.86
1.84
0.98
2.28
1.80
1.68
1.47
1.76
2.13
1.20
1.62
1.40
2.13
1.42
1.62
1.3E
4.16
1.41
1.33
1.71
2.40
2.87
3.76
1.72
1.63
3.91
ECD
X
R.S.D.
90.1
**
4.23
**
10.6
**
2.16
**
**
**
1.90
**
**
6.02
»*
9.96
**
7.79
**
2.40
*»
**
**
*»
11.1
2.02
**
**
*»
*t
3.14
»*
**
**
6.46
6.88
7.19
16.0
4.36
* = not detected.
»* = does not respond.
217
-------
V«ko*lrA«iiilor
wid SM ron v»iv»
Tnplody
Tnploviln
ConntcMltt
Figure 1. Trap/valve subassembly.
Trip
Cylindriully Wound
Tub* H»t*r (2M wattl)
Umpltln
Ttwrmocouptt
For Lint
Irtckttand
Ortridgt
Hcitcri
(2Sw*ns)
Figure 2. Trap assembly.
Figure 3. Sample flow scheme for the automated system.
218
-------
PASSIVE SAMPLING DEVICES AND CANISTERS: THEIR COMPARISON IN
MEASURING AIR TOXICS DURING A FIELD STUDY
J. L. Yarns and J. D. Mulik
Atmospheric Research and
Exposure Assessment Laboratory
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
D. Williams
NSI Technology Services, Inc.
Research Triangle Park, NC 27711
With wide usage of passivated canisters in ambient air
sampling, any acceptance of an alternate or supplemental sampling
technique would first require a favorable comparison with canis-
ter samplings under identical field conditions. A reusable passive
sampling device (PSD) that contains a thermally desorbable sorbant,
such as Tenax-GC®, potentially offers the advantages of unobtrusive
sampling, rapid recycling, and simplified operation. For
laboratory comparisons, canisters and PSDs simultaneoulsy sampled
ambient air containing air toxics (>40 VOCs) that are measured by
the gas chromatographic TO-14 Method. In this case, a prototype
PSD thermal desorber was added to the GC system; the cryogenic
trapping and chromatographic detection (FID; BCD) were common paths
for the VOCs from pressurized canisters or desorbed PSDs. Field
comparisons were warranted after GC backgrounds from desorbed PSDs
were minimized by (1) thermal bulk cleaning under vacuum and (2)
improving the individual PSD storage units. In August, 1989, 11
sampling events (1 canister with 2 PSDs/event) were taken at 4
sites during a Superfund Innovative Technology Evaluation (SITE)
Study in Delaware. Large concentration differences between sites
were measured for benzene, toluene, chlorobenzene, and p-, m- and
o-dichlorobenzene; comparative data indicate this PSD method could
also accurately screen these sites for these VOCs. This paper has
been reviewed in accordance with the U.S. Environmental Protection
Agency's peer and administrative review policies and approved for
presentation and publication.
219
-------
INTRODUCTION
There is continued interest in passive sampling devices (PSD)
because they appear to satisfy the growing need for more simple,
inexpensive methods to sample the low levels of volatile toxics in
ambient air. In many cases, a large data base, resulting from a
large number of samples taken, is required before environmental
problems can be clearly understood.
Although chemical alterations can occur during thermal
desorption from sorbents such as TenaxGC®, there is at least a 200-
fold greater sensitivity to be gained by analysis of the total
desorbed sample, an advantage that is lost with solvent-extracted
PSDs. However, the PSD exposure time versus the retention volume
for a desired chemical species will often limit any ambitions for
a single sorbent to possess wide-spread, quantitative applications.
For these reasons, the selection of PSDs for preliminary initial
screenings of unknown environments may need to be based upon
detecting different chemical classes.
The objective of this paper was to compare canister and PSD
data under field conditions. In this instance, the utility of the
PSD was enhanced by limiting its detection to an aromatic class of
chemicals (substituted benzenes) that demonstrated large retention
volumes for the sorbent.
EXPERIMENTAL
Selection of Sampling Sites and VOCs
Canister grab samples were taken from clustered sites near or
within New Castle County, Delaware and analyzed for volatile
organic toxics (Method TO-14). Based on these data, the follow-
ing four sites were selected for PSD/canister intercomparison: a
superfund site (Site A) ; a landfill (Site B) ; a state monitoring
station (Site C); and an active industrial complex (Site D). The
described PSD/canister comparisons at these sites were part of the
Superfund Innovative Technology Evaluation (SITE) Program.
Benzene, toluene, chlorobenzene, and p-, m- and o-dichloro-
benzene were the VOCs selected for this comparison. These
aromatics were verified by GC-MS in the initial site screenings
and favorable, high retention volumes for these VOCs were
demonstrated in the TenaxGC®-PSD used (1) .
Instrumentation
A Hewlett-Packard 5880 gas chromatograph, equipped with flame
ionization and electron capture detectors and fitted with a cryo-
genic trap for concentrating VOCs (2) , was connected to a prototype
PSD desorber fabricated by Nutech, Inc. As shown in Figure 1, VOCs
from either a canister sample (left side) or a thermally desorbed
220
-------
PSD (right side) share a common cryogenic trap (-150C) for pre-
concentration prior to chromatographic analysis (HP-1 0.53 mm ID
X 30 m capillary column, 0.88 micron thickness; -50 to 150C thermal
program). Multi-level calibration was performed to accommodate the
wide range of VOC concentrations sampled at the selected sites.
PSD Analysis
Passive sampling devices (PSDs), each containing 0.4 g Tenax-
GC® with experimentally measured VOC sampling rates for the
reported VOCs were bulk cleaned in a stainless steel, teflon o-
ring sealed cylinder at HOC under 20 micron vacuum for 8 hours.
The hot evacuated cylinder was brought to atmospheric pressure by
adding nitrogen gas. After cooling, the PSDs were transferred to
their individual air-tight containers under a nitrogen atmosphere
and then stored in nitrogen-purged metal cans until exposed.
Previous tests indicated this bulk cleaning procedure would provide
a. consistent BCD/FID background between PSDs for the duration of
the field study and subsequent period for analysis. After a PSD
was thermally desorbed 10 min in a sealed chamber (Figure 1, far
right) at 18OC, the desorbed VOCs were transferred and recollected
into a -150C cryotrap (Figure 1, far left) by purging the heated
chamber with a 30 cc\min Helium flow for 10 min.
Canister Analysis
Six L canisters were subjected to three cycles of evacuation
and pressurization with ultra-pure air prior to final evacuation
to 20 microns. Random analyses of the cleaned canisters were
performed prior to the field study. After collection of field
samples, the canisters were pressurized to 20 psig and the VOCs
from a 300 cc atmospheric aliquot, were trapped in the —150C
cryotrap at 30 cc/min prior to chromatographic analysis.
Field Sampling
A sampling event was represented by a two hour cocollection
of a single canister with paired PSDs; 11 sampling events were
taken over a four day period between 9:00 and 15:30. Days were
either overcast or sunny during the sampling period; no rainfall
occurred. Canisters were fitted with a linearized sampling orifice
that provided a constant 25 cc/min sampling over the 2 hour
sampling. Readings of residual vacuum in the canisters after
sampling were immediately recorded and these values rechecked prior
to analysis. Three random PSDs remained unopened as controls
during sampling and provided a background average for the sampled
PSDs. The exposed PSDs were placed within 10 cm of the canister
sampling orifice and a metal roof protected all sampling events.
221
-------
RESULTS AND DISCUSSION
Eight of the 11 sampling events (11 canisters and 25 PSDs)
collected during this study were taken at Sites A, B and C and the
analytical data for six volatile organics during the eight events
are presented in Table I. The superfund site, the landfill and the
state monitoring station provided similar VOC levels that did not
exceed 10 ppbv. The screening ability of the PSDs is indicated by
the general agreement to measured canister concentrations.
Assessments of benzene, toluene and chlorobenzene exposure could
be performed using either canister or PSD samples; however, the PSD
consistently provided information on the dichlorobenzenes that was
not always obtained by canister analysis. This was to be expected
at sub-ppbv detection levels because the exposed PSD represented
a 10-fold sampling volume to that used in the canister analysis.
Figure 2 graphically compares PSD and canister data from three
sampling events at an industrial complex having high VOC con-
centrations (5 to 293 ppbv range). The PSDs indicated a general
agreement with the corresponding canister concentrations and they
show promise as a screening tool for locating pollutant sources,
especially when the preselected aromatics have large retention
volumes for the sorbent (1,3). Viewing the general close agree-
ment from the paired PSDs, it not understood why the p-chloro-
benzene data from the first two events showed a wide spread;
however, it can be seen that the PSD averages were similar to the
concomitant canister data. From a verbal report (4) that described
the rapid VOC fluctuation during long path measurements made at
Site D during this same period, it is not clear that errors in PSD
sampling occurred.
CONCLUSIONS:
In this field study, the passive sampling devices agreed with
respective canister data in assessing the targeted aromatic
volatile concentrations at four sampling sites. If only PSDs were
chosen as site screening devices, the following sampling
recommendations are as follows:
1. Several progressive exposure periods (30 -120 min) should
be used, especially if the targeted VOCs have marginal
retention volumes. A three PSD/site minimum should be
taken since analysis of sample aliquots is not currently
possible.
2. In general, agreement between canister and PSD con-
centrations appeared concurrent with close toluene
results; therefore, toluene levels between TenaxGC®-PSDs
should be compared in assessing sample integrity.
222
-------
3. The prototype PSD desorber-GC system produced chromato-
graphy equal to that achieved with canister sampling.
It is recommended that both flame ionization (FID) and
electron capture detectors be used when a wide range of
concentrations may be encountered.
DISCLAIMER
The research described in this article does not neces-
sarily reflect the views of the Agency and no official
endorsement should be inferred. Mention of trade names
or commercial products does not constitute endorsement
or recommendation for use.
REFERENCES
1. R. G. Lewis, J. D. Mulik, R. W. Coutant, G. W. Wooten, and C.
R. McMillin, "Thermally desorbable passive sampling device
for volatile organic chemicals in ambient air," Analytical
Chemistry. 57: 214 (1985) .
2. William A. McClenny, J. D. Pleil, M. W. Holdren, and R. N.
Smith, "Automated cryogenic preconcentration and gas chromato-
graphic determination of volatile organic compounds in air,"
Analytical Chemistry. 56: 2947 (1984).
3. W. R. Betz, "Monitoring a wide range of airborne organic
contaminants," Proceedings of EPA/APCA Symposium, Measurement
of Toxic and Related Air Pollutants (Raleigh, NC, May 1987),
pp. 761-70.
4. G. M. Russwurm, NSI Technology Services, Inc., Research
Triangle Park, NC, personal communications (1990).
223
-------
TABLE 1. Comparison of PSD and canister air sampling at three sites
in Delaware Field Study. Superfund Site (A); Landfill (B);
State Monitor Station (C).
Sampling
Analvte:
Benzene:
Toluene:
Chloro-
Benzene:
m-Cl2-
Benzene:
P-ci2-
Benzene:
o-ciz-
Benzene:
Events:
PSD #1
PSD #2
PSD avg
Canister
PSD #1
PSD #2
PSD avg
Canister
PSD #1
PSD #2
PSD avg
Canister
PSD #1
PSD #2
PSD avg
Canister
PSD #1
PSD #2
PSD avg
Canister
PSD #1
PSD #2
PSD avg
Canister
1
1.20
1.30
1.25
3.75
6.26
7.50
6.88
5.32
2.14
2.90
2.52
2.05
0.65
0.62
0.64
0.71
1.70
4.15
2.92
ND
0.83
1.24
1.04
0.33
SITE A
2
0.80
a
0.80
1.18
1.60
1.60
2.17
0.75
0.75
1.41
1.17
1.17
0.91
1.62
1.62
ND
0.27
0.27
ND
3
0.28
0.44
0.36
0.92
0.95
1.18
1.06
1.88
1.76
2.46
2.11
1.05
0.55
0.58
0.57
0.92
0.49
0.78
0.64
ND
0.25
0.41
0.33
ND
2.
1.
2.
2.
7.
4.
6.
5.
1.
2.
1.
1.
0.
0.
0.
0.
2.
2.
2.
0.
0.
0.
0.
4
75
64
19
11
63
91
27
54
36
42
89
02
97
55
76
60
33
02
18
ND
26
87
57
33
SITE B
5
2.97
2.12
2.54
2.33
7.34
4.18
5.76
3.97
0.68
3.36
2.02
0.96
1.06
0.99
1.03
1.27
1.07
1.41
1.24
ND
0.46
0.24
0.35
1.58
6
0.86
1.20
1.03
0.32
2.57
3.28
2.93
1.15
2.54
2.49
2.52
2.03
0.21
0.48
0.35
1.78
1.70
1.47
1.59
0.98
0.20
0.17
0.19
0.37
7
3.
3.
3.
2.
8.
9.
8.
6.
3.
3.
3.
1.
0.
0.
0.
0.
0.
1.
1.
0.
0.
0.
0.
0.
SITE
15
14
15
35
60
32
96
05
77
72
75
12
50
22
36
59
95
31
13
56
13
28
21
29
C
8
2.48
2.39
2.44
2.33
9.57
9.39
9.48
5.37
0.22
0.23
0.23
0.78
0.35
1.14
0.75
0.59
2.69
2.62
2.66
1.09
0.72
0.76
0.74
0.40
* sample lost during desorption/analysis
ND = Not Detected
224
-------
Plumbing Schematic for Canister - PSD Analysis
v&c
CflYO
TRAP
-19O
Cryogenic Trap/Canlster Svatem
VB4T
PSD Thermal Deaorbcr Svatem
Analysis Mode -
Collect Mode -
Desorber (Chamber) By-pass -
Desorbed Sample to QC trap -
.Figure I. Plumbing pathways for VOC analysis. Collect/analysis valve stages
.for canister samples (Left side) and desorb/analysis stages for PSD samples
(Right side).
225
-------
CANISTER/PSD SAMPLINGS AT SITE D
1989 DELAWARE FIELD STUDY
500
4OO
300
2OO
1OO
30O
g 200
a
100
200
100 -
CANISTER
PSD AVG A
HIGH
JUL 31 13:OO
L laasa
AUG 1 13:OO
AUG 1 15:OO
BENZEhE TCXUEhE OL-BZ M-O.2-BZ P-CL2-BZ O-CL2-BZ
figure 2, Comparative data of canister and paired PSD samples from
sampling events near an active industrial complex during field study (Site D)
226
-------
SECTOR SAMPLING FOR VOLATILE ORGANICS CONTRIBUTIONS TO AMBIENT AIR
FROM INDUSTRIAL SOURCES
Joachim D. Pleil and William A. McClenny
U.S. EPA, Atmospheric Research and Exposure Assessment Laboratory
Research Triangle Park, North Carolina 27711
Michael W. Holdren and Albert J. Pollack
Battelle Memorial Institute
Columbus, Ohio 433201-2693
Karen D. Oliver
NSI Technology Services Corporation
Research Triangle Park, North Carolina 27709
Sector sampling for volatile organic compounds (VOCs) entails the use of an
integrated sampling scheme coupled to a wind direction sensor. Whole air is
collected at a constant rate into one of two SUMMA canisters depending upon
wind direction; when the wind comes from the suspected emissions area, sample
is routed into the "IN" sector canister; otherwise, sample is collected in the
"OUT" sector canister. For this set of experiments, the IN and OUT sectors were
90 and 270°, respectively, with the IN sector centered on the VOC source. Two
samping sites were used, the first about two miles north by northeast of a group
of industrial facilities, and the second was located about one mile south by
southeast of these sources. Sites were alternately operated with duplicate
sampling systems. The resultant data comparisons between IN and OUT
concurrent samples show good correlations to expected VOC emissions, as
determined by grab samples taken within the target area.
227
-------
Introduction
The U.S. Environmental Protection Agency (EPA) has developed
methodology over the past six years for the trace-level determination of certain
nonpolar volatile organic compounds (VOCs) in the ambient air. This is based
upon the use of SUMMA* polished canisters for collecting whole air samples for
subsequent laboratory analysis. Details for the method are available in Method
TO-14 of the EPA compendium of methods.1 Typical applications of TO-14
involve time-integrated point monitoring, usually for 24-h periods, as in EPA's
Toxic Air Monitoring System (TAMS) network2prthe Urban Air Toxics Monitoring
Program (UATMP).* Though point monitoring has proven to be useful in
documenting VOC concentrations, more sophisticated monitoring strategies
have been under development to produce additional information such as time
variability of VOC concentrations and VOC source locations. This has involved
semi-real-time, in situ VOC determination with concurrent wind speed and
direction measurement to allow calculation of back trajectories of VOCs,* time-
dependent measurements in indoor air situations,5-7 and wind-direction-
dependent VOC measurement.8
This work concerns the use of a spatially resolved air monitoring strategy for
inferring VOC emissions from suspect sources. The philosophy of the method is
collection of ambient air into one of two containers, depending upon whether or
not the wind is blowing from a sector containing the suspected VOC source. The
differences between the "IN" sector sample and the "OUT" sector sample at a
single monitoring site will provide candidate VOCs emanating from the IN sector
source. In effect, this "Sector" method provides its own background via the OUT
sector sample. To corroborate this identification, additional sites can be
operated simultaneously at other directions from the suspect source.
This methodology was tested in an airshed containing a cluster of industrial
VOC sources. Two fixed monitoring sites were used, one about two miles north
by northeast of the sources, the second less than one mile south by southeast, as
diagrammed in Figure 1. Each site was operated in duplicate for part of the time.
In addition, various grab samples were taken among the sources. The resulting
samples were analyzed for a variety of nonpolar volatile organic "toxics" (listed
in reference 1) and for CA to CIQ alkanes. A data interpretation method is
presented for the Sector method to deduce the trace-level contributions of the
suspected source area to the receptor sites. The grab sample results are
presented to confirm these deductions.
Experimental Section
All samples were collected in 6-liter volume stainless steel canisters with
internally passivated surfaces. They are often referred to as "SUMMA cans"
because of the proprietary internal (SUMMA) passivation procedure used by the
manufacturers. Canisters are cleaned prior to sample collection by evacuation to
less than SOpm Hg at an elevated temperature of 100°C. Quality assurance of the
clean-up procedure is performed on a subset of these canisters by filling them
with humidified zero air and analyzing for residual contamination. In the field,
samples are always collected by starting at the clean-up vacuum (<50 pm Hg).
Details of the use of various configurations of these sampling devices and the
associated compound storage stability are available in Method TO-141; some
additional representative publications are given in the reference section>11 The
canisters used for this project were from two manufacturers: SIS, Inc. (Moscow,
Idaho) and Biospherics, Inc. (Hillsborough, Oregon).
The sector samplers consist of a conventional "R"-type actively pumped
flow configurations.^ wjth the exception that a three-way valve allows routing
228
-------
the air sample flow between two canisters. The destination of the flow is
dependent upon the wind direction; that is, its routing to one of the two
sampling canisters depends upon a signal from a wind sensor that determines
whether the air is coming from the suspect source (IN sector) or not (OUT sector).
A diagram of this configuration is presented in Figure 2. Three sector samplers
from Xpntech, Inc. (Van Nuys, CA), were used in this study. Sample flow was
maintained at about 10.5 cm3/min, which would result in a final canister pressure
of 20 psig after 24 h if the wind direction never changed. Typically, the IN sector
canister fills at a lower rate because the respective sector is smaller than that for
the OUT canister (90 vs. 270°); sometimes two OUT sector samples are collected
before sufficient volume of IN sample is collected for analysis.
For the first two tests, site 1 was operated in duplicate with samplers #1 and
3, and for the remaining two tests, site 2 was operated in duplicate with samplers
#2 and 3. The details of the sampling program, times, sites, sampler number, and
canister number, are given in Table I.
Two secure sites were provided by the State of Delaware for our
experiments; both had electrical power available for the samplers. The sites were
in the New Castle area near a variety of industrial sources including a coal-fired
power plant, an oil refinery, a petroleum product storage and transfer facility,
various chemical manufacturers, and plastics manufacturing plants. The samplers
were operated with 90° IN sectors (with corresponding 270° OUT sectors). A
diagram of the relative locations of the sources, sampling sites, and the IN sectors
is given in Figure 1.
All analyses were performed by gas chromatography/mass spectrometry
(GC/MS; Hewlett-Packard [HP] model 5880A gas chromatograph and model HP-
5970B mass selective detector, Hewlett-Packard, Avondale, PA, and Palo Alto, CA,
respectively) in selected ion monitoring mode for the TO-14 list of 40 nonpolar
VOCs. In addition, a flame ionization detector (FID) was used in parallel at the
column outlet to determine nontarget compounds, primarily the €4 to CIQ
alkanes. Sample was preconcentrated with a Nutech 320-1 cryogenic focussing
unit (Nutech, Inc., Durham, NC) and separated on an HP 50-m by 0.32-mm-i.d.,
HP-1 fused-silica capillary column. Some of the grab samples taken among the
sources were analyzed by GC/MS scan methods (28.5 to 350 amu) to identify all
major organic constituents. The higher level compounds were then quantified by
using a second GC (also an HP-5880A), equipped with an FID and an electron
capture detector. The general analytical methods for these analyses are
presented in the Iiterature12-i4; the specific analytical protocols used for the
sector sample analyses are identical to those for EPA's TAMS network and are
described in detail in the current TAMS report.2 Most of the grab samples were
taken near a chlorinated compound manufacturer (see Figure 1, source #1);
some additional samples were taken very near (and downwind from) the
petroleum distiller's tank farm (Figure 1, source 4).
Results and Discussion
Grab samples showed two specific groups of compounds, depending upon
location; typical ranges of concentrations are given in Table II for samples taken
near the tank farm and near the chlorinated chemicals manufacturer. These
compounds were extremely variable in time, and we typically collected only grab
samples when there were indications that an "event" was occurring, such as
smells, data from real-time instrumentation, or obvious plumes. Most of these
samples were taken near the chlorinated chemicals plant, and the results are
from mass spectrometry data. The alkane and hydrocarbon data are from a few
grab samples taken on a roadway near the oil storage facility and are based upon
229
-------
approximations from FID traces. Only compounds that exceeded 10 ppbv in at
least one sample are listed in the table.
The data from the sector samplers is broken down into two categories, mass
spectrometry data for "target compounds," the toxic VOCs as listed in Method
TO-14, and data for the alkanes and other hydrocarbons that were taken from
FID traces and quantified with respect to a calibrated benzene response.
Cursory inspection of the concentration data reveals that differences
between the IN and OUT paired sector samples are subtle and that the
concentrations at the sites 1 to 3 miles away from the sources are essentially at
background levels. To allow the extraction of useful information from this aata
set, a parameter for comparing paired IN and OUT sector data on a per
compound basis was defined as
R = (IN-OUTVON + OUT + 10-6) (1)
where IN and OUT are concentrations in parts per billion by volume for specific
sample sets and individual compounds. The addition of 10-° in the denominator
is a calculational aid to avoid dividing by zero.
The parameter R ranges from -1 to +1 and can accommodate
0 concentration values. Positive R values indicate an IN sector prevalence,
whereas negative values indicate an OUT sector prevalence. The absolute value
of R gives a relative measure of the importance of a compound in a sector. For
example, consider the following cases:
If ' Then
IN »OUT R =: +1
IN = 2 OUT ft = +0.33
IN = OUT /? = 0
IN = 1/2 OUT R = -0.33
IN « OUT R a -1
Because this is a "ratio" parameter, some consideration must be given to dealing
with very low concentration values. That is, when the minimum quantifiable
level is approached, at which point is a concentration considered zero or some
minimum value? These decisions can easily change the calculated ft from 0 to ± 1.
This was handled by rounding all data to the nearest 0.1 ppbv (defined as our
minimum quantifiable level for this project), generating all the ft values, and then
generating overall averages of concentrations for each compound from the
sector samples. For compounds that exhibited average concentrations at or
below the minimum quantifiable level, the ft values were not considered as
strong indicators of sector prevalence.
Even in the absence of any sources, there will be scatter in the ft values from
the intrinsic sampling and analytical precision errors. We assumed that such error
is normally distributed and derive a measure of significance by calculating a
relative standard deviation (RSD) and a mean for various categories of ft values.
The resulting data can then be interpreted in light of the RSD values; that is,
different levels of confidence can be assigned to determinations of whether a
compound is emanating from the source sector. In addition, the use of duplicate
sampling allows direct comparisons between nominally identical samples.
Rvalues and their RSDs calculated for such paired samples can also serve to
indicate any system problems with the samplers.
Reduced data from the sector sampling experiments are presented as a
compilation of average ft values and RSDs on a per compound basis for site 1
230
-------
(Table 111), site 2 (Table IV), and for all data combined (Table V). Data from
duplicate samplers are included because these were totally independent
concurrent measurements. A third column is added for convenience, in which the
ratios of the average R value and the RSD are given. This parameter (statistical
significance) is defined as
S = /?(ave)/RSD (2)
and serves as an indicator of the statistical significance of the R value in the
following manner:
If Then
-0.5 1.0 sector, positive S corresponds to IN sector
A similar compilation of information for all the duplicate samples is
presented in Table VI; in this case, the S values are measures of overall
compound-dependent sampling and analytical scatter. Table VII contains the
overall averages of concentrations for each compound.
The information from Tables III through VII is reduced into Table VIM, where
statistically significant sector assignments are made. Compounds are labeled as
associated with the IN or OUT sector for site 1 or site 2; and for the composite
data set, they are labeled both for "moderate" statistical difference and for
"significant" statistical difference.
The following "filtering" was applied to the data to arrive at the results in
TableVIII:
• Dichloromethane data were deleted because duplicate precision data
from Table VI exhibit both a high R value and a large S value; this
indicates a systematic problem, possibly contamination.
• Benzene, toluene, and 1,2,4-trimethylbenzene data exhibit S values
greater than 0.5 for their duplicate comparisons (Table VI). Because the
corresponding R values were all below 0.1 (very dose to neutral bias),
this indicates scatter in side-by-side sampling without appreciable
systematic error. Also, examination of the raw data shows that these
compounds were internally consistent within their respective samplers.
Therefore, this data was considered useful. The methylpentane isomers
were also treated this way.
• Ethyl chloride and vinylidene chloride were deleted from summary Table
VIII because the overall average concentration throughout the study was
well below the minimum quantifiable level (see Table VII), and they
were found during only one sampling episode at low levels.
231
-------
• Though a few other compounds such as trichlorobenzene, n-octane, and
n-nonane exhibit low average concentrations in Table VII, these data
were judged useful and included in Table VIII because they were found
in about half the samples at levels exceeding the minimum quantifiable
level.
The final results in the summary Table VIII show the following:
• The most "moderate" and "significant" correlations are for site 2; this is
expected because this site is closer to the sources, as seen in Figure 1.
• Most of the positive correlation is for the aliphatic and aromatic
hydrocarbons; the source area is dominated by petroleum product
processing and storage.
• Freon 12 exhibits a definite OUT sector prevalence for site 1 and a
moderate IN sector prevalence for site 2. This points out the benefit of
using at least two sites when sector-sampling air from a suspected
source. In this case, the source of the Freon 12 is apparently somewhere
in the IN sector for site 2 but also north or northeast of site 1; therefore,
it is not coming from the source area.
Given the data from Table VIII, one can conclude reasonably that the
following compounds are candidates of VOC source emissions from the suspect
source area:
benzene trichlorobenzene 3-methylpentane
toluene isobutane n-hexane
chlorobenzene butane methylcyclopentane
ethylbenzene isopentane cyclohexane
m,p-xylene pentane isooctane
p-dichlorobenzene cyclopentane n-octane
o-dichlorobenzene 2-methylpentane
When this list is compared to the listing of dominant VOCs found in the grab
samples {Table II), we see that every compound from Table II is identified as an
emission candidate by the sector sampler method. The additional compounds are
likely lower level steady-state source emissions that did not exhibit greater than
10 ppbv levels, or event releases that were missed by the few grab samples.
Conclusions and Recommendations
The sector sampling methodology and the described data reduction
techniques can provide VOC data that indicates probable source emission
identifications, even at distances of;one to three miles. This is a useful technique
for both short-term VOC screening of suspect sources and for long-term
monitoring of the contributions from a specific source. The method is optimized
by the use of multiple sites and by minimizing the distance between the receptor
site and the suspect source; distances of one mile or less are recommended.
Acknowledgments
The authors thank the following individuals for their invaluable assistance
during the performance of this work: Our hosts, Captain Joseph J. Kliment and
Terri H. Brixen of the State of Delaware Department of Natural Resources, made
all arrangements for field sites, provided office space for meetings, and made us
feel welcome in Delaware. Matthias Yoong of Xontech, Inc., provided his expert
232
-------
assistance in preparing the samplers and calibrating flows in the field.
George M. Russwurm of NSI Technology Services Corporation assisted in grab
sampling, in operating the field sites, and in preparing daily meteorological
reports. Although the research described in this article has been funded wholly
or in part by the United States Environmental Protection Agency through
Contract 68-02-4127 to Battelle and Contract 68-02-4444 to NSI Technology
Services Corporation, it has not been subjected to Agency review and therefore
does not necessarily reflect the views of the Agency, and no official endorsement
should be inferred. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
References
"" ^
ids
EPA's Compendium of Methods for the Determination of Toxic Organic
Compounds in Ambient Air. Method TO-14: "The Determination of
Volatile Organic Compounds (VOCs) in Ambient Air Using SUMMA
Passivated Canister Sampling and Gas Chromatographic Analysis",
EPA-600/4-89/017, U.S. EPA, Research Triangle Park, NC, 1989.
2. D. L. Smith, M. W. Holdren, "Method Evaluation of TAMS Network
Sampling", Final Report of Contract #68-02-4127 to EPA-AREAL, Work
Assignment 66, Battelle Memorial Institute, Columbus OH, 1989.
3. R. A. McAllister, W. H. Moore, J. Rice, D. P. Dayton, R. F. Jongleaux, P. L
O'Hara, R. G. Merrill, J. Bursey, "1988 Nonmethane Organic Compound
and Urban Air Toxics Monitoring Programs", Final Report, Contract
68-D8-0014to EPA-AREAL, Radian Corporation, Research Triangle Park,
NC, 1989.
4. W. A. McClenny, K. D. Oliver, J. D. Pleil, "A field strategy for sorting
volatile organics into source related groups," Environ. Sci. TechnoF.
23(11) (1989).
5. J. D. Pleil, W. A. McClenny, K. D. Oliver, "Peak Exposure and Emission
Rate Measurements of Halogenated Volatile Organic Compounds in
Indoor Air", presented at 36tn ASMS Conference on Mass Spectrometry
and Allied Topics, San Francisco, CA, June 5-10,1988.
6. J. D. Pleil, K. D. Oliver, W. A. McClenny, "Time-Resolved Measurement
of Indoor Exposure to Volatile Organic Compounds", Proceedings:
Indoor Air'87, Institute for Water, Soil and Air Hygiene, Berlin, 1987.
7. J. D. Pleil, W. A. McClenny, K. D. Oliver, "Temporal variability
measurement of specific volatile organic compounds," Int. J. Environ.
Anal. Chem. 37: 263 (1989).
8. J. D. Pleil, W. A. McClenny, K. D. Oliver, "Wind Direction Dependent
Whole-Air Sampling for Ambient VOCs", presented at 81st Annual
Meeting and Exhibition of the Air Pollution Control Association, Dallas,
TX, June 19-24,1988.
9. W. A. McClenny, J. D. Pleil, T. A. Lumpkin, K. D. Oliver, "Toxics
Monitoring with Canister-based Systems," Proceedings: 80th Annual
Meeting of the Air Pollution Control Association, Air Pollution Control
Association, Pittsburgh, 1987.
233
-------
10. J. D. Pleil, K. D. Oliver, "Whole-Air Samplers for VOCs Determination in
Ambient Air: Configurations, Parts, and Certification of Successfully
Used Systems," Internal AREAL Report, U.S. EPA, Research Triangle
Park, NC 27711, February 1988.
11. K. D. Oliver, J. D. Pleil, W. A. McClenny, "Sample integrity of trace level
volatile organic compounds in ambient air stored in SUM MA polished
canisters" (and references therein). Atmos. Environ. 20(7): 1403 (1986).
12. W. A. McClenny, J. D. Pleil, M. W. Holdren, R. N. Smith, "Automated
cryogenic preconcentration and gas chromatographic determination of
volatile organic compounds," Anal. Chem. 56: 2947 (December 1984).
13. J. D. Pleil, K. D. Oliver, W. A. McClenny, "Enhanced performance of
Nafion dryers in removing water from air samples prior to gas
chromatographic analysis," J. Air Pollut. Control Assoc. 37: 244 (March
1987).
14. J. D. Pleil, K. D. Oliver, K.D., W. A. McClenny, "Ambient air analyses
using nonspecific flame ionization and electron capture detection
compared to specific detection by mass spectroscopy", J. Air Pollut.
Control Assoc. 38: 1006 (August 1988).
234
-------
Table I. Sector Samples - Times and Locations
Time period Site Sampler
11:00 8/1
17:10 8/3
17:15 8/3
9:20 8/5
10:30 8/5
17:18 8/6
1
1
2
1
1
2
1
2
2
1
3
2
1
3
2
1
2
3
Canistera
IN
11
02
77
23
46
73
75
62
11
OUT
05,50
74,08
24,42
14
32
64
53
35
16
17:258/6 1 1 75 76,31
14:008/8 2 2 71 48,13
23 65 45,44
aBecause of wind conditions, canister #75 serves as IN sample for P4 site for last two time periods.
Table II. Results from Grab Samples: Compounds Found at Levels Higher than
Typical Background Concentrations
Chlorinated chemicals manufacturer
(mass spectrometer data)
Compound
dichloromethane
benzene
chlorobenzene
p-dichlorobenzene
o-dichlorobenzene
trichlorbenzene
ppbv Range
(10 samples)
1-12
1-51
7-260
12-247
4-154
1-41
Petroleum products storage/handling
facility (approx. FID data)
Compound
butane
isopentane
pentane
2-methylpentane
3-methylpentane
methylcyclopentane
ppbv Range
(2 samples)
22-103
13-444
7-200
2-50
2-25
1-13
235
-------
Table III. IN vs OUT Comparisons - Site 1 a
R (ave) RSD Compound 5
-0.230 0.217 FREON12 -1.060
-0.169 0.269 METHYLCHLORID -0.628
-0.333 0.516 ETHYLCHLORIDE -0.645
-0.056 0.078 FREON11 -0.715
0.000 0.000 VINYLIDENECHL ERR
-0.008 0.101 DICHLO METHANE -0.080
0.129 0.231 FREON113 0.557
0.063 0.211 METHYLCHLFORM 0.296
0.067 0.075 BENZENE 0.898
-0.081 0.151 CARBONTETCHLO -0.536
-0.005 0.090 TOLUENE -0.051
0.433 0.496 CHLOROBENZENE 0.873
0.094 0.162 ETHYLBENZENE 0.583
0.073 0.199 m,p-XYLENE 0.367
-0.006 0.266 o-XYLENE -0.022
-0.333 0.816 4-ETHYLTOLUEN -0.408
0.000 0.632 1,3,5-METBENZ 0.000
0.043 0.246 1,2,4-METBENZ 0.176
0.230 0.694 p-DICHLORBENZ 0.332
0.056 0.136 o-DICHLORBENZ 0.408
0.000 0.000 TRICHLBENZENE ERR
0.094 0.170 ISOBUTANE 0.556
0.149 0.129 BUTANE 1.152
0.171 0.107 ISOPENTANE 1.596
0.135 0.080 PENTANE 1.673
0.239 0.428 CYCLOPENTANE 0.559
0.165 0.115 2-METHPENTANE 1.437
0.229 0.140 3-METH PENTANE 1.640
0.096 0.161 n-HEXANE 0.594
0.178 0.150 METHCYCPENTAN 1.185
0.113 0.138 CYCLOHEXANE 0.816
0.029 0.129 ISOOCTANE 0.228
-0.087 0.466 n-HEPTANE -0.187
0.056 0.646 n-OCTANE 0.086
0.134 0.992 n-NONANE 0.135
0.072 0.366 n-DECANE 0.197
aAbbrevtationsare R = (IN - OUT)/(IN + OUT + 10-6), RSD = relative standard deviation,
S = K(ave)/RSD
236
-------
Table IV. IN vs OUT Comparisons-Site 2a
R (ave) RSD Compound S
0.239 0.378 FREON12 0.632
0.016 0.139 METHYLCHLORID 0.114
0.000 0.000 ETHYLCHLORIDE ERR
-0.009 0.075 FREON 11 -0.115
-0.166 0.407 VINYLIDENECHL -0.408
0.039 0.045 DICHLOMETHANE 0.868
-0.175 0.406 FREON 113 -0.431
-0.032 0.086 METHYLCHLFORM -0.374
0.170 0.088 BENZENE 1.925
0.048 0.223 CARBONTETCHLO 0.213
0.144 0.049 TOLUENE 2.952
0.395 0.128 CHLOROBENZENE 3.083
0.122 0.142 ETHYLBENZENE 0.858
0.161 0.048 m,p-XYLENE 3.328
0.142 0.128 o-XYLENE 1.113
0.079 0.137 4-ETHYLTOLUEN 0.580
-0.198 0.633 1,3,5-METBENZ -0.313
0.102 0.128 1,2,4-METBENZ 0.797
0.350 0.203 p-DICHLORBENZ 1.721
0.342 0.198 o-DICHLORBENZ 1.727
0.683 0.402 TRICHLBENZENE 1.700
0.449 0.188 ISOBUTANE 2.383
0.352 0.166 BUTANE 2.120
0.306 0.097 ISOPENTANE 3.164
0.304 0.152 PENTANE 2.001
0.555 0.501 CYCLOPENTANE 1.107
0.291 0.164 2-METHPENTANE 1.777
0.250 0.122 3-METH PENTANE 2.044
0.342 0.159 n-HEXANE 2.148
0.190 0.165 METHCYCPENTAN 1.155
0.250 0.204 CYCLOHEXANE 1.225
0.261 0.169 ISOOCTANE 1.543
0.278 0.389 n-HEPTANE 0.713
0.611 0.443 n-OCTANE 1.380
0.111 0.172 n-NONANE 0.645
0.190 0.268 n-DECANE 0.712
^Abbreviations are R = (IN - OUT)/(IN + OUT + 10-6), RSD = relative standard deviation,
S = fl(ave)/RSD
237
-------
Table V. IN vs OUT Comparisons - Combined Data Seta
R (ave) RSD Compound 5
0.004 0.383 FREON12 0.011
-0.076 0.226 METHYLCHLORID -0.338
-0.167 0.389 ETHYLCHLORIDE -0.428
-0.032 0.077 FREON11 -0.419
-0.083 0.288 VINYLIDENECHL -0.289
0.015 0.079 DICHLOMETHANE 0.197
-0.023 0.353 FREON113 -0.065
0.015 0.162 METHYLCHLFORM 0.094
0.119 0.095 BENZENE 1.254
-0.017 0.194 CARBONTETCHLO -0.086
0.070 0.104 TOLUENE 0.672
0.414 0.346 CHLOROBENZENE 1.196
0.108 0.146 ETHYLBENZENE 0.741
0.117 0.146 m,p-XYLENE 0.804
0.068 0.214 o-XYLENE 0.318
-0.127 0.598 4-ETHYLTOLUEN -0.212
-0.099 0.612 1,3,5-METBENZ -0.162
0.073 0.190 1,2,4-METBENZ 0.384
0.290 0.492 p-DICHLORBENZ 0.590
0.199 0.221 o-DICHLORBENZ 0.901
0.341 0.448 TRICHLBENZENE 0.762
0.272 0.252 ISOBUTANE 1.078
0.250 0.177 BUTANE 1.414
0.239 0.120 ISOPENTANE 1.984
0.219 0.146 PENTANE 1.504
0.397 0.474 CYCLOPENTANE 0.838
0.228 0.150 2-METHPENTANE 1.520
0.239 0.125 3-METHPENTANE 1.906
0.219 0.199 n-HEXANE 1.096
0.184 0.150 METHCYCPENTAN 1.224
0.181 0.181 CYCLOHEXANE 1.002
0.145 0.188 ISOOCTANE 0.774
0.095 0.451 n-HEPTANE 0.211
0.333 0.602 n-OCTANE 0.553
0.122 0.679 n-NONANE 0.180
0.131 0.312 n-DECANE 0.421
aAbbreviations are R= (IN - OUT)/(IN + OUT + 10-6), RSD = relative standard deviation,
S = /?(ave)/RSD
238
-------
Table VI. Duplicate Samples Comparisonsa
R (ave) RSO Compound 5
0.113 0.319 FREON12 0.354
0.001 0.206 METHYLCHLORID 0.007
0.000 0.000 ETHYLCHLORIDE ERR
0.034 0.074 FREON 11 0.466
0.000 0.000 VINYLIDENECHL ERR
-0.350 0.287 DICHLOMETHANE -1.218
0.016 0.433 FREON 113 0.038
0.006 0.110 METHYLCHLFORM 0.058
0.063 0.064 BENZENE 0.972
0.067 0.211 CARBONTETCHLO 0.316
0.029 0.047 TOLUENE 0.624
0.052 0.191 CHLOROBENZENE 0.274
0.000 0.094 ETHYLBENZENE 0.000
0.006 0.054 m.p-XYLENE 0.103
-0.002 0.151 o-XYLENE -0.015
0.020 0.475 4-ETHYLTOLUEN 0.042
0.033 0.483 1,3,5-METBENZ 0.069
0.084 0.161 1,2,4-METBENZ 0.524
-0.156 0.589 p-DICHLORBENZ -0.265
-0.180 0.436 o-DICHLORBENZ -0.412
0.014 0.045 TRICHLBENZENE 0.316
0.021 0.074 ISOBUTANE 0.288
0.011 0.069 ' BUTANE 0.162
0.015 0.063 ISOPENTANE 0.233
0.026 0.064 PENTANE 0.411
-0.036 0.482 CYCLOPENTANE -0.075
-0.044 0.061 2-METHPENTANE -0.723
-0.038 0.068 3-METHPENTANE -0.559
0.015 0.086 n-HEXANE 0.172
-0.048 0.110 METHCYCPENTAN -0.433
0.000 0.094 CYCLOHEXANE 0.000
0.137 0.313 ISOOCTANE 0.437
0.133 0.322 n-HEPTANE 0.414
0.100 0.316 n-OCTANE 0.316
0.000 0.471 n-NONANE 0.000
-0.001 0.195 n-DECANE -0.006
^Abbreviations are /? = (IN - OUT)/(fN + OUT + 10-6), RSD = relative standard deviation,
S = ff(ave)/RSD
239
-------
Table VII. Average Concentrations Measured -Total Data Set
Compound
FREON12
METHYLCHLORIO
ETHYLCHLORIDE
FREON11
VINYLIDENECHL
DICHLOM ETHANE
FREON113
METHYLCHLFORM
BENZENE
CARBONTETCHLO
TOLUENE
CHLOROBENZENE
ETHYLBENZENE
m,p-XYLENE
o-XYLENE
4-ETHYLTOLUEN
1,3.5-METBENZ
1,2,4-METBENZ
p-DICHLORBENZ
o-DICHLORBENZ
TRICHLBENZENE
ave ppbv
0.52
0.49
0.02
0.33
0.00
0.99
19.98
0.48
0.58
0.14
0.99
0.25
0.18
0.56
0.23
0.09
0.09
0.27
0.61
0.13
0.04
Compound
ISOBUTANE
BUTANE
ISOPENTANE
PENTANE
CYCLOPENTANE
2-METHPENTANE
3-METH PENTANE
n-HEXANE
METHCYCPENTAN
CYCLOHEXANE
ISOOCTANE
n-HEPTANE
n-OCTANE
n-NONANE
n-DECANE
ave ppbv
1.20
2.22
2.37
1.05
0.09
0.53
0.37
0.32
0.17
0.15
0.19
0.11
0.07
0.05
0.22
240
-------
Table VIII. Reduced Sector Data - Compounds Assigned to IN or OUT Sectors
Compound
FREON12
METHYLCHLORID
FREON11
FREON113
METHYLCHLFORM
BENZENE
CARBONTETCHLO
TOLUENE
CHLOROBENZENE
ETHYLBENZENE
m,p-XYLENE
o-XYLENE
4-ETHYLTOLUEN
1,3,5-METBENZ
1,2,4-METBENZ
p-DICHLORBENZ
o-DICHLORBENZ
TRICHLBENZENE
ISOBUTANE
BUTANE
ISOPENTANE
PENTANE
CYCLOPENTANE
2-METHPENTANE
3-METHPENTANE
n-HEXANE
METHCYCPENTAN
CYCLOHEXANE
ISOOCTANE
n-HEPTANE
n-OCTANE
n-NONANE
n-DECANE
Moderate stat. sign.
Combined
Site 1 Site 2 data set
OUT
OUT
OUT
IN
IN
OUT
IN
IN
IN
Significant stat. sign.
Combined
Site 1 Site 2 data set
OUT
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
IN
241
-------
to
1. Chlorinated Chemical Manufacturer
2. Power Plant (Coal)
3. Oil Refinery
4. Oil Tank Farm
5. Chemical Company
6. Liquified Gas Manufacturer
7. Plastics Manufacturer
8. Truck Depot
Figure 1. Diagram of relative locations of sector sampling sites and the suspect sources.
-------
Wind Direction Sensor
Sample/ \
|rt|D+ L \
Inlet
(S3
CO
Pump
Electronics: Timer, Solenoid Control,
Wind Direction Decoder
Pressure
Regulator
Vent
Excess Flow
Pressure
Gauge
Illlllllll
Pressure
Gauge
\
SUMMACans
Out
Figure 2. Schematic of sector sampler flow arrangement.
-------
Remote FTIR Measurement of Chemical Emissions
Robert H. Kagann, Ralph DeSimone, Orman A. Simpson
MDA Scientific, Inc.
Norcross, Georgia
and
William F. Herget
Nicolet Analytical Instruments
Madison, Wisconson
We have made long path FTIR measurements as part of a study sponsored by the
Superfund Innovative Technology Evaluation (SITE) program. The purpose of the study was
to compare concentrations measured by the FTIR remote sensor with those measured, by
GC/MS techniques, on samples collected in canisters. The concept of the FTIR system is based
on the Remote Optical Sensing of Emissions (ROSE) system, designed by W. F. Herget.1 The
current system is a design variation which uses a single telescope to both transmit the ir
probe beam and to receive the return beam reflected from a retroreflector placed out in the
field. A Cassagrain telescope with a 37 cm primary mirror is directed at a remotely located
retroreflector. This arrangement simplifies system alignment as compared to the original
ROSE system. A Nicolet Model 730 FTIR system is coupled to the telescope with appropriate
transfer optics. Other advantages of the current system are smaller overall size, much
greater disk storage capacity, and a competerized concentration calculation program which
performs a classical least squares fit of absorption bands to precisely measured reference
spectra.
For the subject program, the FTIR remote sensor was mounted in the rear of an
airconditioned motor vehicle. The telescope was aimed out through an open window. The
measurements were made at seven sites in the vicinity of New Castle, Delaware. Total path
lengths ranged from 250 to 1022 meters. Elevated levels of methane were measured along
with ammonia, in the vicinity of a municipal incinerator landfill site. Chlorobenzene and
p-dichlorobenzene were measured in the vicinity of a chemical plant. Simultaneous canister
collections of the chlorobenzenes were made to test the accuracy of the remote sensor; these
are presented in another paper by Russwurm and McClenny.2
244
-------
File Name
A00731.001
A0073 1.002
A00731.003
A00731.004
A00731.005
A00731.007
A00731 .008
A00731.009
A00731.010
A00731.012
A00731.013
A00731.014
A00731.015
A00731.016
A00731.017
A00731.018
A00731.019
A00731.020
A00731.021
A00731.022
A00731.023
A00731.024
A00731.025
A00731.026
A00731.027
A00731.028
A00731.029
A00731.032
A00731.033
A0073t.034
A00731.035
A00731.036
A00731.037
A00731.038
A00731.039
A00731.040
A00731.041
A00731.042
A00731.043
A0073 1.044
A00731.045
AOO 73 1.046
A0073t.047
A00731.048
Time
—
10:51:03
11:48:28
11:52:53
11:57:36
12:02:27
12:28:20
12:32:47
12:36:52
12:41:05
13:54:59
13:57:16
14:00:48
14:03:32
14:43:59
15:12:22
15:38:06
15:41:14
15:44:12
15:47:18
15:50:17
15:53:18
15:56:16
15:59:22
16:02:25
16:05:33
16:08:50
16:11:57
17:51:28
17:59:34
18:04:36
18:08:20
18:20:37
18:48:08
19:12:36
19:16:23
19:19:34
19:22.33
19:25:34
19:28:44
19:31:43
19:34:48
19:37:47
19:40:47
19:43:45
Pathlength
(meters)
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
374
376
376
376
376
376
376
376
376
376
376
376
376
376
376
376
376
376
No. of Sweeps
256
64
64
64
64
64
64
64
64
16
16
1 6
16
256
256
64
64
64
64
64
64
64
64
64
64
64
64
64
64
64
64
256
128
64
64
64
64
64
64
64
64
64
64
64
Methane
(ppm)
.38(22)
.60(13)
.56(12)
.58(12)
.63(12)
.58(12)
.58(12)
.57(12)
.56(12)
.54(12)
.54(11)
.53(12)
.48(12)
.48(12)
.47(12)
.45(12)
.47(12)
47(12)
.50(12)
.45(12)
.46(12)
.46(12)
.46(13)
.48(13)
.50(13)
.50(13)
.58(14)
.56(17)
.57(17)
.49(15)
.82(18)
.85(19)
.85(19)
.85(19}
.75(18)
.54(16)
.50(15)
.49(16)
.49(15)
1.50(16)
1.48(16)
CO
(PPb)
90(160)
400(240)
290(170)
310(180)
410(230)
190(100)
250(140)
200(100)
390(210)
190(110)
190(100)
280(160)
500(280)
290(170)
300(170)
270(160)
310(170)
260(150)
260(150)
380(200)
280(160)
290(160)
350(200)
200(120)
200(110)
230(120)
180(110)
220(150)
210(170)
200(160)
220(160)
270(190)
260(200)
240(180)
240(190)
300(240)
310(250)
240(190)
260(200)
280(110]
240(180)
300(230)
220(170)
330(260)
p-dichlbenzene
(PPb)
62(11)
37.5(6.0)
31.6(6.3)
31.2(6.3)
51.9(4.9)
62.1(6.3)
41.0(7.4)
44.2(8.3)
47.4(5.6)
40.3(9.8)
61.7(9.1)
70.1(8.4)
45.2(8.4)
58.9(7.0)
76.5(7.7)
108(13)
58.6(9.5)
74.4(9.8)
59(10)
73.3(7.4)
49(10)
77.5(9.5)
54(10)
70(10)
109(10)
121(12)
202(13)
249(13)
260(16)
206(13)
171(16)
219(12)
137(10)
221(12)
183(11)
132(10)
165(12)
177(12)
116(10)
109(12)
120(12)
123(12)
144(11)
155(11)
Chlorobenzene
(PPb)
46(32)
80(31)
36(31)
39(24)
88(33)
43(35)
85(33)
64(32)
52(47)
63(47)
157(46)
129(38)
41(33)
70(41)
92(46)
108(53)
90(46)
64(37)
143(53)
70(51)
50(50)
420(71)
840(81)
398(72)
405(81)
109(60)
110(100)
65(54)
120(100)
130(110)
250(140)
210(140)
130(110)
Table 1 Concentration Values for measurements made at in vicinity of the Standard Chlorine
Chemical plant on July 31,1989. The numbers in the parenthese are the estimated
standard deviation of the least squares fit propagated to the concentrations
determinations.
245
-------
The FTIR remote measurements were made at seven different locations, in the
vicinity of New Castle, Delaware, on ten different days. Figure 1 shows the unistatic
configuration used for these measurements. We were able to measure the ambient levels of
the normal atmospheric constituents, carbon monoxide and methane, in any configuration in
respect to the wind direction. Measurements of the chemicals emitted from local sources,
however, requires that at least part of the ir beam intersect the down-wind flow of air from
a source. Several days of field measurements in the vicinity of the chemical plant gave no
results for local emissions because of unfavorable wind directions. On the second week when
the wind direction shifted, we returned to the chemical plant and made measures at a site
which was clearly down-wind from chemical waste stacks. Analysis of the spectra produced
by the FTIR sensor indicated the presence of chlorobenzene and p-dichlorobenzene, and we
made quantitative measurements for two days at this site. The results are shown in Tables 1
and 2. The numbers in the parenthese are the estimated standard deviations of the least
squares fit propagated to the the concentration determination. These numbers are relative to
the last digit in the concentration. The results for p-dichlorobenzene are the weighted
average of determinations made from a least squares fit of the field spectrum to three
different absorption bands, and the results for chlorobenzene is the weighted average of
determinations made from two different absorption bands.
.
1
1
^ __
1
i
1
1
\
I
I
/
1
1
(
31
1
' Retro
reflector
array
FTIR Remote
Sensor
Figure 1. The Unistatic FTIR Remote Measurement Configuration.
On a subsequent measurement day, we set up the remote sensor at two different sites
at the Pigeon Point incinerator and landfill facility. Here we had flexibility to configure the
measurement downwind of probable emission sources. The results of all the measurement
made at Pigeon Point is shown in Table 3. At the first site, which was down-wind to a
methane burning flare, we measured elevated quantities of methane up to 8.1 ppm. We moved
to the second site, which was in the vicinity of solid waste digestors, and in the first
measurement obtained a concentration of 200 ppb of ammonia in addition to an elevated
concentration of methane of 3.6 ppm. In the subsequent measurements, we no longer obtained
results for ammonia, and the concentration of methane approached ambient levels.
After a cursory investigation, we estimate that the accuracy in the FTIR remoter
sensor determined concentration is ~ 15 percent in cases where the concentrations are not
too close to the minimum detectable limits. This applies to our determinations for
p-dichlorobenzene, ammonia, and methane. We plan a more thorough investigation into the
various sources of error, in the future, in order to better determine the remote sensor's
accuracy. As is discussed by Russwurm and McClenny,2 the results for the comparison of the
p-dichlorobenzene concentrations determined by the FTIR remoter sensor to those
determined from the analysis of simulataneous canister collections showed agreement better
than the estimated uncertainties of either technique.
246
-------
References
W. F. Herget and J. D. Brasher, "Remote measurement of gaseous pollutant
concentrations using a moblile Fourier Transform interferometer system," Applied
Optics, 18, 3404 (1979).
G. M. Russwurm, and W. A. McClenny, "A comparison of FTIR open path ambient data
with method TO-14 canstster data," 1990 EPA/A&WMA International Symposium on
Measurement of Toxic and Related Air Pollutants, Raleigh, NC (April 30 - May 4,
1990).
File Name
A00801.001
A00801.002
A00801.003
A00801.004
A00801.005
A00801.006
A00801.007
A00801.009
A00801.010
A00801.011
A00801.012
A00801.013
A00801.014
A00801.015
Time
10:23:55
10:56:16
11:12:15
11:18:29
11:24:22
11:30:17
11:36:18
12:19:42
12:25:32
12:31:23
12:37:19
12:43:24
14:38:47
15:55:11
Pathlenoth
(meters)
372
372
372
372
372
372
372
372
372
372
372
372
372
372
No. of Sweeps
256
128
128
128
128
128
128
128
128
128
128
128
128
128
Methane
(DDm)
1.75(14)
1.70(14)
1.67(14)
1.58(14)
1,58(13)
1.57(13)
1.61(13)
1.60(13)
1.61(14}
1.61(13)
1.61(13)
1.64(12)
CO
(DDb)
330(190)
320(180)
300(170)
270(150)
320(180)
290(160)
300(170)
270(140)
260(140)
250(130)
260(140)
280(140)
210(110)
200(100)
j-diChlBenzene
(ppb)
295(19)
104(21)
141(25)
56(22)
60(40) I
71(24)
69(23)
45(26)
45(26)
27(26)
78(24)
52(26)
102(20)
Chtorooenzene
(DDb)
77(42)
45(44)
Table 2 Concentration Values for measurements made at in vicinity of the Standard Chlorine
Chemical plant on August 1,1989. The numbers in the parenthese are the estimated
standard deviation of the least squares fit propagated to the concentrations
determinations.
File Name
A00802.001
A00802.003
A00802.005
A00802.006
A00802.007
A00802.008
A00802.009
A00802.011
A00802.013
A00802.014
A00802.016
A00802.017
A00802.018
A00802.019
A00802.020
A00802.021
Site
Pigeon Point 1
Pigeon Point 2
Time
10:04:42
11:29:32
12:01:15
12:07:43
12:13:56
12:19:52
12:25:42
13:18:43
14:26:55
15:48:12
16:38:18
16:52:04
16:57:58
17:03:50
17:09:41
17:15:54
Pathlenalh
(meters)
312
312
312
312
312
312
312 I
312
312
309
309
309
309
309
309
309
No. of Sweets
256
256
128
128
128
128
128
256
256
256
128
128
128
128
128
128
Methane
(PDm)
7.2(4)
8.12(37)
7.04(35)
7.7(4)
6.42(38)
5.74(32)
6.7(4)
3.57(20)
1.80(15)
2.96(32)
2.08(14)
1.87(13)
2.77(18)
1.84(13)
CO
(DDb)
220(150)
200(130)
240(140)
210(130)
220(140)
230(140)
240(140)
130(70)
90(60)
340(150)
260(130)
260(190)
350(160)
310(150)
380(180)
330(150}
Ammonia
(opm)
203(19)
Table 3 Concentration Values for measurements made at the Pigeon Point municipal waste
treatment plant on August 2.1989. The numbers in the parenthese are the estimated
standard deviation of the least squares fit propagated to the concentrations
determinations.
247
-------
A COMPARISON OF FTIR OPEN PATH AMBIENT DATA WITH METHOD
TO-14 CANISTER DATA
George M. Russwurm
NSI Technology Services Corporation - Environmental Sciences
Research Triangle Park, North Carolina
William A. McClenny
AREAL> U.S. Environmental Protection Agency
Research Triangle Park, North Carolina
In order to determine the effectiveness of a Fourier transform infrared (FTIR)
spectrometer used as an open path monitor, a series of experiments were conducted in
the vicinity of New Castle, Delaware. Chlorobenzene and/7-dichlorobenzene data from
the FTIR was compared with data obtained by collecting canister samples and
performing gas chromatography/mass spectrometry (GC/MS) analyses. To adequately
cover the path used by the FTIR, the canister was transported along the path on a
bicycle for approximately one-half hour, and the FTIR data were averaged over this
time period. The p-dichlorobenzene data shows excellent agreement with the canister
data, and the bicycle technique seems to lend itself well to such comparisons.
248
-------
Introduction
One of the requirements of the Superfund Innovative Technology Evaluation
program is to compare new, emerging methods with older more mature techniques. In
this case the new technology was an FTIR long-path monitor, and the more mature
technique was the SUMMA polished canister technique, method TO-14. The FTIR
makes path-integrated measurements over physical paths that can be up to 1 km long,
whereas the canister technique is inherently a point monitoring method. This primary
difference in the two techniques creates difficulties with determining a strategy for
comparison sampling because of the spatial and temporal variations in concentration
along the path. This study was part of the Delaware field program and the data were
taken in an industrial complex south of New Castle.
Results and Discussion
Figure 1 shows a possible configuration of the infrared beam and the plume
emanating from a point source. As the distance between the source and the path
becomes smaller, the ratio of the plume dimensions to the path length also becomes
smaller. This implies that the path-integrated measurement made by the FTIR will also
become smaller, and it can indeed drop lower than the detection limits of the
instrument. The plume can also be moved along the path by the winds. These
conditions were definitely seen during the Delaware study and had been anticipated
beforehand. One condition that had not been anticipated was that the plume can move
vertically in and out of the path rapidly. The initial sampling strategy was to take
relatively long (16-min) FUR scans to enhance the signal-to-noise ratio. However, when
the plume is moving in and out of the path rapidly, the long averaging process can
lower the signal sufficiently to obviate any perceived advantages. Because of this, the
sampling time for the FTIR was changed in the field to a 4-min period, during which
256 scans were coadded with the FTIR.
In an attempt to overcome the anticipated variability of the concentration along
the path for comparison purposes, it had been decided to mount the canister on a
bicycle and ride it along the path. Because the transit time with the bicycle is long
compared to 4 min and we wanted to cover the path many times during any one
sampling period, we extended the sampling time to 32 min. Thus for comparison
purposes eight 4-min runs with the FTIR were coadded, and the bicycle was ridden
along the path for 32 min. It was felt necessary to transport the canister along the path
at a uniform speed, and after a few trial runs this became quite easy to do. As a check,
a stopwatch was mounted on the handlebar of the bicycle so that the rider could keep
track of the total elapsed time and the time for any path traversal.
Table I shows the concentrations of various gases taken with three canisters
equally spaced along the path. These samples were taken simultaneously as grab
samples (about 10 sec) and were at intervals of about 30 m. During this period, some
odors were detected at one end of the path but not the other, and the data clearly
shows concentration variations up to about a factor of 50. Shortly after these samples
were taken, the odor disappeared, being there one instant and gone the next.
During the study, only two compounds were detected at concentrations and
durations that were sufficient to provide comparison data between the FTIR and the
canisters. These were chlorobenzene and/?-dichlorobenzene. Other data was collected
that could be used as a quality control check for the mobile canister technique. This
data allowed three comparisons to be made with the canisters alone. Table II shows
the results of these comparisons. Column 1 gives the ratio of the average value
249
-------
obtained from five canisters placed along the path to the value of the concentration
measured with a canister transported along the path. Column 2 gives the ratio of
various concentrations obtained from two canisters taken simultaneously with the
bicycle. At two of the sites, it was impossible to ride the bicycle because of the terrain,
and we transported the canisters along the path on foot. When this was done, two
persons simultaneously started from opposite ends of the path and carried the canisters
for the 32-min period. The results of this comparison are shown in column 3, and
although the concentrations were all quite low, they compared well. Column 2, the
bicycle duplicate, probably indicates the comparison that can be expected with the
canister method. It thus serves to indicate what should be considered an acceptable
level for other comparisons. The comparison shown in column 3 indicates very little
temporal variability, at least during these sampling periods.
Figure 2 shows the comparison between the canisters and the FUR for
chlorobenzene. Although this shows a higher degree of variability than we would like,
the comparison seems adequate. The variability may very well be caused by the fact
that the concentrations were close to the detection limits for the FTIR.
Figure 3 shows the results of five comparisons between the FTIR and the
canisters. This comparison is excellent and covers a large range of concentrations. It
clearly shows that the FTIR can make very good measurements when the concentrations
are greater than the detection limits and when the spectra are free of interferences.
Conclusions
Two conclusions can be drawn from this work. They are that the bicycle method
for transporting the canisters along the path to make long-path comparisons seems to
work well as long as the plume is fairly well behaved over extended (30-min) periods.
The second is that the FTIR remote sensing technique seems to be a viable technique
for making long-path, open air measurements for some of the volatile organic
compounds.
Disclaimer
Although the research described in this article has been funded wholly or in part
by the United States Environmental Protection Agency through Contract 68-02-4444 to
NSI Technology Services Corporation, it has not been subjected to Agency review and
therefore does not necessarily reflect the views of the Agency, and no official
endorsement should be inferred. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
250
-------
Table I. Grab Sample Concentrations Taken Along Path (ppb)
Compound
Dichlorodifluoromethane
Benzene
Chlorobenzene
m-Dichlorobenzene
p-Dichlorobenzene
o-Dichlorobenzene
1 ,2,4-Trichlorobenzene
Can A
5.4
1.5
7.1
0.5
11.6
3.9
0.8
Can B
0.02
10.5
63.8
2.7
152.5
32.7
21.5
CanC
0.05
50.9
263.2
26.6
247.5
154.1
40.9
Table II. Canister Comparison Ratios1
Compound
Dichlorodifluoromethane
Methyl Chloride
Dichlorotetrafluoroethane
Ethyl Chloride
Trichlorofluoromethane
Dichloromethane
Trichlorotrifluoroethane
1,2-Dichloroethane
1 ,1 ,1 -Trichloroethane
Benzene
Carbon Tetrachloride
Toluene
Tetrachloroethylene
Chlorobenzene
Ethylbenzene
/n-Xylene
o-Xylene
4-Ethyltoluene
1 ,3,5-Trimethylbenzene
1 ,2,4-Trimethylbenzene
m-Dichlorobenzene
p-Dichlorobenzene
Path Avg.2
0.47
1.4
1
3.7*
3.5*
0.10
2.6
0.72
1.0*
0.45*
0.6*
0.4*
1.4*
9.9*
Bicycle
Duplicates
1.0
1.1
3.3*
1.2
1.2
1.3
0.5*
1
0.87
1*
1
1*
1
1
1.1
1
1*
2.8*
1
1
1*
Walking
Duplicates
1
1
1
0.2*
0.83
0.8
0.88
1*
1
1*
1*
1*
Notes: 1. Blank entry indicates concentration below detection limits.
'Denotes that concentrations were < 0.2 ppb.
2. Ratio of the average of 5 cans along the path to bicycle.
251
-------
IR Beam
Retroreflector
Source
Figure 1. Plume/optical beam configuration.
252
-------
200
150
a
a
o
'"§ 100
k.
§
o
o
50
Long Path FTIR
Canister
Figure 2. FTIR - canister comparison for chlorobenzene.
200
150
Q.
Q.
C
O
re 100
•»-»
0)
o
o
o
50
Long Path FTIR
Canister
Figure 3. FTIR - Canister comparison for /j-dichlorobenzene.
253
-------
SHORT-TERM SEQUENTIAL CANISTER-BASED SAMPLING NEAR
SUPERFUND SITES
Karen D. Oliver
NSI Technology Services Corporation
Research Triangle Park, NC 27709
William A. McClenny and Richard E, Berkley
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
A commercially available, 14-port sequential canister sampler was used in a
residential area of New Castle, Delaware, during July and August 1989 to automatically
collect ambient air samples for 20 min at hourly intervals. The samples were
subsequently returned to the laboratory and analyzed for 41 volatile organic compounds
(VOCs) by gas chromatography (GC) with mass selective detection and flame ionization
detection. The portable, battery-operated sampler is more easily deployed in the field
than a GC system, yet it still allows collection of hourly samples so that short-term
fluctuations in concentrations of VOCs in the ambient air can be monitored. These
variations in concentrations would otherwise be averaged out by the collection of, for
example, a 24-hr sample. The resultant hourly concentration data for a set of 25 VOCs
was combined with wind direction and wind speed data to determine which local sources
were impacting the site.
254
-------
Introduction
State and local air monitoring agencies are often faced with the task of
responding to a complaint from the public related to a persistent or noxious smell. To
respond to such a complaint, the agency needs a means of locating the source. In the
Delaware Superfund Innovative Technology Evaluation (SITE) study, a portable,
battery-operated sampling unit was deployed to develop an ambient air data base for
volatile organic compounds (VOCs) that was updated hourly at a sensitive receptor site,
in this case the backyard of a residence located in Llangollen Estates near Superfund
sites and not very far from industrial emission sources. The project goal was to interpret
the temporal variations in VOC concentrations and wind direction (and speed) to
identify local VOC emission sources.
Experimental
An automated, computer-controlled, 14-port sequential canister-based sampler1
(SIS, Inc., Moscow, ID) and 6-liter SUMMA polished canisters (SIS, Inc.) were used
to collect 18 ambient air samples at hourly intervals in a residential area of New Castle,
Delaware. Purge/sample/pause times were programmed into the computer. Laboratory
tests were performed to certify that the sequential sampler was not contaminated. The
certification consisted of passing a mixture of VOCs at 10-ppbv levels through the
canister-based sampler into a preconcentrator/gas chromatograph (GC) for analysis.
The results of analysis were compared to the results obtained by passing the same
sample directly into the GC. Table I shows the results expressed as a percentage
difference referenced to the control results at each of three ports on the sampler. To
further test the sampler, it was placed outside in the sun on a hot day and used to
collect ambient air into a canister. Simultaneously, a canister equipped with a mass flow
controller set to sample at the same flow rate as the sequential sampler was used to
collect ambient air. The agreement as shown in Table II was excellent.
During the experiment in Delaware, the sampler was placed at the backyard site
and programmed to allow for sampling every hour. Prior to the collection of each
sample, all sample lines were automatically purged for 4 min with ambient air. Next,
ambient air was collected for 22 min at a flow rate of 400 cm3/min, resulting in a total
volume of 8.8 liters of sample collected in each canister. After sample collection, the
sampler paused for 34 min before beginning the next purge/sample/pause cycle.
Operation of the sampler was begun at 14:04 on July 25 and was completed at 07:46
on July 26, 1989. Wind speed and wind direction data were recorded every 15 min at
a site approximately 1.6 km southeast of the residential area by using a portable,
battery-operated meteorological data system (Climatronics, Bohemia, NY).
After sample collection, the canisters were returned to the laboratory for analysis
of VOCs using an automated cryogenic preconcentration GC system with flame
ionization and mass selective detectors (GC-FID/MSD). The peak areas of 25 of the
most prevalent VOCs detected by flame ionization (Table III) were compared by using
the Temporal Profile Analysis (TPA) technique2 to determine which volatile compounds
were correlated with each other. Wind direction and wind speed data were then used
to calculate the trajectories of the air masses which moved across the site during and
prior to the sampling periods.3
Results
Analysis results for the 18 samples indicated relatively low levels of VOCs at the
residential site during most of the sampling period. However, significant increases in
255
-------
concentrations of VOCs were observed at 1:30 a.m. and 5:30 a.m. on July 26. During
these occurrences, p-dichlorobenzene, as measured by GC-MSD, increased from an
undetectable level (<0.05 ppbv) at 12:30 a.m. to 5.2 ppbv at 1:30 a.m. and was again
undetectable by 2:30 a.m. /jara-Dichlorobenzene then increased from an undetectable
level at 4:30 a.m. to 14.0 ppbv at 5:30 a.m., and by 6:30 a.m. it was present at less than
0.1 ppbv. The TPA results indicate that a number of compounds were well correlated
with p-dichlorobenzene during these occurrences. All of the VOCs in Table III with
the exception of isoprene, 1,2,4-trimethylbenzene, and unknown 1 correlated strongly
with each other, as indicated by a correlation coefficient of 0.85 or greater.
Plots of the trajectories of air masses during the hour prior to the concentration
increases at 1:30 a.m. and 5:30 a.m. show that the air mass travelled to the sampling site
from the south on both occasions. One of the trajectories is shown in Figure 1. Many
industrial sources such as the Texaco refinery and Standard Chlorine are located
approximately 4 km south of the residential area, and the compounds which increased
may have been carried to the residential area from those industrial sources.
Conclusions
A battery-operated, portable sequential sampler was successfully operated in a
residential community to collect a series of hourly ambient air samples. The data base
of VOC concentrations versus hour was interpreted to give the emission mix apparently
originating from a local source by using temporal profile analysis. The success of this
and other recent, similar experiments implies that VOC emission sources near sensitive
receptor sites can be characterized and located.
Acknowledgement
The authors wish to thank M. W. Holdren of Battelle Columbus Laboratories for
analyzing the samples discussed in this paper.
Disclaimer
Although the research in this article has been funded wholly or in part by the
United States Environmental Protection Agency through Contract 68-02-4444 to NSI
Technology Services Corporation - Environmental Sciences, it has not been subjected
to Agency review and therefore does not necessarily reflect the views of the Agency and
no official endorsement should be inferred. Mention of trade names or commercial
products does not constitute endorsement or recommendation for use.
References
1. M. W. Holdren, D. L. Smith, J. P. Krasnec, "Sequential canister-based
samplers for collection of volatile organic compounds," Proceedings of
Measurement of Toxic and Related Air Pollutants. AWMA VIP-13, Air &
Waste Management Association, Pittsburgh, 1989, pp 60-64.
2. W. A. McClenny, K. D. Oliver, J. D. Pleil, "A field strategy for sorting volatile
organics into source related groups," Environ. Sci. Technol.. 23: 1373 (1989).
3. G. M, Russwurm, W. T. McLeod, "Spatial and temporal correlations of wind
direction and speed over a small region," Proceedings of Measurement of
Toxic and Related Air Pollutants. Air & Waste Management Association,
Pittsburgh, 1990.
256
-------
Table I. Sample integrity test of sequential canister sampler
Difference, %*
Compound Port 5 Port 9 Port 14
Freon 12 13.6 7.33 5.33
Methyl chloride 7.91 0.57 -4.70
Freon 114 3.72 -4.02 -9.13
Vinyl chloride -3.62 -5.79 -7.48
Ethyl chloride 3.85 0.52 -1.46
Vinylidene chloride 2.72 1.28 -1.49
Dichloromethane 4.01 1.72 -2.14
Freon 113 3.98 2.65 -1.88
1,1-Dichloroethane 3.15 2.01 -1.91
cis-l,2-Dichloroethene 3.19 2.99 -0.62
Chloroform 1.63 1.85 0.69
1,2-Dichloroethane 2.73 2.20 0.22
Methyl chloroform 2.48 1.57 -0.66
Benzene 2.24 1.44 -0.45
Carbon tetrachloride -3.62 -0.28 1.95
1,2-Dichloropropane 2.51 1.69 -0.97
Trichloroethene 3.59 3.05 1.05
1,1,2-Trichloroethane 2.12 3.16 0.98
Toluene 1.94 3.47 2.96
1,2-Dibromoethane 0.81 5.47 3.37
Tetrachloroethene 3.01 4.66 2.28
Chlorobenzene 3.08 6.16 4.68
Ethylbenzene 2.74 2.53 1.48
m^-Xylene 2.06 3.91 3.32
Styrene 4.04 6.56 6.78
4-Ethyltoluene 3.95 8.06 8.90
1,3,5-Trimethylbenzene 3.95 10.3 10.6
1,2,4-TrimethyIbenzene 5.39 9.23 11.3
p-Dichlorobenzene 14.2 18.2 24.3
o-Dichlorobenzene 14.2 15.7 19.1
1,2,4-Trichlorobenzene 16.8 25.8 36.4
Hexachlorobutadiene 3.13 6.60 16.2
'Difference was computed as (Port - control)/(control) * 100
257
-------
Table II. Comparison of ambient air samples collected using the sequential canister
sampler and a canister equipped with a mass flow controller
Concentration, ppbv
Compound
Freon 12
Freon 11
1,1,1-Trichloroethane
Benzene
Carbon tetrachloride
Toluene
m,/?-XyIene
4-Ethyltoluene
1,2,4-Trimethylbenzene
PortS
3.42
9.07
0.22
0.84
0.15
1.19
0.55
0.60
0.21
Canister
3.44
9.49
0.23
0.85
0.16
1.08
0.52
0.61
0.25
Table HI. Twenty-five most prevalent compounds at a residential sampling site in New
Castle, Delaware
Propane and propene
Isobutane
Unknown 1
Butene
n-Butane
Isopentane
/i-Pentane
Isoprene
fro/u-2-Pentene
2-Methyl-2-butene
Unknown 2
2-Methylpentane
3-Methylpentane
Hexane
Benzene
3-Methylhexane
Heptane
Toluene
Unknown 3
Chlorobenzene
m^-Xylene
o-Xylene
1,2,4-TrimethyIbenzene
/7-Dichlorobenzene
Substituted benzene
258
-------
X*
Figure 1. Trajectory of air mass prior to detection of high concentrations of VOCs
at Llangollen Estates.
259
-------
SPATIAL AND TEMPORAL CORRELATION OF WIND DIRECTION AND
SPEED OVER A SMALL REGION
George M. Russwurm
NSI Technology Services Corporation - Environmental Sciences
Research Triangle Park, North Carolina
Willie T. McLeod
AREAL, U.S. Environmental Protection Agency
Research Triangle Park, North Carolina
During a recent field study in Delaware a set of wind direction and speed data
was collected from four sites spread over about 10 km. An attempt was made to
determine whether the data from two of these sites separated by 3 km show any
correlation between the wind speed and direction. Although the back trajectories
calculated from a subset of these look very similar, it is difficult to show a definite
correlation between the two sets. The data show that when the wind speed is less than
about 3 mph careful consideration must be given to its use for these calculations.
260
-------
Introduction
During the Delaware study, a set of wind speed and direction data was taken at
two stations separated by 3 km. These data were taken continuously as 15-min vector
averages for a 10-day period. One of the sites was on top of a one-story building that
was situated at the top of a small hill overlooking the Delaware River. The second was
3 km south at about the same elevation (perhaps 20 ft lower). This station overlooked
the Delaware River also, and both were about one-half mile from it. A subset of the
total data representing a continuous block of four days taken at both of these stations
is used for the remainder of this discussion.
Results and Discussion
One way to present wind data is to plot wind roses, and these are shown for the
two sites for the four-day period in Figure 1. These are plotted with 22.5 ° segments in
azimuth and 5-percent increments radially. For these two stations, the calm periods
represented about 5 and 15 percent of the total time. Other than giving some general
information about the character of the winds, these diagrams do not appear to be too
informative, however. They do not, for example, give the sequence of events of when
the wind was coming from what direction and with what speed. These plots do show
that the predominant winds for this period were from the two northerly quadrants at
both stations. The distribution of the angle of the wind at P4 is significantly different
than that at the substation. The implication is that if these sets of wind data are used
for back trajectory calculations, differences in the air parcel paths should be expected.
For some of the experiments conducted in Delaware, an attempt was made to
use the wind information to track the air parcel back from the instruments to an
industrial complex. An implicit assumption in doing this is that if the wind speed and
direction are measured at point A (the monitor), there is some confidence that this wind
pattern is the same at point B (the source), a short distance away. It was hoped that
these two data sets would allow that assumption to be tested, and the four-day back
trajectories are shown in Figure 2. It should be pointed out that we were interested in
trajectories going back in time for a few hours and not days; the four-day span is used
to show the trajectories more clearly. These plots are on a scale of 1 to 2.5 million in
order that the trajectory would fit on an 8^- x ll-in. sheet of paper; thus, the
geographical area covered is quite large compared to the region we were interested in.
Also, all the calculations were done by using vector addition or averages, and no scalar
mathematics was used.
At first glance, these curves appear to be quite similar and to be correlated.
Proving this is another matter, however. There are two sections to these curves; one
is the lower, more or less horizontal section and the other, the vertical section. If the
scale is kept in mind, the two lower sections are seen to diverge quite rapidly. Although
there seem to be differences in the wind speed, the angular divergence predominates
in this region of the trajectories, at least initially. The reverse is true for the vertical
sections of the trajectories, where the wind directions appear to be in much better
agreement.
This interpretation is borne out in Figure 3, which shows the geographical
distance between the two trajectories as a function of time. This graph shows that the
two trajectories diverge to about 70 miles over the first 24 hr, but then no further
significant divergence occurs.
To further analyze this data set, use was made of the fact that these data points
261
-------
are vectors and the dot product could supply useful information. The interpretation of
the dot product is similar to that for the correlation coefficient used with scalar data.
The average wind speed for each pair of data points was calculated, and then
the ratio of the magnitudes versus average wind speed was plotted. The plot is shown
in Figure 4. Actually, this plot was done by sorting the data on average speed and then
plotting it as a function of sequence number. This makes viewing the data clearer, but
it should be noted that only 15 percent of the total data lies above 6 mph. There are
two things to be noted about this curve. The first is that at less than about 3 mph the
data are quite variable. If the wind speed were perfectly correlated, this plot should be
a horizontal line with a value of 1. The second point to notice is that there seems to
be a lower value to the data, even at higher speeds, that is different from 1. From the
actual data this value is seen to be 1.6. The first point implies that, when the wind
speed is less than about 3 mph, using the data for back trajectory calculations should
be done only with the recognition that significant error may be present. The second
point indicates that there may be something wrong with the instruments in that there
is a constant bias between them. When the data for the trajectories are replotted after
all the wind speeds from the substation are multiplied by the factor 1.6, as shown in
Figure 5, the resulting coincidence is at first much better. Note, however, that the lower
portion of the curves still diverges almost as much as it did in the original. If the
trajectories are replotted by using only 2h days of the data, as shown in Figure 6, the
vertical portions of the curves are in good agreement, and we do not have to invoke the
problems of large instrumental bias. At this time it is felt that the instruments were
functioning properly, and the differences in the first part of the curves are real.
This raises an interesting point, however, about the calibration of wind
measurement devices. The only accepted procedure for calibrating the anemometer
is to put it in a wind tunnel Although portions of the instruments (electronics) can
be checked in the field, there are apparently no field procedures to verify the accuracy
of the entire system under field conditions. This is indeed a severe handicap.
The last graph, Figure 7, shows the cosine of the difference of the two wind
directions plotted as a function of average wind speed. This should have a value of 1
if the two directions are colinear and a value of -1 if they are 180° apart. Again, the
data is quite variable at wind speeds less than about 3 mph.
Conclusions
This data set implies that a lower limit exists for the use of wind information in
the calculations of back trajectories. When the wind speed is less than about 3 mph,
the assumption that a single data set holds for even a small region surrounding the site
probably does not hold true.
262
-------
P4
Sub
Figure 1. Wind roses P4 and substation.
Figure 2. Four-day back trajectories.
Time (days)
Figure 3. Separation distance versus time.
263
-------
26
"O
3
I
o
1
cc
Average Speed
Figure 4. Ratio of magnitudes versus average speed.
Sub
Figure 5. Four-day back trajectories (sub x 1.6).
264
-------
Sub
IN
Figure 6. Back trajectories for 2k days.
+1:
Average Speed
-1
Figure 7. Cos (e - *) versus average speed.
265
-------
SPECIATION OF ORGANIC COMPONENTS OF
MOBILE SOURCE EMISSIONS
Kenneth Knapp, John E. Sigsby Jr., Fred D. Stump, and David
Dropkin, U.S. Environmental Protection Agency, Research Triangle
Park, NC 27711 and Charles Burton and William Crews, NSI Technology
Services Corporation, Research Triangle Park, NC 27709
Both the evaporative and tailpipe emissions from mobile
sources are complicated mixtures of organic and inorganic
compounds. To get a better understanding of these emissions and to
determine effects of fuel, vehicle and operating conditions on the
emissions, speciation of the many, sometimes over 250, organic
components is needed. This paper describes the four different gas
chromatographic methods used to get the maximum speciation
information. The methods for diluted emission samples collected in
Tedlar bags are: Most of the compounds from C4 to C12 are analyzed
on a 60m DB-1 capillary column programmed from either -60°C or -
50°C to about 180°C with a hydrogen flame ionization detector (FID) ;
methane is analyzed on a molecular sieve-Poropak QS column at 30°C
with an FID; C2 - C3 are analyzed with a silica gel column at 30°C
with an FID; and the seven C4's are analyzed with a 0.19% picric
acid on Graphpac at 43°C with an FID. GC methods have also been
developed to analyze fuels including oxygenated compounds such
methanol, ethanol, methyl tertiary butyl ether (MTBE) and ethyl
tertiary butyl ether (ETBE), compounds used in alternative fuels.
266
-------
INTRODUCTION
Gaseous hydrocarbon emissions from motor vehicles are of
concern primarily because many of these hydrocarbons are oxidant
precursors in atmospheric photochemical processes1"4. These
transportation related pollutants have been characterized as one
of the most significant air pollution problems. In most Urban
areas, the hydrocarbon emissions from mobile sources account for
more than half of these precursors. Quantitative speciation of
vehicle hydrocarbon emissions is needed, to elicit a better
understanding of the Mobile Source contribution to this ozone
(oxidant) nonattainment problem. Another area of interest for
these analyses is the potential toxicity and carcinogenicity of
the components in Mobile Source emissions5. Benzene and 1,3-
butadiene, which are regarded as hazardous pollutants, are two
examples of compounds found in large amounts in automobile
exhaust. Accurate determination of these and other compounds are
difficult because of the large numbers of similar compounds many
of which coelute from gas chromatographic columns when analyzing
vehicle emissions.
In the Mobile Sources Research Branch (MSERB) of the U.S.
Environmental Protection Agency several different gas
chromatographic (GC) methods are used to obtain the maximum
speciation information of vehicle emissions. Two different gas
chromatographs are currently in use in this laboratory to
analyze for hydrocarbons. The first instrument has been modified
into two sections. In the first section an aliquot of the
collected emissions sample is taken to measure the compounds
methane, ethane, ethylene, and acetylene. A modification
recently made to this section requires two samples for analysis
but extends the analysis so that propane and propylene are also
measured. The second section of this instrument is used for the
detailed analysis of the C, to C13 hydrocarbons. The second
instrument is configured into four sections and four samples
are, therefore, required for complete speciation analysis. In
one section of this instrument only Methane is measured. Ethane,
ethylene, acetylene, propane, and propylene are measured in
another section of this instrument. The seven C4's that are found
in Mobile Source emissions are measured in the third section6.
The detailed analysis of the C4 to C13 hydrocarbons is done in the
fourth section7.
Gasoline and other fuels are run on another GC instrument
similar to that used for the speciation of the C4 to C13
hydrocarbons. Liquid samples are introduced to the separating
columns by flash vaporizing the sample in a heated injection
port on the GC. In addition, liquid samples can be analyzed by
vaporizing an aliquot of the sample into a heated manifold which
is continually being purged with a metered amount of zero grade
air.
Gas Chromatography/ Mass Spectrometry is used for positive
identification of the peaks and to determine if peaks contain
more than one component and whenever possible identify the
components and their compositional ratios.
Another gas chromatographic analytical system is used to
267
-------
analyze samples for oxygenates such as ethanol, and methyl
tertiary butyl ether (MTBE) when alternative fuels are tested.
This system involves cryogenically concentrating an aliquot of
the sample and analyzing the ethanol and MTBE through a dual
column arrangement with an additional cryofocusing step8.
Methanol is measured with another gas chromatograph since
emissions samples known to contain methanol are pumped through
water and then an aliquot of the water is taken for analysis.
The carbonyl compounds which may be present in vehicle
emissions samples, such as aldehydes and ketones, are collected
on cartridges containing silica gel treated with
2,4- dinitrophenylhydrazine. The derivatized aldehydes and
ketones are then analyzed by high performance liquid
chromatography (HPLC).
EXPERIMENTAL
Sampling
Gas samples were collected in Tedlar bags (E.I. du Pont
de Nemours and Co. Wilmington, Del.) from the Constant Volume
System (CVS) and evaporative emissions test cell. New bags are
flushed four to five times with zero air to eliminate
contamination problems.
Sample Introduction
Two different methods are used to introduce the samples
into the chromatographic system from the bags. One method
involves the use of a sample loop constructed of clean
chromatographic grade stainless steel tubing in 1 ft. , 5 ft. , or
10 ft. lengths and 1/8 OD. The loops are coiled to fit inside a
controlled temperature sample housing unit or to fit inside the
gas chromatograph. The loop has been prepared for sample
analysis by flushing it with methylene chloride and then dried
with zero grade air prior to use.
The other method used for sample introduction involves
the use of a hydrocarbon trapping system consisting of a
stainless steel tube 6 in x 1/8 in OD packed with 4 in. of
untreated glass wool. This trapping system can be used to
analyze low concentrations of mobile sources exhaust. Sample
collection volumes of up to 1500 cm3 have been used with this
trapping system. This trap extended the detection limits for
most hydrocarbons to the 15.0 parts per trillion carbon (pptrC)
level.
A Nutech Model 320 controller (Nutech Inc. Durham,
N.C.) maintains concentration and desorption temperatures in the
individually controlled traps for the C2 to C3 C4's, and C4 to C13
analyses and heats the 1/16 OD stainless steel transfer lines.
The traps and all the heated lines are wrapped with insulation
tape. The transfer lines and Seiscor 6-port switching valves are
maintained at 120°C while the Nutech controller holds the sample
loop at -170°C with liquid nitrogen for the C2 and C3
concentrations, -155°C for the C4's, and -135°C for the C4 to C13
concentration. During the C4 to C13 analyses the trap is
backflushed with helium at a trap temperature of 150°C.
268
-------
Gas Chromatographic Methods
The two gas Chromatographic analytical systems that are
currently in use in this laboratory to analyze for hydrocarbons
are a Hewlett Packard model 5880 and a Perkin-Elmer model Sigma
2000. Both GC systems use the same basic methodology but with
different columns or slightly different oven temperature
programming parameters due to the basic differences in the
instruments physical configuration. Both GC systems utilize a
flame ionization detector (FID) and transfer raw data to a PC-
AT or compatible computer where the data is verified. Many of
the same samples are analyzed by both systems in order to
maintain quality control/quality assurance of our methods.
The C, to C4 analytical procedure, shown in Figure 1, is run
on the Hewlett Packard GC. Methane is measured by directing a
1 cm3 aliquot of the bag to flow through a 6 ft. x 1/8 in. o.d.
stainless steel column packed with 60/80 mesh 13x Molecular
Sieve. The column is operated at ambient with a 30 cm3/min helium
carrier flow. Ethane, ethylene acetylene, propane and propylene
(C2 to C3) are analyzed by taking a 5 cm3 aliquot of the bag and
directing the sample to a 6 ft. x 1/8 in. OD stainless steel
column packed with 40/60 mesh type 58 Silica Gel. The column is
operated at ambient with a 30 cm3/min helium carrier flow. The
C4 hydrocarbons, iso-butane, n-butane, 1-butene, isobutylene,
1,3-butadiene, cis-2-butene, and trans-2-butene are separated on
a 9 ft x 1/16 in. OD teflon coated stainless steel column packed
with 0.19% picric acid on 80/100 mesh Graphpac ( Altech Assoc.
Inc. Waukegan, Rd, Deerfield 111.). The column is operated
isothermally at 43°C with a 30 cm3/min helium carrier flow.
The C1 to C2 hydrocarbons are analyzed with the Perkin Elmer
GC as shown in Figure 2. A 10cm3 sample is directed through a 6
ft x 1/8 in. OD stainless steel column packed with 40/60 mesh
Poropak Q/QS and a 6 ft x 1/8 in OD stainless steel column
packed with 40/60 mesh type 58 Silica Gel. The column is
operated at ambient with a 30 cm /min helium carrier flow. After
16 min the columns are backflushed to remove the higher
molecular weight hydrocarbons.
A recent modification to the C1 to C2 analytical system
that allows for the measurement of propane and propylene is
shown in Figure 3. The main differences between the c, and C2
analytical method and the Ct to C3 method is that a 6 ft x 1/8
in. OD stainless steel column containing 60/80 mesh Molecular
Sieve 13x is connected in series with the Poropak Q/QS column
and that two samples are required for analysis. Methane is
measured first by the Molecular Sieve/Poropak Q/QS columns and
then the system is backflushed so that the second sample can be
analyzed for the C2 and C3 compounds with the column containing
Silica Gel.
The analytical systems used to do the C2 or C, to C12 or C13
separations are shown in Figure 4, using the Perkin -Elmer GC,
and in Figure 5, using the Hewlett Packard GC. Both GC's use a
60 meter x 0.32 mm ID fused silica capillary column to separate
the relatively higher molecular weight hydrocarbons. The
capillary column is coated with a 1.0 urn film thickness bonded
methyl silicone liquid phase (DB-1, J&W Scientific. Folsom
269
-------
Calif.) The oven temperature programing parameters are as
follows:
Initial oven temperature: -60°C
Hold at initial temperature: 6.5 min.
Oven temperature programming rate 1: 30°C/min. to -45°C.
Oven temperature programming rate 2: 6°C/min.
Final oven temperature: 175°C.
Detector temperature: 225°C.
Ethanol and MTBE are analyzed on a separate GC system shown
in Figure 6. An aliquot from a bag sample is cryogenically
concentrated at -155°C and then desorbed at 150°C. The sample is
directed first into a 7 ft x 1/8 in. OD teflon column packed
with 15% 1,2,3-tris cyanoethoxy- propane (TCEP) on 80/100 mesh
firebrick with hydrogen as the carrier gas at a flow rate of 5.0
cm3/niin. After the light hydrocarbons are vented (0.2 to 0.3 min)
the column is backflushed to the head of another column where
the sample is cryofocused with liquid CO2. The alcohols and MTBE
are then analyzed with a 25 meter x 0.53 mm ID fused silica
capillary column coated with a 5.0 um film thickness methyl
silicone bonded liquid phase. The columns are operated
isothermally at 60°C and the detector temperature is set at
200°C.
Methanol is analyzed with a separate analytical system.
This method requires that methanol be collected in water. The
automobile exhaust sample is pumped through two impingers
connected in series in order to trap the methanol in the water.
A 1.0 ul aliquot of the water in the impinger is then analyzed
by gas chromatography. The column used to separate the methanol
from the water is a 25 m x 0.53 mm ID fused silica capillary
column coated with a 5.0u film thickness methyl silicane bonded
liguid phase. The carrier gas is helium with a flow rate of 20
cmvroin. The injection port temperature is 118°C, the detector
temperature is 150°C and the oven temperature is at 100°C.
Calibration of systems
Both the oxygenate and detailed hydrocarbon analyses are
calibrated by external standard methods. The C, to C12 hydrocarbon
analysis is calibrated by using a cylinder containing 16
hydrocarbons ranging from C2 to Cn. Individual hydrocarbons are
also analyzed by injecting the hydrocarbons into a Tedlar bag
using the apparatus shown in Figure 7 to verify each component
in the calibration cylinder as well as to confirm
identifications made by GC/MS. Ultra zero grade air is swept
through the inj ector where the hydrocarbons and oxygenates are
flash vaporized at 150°C into -an evacuated Tedlar bag. The
calibration mix undergoes the same analysis as the unknown
sample. The area counts of the standards are compared to the
area counts of the unknown to calculate concentrations of the
unknowns. Weekly calibrations are performed on each instrument,
with at least five consecutive analyses to calculate precision
and check for instrument drift. Daily span checks are compared
to the control chart to ensure that the instrument is operating
within the control limits on any given day. If the daily span
checks are more than two standard deviations from the mean of
the weekly calibrations then a second span check is performed.
Corrective action is taken if the second check is also out of
270
-------
control. The first out of control check is treated as an anomaly
if the second check is within the control limits.
SUMMARY
Several GC methods have been described that permit maximum
speciation of mobile source emissions. These methods entail the
separation of the C1 to C13 hydrocarbons as well as the separation
of oxygenate fuel additives such as methanol, ethanol, and MTBE.
These methods are applicable to the analyses of diluted exhaust
and evaporative emissions. In addition, with the use of
cryogenic concentrating traps the limits of detection can be
lowered to the low ppbC levels required for ambient air
analysis. Gasoline, other fuels and pure liquid compounds are,
also, analyzed by these methods using either direct liquid
injection or by vaporization of the liquid into a heated and
zero air purged manifold that is connected to a bag.
REFERENCES
1. Cox, R.D.; Devitt, M.A.; Lee, K.W.; Tannzhill, G.K.;
"Determination of Low Levels of Total Nonmethane Hydrocarbon
content in Ambient Air," Environ. Sci. Technol. 16: 57-61
(1982) .
2. Lonneman,W. A.; Kipczunske, S.L.; Barely, P.E.; Sutterfield,
F. D.; "Hydrocarbon Composition of Urban Air Pollution,"
Environ. Sci Technol. 12: 459-463 (1978).
3. Black, F.M.; High, L.E.; Fontijn, A.; "Chemiluminescence
Measurement of Reactivity Weighted Ethylene-Equivalent
Hydrocarbons," Environ. Sci. Technol. 11: 597-601 (1977).
4. Cox, R.D.; Earp, R.F.; "Determination of Trace Level
Organics in Ambient Air by High Resolution Gas
Chromatography with Simultaneous Photoionization and Flame
lonization Detection," Anal. Chem. 54: 2265-2270 (1982).
5. Loffe, B.V.; Isidorov, V. A.; Zenkevich, I. G.; "Certain
Regularities in the Composition of Volatile Organic Pollutants
in the Urban Atmosphere," Environ. Sci. Technol. 13: 864-868
(1979).
6. Crews, W.S.; and Stump, F.D.: "Gas Chromatographic Method
for the Quantitation of 1,3-Butadiene and Other C4
Hydrocarbons in Automobile Exhaust," Submitted for
publication to Anal.Chem. October 1988.
7. Duncan, J.W.; Burton, C.D.; and Crews, W.S.; " A Method for
the Measurement of Methanol, Ethanol and Methylterbutyl
Ether Emissions from Motor Vehicles," Submitted for
publication to J, Chrom. Sci^ September 1988.
271
-------
Figure 1. C to C Analytical System
Valve 2
10 cm3sample loop
,GB>,
Vent
ro
1 cm3sample loop
Sam
Column
Flush
Vent
Cryogenic Trap
Valve 1
Valve 3
Column (Methane)
Flush Molecular Sieve
13X Column
Valve 4
to C3)
Column Silica Gel
Flush Column
olumn
Flush
Vent
0.19% Picric Acid Column
FID (1,3-butadiene)
-------
-Sample
[•O
-J
w
V2
Sample
Pump
Poropak
Helium
Vent Helium Carrier
Carrier
Silica
JGel
FID
<
Air
Figure 2. C1 to C2Hydrocarbon Analysis
-------
0
5 cm
Sample
Molecular Sieve
Poropak QS
Silica Gel
Sample
Pump
Helium
Vent Helium Carrier
Carrier
Vent Helium
Carrier
Figure 3. C.,to C3Hydrocarbon Analysis
-------
1 1
0) CO
o
.a>
a
I £
o
CO
O
CO
"co
CD
c
o
CO
o
o
CD
D
O)
275
-------
Sample
Helium
flush
n
Helium
carrier gas
. auxiliary
vent Hel urn
Hydrogen
carrier gas
vent
ss cylinder
pump
Silica Get
column
valve oven
Figure 5. Cpto C^ ^Hydrocarbon Analysis
-------
Nitrogen
Gas
Nutech
Temperature
Controller
Cooling
Coil '
Hydrogen
Flush
Toggle
Valve
Vacuum
Gauge
©—o
Toggle
Valve <™2>
Vacuum
Pump
Toggle
Valve
OV1)
Vacuum
Reservoir
Jumper
o-
Hydrogen
Flow
Togg|e
(TV4) Valve
15%TCEP
o Firebrick
2.15m
CO2
(Cryogenic
Focusing)
FID
o
5 I P-E 25 m
0 Methyl Silicone
Column
Figure 6. ETOH-MTBE Analytical System
-------
Zero Grade Air
Heated Manifold
Metering
Valve
NS
-4
00
Pump
Electronic
Flow
Controller
Septum Injector
15(fC
Tedlar
Sample Bag
Figure 7. Heated Manifold for the Preparation
of Calibration Mixes and Standards
-------
COMPARISON OF DATA FROM A FOURIER TRANSFORM INFRARED AUTOMOTIVE
EMISSIONS SAMPLING AND ANALYSIS SYSTEM WITH THOSE OBTAINED FROM
CONVENTIONAL AUTOMOTIVE EMISSIONS ANALYTICAL INSTRUMENTATION
Alexander 0. McArver. Richard. F. Snow
NSI Environmental Sciences
Research Triangle Park, NC 27709
John E. Sigsby, Jr.
Atmospheric Research and Exposure Assessment Laboratory
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
Comparisons were made of a Mattson Fourier transform
infrared spectroscopy based Real time Automotive Emissions
Analyzer (REA) and classical analytical techniques. Data from
the REA for CO, CO2, and NOx were compared to both integrated
real time and bag results from conventional emissions analysis
instrumentation. REA formaldehyde data were compared to results
from the DNPH derivatization/LC technique. REA methanol data
were compared with GC analysis of samples collected in water
impingers. The exhaust emissions from automobiles fueled with
gasoline, methanol, and methanol blends were analyzed as well
as calibration mixes. The REA also allowed the acquisition of
data on the emission composition of a variety of other compounds
such as N20 and NO. Generally good agreement was obtained
between the REA results and those from the established
techniques.
279
-------
Introduction
The real time analysis of automotive emissions for
regulated and unregulated emissions has received increased
interest in recent years, especially with regard to
alternatively fueled vehicles. Several research groups have
developed FTIR based real time analysis systems recently.1234 A
FTIR based real time automotive emissions analysis system has
been acquired and emissions data obtained from it have been
compared to those obtained from conventional analytical
instrumentation.
Experimental
All FTIR real time data were collected on the Mattson
Instruments (Madison, Wi) REA (real time emissions analyzer).
The REA consists of a Mattson Nova Cygni 120 spectrometer,
Masscomp 5300 computer and proprietary software developed by
Ford Research Labs and Mattson Instruments. The sample flows
through a 21.75 meter variable pathlength Wilkes gas cell made
by Foxboro. Spectra are taken at 0.25 cm-1 resolution,
transformed and concentrations calculated such that
concentration values are produced once every 3 seconds for
approximately 22 selected compounds. Below is a listing of
currently available compounds.
Table I. Possible Compounds from the REA.
Hydrocarbons Inorganics Oxygenates Others
Methane Carbon Monoxide Formaldehyde Ozone
Ethane Carbon Dioxide Acetaldehyde Carbonyl Sulfide
Ethylene Water Methanol Hydrogen Chloride
Acetylene Nitric Oxide Ethanol Hydrogen Cyanide
Propane Nitrogen Dioxide Ketene Ammonia
Propylene Nitrous Oxide Formic Acid Sulfur Dioxide
1,3-Butadiene Nitrous Acid Carbon tetra-
i-Butylene Nitric Acid Chloride
trans-2-Butene
i-Octane
These concentration data were transferred to an IBM-AT computer
where they are averaged and summarized for each bag of the
driving cycle using Lotus macros and Pascal programs written in-
house. A Horiba CVS (constant volume sampling) system supplied
the diluted automotive exhaust samples. Approximately 70
standard FTP driving cycles (UDDS) were performed to produce the
automotive emissions data. A 1988 GM flex fueled Corsica was
operated on MO, M25, M50, M85, and M100 methanol fuel at 40, 75,
and 90F for these FTP cycles. Conventional analysis data of C02,
CO, and NOx were acquired from the CVS analyzers. A real time
data acquisition system operated on an IBM-AT computer by Data
Translation Notebook (DTN) provided l second time resolved data
from the CVS analyzers. Formaldehyde analysis was performed by
the LC-DNPH method described by Tejada.5 Methanol analysis was
performed by the GC-impinger technique.6 Bag samples were taken
from the CVS system in Tedlar bags and analyzed after the test
was completed. The general layout of the testing facility and
sample collection has been described elsewhere.7
280
-------
Discussion
Comparisons of the REA results show generally good
agreement with those from the conventional analysis
instrumentation. Specific comparisons are in the following
table. Ratio averages for Bags 1, 2, and 3 are listed in
descending order for each compound.
Table II. Ratio Comparisons REA vs Conventional Analysis
DTN RT/CVS
REA RT/CVS
Carbon dioxide
1.202(.028) 1.007(.042)
1.023(.015) 1.033(.059)
1.028(.018) 1.021(.055)
Carbon monoxide
1.069(.041) 0.981(.096)
1.030(.056) 0.987(.117)
1.033(.087) 1.030{.141)
Nitrogen oxides
0.996(.077) 1.088(.051)
1.036(.197) 1.042(.144)
1.025(.104) 1.063(.078)
Methanol
Formaldehyde
REA RT/LC
1.035(.094)
1.055(.285)
0.771(.213)
1.088(.127)
1.086{.180)
0.984(.253)
REA TN/CVS
1.076(.076)
1.035(.104)
1.025(.088)
0.975(.126)
1.065{.158)
0.915(.300)
1.035(.136)
1.222(.215)
1.019(.199)
REA TN/LC
1.021(.148)
0.818(.315)
0.764(.377)
0.987(.165)
1.033(.230)
0.997(.260)
REA CVS/CVS
1.023(.079)
0.984(.091)
1.035(.086)
0.961(.120)
1.133(.204)
1.035(.233)
1.033(.143)
1.063(.293)
1.057(.181)
REA CVS/LC
0.984(.136)
0.919(.377)
0.713(.212)
0.972(.297)
1.028{.205)
1.051(.189)
The above carbon dioxide data show excellent agreement of the
REA with the CVS analyzer for all three bags, both for the
integrated real time analysis as well as the tunnel (TN) and CVS
bag analysis. The carbon monoxide data and the oxides of
nitrogen data also show excellent agreement for all the bag
comparisons. The REA methanol data shows excellent agreement
with the GC data for all the Bag 1 data, but Bag 2 and 3 data
do not agree as well. This degree of disagreement can be
attributed to the lower concentrations of methanol (0 to 10 ppm)
found in Bags 2 and 3. Similarly, the formaldehyde data shows
excellent agreement for the Bag 1 data, but Bags 2 and 3 data
do not agree as well. Again, this can be attributed to lower
concentrations in Bags 2 and 3 (10 to 200 ppb) than in Bag 1
(200 to 1500 ppb).
The REA also has the potential for modal analysis. Fig 1
shows how well the REA concentration trace for C02 compares with
the DTN trace. As can be seen in the figure, concentration
increases and decreases compare well. However, the peak
concentrations in the REA trace are not as high and the low
concentrations are not as low. This is due to the 5 L volume of
the cell and the 25 L/min flow rate which combines for an
approximately 10 second sample residence time in the cell. An
281
-------
example of how modal analysis can be used is seen in Fig. 2. The
CO concentration can be seen to greatly decrease at
approximately the two minute mark, indicating the warming up of
the car engine and the "lighting off" of the catalyst. As can
be seen, the vast majority of formaldehyde and methanol in Bag
1 is produced in these first two minutes. Superimposing the
driving trace upon the concentration trace in Fig 2, the effect
of the rapid accelerations and decelerations have upon the
emission concentrations can be seen.
Conclusions
The REA results have been shown to compare well with those
obtained from conventional automotive emissions instrumentation
for the compounds thus far studied. Additional compounds need
to be studied and the potential for modal analysis fully
explored.
1. J. W. Butler, P. D. Maker, T. J. Korniski, L. P. Haack.
"On-line characterization of vehicle emissions by
FTIR and mass spectrometry", SAE Paper No. 810429 (1981).
2. J. W. Butler, P. D. Maker, T. J. Korniski, L. P. Haack, F.
E. McKelvey, and A. D. Colvin, "A system for on-line
measurement of multicomponent emissions and engine
operating parameters", SAE Paper No. 851657 (1985).
3. J. Staab, H. Klingenberg, H. Pluger, W. F. Herget, and M.
I. Tromp, "First experiences in testing a new
multicomponent exhaust gas sampling and analyzing
system", SAE Paper No. 851659 (1985).
4. B. Heller, H. Klingenberg, G. Lach, and J. Winckler.
"Performance of a new system for emission sampling and
measurement (SESAM)1', SAE Paper No. 900275 (1990).
5. S. B. Tejada, "Evaluation of silica gel cartridges coated
in situ with acidified 2,4-dinitrophenyl-hydrazine for
sampling aldehydes and ketones in air", Int. J. Envir.
Anal. Chem, 26, 167-185. (1986).
6. L. R. Smith and C. Urban, "Characterization of exhaust
emissions from methanol and gasoline fueled automobiles",
Final Report, EPA 460/3-82-004. US EPA, RTP, NC. 1982.
7. R. Snow, L. Baker, W. Crews, C. 0. Davis, J. Duncan, N.
Perry, P. Siudak, F. Stump, W. Ray, and J. Braddock.
"Characterization of emissions from a methanol fueled
motor vehicle", JAPCA 39:48 (1989).
282
-------
00
w
CsJ
o
o
LD
O
cc
111
Q_
1.5 -
1 -
0.5 -
3.0
TIME (MINUTES)
FTIR +
DTN
Figure 1. REA and DTN Real Time Trace
CO2 Bag 1 of FTP
-------
1.5 -
tN»
00
0
CO, PPM
23456
TIME, MINUTES
SPEED,MPH*15 o METHANOL, PPM A
8
FORM, PPM*50
Figure 2. FTP Bag 1. 1988 Corsica. Speed vs.
Carbon dioxide, Methanol, and Formaldehyde
-------
REMOTE SENSING MEASUREMENTS OF
CARBON MONOXIDE EMISSIONS
FROM ON-ROAD VEHICLES
Robert D. Stephens
Steven H. Cadle
Environmental Science Department
General Motors Research Laboratories
Warren, MI 48090-9055
Failure of many urban areas to meet clean air standards for CO has
increased pressure for stricter vehicle emission controls. To understand
the impact of this strategy and/or to propose alternative, more effective
control strategies, requires a better understanding of the emissions from
vehicles during normal, on-road operation. This report describes
instrumentation that is capable of remotely measuring the CO emissions from
thousands of vehicles per day with a sensitivity of ±\% CO, which, for new
vehicles, is approximately 10 grams per mile of CO. A prototype of this
instrument was used in Denver, CO in January of 1989 during a study
conducted in conjunction with researchers from the University of Denver,
who have developed similar instrumentation. Emission measurements were
made on approximately 4000 vehicles that were identified by make and model
year from state vehicle registration records.
285
-------
Introduction
Failure of many urban areas to meet clean air standards for CO has
increased pressure for stricter vehicle emission controls. To understand
the impact of this strategy and/or to propose alternative, more effective
control strategies, requires a better understanding of the emissions from
vehicles during normal, on-road operation. This report describes
instrumentation that is capable of remotely measuring the CO emissions from
thousands of vehicles per day with a sensitivity of ±1% CO, which, for new
vehicles, is approximately 10 grams per mile of CO. A prototype of this
instrument was used in Denver, CO in January of 1989 during a study
conducted in conjunction with researchers from the University of Denver,
who were contracted by GM to use similar instrumentation, which was
developed there. Emission measurements were made on approximately 4000
vehicles that were identified by make and model year from state vehicle
registration records.
Experimental
The systems operate by transmitting an infrared beam through the
exhaust plume of each passing vehicle (see Figure 1). Three detectors
monitor the infrared intensity in separate infrared spectral regions. The
detector signals were measured immediately before and after a vehicle
entered the infrared beam. Transmittance values for each detector channel
were obtained by determining the ratio of post-car to pre-car signals.
Concentrations of CO and COn were derived from transmittances using
polynomial equations which were best fits to calibration measurements of
transmittance as a function of known concentrations. The most notable
difference between the DU and GM instruments ..is the use of a spinning gas
filter correlation cell in the DU instrument . This cell enables the use
of one detector to alternately measure the CO and reference signals,
whereas the GM system uses separate detectors and measures all signals
simultaneously. The DU system uses 16,6 millisecond time resolution for
the CO and reference measurements and 8.3 milliseconds for the COg
measurement; the GM time resolution is 8.3 milliseconds for all channels.
Signal to noise ratios for the DU system is approximately 1000 to one
versus 250 to one for the GM system.
The instruments were used to measure emissions from vehicles exiting
1-25 onto southbound Speer Blvd., a major Denver traffic artery. This
uphill ramp assured that most vehicles were modestly accelerating during
measurement. The nearest on-ramp to this freeway was approximately two
miles away, which guaranteed that all vehicles would be at operating
temperature when our measurements took place. A video system recorded
license plate numbers of all vehicles measured. Registration information
provided age and make information for each vehicle.
The CO and COn within an exhaust plume from a vehicle will disperse
at an equal, but unEnown rate. Although CO and CO., concentrations will
vary rapidly with time, the C0/C02 ratio should be constant.
Concentrations were therefore determined relative to a fixed pathlength
(203 mm) and values of CO and C00 are reported as C0/C00 ratios.
a fi
Although the CO/CO,, ratio measurements are good indicators of vehicle
emission levels, the ratios can also be converted to percent CO levels, or
also to gram-per-mile emission rates. Gram-per-mile (gpm) emission rates
are a more conventional unit because it accounts for fuel economy.
286
-------
Conversion from measured CO/COn to gpm of CO requires knowledge of
fuel density, the fraction of fuel weight present as carbon, and the fuel
economy (miles per gallon) of the vehicle during the measurement. During
the time of this study, Denver was participating in the oxygenated fuel
program, whereby the fuel contained 2% oxygen by weight. Average fuel
density was measured as 2745.6 grams/gallon and the average fraction of
fuel weight present as carbon was 0.844. The most uncertain aspect of the
conversion to gpm emission rates is the fuel economy of the vehicle. To
minimize the uncertainties of this conversion, the measurements were
carefully sorted by vehicle type, age and make and conversions were then
based on fuel economy data for each subset of vehicles. The emission rates
reported in this study, then, should be viewed as average, relative rates
that could have substantial errors for any one vehicle.
The following equation was used to convert from CQ/CCU ratios to gpm
of CO emitted:
CCi(gpm) = 2747.fl (g/gallon) * 0.844 (g-C/g-fuel) * Q/(l + 1.175Q) * 28/12 (g-CO/g-C) + I (mpg)
In this equation, Q represents the measured CO/COg ratio. Note that
the emission rate is corrected for an assumed, but unmeasured, rate of
hydrocarbon emission (1.175 * QJ .
Results
Direct comparisons between the DU and GM measured CO/CCL ratios for
294 vehicles show good agreement. The correlation (r) between the two
measurements is 0.84 and a least squares fit of GM versu DU data yields a
slope of 0.88±0.04. Likely contributors of scatter in this plot are the
low signal to noise ratios of the GM instrument and the susceptibility of
the DU system to errors induced by the reference correction technique when
particulate matter is present during measurements.
Three potential problems that might affect the accuracy of the remote
measurements of CO/CCL ratios are: 1) Interference due to mixing of
exhaust plumes from different vehicles 2) Infrared emission from hot
exhaust 3) The technique by which transmittance measurements are
corrected for scattering by particulate matter.
The GM and DU measurements each have the potential for inaccurately
measuring an exhaust plume when remnants of a previous plume of different
concentration remains. In the DU instrument, a software routine converts
CO/COn ratios to percent CO and C02 and plots each measured concentration
of CO vs COn- Measurements are discarded if the standard deviation of the
linear fit to the slope exceeds 20?! . Such a test is meant to detect
changing CO/COg ratios that would occur due to mixing of two plumes of
different concentration ratios.
If a residual plume effect occurs, the effect would be expected to be
most pronounced when measurements are made with short time separations
between vehicles that emit very different concentrations of CO.
Data from this study show that vehicles which pass the
instrumentation within one second of a high emitting (>555 CO) car, are less
likely to be measured as low emitting (<1% CO) than cars which are 4 or
more seconds behind high emitting cars. Also, cars within one second of a
287
-------
high emitting car are more likely to have nonconstant CO/CO^ ratios than
are cars which are 4 or more seconds behind a high emitting car. The
magnitude of these shifts in the frequency distribution of measurements is
time dependent, i.e. the frequency (f) of measuring a clean car immediately
after a dirty car follows the following relationship: f(l sec) < f(2 sec) <
f(3 sec) < f(4 sec) « f(10 sec). Also the frequency of measuring
nonconstant CO/CO^ ratios follows the relationship: f(l sec) > f(2 sec) >
f(3 sec) > f(4 sec) ~ f(10 sec). Nonconstant C0/C02 ratios are the
consequence of the mixing of two plumes of differing concentrations.
This data suggests that a residual plume effect does occur and that
the effect seems to last, on average, at least one second and possibly as
long as 4 seconds. The DU nonlinearity test identifies plume nonlinearity,
at least for many instances where a 0-1% CO vehicle follows closely behind
a >5% CO vehicle. However, rejection of these measurements results in a
sampling bias; a clean car (low CO) might not be measured if it follows
closely behind a dirty car. The effect of this sampling bias is that the
median CO measured in an exhaust plume is higher when a residual plume of
>5% CO is present. It is impossible to determine from this data whether
this increase in median CO is due only to sampling bias or is partially due
to measurement error induced by the presence of a high level of CO.
The GU system was occasionally operated with no infrared source to
determine if infrared emission from hot exhaust was detectable. Emission
signals were detected in 17 of 220 measurements taken in this manner. The
effect of infrared emission from hot gases or particulate would be to
induce inaccuracies in the measurement of CO and CCL by both systems. Of
the 17 vehicles that generated detectable infrared emission, 16 were either
vans, pickup trucks. The factor most common to these vehicles seems to be
the orientation of the tail pipe. Vans and pickup trucks frequently have
tail pipes that direct hot exhaust perpendicular to the direction of
travel, i.e., toward the instruments used for this study. This
interference can be minimized by modifications to the instrumentation.
Alternately measuring CO and reference signals, as is done in the DU
system, can potentially be a problem. In effect, each CO data point and
every other COg data point is being corrected by a reference data point
being measured 8.3 milliseconds later. The turbulence behind most vehicles
causes significant plume concentration fluctuations on this time scale.
Simultaneous measurements of CO, C02 and reference signals will provide
more accurate measures of CO and CO,, concentrations in cases where
particulate is present.
During this study, CO emissions from 3243 passenger cars and 887
light-duty trucks were measured. Figure 2 shows the fraction of the 3243
car measurements that are associated with each model year and also the
fraction of all passenger car CO emitted by vehicles of each model year.
It is clear that older vehicles emit a large fraction of the total CO. For
example, 6.9% of the total number of passenger cars measured in this study
were pre-1975 models (with a mean model year of 1970), yet they contributed
28.058 of all CO emitted by cars. Alternatively, 1987 through 1989 models
contributed only 4.1% of CO from cars, but represented 23% of all cars
measured.
An examination of the highest emitting passenger cars indicates that
50% of CO from cars is emitted by 8.1% of the cars. These cars, on
average, were 12.3 years old and emit 90.4 gpm of CO. If all passenger
288
-------
cars emitted at the rate measured for 1989 cars, emission rates would be
reduced by 86.5%.
Discussion
The remote sensing technique described here can provide accurate
measurements of CO/COo, with only minor interferences that can be
eliminated with instrument modifications. These ratios are, in themselves,
useful indications of relative CO emission rates. The ratios can be
converted to values of %CQ with reasonable accuracy (±1% CO). Conversions
of the ratios to emission rates (gpm) requires knowledge of hydrocarbon
emission rates and instantaneous fuel economy. Using careful estimates of
these values probably provides reasonably accurate fleet average emission
rates when large numbers of vehicle measurements are available.
This study was also affected by factors that cannot be assessed; for
example, the effect of site selection, the Denver oxygenated-fuel program,
altitude, and the effect of ambient temperature. These measurements are
also not representative of other cities due to differences in vehicle
fleets.
Conclusions
The data acquired during this study has provided important insight
into fleet average CO emission rates at a site located in Denver, CO. To
the extent that this site is typical of other urban areas within the United
States, the results of this study can be useful in identifying cost-
effective strategies for reducing urban CO pollution.
We have found that the majority of CO is emitted by a small minority
of all vehicles. By examining the highest emitting vehicles, we find that
50S of the CO is emitted by 8.9?£ of all vehicles. By examining the oldest
vehicles, we find that 57JS of the CO is emitted by pre-1980 vehicles. In
contrast, 1988 and 1989 vehicles contributed a combined total of only 2% of
all CO emitted.
For CO control strategies, this data suggests that ambient CO levels
can be impacted most by reducing emissions from the highest emitting
vehicles rather than further reductions in emissions from new vehicles.
References
1. G. A. Bishop, J. R. Starkey, A. Ihlenfeldt, W. J. Williams, and D. H.
Stedman, Anal. Chem., 61, 671A,(1989).
2. G. A. Bishop, University of Denver, Denver, CO, Private Communication,
(1990).
289
-------
MIRROR
BEAM
^SPLITTER
BEAM
^. MIRROR
J^
KMM3-* OPTICAL FILTERS
VIDEO
CAMERA
Figure 1. Schematic diagram of the GMR instrument for measuring CO/CO,
ratios.
0.30 r
0.25
0.20
o
ts
0.10
0.05
0.00
Fraction of Cars
Fraction of CO from Cars
n = 3243
ffffff/ft///////
Model Year
Figure 2. Distributions of CO emission rates and vehicle fraction by
model year.
290
-------
The Effect of Oxygenated Fuels
on Automobile Emission Reactivity
Charles D. Burton
Mobile Sources
NSI Technology Services, Inc.
Research Triangle Park, North Carolina
Hydrocarbon emissions of vehicles are currently regulated on the basis
of the total mass emitted without regard to the individual hydrocarbon's
reactivity toward ozone formation. In this study individual hydrocarbons
from tailpipe and evaporative emissions are weighted, relative to ethane, and
summed according to their potential for ozone formation to determine the
effect of temperature and fuel composition of the relative reactivity of the
vehicle emissions. Two base fuels and two oxygenated fuels were studied at
three temperatures (40, 75, and 90°F) following vehicle certification
procedures as described in the Federal Register. The oxygenated fuels
contained either ethanol or methyl tert-butyl ether(MTBE), blended to an
oxygen content of 3% by weight. Both the evaporative emissions and the
reactivity of these emissions were significantly greater for the oxygenated
fuels than for the base fuels. Ethanol increased the total tailpipe
emissions and the reactivity of those emissions compared to the summer base
fuel. Lower reactivity of the tailpipe emissions when the MTBE blend was
used resulted from lower total mass emissions including lower aromatic
content. A non-oxygenated fuel was prepared from a low aromatic blend stock,
used to produce the MTBE blend, by adding toluene to give the same octane
rating as the MTBE blend for a comparison base fuel. The use of toluene, a
low relative reactivity aromatic instead of the higher relative reactivity
aromatics, produced a fuel that when used had a lower relative reactivity
emissions than the summer grade fuel. The olefinic fractions were the major
contributor to the reactivity of the emissions for all test conditions,
regardless of the fuel used accounting for as much as 75% of the tailpipe and
evaporative emissions reactivity.
291
-------
Introduction
A primary reason for controlling automobile emissions is their role as
precursors to ozone formation. Historically, the emissions have been
regulated on the total hydrocarbon mass emitted. This approach fails to take
into account the atmospheric reactivity differences of the individual
hydrocarbons. Some compounds, such as 2-methyl-2-butene, have low emission
rates, yet are highly reactive for ozone formation. Other compounds such as
n-butane have high emission rates, but are relatively non-reactive. If in
compliance with requirements for reduced total hydrocarbon mass emissions,
compounds of relatively high reactivity remain, the ozone benefit will not be
commensurate with the mass emission reductions.
Since the introduction of emission controls, tailpipe emissions of
carbon monoxide and hydrocarbons have been reduced by over 90% (1).
Significant tailpipe hydrocarbon (HC) and carbon monoxide (CO) reductions
occurred with the introduction of oxidation catalytic converters and oxides
of nitrogen (NOx) reductions with the introduction of exhaust gas
recirculating valves (EGR) and later 3-way catalytic converters. Today's 3-
way catalyst tailpipe control systems also include exhaust oxygen sensors for
computer control of the Air/Fuel Ratio. By maintaining the Air/Fuel Ratio at
stoichiometric levels, emissions HC, CO, and NOx can be controlled for
maximum reduction.
Reductions in total evaporative emissions can be realized by the
reduction of the fuel vapor pressure (RVP) of the fuel which is largely
determined by the amount of C4 and C5 paraffins present. Because these
paraffins have relatively low reactivities, they may not contribute
significantly to the overall reactivity of the emissions. Therefore, simply
reducing the RVP of the fuel by reducing the C4 and C5 paraffins may not
reduce the rate of ozone formation, commensurate with reductions of
hydrocarbon mass emissions.
The addition of oxygenates such as methyl tert-butyl ether (MTBE) to
gasoline is being used to reduce carbon monoxide emissions and to elevate
fuel octane. Previous studies have examined the effect of oxygenated blends
on hydrocarbon mass emission rates (2). The purpose of this paper is to
examine the impact of oxygenated fuel additions on the ozone potential of
motor vehicle emissions.
The reactivity scheme used in this study is based on the reaction rate
of the organics with OH radicals. Reaction rate constants and relative
reactivity data for each compound were obtained from the California Air
Resources Board (3). Each compound was assigned a reactivity value relative
to ethane which is assigned a relative reactivity of 1.0. A selected list of
compounds and their reactivity values on both a volume and a weight basis are
given in table 1.
Experimental
A 1988 Chevrolet Corsica was used for all tests. The car was equipped
with current technology emission control equipment which includes a catalyst,
an EGR valve, an Exhaust Oxygen Sensor (EOS) and computer control of the
Air/Fuel ratio (A/F). The vehicle was powered by a 2.0L 4 cylinder engine
with a throttle-body fuel injection system. No modifications were made to
the engine, drive-train, or computer parameters. However, the fuel tank was
modified by the insertion of a thermocouple plug to measure the fuel
temperature during evaporative testing and a drain line to simplify the
removal of waste fuel. The car arrived with 17 thousand miles and was driven
781.2 miles during the testing. All testing was done in accordance with the
292
-------
rules and regulations as described in the Federal Register (4). A more
comprehensive description of the test procedures and analytical techniques
was reported by Stump, et al (5).
All evaporative and exhaust samples were analyzed for speciated
hydrocarbons, oxygenates, and aldehydes by the methods reported by Stump, et
al (6). The data was first reduced to the standard reporting format of
grams/mile for exhaust Diurnal+3 . 01 ( trips/day) xHot Soak
data, of grams/test for ***& 3i.imiJes/day
diurnal and hot soak
samples, and a Evap- Evaporative Equivalent (grams /mile)
o a p oae Diurnal-Diurnal Emissions (grams/day)
emissions. The Hot Soak- Ho t Soak Emi ssi ons ( grams/ trip)
evaporative equivalent
value was calculated
from the diurnal and hot soak results for each compound found in the
evaporative samples. Each reported compound in the evaporative and tailpipe
samples was multiplied by the associated relative reactivity index
{ethane=l, toluene=7.3) and summed to produce an emission reactivity index
(ERI). Tailpipe and evaporative ERI results are in the units of relative
reactivity * Emission
grams/mile. Class Reactivity Y^ HC a/mile x Relat.iv.e
.... - . , .i\e«Lr uj. vj. uy • / ., .nv-/ y/ IIU..L e? A ijc,s/-i--i T^V t-tr
reactivity fraction index (ERI) allHC's Reactivity
and class mass
fractions were also
calculated for all evaporative and tailpipe emissions. The data were then
sorted by class and Carbon number from which the tables were prepared.
Calculations for ^ /m-7~ Relative
the fuels were Class f^ u^,g/mij.eclm xReactivity
based on volume Reactivity - c±ass - xlOO
percentages and Fraction —, H^^/rn-10 Relative
consequently the 2^ . HC'^/inj-Le ^Reactivity
relative all HC's
reactivities ^
used were on a Z^ HC, g/mileclaBS
volume basis. Mass. _ class
Fraction
Four fuels HC,g,mile
were chosen for all He's
testing based on
present use and
predictions of what fuels may be available if the use of oxygenated fuels
becomes mandatory. The fuels were: a locally available unleaded summer grade
base fuel, a J"f £ . -^ Relative Reactivity
"splash blend" of ^ex^FRI)' 11^ (by Volume)
8.1% (v/v) ethanol Index (FRI} a11 Hc s
fuel, a high
aromatic base fuel and a low aromatic reformulations with 16.5% (v/v) MTBE
blend. Both of the oxygenated fuels contain 3% (wt) oxygen. Table 2
summarizes the characteristics of each fuel along with a listing of current
national averages.
The regular unleaded summer grade gasoline was obtained from a local
gasoline distributor and represents a regular unleaded gasoline distributed
in North Carolina during the spring and early summer months. The fuel had an
octane rating, (R+M)/2, of 87.6 and a RVP of 10.2 psi. The aromatic, olefin,
and paraffin fractions of 46.7%, 11.5%, and 37.7% by volume respectively.
A portion of the summer grade fuel was splash blended with ethanol at
293
-------
the 8.1% v/v to yield 3% oxygen by weight. In order to simulate possible
actions of a local distributor, no attempt was made to reduce the RVP. As a
result, a slight increase in the RVP (2.94%) and slight increase in the
octane Rating (1.71%) was obtained. The resultant blend is believed to
represent an ethanol blend that local distributors might produce.
Sun Oil Company was commissioned to prepare the MTBE blend used for
this study. For this purpose, a low aromatic blend stock (LABS) was prepared
and used for blending. MTBE was added at the 16.2% v/v level for a 3% by
weight oxygen blend. The resultant fuel had a slightly higher RVP (0.95%)
and octane (0.56%) than the unleaded summer grade base fuel. The 16.2% MTBE
fuel contained 20.9% aromatic, 52.6% paraffin, and 10.3% olefins. An
aromatic blend stock, consisting primarily of toluene, was added to the LABS
to match the octane rating of the MTBE blend. This high aromatic fuel was
tested as a base fuel for the MTBE blend.
Results and discussions
Relative reactivities based on volume, as opposed to the weight basis
used for emissions data, were used to calculate the reactivity contribution
for each compound in the fuel. Class and carbon number reactivities were
summed and normalized for comparisons to the volume percentages (Table 3).
A fuel reactivity index (FRI) for each fuel was calculated by weighting the
volume fraction of each compound by the compound's volume reactivity and
summing the results. The summer grade fuel had the highest FRI, 70.69. The
8.1% ethanol fuel prepared from the summer grade fuel had a FRI of 69.07,
only slightly lower than the base fuel. The decrease in reactivity is
attributed to the relatively low reactivity of ethanol.
The high aromatic fuel and the MTBE fuel were both prepared from a low
aromatic blend stock which contained low concentrations of the reactive C9
aromatics. For both the high aromatic fuel and the MTBE fuel, the LABS was
diluted with relatively low reactivity compounds, 16.2% MTBE and 26% toluene.
The olefinic reactivity of the MTBE and the base fuel is similar to that of
the summer base fuel, however the lower C9 aromatic content of the prepared
fuels resulted in an overall FRI of 57.96 for the MTBE fuel and 48.86 for the
high aromatic base fuel compared to the 70.69 for the summer grade fuel. The
high Aromatic fuel was prepared by the addition of toluene to the LABS such
that the octane rating matched the 89.6 octane rating of the MTBE fuel.
Toluene accounted for 30.69% of the total volume compared to only 6.67% in
the MTBE fuel. However, toluene has a low relative reactivity and did not
contribute significantly to the overall FRI.
The olefin fractions of the fuels are only 8.5- 11.5% of the total fuel
by volumes, but represent 59.4-66.5% of the FRI. The major portion of the
olefin reactivity comes from the C5 olefins which have a very high relative
reactivity value. The C5-C8 olefins as a whole are 4 times more reactive
than any of the aromatics and almost 10 times more reactive than the xylenes
which contribute roughly 8% of the total volume. This amount is almost the
same volume percent as the total olefin fraction. The total aromatics, while
contributing 20.9-37.7% of the volume, contribute 16.7-28.1% of the FRI. The
paraffins which are the major portion of the fuel on a volume basis, but
represent only 10.6-14.7% of the total FRI. The oxygenates have relative
reactivity's similar to the paraffins and did not contribute significantly to
the overall fuel reactivity.
As expected evaporative mass emission rates significantly increased at
test temperatures of 90°F (Table 4). At 90°F the high aromatic fuel had lower
mass emission rates than did the other fuels. At 40°F and 75°F the
evaporative mass emission rates for all fuels were within 0.013 grams/mile
294
-------
and indicated no significant trends between oxygenated and non-oxygenated
fuels. At 90°F however, the summer base fuel had a three fold increase in
evaporative mass emissions compared to the 75°F test, while the ethanol fuel
had an evaporative mass emissions increase of almost tenfold. A similar
tenfold increase occurred with the MTBE blend, but not with the non-
oxygenated high aromatic base fuels. In all cases the increase in
evaporative emissions resulted in higher ERI values for the evaporative
emissions, due to the increased mass emission rates.
Since the evaporative emissions are predominately paraffins, the ERI
increases are not directly proportional to the mass emission rate increases.
Paraffins typically account for as much as 73% of the evaporative emission,
with n-Butane contributing 23% to the total mass. At 90°F, n-butane accounted
for 44.5% of the total evaporative mass, but on a reactivity scale, n-butane
accounted for less than 20% of the evaporative ERI. The major contributor to
the evaporative ERI are the olefins, which account for an average of 68.92%
of the ERI at 90°F (Tables 5 and 6).
The evaporative mass and consequently ERI values of the 90°F tests were
significantly higher for the oxygenated fuels than for either of the non-
oxygenated fuels. Even though the oxygenates represented significant
fractions of the 8.1% ethanol and 16.2% MTBE fuels they represented less than
1.2% of the total mass in the evaporative samples (Table 7). The absence of
the oxygenates in the evaporative samples and the elevated mass emission
rates indicate that the evaporative canister may be selectively retaining the
oxygenates and displacing the hydrocarbons.
At all three temperatures the tailpipe emissions of the 8.1% ethanol
fuel were found to have greater ERI rates than the summer grade exhaust due
to the greater mass emission rates for the ethanol fuel (Table 8). At 75°F,
the ethanol fuel had a higher total mass emission rate yet, because of the
reduced aromatic emission rates, the resultant exhaust was less reactive. At
40°F and 90°F the tailpipe mass emissions and ERI values were greater for the
ethanol fuel than for the summer grade fuel.
At all three temperatures the high aromatic fuel and the 16.2% MTBE
fuel had lower total mass emission rates. The lower mass emission rates
coupled with the inherently lower relative reactivity of the fuel components
resulted in lower tailpipe ERI values. In all tests the aromatic ERI rates
were reduced when these two fuels were used. This is a direct result of the
lower C9 aromatic content of these fuels. Total aromatic concentrations in
the emissions for the high aromatic fuel and the summer grade fuel were
similar. The high aromatic emissions contained large percentages of toluene
whereas the summer grade emissions contained large C9 aromatic fractions,
reflecting the aromatic composition of the fuels. Since toluene has a low
relative, reactivity, it did no raise significantly the ERI rates of the
tailpipe emissions for the high aromatic fuel.
The high aromatic fuel demonstrated most drastically the effect of the
fuel composition on tailpipe ERI (Tables 11, 12, and 13). Mass emission
rates for the high aromatic fuel were 9.42, 18.43, and 33.81% lower than
those for the summer grade fuel for the respective test temperatures, 40,75,
and 90°F. Yet those emissions were 30.75, 23.21, and 42.67% less reactive
than those of the summer grade fuel. The aromatic mass fraction of the high
aromatic exhaust was of similar proportions to the summer grade fuel, but
because the aromatic fraction is predominately toluene and not the C9
aromatics found in gasoline, the exhaust from the high aromatic fuel produced
lower ERI values.
Similarly, the MTBE blend contained low aromatic concentrations, which
295
-------
was reflected in the exhaust. When the MTBE fuels (20.92% aromatic) was
used, the aromatic content of the total tailpipe mass emission ranged from
13.12% to 25.37%. The low total mass emission rates coupled with the reduced
FRI resulted in lower tailpipe ERI rates for the MTBE fuel.
The MTBE fuel also contained low C9 aromatic concentrations, but did
not contain large concentrations of toluene. Thus the aromatic mass emitted
by the MTBE fuel was less than the other fuels. However the ERI decrease
caused by the low aromatic emission rates were offset slightly by increased
isobutylene formed during combustion of MTBE (Table 14). Isobutylene ERI
values increased an average of 420% when the 16.2% MTBE fuel was used. Yet
even with the isobutylene increases the 16.2% MTBE fuel resulted in the
lowest olefinic ERI at test temperatures of 40°F and 75°F. The low olefinic
fraction of the fuel and the lower mass emission rates combined to produce
the lower olefinic ERI rates.
The average relative reactivity index (ARRI) was calculated for each
sample (Table 9). Generally, the data showed that the high aromatic and
16.2% MTBE fuels ^ vf< nim*i* Relative
produced lower ARRI Average L . HC'fffmiJ-e ^Reactivity
values than the Rellve - a11 HC s _
summer grade and T, ^
ethanol fuels for Index (ARRI) E HC.g/mile
reasons discussed all He's
previously. However
the data also showed a marked decrease in the ARRI of the ethanol tailpipe
emissions at 75°F. Total mass emissions and class distribution values were
within expected ranges. However on a relative reactivity scale, the 75°F test
with the ethanol fuel failed to give an increase in ERI commensurate with the
mass emission rate increase as compared to the 75°F summer grade test. As a
result the ARRI was drastically lower then was expected. For all tests the
ARRI values failed to show any distinct trends between temperature or fuel.
A closer examination of the class and carbon number distributions of
the tailpipe emissions from the 75°F test with the ethanol fuel revealed that
the C8 olefinic contribution to the ERI was less than with the other fuels
and temperatures tested (Table 10). For all of the tests, the C8 olefin
emissions account for less than 1% of the total mass. However when that mass
is multiplied by the large relative reactivity value of the C8 olefins
(194.9) their importance greatly increased. On the relative reactivity basis
the C8 olefins accounted for as much as 6.2% of the total ERI. The
inconsistency found in this data is possibly attributable to missing or
incomplete data used in interpreting the chromatographic results.
Chromatographically the C8 olefins appear to be minor peaks when compared to
the relatively large surrounding peaks such as iso-octane or 2,5 dimethyl
hexane which all elute within a five minute window. Within this five minute
window there are ten compounds tentatively identified as C8 olefins half of
which are normally are normally observed in the samples. One might be
tempted to dismiss the "minor" peaks as having very little importance to the
overall analysis however, this study underscores the value of accurately
identifying and quantifying even the minor compounds found in the samples.
Conclusions
Reducing the Reid Vapor Pressure by lowering the concentration of n-Butane
may reduce the overall evaporative mass emissions, but may not lower the
ozone potential of the hydrocarbon emissions commensurately.
Oxygenates increase the evaporative emission rates, possibly attributable to
the evaporative canister selectively retaining the oxygenates.
296
-------
The ERI rates increased with the use of ethanol due to the increased mass
emission rates that occur with the use of ethanol.
Total tailpipe emission rates and ERI values decrease with the use of MTBE,
in spite of the increased isobutylene emissions. The lower overall mass
emission rates of the MTBE fuel combined with the lower aromatic and olefinic
concentrations, of the fuel reduced the tailpipe reactivity index.
References
1) MVMA Motor Vehicle Facts and Figures '89. Motor Vehicles Manufacturers
Association of the United States. 1989, page 88.
2) Lang J. and Black F., "Impact of Gasohol on Automobile Evaporative and
Tailpipe Emissions", SAE Paper No 810438. (1981)
3} Unpublished data obtained from the California Air Resource Board
4) "Code of Federal Regulations", Title 40, Part 86, U.S. Government
Printing Office, Washington, D.C., July 1983.
5) F. Stump, D. Dropkin, "Gas Chromatographic Method for Quantitative
Determination of C2 to C13 Hydrocarbons in Roadway Vehicle Emissions",
Analytical Chemistry 57: 2629 (1985)
6) F.Stump, Knapp, K., Ray W., "Seasonal Impact of Blending Oxygenated
Organic Gasoline on Motor Vehicle Tailpipe and Evaporative Emissions",
Journal of Air and Waste Management Association, Accepted for
Publication June 1991
7) J.Duncan, C. Burton, W. Crews, "A Method for Measurement of Methanol,
Ethanol, and tert-Butyl Ether emissions from Motor Vehicles", Presented
Pittsburgh Conference, (1988)
8) Black F., High L., "Chemilumenscence Measurement of Reactivity Weighted
Ethylene-Equivalent Hydrocarbons", Environmental Sciences and
Technology. 11, page 1977
9) Black F, High L., Lang J, "Composition of Automobile Evaporative and
Tailpipe Hydrocarbon Emissions", Journal of the Air Pollution Control
Association. 30, page 1216, (1989)
10) Dunker A, "The Relative Reactivity of Emissions from MEOH-fueled and
Gasoline-Fueled vehicles in forming Ozone", Presented at Air and Waste
Management Association. (1981)
Table 1. Selected Relative Reactivity Values
Compound Reactivity (related to Ethane)
by Volume by Weight
C8 Otefins 727.3 194.9
C5 Olefins 545.5 233.9
Isoprene 363.6 160.5
1,3-Butadiene 242.9 135.0
trans-2-Butene 231.6 124.2
Terpenes 206.5 45.6
eis-2-Butene 204.0 109.3
C9 Aromatics 109.1 24.4
Ethanol 10.3 6.9
Toluene 22.5 7.3
n-Butane 9.2 4.8
Benzene 4.7 1.8
Ethane 1.0 1.0
Methane 0.0 0.1
297
-------
Table 2. Test Fuel Specifications
National
Fuel Property Summer
Regular Grade (1)
Unleaded Average
Specific Gravity
RVP. psi (2)
Octane ((R+t1)/2)
Distillation, Deg. F
IBP (3)
10 X
50 X
90 X
End Point
Volume X
Paraffins
Olef ins
Aromatics
Benzene
Toluene
NTBE
Ethanol
0.7406
10.0
87.1
92.0
119.0
205.0
341.0
420.0
58.1
12.4
29.5
1.6
(4)
<0.5
<0.9
Commercial
Suimer
Grade
Range
(0.718-0.7594)
(7.7-12.2)
(84.4-90.7)
(85-106)
(107-138)
(143-231)
(306-367)
(378-482)
(35.0-72.9)
(0.8-37.0)
(15.8-41.5)
(0.6-4.7)
(4)
(0.0-5.7)
(0.0-9.0)
0.7428
10.2
87.6
89.6
118.4
212.0
356.0
413.6
48.62
10.33
37.93
1.39
6.19
ND (3)
NO (3)
Summer
Grade/
8.1 X
Ethanol
0.7465
10.5
89.1
84.0
114.0
196.0
360.0
420.0
44.68
9.49
34.86
1.28
4.28
ND (3)
8.10
High
Aromatic
0.7383
10.1
89.4
96.8
131.0
199.4
269.9
399.2
45.58
8.42
45.60
0.45
6.66
ND (3)
NO (3)
16.2 X
HTBE
0.7146
10.6
89.6
96.8
123.8
177.8
284.0
395.6
52.61
9.90
20.92
0.53
30.69
16.20
ND (3)
(1) Source: MVMA (1987) (2) RVP-Reid Vapor Pressure (3) ND-Not Detected (4) Data unavailable
(5) Oxygenates and Unknowns represent the remainder of the percentages in all tables
Table 3.
Suimer Grade
8.IX Ethanol
High Aromatic
16.2X MTBE
Fuel CoMpositions. by Class
Paraffins
46.7 (10.6)
57.9 (13.0)
45.4 (13.2)
52.6 (14.7)
Volm
Olef ins
11.5 (59.4)
11.5 (65.3)
8.5 (59.6)
9.9 (66.5)
K X. (FRI X)
Aromatics
37.7 (28.1)
26.6 (20.4)
45.8 (26.9)
20.9 (16.7)
Oxygenates
and Unknowns
4.1 (1.9)
9.0 (1.3)
0.4 (0.2)
16.6 (2.1)
Fuel Reactivity
Index
70.69
69.07
48.86
57.96
Table 4.
Fuel
Summer Base
8.1 X Ethanol
High Aromatic
16.2 X HTBE
Evaporative Enission Rates
graas/aile (Emission Reactivity Index)
40°C
0.0159 (0.459)
0.0261 (0.703)
0.0190 (0.415)
0.0178 (0.512)
75°C
0.0542 (1.169)
0.0516 (1.142)
0.0444 (1.112)
0.0417 (0.853)
90°C
0.1562 (2.740)
0.4918 (8.959)
0.2450 (3.677)
0.3581 (6.636)
Table 5.
40°F
paraffins olefins
Simmer Base 11.29 61.33
8.1 X Ethanol 13.84 57.59
High Aromatic 14.39 64.97
16.2X HTBE 11.23 40.01
Class percentages of Evaporative Emissions
by Reactivity
75UF
aromatics: paraffins olefins aromatics:
23.46 : 11.55 48.94 37.35 :
24.81 : 16.53 48.47 33.08 :
18.98 : 12.74 62.43 24.08 :
47.18 : 18.02 65.81 15.75 :
90°F
paraffins olefins
23.05 62.74
23.13 73.70
26.53 63.83
21.90 75.39
aromatics
13.36
2.92
9.28
2.53
Table 6.
Organic class Reactivity Distribution of Evaporative Emissions
Emission Reactivity Index
40°F
paraffins olefins
Suimer Base 0.5183 0.2815
8.IX Ethanol 0.0974 0.4051
High Aromatic 0.0598 0.2698
16.2X HTBE 0.0574 0.2047
75°F
aromatics: paraffins olefins aromatics:
0.1077 : 0.1350 0.5721 0.4367 :
0.1745 : 0.1888 0.5536 0.3778 ;
0.0788 : 0.1416 0.6941 0.2678 :
0.2426 : 0.1536 0.5611 0.1343 :
90°F
paraffins olefins
0.6317 1.7193
2.0725 6.6027
0.9755 2.3466
1.1020 5.0971
aromatics
0.6332
0.2615
0.3415
0.6032
Table 7. Oxygenate Evaporative Concentrations
8.IX Ethanol Fraction
DIURNAUX TOTAL) HOT SOAK (X TOTAL) EVAP EQUIV (X TOTAL)
40 0.00696 (1.461)
75 0.00783 (0.510)
90 0.01044 (0.868)
0.00609 (1.011)
0.00624 (0.868)
0.01363 (0.720)
0.000817 (1.009)
0.000855 (0.714)
0.001653 (0.112)
16.2X MTBE Fraction
DIURNALCX TOTAL) HOT SOAK (X TOTAL) EVAP EQUIV (X TOTAL)
40 0.006573 (2.247)
75 0.007750 (0.567)
90 0.183667 (0.750)
0.00445 (1.035)
0.00290 (0.371)
0.01092 (0.476)
0.000648 (1.260)
0.000508 (0.421)
0.006712 (0.067)
298
-------
Table 8.
Fuel
Sunnier Base
8.1X Ethanol
High Aromatic
16.2X MTBE
Tailpipe Emission Rates
Reactivity Index)
40°C
0.5552 (14.387)
0.7880 (18.566)
0.5029 (10.035)
0.3386 ( 8.818)
75°C
0.3597 (10.677)
0.5674 ( 8.668)
0.2934 ( 8.199)
0.2554 ( 6.421)
90°C
0.3195 ( 7.474)
0.5316 (13.667)
0.2114 ( 4.285)
0.3041 ( 6.862)
Table 9.
Summer Base
8. IX Ethanol
High Aromatic
16. 2X KTBE
40°F
28.87
26.93
21.84
26.04
Evaporative
75°F 90°F
21.57
22.13
25.02
25.14
17.54
18.22
15.01
22.56
Average Relative Reactivity Index (AMIl
40°F
25.91
23.56
19.95
28.76
Exhaust
75°F 90°F
29.68 23.39
15.28 25.71
27.94 20.27
20.46 18.53
Table 10.
C8 Otefin Reactivity Emission Rates
X of ER1 fX of Total Haw)
Sunnier Base
8.1% Ethanol
High Aromatic
16. 2X HTBE
40°F
5.057 (0.679)
4.042 (0.487)
3.172 (0.324)
4.973 (0.649)
75"
4.503 (0.708)
1.004 (0.108)
2.853 (0.266)
5.463 (0.693)
90°F
0.783 (0.085)
4.706 (0.622)
6.217 (0.635)
1.846 (0.195)
Table 11.
Class percentages of Tailpipe Emissions
X of Emission Reactivity Index
40°F
paraffins olefins aromatics:
Sunnier Base 10.66 58.89 29.72 :
8.1 X Ethanol 10.34 54.96 32.01 :
High Aromatic 11.12 68.16 20.19 :
16.2X HTBE 11.63 59.49 23.48 :
75°F
paraffins olefins
9.936 49.65
8.907 56.51
14.67 56.04
9.576 60.34
aromatics:
38.87 :
31.91 :
27.43 :
28.84 :
90°F
paraffins olefins
12.2 59.5
14.56 63.5
11.38 59.51
16.06 74.28
aromatics
26.45
21.05
28.40
8.79
Table 12.
Claw Reactivity Distribution of Tailpipe Emission*
40°F
paraffins olefins
Summer Base 1.5337 8.4725
8.U Ethanol 1.9197 10.2039
High Aromatic 1.1159 6.8399
16.2X MTSE 1.0255 4.8663
ctivitv Index
aromatics: paraffins olefins
4.2758 : 1.0609 5.3011
5.9430 : 0.7721 4.8983
2.0261 : 1.1861 4.5947
2.0250 : 0.6149 3.8742
90°F
aromatfcs: paraffins olefins aromatics
4.1501 : 0.9118 4.4470 1.9769
2.7660 : 1.9899 8.6785 2.8769
2.2490 : 0.4876 2.5500 1.1269
1.8518 : 1.1020 5.0971 0.6032
Table 13.
40°F
paraffins olefins aromatics;
Summer Base 46.76 19.98 33.25 :
8.IX Ethanol 46.65 17.64 35.72 :
High Aromatic 45.78 19.92 34.75 :
16.2X MTBE 52.89 22.69 24.42 :
by Organic class
75UF
paraffins olefins aromatics:
47.70 17.99 34.31 :
45.87 19.81 34.32 :
42.27 19.42 38.3 :
53.17 21.45 25.37 :
90°F
paraffins olefins
55.44 19.23
59.66 17.14
46.04 18.73
64.44 22.48
aromatics
25.53
23.21
35.73
13.12
Table 14.
Isobutvlene Reactivity Emission Bate
Emission Reactivity index iX ERI)
40°F
Summer Base 0.348 ( 2.42)
8.IX Ethanol 0.380 ( 2.05)
High Aromatic 0.491 ( 4.89)
16.2X HTBE t.266 (14.35)
75°F
0.206 ( 1.93)
2.820 ( 3.32)
0.231 ( 2.82)
0.875 (13.62)
90°F
0.231 ( 3.09)
0.427 ( 3.12)
0.277 ( 6.47)
1.085 (15.82)
299
-------
FTIR: FUNDAMENTALS AND APPLICATIONS IN THE ANALYSIS OF DILUTE VEHICLE EXHAUST
C. A. Gierczak, J. M. Andino, T. J. Korniski and J. W. Butler
Chemistry Department
Scientific Research Laboratories
Ford Motor Company
Dearborn, Michigan
Fourier transform infrared (FTIR) spectroscopy has been shown to be a
valuable tool in the analysis of complex gaseous mixtures, such as dilute
vehicle exhaust. Regulated and non-regulated vehicle emissions have been
routinely sampled and analyzed using prototype instrumentation developed in
this laboratory, and in several other laboratories over the last decade.
More recently, commercial versions of these FTIR analyzers have become
available through several manufacturers. This paper reviews data acquisition
and processing techniques employed by several of the prototype and commercial
FTIR systems. Techniques utilized by the FTIR emissions analyzer developed
in this laboratory, and several of its unique capabilities, are also
addressed in detail. In addition, experiments designed to investigate the
effects of instrumental and environmental parameters that alter spectral line
widths, and consequently affect the accuracy of this high resolution system,
are described. The results of these studies indicate that routine variations
in parameters, such as instrument resolution, sample temperature and pressure
have minimal effects on the accuracy of the analysis of low concentrations
of nitric oxide.
300
-------
Introduction
Fourier transform infrared (FTIR) spectroscopy has been shown to be a
valuable tool in the analysis of complex gaseous mixtures, such as dilute
vehicle exhaust. Regulated and non-regulated vehicle emissions have been
routinely sampled and analyzed using prototype instrumentation developed in
this laboratory, ''' and in several other laboratories over the last de-
cade.5'6'7'8'9 There are several reasons which explain the increase in the
popularity of FTIR spectroscopy as a technique for the analysis of vehicle
emissions. Unlike conventional emissions analyzers, the FTIR is a single
instrument capable of performing simultaneous multicomponent analyses. In
addition, FTIR systems have been shown to be more practical, and as sensitive
as, wet chemical techniques for the analysis of non-regulated emissions such
as methanol and formaldehyde. In more recent years, commercial versions of
these FTIR analyzers have become available through several manufacturers.
This paper reviews data acquisition and processing techniques employed by
several of the prototype and commercial FTIR systems. Techniques utilized
by the FTIR emissions analyzer developed in this laboratory, and several of
its unique capabilities, are also addressed in detail. In addition,
instrumental and environmental parameters that alter spectral line widths,
and consequently affect the accuracy of this high resolution system, are
explored.
The FTIR analyzers currently in use employ a variety of data acquisi-
tion and processing techniques to quantify gaseous species. Most often, the
frequencies of unique absorption bands are used to identify the chemical
species, and peak heights or peak areas are used to determine their
concentrations. Linear processing techniques, such as masking, subtrac-
tion,7'9 and linear regression programs5'6'8 are used if the behavior of such
absorption bands obeys Beer's Law. If the behavior of these bands deviates
from Beer's Law, nonlinear techniques such as least squares fitting programs,
which rely on higher order regressions, must be employed.8'11
Selecting "interference free" absorption lines for the quantification
of gaseous species in complex mixtures is often a difficult task. If the
resolution of the spectrometer is not sufficient, interferences can arise
from overlapping lines. Often times, it is difficult to correct for such
interferences. It is for this reason that high resolution FTIR spectroscopy
is preferred for quantification. Least squares fitting programs have been
used in conjunction with lower resolution spectroscopy with reasonable
success. There are several advantages and disadvantages associated with
high resolution spectroscopy. Typically, high resolution spectra require
greater data acquisition times, are noisier than the lower resolution
counterparts, and require the use of computers with greater memory and
processing capabilities. Higher quality (higher cost) interferometers are
needed to acquire precise, high resolution infrared spectra. In addition,
to maximize resolution infrared beam diameters must be keep to a minimum,
thus affecting instrumental throughput and sensitivity.
The advantages of high resolution systems are illustrated using an
infrared spectrum of dilute vehicle exhaust (Figure 1). It is evident that
absorption lines associated with water and carbon dioxide dominate the
spectrum. Without the ability to resolve the unsaturated lines from the
saturated lines, only a few, rather narrow spectral regions remain free from
interferences. The physical or chemical removal of water or carbon dioxide
from such samples improves the utility of the spectral region, but presents
several problems. Techniques used to eliminate water will also remove water
301
-------
soluble components, such as methanol and formaldehyde. In addition, it is
difficult to perform quantitative analyses after "drying" a sample unless the
amount of water and water solubles removed can be accurately determined. An
alternate technique used to remove water and carbon dioxide absorption bands
is spectral subtraction. Most often, in the case of raw or diluted exhaust
samples, the water and carbon dioxide absorption lines are saturated and
cannot be completely eliminated by linear subtraction methods. If used
properly, high resolution systems can take advantage of the unsaturated
regions of the spectrum and perform interference free analyses of several
regulated and non-regulated emissions components (Table I).
In addition to the identification and quantification of exhaust compo-
nents, several of the commercially available systems provide time-resolved
emission data. Coupled with direct sampling devices, such as a dilution
tube3'4 or Constant Volume Diluter (CVD),12 these systems are useful for
evaluating factors that affect emission levels during transient phases of
vehicle tests. Relationships between variables such as road speed or post
catalyst temperatures and emission levels of several components can be
established. Examples of some of the information available using time-
resolved data acquired on the FTIR emissions analyzer developed in this
laboratory are presented in Figures 2 and 3.
Restricted by computer processing and storage capabilities, some FTIR
analyzers are limited to acquiring integrated or signal averaged data, as
opposed to real-time data. Consequently, such systems are often coupled with
indirect sampling devices, such as Tedlar bags or other sample containers. '
These systems have also been used with direct sampling devices to analyze
emissions during steady state vehicle experiments, or to obtain integrated
emission values for transient cycles.
Experimental Methods
Instrumentation
The FTIR emissions analyzer developed in this laboratory is comprised
of three basic subsystems; 1) the infrared spectrometer and optical
accessories for gas phase measurements, 2) the data acquisition and
processing system, and 3) the associated sampling hardware. The high
resolution Fourier transform infrared spectrometer (Nova-Cygni 120 model,
Mattson Instruments, Inc., Madison, WI) is a research grade instrument
capable of acquiring quality, eighth wavenumber (0.125 cm ) spectra. (Due
to computer limitations and time resolution constraints, the system is
operated at a resolution of 0.25 cm" , and the interferograms are zero filled
to an effective resolution of 0.125 cm ) . The spectrometer is equipped with
a water cooled, glow bar source and a narrow band MCT detector with a
linearized preamplifier. A variable pathlength, multi-pass gas cell (Wilks
20 Meter - Model 9020} with potassium bromide windows, is used exclusively
at the 14th order setting (21.75 meters) to improve the sensitivity of the
system for all chemical species.
Data acquisition and processing is controlled by a Concurrent computer
equipped with an array processor (Model 5450). The computer is a multi-user
system based on the UNIX operating language. The data acquisition and
processing routines were developed at the Ford Motor Company Scientific
Research Laboratories. This software was designed for the specific purpose
of (but not limited to) analyzing multicomponent gas phase samples composed
302
-------
of chemical species typically found in dilute vehicle exhaust. The system
can be used to acquire either time-resolved or signal averaged data. It is
capable of co-adding, transforming, and analyzing high resolution spectra in
3 second intervals. Currently, 24 distinct components are monitored on a
real-time basis. The real-time detection limits for these components are
reported in Table II. It should be noted that the detection limits for the
concentration ranges most often encountered during vehicle testing are
listed. Other ranges, which are more appropriate for atmospheric concentra-
tion levels, provide greater sensitivity for a limited number of species.
Signal averaging can be employed during steady-state engine or vehicle
experiments to improve detection limits. The hardware and software used with
the system are capable of monitoring up to 50 individual components. Species
other than those listed in Table II can be added to the system by acquiring
a representative reference spectrum, developing an effective mask, and
incorporating the calibration factor into the existing software. This
operation is performed routinely in our laboratory.
The sampling hardware consists of high flow rotary vane and direct
drive high vacuum pumps, an absolute pressure meter, and 0.5 in. OD stainless
steel transfer lines and valves. Direct sampling of dilute exhaust is
performed using the high flow pump (Model 5KC48PG656BS, General Electric,
Fort Wayne, IN). The high vacuum pump (Model E2M-18, Edwards, Sussex,
England) accommodates samples acquired indirectly, such as those contained
in Tedlar bags or other sample vessels. Either the high flow or the high
vacuum pump can be used to sample pressurized gas cylinders directly. The
reference spectra used for quantitative analyses were acquired at sample
pressures of approximately 933 mbar (700 torr). To prevent inaccuracies due
to line broadening, the absolute pressure meter (Balzers - Model APG 010) is
used to monitor sample pressures, so that they can be maintained at or as
close to 933 mbar as possible.
Instrumental and Environmental Parameters and Their Effects
on the Accuracy of this System
As mentioned above, the quantitative capabilities of the FTIR emissions
analyzer developed in this laboratory are based on high resolution infrared
spectroscopy. A procedure called "masking"10 is used to both identify and
quantify absorptions in the sample spectra. A mask consists of select
groups of spectral channels called segments. The segments are chosen from
the unsaturated regions of the spectrum. Several channels are used to form
either the background or signal segments of the mask. Background segments
are used to determine the baseline of the sample spectrum in a given region.
Ideally, background segments contain data channels to the immediate right and
left of the signal channels. Signal segments are used in combination with
the background segments to determine the area under an absorption band.
Instrumental and environmental factors that alter spectral line widths can
affect the accuracy of the high resolution masking technique. The general
effect of variations in spectral line widths on the accuracy of the masking
technique is illustrated in Figure 4. The extent of these effects vary
depending on the nature of the absorbing species and the mask used. For
example, masks for chemical species which have very narrow absorption bands,
such as C02 and NO, may contain as few as one or two channels in their signal
segments. If the instrumental resolution is insufficient, or if natural line
widths are broadened due to an increase in sample pressure or temperature,
the spectral line widths of such species will increase. In such cases, the
channels chosen for the background segments may be skewed or shifted upward
due to the line broadening, while the signal channels are simultaneously
303
-------
shifted downward. Consequently, the areas determined by the mask segments
will be less than those areas established using the reference spectra. In
such cases, the accuracy of the masking technique is affected.
Several experiments were designed to investigate the effects of varia-
tions in spectral line widths on the accuracy the system. The experiments
were performed using static gas phase samples from a high pressure cylinder
containing a low concentration nitric oxide diluted with nitrogen. The
natural line widths of nitric oxide absorptions are less than the resolution
of the analyzer, therefore, the effects of variations in spectral line widths
are expected to be greatest for this species. The cylinder (19.8 ppm ±1% NO)
was obtained from AIRCO {Murray Hill, NJ), and its content was certified by
the Ford Motor Company Gas Standards Laboratory. All studies were performed
using the prototype hardware and the software developed at Ford Motor
Company.
Results
Temperature Study
The temperature study encompassed temperature fluctuations that might
be seen under normal atmospheric variations and under more extreme varia-
tions. (The use of heated raw exhaust samples versus room temperature dilute
exhaust samples has been a issue of recent debate). A heating jacket
(Foxboro Company - Model 500-0109) designed specifically for the Wilks gas
cell, was used to control the sample temperature. Thermocouples were placed
between the blanket and the cell, and inside the outlet port of the cell to
monitor the temperatures of the exterior of the gas cell and of the sample
gas. Temperature differences as great as -35°F between the exterior of the
cell and the sample gas were seen at higher sample temperatures (~160°F) .
Although the gas cell is rated for temperatures as high as 230°F, the sample
temperatures were maintained below ~150°F to safely accommodate these
temperature differences. Spectra were acquired at four temperatures, ranging
from room temperature to -140°F.
If the gas obeys the ideal gas law, a inverse linear relationship
between sample temperature and instrument response is expected. (If the
pressure and the volume of the cell remain the same, an increase in
temperature will result in a decrease in the number of molecules in the
lightpath). The theoretical relationship between instrument response (R) and
sample temperature (T) developed using the ideal gas law and the instrument
response at 84°F is the following:
R -. (-0.034 ppm/°F)T +23.1 ppm.
The plot of instrument response as a function of sample temperature is
presented in Figure 5. A linear regression of these data was performed and
it was determined that the experimental relationship between instrument
response and sample temperature is equivalent to the theoretical relationship
(within experimental error) for the given concentration of nitric oxide.
(slope - -0.037 ppm/°F, intercept =23.4 ppm, and a correlation coefficient
- 0.996). In addition, the linear subtraction of a spectrum acquired at room
temperature from one acquired at 140°F indicates no appreciable line
broadening with increasing temperature.
304
-------
Pressure Study
The direct sampling devices routinely used with the FTIR emissions
analyzer developed in this laboratory are open to the atmosphere. Thus,
significant ambient pressure changes during a vehicle or engine test, such
as passing storm fronts, will cause sample pressure changes which may impact
the system's accuracy. In addition, there is some concern that results
obtained using the commercial version of this instrumentation in testing
facilities that are located at higher altitudes (with ambient pressures below
the recommended sampling pressure of 700 torr) such as Denver, Colorado, will
not be comparable to results obtained at other testing facilities. To
address these issues, a pressure study was undertaken to establish the
affects of ambient pressure changes on the accuracy of system. The
analyzer's performance was characterized for nitric oxide at pressures
ranging from approximately 933 mbar to 150 mbar.
If the ideal gas law is applied to the case of the pressure study, a
direct linear relationship between sample pressure and raw instrument
response is expected. If the raw instrument response was corrected using
the ratio of the reference pressure (933 mbar) to the sample pressure, the
system should provide accurate analyses for samples at pressures other than
933 mbar. Unfortunately, in practice, this does not hold true. A plot of
raw and corrected instrument responses obtained experimentally (Figure 6)
reveals that the relationship is somewhat linear over a narrow pressure
range, but becomes fairly nonlinear over a broader pressure range (933 mbar
to 150 mbar). The relationship over this broader range is more accurately
described by a second order polynomial equation. For the given concentration
of nitric oxide, the following polynomial relationship has been established
between raw instrument response (R) and sample pressure (P):
R = [-9.59 x 10'6 ppm/(mbar)2]P2 + [2.99 x 10'2 ppm/mbar]P + 0.55 ppm.
Several reasons for this departure from linearity may be considered.
Instrumental factors, such as changes in detector linearity with varying
throughput, might be responsible. In addition, inter- or intra-molecular
interactions may change at lower pressures, thereby resulting in different
Line shapes and intensities. The band shapes that result from a linear
subtraction of a spectrum acquired at 150 mbar and one acquired at 933 mbar
indicate a reduction in spectral line width with a reduction in pressure,
supporting the above theory.
Aperture Study
Factors, such as beam diameters, and the quality and alignment of
optical components can affect the resolution of FTIR spectrometers. It has
been demonstrated in the pressure study, that variations in spectral line
widths can effect the accuracy of the system. An aperture study was
undertaken to determine the relationship between instrumental resolution and
che performance of the analyzer. To determine the effect of aperture size
(beam diameter) on the accuracy of the nitric oxide analysis, instrument
response was monitored for several aperture settings. Adjustments of the
diaphragm aperture are controlled by a stepper motor. Aperture size can be
changed by selecting the number of steps the motor undergoes relative to a
zero point. This mechanism is very reproducible, but the relationship
between motor steps and aperture area is not a linear one. Spectra were
acquired at three stepper motor setting: 30, 40, and 50 steps, which
correspond to aperture areas of approximately, 0.30, 0.52, and 1.27 mm ,
305
-------
respectively. (The increase in throughput at a setting of 50 steps prompted
the use of a 50% transmitting, neutral density filter to prevent saturation
of the detector). A plot of instrument response as a function of aperture
area is presented in Figure 7. A linear regression was performed on these
data, and it was established that the relationship between instrument
response and aperture area is linear for the given concentration of nitric
oxide, for the range of aperture areas studied (slope - -1.85 ppm/mm ,
intercept - 21.0 ppm, and a correlation coefficient - 0.998). In addition,
the linear subtraction of a spectrum acquired using an aperture setting of
30 from a spectrum acquired using a setting of 50 was performed. The
resultant band shapes indicate loss of resolution with increasing aperture
setting.
Conelusions
The quantitative capabilities of this FTIR emissions analyzer are based
on the use of high resolution infrared spectroscopy. Therefore, variations
in spectral line widths may affect the accuracy of the system. A study
investigating the extent of such effects on the accuracy of the nitric oxide
analysis was performed. It was determined that linear equations can be used
to describe the relationships between instrument response and parameters such
as sample temperature and aperture area over the ranges studied. On the
other hand, the relationship between instrument response and sample pressure
is somewhat nonlinear over the pressure range studied. For narrower pressure
ranges the relationship can be described using linear equations.
It has been determined from this work, that ambient temperature
fluctuations do not appreciably alter spectral line widths. Variations in
sample temperature of ±10°F are expected to result in inaccuracies on the
order of ±2% for samples containing nitric oxide levels similar to those used
in this study. If accuracies greater than ±2% are needed, the regulation of
sample'temperatures should be considered.
It has also been determined, that ambient pressure fluctuations
measurably alter the widths of NO absorption lines. Variations in sample
pressure of ±50 mbar result in errors on the order of ±3% for samples
containing lower levels of nitric oxide. Feed back circuity has been used
successfully in this laboratory to regulate the sample pressures during
vehicle testing, and is recommended if system accuracies greater than ±3%
are required. For locations that have naturally low ambient pressures,
correction factors, which compensate for lower sample pressures can be
readily incorporated into the software.
It is apparent from this study, that the aperture size (beam diameter)
has a noticeable effect on instrument resolution, and hence instrument
response in the specific case studied. A maximum instrument response error
of 6% for nitric oxide is observed for changes in aperture area from 0.52
mm to 1.27 mm . Since the changes in instrument response behave linearly
over the range considered, it is conceivable that the data could be corrected
for the variations in aperture size. However, to eliminate the effects of
loss of instrument resolution due to beam diameter fluctuations, it is
strongly recommended that the resolution of the reference spectra and the
sample spectra be matched as closely as possible, and that aperture size
remain constant.
306
-------
The results of this study Indicate that routine variations in
parameters, such as instrument resolution, sample temperature and pressure
have minimal effects on the accuracy of the FTIR analysis of low
concentrations of nitric oxide. Studies that may further clarify the these
effects for nitric oxide, as well as for several other emission components,
are in progress.
Acknowledgement
The prototype instrumentation developed in this laboratory was commercialized
in 1988 by Mattson Instruments, Inc., Madison WI. This instrumentation may
be further developed, in the near future, by another manufacturer.
References
1. P.O. Maker, H. Niki, C.M. Savage, L.P. Breitenback, "Fourier
Transform Infrared Analysis of Trace Gases in the Atmosphere,"
Monitoring Toxic Substances". 94, ACS, Washington, D.C. 1979, pp. 161-
175.
2. J.W. Butler, P.O. Maker, T.J. Korniski, L.P. Haack, "On-Line
Characterization of Vehicle Emissions by FT-IR and Mass Spectrometry,"
SAE Technical Paper Series 810429: (1981).
3. J.W. Butler, P.O. Maker, T.J. Korniski, L.P. Haack, F.E. McKelvy,
A.D. Colvin, "A System for On-Line Measurement of Multicomponent
Emissions and Engine Operating Parameters," SAE Technical Paper Series
851657: (1985).
4. L.P. Haack, D.L. LaCourse, T.J. Korntskt, "Comparison of Fourier
Transform Infrared Spectrometry and 2,4-Dinitrophenylhydrazine Impinger
Techniques for the Measurement of Formaldehyde in Vehicle Exhaust,"
Anal. Chem. 58. 68. (1986).
5. J. Staab, H. Klingenberg, W.J. Riedel, "Progress in the Prototype
Development of a New Multicomponent Exhaust Gas Sampling and Analyzing
System," SAE Technical Paper Series 840470: (1984).
6. J. Staab, H. Klingenberg, H. Pfluger, W.F. Herget, M.I. Tromp, "First
Experiences in Testing a New Multicomponent Exhaust Gas Sampling and
Analyzing System," SAE Technical Paper Series 851659: (1985).
7. P.M. Hanst, E.R. Stephens, "Infrared Analysis of Engine Exhausts:
Methyl Nitrite Formation from Methanol Fuel," Spectroscopy. 4, 33:
(1989).
8. B. Heller, H. Klingenberg, G. Lach, and J. Winckler, "Performance
of a New System for Emission Sampling and Measurement (SESAM)," SAE
Technical Paper Series 900275: (1990).
9. P. Reiger, "Application of FTIR to Motor Vehicle Exhaust
Measurements," Proc. 82nd Ann. Meet. Air and Waste Mang. Assoc..
#89-4B.5: (1989).
307
-------
10. J.W. Butler, P.D, Maker, T.J. Korniski, A.D. Colvin, (Ford Motor Co.)
U.S. Patent 4,801,805 (January 31, 1989).
11. B. Herget, "Spectral Resolution and Species Quantification," Madison,
WI, written communication, (1989).
12. J.W. Butler, P.D. Maker, T.J. Korniski, A.D. Colvin, (Ford Motor Co.)
U.S. Patent Appl. #88-497.
13. P. R. Griffiths, Chemical Infrared Fourier Transform Spectroscopy.
John Wiley and Sons, New York. 1975, pp.31-40.
308
-------
Table I. Spectral regions used for quantitative analysis of some exhaust components.
Component
carbon monoxide
nitric oxide
nitrogen dioxide
nitrous oxide
sulfur dioxide
methane
Spectral Region (cm-1)
3010 - 2220
1800 - 1940
1580 - 1630
2185 - 2220
1330 - 1385
1240 - 1345
Table II. Real-time detection limits of the FTIR analyzer
Component
CH20 (Formaldehyde)
CH30H (MethaneL)
CO (Carbon Monoxide)(4)
C02 (Carbon Dioxide)(4)
SUM_HC(5)
SUMJJOXC6)
MO (Nitric Oxide)
N02 (Nitrogen Dioxide)
N20 (Nitrous Oxide)
HONO (Nitrous Acid)
CH4 (Methane)
C2H2 (Acetylene)
C2H4 (Ethylene)
C2H6 (Ethane)
C3H6 (Propylene)
1C4H8 (Isobutylene)
13C4H6 (1,3-Butadiene)
HCIO (nonspeciated HC)(5)
CH3CHO (Acetaldehyde)
C2H60 (Ethanol)
HCOOH (Formic Acid)
S02 (Sulfur Dioxide)
H20 (Water)
HCN (Hydrogen Cyanide)
NH3 (Ammonia)
CF4 (Carbon Tetrafloride)
(1): Limits for UDDS cold start transient phase (Bag 1) or the hot start
transient phase (Bag 3) at a dilution flow rate of 700 CFM.
(2): Limits for UDDS cold start stabilized phase (Bag 2)
for a dilution flow of 700 CFM.
(3): Limits for the Cold/Hot calculation for a
dilution flou rate of 700 CFM.
(4): Ranges appropriate for atmospheric levels provide
greater sensitivity for these components.
(5): Molecular weight based on a hydrogen to carbon ratio of 1.85:1.
(6): Based on the molecular weight of N02.
Detection Limit
(ppm)
0.54
1.8
8.9
470
19 ppmC
3.6 ppmN
0.39
2.8
0.10
2.4
1.7
0.62
1.3
0.54
4.2
1.3
1.8
5.7 ppmC
3.6
0.62
0.39
0.73
3200
0.44
0.85
6.4E-03
Bags 1 or 3 (1)
(ma/mile)
700 CFM
2.6
9.1
39
3.2E+03
42
23
1.6
18
0.7
16
4.2
2.5
5.6
2.6
28
11
15
12
25
4.5
2.8
7.4
9.0E+03
1.9
2.3
8.8E-02
Bag 2 (2)
(ma/mile)
700 CFM
2.0
6.9
30
2.5E+03
32
17
1.2
13
0.5
12
3.2
1.9
4.2
2.0
21
8.7
12
10
19
3.4
2.2
5.6
6.9E+03
1.4
1.7
6.8E-02
Cold/Hot (3)
(ma/mile)
700 CFM
2.3
8.0
34
2.9E+03
37
20
1.4
15
0.6
14
3.7
2.2
4.9
2.3
24
10
14
11
22
3.9
2.5
6.5
7.9E+03
1.6
2.0
7.8E-02
309
-------
3.5
Hav«length(urn)
5
A
b
s
o
r
b
a
n
c
e
3200
2800 2400 2000
Havenumbers
1600
1200
BOO
Figure 1. High resolution spectrum of dilute vehicle exhaust from an
alternate fuel vehicle running on a mixture of 85% methanol and 15% gasoline.
Methane Emissions / Post-Catalyst Temperature
Gasoline Fuel
15 , , 600
10 -,
W
W
<0
0 12345678
Time (minutes)
• Methane Post-Catalyst
Figure 2. Methane emissions and catalyst temperature as a function of time
for a gasoline fueled vehicle during the cold start transient phase (Bag 1)
of the Urban Dynamometer Driving Schedule (UDDS).
310
-------
Formaldehyde Emissions for Bag 1
25
20
6 15
CO
M-85 Fuel
234587
Time (minutes)
IHCHO Road Speed
8
70
60
50
40
a.
>*
4->
f*
o
o
"30 '3
20
10
9
Figure 3. Formaldehyde emissions and road speed as a function of time for
an alternative fuel vehicle running on 85% methanol and 15% gasoline during
Bag 1 of the UDDS.
Normal Line
Broadened Line
I i Background (B) and
1 Signal (S) Channels
BBBBSSSBBBB
Figure 4. The effect of a variation in the spectral line width on the
accuracy of the masking technique.
311
-------
TEMPERATURE STUDY
Nitric Oxide
25
23--
f -^
a
-------
r "^
a
ex
,3
0)
on
K
22
21
20--
19--
18--
17
0.000
mis STUDY
Nitric Oxide
R = aA + b
a = -1.85 ppm/mm
b = 21.0 ppm
1.500
Figure 7.
oxide.
0.500 1.000
Aperture Area (mm2)
Instrument response as a function of aperture area for nitric
313
-------
EMISSIONS FROM A FLEXIBLE-FUELED VEHICLE
Peter A. Gabele
U.S.Environmental Protection Agency
Research Triangle Park, North Carolina
William Crews and Paula Siudak
NSI Technology Services, Inc.
Research Triangle Park, North Carolina
In anticipation of a shift from gasoline to methanol,
flexible-fueled automobiles capable of operating on various
mixtures of both fuels are being developed. This study
examines both the exhaust and evaporative emissions from a
prototype General Motors Variable Fuel Corsica. Results are
reported for tests conducted at temperatures of 40°, 75°, and
906F, and for fuels MO, M25, M50, M85, and M100. In addition
to regulated emissions and fuel economy, emission rates for
methanol, aldehydes, and a large number of hydrocarbon
compounds were measured. The data indicate that as fuel
methanol content is increased, formaldehyde and methanol
comprise increasingly greater portions of the organic material
while hydrocarbons comprise less. Increasing fuel methanol
content has no significant effect- on exhaust regulated emission
rates (organic material, carbon monoxide, and nitrogen oxides)
nor on the composition of total hydrocarbons, except for
methane, which increases substantially. The effect of ambient
temperature on both exhaust and evaporative emissions is
similar to its effect on gasoline cars: organic and carbon
monoxide exhaust emissions increase substantially at the lower
temperatures, and evaporative emissions increase steadily with
increases in temperature.
314
-------
INTRODUCTION
The lack of progress toward attainment of the ambient
ozone standard has prompted many areas of the country to press
for stricter emission standards and other abatement strategies.
In Southern California, regional officials have voted for a
plan that would, by 1998, convert 40% of the cars to cleaner
fuels. If enacted, the plan would ban the use of conventional
gasolines in Southern California by 2007.
Currently, the most promising alternative fuel for
automobiles appears to be methanol. In anticipation of a fuel
switch to methanol, some automobile companies have already
developed vehicles which are designed to operate on methanol
and/or gasoline. One scenario suggests that these flexible-fuel
cars could proliferate during a transition of fuels to comprise
a significant fraction of the consumer fleet. For this reason,
a study was undertaken to characterize both the exhaust and
evaporative emissions from a flexible-fuel automobile. The
data obtained in this study will be used in air quality models
to predict the impact of methanol usage on ambient ozone
levels.
EXPERIMENTAL PROCEDURES
Test Vehicle
The test vehicle used in this study was a 1988 General
Motors (GM) Variable Fuel Corsica having a conventional, closed
loop, three-way catalyst. Its engine is a port fuel injected,
2.8-1 six-cylinder configuration, having a compression ratio
of 8.9:1. A fuel sensor electronically monitors the oxygen
content of the fuel being injected to continuously adjust
engine parameters, enabling optimum operation with respect to
fuel alcohol content.
The Variable Fuel Corsica was obtained on loan from the GM
Advanced Engineering Staff (AES) to EPA for testing. Prior to
delivery, the Corsica was tested for both exhaust and
evaporative regulated emissions by GM. Much of these data are
included in the paper for comparative purposes when regulated
emissions data are presented.
Fuel Description
The five fuels examined in this study were MO, M25, M50,
M85, and M100. In these five descriptors, the prefix "M" is
followed by a number representing the volume percent of
methanol mixed with gasoline. For example, an M85 fuel
contains 85% methanol and 15% gasoline. The methanol used was
of a laboratory grade specification and the gasoline with which
it was mixed was indolene, the standard certification fuel. The
Reid vapor pressures (RVP) measured for the MO, M25, M50, M85,
and M100 fuels were 9.0 psi, 10.9 psi, 10.1 psi, 8.0 psi, and
4.6 psi, respectively.
Facility Description and Emission Measurement
All tests were conducted at the EPA Environmental Research
315
-------
Center Annex, Research Triangle Park, North Carolina. The
chassis dynamometer used for vehicle road load simulation is
enclosed within a temperature controlled test chamber (TCTC),
which permits vehicle soak and operation at temperatures
ranging from 20°F to 100°F.
Emissions were examined for both regulated (organic material
(OM) , carbon dioxide (CO), and nitrogen oxides (NOx)) and
unregulated pollutant emissions. Regulated emissions were
measured using the procedures set forth in the Federal Register
(1) . Nonregulated pollutants, consisting of aldehydes and
speciated hydrocarbons, were measured using methods previously
published (2,3).
Run Schedule and Test Descriptions
The overall test program was broken down into test sets and
subsets. The sets consisted of all runs with a given test
fuel, and the subsets were all runs at a given test
temperature. Because there were five fuels being examined,
there were a total of five test sets which were run in the
order of MO, M85, M100, M50, and M25. There were three subsets
run in the order of 75°F, 40°F, and 90°F, and three replicate
tests were run within each subset. The subset for M100 at 40°F
was not run because the vehicle could not be cold started on
M100 at temperatures below 60°F.
RESULTS AND DISCUSSION
Exhaust Emissions
Exhaust emission rasults are presented in Table 1. These
presentations enable emission rate comparisons for the two
principal variables examined in the study: fuel methanol
content and ambient temperature.
Emission rates, except for NOx, are significantly higher at
the lowest temperature examined (40°F) . This is expected
because organic and CO emissions typically increase as ambient
temperature decreases. In a similar study involving a dedicated
methanol car (4) , the increases in HC, CO, and methanol
emissions were even more pronounced than those presented here
with M85 fuel.
The effect of fuel type (fuel methanol content) on CO,
NOx, and total organic exhaust emission rates is not
significant; however, fuel type does strongly influence the
composition of organic material. Increases in fuel methanol
content result in increases in methanol and formaldehyde
emission rates with corresponding decreases in HC emission
rates. In addition to these compositional changes, hydrocarbon
compositions are also affected. As fuel methanol content is
increased, steady but significant increases in the methane
fraction occur.
The fractions of hydrocarbon compounds other than methane
are much less sensitive to either fuel methanol content or
ambient temperature. Table 2 provides fractions as percentages
for 6 of the more important hydrocarbon compounds contained in
316
-------
exhaust emissions. The final column in the table gives average
values from recent exhaust emission studies conducted with
gasoline vehicles (5,6). In general, the hydrocarbon
composition from gasoline vehicle exhaust emissions appears
similar to those from the flexible-fuel vehicle, except for
methane, as discussed above.
An additional formaldehyde measurement was made during the
first 125 s of the test to determine the portion of
formaldehyde emissions occurring immediately after the cold
start. Formaldehyde emissions, which occur during the cold
transient test phase (first 505 seconds), generally comprise
a vast majority of the total test formaldehyde emissions.
Table 3 reports the emission rates and percentages of
formaldehyde emitted during the first 125 s and the first 505
s of operation.
Evaporative Emissions
Evaporative emission summaries are presented in Table 3. As
expected, evaporative emissions increase with increase in
either test temperature or fuel volatility. For the 90°F
tests, the order of increasing evaporative emission rates
corresponded precisely with order of increasing fuel
volatility: M100 < M85 < MO
-------
REFERENCES
4. R. Snow, L. Baker, W. Crews, C.O. Davis, J. Duncan, N.
Perry, P. Siudak, F. Stump, W. Ray, J.Braddock,
"Characterization of Emissions from a Methanol Fueled
Motor Vehicle", JAPCA, Vol. 39, pp 48-54 (1989).
1. Federal Register, 40CFR Part 86, "Standards for Emissions
from Methanol-Fueled Motor Vehicles and Motor Vehicle
Engines" Final Rule, April 11, 1989, Vol. 54, No. 68.
2. S.B. Tejada, "Evaluation of Silica Gel Cartridges Coated
in Situ with Acidified 2,4-dinitrophenyl-hydrazine for
Sampling Aldehydes and Ketones in Air", Intern. J. Environ.
Anal. Chem., Vol. 26, pp. 167-185 (1986).
3. F.D. Stump, D.L. Dropkin, "Gas Chromatographic Methods for
Quantitative Determination of C2 to C12 Hydrocarbons in
Roadway Vehicle Emissions", Anal. Chem., Vol. 57, pp.
2629-22634 (1985).
5. F. Stump, S. Tejada, W. Ray, D. Dropkin, F. Black, W.
Crews, P. Siudak, C.O. Davis, P. Carter, "The Influence
of Ambient Temperature on Tailpipe Emissions from Late
Model Light-Duty Gasoline Motor Vehicles", Atmospheric
Environment, Vo. 23, p. 307, February 1989.
6. F. Stump, S. Tejada, W. Ray, D. Dropkin, F. Black, R.
Snow, W. Crews, P. Siudak, C.O. Davis, P. Carter, "The
Influence of Ambient Temperature on Tailpipe Emissions
from 1985-1987 Model Year Light-Duty Gasoline Motor
Vehicles - Part II", to be published in Atmospheric
Environment, also available through EPA, MD-46, RTF, NC
27711.
318
-------
Table 1. Exhaust emissions
Table 4. Evaporative emissions
MO
40 F
75 F
90 F
M25
40 F
75 F
90 F
M50
40 F
75 F
90 F
M85
40 F
75 F
90 F
M100
75 F
90 F
OMHCE
(a/mi)
0.92
0.32
0.36
0.78
0.28
0.30
0.67
0.27
0.25
1.00
0.24
0.32
0.42
0.31
HC
(d/mi)
0.92
0.32
0.36
0.70
0.22
0.25
0.47
0.17
0.15
0.35
0.11
0.17
0.04
0.03
MeOH
(a/mi)
0.00
0.00
0.00
0.18
0.11
0.11
0.44
0.21
0.20
1.45
0.29
0.31
0.82
0.60
HCHO
(ma/mO
5.6
4.8
4.7
13.0
8.8
7.1
23.2
14.1
12.9
58.0
27.3
28.5
40.6
31.7
CO
(0/mO
8.80
2.60
2.80
8.46
2.50
3.60
7.50
2.90
3.00
8.50
2.60
2.40
3.10
3.10
NOx
(0/mi)
0.20
0.22
0.26
0.22
0.24
0.26
0.24
0.26
0.24
0.28
0.26
0.31
0.27
0.24
notes (a) Oganta-HC+13.8/32(NteOH)+13.8/30(HCHO)
(b) no 40 F M100 tests.
Table 2. Fractions for six selected exhaust
HC compounds. (% of total HCs)
MO
40 F
75 F
90F
M25
40 F
75 F
90 F
M50
40 F
75 F
90 F
M85
40 F
75 F
90 F
M100
75 F
90F
OMHCE
(arams)
0.45
0.82
1.45
0.56
1.45
5.38
0.25
0.58
1.67
0.28
0.64
0.65
0.40
0.47
HC
(qrams)
0.45
0.82
1.45
0.47
1.34
4.83
0.17
0.46
1.47
MeOH
(arams)
0.00
0.00
0.00
0.20
0.26
0.62
0.19
0.28
0.43
1
0.17 0.16
0.48 0.25
0.50
0.12
0.10
0.31
0.58
0.85
METHANE
ETHYLENE
PROPYLENE
BENEZENE
TOLUENE
1,3-BUTADOJE
THC
(g/mi)
Test Temp
rn
40
75
90
40
75
90
40
75
90
40
75
90
40
75
90
40
75
90
40
75
90
MO
10
10
13
5
5
5
3
4
4
3
3
3
14
14
13
0.3
0.5
0.4
0.9
0.3
0,3
M25
10
12
13
5
6
5
3
4
4
3
3
3
14
14
11
0.6
0.6
0.6
0.7
0.2
0.3
M50
11
15
17
5
6
5
3
4
4
3
3
3
15
12
12
0.4
0.6
0.8
0.5
0.2
0.2
Mas
15
25
19
5
5
4
2
3
2
3
3
3
14
12
12
0.2
0.4
0.6
0.3
0.1
0.1
M100
50
56
1
2
0.5
0.0
0.7
1.5
2.0
3.0
0.0
0.0
0.02
0.01
20 gas
vehicles
17
5
2
4
9
0.3
notes:(a) Organic-HC+138/32(MeOH)
(b) no 40 FM100 tests.
Table 3. Percent of FTP formaldehyde
emission at 125 s. and 505 s.
Amb.
Temp.
(F)
M85
M100
40
75
90
% Total
125s.
51
59
51
505s.
81
88
85
75
90
60
63
92
90
319
-------
A RETROSPECTIVE ANALYSIS OF A BASELINE AIR PATHWAY
ASSESSMENT AT A PRE-REMEDIAL SUPERFUND SITE
Richard W. Tripp, Jody Hudson, Harry Kimball,
U. S. Environmental Protection Agency-Region VII,
25 Funston Road, Kansas City, Kansas 66115
A Baseline Air Pathway Assessment (APA) was performed at a
Pre-Remedial Superfund Site containing surface and subsurface
chemical contamination. The objectives of the baseline APA were
to: 1) measure the average upwind and downwind concentrations
of selected VOCs at the 95% confidence level in the ambient air
around the site and compare these concentrations to the Ap-
plicable, Relevant, and Appropriate Requirements (ARARs) estab-
lished for this project; 2) determine whether the site is a
source of the VOCs measured in ambient air based on statisti-
cally significant differences at the 95% confidence level
between average upwind and downwind concentrations; and 3)
measure average concentrations of selected VOCs at the 95%
confidence level within selected dwellings and compare these
concentrations to the ARARs established for the project. A
Quality Assurance Project Plan (QAPP) was developed describing
the strategy and procedures used to perform the APA. The
project was implemented, whole air samples were collected and
analyzed, the data were assessed, and a report was prepared.
This paper describes and evaluates the study and discusses
alternate approaches for performing data analysis using the
pollutant and weather data obtained during the study.
320
-------
INTRODUCTION
A Superfund site containing a variety of stored chemicals
and both surface and subsurface soil contamination was
investigated to determine whether there was an ongoing air
release. The investigation consisted of performing a baseline
APA utilizing air monitoring procedures. A statistically
significant number of samples (6) were collected at ecich site
comprising the monitoring network around the site perimeter.
Another set of six samples were collected in crawl spaces of
nearby houses to determine if volatile organics were being
emitted in soil gas resulting from subsurface contamination
around the site. Summa polished spheres with flow control
valves were used to collect the samples, which were then ana-
lyzed by Gas Chromatography/ Matrix Isolation-Fourier Transform
Infrared Spectroscopy.
Sample results were grouped by sampling location in rela-
tion to the wind, and averaged. The averages of the grouped
results were then statistically compared to determine if meas-
ured VOC concentrations were significant.
EXPERIMENTAL METHODS
The study plan contained an air monitoring network which
was designed so that the monitoring sites would be located
upwind and downwind of the site according to the current and
predicted wind conditions. The network consisted of one
upwind monitoring station and three downwind monitoring
stations all of which were located near (but outside of) the
site fenceline. The downwind monitoring stations were located
directly downwind and at 45° right and left of the downwind
monitoring station. Sampling periods were limited to four
hours due to sampling equipment constraints. In order to
average out diurnal effects and other variables which may cause
short term concentration fluctuations, multiple 4 hour sampling
events were performed with the individual events collected over
different periods of a 24 hour cycle. A total of six sampling
events were performed. An on-site meteorological station was
used to measure wind direction and velocity. A sampling event
was considered valid only if the actual wind direction was
appropriately aligned with the monitoring network for a minimum
of 70% of the 4 hour event. If the winds did not meet this
criteria, the samples were voided and the event repeated.
Sampling operations began on August 17, 1989, and were
completed on September 19, 1989. Collected samples were pre-
pared for analysis by cryogenic preconcentration with qualita-
tive and quantitative analysis accomplished using Gas Chroma-
tography/ Matrix Isolation-Fourier Transform Infrared Spectos-
copy/Flame lonization Detection (GC/MT-FTTR/FID). The samples
were analyzed for 17 target VOCs. The measured average
concentrations from each site within the monitoring network
were compared to the Applicable, Relevant, and Appropriate
Requirements (ARARs) contained in the Quality Assurance Project
Plan (QAPP). The ARARs for this project were based on the
321
-------
American Conference of Governmental and Industrial Hycfienists'
(ACGIH) Threshold Limit Values (TLVs) for the respective VOCs
reduced by a factor of 1/420.
RESULTS
The ambient air results showed only 1,1,1-trichloroethane,
trichloroethylene, methylene chloride, and tetrachloroethylene
at concentrations above the Method Detection Limits (MDL). The
mean concentrations and bounding 95% confidence limit at each
monitoring site were statistically compared to the ARARs and to
one another. This was done using the student t test for the
predicted upwind, downwind, downwind 45° right, and downwind
45° left, sites. Statistically, both methylene chloride and
tetrachloroethylene were at the same concentration at all
monitoring sites. Using this same test, 1,1,1-trichloroethane
was detected at a higher concentration 45 downwind left and
downwind center than upwind. Trichloroethylene was detected at
a statistically higher concentration 45° downwind left: than
upwind (Figure 1-4). Both l,l,l-trichloroethane and trichloro-
ethylene were below their respective ARARs at the 95% confi-
dence level.
The three nearby residences were sampled with six samples
collected from each crawl space. All crawl space samples were
collected from 8:00 am to 4:30 pm. Methylene chloride,
cis-l,2-dichloroethylene, hexane, 1,1,1-trichloroethane,
trichloroethylene, toluene, and tetrachloroethylene were found
above the MDLs. The average concentrations for each residence
was compared against the average blank value at the 95%
confidence level. There was no significant difference between
the blank concentrations and the concentrations found in the
crawl spaces.
ALTERNATE DATA ANALYSIS PROCEDURE
An analysis of the data was performed using the monitoring
data from the actual downwind rather then the predicted
downwind. This approach uses the predicted downwind Kite (i.e.
network centerline site) for the 1st and 6tn event and the 45°
left site for the remaining data points. The mean concentra-
tion was calculated using the individual concentrations meas-
ured from the sites.
A plot of the actual downwind VOC concentrations vs the
sampling time (Figure 5) shows that during daylight hours the
concentrations are significantly higher than at night. Two
possible causes were identified. One cause could be increased
daytime emissions from the contaminated soil on-site which was
found to contain trichloroethylene and tetrachloroethylene.
The higher daytime concentrations could result from higher soil
temperatures causing the soil to de-gas at a higher rate during
the daytime hours. The higher daytime concentrations could
also have been from emissions within an on-site building
322
-------
(chemical storage warehouse) where a cleanup effort was going
on during daylight hours. Probably, the higher daytine values
were caused by a combination of both possibilities.
In order to determine if the increased concentrations were
the result of emissions from the building, the concentrations
of airborne VOCs found within the building were compared to
concentrations found -in the soil, at the downwind site perime-
ter, and in the crawl spaces (Figure 6). Trichloroethylene is
the highest of the VOC in both building samples, the downwind
sample and the residence. Benzene, 1,1,1-trichloroethane and
toluene concentrations varied for second highest VOC :.n the
building samples, the downwind sample and the residence. Meth-
ylene chloride concentration was always the lowest. This fact
coupled with the fact that the warehouse doors were only open
in the daytime could explain the higher daytime ambient air
concentrations.
The total VOC concentrations were also then plotted along
with the days in which rain fall occurred. {Figure 7 & 8).
This appears to show that the rain had little effect on the
emissions.
CONCLUSION
Use of the data analysis procedures contained in the
project QAPP were appropriate to: 1) determine if VOCs were
leaving the site and at what concentration; 2) measure concen-
trations of VOCs in the crawl spaces; and 3) determine whether
the project ARARs were exceeded. Use of alternative techniques
of data analysis were effective in obtaining additional infor-
mation which facilitates an understanding of various factors
affecting the ambient air concentrations around the site.
REFERENCES
1. J. Hudson, "Quality Assurance Project Plan For Air
Toxics Monitoring at ..., U.S. Environmental Protection
Agency, Region VII.
2. R. Tripp, "Toxic Air Monitoring in Ambient Air at ...",
U.S. Environmental Protection Agency, Region VII.
3. R. Tripp, "Toxic Air Monitoring at Residences Near ...",
U.S. Environmental Protection Agency, Region VII.
323
-------
1,1,1- Trichloroethane
35—
-------
Tetrachloroethylene
35 -,
30 -
c 20 -
o
o
5
O 15 H
10 -
5 -
0
[I
[]
Upper 95%
Mean
Lower 95%
Upwind Downwind Downwind Downwind
45% Left Centerline 45% Right
Methylene Chloride
14 -
12 -
10 -
8 -
o
O 6 H
4 -
2 -
I)
(1
O Upper 95%
D Mean
Lower 95%
Upwind Downwind Downwind Downwind
45% Left Centerline 45% Riflht
Figure 3 & 4.
Statistical representation of the mean and 95%
confidence level for two compounds which were
reported at levels above the method detection
limit.
325
-------
Downwind Concentration Verses Time
350 -i
1
s
Concentration
3OO -
250 -
200 -
150 -
1OO -
50 -
A
^ \
-^ ^llll- "---rr
Midnight 4am
Trichloroethylene
trVTrichloroethane
B am Noon
Start time
4 pm 8 pm
Methylene Chloride
Tetrachloroethylene
Figure 5
Concentrations of the ambient results verses time
for the individual downwind site.
Comparison of Sampling Points
104-
a 103-
I 102-
!
o
o
O
X::J
wjj;
«*
*ir
«g
l«r
m
Fl
r
1
•'
Building 8/29 Yard Residence 1
Building "D/6 Downwind Event 1
• Methylene Chloride
D HVTrichloroethane
O Benzene
Figure 6.
D Trichloroethyiene
0 Toluene
(3 Tetrachloroethylene
Comparison of the building and soil results verses
the two highest events El & R4).
326
-------
Concentration and Rain
Downwind Ambient Data
OUw —
500 -
1
? 400 -
o
~ 3OO -
o
g
O 2OO -
10O -
flfllnl „
.
flfl I
Hi)
n J i
-5.0
-4.0
£
a
oc
-3,0 =
e
o
-2.0 1
-1.0
m ^_«
voc
2468 101214161820222426283032343638
Days
Figure 7. A comparison of the ambient VOC concentration
and the amount of rain.
Concentration and Rain
Residence 1
6OO -
5OO -
1
^400 -
Q
1
§300-
|
o
200 -
100 -
• VOC
RflrljL
iJ
2 4 6 8 10 12 14 16 18 20
nfl
p6.O
-5.0
-4.0 _
£
£
-3.0 °
1
o
-2.0
-1.0
22 24 G Rain
Days
Figure 8. A comparison of the residential VOC concentration
and the amount of rain.
327
-------
AMBIENT AIR MONITORING OF A SARA TITLE III
FACILITY USING THE TAGA 6000E MS/MS
David B. Mickunas and Vinod Kansal
International Technology Corporation
REAC Project
GSA Raritan Depot (MS-802)
Edison, NJ 08837
Thomas Pritchett
US EPA Environmental Response Team
GSA Raritan Depot (MS-101)
Edison, NJ 08837
SARA Title III legislation requires facilities to report the quantities of regulated chemicals that are
used on-site in an effort to determine the chemicals' fate. One pathway by which a chemical may leave
a facility is through volatilization into the atmosphere. These chemical emissions pollute the
environment and may be a potential health problem. The TAGA 6000E was used to investigate a
specific site for chemical losses due to volatilization by analyzing the ambient air on and off the
facility's property.
328
-------
Introduction
In 1986, the Emergency Planning and Community Right-to-Know Act, also known as Title HI,
was established. This act has four major sections, with Section 313 providing regulations for toxic
chemical release reporting. Facilities subject to this reporting are required to complete Toxic Chemical
Release Form (Form R) for specified chemicals. Information necessary to complete this form includes
the quantity of the chemical entering air, land, and water annually.
In 1989, the Research Engineering and Analytical Contract (REAC) was directed by the US
EPA Environmental Response Team (ERT) to provide analytical support to US EPA Region I in its
efforts to conduct an air monitoring study in the vicinity of a Title III facility in North Haven,
Connecticut. Monitoring was being conducted by Region I to determine if this facility was introducing
regulated chemicals into the air. The study was designed to monitor off site locations on the perimeter
of the facility for compounds whose presence in the ambient air would strongly suggest origination at
the facility. The air study also included the performance of on-site analyses for compounds, to locate
their sources.
Experimental Methods
Eight target compounds (Table I) were selected for continuous monitoring by the TAGA
6000E (Figure 1). Five of these compounds were selected based on their identification in the facility's
Title III report. The other three compounds were selected because they represent common solvents,
and because one is an indicator of target compound contribution due to vehicular exhaust. The
sampling was performed during periods when it was believed that the meteorological conditions would
permit observation of the maximum concentrations of the targeted compounds.
The TAGA 6000E performed direct-air sampling. Outside air was continuously drawn through
a port in the roof, at a flow rate of approximately one and one-half liters per second. The sample
passed over a glass splitter, where a representative sample entered the transfer line leading to the
source. The remaining flow was vented from the bus.
While continuously analyzing for the target compounds, mobile monitoring was performed
along public roads and over lanes around the commercial facilities surrounding the Title III plant
(Figure 2). As the mobile monitoring proceeded, the computer was flagged with a letter to denote an
event or a location; these flags were also recorded on the TAGA operator's log sheet. The flags are
indicated on the ion profiles (Figure 3) and correspond to the map locations associated with that file.
Once a plume was located, an effort to isolate the source was conducted by monitoring around the
suspected source to determine if any upwind contributions existed.
Summa canisters were used as the medium to collect whole-air samples. The canisters utilized
for this field activity had been cleaned and certified prior to release to the sampling crew; before any
sample was collected, each canister was checked for leaks with a pressure gauge to ensure that a proper
vacuum existed.
Once the TAGA identified a plume, the sampling strategy was to collect a Summa. The
collection was accomplished by connecting a Summa to the TAGA sample air flow tubing via a glass
splitter and a section of Teflon tubing. No attempt was made to regulate the flow rate into the
canister. After the Summa had achieved ambient pressure, the valve was closed, the tubing was
disconnected, and both the canister and the glass splitter were capped. The time, date, and location
of sampling were noted on the canister, and also included on the TAGA operator's log sheet.
The Summas were returned to Edison, New Jersey, where their pressure was recorded; ultra-
high purity nitrogen was added until the canister's final pressure was twice its initial pressure; and the
canister's contents were analyzed by gas chromatograph/mass spectrometer (GC/MS) for targeted and
non-targeted compounds. The Summas were also analyzed by the TAGA for the targeted compounds.
Results and Discussion
This investigation had three goals: 1) identify off site plumes using target compounds; 2)
329
-------
attempt to locate the sources of plumes, using meteorological data collected concurrently with sampling;
and 3) collect whole-air samples using Summa canisters for target and non-target compound analysis
at a later date using conventional methods.
Sampling was conducted during periods when it was believed that the meteorological conditions
would produce an atmospheric inversion, resulting in observation of maximum concentrations of target
compounds. These sampling periods were:
Sampling Period I August 8, 1989 04:29 - 07:42
Sampling Period II August 8, 1989 20:17 - 23:52
Sampling Period III August 9, 1989 12:34 - 16:55
Sampling Period IV August 10-11, 1989 23:40 - 03:19
Four distinct plumes were observed off site. One contained the target compounds benzene,
toluene, and chlorobenzene. This plume is believed to have originated from the Title III facility.
Another plume contained the target compound 1,1,1-trichloroethane. This plume is believed to have
originated from an electrical part manufacturing plant. The third plume contained methylene chloride
and is believed to have originated from a parcel delivery service. The fourth plume is suspected to
have originated from a graphic arts building. The maximum concentration observed off site for the
target compounds were:
Benzene 300 ppb
Chlorobenzene 30 ppb
Methylene Chloride 120 ppb
1,1,1-Trichloroethane 1100 ppb
Toluene 180 ppb
Xylene 80 ppb
1,4-Dioxane 18J ppb
No 1,2-dichloroethane was detected off site. The J associated with the 1,4-dioxane
concentration denotes that the value is above the detection limit but below its quantitation limit. The
maximum concentration observed on-site for the target compounds were:
Benzene 5500 ppb
Chlorobenzene 350 ppb
1,2-Dichloroethane 70 ppb
Toluene 1700 ppb
No methylene chloride, 1,4-dioxane, 1,1,1-trichloroethane, or xylene was detected on-site.
REAC provided a meteorological station for the examination of the micro-meteorology at
various locations near the Title III facility. There was considerable variation between the REAC data,
generated off site, and the on-site data available from the facility. The differences in the
meteorological conditions were believed attributable to local topographical features, which allowed
channelling and eddy formation. The meteorological data generated at the facility was used to correlate
the wind conditions and locations of the source of the plumes.
Summa canister samples were taken in several plumes so analyses for targeted and non-targeted
compounds could be performed by conventional GC/MS methods. The canisters were also analyzed
by the TAGA 6000E. The results of these analyses are located in Table II. The results between the
two techniques showed good agreement for the target compounds, particularly when concentrations
were elevated.
Conclusion
The goals of this investigation were successfully met. The TAGA 6000E identified off site
plumes using target compounds; the sources of these plumes were located using meteorological data
collected concurrently with the sampling; and the TAGA 6000E located plumes so whole-air samples
could be collected in Summa canisters for target and non-target compounds analyses using conventional
330
-------
GC/MS methodology. Furthermore, the Summa canister analysis for target compounds by both the
GC/MS and the TAGA showed good agreement.
TABLE I. TARGET COMPOUND LIST
benzene*
chlorobenzene*
1,2-dichloroethane*
1,4-dioxane
methylene chloride*
1,1,1-trichloroethane
toluene*
xylene
* Identified from facility's Title III report.
331
-------
TABLE II SUMMA CANISTER ANALYTICAL RESULTS (concentrations in ppb)
COMPOUND
1,2-dichloroethane
benzene
methylene chloride
1,4-dioxane
toluene
1,1,1-trichloroethane
xylene
chlorobenzene
SUMMA #41
GC/MS TAGA
SUMMA #69
GC/MS TAGA
ND
QL=10
ND
ND
QL=10
QL=10
QL=10
ND
DL=4
DL=6
DL=28
DL=6
DL=13
DL=4
DL=8
DL=2
ND
95
ND
ND
18
QL=10
OL=10
9J
7J
74
DL=28
10J
22J
9J
10J
6
SUMMA #67
GC/MS TAGA
ND
QL=10
ND
ND
QL=10
39
QL=10
ND
6J
DL=6
30J
8J
DL=13
42
9J
DL=2
COMPOUND
1,2-dichloroethane
benzene
methylene chloride
1,4-dioxane
toluene
1,1,1-trichloroethane
xylene
chlorobenzene
SUMMA #52
GC/MS TAGA
ND
QL=10
49
ND
QL=10
ND
QL=10
ND
6J
DL=6
71J
8J
DL=13
5J
10J
DL=2
SUMMA #1
GC/MS TAGA
ND
117
QL=8
ND
154
QL=8
ND
13
6J
139
DL=28
DL=6
197
5J
DL=8
17
SUMMA #48
GC/MS TAGA
ND
235
QL=10
ND
92
QL=10
QL=10
27
6J
229
DL=28
8J
86
5J
9J
19
COMPOUND
1,2-dichloroethane
benzene
methylene chloride
1,4-dioxane
toluene
1,1,1 -trichloroethane
xylene
chlorobenzene
SUMMA #12
GC/MS GC/MS TAGA
(DUP)
ND
20
QL=10
ND
18
ND
QL=10
3J
ND
20
QL-10
ND
17
ND
OL=10
ND
6
22
35J
8J
26J
9J
9J
4J
SUMMA #58
GC/MS GC/MS TAGA
(DUP)
ND
QL=10
QL=10
ND
QL=10
QL=10
QL=10
ND
ND
QL=10
QL=10
ND
QL=10
QL=10
QL=10
ND
6J
DL=6
32J
10J
14J
10J
10J
DL=2
ND = Not detected
DL = Detection Limits
J = Value between Detection Limits and Quantitation Limits
QL = Quantitation limits
DUP = Duplicate
332
-------
TAGA 6000E TANDEM TRIPLE QUADRUPOLE MASS SPECTROMETER
CALIBRATION
INLET
CAD GAS
co
co
CRrOGENH
VACUUM PUMP
CEM
(pulse counting]
Q3
US EPA ENVIRONMENTAL RESPONSE TEAM
RESPONSE ENGINEERING AND ANALYTICAL CONTRACT
68-03-3482
Figure 1. The TAGA 6000E Mass Spectrometer.
-------
CO
CO
ORTH HAVEN, CONNECTICUT
US EPA ENVIRONMENTAL RESPONSE TEAM
RESPONSE ENGINEERING AND ANALYTICAL CONTRACT
68-03-3482
Figure 2, A Route Traversed To Perform Air Monitoring.
-------
70
BENZENE CUPJ0153
6D -
50 -
40 -
3D -
20 -
10 -
0
BCD EFGH IJK L HHOPQR ST IT VI
0.00 2.06 4.14 6.21 6.29 1036 12.44 14.51 16 60 18.67 20.75 22.82 24.90 2697
METHYLENE CHLORIDE CUPJ015}
L MNOPQR ST
D.OO 2.D6 4,14 6.21 8.29 10.36 12.44 14.51 16.60 18.67 20.75 22.82 24.90 26,97
1,1,1-TRICHLOROETHANE
300 -
250 -
2QD -
150 -
100 -
50 -
iBCDZFGH I J K 1 BNOPOR ST U VI
A
. . _ J
\
0.00 -Z.D5 1.14 6.21 B.29 10.36 12.44 14.51 16.60 18.67 20.75 22.82 24.90 26.97
CHLOROBENZENE CUPJ015}
BC D E F G H IJK L tlBOPQR ST U VI
0,00 2.06 4,14 6.21 8.29 10.36 12.44 14.51 16.60 18.67 20.75 22.82 24.90 26,97
TIKE (Bin)
Figure 3. Ion Profiles For Target Compounds.
335
-------
DEVELOPMENT OF A "STATEMENT-OF-WORK (SOW)
FOR THE ANALYSIS OF AIR TOXICS AT SUPERFUND SITES"
AS PART OF THE CONTRACT LABORATORY PROGRAM (CLP)
Russell McCallister
U.S. Environmental Protection Agency
Office of Emergency and Remedial Response (OERR)
Hazardous Site Evaluation Division (HSED)
Analytical Operations Branch (AOB)
Washington, DC 20460
Emile I. Boulos
U.S. Environmental Protection Agency
Office of Emergency and Remedial Response (OERR)
Hazardous Site Evaluation Division (HSED)
Analytical Operations Branch (AOB)
Washington, DC 20460
William T. "Jerry" Winberry
Engineering-Science, Inc.
Gary, NC 27513
Linda Forehand
Engineering-Science, Inc.
Gary, NC 27513
ABSTRACT
As specified in the Comprehensive Environmental Response, Compensation
and Liability Act of 1980 (CERCLA) and the Superfund Amendments and
Reauthorization Act of 1986 (SARA), EPA has the responsibility for assessing the
potential of air emissions and air quality impacts prior to and during Superfund
hazardous waste site cleanup. CERCLA and SARA mandate the characterization
of all contaminant migration pathways from waste to the environment and of the
resulting environmental impacts. Specifically, they mandate that "all potential
migration pathways for contaminants" be characterized. Heretofore, the U.S.
Environmental Protection Agency (EPA) has developed specific Statement-of-Work
(SOWs) to provide laboratories with specific analytical techniques in the analysis of
both organic and inorganic analytes in soil, sediment and water as an integral part of
the Contract Laboratory Program (CLP). In addressing the regulatory initiatives for
assessing all pathways, the Analytical Operation Branch (AOB) of USEPA has
initiated a project to develop a "Statement-of-Work (SOW) for the Analysis of Air
Toxics at Superfund Sites" as part of the CLP. The objective of this paper is to
outline the development of the SOW and briefly describe the analytical
methodology for analysis of organics and inorganic air toxics.
336
-------
INTRODUCTION
As specified in the Comprehensive Environmental Response, Compensation
and Liability Act of 1980 (CERCLA) and the Superfund Amendments and
Reauthorization Act of 1986 (SARA), the U.S. Environmental Protection Agency
(EPA) has the responsibility for assessing the potential for air emissions and air
quality impacts prior to and during Superfund hazardous waste site cleanup.
CERCLA and SARA mandate the characterization of all contaminant migration
pathways from waste to the environment and of the resulting environmental
impacts. Both CERCLA and SARA regulations were developed within the Office
of Emergency and Remedial Response (OERR).
Within the Hazardous Site Evaluation Division (HSED) of OERR is the
Analytical Operation Branch (AOB), which administers the Contract Laboratory
Program (CLP). The CLP provides analytical support service to Superfund and
other EPA programs. The current SOWs address standardized analytical
procedures for approximately 23 inorganic and 100 or more organic chemicals
comprising the CLP hazardous substances list (HSL).
Within the CLP, the contract laboratory must adhere to defined methodologies
associated with quality assurance/quality control (QA/QC), contract required
detection limits (CRDL), reporting requirements and timeliness of data reporting as
directed by the SOW. Present SOWs address organic and inorganic constituents in
multi-media and at varying concentration levels. SOWs have been used in the CLP
since its conception in 1980.
While the agency has numerous SOWs for soil and water, there was no guidance
available for air toxics analysis of samples acquired during Superfund investigations.
In order to provide technical support to EPA Regional Remedial Program
Managers (RPMs) and Environmental Program Managers (EPMs) utilizing the
CLP, AOB has developed a "Statement-of-Work for Analysis of Air Toxics at
Superfund Sites" to be used as part of the CLP. This paper will outline its
development in becoming a part of the CLP.
BACKGROUND
The development of the air toxics SOW involved several tasks. These tasks
involved:
• Task 1 - Survey Involving Air Toxics Identification for Inclusion into a
Target Compound List (TCL)
• Task 2 - Selection of Analytical Methodology Associated with TCL
• Task 3 - Preliminary "Statement-of-Work for Analysis of Air Toxics at
Superfund Sites"
• Task 4 - Preliminary Multi-Lab Methods Validation for Primary Air Toxic
Target Compound List
• Task 5 - Final Multi-Lab Methods Validation for Primary Air Toxic Target
Compound List
337
-------
• Task 6 - Final "Statement-of-Work for Analysis of Air Toxics at Superfund
Sites"
To insure that Regional, state and commercial laboratory concerns were
addressed AOB formed an Air Toxics Workgroup Committee (ATWC). The
ATWC provided input and direction to the TCL (Task 1) and the selection of
analytical methodology (Task 2) for those on the TCL.
The TCL consist of approximately 257 target compounds. Of the 257
compounds, 43% are volatiles thus having vapor pressure greater than 0.1 mm Hg.
Within the volatile group, there are two basic sampling options: canister sampling
and adsorbent methodology. A closer examination of the TCL illustrates that while
most volatiles can be sampled by both canister and adsorbent, there are some
volatiles which are amenable to one or the other methodology. Similarly, the
detection limits are vary between the methods for specific volatiles on the TCL.
Approximately 32.4% of the TCL are classified as semi-volatiles with vapor
pressures ranging from 101 to 106 mm Hg. Historically, the sampling methodology
involves a filter combined with a combination polyurethane foam and XAD-2 plug
to retain the semi-volatiles.
Metals comprise approximately 28% of the TCL. The sampling techniques
normally involve filtration.
SELECTION OF ANALYTICAL METHODOLOGY AS PART OF THE AIR
TOXICS SOW
The selection of the proper analysis method for an analyte is dependent on
many important interrelated factors. These include the compound or compounds of
interest, the level of detection required, the degree of selectivity needed, and the
purpose of the data collected. Other factors which may be as important as the
above are cost, the accuracy and precision required, and the number of samples to
be sampled and analyzed.
In general, analytical methods selection were reviewed for inclusion into the
Statement-of-Work from the following analytical publications:
• "Compendium of Methods for the Determination of Toxic Organic
Compounds in Ambient Air,"
• "Ambient Air Monitoring at Superfund Sites,"
• SW-846 entitled: "Test Methods for Evaluation of Solid Waste," and
• Methods for Organic Chemical Analysis of Municipal and Industrial
Wastewater.
Based upon the survey results, committee input and best available analytical
techniques, the following analytical methodology, as outlined in Table 1.0, was
selected to address groups of air toxics found on the TCL.
338
-------
Table 1.0
SOW Analytical Methodology
Classification
Volatiles
SOW
Method
Method 1.1
Method 1.2
Semi-volatiles,
Including
Pesticides and
PCBs
Metals
Method 2.0
Sampling
Technique
TenaxR
Adsorbent
(Reference
Method)
Sample
Conditioning
Thermal
Desorption,
Cryogenic
Trapping and
Focusing
Analysis
Technique
GC/MS/SCAN
Other Adsorbent
(Equivalent
Method)
SUMMA*
Canister
Nafion Dryer, GC/MS/SCAN
Cryogenic
Trapping
(Reference
Method)
Modified
Water Purge,
Adsorbent Trap,
Then Thermal
Desorption
(Equivalent
Method)
Filter
Followed by
PUF/XAD-2
Adsorbent Trap
Using Hi-
Volume Sampler Gel Clean-up
10% Ether/ GC/MS/SCAN
hexane
Soxhlet
Extraction,
Silica
Method 3.0 Filter
Microwave
Extraction
Using HN03/
HC1 Acid
Solution
ICAP
SOW ANALYTICAL METHODOLOGY DESCRIPTION
• Method 1.1 - Volatile Organics Utilizing Solid Adsorbent
Conventional methods for VOC determination have relied on solid adsorbent
techniques such as carbon and, more recently, Tenax". For example, the U.S.
Department of Health and Human Services, National Institute for Occupational
Safety and Health base many of their sampling procedures on the use of carbon
adsorption techniques. As with many solid adsorbents, there are many limitations to
339
-------
the use of TenaxR. The more significant problems in utilizing TenaxR and other solid
adsorbents are:
• Artifact formation,
• Contamination,
• Analyte breakthrough, and
• Solvent extraction
Although sorbent techniques demonstrate problems, several advantages can be
gained through their use. First, integrated sampling over a period of 8 to 12 hours is
easily performed. Because of the small size and portability of the sample tubes and
pumps, they are easily located in many sampling application. Finally, multi-bed
adsorbents can be used to sample different groups of volatiles, thus making their
utility a major function of a monitoring protocol.
While there is reliable data to support a multibed tube, the reference method
(RM) procedure utilizes TenaxR for the adsorptive media. This selection was based
primarily on the years of experience the USEPA has on characterizing its
weaknesses and strengths.
Sampling involves drawing ambient air through an adsorbent cartridge
containing approximately 1-2 grams of TenaxR. While highly volatile organic
compounds and most inorganic atmospheric constituents pass through the cartridge,
certain organic compounds are trapped on the resin bed.
After the organics are trapped on the resin bed, the cartridge is tagged and
transported back to the lab for analysis.
Upon receipt at the laboratory, the cartridge is logged into the lab book and the
chain-of-custody form completed. The cartridges are stored under refrigeration
until analysis.
As illustrated in Figure 1.0, the cartridge is submitted for analysis by GC/MS.
During analysis, the cartridge is removed from the refrigerator, an internal standard
is added to permit quantitative analysis, and the organics trapped on the TenaxR are
thermally desorbed. The organic vapors are removed from the TenaxR by heating
the sample cartridge to 275 C under a flow of helium. The desorbed vapors are
collected in a cryogenic trap which is cooled to liquid nitrogen temperature. The
use of the cryogenic trap allows the carrier gas flow, needed for the GC/MS, to be
balanced.
The cryogenic irap containing the organics is then heated to transfer the sample
to the head of the capillary GC column which is cooled to liquid nitrogen
temperatures. T.iis step is essential to focus the organic compounds and allow their
application to the head of the capillary column in a discrete band.
The scan of the mass spectrometer is initiated and the analytical procedure is
begun. Under a flow of helium, the GC column is programmed to a temperature to
allow the elution of all of the organic compounds while the mass spectrometer is
scanning. Data are recorded by the computer for subsequent processing.
Quantitation is performed by the method of relative response factors, where the
340
-------
proportionate system responses for analyte and standard are determined prior to the
analysis of the sample and this relative system response is used to determine the
quantity of compound present on the sample cartridge.
Component identification is normally accomplished, using a library search
routine, on the basis of the GC retention time and mass spectral characteristics for a
single ion for each target compound.
The quantitative analysis is performed by a combination of manual and
computerized procedures: the computer is instructed to seek characteristic ions in a
previously determined retention window. At this, point the operator intervenes to
determine if the compound of interest has been located correctly. If the compound
identification is correct, the computer then performs the quantitative calculation
using the method of relative response factors. Data are reported as ng/sample, and
can be subsequently converted to whatever units are desired.
While the RM outlines the use of TenaxR as the adsorbent, the Statement-of-
Work (SOW) will allow alternative adsorbent, i.e., multibed adsorbents, as
equivalent methods. The Contract Laboratory must demonstrate accuracy and
precision limits as well as sorbent characteristic data for selected analytes inorder
for other adsorbents to be classified as equivalent methods. In addition, the analysis
of Performance Evaluation (PE) samples must meet the accuracy and precision
limits. These limits are:
• Accuracy ± 30%
• Replicate Precision ± 30%
The accuracy and replicate precision limits were based upon data acquired over
the last five (5) years from the Toxics Air Monitoring System (TAMS) utilizing the
TenaxR solid adsorbent for monitoring volatile organic compounds.
• Method 1.2 - Volatile Organics Utilizing Canister Methodology
Both subatmospheric pressure and pressurized sampling modes use an initially
evacuated canister and a pump-ventilated sample line during sample collection.
Pressurized sampling requires an additional pump to provide positive pressure to
the sample canister. A sample of air is drawn through a sampling train comprised of
components that regulate the rate and duration of sampling into a preevacuated
SUMMAR passivated canister.
After the air sampled is collected, the canister valve is closed, an identification
tag is attached to the canister, and the canister is transported to a predetermined
laboratory for analysis.
Upon receipt at the laboratory, the canister tag data is recorded and the
canister is attached to the analytical system. During analysis, water vapor is reduced
in the gas stream by NafionR dryer (if applicable), and the VOCs are then
concentrated by collection in a cryogenically-cooled trap. The VOCs originally
collected in the trap are revolatilized, separated on a GC column, then detected by
one or more detectors for identification and quantitation.
The Reference Method (RM) involves using a micro-bore capillary column
coupled to a mass spectrometer (MS) operated in the scan mode. The use of the
GC-MS-SCAN detector system is recommended for positive identification and
341
-------
quantitation of targeted compounds to assure that high-quality data is acquired.
This configuration, as outlined in Figure 2.0 and Figure 3.0, will allow for the
identification of target and non-target compounds. This methodology will be the
reference method.
While the above procedure outlines the use of a PermapureR dryer to remove
moisture prior to analysis by the GC/MS system, the SOW will allow alternatives to
the NafionR dryer/cryogenic trapping sample concentration step, if they can
demonstrate meeting performance criteria by analysis of PE samples. Based upon
historical data from the Urban Air Toxics Program (UATP) and Toxics Air
Monitoring (TAM) monitoring systems operated by USEPA, the PE limits are:
• Accuracy: ± 30%
• Replicate Precision: ±25%
These methods will be classified as equivalent methods. The SOW addresses the
use of a modified purge-and-trap technique, as outlined in SW-846, Method 5030
entitled: "Organic Analysis-Purge-and-Trap Preparation and Extraction."
In one modification of the purge-and-trap technique (see Figure 4.0), the
canister is evacuated to a large cryogenic trap at a rate of -~ 400 mL/min for a
period of 10 minutes. The larger trap allows the analyst to acquire a larger sample
for analysis to meet required detection limits. Once trapped, the gas stream is
bubbled through a 1 mL water solution (to remove excess moisture) where the
volatiles are efficiently removed from the aqueous phase to the vapor phase. The
vapor is swept onto an sorbent column (analytical trap) where the volatile
compounds are adsorbed. The analytical trap can be a combination of different
sprbents (multi-bed trap). The analytical trap is backflushed for a short period of
time (~2 minutes) to remove moisture. After purging is completed, the analytical
trap is heated and flushed with inert gas to desorb the volatile organics onto the
head of the GC column where they are cryofocused to improve peak shape and
separation. After cryofocusing, the organics are separated on the GC column and
identified and quantified by the MS.
Another modification to EPA methodology would allow the use of an analytical
trap without the cryogenic trap, as illustrated in Figure 5.0.
The above modifications can all be used in the SOW as long as the analytical
laboratory can meet Performance Specifications of precision and accuracy for PE
samples. Figure 6.0 illustrates the various configurations allowed under the SOW
equivalency program.
• Method 2 - Semi-volatile, Pesticides and PCB by Combination Filter/PUF/
XAD-2 Adsorbent Cartridge
Semi-volatiles, pesticides and PCBs have received increased attention in recent
years at Superfund sites because some of these compounds are highly carcinogenic
or mutagenic. In particular, benzo[a]pyrene (B[a]P) has been identified as being
highly carcinogenic. To evaluate the extent of human exposure to B[a]P and other
semi-volatiles, reliable sampling and analytical methods are necessary.
Traditionally, a filter has been used to collect the semi-volatiles while
polyurethane foam has been used to collect the pesticides and PCBs: Recent
studies have indicated that during sampling non-volatile PAHs (vapor pressure
342
-------
< 10"* mm Hg), PCBs and pesticides may be trapped on the filter or adsorbent, but
post-collection volatilization problems may distribute them downstream of the filter.
To address this weakness, a combination filter followed by an adsorbent has been
proposed as part of the Reference Method (RM).
A wide variety of adsorbents such as TenaxR, XAD-2 and polyurethane foam
(PUF) have been used to sample B[a]P and other PAH, pesticides and PCBs. All
adsorbents have demonstrated high collection efficiency for B[a]P in particular. In
general, XAD-2 resin has a higher collection efficiency for volatile PAHs than PUF,
as well as a higher retention efficiency. However, PUF cartridges are easier to
handle in the field and maintain better flow characteristics during sampling.
Likewise, PUF has demonstrated its capability in sampling prganochlorine
pesticides, polychlorinated biphenyls and polychlorinated dibenzo-p-dioxins.
However, PUF has demonstrated a lower recovery efficiency and storage capability
for naphthalene and B[a]P, respectively, than XAD-2. There have been no
significant losses of PAHs up to 30 days of storage at room temperature (23 C) using
XAD-2. It also appears that XAD-2 resin has a higher collection efficiency for
volatile PAHs than PUF, as well as a higher retention efficiency or both volatile and
reactive PAHs. Consequently, while the literature cites weaknesses and strengths of
using either XAD-2 or PUF, the RM recommends the sampling system to consist of
a filter followed by a combination PUF/XAD-2 adsorbent. Inorder to reach needed
detection limits, a high volume air sampler (10 cfm) has been recommended.
The analytical technique is a modification of EPA Test Method 610 and 625,
Methods for Organic Chemical Analysis of Municipal and Industrial Wastewater;
Methods 8000, 8270, and 8310, Test Methods for Evaluation of Solid Waste; and
Method 680, Determination of Pesticides and PCBs in Water and Soil/Sediment by
Gas Chromatography/Mass Spectrometry. The choice of the gas chromatography
coupled with mass spectroscopy was influenced by their sensitivity and selectivity,
along with their ability to analyze complex samples.
Filters and adsorbent (PUF/XAD-2) cartridges are cleaned in solvents and
vacuum dried. The filters and adsorbent cartridges are stored in screw-capped jars
wrapped in aluminum foil (or otherwise protected from light) before careful
installation on the sampler.
Approximately 360 m3 of air is drawn through the filter and combination
adsorbent cartridge.
The amount of air sampled through the filter and adsorbent cartridge is
recorded, and the filter and cartridge are placed in an appropriately labeled
container and shipped along with blank filter and adsorbent cartridges to the
analytical laboratory for analysis.
The filters and adsorbent cartridge are extracted by Soxhlet extraction using
10% ether/hexane. The extract is concentrated by Kuderna-Danish (K-D)
evaporator, followed by silica gel cleanup using column chromatography to remove
potential interferences prior to analysis by GC/MS, as illustrated in Figure 8.0.
The eluent is further concentrated by evaporation, then analyzed by GC-MS
detection, as illustrated in Figure 7.0. The analytical system is verified to be
operating properly and calibrated with five concentration calibration solutions using
the internal standard approach. The mass spectrometer is operated in the SCAN
mode so both target and tentatively identified compounds (TICs) can be addressed.
343
-------
• Method 3 - Metals
Paniculate metals are defined as those elements that can be collected by
standard filter media at ambient temperatures. Sampling methods for metal
particulate are generally based on capture of the paniculate on filter media. For
the most part, glass fiber filters are used; however, organic and membrane filters
such as cellulose ester and Teflon have also been used. The membrane filters
demonstrate greater uniformity of pore size and, in many cases, lower background
levels of trace metals than are found in glass fiber filters.
The basic particulate air sampler is the high volume sampler which collects a
nominal m3 sample over a 24-hour period and captures particulate on an 8 x 10 inch
filter (glass fiber). The nominal cut point is 100 jim for the maximum diameter
particle size capture. A size-selective inlet can be placed ahead of the filter to
separate respirable and nonrespirable particulate matter.
After collection, the filter can be stored in a manila folder and a protective
envelope. The filter should be folded upon itself with the loading folded inside.
Care should be taken so as not to damage or touch the filter and not to dislodge the
loading.
The extraction procedure involves cutting a 1" x 8" strip from the Hi-Vol filter
using a template as described in the Federal Reference Method entitled:
"Determination of Lead in Suspended Particulate Matter Collected From Ambient
Air." The strip is extracted by a microwave extraction procedure with a HNO3/HC1
acid solution for 23 minutes and analyzed by ICP, as illustrated in Figure 8.0,
The analytical system utilizes the inductively coupled argon plasma
spectrometer (ICAP). ICAP enables multielement analysis (10-40 elements per
minute) thus lending itself to the multielement requirements of the Contract
Laboratory Program (CLP).
CONCLUSIONS
In response to CERCLA and SARA mandate programs to characterize
potential air emissions and air quality impacts of air toxics through the air pathways,
the USEPA/AOB has developed a SOW for the analysis of air toxics from
Superfund sites to be used in the CLP. The SOW was developed through the
guidance of the air toxics workgroup committee to insure that the needs of regional,
state and local air pollution programs were addressed.
DISCLAIMER
Although the research described in this article has been funded wholly or in
part by the United States Environmental Protection Agency through Contract 68-02-
4398 to Engineering-Science, it has not been subjected to Agency review and
therefore does not necessarily reflect the views of the Agency, and no official
endorsement should be inferred. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
344
-------
/
TENAX
TUBE
2
I
K
f
/
THERMAL
DESORPTION
/
i
Y
/
CRYOGENIC
TRAPPING
t
^
S,
1(>
£
t>
Chalri-ol
Custody Form
Enltr Into Logbook
Inttrnal Standard
(PFT-FIB)
Background Carllflcatlon
Hilium Flaw
370° C Mil. T»mp.
B mlnutts
/ £
MASS
SPECTROMETRY
b
-J
/
GAS
CHROMATOGRAPHY
Axs[gnm«nt of Mass
Numb.r with BFB
Tuning with BFB
P.liClvt Abundincl
p»l*rmlnatlon with
p«r!luoroto!u»n§ (PFT)
Inlirnil Standard
of fluoro-Z-1odob«nz>n«
(FIB)
Column OV-1 Capillary
Program 4 C/mln to
lo 210 °C
Htllum Carrlir Ga>
R*sponai Factor
Evaluation
Liquid Nllrogin
@ - 1SO°C
Artir inpplng,
rapid dialing
to Z7tP C
/~~ /:
CRYOGENIC jg
FOCUSING Sf
X
Liquid Nltrogan
@ -1SO° C lor
49 itconds
Figure 1.0. Strategy For Analyzing Volatile Organic Compounds Utilizing Gas
Chromatography Coupled to A Mass Spectrometer from TenaxR Adsorbent Tubes.
Pressure
Regulator
Vent
Nafion
Dryer
Optional
Pressure
Gauge
Exhaust
Vent
... c*, (Excess)
Mass Flow *
Co rrt roller
«-Dry
Forced
Airln
f
~^
uj
6-Port
Chrexnatcigraphic
Vafve
Cryogenic
Trapping
Unit
Tee
Connection
OV-1
Capillary
Column
' {0.32mm x 50m)
Mass Spectrometer
In SCAN Mode
Figure 2.0. Strategy for Analyzing Volatile Organic Compounds Utilizing
SCAN From SUMMAR Passivated Canisters As Reference Method 1.2.
GC-MS-
345
-------
Canister
P*PI ^
Nation9 Dryer
• Initial Pressure Checked • Automatic Dry<
Clean-up
Cryogenic
Trap
Pressurized to 20 pslg
• Sampled at 20 m Urn In
for 10 Minutes
Gas
Chromalography
OV-1 Capillary Column
(50-m X 0.31 ID)
Program at 8'C/mln
to 150*C
Helium Carrier Gas
Mass
Spectroscopy
Liquid Nitrogen
@ - 150.0*C
After Trapping,
Rapid Heating
to 120*C In
60 Seconds
Assignment ot Mass
Number with PFTPA
Tuning with PFTPA
Relative Response
Factors
• MS-SIM for 40 Analytes
• Static/Dynamic
Calibration
Figure 3.0. Reference Method 1.2 Analytical Method For Trace Organics In Gas
Samples Collected By Canister.
Canister
• Sampled @
400 mUmlnute
lor 10 Minutes
Gas
Chromatograpny
Cryolocuslng with
Liquid Nitrogen
to -150°C
Vocol Megabore
Column
(60 m x 0.75 mm ID)
Analytical Trap
• Carbopak B/
Carbosleve S-lll
• Helium Purge to
Remove Moisture
• Desorptlon at 22o°C
lor 3 Minutes
Mass
Spectroscopy
Cryogenic
Trap
• Quartz Trap
• Liquid Argon
Water Purge
• Assignment ot Mass
Number
• Tuning
• Relative Response
Factors
• MS-SCAN for 55 Analytes
• Static/Dynamic
Calibration
Figure 4.0. Equivalent Analytical Method For Trace Organics In Gas Samples
Collected By Canister As A Modification To Reference Method 1.2.
346
-------
-Internal STO Added
Water Purge
Canister
Initial Pressure
Checked
Canister Pressurized
to -3D pslb
Gas
Chromatography
Cryofocuslng
to -50°C
Analytical Trap
Carbopak B/
Carbosteve S-lll
30 mUmln for
10 minutes
Mass
Speclroscopy
• Assignment ol Mass
Number with BFB
• Tuning with PFK
• Relative Response
Factors
• MS-SCAN lor 63 Analytes
• Static/Dynamic
Calibration
• -1 ppb Limits of Detection
Figure 5.0. Equivalent Analytical Method For Trace Organics In Gas Samples
Collected By Canister As An Alternative To Reference Method 1.2.
Canister
Mass
Spectroscopy
• SIM Mode
• SCAN Mode
• Ion Trap MS
• Quadrupole MS
• Other MS Types
Sample
Conditioner
• Pressure Adjustment
• Perma-Pure Dryer
• Add Surrogates
and Standards
Gas
Chromatography
Column
• Fused Silica
Cap. Col.
• Packed Column
• Two-D
Chromatography
Data Analysis
Preconcentrator
• Auto Cryogen Trap
• LAr In Flask
• Solid Sorbent Tube
Sample
Conditioner
• Cryofocuslng
— Separate Assembly
— Head of Column
• Water Purge to
S. Sorb.
• He Purge of S. Sorb
Figure 6.0. Generic Equivalent Analytical Method For Trace Organics In Gas
Samples Collected By Canisters As An Alternative To Reference Method 1.2.
347
-------
Filter
Adsorbent
PUF/XAD-2
Surrogate Standard
Addition for
GC/MS Analysis
Soxhlel Extraction with
10% Ether/Hexane
(18 Hours/3 Cycles/Hr)
Drying with Anhydrous
Sodium Sulfate
T
Hexane Rinse
Hexane Rinse
Kuderna-Danlsh (K-D)
Evaporator Attached with
Macro Synder Column
Water Bath at 50°C
Volume Adjusted
to 10 mL
V
Concentrate to 1 mL
with Nitrogen Slowdown
GC/MS Analysis
Figure 7.0. Strategy For Analyzing Semi-Volatiles and Pesticides Utilizing
Combination Filter/Adsorbent (PUF/XAD-2 Combination) Followed By Gas
Chromatography Separation With Mass Spectroscopy Identification.
Microwave Extraction
Using HNO-j/HCL
Acid Solution
ICP - Saaipled Introduced
by a Nebulizer Through
Center of a Plasma
Torrid (7000K)
Detector (photoraultiplier tubes) - Observe
Spectrum of Elements
Spectrometer - Resolves the
Observed Spectrum
Processor - Calculates Observed
Elemental Concentrations
Figure 8.0. Strategy For Analyzing Metals On Glass Fiber Filters Utilizing
Inductively Coupled Argon Plasma Emission Spectroscopy
348
-------
A PROGRAM TO VALIDATE AIR MONITORING METHODS
FOR THE SUPERFUND PROGRAM
W. J. Mitchell, L. J. Smith, H. L. Crist, and J. C, Suggs
U. S. Environmental Protection Agency
Atmospheric Research and Exposure Assessment Laboratory
Quality Assurance Division
Research Triangle Park, NC 27711
EPA plans to accelerate its remediation and closure activities at Superfund
sites and is committed to monitoring around these sites after closure to ensure
that the remediation was successful. The increased activities at Superfund sites
will increase the chances that toxic gases and contaminated soil particles
(fugitive emissions) will be released to the atmosphere. Thus, it will be
necessary to monitor the air over and around the sites using reliable methods
to assure the site workers and the general population that they are not in any
clanger. Reliable methods will also be needed to ensure that the site remains
harmless after closure. This paper describes EPA's efforts to validate its draft
e.ir monitoring methods for measuring the particle, semi-volatile organic, toxic
metal and volatile organic concentrations over and around Superfund sites. The
first phase of this effort will determine the precision and accuracy of the
Analytical portion of these methods through round robin testing. The experi-
mental design, the types of samples, the analytes and the concentrations selected
and the basis for their selection are described. The results obtained to date
are also described.
INTRODUCTION
As Russ McAllister of our Office of Emergency and Remedial Response (OERR)
discussed earlier today, EPA will establish a Contract Laboratory Program (CLP)
for air samples collected over and around Superfund sites. In many ways this
CLP program will resemble the current CLP program for soil and water samples.
For example, each CLP laboratory must use an OERR-approved analytical method.
Also, the analytical equipment and the analytes will be similar. There will be
one significant difference, however. In the air sample CLP program the labo-
ratory that will analyze the samples will provide the sample collection medium
to the contractor who will collect the sample.
OERR is now developing four analytical methods for the air sample CLP
program. OERR has a corresponding effort underway to develop air sampling
nethods compatible with these analytical methods. The four analytical methods
and the material used to collect the air sample are:
(1) VOC (volatile organic compounds) by Canister -- VOC's collected in
a stainless steel 6-L canister, released into a cryogenically cooled cold
trap for concentration and then quantitated by GC/MS;
349
-------
(2) VOC by Tenax -- VOC collected on Tenax, thermally desorbed from the
Tenax into a cryogenically cooled trap for concentration and then quanti-
tated by GC/MS;
(3) Semi-VOC (semi-volatile organics) -- Semi-VOC collected by pulling
air through a cartridge containing a filter, polyurethane foam and XAD-
2. Semi-VOC's Soxhlet extracted, extract concentrated and semi-VOC's then
quantitated by GC/MS. (Note: Semi-VOC's include pesticides and poly-
chlorinated biphenyls.)
(4) Metals by Hi-Vol -- Metals collected on a quartz or glass-fiber
filter, filter extracted with a microwave acid extraction and analyzed by
ICP (preferred), FAA, AA or graphite furnace AA.
EPA's Atmospheric Research and Exposure Assessment Laboratory (AREAL) has
agreed to validate the draft OERR analytical methods to establish for each
analyte:
(1) the precision and accuracy (bias) that can be expected under routine
use by a qualified, experienced laboratory;
(2) the clarity and applicability of the draft analytical method;
(3) significant sources of error so that appropriate QC procedures can
be included in the final draft of the method; and
(4) target qualifying goals for OERR to use when selecting laboratories
for the CLP program.
AREAL will: design the study; prepare, certify and distribute the samples;
statistically analyze and reduce the data; and prepare a report on the results
achieved. This paper describes how AREAL plans to conduct the validation study.
STUDY DESIGN
Six to ten laboratories will use the draft OERR analytical methods to
analyze synthetic samples that closely simulate air samples that CLP laboratories
will routinely analyze. Each laboratory will receive the same type and quantity
of samples to analyze and report the results to AREAL using an AREAL-supplied
form.
SELECTION OF PARTICIPANTS
Only laboratories that have the required analytical equipment on hand and
have experience with the analytical method, or one very similar to it, will be
considered. These restrictions are required for the validation study to provide
estimates of the analytical method's performance in routine use. OERR's Sample
Management Office (SMO) has already identified 15 to 20 qualified laboratories
who are interested in participating. However, SMO cannot select the six to ten
participating laboratories until the analytical methods become available later
this month. Participants will be selected based on their competence, on their
ability to analyze the samples in the 7 to 21 days specified and on the cost to
OERR for them to participate.
350
-------
SELECTION OF ANALYTES
The following approach was used to select the analytes. First, the OERR-
ranked target compound list for the proposed CLP program was examined and the
analytes OERR considered most important were identified. Then, an OERR data base
that lists the number of times each proposed analyte was found in soil, water
and waste samples from Superfund sites was examined. This was done to ensure
that all analytes frequently encountered had been considered. (It is expected
that the analytes in the soil, water and waste at a Superfund site will be the
same as those that will likely be found in air samples from the site.) These
two lists were compared to each other and to the inventory of available reference
materials for the proposed analytes. The final list of analytes was then
selected based on the above information and on AREAL's experience with the
stability of the analytes in the sample collection medium (e.g. , canister, Tenax,
filter).
SA3WIZ TYPES
As mentioned, samples that closely simulate an actual air sample will be
used for each method. For the semi-VOC and metals methods, samples that repre-
sent a portion of the analytical method will also be used. These latter samples
will provide OERR a rough estimate of the extraction efficiency of the analytical
method. The number, type, and sample concentrations to be used are:
(1) VOC-canister -- 6-L canisters (4 to 7) supplied by the participating
laboratories will be spiked with up to 20 of the 36 candidate analytes.
The concentrations will range from 1 to 40 ppb.
(2) VOC-Tenax -- Tenax cartridges (4 to 7) supplied by the participating
laboratories will be spiked with up to 20 of the 31 proposed analytes.
Between 25 and 300 ng of each analyte will be placed on a cartridge.
(3) Semi-VOC (pesticides) -- Three, filter/PUF/XAD-2 cartridges supplied
by the participants will be spiked with up to 15 of the 21 proposed
pesticide analytes. Each participant will also receive three solutions
containing some of the same pesticides to evaluate the efficiency of their
extraction. Cartridges will contain between 400 and 4000 ng and the solu-
tions will contain between 400 and 4000 ng per mL.
(4) Semi-VOC (non-pesticides) -- Same as for pesticides except the
cartridges will contain between 50 and 350 jUg and the solutions will
contain between 10 and 100 /*g per mL.
(5) Metals -- Three solutions, three filters spiked with aqueous
solutions of metals, and two soil samples will be supplied to each
participant. Up to 14 of the 22 proposed analytes will be contained in
the samples. The solutions will contain between 50 and 500 ng per mL; the
filters will contain between 3 and 50 /ig; and the soils will contain
between 50 and 200 /Jg per g.
351
-------
REPORTING THE RESULTS TO AREAL
All participants will report their results using a form provided by AREAL,
They will also report the source, the physical state, and the purity of all
materials used to calibrate their analytical equipment.
STATISTICAL ANALYSIS OF THE DATA
For each method and analyte an analysis of variance (ANOVA) will be
done(1'2>. This assumes that the variances are found to be homogeneous. Con-
fidence limits for precision and accuracy will then be derived for the analyte.
For each method, a multi-variant ANOVA will also be done for all analytes to
obtain additional information on the performance of the method. A ranking
approach such as that described by Youden<3> may also be done, particularly if
the variances are found not to be homogeneous.
REFERENCES
1. Mandel, J., "Repeatability and Reproducibility." Materials Research and
Standards. U (1971), pp. 8-16.
2. Mandel, J., "The Measuring Process." Technometrics. 1, (1959), pp. 251-
267.
3. Youden, W. J., "The Collaborative Test." Proceedings of the 76th Annual
Meeting of the Assoc. of Official Agricultural Chemists, October 1962.
DISCLAIMER
The information in this document has been subject to the United States
Environmental Protection Agency's peer review and has been approved for publica-
tion as an EPA document. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
352
-------
REDUCING UNCERTAINTIES IN INHALATION RISK ASSESSMENTS
Sally A. Campbell
S.A. Campbell Associates
Columbia, Maryland
Kenneth L. Zankel
Versar, Inc. ESM Operations
Columbia, Maryland
Assessment of inhalation risks at hazardous waste sites requires
estimation of ambient air concentrations of toxic contaminants at loca-
tions near the site. This estimation may be developed through ambient air
monitoring or emissions and dispersion modeling. Either approach may be
uncertain by more than an order of magnitude. Uncertainties in ambient
air measurements are related to difficulties in establishing the represen-
tativeness of the measurement time and location, particularly for sampling
programs of short duration. Uncertainties in modeling approaches are
dominated by difficulties in obtaining emissions information, but may also
be affected by terrain features. Hybrid techniques in which the strengths
of each approach are exploited may lead to a substantial reduction in the
uncertainty of the final risk estimate. Application of these techniques
require close collaboration between the modeling and measurement com-
munities .
353
-------
Introduction
Air pathway analyses (AFA) are conducted at Superfund sites to evaluate
inhalation exposures to the toxic chemicals found at the site. Both
baseline risks and risks associated with remediation are generally as-
sessed. A number of guidance documents are available to assist the
analyst1'5.
Air emissions mechanisms at hazardous waste sites include: volatiliza-
tion through the jsoil cap of chemicals presently in soils, condensed
phases, and ground water; release of chemicals from ground water to
surface water with subsequent volatilization both at the ground water
release point and at downstream locations; and resuspension of chemicals
on soils and sediments.
•
Risk assessors must evaluate annual average and maximum short term (8-
or 24-hour) inhalation exposures and risks attributable to each emitting
area and for the site as a whole at nearby residences, work sites, and
recreational areas. In many states, fenceline concentrations of toxic air
contaminants (TACs) for various averaging times are needed for comparison
with state reference values.
Each of these inhalation risk assessment activities requires the
assessor to develop a representation of the atmospheric concentration
field for each component attributable to each site of contamination. This
field is combined with human activity factors to yield exposures, and with
unit risk factors (where available) to yield risks. Uncertainties in-
herent in the unit risk factors are well known but not particularly
interesting as long as the same factors are used in all risk assessments.
These remarks will be addressed, therefore, to improving exposure es-
timates.
At present, inhalation exposures are generally assessed through either
air quality modeling or monitoring. Interested parties frequently suggest
that monitoring results, where available, are preferable to modeled
results. However, this is seldom true. Monitoring data may yield point
concentration estimates with remarkable accuracy, but establishing the
spatial and temporal representativeness of these results is very dif-
ficult. On the other hand, modeling approaches are not reliable when
emissions are not well known, where concentrations must be estimated at a
particular site, or where complex terrain is an issue.
In practice, a combination of modeling and monitoring activities is
needed to produce exposure estimates of acceptable certainty. This paper
provides a description of the limitations of each approach alone and
suggests several ways these activities can be integrated to produce
exposure estimates of lower uncertainty than those produced by either
method alone.
Limitations of Ambient Monitoring Data
The principle limitation of monitoring data is the question of represen-
tativeness. Ambient air measurement techniques are rapidly becoming very
good so that point estimates for many contaminants can be obtained with
uncertainties in the range of one to ten percent for many contaminants,
even at very low concentrations. However, these point source estimates
generally do not provide adequate measures of average and maximum ex-
posures, even-at the measurement point, unless a very extensive monitoring
program is undertaken. Difficulties in obtaining representative data in
short term sampling programs occur both because of meteorological variabi-
lity and source strength variability in both time and space.
A simple thought experiment indicates that the minimum uncertainty in
annual average ground level concentrations obtained through a one week
sampling campaign at a typical hazardous waste site is likely to be on the
order of six to eight. The corresponding maximum 24-hour concentration
actually cannot be assessed with any certainty in such a campaign.
354
-------
We have estimated, using the ISCST model, the average annual and short
term ground level concentrations (GLCs) and GLC variability associated
with unit emissions from two typical uniform ground level area sources
having widths of 300 and 10 m respectively. In each case, the concentra-
tions are assessed at a distance of 300 m from the source center using
meteorological data from Tulsa, Oklahoma.
The projected maximum annual average concentration for the small source
is about twice the estimated concentration for the large source (Figures
la and Ib). The ratio of maximum daily average concentrations for the two
sources is about 5 (Figures Ic and Id). Weekly average concentrations
(block averaged) differ from the average concentrations by a factors up to
4 for the large source and 8 (excluding two near-zero values) for the
smaller source (Figures la and Ib). Examination of the frequency dis-
tributions of the weekly averages indicate that the 95Z confidence limits
GLCs for the two sources differ from the average by about the same factor
(figs le and If). Thus, the measurement results have an minimum inherent
uncertainty of a factor of at least 6-8 for the small source at any one
location. In fact, for any randomly selected week, the probability is
about equal for the small source that an average concentration will be
obtained that is half the actual annual average GLC as that the actual
annual average will be obtained.
The differences associated with source size arise because wind direction
variability is more significant for the small source. The directional
variability of the location of maximum concentration is uncertain by about
a factor of two (not shown), so the overall minimum uncertainty in a
measurement program intended to determine the maximum average off-site
concentration for this type of source is about an order of magnitude.
For cases where emissions are not constant (e.g. fugitive dust) or not
uniform, or where terrain complications are important, the overall uncer-
tainty in the maximum average concentration will be considerably more than
an order of magnitude.
Limitations in Modeling Approaches
Assessment of exposures via modeling requires estimating both an
emissions term and a diffusion parameter in air.
Emission estimates
Emissions estimates for the various types of sources can be derived
using methods described in several EPA guideline documents 1>4>s. Emissions
of fugitive dust from undisturbed soil and from transportation and earth-
moving activities are well characterized. Emissions estimates can be
obtained from the general equations (rated A) in Chapter 11 of AP-42, from
data on limestone and other surface mines (given elsewhere in AP-42), or
from rules of thumb given in the Superfund guidance documents. Unfor-
tunately, the Superfund data sometimes yield emissions estimates that may
be as much as several orders of magnitude lower than the AP-42 data.
Emissions from lagoons and other water bodies can be obtained with some
assurance given detailed information on the nature and concentrations of
the contaminants using existing information on the rate of loss of con-
taminants from water. However, for concentrated lagoons, loss rates may
be reduced by contaminant complexation within the water. Models have not
been developed to address either concentrated sludge or the issues of
multi-layer lagoons having dilute water covers on sludge bottoms.
Emissions through the soil cap of volatile components in contaminated
soils and groundwater through the soil cap may be computed using the Shen
transport equation4'5. This method assumes that each contaminant is
transported vertically to the surface at a rate determined by its subsoil
equilibrium vapor pressure, its diffusivity in air, the soil thickness,
and the dry porosity of the soil. For many sites, especially in the
baseline condition, through-soil volatilization yields the most signifi-
355
-------
cant exposure estimates and drives the risk assessment. Unfortunately, it
is also the area in which emissions calculations are least certain.
For this computation, the largest uncertainties are in the areas of soil
moisture and soil porosity. Projected emissions for dry soil are about
2000 times those for soil in which 90Z of the soil pores are filled with
water. Packed soil and cement caps may reduce emissions, but they cannot
be considered an absolute barrier. For instance, emissions will be
released from under the concrete pad whenever the pad is broken for
construction (and these emissions would contribute to the 70 year ex-
posure). Continuous emissions should be anticipated from cracks and other
imperfections in all of the hard covers. Uncertainties in the nature of
the subsoil contaminant (and therefore, in subsoil vapor concentrations)
may also be important. Through soil emissions are also sensitive to the
hypothesized source size. Because total emissions are computed using
average and maximal emissions rates per unit area and hypothesized emiS-
sions areas, uncertainties in total emissions for through soil emissions
are proportional to the uncertainty in this parameter.
Atmospheric Transport and Dispersion
The availability of well characterized atmospheric transport models
allows computation of annual average and maximum ground level concentra-
tions with some assurance. However, the locations at which these maximal
values occur and the maximal concentrations as a pre-determined point
(such as a specific residence) are projected with much less certainty.
Difficulties in projecting GLCs occur when terrain issues intervene, and
may be significant when emissions are released in ravines, from elevated
lagoons, or in mountainous terrain.
Integrating Modeling and Measurement Activities
Major reductions in the uncertainty of risk assessments may be made by
combining modeling and measurement efforts. Two major areas of oppor-
tunity exist: use of measurement techniques to improve emissions es-
timates and use of dispersion modeling for designing monitoring activities
and interpreting the results.
Measurements that can be made to improve emissions estimates include
studies of soil porosity and moisture, surface and subsoil contaminant
states, soil absorption coefficients, and gross though soil emissions at
well characterized locations. If these measurements are conducted with a
particular modeling approach in mind, they can lead to a substantial
improvement in emissions modeling results. However, measurements of soil
emissions at point locations are not, by themselves, adequate unless their
representativeness can be demonstrated through concurrent modeling or
correlative measurement activities. Likewise, the usual soil gas surveys
are of little use since they are a measure of static, not dynamic trans-
port. However, measurements of soil vapor as a function of depth can be
fitted with soil transport models to yield emissions estimates.
Ambient air measurements can also be designed to enhance transport
modeling. Two considerations are necessary. First, the monitoring
network should be established with the aid of modeling results similar to
those in Figure 1 for the situation to be evaluated. These should be
developed using the best possible representation of the emitting sources
(including stack emissions points) and local meteorological data. This
analysis provides information both'on the best locations for samplers, and
on the number of sampling units and length of record necessary to obtain
data of adequate reliability. Second, the monitoring should be designed
to feed information back to the model (so that short term maximum con-
centrations can be estimated). Examples of such designs include:
o Measurements of GLCs and concurrent meteorological conditions adjacent
to large lagoons or fugitive dust sources to obtain data that can be
back-calculated using appropriate modeling tools to give area-averaged
emissions rates. THese measurements can be enhanced by the use of two
356
-------
pairs of wind-controlled samplers that operate concurrently as upwind-
downwind monitors (with time charts so the sampling period can be
determined). These samples can be integrated over multiple days. Use
of narrow wind acceptance criteria improves the usefulness of the data,
but may lengthen the elapsed time needed to get an adequate sample.
o Measurements using wind-controlled sampling devices to give information
on ambient concentrations at specific locations for various specific
meteorological conditions (shown by modeling to be diagnostic), the
frequency of occurrence of which can be obtained from nearby meteorolog-
ical records to obtain an estimate of annual averages.
o Measurements, as a rough check on the models, of particular cases
expected to give high concentrations. For instance, measurements of
concentrations within a ravine under light nocturnal flow.
o Measurements of TAG flux across a fenceline or other boundary obtained
by suitably designed long-path length measuring devices. Again, the
number and location of such devices should be determined with a par-
ticular modeling approach in mind.
o Real time measurements of emissions from during preliminary site
preparation activities using micro-balance or optical monitors to obtain
data useful in projecting the risks of full remediation
Each of these designs should be developed to maximize the number of
independent measures of the parameter to be tested within the cost
restraints of the project. For instance, two days of wind-controlled
sampling with two pairs of samplers could yield four to six meteorologi-
cally distinct conditions at the edges of a lagoon, each of which could be
used to generate an independent estimate of the emission strength of the
lagoon. In contrast, normal upwind-downwind 8-hour samples would yield,
at best, two independent measures. Two 24-hour samples without direction
control would probably yield no usable data because of the variability of
wind conditions during each period.
Conclusion
Both ambient monitoring and modeling approaches to risk assessment at
hazardous waste sites may be uncertain by more than an order of magnitude.
However, by combining the approaches in a thoughtful way, the strengths of
each may be exploited to resolve the major areas of uncertainties in the
individual approaches and produce a substantial reduction in the uncer-
tainty associated with either approach alone.
References
1. EPA (1977) Guidelines for Air Quality Maintenance Planning and
Analysis Volume 10 (revised). U.S. EPA Office of Air Quality
Planning and Standards. EPA-450/4-77-001.
2. EPA (1985) Compilation of air Pollutant Emissions Factors. U.S.
EPA Office of Air Quality Planning and Standards. AP-42, Fourth
edition.
3. EPA (1986) Guideline on Air Quality Models (Revised). U.S. EPA
Office of Air Quality Planning and Standards. EPA-450/2-78-027R.
4. EPA (1987) Hazardous Waste Treatment Storage and Disposal Facili-
ties (TSDF) - Air Emission Models, Documentation. U.S. EPA Office
of Air Quality Planning and Standards. EPA-450/3-87-026.
5. EPA (1989) Air/Superfund National Technical Guidance Study Series,
Vol. II. U.S.EPA Office of Air Quality Planning and Standards.
EPA-450/1-89-002
357
-------
a)
b)
c)
d)
110
120
100
to
10
to-
20 •
0-
GLC (4/eu.m.)
0 40 M l]0 110 100149 ]•) 110 3(0 400 440 480 UO
QIC (ugta/.m.)
e)
f)
5 10 IS 10 25 30 55 40 45 SO 55
QCC (ugfcu.m.)
0 10 20 u 40
M 10 70 M M 100 MO 1» 1U 140
3UC (utfcu-m.)
Figure 1. Summary statistics of atmospheric concentrations projected for
unit emissions from uniform, ground level area sources: a) weekly block
average GLCs and annual average concentrations for 300 m wide area source;
b). Same as a, but for a 10 m wide source; c) and d), frequency of
occurrence of daily average concentrations for the 300 m and 10 m sources,
respectively; e) and f), frequency of occurrence of weekly average con-
centrations for the two sources.
358
-------
DESIGN AND IMPLEMENTATION OF AN AIR MONITORING
PROGRAM AT A SUPERFUND SITE
Chris G. Harrod, Ebrahim Khalili, Ph.D., and
Bruce E. Dumdei, Ph.D.
ENSR Consulting and Engineering
Westmont, Illinois
A comprehensive air monitoring program was conducted during remediation activities at a Superfund
site in New Jersey. The site was a chemical reclamation operation where chemicals such as hydraulic fluids,
paints, varnishes, solvents, oils, plasticizers, and printing inks were processed between 1965 and 1979. By-
products of processing operations were disposed of at the site and, as a result, the surface and subsurface
soils were significantly contaminated with various organic and inorganic compounds.
An air monitoring program was designed to collect valid, representative measurements of compounds
present in the air during remediation/clean-up activities. Air monitoring was conducted in the immediate vicinity
of the site to protect the health and safety of site personnel and of the general public by ensuring that the
ambient air quality was not significantly affected during remediation activities. During clean-up operations, air
monitoring was conducted along the perimeter of the site using two sampling strategies. First, portable
monitors for total volatile organic compounds and particulates were used to conduct real-time air
measurements. Subsequent analysis of whole air samples was conducted using field gas
chromatography/photoionization detection (GC/PID). These measurements provided real-time indications of
off-site migration of the various target compound groups during remediation activities. Second, in order to
validate the real-time data and provide more detailed documentation of site emissions, time-weighted average
sampling methods using filter collection and/or solid sorbents were employed in the field. Samples collected
were subsequently analyzed at specialized analytical laboratories.
The presented paper provides a discussion of the design of the air monitoring program, the collection
of air quality data, and the interaction between air monitoring activities and remediation activities. The air
monitoring program conducted at the site proved to be effective in controlling emissions. The ambient air
quality was not significantly affected during the remediation, and as a result, the health and safety of both on-
site personnel and the general public were protected.
359
-------
Introduction
Under the direction of the United States Environmental Protection Agency (U. S. EPA)
Region II, a Consent Order was issued to conduct remedial construction work at a former
chemical reclamation facility located in southeast New Jersey. Operations at the facility
included buying, selling, and processing chemical products such as hydraulic fluids, paints,
varnishes, solvents, oils, plasticizers, and printing inks. By-products of these materials
were disposed of on-site, and as a result, surface and subsurface soils were contaminated
with various organic and inorganic chemicals. The abandoned reclamation facility
consisted of a concrete block building, a distillation house, primary and secondary
lagoons, a septic system, a tank farm, and an underground storage tank. The numerous
remediation activities, including the demolition of existing buildings, the cleanup and
removal of storage tanks, and the excavation of on-site and off-site soils, created the
potential for off-site release of volatile organic compounds (VOCs), particulate matter, and
trace elements.
The purpose of the air monitoring program was to design and implement an air monitoring
strategy to control off-site emissions resulting from remedial activities, as well as to protect
the health and safety of site personnel and the general public. Both real-time emissions
measurements and time-weighted average (TWA) sampling were conducted to meet the
objectives of the air monitoring program. This paper presents a discussion of the air
monitoring design and strategy for selection of target parameters, selection of air sampling
techniques, selection of sampling locations, and determination of corrective action
measures for the subject site.
Design Strategy and Implementation
Selection of Target Parameters
A number of compounds, including VOCs, semi-volatile organic compounds (SVOCs),
polychlorinated biphenyls (PCBs), lead, and chromium, were detected in samples collected
at the site from numerous soil borings during previous site investigations. A modified risk
assessment was conducted to select target compounds that would be monitored at the
site. Specific target compounds were selected based on the following criteria:
Concentration in the soil
Frequency of detection in soil samples
Toxicity
Exposure limits
Volatility
A ranking system was developed to make the final selection of target compounds.
Selection of Sampling Techniques
Based on the selection of target parameters and compounds, both real-time and TWA air
sampling techniques were used to monitor air at the perimeter of the site. Instantaneous
(real-time) field measurements for total VOCs were performed using an HNu Model PI-101
photoionization detector (PID). A real-time aerosol monitor, the GCA MINIRAM, was used
360
-------
to measure real-time particulates. The MINIRAM is a portable, battery-operated instrument
that measures airborne particulates. The instrument has a detection limit of approximately
0.01 mg/m3. A field gas chromatograph (GC) equipped with a PID was also used to
identify real-time target compounds during the remediation activities. The GC/PID was
calibrated daily. Samples for on-site GC/PID analysis were collected in gas-tight glass
syringes and immediately analyzed.
Real-time monitoring program activities are summarized in Table I. The table lists the real-
time parameters monitored for each of the 12 remediation activities. As shown in this
table, the appropriate real-time monitoring was selected based on the nature of each
specific remediation activity.
In order to validate the real-time data and to provide more detailed documentation of
emissions from the site, manual, TWA sampling methods using filter collection and/or solid
sorbents were conducted in the field, and subsequent laboratory analysis of the samples
was performed. The sampling methods included TWA volatile sampling with Tenax
adsorbents; total suspended particulates (TSP), lead, and total chromium sampling with
Hi-Vol filters; and semi-volatile sampling for polychlorinated biphenyls (PCBs), naphthalene,
and phenol with polyurethane foam {PUF)/XAD-2 sorbent cartridges. TWA volatile organic
were collected, using a personal sampling pump, on a Tenax sorbent cartridge followed
in series by a Tenax/charcoal cartridge.
TWA sampling program activities are summarized in Table II. The table lists the TWA
sampling conducted for each of the 12 remediation activities. As with the real-time
monitoring, the selection of appropriate TWA monitoring was based on the nature of each
specific remediation activity.
Selection of Sampling Stations
Real-time monitoring stations were selected each day based on National Weather Service
local wind forecast, prevailing wind direction recorded on-site by the meteorological
monitoring station, and locations of on-site remediation activities. Real-time measurements
for total VOCs using the HNu were conducted hourly at one upwind location and three to
four downwind locations, as necessary, to determine total VOCs representative of
conditions downwind of site activities. Additional locations were selected if needed. Real-
time monitoring for particulates was conducted at the same times and locations as the
VOC measurements using a MINIRAM monitor. More frequent monitoring (30-minute
intervals) was performed when action level exceedances were encountered.
Locations for TWA samples were selected based on the criteria used to determine real-time
monitoring locations. TWA samples, including specific VOCs, PCBs, naphthalene, phenol,
TSP, lead, and total chromium were collected each working day. The required samplers
were established at the selected locations and allowed to collect samples for approximately
8 hours.
Determination of Corrective Action Measures
Action levels were established for the real-time monitoring program based on negotiations
with the U. S. EPA. Reaf-time measurements provided instantaneous indications of off-site
361
-------
migration of various target compounds during remediation activities. Figures 1 through 4
summarize action levels and the required responsive steps.
If the HNu reading was 1 ppm or greater above the background level (upwind
concentration), the on-site safety officer and project manager/field coordinator were
immediately notified. In such cases, the frequency of monitoring was increased to 30
minute intervals, and downwind grab samples were analyzed using the Photovac GC/PID.
The intervals between measurements remained at 30 minutes until the difference between
the upwind and downwind measurements was less than 1.0 ppm for three consecutive 30-
minute readings.
When the VOC levels at any downwind site exceeded those measured at the upwind site
by 2.5 ppm or greater, the GC/PID was used to evaluate which real-time VOCs were
present. The site coordinator was notified of HNu readings that remained at 2.5 ppm
above background for 10 minutes so that a course of action could be chosen to bring the
level of emissions below action limits. Possible courses of action included the use of
vapor suppressant foams, work slowdown, and if necessary, work shutdown.
If the real-time paniculate level at any downwind site exceeded that measured at the
upwind site by 1.0 mg/m3 or more, the project manager was notified so further action
could be taken to reduce emissions. No such action was necessary during the air
monitoring program.
Criteria for Sample Analysis
TWA samples with the highest potential to provide useful, defensible, and representative
data of the highest quality were prioritized each day for analysis. Samples were validated
individually upon collection for analysis based on the following criteria:
Meteorology
Sampler operation
Physical integrity of sample
Proper sample documentation
Field observations
Exceedance of action levels
Although daily sampling was conducted to provide data for the thorough coverage of a!)
potential emissions events, the above criteria were used to assure a cost-effective air
monitoring program.
Results
The results of the real-time measurements for VOCs are shown in graph form in Figure 5.
As shown, for the majority of the days measurements were taken, VOC concentrations
were below 1 ppm.
The results of the GC/PID analyses also indicated that, for the majority of days monitored,
no concentrations of VOCs above detection limits were identified. On days when one or
more compounds were detected, for the majority of the day, VOC concentrations for all
real-time target compounds were below detection limits. The compounds most frequently
362
-------
detected by GC/PID were trichloroethene, toluene, and tetrachloroethene. These results
are consistent with the results of the TWA sampling for VOCs. Concentrations of TWA
VOCs recorded for an 8-hour average were, as expected, much lower than the
instantaneous grab samples taken for on-site GC/PID analysis.
Several compounds were detected during the TWA sampling program. VOCs were also
detected at the upwind sites at background levels. All downwind concentrations were
below 130 ppb. The majority of concentrations were below 10 ppb.
The results of the real-time particulate measurements indicated that the highest
intantaneous particulate concentration encountered was 0.4 mg/m3. Readings were
therefore lower than the 0.5 mg/m3 action level for the entire project. The majority of all
readings were at or near background levels for the duration of the project. These results
are consistent with the TWA TSP results. The highest concentration of TSP detected was
258 /ig/m3. The highest concentration of lead detected was 0.26 /ig/m3, and of
chromium 0.02 jig/m3. TSP levels for the majority of samples were below 100 /ig/m30,
lead was below 0.1 /jg/m3, and total chromium was below detection limits.
The results of the analyses of TWA samples for PCBs indicated that only low levels of
PCBs (< 15 ng/m3) were detected in the course of monitoring conducted during any of the
site activities. Two other SVOCs (naphthalene and phenol) were collected on the same
cartridge as PCBs. The results of the analyses for these compounds indicated only low
levels of these compounds. The highest detected concentration of phenol was 0.19
ng/m3, and of naphthalene, 21.8 ng/m3.
Conclusions
The air monitoring program conducted at the Superfund site proved to be effective in
controlling site emissions. The real-time monitoring program enabled the continuous and
timely generation of data that provided useful emissions information during daily site
remediation activities. This information allowed the field site coordinator and site safety
officer to determine when appropriate corrective measures were required to reduce site
emissions.
Because of the conservative nature of the real-time monitoring action levels, off-site
impacts were minimized. The use of the GC/PID supported the HNu readings by
identifying and quantifying the real-time target VOCs of concern at the site. The manual
TWA monitoring program provided detailed information regarding the magnitude of
emissions over 8-hour periods throughout the project. The data produced from the manual
TWA monitoring correlated well with the real-time data. Analysis of Tenax, PUF/XAD-2,
and Hi-Vol samples indicated that only low to average emission levels were detected at any
time during the project.
The air monitoring program conducted at the site successfully accomplished its intended
goals. The program demonstrated that the ambient air quality was not significantly
affected during remediation activities and, as a result, the health and safety of both on-site
personnel and the general public were protected.
363
-------
TABLE!
Real-Time Monitoring Program
Slt« Remediation ActfvKv
1) On-Site Soils
2) Off-Site Soils
3) Concrete Block Building
4) Distillation House
5) Interior Process Tanks and Vessels
6) Exterior Process Tanks and Vessels
7) Underground Tanks
8) Tank Farm
9) Septic System
10) Excavation Area
11) Secondary Lagoon Area
12) Buried Waste Area
Monitoring Conducted
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Real-Time Particuiates (MINIRAM)
Real-Time Particuiates (MINIRAM)
Total VOCs (HNu PID), Target VOCs (Photovac 10S
GC/PID)
Total VOCs (HNu, PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Real-Time Particuiates (MINIRAM)
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
TotaJ VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
Total VOCs (HNu PID), Real-Time Particuiates
(MINIRAM), Target VOCs (Photovac 10S GC/PID)
364
-------
TABLE II
Time-Averaged Sampling Program
Sfte Remediation Actlvftv
1) On-Site Soils
2) Off-Site Soils
3) Concrete Block Building
4) Distillation House
5) Interior Process Tanks and Vessels
6) Exterior Process Tanks and Vessels
7) Underground Tanks
8) Tank Farm
9) Septic System
10) Excavation Area
11) Secondary Lagoon Area
12) Buried Waste Area
Monitoring Conducted
Target Volatiles (Tenax) PCBs, Phenols, Naphthalene
(PUF/XAD) Particulars, Metals (Hi-Vol)
Target Volatiles (Tenax) PCBs, Phenols, Naphthalene
{PUF/XAD) Particulates. Metals (Hi-Vol)
Particulates (Hi-Vol)
Particulates (Hi-Vol)
Target Volatiles (Tenax)
Target Volatiles (Tenax) Particulates, Metals (Hi-Vol)
Target Volatiles (Tenax) Particulates, Metals (Hi-Vol)
Target Volatiles (Tenax) Particulates, Metals (Hi-Vol)
Particulates (Hi-Vol)
Target Volatiles (Tenax) PCBs, Phenols, Naphthalene
(PUF/XAD) Particulates, Metals (Hi-Vol)
Target Volatiles (Tenax) PCBs, Phenols, Naphthalene
(PUF/XAD) Particulates, Metals (Hi-Vol)
Target Volatiles (Tenax) PCBs, Phenols, Naphthalene
(PUF/XAD) Particulates, Metals (Hi-Vol)
365
-------
CO
OT
REAL-TUE UOMTORINC
TOTAL
VOLATILES
CONSTRUCTION LOCATION®
NW FORECAST
ON-SITE WTO DIRECTION
4-6 SITES
P ARTICULATES
• IDENTIFY CONSTRUCTION UNTT(S)
• DETERMINE UONTORMC REQUIRED
TARGET
VOLATLES
Figure 1. Air Monitoring Plan: Overview
-------
TARGET VOLATILES
CALIBRATE WTH STANDARDS
FOR TARGET VOLATILES
SAMPLE DAILY OR WHEN
TOTAL VOLATILES > £5
TARGET VOLATILES
> 0.5 ppm EACH
I
ANY TARGET
VOLATILE > 1 ppm
NOTIFY PROJECT MANAGER
AND
SITE SAFETY OFFICER
1
• NOTIFY PROJECT MANAGER AND
SITE SAFETY OFFICER
• TAKE ACTION UNTIL TOTAL
VOLAmiS < 1 ppm FOR 2
CONSECUTIVE 30 UIN. READINGS
Figure 2. Air Monitoring Plan: Target Volatiles
367
-------
PARTICULATES
1
L
CALIBRATE MINIRAM
J
MONITOR AND RECORD
DATA AT EACH SITE HOURLY
I
• NOTIFY PROJECT MANAGER AND
SITE SAFETY OFFICER
• INCREASE FREQUENCY TO 30 KIN.
UNTIL DOWNWIND MINUS UPWIND
< 0.5 mg / m3 FOR 3
CONSECUTIVE 30 MM. READINGS
| NOTIFY PROJECT MANAGER AND
SITE SAfETY OFFICER
MODIFY CONSTRUCTION ACTIVITY
UNTiL DOWNWIND MINUS UPWIND
<0.5 ma/m3 FOR 2
CONSECUTIVE 30 MM. READINGS
Figure 3. Air Monitoring Plan: Participates
368
-------
I
TOTAL VOLATILES
1
CALIBRATE WITH
5 ppm IS08UTYLENE
1
MONITOR AND RECORD
DATA AT EACH SITE
HOURLY
DOWNWIND MINUS
UPWIND < 1 ppm
DOWNWIND MINUS
UPWIND > 1 ppm
DOWNWIND MINUS
UPWND > 2.5 ppm
• NOTIFY PROJECT MANAGER AND
SITE SAFETY OFFICER
• INCREASE FREQUENCY TO 30 MIN.
UNTIL DOWNWIND MINUS UPWIND
< 1 ppm FOR 3 CONSECUTIVE
30 MIN. READINGS
USE HELD C.C. TO MEASURE
SPECIFIC VOLATILES
• NOTIFY PROJECT MANAGER AND
SITE SAHTY OFT1CER F > 2.5
ppm FOR 2 CONSECUTIVE 30 MIN.
READINGS
' TAKE ACTION UNTIL DOWNWIND
MINUS UPWIND < 1 ppm FOR 2
CONSECUTIVE READINGS
Figure 4. Air Monitoring Plan: Total Volatiles
369
-------
03
-3
O
g,
20.0
Maximum VOC Concentration Range (ppm)
(BDL - Below Detection Limit)
Figure 5. Maximum HNu Readings Histogram
-------
EVALUATION OF METHODS FOR DETECTING DIMETHYL
MERCURY IN AMBIENT AIR AT A SUPERFUND SITE
Thomas H. Pritchett
U.S. EPA Environmental Response Team
GSA Raritan Depot (MS-101)
Edison, NJ 08837
Larry Kaelin
Roy F. Weston - REAC
GSA Raritan Depot (MS-802)
Edison, NJ 08837
ESrian Brass & Vinod Kansal
International Technology, Inc. (REAC)
GSA Raritan Depot (MS-802)
Edison NJ
Dimethyl mercury, which can be formed by biological detoxification processes for
elemental and inorganic mercury, is ten times more toxic than elemental and inorganic
mercury, and is more than 10^4 times more volatile than elemental mercury. Thus even
trace levels of dimethyl mercury relative to elemental mercury can be of significant
concern during soil excavations. This past summer, our organization was requested to
develop both a real-time analytical method capable of detecting organomercury
compounds at the TLV of 10 ug/m^S and a time-weighted-average sampling method
with a detection limit less than .25 ug/m^S.
The real-time method was developed built around the Jerome gold film mercury vapor
analyzer. The time-weighted-average methods that were initially tested for applicability
to dimethyl mercury included use of potassium permanganate based impinger solutions,
adsorption/thermal desorption with carbon tubes, and adsorption/thermal desorption
with Tenax tubes. Tenax was the only method to give reliable results. Additional work
was then done with Tenax to determine safe sampling volumes under field conditions, to
estimate approximate detection limits at three local laboratories, and to determine the
degree that the detection limits could be increased by using GC/MS-selected ion
monitoring versus GC/MS-full scanning.
371
-------
INTRODUCTION
In June 1989, U.S. EPA Region I requested assistance from the U.S. EPA
Environmental Response Team (ERT) because of the potential for dimethyl mercury
emissions at the Nyanza Chemical site in Ashland, Massachusetts, The current cleanup
of mercury contaminated soils, which were being performed under contract for the Army
Corps of Engineers, had then been shut down for three weeks because of the lack of
confirmed real-time and time-weighted-average air monitoring methods. Until such
methods could be developed, no excavations of mercury laden soils would be allowed by
the Corps, by the on-site contractor's industrial hygienist and by the surrounding
community. This shut down was costing the government an estimated $10,000 a day.
The ERT*s preliminary assessment of the situation indicated that indeed releases
of dimethyl mercury (DMM) could conceivably occur during the remediation of the
mercury contaminated, wetland portions of the site. Dimethyl mercury had previously
been documented in the literature as a by-product of anaerobic degradation of mercury.
The site contained several wetland areas in which such degradation could occur and in
which the standing water could act as a potential barrier to DMM escaping to the
atmosphere. Therefore, it was conceivable that pockets of DMM vapor could be bound
to the subsurface mercury contaminated soils in these areas. In addition, because of its
greater than four orders of magnitude higher volatility (vapor pressure of 63 mm Hg at
25°C) relative to elemental mercury and its ten-fold greater toxicity, even minute relative
quantities of DMM elemental mercury could be a more significant health concern than
the elemental mercury itself - especially considering that the TLV for DMM is 10 ug/m3
or approximately 1 ppb(v).
Because the potential DMM problem was a valid concern and because the lack of
any confirmed method was costing the government an appreciable amount of money, the
ERT agreed to develop the analytical methods that could be used by the on-site cleanup
contractor. Since the sampling and analyses would ultimately be performed by the
cleanup contractor and not by the ERT, the developed methods had to meet several
criteria normally not found in pure research, method development. First, all time-
weighted-average (TWA) sampling methods had to use commercially available sampling
equipment and media that could be analyzed by a typical environmental or industrial
hygiene laboratory. Second, the real-time analytical methods had to be developed
around commercially available portable instruments. Third, both methods had to be
selective for DMM over elemental mercury. Fourth, the initial desired sensitivities of
the real-time and TWA methods were 10 ug/m3 and O.lug/m3, respectively.
The real-time and semi real-time methods that were initially considered included
the Jerome gold film mercury vapor analyzer, the Bacharach MV-2 mercury vapor
analyzer, and gas chromatography with a photoionization detector (PID). The TWA
methods initially considered were absorbent tubes with thermal desorption to a GC/MS
and impinger solutions. The two absorbent media evaluated were Tenax and
Carbosphere.
372
-------
REAL-TIME ANALYSES
The Bacharach MV-2 is a UV instrument which responds to compounds such as
elemental mercury vapors, that absorb UV light at the wave length of 254 nm. As was
ejected, initial work with the Bacharach MV-2 at the Region laboratories indicated
that the instrument would not respond to DMM. Before any organo-mercury compound
can be detected by UV absorption, it must first be converted into elemental mercury. It
may have been possible to use an impinger to initially trap the DMM, then convert it to
elemental mercury, and finally purge the elemental mercury vapors into the instrument.
However, since the prime objective was to develop real-time monitors for field
applications, such an impinger setup for the MV-2 was not pursued.
The Jerome meter is a gold film sensor which is set into one side of a wheatstone
bridge circuit Elemental and organo-mercury vapors will contact the gold film and form
an amalgam with the gold surface. The formation of the amalgam changes the resistance
of the film detector and unbalances the circuit. The result is a signal which is
proportional to the mercury vapor concentration. The amalgam is reversible and the
gold film returns to its initial resistance once the mercury or organomercury vapors are
removed. With elemental mercury the film must be heated in order to revert the
amalgam back to pure gold. However, with DMM several runs of clean air are typically
sufficient to restore the film surface and will result in a decreased per cent film
saturation reading. This decrease in film saturation without heating the film was only
seen with DMM and not with elemental mercury.
Initial consultation with the manufacturer's technical staff indicated that the
instrument would have a response to DMM but that response had not been reproducible
or linear in their experiments. Initial work at the Region 1 laboratory using an
instrument configured as per manufacturer's recommended conditions confirmed that the
response on the Jerome to DMM to be erratic and unreliable. Subsequent work found
that following each positive readings with four successive clean air purges resulted in a
linear and reproducible response. Evidently, the methyl groups prevent DMM from
forming a true amalgam and cause the DMM-gold interaction to remain a surface
phenomenon. Because the response is due to a surface interaction and not a true
amalgam, the response was found to be extremely sensitive to the cleanliness of the
film's surface. As can be seen in Table I, the maximum sensitivity of the instrument is
obtained with freshly cleaned films. As the film becomes dirty from use, the sensitivity
will decay over time. As can be seen in Figure 1, this decay is also a function of the
concentration of DMM that the instrument has been seeing.
Finally because the methyl groups force the DMM to have certain orientations
prior to interacting with the gold film, the overall sensitivity for the instrument is less for
DMM than for elemental mercury. Modifications to the instrument were required to
increase the sensitivity such that it would read the correct mg/m3 reading for DMM.
These modifications included increasing the sampling flowrate, doubling the sampling
time to 20 seconds, and increasing the signal amplification by increasing the resistance
between two test points (TP2 and TP3) from approximately 65 Ohms to 98 Ohms.
373
-------
Selectivity for DMM over elemental mercury was obtained by placing a NIOSH
6000 silver impregnated Chromosorb tube in the inlet of the instrument. These tubes
were found to selectively absorb the elemental mercury without affecting the transport of
DMM at all. In addition, the suppliers specifications were such that changing tubes did
not affect the instrument's calibration by significantly changing the sampling flowrate.
The other "real-time" method, which was suggested by Dr. Tom Spittler, involved
the use of a portable GC equipped with a PID detector. The GC was able to detect
DMM and apparently had no response for elemental mercury. However, the GC could
just barely resolve DMM from benzene. Unfortunately, low ppb(v) levels of benzene
would be expected immediately downwind of an area in which diesel-powered earth
moving equipment are be used. In addition, the instrument detection limit had a
detection limit of approximately 0.3 - 0.5 ppb(v). Therefore, because of the potential
problems with benzene interferences and because of the initial successes with the
Jerome, the ERT did not pursue this option further. Since that time, researchers at Oak
Ridge National Laboratories have continued work with the portable GC using a Tenax
based sample preconcentrator.
TWA SAMPLING METHODS
Trapping with Spherocarb followed by thermal desorption using TO-2 conditions
did not work. The affinity of Spherocarb for DMM was too great to be overcome by the
temperatures being used.1
The second method used was an impinger train which was designed to trap the
DMM by oxidizing it to inorganic mercury. After samples were pulled, the solutions
were taken to a laboratory which analyzed them by cold vapor AA. The impinger
solution had the following composition: 5% v/v of 5N H2SO4, 2.5% v/v of concentrated
nitric acid, 15% v/v of a 5% w/w potassium permanganate solution, and 8% of a 5%
w/w potassium persulfate solution. However, after spiking experiments with different
flowrates, this method was abandoned. Because of its high volatility and low water
solubility, the DMM was being purged through the impinger solutions before it had a
chance to be oxidized and thereby immobilized.
Initial work performed in conjunction with Dick Siscanaw of Region I
demonstrated better than 85% recoveries of DMM spiked onto the Tenax and then
thermally desorbed. Spiked samples sent out to local environmental laboratories
indicated that levels as low as 5 ng/tube to 25 ng/tube, could be detected using a
GC/MS in the full scan mode. Additional work with one laboratory demonstrated that
using selected ion monitoring (SIM) enabled the laboratory to improve the effective
detection limits five-fold.
After determining that DMM could be quantitatively desorbed from the tubes and
that the GC/MS methods could detect low quantities of DMM, the actual sampling
parameters were established. This consisted of spiking 1.5 gram Tenax tubes with
vaporized injections of a DMM solution in methanol. On a hot summer day these tubes
were then taken on-site and placed directly above dry areas of known high levels of
374
-------
elemental and inorganic mercury vapors. Three different flowrates were used to pull
different volumes of the site air through each set of tubes. The tubes were then taken
back to the laboratory for GC/MS analyses. As can be seen in Table II, no
breakthrough was seen with the 24 liter sample. This sample volume combined with the
worst GC/MS/SIM ng/tube detection limit translated into a method detection limit in
i:he range of 0.2 ug/m3.
CONCLUSIONS
Based upon this work, the following three tier sampling approach has been
recommended to the Corps of Engineers for monitoring DMM in the air during the
Nyanza cleanup operations. The first tier involves the use of the Jerome gold film
analyzer with the silver impregnated chromosorb pre-filter to obtain real-time readings
for DMM. Care should be taken to insure that the instrument's calibration is checked
and re-adjusted throughout the day. Whenever a positive reading is obtained by the
Jerome, a confirmation air sample should be taken and concentrated approximate 5-10
fold for analysis by the PID portable GC, the second tier. The third tier would consist of
both perimeter Tenax samples and at least work zone Tenax sample for subsequent off-
site analyses. As more sensitive and selective GC detectors, such as the atomic
fluorescence detector, are developed and become more common, then these instruments
would be appropriate as replacement tier 2 analytical methods.
375
-------
TABLE I. JEROME METER RESPONSE AND HEAT CYCLES
DMM STANDARD @ 6.4 ppbv
Time (minutes)
Start - 0
End - 345
After heat cycle -
% recovery -
DMM STANDARD @ 13.7 ppbv
Time (minutes)
Start - 0
End - 120
After heat cycle -
% recovery -
meter reading
0.040
0.014
0.038
95.0%
meter reading
0.188
0.120
0.177
94.1%
NOTE: Above standards analyzed at 5 minutes intervals until end time.
TABLE II. TENAX TUBE BREAKTHROUGH RESULTS
Tube #
1
2
3
4
5
6
7
8
9
10
11
ng DMM spiked
2000
2000
2000
2000
2000
2000
2000
2000
2000
0
0
Volume of Air
Pulled Thru tube
95
95
48
48
24
24
0
0
0
0
0
ng DMM
GC/MS Data
< 25
57
970
1200
1800
2500
1900
1900
2000
< 25
< 25
% Recovery
0
2.85
48.5
60.0
90.0
125.0
95.0
95.0
100.0
376
-------
FIGURE 1. DECAY OF DIMETHYL MERCURY RESPONSE OVER TIME AS OBSERVED AT
THREE SEPARATE CONCENTRATIONS.
.20 —i
to
-J
-J
.10 —
O
Q_
(/)
LJ
(X.
.040
.030 —
.020 —
.010 —
.000
50
150
200
MINUTES
LEGEND
O - 13.7 ppbv OMM
D - 6.4 ppbv DMM
^ - 1.3 ppbv DMM
I I I T T I
250
300
350
-------
HAZARDOUS WASTE TSDF PMln EMISSIONS
10
C. Cowherd, Jr.
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110
W. L. Elmore
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 47711
This paper summarizes a guidance document that provides regulatory
and industrial personnel with necessary information to identify sources of
contaminated fugitive PM10 emissions, estimate the magnitude of emissions,
select viable control measures, and estimate the effectiveness of those
measures. Sources of particular concern are landfills, land treatment,
waste stabilization, dry surface impoundments, and roads. Included in the
document are detailed descriptions of PM10 emission models and control
performance models that apply to these sources. These models were devel-
oped by plume profiling of representative sources before and after control
application using monitoring techniques described in the paper. Because
of the similarity between various Superfund site activities and treatment,
storage, and disposal facility (TSDF) operations, these technological
approaches are in large part transferable to Superfund sites.
378
-------
Introduction
Earlier EPA studies have evaluated fugitive particulate matter (PM)
emissions from hazardous waste Treatment, Storage, and Disposal Facilities
(TSDFs). Sources of particular concern are landfills, land treatment,
waste stabilization, dry surface impoundments, and roads. These sources
of emissions are regulated by a site-specific permitting system for TSDFs
that has been established under the authority of the Resource Conservation
and Recovery Act (RCRA).
This paper summarizes a guidance document1 that provides regulatory
and industrial personnel with sufficient information to identify sources
of contaminated fugitive PM emissions, estimate the magnitude of emis-
sions, select viable control measures, and estimate the effectiveness of
those measures. The PM particle size fraction of interest, designated as
PM10, consists of particles with an aerodynamic diameter equal to or less
than 10 micrometers (ym).
Major TSDF Source Categories
The following source categories may be identified in relation to
TSDFs:
Paved and Unpaved Roads
Open Waste Piles and Staging Areas
Dry Surface Impoundments
Landfills
Land Treatment
Waste Stabilization
Unlike the other source categories listed above, the emissions from
paved and unpaved roads at hazardous waste (HW) treatment, storage, and
disposal facilities (TSDFs) are not due to an identifiable unit operation
at each facility. Rather, the roads serve as "linkages" between the vari-
ous unit operations (e.g., access from the gate to the active disposal
area, transfer of solidified wastes from the solidification unit to a
disposal unit, etc.).
Particulate emissions occur whenever a vehicle travels over a paved
or unpaved surface, such as a road or an area proximate to a TSDF unit
operation. Note that travel areas adjacent to TSDF operations may have a
high potential for contaminated particulate emissions, with the potential
inversely related to the level of "good operating practices" at the unit
operation. In the case of paved roads, emissions originate with
(a) resuspension from vehicle tires and undercarriages or (b) material
deposited onto the surface. Material tracked onto the road from poten-
tially contaminated travel areas surrounding TSDF operations will be
spread along the road's length and eventually some of that material will
become airborne. While track-on material is usually more noticeable on
paved roads, the spreading of potentially contaminated material also
occurs on unpaved surfaces.
Past surveys of TSDFs have in fact indicated that unpaved road
surfaces at these types of facilities may have RCRA metal and semivolatile
organic contamination levels essentially comparable to levels in samples
taken from process-related surfaces (e.g., landfill areas, areas surround-
ing a stabilization/solidification unit, etc.).2 Because of the large
volume of traffic associated with many facilities, those surveys suggest
that general vehicular traffic on plant roads may be the principal source
379
-------
of contaminated emissions associated with TSDFs. While the source of road
surface contamination cannot be definitively identified, the two most
likely causes are (a) spillage from vehicles hauling the waste or
(b) "track-out" of material from surfaces surrounding process operations
at the TSOF.
In many respects, control measures commonly applied to alleviate
"gross" (i.e., uncontaminated) particulate emissions may actually compound
contaminated emissions. For example, travel areas surrounding an active
TSDF unit operation may be watered to control emissions. However, water-
ing tends to increase the amount of track-out onto an adjacent paved road,
and the deposited material soon dries out and becomes airborne under
normal plant traffic conditions.
Each section of the guidance document begins with an overview of a
specific source category, describing emission characteristics and mech-
anisms. Following this, available emission factor models (Table I) are
presented to provide a basis for analyzing the operative nature of control
measures. Next, demonstrated control techniques are discussed in terms of
estimating efficiency and determining costs of implementation. Suggested
regulatory formats explain the "philosophy" used in implementing the
preceding technical discussions in viable regulations and compliance
actions.
Particulate Emission Factor Models
In developing particulate control strategies for "traditional"
pollutant concerns (i.e., to meet National Ambient Air Quality Standards
[NAAQS]), gross particulate emissions from open sources are estimated
using the predictive emission factors presented in Section 11.2 of EPA's
"Compilation of Air Pollutant Emission Factors" (AP-42).3 These factors
cover the generic source categories:
Unpaved travel surfaces
Paved travel surfaces
• Exposed areas (wind erosion)
Materials handling
Note that these factors are updated periodically and that users of this
document should use updated factors as they become available.
These emissions factors share many common features. For example, the
models are formulated as empirical expressions that relate variations in
emission factor (e) to differences in the physical properties (p) of the
material being disturbed and the mechanical energy (m) responsible for the
generation of particulate according to the general form:
e = Kpamb (1)
As empirical models, open dust source emission factors have adjustable
coefficients (K,a,b) that reflect relationships determined from actual
open dust source testing.
Much of cited guidance document centers on the application of the
open dust factors to estimate particulate emissions from permitted TSDF
units. Table I summarizes the applicability of AP-42 emission factor
380
-------
models to TSDF source categories. Estimation is accomplished according to
the general model form:
R = j: a. • e. • A, (2)
where: R = emission rate of contaminated airborne participate (kg/yr)
for a given TSDF, consisting of n identifiable unit
operations
a,- = fraction of contaminant in particulate emissions for the
jth operation (ug/g); for uncontaminated particulate
emissions, a- = 0
J
e.: = emission factor(s) (mass/source extent)
J
A.: = source extent(s) (source dependent units)
J
This approach is analogous to that used by MRI in recent work on
estimation of emissions from surface contamination sites.1* The approach
is also consistent with techniques used in air pollution assessments.
In the TSDF content it is important to recognize the following:
1. The a term refers to the "level of contamination" and is usually
represented as the concentration (e.g., yg/g) of the metal (s) or
organic compound(s) of concern in the disturbed material.
2. Permitted TSDF units may consist of multiple, identifiable unit
operations; in the case of landfills, for example, unit opera-
tions include loadoat of bulk hazardous waste, loadout of
temporary cover, lift construction, and general vehicle traffic
proximate to the landfill face.
3. The major assumption implicit in Eq. (2) is that TSDF processes
that generate particulate emissions can be adequately repre-
sented by existing open dust source emission factor models.
Item 2 above is relatively straightforward and is demonstrated in the
individual chapters. Items 1 and 3 require further clarification as given
below.
Specification of the level of contamination (a) probably is the
single most difficult aspect of applying the emission factors. For a
given TSDF source, "representative" value(s) of o may depend upon a number
of factors including principally:
1. The nature of waste streams handled and whether the wastes are
metal-containing or organic waste streams.
2. Operational practices that influence the availability of waste
material for entrainment into the atmosphere.
3. The age of the facility to the extent that it reflects
"residual" contamination associated with long-term disposal and
typical waste stream volumes.
381
-------
Note that the term representative is used here to connote a long-term
value, such as an annual mean, that adequately defines a over the entire
TSDF source.
There are essentially two methods for specification of a. The
preferred method is through source-specific sampling and analysis (S&A).
In some cases it may be feasible to develop estimates of a for a specific
source based on existing information other than actual S&A data. In
general, these estimates should be developed from the perspective of a
"worst-case scenario," because of the greater uncertainty involved. In
other words, they should be conservatively high and thus tend to protect
against any potential underestimation of risk associated with particulate
emissions.
The principal sources of information for worst-case estimates are
expected to be:
1. Waste manifests and any "tracking" information that a facility
routinely generates to characterize its receipts.
2. Conversations with facility personnel.
Because the emission factor models are empirical expressions, in a
strict sense their degree of applicability relative to TSDF sources is
related to the degree of compliance with two conditions:
1. How closely a given TSDF unit operation resembles the source
conditions underlying test data used in development of the
emission factor.
2. How closely the physical characteristics of the disturbed
material resemble those tested in development of the emission
factor.
Based on surveys described in reference 2, the general conclusion is
that TSDF unit operations are reasonably comparable to those tested in
development of the emission factor relationships.
Results of the surveys described in reference 4 also clearly point
out that the physical characteristics of waste and disturbed surface
materials at TSDFs do not always conform closely to those materials tested
in development of the emission factors. More specifically, the moisture
and/or "oily nature" of waste streams may be markedly different from the
dry, finely divided materials upon which the field tests were performed.
The current emission factor models do not fully account for the inherent
mitigation provided by the physical binding effect of oily substances,
which may add conservatism to the predictive emission rates. However, at
any given facility, the potential for increased spreading of material may
offset this conservatism.
Preventive and Mitigative Control Options
Typically, there are several options for control of fugitive
particulate emissions from any given source. This is clear from the
mathematical equation used to calculate the emissions rate:
R = A e (1 - c) (3)
382
-------
where: R = estimated mass emission rate
A = source extent (i.e., surface area for most open dust
sources)
e = uncontrolled emission factor, i.e., mass of uncontrolled
emissions per unit of source extent
c = fractional efficiency of control
To begin with, because the uncontrolled emission rate is the product of
the source extent and uncontrolled emission factor, a reduction in either
of these two variables produces a proportional reduction in the
uncontrolled emission rate.
Although the reduction of source extent results in a highly
predictable reduction in the uncontrolled emission rate, such an approach
in effect usually requires a change in the process operation. Frequently,
reduction in the extent of one source may necessitate the increase in the
extent of another, as in the shifting of vehicle traffic from an unpaved
road to a paved road.
The reduction in the uncontrolled emission factor may be achieved by
process modifications {in the case of process sources) or by adjusted work
practices (in the case of open sources). The degree of the possible
reduction of the uncontrolled emission factor can be estimated from the
known dependent of the factor on source conditions that are subject to
alteration. For open dust sources, this information is embodied in the
predictive emission factor equations for fugitive dust sources as pre-
sented in Section 11.2 of EPA's "Compilation of Air Pollutant Emission
Factors" (AP-42).z
Control techniques can be divided into two broad categories—preven-
tive and mitigative. Although differences between the two are not always
clear, in general, preventive measures involve techniques that reduce
source extent or improve mechanical source operations relative to the
generation of particulate emissions. By contrast, mitigative techniques
typically focus on altering the surface/material conditions that consti-
tute the source of particulate emissions.
In the TSDF context an example of a preventive control technique
involves traffic routing of haul truck vehicles in the staging areas
proximate to an active landfill face. The objective of traffic routing is
to minimize the contact between vehicle wheels and waste material and thus
limit the potential spreading of contamination material to the facility
roadways. In this fashion the source extent of contaminated material is
effectively reduced.
An example of a mitigative measure involves the application of water
to unpaved travel surfaces in order to suppress the entrainment of gross
or contaminated particulate by vehicle traffic. The important point here
is that relative to contaminated particulate, it is assumed that the
roadways are characterized by some (albeit unknown) level of contamination
(a) associated with long-tern waste disposal operations, and that the
objective is to minimize the entrainment of this material into the
atmosphere.
383
-------
References
1.
2.
3.
4.
C. Cowherd, P. Englehart, G. E. Muleski, J. S. Kinsey, "Hazardous
Waste TSOF Fugitive Particulate Matter Air Emissions Guidance
Document," EPA-450/3-89-019, prepared for the U.S. Environmental
Protection Agency, OAQDS, Research Triangle Park, NC (May 1989).
P. Englehart, D. Wallace, "Assessment of Hazardous
Particulate Emissions," Final Report, EPA Contract No.
Work Assignment Nos. 5 and 13 (Oct. 1986).
Waste TSDF
68-02-3891,
U.S. Environmental Protection Agency, "Compilation of Air Pollution
Emission Factors (AP-42)," Fourth Edition, September 1985; Supple-
ment A, October 1986; Supplement B, September 1988. Research
Triangle Park, North Carolina (Sept. 1988).
C. Cowherd, G. E. Muleski, P. J. Englehart, D. A. Gillette, "Rapid
Assessment of Exposure to Particulate Emissions From Surface Con-
tamination Sites," EPA/600-8-85/002, prepared for U.S. Environmental
Protection Agency, Office of Health and Environmental Assessment,
Washington, D.C. (Feb. 1985).
TABLE I
APPLICABILITY OF AP-42 EMISSION FACTORS
Emission model
TSDF source
Paved/unpaved
roads
Unpaved
roads
X
Paved
roads
X
Mater ia Is
handl ing
Ind. wind Dozer
erosion operations
Land
Open waste
piles/staging
areas
Dry surface
impoundments
LandfilIs
Land treatment
Waste stabiI-
ization
384
-------
A METHODOLOGY FOR IDENTIFYING AND RANKING TOXIC
AIR POLLUTANTS AT SUPERFUND SITES
Emile I. Boulos
U.S. Environmental Protection Agency
Office of Emergency and Remedial Response (OERR)
Hazardous Site Evaluation Division (HSED)
Analytical Operations Branch (AOB)
Washington, DC 20460
Bruce S. Carhart and Andrea J. Randall
Engineering-Science, Inc.
Fairfax, VA
Donald L. Decker
Engineering-Science, Inc.
Gary, NC
ABSTRACT
This paper describes a methodology for identifying toxic air pollutants of concern at
Superfund sites, as a preliminary step to developing an air toxics Statement-of-Work for
USEPA's Contract Lab Program. The methodology ranks the toxic air pollutants of
concern in order of relative importance, as determined by frequency of occurrence at
Superfund sites and potential for threat to human health.
A master list of 260 potential toxic air pollutants was formed from USEPA's
Hazardous Substances Priority Lists and other authoritative lists. Primary criteria for
selection of available data describing the target compounds were toxicity, carcinogenicity
and potential for human exposure at or in the vicinity of Superfund sites.
Secondary criteria included consideration of regional, state and local regulatory needs and
availability of existing analytical methods and reference standards. In evaluating the
acceptability of candidate analytical methods, a reference ambient level was determined for
a number of target compounds lacking unit risk estimates. Data describing these criteria
were arrayed in a computer spreadsheet and processed in a ten-term algorithm, to arrange
the toxic air pollutants in order of priority. Each algorithm term described one of the
primary or secondary criteria and was assigned a relative weight by USEPA.
385
-------
INTRODUCTION
As specified in the Comprehensive Environmental Response, Compensation and
Liability Act (CERCLA) and the Superfund Amendments and Reauthorization Act
(SARA), EPA has the responsibility for assessing the potential for air emissions and air
quality impacts prior to and during Superfund hazardous waste site cleanup. CERCLA
and SARA mandate the characterization of all contaminant migration pathways from waste
to the environment and of the resulting environmental impacts. In particular, a remedial
investigation must provide data on air emissions from the site in undisturbed and disturbed
states. CERCLA and SARA also require the development of data that are "necessary and
sufficient" to characterize the "nature and extent" of contamination on site. Consequently,
regulatory agencies, site managers and Remedial Program Managers (RPMs) must develop
site-specific objectives in characterizing air emissions at these sites.
To address the absence within EPA of standardized sampling and analytical
procedures for characterizing air emissions from pre-remedial/removal activities at
Superfund sites, EPA's Office of Emergency and Remedial Response (OERR) is developing
a "Statement-of-Work for Analysis of Air Toxics at Superfund Sites."
One of the early, tasks in the development of the Statement-of-Work is the
preparation of the target compound list (TCL), a prioritized list of compounds for which
analytical protocols are to be included in the Statement-of-Work. The TCL comprises
compounds most commonly found at Superfund sites and which pose the most significant
threat to human health, and which are likely to enter the ambient air at Superfund sites.
Many of the target compounds do not have EPA published unit risk values. Unit risk
values are significant because they allow the calculation of ambient levels which relate to
any specific risk level of concern (e.g., 10"6) for carcinogens, data which in turn predict the
required level of analytical detection limits for many of the analytical data uses. If
substances are to be the target of sampling activities at NPL sites, it is clearly desirable to
have sampling and analytical methods which are capable of detecting concentrations of
substances at these levels. To compensate for the lack of unit risk estimates for many
target compounds, Reference Ambient Levels (RALs) were developed to represent the
lowest ambient concentration of concern at Superfund sites.
DEVELOPMENT OF A RANKED TARGET COMPOUND LIST
Since no generally accepted list of air toxics at Superfund sites is available, the first
task in the development of the ranked TCL is preparation of a master list of candidate
compounds. The Hazardous Substances Priority Lists* were selected as the starting point
in preparing this list.
'CERCLA requires that the USEPA and the Agency for Toxic Substances and Disease Registry jointly list in order of priority, hazardous
substances which are most commonly found at NPL facilities and which the agencies determine are posing the most significant potential
threat to human health.
386
-------
It was necessary to augment the Hazardous Substances Priority Lists of 200 of the
most hazardous substances found at Superfund sites because these lists were compiled with
only minimal consideration of the air pathway. Accordingly, 60 air toxics compounds were
selected from other authoritative lists.
After the master list was compiled, a straightforward method was developed to rank
these compounds in order of importance as air toxics at Superfund sites. General
considerations in developing the ranking scheme were:
• Frequency with which compounds were mentioned in a survey of EPA
Regions.
• Frequency with which compounds have been found at Superfund sites.
• Compounds which present a risk of exposure by inhalation and which are
highly toxic or carcinogenic.
• Frequency with which compounds are requested under Federal or state
regulations (ARARs and TBCs) to meet cleanup goals.
• Availability of sampling/analysis methods and reference standards for the
TCL.
• Relative volatility of the candidate compounds.
• Availability of health-based data (unit risk, reference doses and acceptable
ambient levels) for the TCL.
These criterias can be grouped into three major areas; health effects, regional needs,
and potential for human exposure.
HEALTH EFFECTS
Health effects were addressed in detail through a variety of descriptive indicators,
including unit risk factors, cancer potency slopes, and reference dose values, and other less
specific descriptors in cases where data of primary quality were not available.
In considering health effects we used available data developed by USEPA's Pollutant
Assessment Branch (PAB) and generally contained in the list of unit risk factors for the
inhalation of carcinogenic air contaminants. This list is maintained by PAB for air
assessments performed with EPA's Office of Air Quality Planning and Standards (OAQPS).
The PAB also maintains a separate list of compound "cancer potency slopes" which in most
c;ases are based upon ingestion routes of exposure. Because in many cases these cancer
potency slopes have been and will continue to be converted to inhalation factors for use
in air toxics risk assessments, we included these data in our assessment and ranking of
health effects. For non-carcinogens we referred to lists maintained by EPA's non-
carcinogen workgroup. These are compounds for which EPA has determined a need for
the development of "reference dose" (RfD) values. Reference dose is used by USEPA as
threshold value in evaluating non-carcinogenic health effects.
387
-------
For other compounds on the list not described by any of these data, various health
effects indicators such as threshold limit values and reportable quantity data from SARA
Title III were used. In all cases, the end product of the health effects data assessment was
a health effects ranking number between one and ten, so that each compound on the
master list could be ranked on a common basis. Health effects was designated the most
important of the ten ranking criteria, and was accordingly most heavily weighted in the
ranking scheme.
REGIONAL NEEDS
In assessing regional needs for sampling guidance and analytical methods for specific
air toxics, we relied on responses to the USEPA Regional air toxics survey conducted in
March, 1989. Many regions provided lists of important air toxics compounds; frequency
with which specific compounds appeared on these various lists was the second most
important ranking criteria.
Of next major importance as an indicator of regional needs is the regulation of
specific air toxics compounds by the various states. Since the limited scope of the project
precluded an in depth review of all applicable state regulations, we relied on the data base
developed by the National Air Toxics Information Clearinghouse as an indicator of state
regulatory activity for specific air toxics chemicals. For various states regulating on the
basis of acceptable ambient levels (AALs), frequency of occurrence of regulations for
specific chemicals was the third most important ranking criteria.
Frequency cf occurrence on various credible lists of hazardous materials was also
considered to be a useful ranking indicator. The California Air Resources Board (CARB)
publishes a "Lists of Lists" which shows the frequency with which specific chemicals are
listed in 12 authoritative lists of hazardous chemicals. The New York Air Guide I also
categorizes specific air toxics compounds as high, medium, or low toxicity. SARA Title III,
Section 302 also lists hazardous pollutants. Frequency of occurrence in each of these lists
was used as an indicator of relative importance of these compounds, occupying the 4th, 5th,
and 8th positions in order of importance of ranking criteria.
POTENTIAL FOR HUMAN EXPOSURE
Indicators for the potential for human exposure were incorporated by considering
both the frequency of occurrence at Superfund sites and the volatility of each of the listed
compounds. Frequency of occurrence at Superfund sites was obtained directly from the
August 1988 list entitled "Frequency Distribution of Substances Present at Final and
Proposed NPL Sites." Each target compound was assigned a volatility ranking number
between 0.5 and 3, derived from boiling point and/or vapor pressure data as available.
These indicators are generally considered to represent potential for human exposure
through the air pathway at Superfund sites and were assigned the 6th and 7th positions in
the ranking scheme.
388
-------
AVAILABILITY OF ANALYTICAL METHODS AND REFERENCE STANDARDS
The 9th and 10th weighing positions were assigned to availability of analytical
method and standard, respectively. We felt that for marginal compounds the availability
or lack of analytical methods and reference standards could be a factor in the decision to
include or exclude such compounds from the priority target compound list.
In evaluating the suitability of available anaytical methods we decided to compare
method detection limits with ambient levels of concern determined by unit risk values and
an acceptable risk of 106.
Many of the substances of the 257 on the original TCL do not have EPA published
unit risk values. To provide a comparable measure of ambient level of concern for these
target compounds, we devised the reference ambient level (RAL). RALs were designed
to be an approximation of potential Applicable or Relevant and Appropriate Requirements
(ARARs) or "To-Be-Considered" materials (TBCs) in establishment of air cleanup standards
for remedial actions at National Priority List (NPL) sites. If substances are to be the target
of sampling activities at NPL sites, it is clearly desirable to have sampling and analytical
methods which are capable of detecting concentrations of substances at these levels.
The concentration-based levels of concern were obtained from state and local air
toxics programs. There has been much controversy over appropriate methodologies for
determining acceptable ambient levels (AALs). Many states use Threshold Limit Values
(TLVs) divided by a safety factor, despite the fact that the publisher of TLVs, the American
Conference of Government and Industrial Hygienists, explicitly recommends against the use
of TLVs for such an application. Nevertheless, under USEPA's developing guidance for
ARARs and TBCs, any state regulation or guidance potentially could be used as the basis
of an air clean-up standard, regardless of the basis of that regulation or guideline.
Accordingly, it is important to know what states currently are using as AALs to determine
what sampling and analytical methods are needed at NPL sites. Having an evaluation of
current practice for AALs appears to be the best starting point for determining an
appropriate approach for recommending practical sampling and analytical methods.
Current practice for AALs, however, is not uniform from state to state. In order to
focus this effort, we decided to concentrate on states with the most sites on the NPL. Each
state was contacted for documentation on their methodology for developing AALs, along
with a listing of pollutants and applicable AALs (if one exists). Some of the states provided
methodologies only, and calculations of AALs based on designated input data (such as
TLVs) was necessary.
AALs were then tabulated by averaging time, and if more than one state regulates
a single pollutant for the same averaging time, AALs were presented as a range. No
conversion factors were used to modify AALs from one averaging time to another, because
applicable ARARs and TBCs arguably must use.the state or local guideline as it exists.
RALs based on the lowest concentration of any lO'* concentration, modified RfD, or AAL
389
-------
were then tabulated.
In all cases, if an Acceptable Air Risk (AAR) was given, it was selected as the RAL.
If no AAR value was available, the most restrictive value represented by the states and the
corresponding Reference Dose (RfD) value was selected. In cases where neither an AAR
nor RfD was available, the most restrictive state value was selected as the RAL.
Information used to develop the RAL table was obtained from the states of
California (Bay Area), Connecticut, Indiana, Maryland, Massachusetts, Michigan, New York,
and Wisconsin. The Reference Doses (RfD) were taken from USEPA's IRIS system.
DESCRIPTION OF RANKING PROCESS
To complete the ranking process, each of the candidate chemicals on the expanded
master list was entered into a Lotus 1-2-3 spreadsheet and arrayed with corresponding
numerical data describing each of the ten ranking criteria. An algorithm was devised which
would position the maximum value of each of the ranking criteria terms in its relative
weighted position. The algorithm, the ranges of the numerical data in the spreadsheet, and
the relative maximum term values are shown in Figure 1. For example, AOB designated
Health Effects as the most important of the descriptive criteria and was accordingly ranked
number one (1); Availability of a Reference Standard was the least important, ranked
number ten (10). These rankings are shown in Figure 2, column 2. For convenience in
developing a ranking index (RI) algorithm, the ten (10) ranking parameters were assigned
corresponding maximum values of the algorithm terms representing these parameters. For
Health Effects, a maximum term value of 100 was selected, while for Availability of
Reference Standard a maximum term value of 30 was selected. Intermediate term values
for other descriptors were selected so that each algorithm term representing a descriptor
was retained in the relative positions specified by AOB.
Data in the Lotus 1-2-3 spreadsheet were arrayed in various ways for the various
descriptors. For Health Effects, each chemical was assigned an index value between 1 and
10, ten being the maximum value of the health effect descriptor and representing the
greatest level of health hazard. For Availability of Reference Standard, each chemical was
assigned an index value between 0 and 2, 0 representing no standard available and 2
representing availability of an EPA-certified standard. The algorithm constant was derived
by dividing the maximum term value by the maximum numerical index value arrayed in the
spreadsheet. For health effects, the algorithm constant was 10 (100/10 = 10), while for
Availability of Reference Standard, the algorithm constant was 15 (30/2 = 15).
As shown in Figure 1, the ranking index was designated as the sum of the descriptor
terms, with a maximum summed value of 645. Figure 2 illustrates the scoring for vinyl
chloride, the highest ranked chemical on the TCL with a score of 515. Table 1 lists the
data sources used in development of the target compound list.
390
-------
CONCLUSIONS
Although no absolute measure of the validity of the ranking methodology is
available, our preliminary conclusion at this stage is that the list adequately predicts the
range of compounds likely to be encountered at Superfund sites. We also feel that it
represents the best compromise to providing a target compounds list which addresses health
effects, regional needs, potential for human exposure, and regulatory requirements. By
considering unit risk estimates and RALs in comparison with analytical detection limits
for priority target compounds, it was possible to select appropriate general analytical
methods for nearly all target compounds.
As greater experience is gained in analysis of air toxics at Superfund sites, changes
to this target compound list and possibly to the general analytical protocols are expected.
We are, however, confident that we can move forward in development of the Statement-
of-Work with the expectation that the initial Statement-of-Work for Superfund site air toxics
analysis will adequately address the program requirements.
DISCLAIMER
Although the work described in this paper has been funded wholly or in part by the
United States Environmental Protection Agency through Contract No. 68-02-4398 to
Engineering-Science, Inc., it has not been subjected to Agency review and therefore does
not necessarily reflect the views of the Agency, and no official endorsement should be
inferred.
391
-------
TABLE 1
LIST OF DATA SOURCES
FOR DEVELOPMENT OF TARGET
COMPOUND LIST
EPA Health Effects Summary Tables
PIPQUIC (IRIS) Data Base
March '89 Survey Responses
New York Air Guide II
California Air Resources Board List of Lists
SARA Title III Section 313 List
(and Reportable Quantity Table)
National Air Toxics Information
Clearing House Report on Air Toxics
Activities
Frequency Distribution of Substances Present
at Final and Proposed NPL Sites
Superfund Public Health Evaluation Manual
USEPA Technical Guidance for Hazards Analysis
CRC Handbook of Chemistry & Physics
Merck Manual
392
-------
CO
w
FIGURE i KAJUU.NCi INDEX (Rl)
ALGORITHM AND DERIVATION OF TERM VALUES
RI = 10G + 11.3B + 120M + 7.5D + 23.3F + 10K + 20L + 40E + 35C + 15J
Data
Descriptor
Health Effects Index
Number of Listings in
Survey Responses
Frequency of Listings
in State AAL Regs.2
Number of Entries in
GARB List of Lists
Category in NY Air Guide
Frequency of Occurrence
at NPL Sites3
Volatility Index
Listing in SARA Title III
Availability of
Analytical Method
Availability of
Reference Standard
Source
Rank1
1
2
3
4
5
6
7
8
9
10
Selected
Max. Value
of Descriptor
(a)
100
90
80
75
70
65
60
40
35
30
Lotus 1-2-3 Spreadsheet
Ranae of Numerical Data
Algorithm
Variable
G
B
M
D
F
K
L
E
C
J
Min.
Value
1
0
0
0
0
0
0.5
0
0
0
Max.
Value
fb)
10
8
0.667
10
3
6.5
3
1
1
2
Derived
Constant
for
Algorithm
fa * b)
10
11.3
120
7.5
23.3
10
20
40
35
15
Maximum value of Algorithm
645
relative importance of descriptor, as specified by AOB, HSED, OERR
numerical data expressed as decimal fraction
numercial data expressed as decimal fraction x 100, for multiple listings only
-------
FIGURE 2. SAMPLE CALCULATION
FOR VINYL CHLORIDE
Value* of
Variable
in Algorithm
Descriptor
Vinyl Chloride
Term Value2
6
B
4
8
M = 0.5
D = 10
F
K
L
E
C
3
1.51
3
1
1
J = 2
Health Effects index
Frequency of Listing in
Survey Responses
Frequency of Listing
in State AAL Regs.
Number of Entries in
GARB List Of Lists
NY Air Guide List
Frequency of Occurrence
at NFL Sites
Volatility Index
Listing in SARA Title III
Availability of Analytical
Method
Availability of Reference
Standard
Total Algorithm Value
for Vinyl Chloride
40
90
60
75
70
15
60
40
35
30
515
RI = 106 + 11.3B + 120M + 7.5D + 23.3F 4- 10K + 20L + 40E +
35C + 15J
= 10(4) + 11.3(8) + 120(0.5) + 7.5(10) + 23.3(3) +
10(1.51) + 20(3) + 40(1) + 35(1) + 15(2)
=40+90+60+75
30
+ 70 + 15 + 60 + 40 + 35 +
= 515
1 values for vinyl chloride terms from Lotus 1-2-3
spreadsheet
2 rounded off to nearest whole number
394
-------
AIR TOXICS CONSIDERATIONS
AT AN ACTIVE
FIREFIGHTER TRAINING FACILITY
Michael J. Barboza, P.E.
Malcolm Pirnie, Inc.
100 Eisenhower Drive, P.O.Box
Paramus, New Jersey 07653
36
Catherine Bobenhausen, CIH
Malcolm Pirnie, Inc.
2 Corporate Park Drive
White Plains, New York 10602
This paper describes the concerns, study approach and the results of air sampling
at an active firefighter training facility. Air sampling was performed of the
buildings, mockup structures, and gas wells (installed near mock up structures)
during live burns for parameters including VOCs and combustible gas,
The concerns related to the potential for vapor emissions from site contamination
and methane from an adjacent landfill. Groundwater contamination was likely a
result of previous use of dry wells. During previous site investigations,
combustible gas was detected in groundwater wells, triggering concern for the
safety of on-site personnel, and the safety of firefighters using the training
facility.
The study was developed to evaluate the presence of gases and vapors in the
buildings and the potential for any effects of live burns on gas and vapor
levels. The air sampling program included organic vapor collection on adsorbent
tubes, and use of combustible gas meters, photoionization detectors, and a
portable GC. Since site conditions and concerns may be representative of fire
training facilities throughout the country, the air sampling approach utilized
1s applicable to similar studies.
395
-------
INTRODUCTION
Fire training facilites provide a valuable service to communities throughout
the country by providing a means of training and exposing firefighters (in a
controlled setting) to the harsh and dangerous realities of fighting fires.
Unfortunately some of these facilities may be on sites contaminated from
spillage of waste liquids or loss of fuels through inadeqaute drainage
systems. The presence of contamination may pose a threat to air quality and
a potential hazard to personnel who occupy and use such facilities.
This paper describes the concerns, study approach and the results of air
sampling at an active firefighter training facility. Air sampling was
performed of the buildings, mockup structures, and gas wells {installed near
mock up structures) during live burns. Parameters of interest included
selected volatile organic compounds (VOCs) and combustible gas.
BACKGROUND
The site houses fire training and administrative functions. The facility is
occupied by full and part time administrative and maintenance personnel and
periodically (during training sessions) by instructors and students from over
50 volunteer fire districts. Advanced training consists of classroom and
controlled live burn exercises.
Structures on the 10 acre plus site include a number of buildings i.e.,
administration, first aid, classroom, pumper test, maintenance and pump house.
In addition training facilities include building mockups (dwelling,
commercial, and tower) and fire extinguisher, open pit, tower/ladder, and
propane training areas.
Adjacent to the site is a municipal solid waste landfill that has landfill
gas control systems (i.e., recovery, flaring and venting) that extract gas
from the landfill and burn it for power generation, flare it for disposal and
vent it to control subsurface migration, near the fire training site.
Ground water investigations of the site indicated the presence of floating
product (chiefly No. 2 fuel oil) resulting from past use of dry wells. Trace
contaminants related to fuels and solvents detected in ground water were the
focus. These included benzene, toluene, xylenes, carbon disulfide, vinyl
chloride, methylene chloride, 1,1-dichloroethene, 1,1-dichloroethane, trans-
1,2-dichloroethene,1,1,1-trichloroethane,trichloroethene,tetrachloroethene,
chlorobenzene, ethylbenzene, 2-butanone and acetone. Concern was raised when
significant levels of combustible gas were detected in several of the ground
water monitoring wells.
Maintenance personnel at the site were reported to have been exposed to
noxious gases, suspected as originating from the landfill. Other reports
indicated occasional instances of unusual fire conditions in the mock up
buidings during training.
Concern over potential hazards associated with air quality onsite led to
questions related to safe operation of the facility, i.e.,
- are toxic vapors or gases present at the facility at potentially
harmful levels;
- can explosive conditions arise from combustible gas accumulation
in the buildings; and
396
-------
- is there migration of potentially harmful or dangerous vapors or
gases from the ground into the mockup buildings during fire train-
ing exercises?
APPROACH
The diversity of the concerns and the different variables associated with
routine site activities called for an approach consisting of careful planning
and clearly identifying the study objectives (1). The identified objectives
of this study were to:
- quantify airborne concentrations of volatile organic contaminants
(VOCs) during various site activities, and assess potential hazards to
personnel and firefighters;
- develop a program for continued operation of the facility while
reducing exposures, consisting of a combination of appropriate engineering
controls, administrative procedures and/or continuous site monitoring; and
- establish a framework for emergency and contingency planning.
A number of tasks were performed to meet these objectives, that included:
- air quality monitoring with various meters (oxygen, combustible
gas, photoionization and organic vapor detectors);
- air quality screening with a Photovac field gas chromatograph;
- air sampling with adsorbent tubes with laboratory analysis for
VOCs;
- installation of gas monitoring wells;
- measurement of differential well pressure and temperature during
fire training activities and screening the constituents of soil
gas;
- observation of administrative, classroom, and fire training
activities; and
- interviewing of various site personnel regarding standard
procedures and site history.
Grab (air) samples (almost real-time) were analyzed for six VOCs using a
Photovac {GC Model 10S50) portable gas chromatograph onsite to monitor air
quality over a 72-hour period. The calibration gas used with the Photovac GC
consisted of 1 ppm each of vinyl chloride, methylene chloride, 1,1-
dichloroethene, trichloroethene, benzene and tetrachloroethene, based upon
presence in ground water and relative toxicity.
In addition, over thirty samples were collected with tenax/carboxen 569 (a
Supelco product consisting of carbonized molecular sieve) adsorbent tubes at
flowrates of 80 cc/min for representative sample periods. Samples were
collected during normal administrative work hours, as well as during the
evening training sessions involving live burns within mock-up structures.
Analysis was by GC/MS for volatile organic compounds including a library
search for non-target compounds, using EPA Methods T01 and T02 (2).
397
-------
Temperature and pressure measurements were made using temperature probes, an
inclined manometer and magnahelic gauges. Observations were made of the
potential effects of barometric pressure on possible gas movement. Significant
changes in well pressure difference may suggest the occurence of a chimney
effect in the mockups during burns, potentially causing gas migration from the
ground into the mockups.
RESULTS
Results indicated that, while there was evidence of VOCs in ambient air at
the site, the concentrations found were in the low ppb range, and were not
considered an immediate threat to site personnel. Table 1 summarizes the
maximum values detected. The compounds detected in the air were similar to
those found in soil and ground water on the site and some are typical of
landfill gas. There is often difficulty in assessing the results in regard to
appropriate safe levels for VOCs especially for samples of indoor air (3).
The levels found were below IDLH levels, OSHA PELs, and ACGIH TLVs. However,
given the presence of subsurface gas and the potential controlling effect of
the landfill gas extraction systems, there is potential for air quality
deterioration. Some of the highest values of VOCs were detected in the
maintanence garage.
Sampling in the mockups during the live burns with adsorbent tubes was limited
due to the harsh environment (high temperatures, water, heavy smoke and
combustion products) within the mockups during live burns. Low {generally less
than 10 ppb) concentrations of seven VOCs were detected at this location.
Differential pressure measurements in gas monitoring wells near the mockups
during fire training exercises showed no significant variability attributable
to fire training activities involving live burns. A pressure gradient was
detected from measurements in gas monitoring wells with greater negative
pressures near the landfill boundary and lower negative pressures measured
near the mockups, due to the influence of the landfill gas extraction systems.
The landfill gas extraction systems were likely influencing movement of
subsurface gases at least as far from the landfill as the mockup buildings,
as indicated by negative pressure measurements in wells near the mockups
during normal operations (i.e., when live burn exercises were not ongoing).
DISCUSSION
Under normal conditions, measured VOC levels were generally low, but results
were obtained from only a few discrete periods, when the landfill gas
extraction systems were operating. Periodic air sampling was recommended to
observe any changes in contaminant concentrations over time (i.e., months,
years).
The landfill gas control systems were shown to exert subsurface influence on
the site at least as far as the mockup buildings, likely controlling vapors
on the site (via the subsurface). Increased concentrations of gases could
result if the landfill gas control systems failed or were operated with
reduced effectiveness. Some of the risks involving this situation are
attributed to the uncertainties associated with variations in system
performance on gas and vapor migration patterns, atmospheric pressure effects
and also on what effect duration of upsets would have on the gas migration.
398
-------
RECOMMENDATIONS
The greatest risk appeared to be associated with the uncertainties and
possible variabilty in the effectiveness of the landfill gas control systems,
it was recommended that air monitoring be performed along the perimeter of the
site and at some specific locations onsite.
An automated detection system was recommended for this consisting of remote
sensors installed at key locations, a central display and recording system to
provide easy and continual identification of the status of hazard conditions,
and warning devices (alarms/lights) to alert personnel of potentially
hazardous conditions. The primary constituent to monitor would be combustible
gas. The output of the monitoring system would be used to curtail or restrict
activities in parts of the site or suspend operations until the alarm
condition subsides and/or the cause identified and the hazard evaluated.
The results from the monitoring system may be used to curtail or restrict
activities in parts of the site or suspend operations unitl the alarm
condition subsides and/or the cause identified and the hazard addressed.
In addition to monitoring, the following procedures {some of which had already
been implemented) were recommended to minimize potential risks:
- Check mockup buildings for combustible gas and organic vapors before
and during any maintenance work, especially any welding or similar activity.
Suspend work if measurements exceed prescribed action levels;
- Vent mockups when not specifically in use for fire training and during
maintenance and recharging activities;
- Monitor wind direction/speed with a fixed system with recording
devices for historical record and also with wind socks at various locations
visible from different areas of the site;
- Encourage use of communications systems (i.e., radios) for personnel
working on different parts of the site;
- Encourage reporting and documentation of monitoring and the occurrence
of unusual events (ie accidents, incidents, fires, odors, gases, etc.)
- Establish a safety organization to address safety issues of site
contamination and enforce the operational safety plan which was developed
based upon the findings of this study.
CONCLUSIONS
The study involved unusual site conditions, diverse concerns requiring a
monitoring program that was planned and designed, to address the major
concerns; and to provide recommendations for mitigation and monitoring. Since
immediate hazards were not detected the resultant concern was for potential
hazards due to changes in the situation associated with the landfill gas
extraction systems; therfore the primary recommendation was to continue
monitoring for indicator parameters that would provide an early warning of
changes in the onsite gas situation.
399
-------
REFERENCES
1 Barboza , M.J., and T.T. Shen. Planning Air Monitoring for VOCs at Waste
Sites. Presented at EPA/APCA Symposium on Measurement of Toxic and Related
Air Pollutants, Raleigh, N.C. May 2-4, 1988.
2 Winberry, W.T., N.T. Murphy and R.M. Riggin. Compendium of Methods for
Determination of Toxic Organic Compounds in Ambient Air. EPA-600/4-89-
017, U.S.Environmental Protection Agency, Research Triangle Park, N.C.,
June 1988
3 Bobenhausen C. and M.J. Barboza. Action Levels for IAQ Contaminants. HazMat
88, Atlantic City, N.J. June 1988.
Table 1 SUMMARY OF MAXIMUM CONCENTRATIONS MEASURED
COMPOUND MAX CONCENTRATION (PPB)
ACETONE 63.3
ACRYLONITRILE 18.7
BENZENE 48.1
2-BUTANONE 57.2
CARBON DISULFIDE 4.3
CHLOROBENZENE 0.9
CHLOROFORM 1.6
CHLOROMETHANE 47.4
1,2-DICHLOROETHENE 5.2
ETHYLBENZENE 28.4
2-HEXANONE 638.1
4-METHYLPENTANONE 66.8
METHYLENE CHLORIDE 12.3
STYRENE 86.1
1,1,2,2-TETRACHLOROETHANE 4.6
TETRACHLOROETHENE 5.3
TOLUENE 139.0
1,1,1-TRICHLOROETHANE 45.1
TRICHLOROETHENE 12.9
XYLENES 289.6
400
-------
Tiaverse-Monitoring for Superfund Sites
A. J. Cimorelli
Air Programs Branch
T.A. Casey
Air Programs Branch
R.L Smith
Superfund Programs Branch
EPA Region ill
841 Chestnut Street
Philadelphia, PA 19107
Awareness is increasing that activities related to the clean-up of Superfund sites can lead to
air emissions having significant effects off-site. Because these emissions are often unplanned or
temporary, project managers are frequently reluctant to dedicate resources to what is thought to be
an unlikely public health concern. Without further information fitting the monitored concentrations into
the structure of the plume, a single monitor at fenceiine is, in most cases, inadequate. A single
measurement of concentration, without an understanding of where it resides in the plume distribution,
d<:>es not enable an adequate estimation of downwind dilution.
This paper introduces an inexpensive technique- entailing the use of hand-held monitors, on-
site wind data, and a priori modeling and toxicological analyses- which enables on-site personnel to
make informed decisions concerning evacuation and other potential response to ground-level emissions
from Superfund sites. Additionally, a method of bridging the gap between information and decision-
making is offered.
V\fhite the techniques offered in this paper may be used to fulfill EPA policy, they are presented here
for informational purposes only. This paper is not to be construed as official EPA policy or guidance.
Mention of brand-names HI no way constitutes endorsement
INTRODUCTION
Activities related to the remediation of Superfund sites can lead to air emissions having
significant impact off-site. During the Fall of 1989, the authors of this paper were asked to develop a
contingency plan as part of an effort to protect public health in case of accidental chemical spills during
remediation activities at the Maryland Sand, Gravel, and Stone Superfund site. This paper describes
a technique enabling on-site personnel to make informed judgments concerning evacuation and other
response to an accidental spill of volatile organic chemicals. Not surprisingly, the crux of this plan
involves quantifying, in a timely and accurate manner, the parameters important to decision-making.
This is accomplished by performing as much a priori analysis as possible and delineating straightforward
procedures for the gathering of necessary information. The technique developed for Maryland Sand
included two novel concepts: the integration of modeling with traverse-monitoring to assess impacts,
and the development of Response Indices explicitly linking decision-making with data collection and
analyses.
Specifications
To be useful, the technique must (1) enable decision-making to be based on estimates of risk
that are neither overly conservative nor non-conservative; (2) provide real-time information; (3) not
represent an undue hardship to the potentially responsible party in terms of resource requirements. The
401
-------
development of this technique can be divided into two tasks. First, it is necessary to determine what
decision needs to be made to made and what criteria are to used to make it (action levels). This
requires the identification of information that will be important in the decision-making process before the
procedures for gathering this information can be developed. The second task is the development of
a means of data collection and analysis to supply the needed information.
Bases (or Decision-Making
Three factors were quickly identified as directly impacting decision-making: pollutant
concentration at sensitive receptors (as opposed to an arbitrary point along the fenceline), chemical-
specific toxicity, and duration of exposure.
The next step was to bridge the gap between data collection/analysis and on-site decision-
making. This was establish by developing new quantities called Response Indices. We developed
these Indices to employ toxicological and atmospheric analyses to relate measurable on-site phenomena
with proper responses. Each of the Response Indices was a function of factors identified above such
that
Rl = f(contaminant concentration, toxicity, duration).
Next, we developed methods of data collection and analysis to quantify each parameter.
DATA COLLECTION AND ANALYSIS
Estimating Concentrations at Downwind Receptors
Before the traverse-monitoring method was developed, on-site monitoring and modeling were
considered for use in estimating impacts. It was soon apparent that these techniques were, for various
reasons, insufficient. Brief descriptions of these alternatives are offered below.
On-site Monitoring
It is common practice to place a monitor or two at the fenceline of a site with the expectation
that off-site impacts will be quantified. While a monitor can offer reliable, real-time information
concerning pollutant concentrations at a single place and time, this information alone, without fitting the
monitored concentration into the structure of the plume, is inadequate to quantify impacts elsewhere.
(This is true even for the ground-level sources under consideration here.) Figure 1 shows three
instances in which a monitor at the fenceline of a site would record identical concentrations, even
though impacts off-site are quite different for each case. Without knowledge of the source-strength, the
width of the plume, and where the monitored concentration fits within the concentration distribution of
the plume, downwind impacts may be underestimated.
Modeling
Air dispersion modeling can often provide information that monitoring cannot. Unlike a single
monitor, a model applies theoretical and empirical knowledge of how pollutants are transported and
diffused through the air to estimate impacts. Additionally, models are quite able to predict the maximum
impacts that can be expected from a given source of emissions. Monitors cannot make predictions,
especially of the maxima that are relevant for most regulatory purposes.
For the present application, however, modeling has several drawbacks. First, modeling requires
both source and site specific information: One must characterize the source (emission rate, size, and
location relative to receptors) and the atmosphere (wind speed and direction and atmospheric stability).
These requirements make (near) real-time estimates of impact impossible without collecting real-time
402
-------
meteorological data Second, in the event of accident, such as the rupturing of a drum with unknown
contents, the source strength will be unknown, making the estimation of impacts impossible. Finally,
the screening procedures that are often used to simplify modeling exercises are not applicable to this
case. Over-conservatism is likely to induce overreaction, including needless evacuation, and the
{invoking of unwarranted measures may not only cause unnecessary inconvenience and trauma, but also
diminish the sense of urgency and when, actual emergencies develop.
Figure 1
El
C
Tliis figure describes three vary different situations when a single monitor at fencellne would show the same impacts, showing
that downwind impacts cannot be predicted with a single, stationary monitor. In this case, emissions from source A an less
than those from source B, which are, in turn, loss than those from source C.
Traverse-Monitoring
The traverse monitoring technique is a marriage of monitoring and modeling developed to
exploit the strengths of both methods. It is based upon the incorporation of data reported by a
technician who walks through the plume while carrying a hand-held VOC monitor. The smooth,
Gaussian distribution that is often used to characterize the distribution of a pollutant within a plume
exists only as a result of the time-averaging of many instantaneous pictures; each of which would be
characterized by some maximum concentration x^. As a rule, x^, is greater than x^.*, the time-
averaged maximum found along the time-averaged centeriine (that is, the average of maxima unpaired
in space); thus x^, can be used to develop conservative estimations of x«vgmw. Quasi-instantaneous
maxima can be measured by traversing the me with an Hnu meter or some other device capable of
continuous or near continuous reports of total VOC concentrations (with very short lag time). In this
way, concentrations can be monitored at a known relative position within the me.
The time-averaged maximum can then be approximated as follows:
n
(1)
i
where: n • the current traverse number (e.g., 50 for the
50th traverse)
403
-------
N = the number of traverses to be averaged
i = n - N
i = instantaneous maximum concentration in the
me at the downwind distance of the traverse
(real-time, on-site measurement).
Thus, Xngmix is updated with each traverse, and the variable n behaves much like a dimensionless time.
There are two factors to be weighed in the selection of an appropriate N. First is response-time
toxicology (What time-frame is important with respect to the chemicals present and the protection of
health, and how much time is necessary to take corrective action? This method is inappropriate where
"single-breath1 concentrations are important). Second is the level of stability required. (The
instantaneous maximum will vary significantly with each traverse; the more traverses that this "running
average' encompasses, the less sensitive it will be to fleeting fluctuations which could induce
overreaction.) Clearly, the role of the toxicologist is important in selecting a suitable N.
The following procedure was recommended for the Maryland Sand, Gravel, and Stone site:
At the commencement of the incident the person who is to perform the traverses will
note the average flow vector of the wind (the direction toward which the wind is blowing)
and position himself 100 meters directly downwind; he will then walk 50 meters in a
cross-wind direction (That is, if he positioned himself by starting at the source of
emissions and walking 100 meters downwind, he will now turn 90 degrees to his right
or left and walk for 50 meters). Now he is ready to traverse the plume. Each traverse
consists of a 100 meter walk (at a deliberate pace) perpendicular to the wind direction
and beginning and ending 50 meters from the plume centerline. During each traverse,
the traverser will monitor his Hnu meter and report the maximum observed concentration
(in ppmv). The time-averaged, centerline concentration is then approximated as in
Equation (1) letting N, in this case, equal 10.
Once the centerline concentration at a known distance is estimated, a Dilution Factor, D, can
be used to estimate concentrations farther downwind. D is a function of wind speed, stability and
distance from the source. The Dilution Factor can be estimated using standard air diffusion models
by predicting centerline concentration near the traverses and at the point of interest downwind. The
Dilution Factor is the ratio of these predictions such that
XH = D * x,^ (2)
where XR is the concentration at the receptor of interest.
The Dilution Factor is estimated as a function of atmospheric stability category and wind
direction (the wind direction dictates if the me will travel in the direction of a nearby home or a distant
one; the locations of sensitive receptors are built into the Dilution Factors, which are direction-specific).
The stability category can be inferred from lateral turbulence in the atmosphere (EPA, 1986) which,
along with wind direction, will be measured on-site. For the case of Maryland Sand, Dilution Factors
were calculated and tabulated for use on-site.
Time-Frame
Time-dependency is incorporated into decision-making in two ways- through the selection N
(number of traverses to be averaged) and through a priori period-of-exceedence minimums attached
to the Response Index (e.g. 'If the Rl remains greater than 1.0 for 20 minutes, then initiate emergency
measures').
404
-------
Toxicology
The lexicological Factor, T, relates the concentrations of individual VOCs at the downwind
receptor to the Response Index, and must be defined in such a way as to make Rl understandable
and useful. Essentially, T encompasses the important lexicological information, such as chemical
composition and exposure limits, that is necessary to evaluate the significance of concentrations of
total VOCs predicted using the Dilution Factor. The units used to describe T should be the inverse
of those used to describe XR so that the Response Indices will be dimensionless.
For our application, T was developed as a function of chemical composition and tolerance
(limits only, and we use an additive response approach to define the lexicological Factor for each
chemical as the ratio of its mole fraction (as a proportion of all VOCs) and its tolerance level and has
units of ppmv1. The overall T is the sum of these ratios:
x
T = z(mole fraction/ tolerance^ over all chemicals (3)
i=1
where x = the number of chemicals detected
Tolerance levels and molecular weights for chemicals suspected to be on-site were tabulated for use
on-site.
Mole fractions must be measured on-site with a gas chromatograph analyzer or similar machine.
Integrated, 5-minute bag sampling for GC analyses can be performed directly downwind of the source.
GC analyses are performed after the first few instantaneous maxima have been found by traversal and,
subsequently, every 30 minutes using a portable GC (Qualitative analyses are updated periodically due
to the non-uniform rate of volatilization of VOCs).
THE RESPONSE INDICES
We are now in a position to quantitatively define the Response Indices as follows.
* D * T (4)
D * T (5)
where: x^, = instantaneous, maximum concentration
across the me at a prescribed downwind
distance (from real-time, on-site
measurements).
~ time-averaged centertine total contaminant
concentration at a prescribed distance
(concentration estimated from real-time,
on-site measurements).
D • Dilution Factor which relates x.^^ to
concentrations of the pollutant at sensitive
receptors farther downwind (XR), s.t. x,, = D *
T • Toxicological Factor which relates estimated
concentrations of individual contaminants at receptors to the Response index,
s.t. Rl, = T * XH.
Note that the Response Index is 'self-normalized' to 1.0 through the assumption of additive response
toxicology.
405
-------
By substituting Equations (1) and (5) into equation (4) above we obtain the time-averaged
Response Index
Rln = (I/NMRU (6).
i=n-N
Through the Response Indices, Rln and Rl,, we may relate events on-site to predetermined, thoroughly
considered responses. For the Maryland Sand site, the following protocol was used.
Traverses will commence if, at any time, a total VOC concentration of 20 ppmv
is measured in the pit. If no VOCs are detectable by the traverser in 30 minutes, then
traversals may be discontinued. If Rl, is not increasing then traversals may cease if (a)
Rl, fails to exceed 1.0 within tfre first 30-minutes, or (b) if Rl, fails to exceed 1.0 in any
subsequent hour. However, traversals may never by discontinued if Rl, is increasing.
CONCLUSIONS
The development of Response Indices to bridge the gap between data collection/analysis and
decision-making is inherently flexible; the parameters can be weighted or entirely different parameters
can be used. A Response Index approach can be helpful in almost any situation that is too
complicated for intuition or common sense or simple guidelines to rule. The exercise of developing
Response Indices has great merit in itself: it forces the explicit examination of the problem for
parameters of importance.
The traverse-monitoring technique has wide applicability and can be used in most situations
were an unknown quantity of volatile organics are emitted from a small, ground-level source. As noted
above, this technique is not applicable in cases where single-breath doses are of concern.
At this juncture, the traverse monitoring technique can benefit from development in two areas.
One is automating the process of manipulating the several sets of data that are necessary to calculate
the Response Index during each traverse. This would consist of a computer program that incorporates
the a priori modeling and lexicological analysis with the data that are collected on-site. A second
means of improvement is a field study to estimate the difference between the monitored center-line
concentration and that which would be estimated using conventional modeling techniques.
Although the traverse monitoring technique has only been employed at the Maryland Sand
site, it could potentially gain great popularity as a convenient, inexpensive method of quantifying off-
site impacts from uncontrolled emissions of volatile organics.
EPA, 1986. Guideline on Revised Air Quality Models (Revised) (EPA-450/2-78-027R) U.S. Environmental
Protection Agency, Office of Air Programs, Research Triangle Park, North Carolina
List, R.J., 1951. 'Smithsonian Meteorological Tables," Sixth Revised Edition, p.497-505, Smithsonian
Institution Washington, D.C.
Turner, 1970. Workbook of Atmospheric Dispersion Estimates. U.S.
Environmental Protection Agency, Office of Air Programs, Research Triangle Park, North
Carolina
406
-------
THE U.S. AND CALIFORNIA
CLEAN AIR ACTS: IMPLICATIONS FOR THE FUTURE
By John T. Ronan III
Sedgwick, Detert, Moran & Arnold
San Francisco, California
The "Decade of the Environment" will bring with it
stringent environmental controls and pervasive regulation
affecting most areas of the national economy.
All of the "easy" technical solutions for environmental
control have been used to meet current environmental laws.
Future controls must necessarily involve restrictions on
manufacturing, transportation and the consumption of energy.
Stringent environmental regulation of the 1990's will
therefore impact upon the economy, personal lifestyle and
politics.
The environmental controls of the 1990's will create a
collision between economic costs, political realities and
environmental benefits. Much of the regulatory activity in
the 1990's will involve the resolution of this conflict.
Federal Clean Air Act Amendments of 1990
Key Provisions
Title 1 - Attainment and Maintenance of Ambient Air
Quality Standards
Non-attainment areas rated Moderate to Extreme. Most
areas are to achieve federal standards by the year 2000 and
the six worst areas by no later than 2010.
Title II - Mobile Sources
Stringent controls on automobiles and other vehicles.
Mandatory use of methanol/ethanol fuels for the worst
non-attainment areas.
Title III - Air Toxics
Stringent controls upon manufacturing plants emitting
air toxics with regulation based upon risk assessment.
Noncompliance will result in forced plant closures.
407
-------
Title IV - Acid Deposition Control
Stringent acid rain controls directed towards 50%
reduction in nationwide sulfur dioxide emissions/ with a
heavy impact upon Mid-western electric power generation.
Some 101 cities have failed to meet the 1987 deadline
for the ambient air quality standard for ozone, the main
consituent of urban smog.
Compliance Deadlines for Non-Attainment Areas;
2010 - Extreme Area (Los Angeles)
2005 - Severe Areas (San Diego, Chicago, Houston,
New York, Philadelphia)
2000 - Serious Areas (Washington, D.C., Boston, Atlanta,
Sacramento, Fresno and 30 other cities)
1995 - Moderate (San Francisco Bay, Santa Barbara and 60
other cities)
Increments of progress;
Initially 4% annual reductions in non-attainment
pollutants.
Later 3% annual reductions until standards achieved.
In severe and extreme areas, additional reductions in
smog-forming chemicals (reactive volatile organics) from
plants emitting over 50 tons a year.
Bus fleets in urban areas to run on natural gas. Car
pool, parking and other transportation restrictions and
controls.
Economic Impacts
Annual Costs
Total cost $21 to $30 billion annually.
Urban Smog Reduction
Total cost from $11.5 to $20.5 billion annually
including $3.0 to $12.0 billion for cleaner automobiles and
cleaner fuel.
408
-------
$2.5 billion for emission reduction from landfills, dry
cleaners, chemical plants and stationary sources.
$1.0 billion in local enforcement programs.
$5.0 billion for state programs.
Air Toxics Reduction
At least $5.5 billion and probably greater depending
upon implementing regulations to be promulgated in the future
Acid Rain Reduction
$4.0 billion for electric utilities to install
scrubbers, to convert facilities and to purchase low sulfur
coal.
California Clean Air Act of 1988
Substantially more stringent than Federal Clean Air Act
Amendments.
Smog-related controls relating to reactive volatile
organics.
Transportation controls.
Consumer products regulation and restrictions.
District wide emissions shall be reduced 5% or more a
year for each non—attainment pollutant or its precursor
(averaged every three-year period) until federal standards
are achieved.
Non-Attainment Districts Are Classified As
Moderate Air Pollution - Attainment of federal standards
by December 31, 1994.
Serious Air Pollution - Attainment of federal standards
by December 31, 1997.
Severe Air Pollution - (Los Angeles) Reduction of
ambient pollution levels by 25% by December 31, 1994.
Reduction by 40% by December 31, 1997. Reduction by 50% by
December 31, 2000, based upon the average ba'se line of 1986
through 1988.
409
-------
Stringent regulation will be imposed upon Districts
which though listed as Moderate, contribute to pollution in
other Districts through air transport of pollutants.
(Example, San Francisco Bay Area.) A District attainment
plan will include allowances for pollution transport both
upwind and downwind of the District.
Hayden-Van de Kamp Initiative on November 6, 1990 Ballot
Greenhouse Gas Reduction Plan "to reduce annual
emissions of any gases which may contribute to global
warming" (in the judgment of the Air Resources Board).
Maximum feasible reductions are to be achieved, with a
mandatory reduction of 20% by January 1, 2000, and 40% by
January 1, 2010, based upon 1988 levels. These percentages
are to adjusted to reflect any difference between the
projected rate of population growth in California and that of
the United States.
Greenhouse gases include carbon dioxide,
chlorofluorocarbons, halons, nitrogen oxide, methane and
any other gases so designated by the State. Gases are to
be controlled in proportion to their respective contributions
to global warming.
Imported electric power generated out-of-state will be
curtailed to the extent that these generating plants
contribute to global warming.
Products whose manufacture contribute to global warming
will be banned and their import into the state prohibited.
The Greenhouse Gas Reduction Plan would be the most
important economic event in California since the "energy
crises" in 1979. Energy prices would increase by $10 to $15
billion annually by the year 2000, based upon California
Energy Commission assumptions. California would be able to
support one million fewer jobs by the year 2000, principally
in the manufacturing and construction areas. Annual energy
prices are estimated to increase from $25 to $45 billion by
the year 2010.
Economic Costs of National Greenhouse Gas Controls
The following economic projections are based upon
economic models by Alan Manne, Stanford University, and
Richard Richels of the Electric Power Research Institute.
The model is based upon the limitation of carbon dioxide
emission to 1990 levels through 2000 and then the gradual
reduction to 20% by 2030.
410
-------
Assuming no replacement technology, the annual cost
would reach $500 billion. If replacement technology were
available, the annual cost would be $50 billion.
Cumulative Costs through 2100 (present value).
Manne-Richels Model
Scenario I - Pessimistic . . . $3.6 Trillion
No practical way to shift to clean energy technologies;
no automatic increase in economic-wide energy efficiency.
Scenario II - Moderate . . . $1.8 Trillion
Automatic energy-efficient adjustment of one percent per
year.
Scenario III - Optimistic . . . $800 Billion
Cost effective substitutes for fossil fuel energy
available; one percent energy efficient adjustment per year.
Conclusions
Clean Air legislation in the 1990's will have a major
impact on the national and state economies, politics, and
personal standards of living.
Existing control technology has been utilized to the
extent available to meet current standards. Future standards
must be met largely by curtailment and restrictions. Future
Clean Air legislation will mandate public and private
transportation controls. Clean Air Act legislation will
reach as far as consumer product regulation. Even if new
technology can be developed, the deadlines imposed by pending
legislation may be too short for adequate new technology to
be developed.
If Greenhouse gas reduction legislation is adopted, a
reduction in energy consumption will be mandated as nuclear
power is currently a politically unacceptable alternative.
Depending upon the degree of regulation, Clean Air Act
regulation could range from costly to recessionary in its
economic impact.
As the government implementation of Clean Air Act
legislation evolves, conflicts and tradeoffs will necessarily
require a resolution between economics, politics, and
environmental benefits. The 1990'sr the "Decade of the
Environment," will therefore involve difficult and costly
decisions as we confront the reality and the costs of
environmental control.
411
-------
SUPERFUND COMMUNITY RELATIONS PLANS:
TAKING ADVANTAGE OF THE REQUIREMENT
Deborah C.Z. Hirsch
ERM-West
Sacramento, California
Community Relations Plans are now treated as serious components
of Superfund site activities. Community relations planning can be
approached in two ways: grudgingly, in conformance with minimal
agency requirements, or proactively, with long-term project goals in mind.
When Community Relations Plans (CRPs) are used to encourage the
participation of concerned parties in the early stages of remedial action
planning, time and money may be saved during the later stages of the
project approval process.
This paper will make the case for taking advantage of the
requirement and undertaking a proactive approach to community relations
planning. It will describe the regulatory basis and the implementation
requirements of Community Relations Plans. Conclusions will be drawn
as to how effective community relations planning can contribute to the
favorable resolution of Superfund projects.
412
-------
Introduction
In 1980, Congress enacted the Comprehensive Environmental
Response, Compensation, and Liability Act (CERCLA), the landmark
legislation now known as "Superfund". Superfund responded to the
American public's growing perception that not enough had been done to
manage the public health and safety implications of toxic and hazardous
waste dumping.
Two years earlier, extensive media coverage had exposed people in
communities all over the country to the environmental and human
tragedy of unrestricted toxic dumping at the Love Canal site in New York.
Television broadcasts showed public officials who seemed powerless as
industry spokespeople sought to avoid responsibility for this disaster.
Voters clamored for redress. By enacting CERCLA/Superfund, Congress
gave the Environmental Protection Agency (EPA) authority to assign
responsibility and exact payment for the clean-up of contaminated toxic
and hazardous waste dump sites. Superfund overturned years of
"business as usual" practices for both industry and the agencies suddenly
charged with implementing Superfund.
The finer points of Superfund enforcement are still being resolved.
The legislation was interpreted by the National Oil and Hazardous
Substances Pollution Control Contingency Plan (NCP) of 1982, which
defines and authorizes enforcement and implementation responsibilities.
The Superfund Amendments and Reauthorization Act (SARA) of 1986 set
forth additional requirements and guidelines. The practical
implementation of Superfund has also been refined through time and
experience.
The Superfund study process comprises several formally defined
phases: the Remedial Investigation/Feasibility Study (RI/FS), the
Remedial Action Plan (RAP), and Remedial Design/Remedial Action
(RD/RA). Community Relations Plans (CRPs) are designed to be
implemented concurrent to these phases of Superfund activities.
Interim handbooks detailing guidelines for the implementation of
CRPs were issued in March and June, 1988, by the EPA's Office of
Emergency and Remedial Response. The June, 1988, handbook, entitled
"Community Relations in Superfund", currently serves as the agency's
formal guidance document.
Since the first interim handbook was published in 1988, Federal and
State Superfund enforcement agencies have expanded the requirements
for preparation and implementation of formal, extensive CRPs as part of
most Superfund projects. CRPs are a logical vehicle for addressing the
increased sophistication and expectations of the public regarding
Superfund projects and toxic contamination issues.
413
-------
Objectives and Implementation Requirements
Objectives
The Superfund community relations effort promotes
communications between the public, potentially responsible parties
("PRPs" - site owners and others who may be responsible for clean-up
costs), and the lead government agency responsible for Superfund
enforcement actions.
EPA's current handbook (June, 1988) on CRP implementation lists
the overall objectives of community relations as follows:
• Give the public the opportunity to comment on and provide input to
technical decisions. An ongoing community relations effort should
encourage and assist the local public to contribute to agency
decisions that will have long-term effects on their community.
• Inform the public of planned or ongoing actions. Community
relations activities should inform the local public of the nature of
the environmental problem, the threat it may pose, the responses
under consideration, and the progress being made.
• Focus and resolve conflict. Conflict may be unavoidable in some
circumstances, but it can be constructive if it brings into the open
alternative viewpoints based on sound reasons for criticism or
dissent. An effective community relations effort channels conflict
into a forum where it can serve a useful purpose.
EPA has identified two ways in which its ability to make useful
decisions can be enhanced by public input:
1. Communities are able to provide valuable information on local
history, citizen involvement, and site conditions; and
2. Identifying the public's concerns enables EPA to fashion a response
that is more responsive to community needs.
The concerned community often maintains its own set of
community relations objectives. These might include:
• to be included in decisions that impact the community; and
• to influence decision-making by regulatory agencies and PRPs so
that final Remedial Action Plans safeguard the health and safety of
community members, and protect property values and quality of
life.
414
-------
Site owners and others who may be financially responsible for
remedial action often enter the CRP process with objectives that are very
different from the community's. Remedial action alternatives can vary
drastically in terms of cost in time and money, and responsible parties
can have thousands or millions of dollars at stake. Site owners and PRPs
may fear the CRP process as cumbersome at best; at worst, the
dissemination of "scary" technical information to a comparatively
uninformed public may seem a risk fraught with dangers to the bottom
line. Therefore, PRPs might conclude that the best objective is to avoid the
community relations gauntlet by minimizing contributions to the process.
It is the contention of this paper that the proactive, effective use of
the CRP process should not increase financial risk or liability to site
owners and PRPs; rather, it should serve as an opportunity to manage
that risk. This issue will be discussed in further detail in the
CONCLUSIONS section.
Consultants and community relations specialists are often retained
to manage the CRP process for site owners and PRPs. As their client's
representatives, these consultants must necessarily focus on their client's
objectives. Given that participation in the CRP process is mandated by
law, some consultants would advise their clients to take advantage of the
requirement by adopting the following objectives:
• communicate effectively with the interested community so that
upset and controversy can be avoided;
• coordinate effectively with regulatory agencies so that delays,
revisions of documents, etc., can be minimized; and
• generate community and regulatory agency support for a preferred
remedial action alternative.
PRPs should recognize the potential value of a process that can be
used proactively to influence the selection of a remedial action alternative.
Implementation Requirements
In overview, EPA's CRP implementation requirements follow this
schedule of key tasks:
1. Conduct community interviews.
2. Prepare the Community Relations Plan (CRP).
3. Establish locations for information repositories and administrative
records.
4. Complete the Remedial Investigation/Feasibility Study (RI/FS) and
proposed Remedial Action Plan (RAP).
415
-------
5. Provide for public comment and the opportunity for a public
meeting on the proposed RAP.
6. Prepare a responsiveness summary of significant public comments
on the RI/FS and the RAP.
7. Prepare an explanation of any differences between the Final RAP
and other actions taken.
8. Provide for public notice of the selection of the remedial action
alternative.
9. Revise the CRP, if necessary, to provide for community concerns
regarding remedial design/remedial action (RD/RA).
10. Prepare a fact sheet for the public explaining final erigineering
design.
The CRP should include a description of the site background,
history of community involvement at the site, community relations
strategies, a schedule of community relations activities, and a list of
affected and interested groups and individuals. EPA recognizes that
details of the CRP process will vary in response to the unique conditions of
each project.
Conclusions - How to Make Community Relations Plans Work
In a society that is increasingly centered on information
exchange, the best long-term strategy for business, government and the
public is to promote the exchange of information that is accurate,
relevant and complete. The CRP process is most effective when:
• communication is clear and frequent between all concerned
parties: PRPs, regulatory agencies, the community and any
consultants who may be involved;
• the PRPs (and consultants) take a proactive approach by
identifying and contacting community leaders/public interest
groups early in the process and presenting information in a way
that is meaningful to the audience;
• the regulatory agencies provide consistent direction and timely
review of documents; and
• the community reads and understands pertinent technical
information so that comments and questions are germane and
appropriate.
416
-------
When CRPs are used to encourage the participation of concerned
parties in the early stages of remedial action planning, time and money
may be saved during the later stages of the approval process. Early
public input to remedial action planning provides for the development of
documents that will "fly" and minimizes the need for costly revisions
and delays.
CRPs give the public a say in actions that may impact its future.
Conversely, CRPs provide a forum that allows PRPs to present their
perspectives and make the case for preferred alternatives and project
resolutions. For example, discussions during early stages of the CRP
process might lead to public support for a less costly on-site remediation
alternative, should this alternative reduce risk that might result from
transportation of hazardous materials or should this alternative allow
for more timely remediation and re-use of the site.
The CRP process enables PRPs to explain why a publicly
supported alternative may be prohibitively expensive or otherwise
unworkable, and it allows the public to clarify priorities and ensure that
key concerns are addressed. CRPs create additional work for
implementing agencies, but they can also shorten and simplify the RAP
and RD/RA processes by managing controversy and building
compromise.
Community Relations Plans are a natural response to the times.
Given that the public is no longer conditioned to accept the conclusions of
site owners and regulators on faith, CRPs provide a mechanism that
allows both PRPs and the community to work with regulators and
contribute to remedial action planning so that compromises leading to
ultimate project resolution can be reached. In conclusion, the community
relations process can and should be used to develop workable, viable
Superfund Remedial Action Plans that will result in better environments
for all of us.
417
-------
THE EVALUATION OF BREATH VOCs RESULTING FROM
HUMAN EXPOSURE TO MICROENVIRONMENTS
James H. Raymer, Kent W. Thomas, Stephen D. Cooper,
and Edo D. Pellizzari
Analytical and Chemical Sciences
Research Triangle Institute
3040 Cornwallis Road
Research Triangle Park, North Carolina
Abstract
Breath measurements offer the potential for a direct and noninvasive
evaluation of human exposure to volatile organic compounds (VOCs) in the
environments in which people live and work. This research study was
conducted to further develop the potential of this exposure assessment
methodology. Air samples were collected in 32 microenvironments to
determine a few possible sources of human exposure to selected VOCs.
Several people were exposed to the atmosphere in six microenvironments for
several hours. Air concentrations of VOCs were measured during these
exposures and breath samples were collected and analyzed at multiple time
points after the exposure to evaluate elimination kinetics. Elimination
half-lives were estimated using mono- and biexponential pharmacokinetic
models. Analysis of microenvironment air samples from homes, workplaces,
vehicles, etc., revealed a wide range of potential human exposures to VOCs
at concentrations from 1 to 16000 /ig/m . In general, a biexponential
function provided a better fit to the decay data than did the
monoexponential function. One-compartment half-lives ranged from 0.08 h
for n-nonane to 4.3 h for 1,1,1-trichloroethylene. Two-compartment
half-lives ranged from 0.03 h for m,£-xylene to 0.78 h for one
dichloromethane measurement for the first half-life, and from 0.61 h (n-
octane) to 21 h (j>-dichlorobenzene) for the second half-life.
418
-------
Introduction
During the past decade, the U.S. EPA has conducted studies designed to
assess personal exposure to volatile organic compounds (VOCs) through its
Total Exposure Assessment Methodology (TEAM) program. In addition to indoor
and outdoor air samples, samples of exhaled breath were also collected '
and statistically significant correlations were observed between breath
concentrations of specific chemicals and a person's activity or presence in
a distinct microenvironment . Micronenvironments are defined as discrete
locations where people live, work, or visit with distinct sets of activities
or characteristics that effect VOC emmissions and airborne concentrations.
Consequently, two general questions have arisen: (1) what are the common
personal activities and microenvironments that may lead to elevated human
exposure to VOCs and (2) can breath measurements provide a quantitative
measure of VOC exposure?
The analysis of breath for VOCs has a number of advantages over the use
of personal (exposure air) monitoring. First, the collection method is
simple and can be completed in approximately one minute. Thus the burden
associated with the collection of a breath sample is less than that of a
personal air sample. It can also provide a noninvasive alternative to blood
collection and analysis methods in the estimation of body burden. The
presence of the VOC in breath serves as a biological marker of exposure and
is more reflective of dose than is an exposure air measurement. If the
uptake and elimination characteristics of the VOC by the body are
understood, the possibility exists that breath measurements made at a known
time after the exposure can provide an estimate of the exposure (air)
concentration. Although breath excretion rates of a few VOCs have been
measured in a few subjects , further information is needed for more
compounds, subjects, and exposure levels for the development of accurate
predictive models.
The purpose of this study was to evaluate some common personal
activities and microenvironments that might lead to elevated VOC exposures,
and to determine the breath VOC concentrations that result from exposure,
including low level exposure, to a variety of chemicals. We also wished to
test a new breath sampling device and to use it, whenever possible, to
follow the decay of VOCs from the body and to calculate pharmacokinetic
parameters, such as the elimination half-lives, that can be used for the
development of exposure-prediction models.
Experimental Methods
Air, Breath Sampling and Analysis
Most of the air and breath samples described in this manuscript were
collected into evacuated SUMMA polished stainless steel canisters fitted
with a shutoff valve. For screening of the microenvironments, "grab"
samples of air were collected in most locations by opening the canister for
approximately one minute without a flow restricting orifice. Longer-term
Integrated samples (restricted inlet) were collected in some cases when the
VOC emissions were unpredictably time dependent. In some cases, the
analyses of less volatile organic compounds were facilitated through the
retention of the chemicals on a Tenax-GC sorbent bed as a known volume of
air was passed through the system . The VOCs in exhaled breath samples were
collected onto Tenax-GC or into canisters from either a Tedlar bag or a
new, miniaturized spirometer capable of collecting the breath sample in
approximately one minute. The Tedlar bag spirometer collected whole breath
419
-------
Q
and the miniaturized spirometer collected predominantly alveolar air .
Their side-by-side use here was an aspect in the development of the
miniaturized device .
Analysis of VOCs from canister samples (air and breath) was conducted
using a gas chromatograph/mass spectrometer (GC/MS) with a cryofocussing
trap similar to that of McClenny and completely described elsewhere . In
general, the analysis of canister samples for polar analytes was not
performed because of the losses associated with the water-permeable Nafion
membrane in the inlet system. The removal of water from breath samples was
found to be necessary in order to prevent freezing of the cryogenic trap.
Adsorbed VOCs were recovered from Tenax-GC via thermal desorption. The
released vapors were focussed onto a cryogenically cooled trap (-196*0) with
subsequent introduction in a GC/MS system as described elsewhere .
Exposure Experiments and Data Treatment
Various microenvironments were chosen for air sampling based on the
anticipated presence of elevated levels of one or more VOCs. Through the
evaluation of a range of microenvironments, insight into the types of
environments that contribute to human exposure would be obtained. Based on
the VOC levels found, a subset of these environments was selected for human
exposure experiments to see if these microenvironments led to elevated
levels of VOCs in exhaled breath. For each of the human exposure
experiments a volunteer spent an average of four hours in the selected
environment. After this time, the person was removed from the exposure
situation and samples of exhaled breath were collected at multiple times
over the next 3.5 h. When collecting alveolar breath samples using the new
spirometer mentioned above , these time points were 3, 8, 18, 28, 38, 53,
68, 98, 128, 173, and 218 minutes after exposure. Whole breath samples were
also collected in some experiments for either a comparison to those
collected using the alveolar device or for those chemicals more
appropriately sampled using Tenax-GC. In addition to collection of a breath
sample just before the start of the exposure, canister air samples were
collected the night before the exposure, during the exposure, and during
sample collection, to help identify potentially confounding exposures.
The VOC concentrations in these breath samples were studied as a
Et F t
function of time, and fit to equations of the form A = Ce and A = Cje 2
+ C2e~ 2 corresponding to a one and two compartment pharmacokinetic decay,
respectively, where A is the concentration in the breath at any time t; C,
C]_, and G£ are constants; and E, E^ and £2 are the exponential constants
that reflect the rate of decay. This curve fitting was accomplished using
NUN, a nonlinear curve fitting routine incorporated into SAS software (SAS
Institute, Gary, NC). By defining the half-life (t^/2) to be. the time
needed for the breath concentration to be reduced by one half, the
calculated exponential constant was used to solve for ^-\J2.' The half-lives
for the two compartment model were found in the same manner using E^ and £2
independently.
Results
Microenvironmental Screening
Air samples from a total of 32 microenvironments were analyzed in this
study. The environments included a copy center, areas with wood staining
activities, areas with oil-based painting activities, wood and metal shops,
420
-------
a furniture stripping shop, two hardware stores, two interior decorating
stores, two beauty schools, homes using consumer products such as moth
crystals, a home garage, an auto and mower refueling area, a paint and body
shop, an indoor swimming pool, a bar/club with smokers, and diapers soaking
in bleach. The concentrations of the target compounds found in some of
these environments are shown in Table I. It can be seen that a wide range
of aliphatic, aromatic and chlorinated chemical vapors was detected. This
demonstrates that areas with the potential for exposure to a variety of VOCs
are ubiquitous.
Exposure Experiments
Exposure experiments were conducted in the furniture stripping shop,
indoor swimming pool, hardware store, a home garage with refueling and wood
staining operations, and in an environment where consumer products were in
use. This last exposure was staged because of the desire for Information on
p-dichlorobenzene, fl-pinene, and limonene. The first of these three
chemicals is found in moth crystals, and the last two can be found In lemon
scented wood polish, products commonly used in the home. In general,
exposure air concentrations of greater than 100 /lg/m resulted in
measureable levels of the compound in breath. Even at this low exposure
level, it was often possible to measure a decay of the compound as it was
eliminated from the body through breath. The short time required for the
alveolar sampling device to collect the sample resulted in a greater number
of data points than did the whole breath spirometer. The rapid collection
of samples during the initial phase of the decay is necessary to
characterize the elimination of the VOC from the blood, or first compartment
in the compartmental description of the body.
The body is often thought of as consisting of a series of compartments.
As a chemical is inhaled, it can partition into the blood through the lung.
From here it can be carried to other tissues or groups of tissues, e.g,
vessel-rich group or fat tissue, and stored at rates dependent upon the
relative solubility of the VOC in the tissue (blood:tissue partition
coefficient) and the rate at which that tissue is perfused. Upon
termination of the exposure, the remaining parent VOC is released (some can
be lost to metabolism) via the reverse process. Through a classical
pharaacokinetic analysis of the data, the relative compartmental
distributions can be determined from the C values (not presented here) in
addition to the information about residence times already mentioned. This
type of analysis begins to address the issue of dose, that is, concentration
times time, to a compartment; this is the most relevant toxlcologlcal
Information. If the appropriate partition factors and perfusion rates are
known, physiologically based pharmacokinetic (PBPK) modeling can be used to
study the doses various target organs. Such models can become quite
complex and were beyond the scope of the current study.
Residence half lives calculated from the exponential coefficients for
the VOCs in this study are shown in Table II. Both one-and two-
compartment models were used. Data were not collected at sufficiently long
times after exposure to permit mathematical analysis with a number of
compartments greater than two. Values presented are derived from the
alveolar decay data except for fl-pinene, limonene, and p-dichlorobenzene,
which were collected and analyzed using whole breath with Tenax-GC based
collection and analysis. Based on an analysis of the data using the F-test,
the two compartment model provided for a better mathematical representation
421
-------
of the data . The generally short first half lives represent the
elimination of the VOC from the blood. Such rapid changes require frequent
measurement; the alveolar spirometer allowed for increased sample collection
capability relative to the whole breath spirometer.
Conclusions
The general results of the microenvironmental screening provide an
interesting snapshot of potential VOC exposure possibilities, but one which
should not be extended to all similar microenvironments or activities. The
results of the exposure experiments show that information on VOC elimination
through breath can be obtained for low-level exposures without the use of an
exposure chamber. The fact that decays were measured indicates that low
levels of VOCs can accumulate in the body. The toxicological significance
of such doses remains to be determined.
Acknowledgements
Although the research described in this article has been funded by the
United States Environmental Protection Agency under Contract No. 68-02-
4544 it has not been subjected to Agency review and therefore does not
necessarily reflect the view of the Agency and no official endorsement
should be inferred.
References
1. L. A. Wallace, E. D. Pelllzzarl, T. 0. Hartwell, C. M. Sparaclno, L. S. Sheldon, H. Zelon,
"Personal exposures, I n'door-outdoor relationships, and breath levels of toxic atr polutants
measured for 355 persons In New Jersey," Atmos. Environ. 13: 1651 (1985).
2. L. A. Wallace, E. 0. Peillzzarl, T. D. Hartwell, R. Hhltmore, H. Zelon, R. Perritt, and L. S.
Sheldon, "The California TEAM study: breath concentrations and personal exposures to 26
volatile compounds In air and drinking water of 188 residents of Los Angeles, Ant loch, and
Plttsburg, CA," Atmos. Environ. 22: 2141 (1988).
3. L. A. Wallace, The Total Exposure Assessment Methodology (TEAM) studv: Summary and Analysis.
Volume 1. Report No., EPA/600/6-87-002a, U. S. Environmental Protection Agency, Office of
Acid Deposition, Environmental Monitoring and Quality Assurance, Washington, DC 20460
(1987).
4. S. M. Gordon, L. A. Wallace, E. D. Pelllzzarl, H. J. O'Neill, "Human breath measurements in a
clean-air environment to determine half-lives for volatile organic compounds," Atmos.
Environ. 22= 2165 (1988).
5. K. J. Krost, E. D. Pelllzzarl, S. G. Waiburn, S. A. Hubbard, "Collection and analysis of
hazardous organic emissions," Anal. Chem. 54: 810 (1982).
8. E. D. Pellizzarl, R. A. Zweidinger, L. S. Sheldon, "Breath sampling," In I ARC Mannual on
Environmental Carcinogens: Selected Methods of Analysis. Vo]_. 10 - Benzene. Toluene, and
Xylene. L. Gishbein and I. K. O'Neill, Eds., Oxford University Press, 1988, p 399.
7. K. W. Thomas, E. D. Pellizzari, S. D. Cooper, "A canister based method for collection and
GC/MS analysis of volatile organic compounds in human breath," iL. Anal. Toxicol.. submitted
1990.
8. J. H. Raymer, K. H. Thomas, S. 0. Cooper, D. A. tthitaker, E. D. Pellizzari, "A device for
sampling human alveolar air for the measurement of expired volatile organic compounds," jL
Ana I. Toxicol.. in press, 1990.
9. W. A. McClenny, J. 0. Pleil, M. W. Holdren, R. N. Smith, "Automated cryogenic
preconcentration and gas chromatographic determination of volatile organic compounds in air,"
Anal. Chem. 56_: 2947 (1984).
10. E. D. Pellizzari, K. W. Thomas, J. H. Raymer, D. J. Smith, S. D. Cooper, Breath Measurements
of Individuals Exposed to Chemicals During Personal Activities. Final report on U. S. EPA
Contract 68-02-4544, Research Triangle Park, NC (1989).
11. M.E. Anderson, "A physiologically based toxicokinetic description of Inhaled gases and
vapors: analysts of steady state," Tox I co I. ADD!. Pharmacol, £Q: 509 (1981).
422
-------
TABLE I. AIR CONCENTRATIONS (ug/i3) 1H MICROENVIRONMENT SCREENING
CANISTER SAMPLES
Compound
Vinyl choMde
Isopentane
n-Pentane
Vinyl idena chloride
2-Methylpentane
OichloroMthane
Chloroform
1,1.1-Trlchloroethane
Carbon tetrachlorlde
Benzene
Trichloroethylene
Tolucn*
jpOctane
Tetrachtoroethylene
Ethylbenzene
l.B-Xylene
ji-Nonane
0,-Xylene
Styrtnt
fl-0«cane
jj-0 i ch lorobenzene
fl-Oodecane
Photocopy t
Print Center
N0»
NO
180
NO
2
10
SO
5
NO
6
NO
9
ND
NO
1
5
2
4
ND
ND
ND
NO
Oil-Based
Painting
ND
ND
150
ND
HO
25
77
3
ND
ND
5
20
16
NO
24
88
230
39
ND
1200
ND
46
Metal
Shop
ND
NO
62
4
12
23
36
21000
NO
ND
8
130
27
1200
4
n
26
4
ND
63
ND
NC&
Wood
Staining
ND
ND
1100
NO
58
2
NO
IB
ND
10
5
2700
350
2
11
30
340
11
2
810
ND
NC
Hone No. 2
N1th Moth
Crystals
ND
3
3
ND
3
77
HO
34
ND
2
ND
61
1
NO
47
180
5
11
ND
9
>540
3
Indoor
SMlMing
Pool
ND
24
15
ND
7
ND
240
2
ND
6
NO
7
1
NO
3
to
2
4
ND
4
18
ND
Furniture
Stripping
Shop
MO
10
6
3
26
7100
2
280
ND
4
120
2500
29
23
120
430
61
160
68
160
ND
35
(continued)
TABLE I. AIR CONCENTRATIONS lug/in3} [N MICROENVIRONHENT SCREENING
CANISTER SAMPLES (continued)
Vinyl clorloe
Isopentane
fl-Pentane
Vinyl idene chloride
2-Methylpentant
Dichloroattthane
Chloroform
1.1. 1-TMchloroethane
Carbon tetrachlorlde
Benzene
THchloroethylene
Toluene
n. -Octane
Tetrachloroethylene
Ethylbenzene
I.B-Xylene
Q-Nonane
£-Xy1ene
Styrene
Q-Decane
fi-Dichloroben/ene
fl-Oodecane
Hardware
Store
No. 1
NO
29
16
2
41
900
ND
210
ND
9
ND
650
80
27
590
1700
290
110
38
570
39
57
Interior
Decorating
Store No. 1
ND
35
19
ND
12
240
NO
22
ND
9
NO
310
21
9
28
93
380
22
6
700
ND
NC
Beauty
School
No. 2
ND
43
11
ND
3
NO
6
8
ND
8
7
320
ND
4
2
8
3
2
ND
2
3
2
Bar/Club
with
Snokers
NO
74
27
ND
22
6
6
3
ND
20
ND
54
2
1
10
31
E
13
6
7
NC
NC
Auto 4
Mower
Refueling
1
>1500
>3600
1
>1900
NC
NC
2
NC
>380
NO
920
22
NO
110
340
20
120
13
10
NC
NC
Paint 4
Body
Shop
NO
260
110
NO
61
7
1
3
ND
68
NO
2100
35
NO
67
220
36
60
19
5
NC
NC
Hone
Diapers
In Bleach
ND
20
16
NO
ND
41
94
ND
NO
4
NO
11
NO
ND
1
7
2
2
2
3
NC
NC
•NO • not detected.
bNC • not calculated.
423
-------
TABLE II. CALCULATED HALF-LIVES FOR VOCs IN THIS STUDY
Compound
Exposure Air
Level
-------
AUTOMATED SPECIFIC ANALYSIS OF BREATH WITH A PORTABLE MONITOR
William R. Penrose, Li Pan, Melvin W. Findlay,
G. Jordan Maclay, and Joseph R. Stetter
Transducer Research, Inc., 1228 Olympus Drive
Naperville, IL 60540
Breath analysis can be an important and often useful measurement in assessing
exposure to certain pollutants. Alcohol and carbon monoxide are two examples of
compounds that are accumulated in the body and excreted slowly when exposure ceases.
In the quest for a better breath alcohol monitor, we have studied the development of
techniques for self-calibration that can be incorporated into analytical instruments of all
kinds. One of these promising techiques is dynamic coulometry.
Amperometric electrochemical sensors operate in a destructive mode: the sample is
ojddized or reduced, producing a current at the working electrode which is the output
sijimal. Dynamic coulometry is carried out with two sensors connected in series. If the flow
through them is made reversible, they can be made to calibrate one another. In this paper,
we investigate some of the assumptions of dynamic coulometry. It appears that it may be
possible to construct an instrument that need only be calibrated when sensors are installed,
even though not all the assumptions of the technique are supported.
425
-------
Introduction
We have been pursuing the development of dynamic coulometry, a modification of
amperometric gas detection, that promises to reduce the need for frequent calibration of
analytical instruments. The work discussed in this paper concerns portable alcohol
analyzers, but the method may find application to any electrochemically-active analyte.
Blood alcohol monitors are needed by agencies involved in the detoxification and
treatment of alcohol abusers. In the former case, the care provider needs objective
assurance that the patient is adhering to an abstinence regime; many patients resort to
elaborate strategies to obtain alcohol and to disguise their use of it.
Although blood alcohol levels are the most accurate indicator of exposure, in the
case of alcohol and carbon monoxide, the blood levels are directly related to the
concentrations of these compounds in alveolar air (1). Monitoring the breath is faster,
easier, cheaper, and much less invasive than a blood test.
Breath alcohol instruments, or "breathalizers", have been marketed for many years,
but there are many areas for improvement (2,3,4,5). An instrument that is resistant to
interferences and can perform automatic calibration (or does not need calibration) is very
much needed in many applications. Recent newspaper reports, for example, have described
the dismissal of DUI cases because police officers were calibrating their breathalyzers at
home. Employers, such as managers of truck and taxicab fleets, also require instruments
of this kind, whose very presence in the workplace will reduce the incidence of alcohol
abuse on the job. A device that is useful for this purpose must be very selective in order
to distinguish between alcohol and common interferents such as acetaldehyde and carbon
monoxide, as well as to foil deliberate attempts to trick the meter. Since the user
population is medically- rather than technically-oriented, the device should not involve the
user unnecessarily in the details of its operation.
Dynamic coulometry makes use the fact that amperometric gas sensors consume a
fraction of the analyte that enters them (6,7). This fraction is the "efficiency", since the
oxidation of the analyte produces an electric current that constitutes the signal output of
the sensor. The higher the efficiency, the greater the current, and the more analyte is
consumed. If two such sensors are connected in tandem and configured for reversible flow,
each can measure the efficiency of the other. When the output current of the sensor is
divided by its efficiency, an estimate of the current generated by oxidation of 100% of the
analyte is obtained (Equations 1-3). This can be converted directly to an absolute
measurement of concentration.
An important feature of this project has been that the proposed instrument will be
able to perform a true "calibration" automatically and without the need of on-board
standards or user intervention. In addition, it will be sufficiently selective to eliminate most
common accidental or deliberate interferences. Dynamic coulometry developed for alcohol
monitoring might be applied to other measurements such as NO2, NO, formaldehyde, and
US. It may significantly reduce the time and money spent in daily, weekly or monthly
calibration of health and safety instrumentation.
426
-------
Ejcperimcntal Methods
An experimental test bed was set up according to Figure 1. The twin solenoids
caused the reversal of flow when actuated. To find low-power solenoids with low airflow
resistance was not a trivial task; the solenoids used were pre-production prototypes supplied
by The Lee Company. The sensors were standard carbon monoxide sensors manufactured
by Transducer Research, Inc. They were selected by choosing pairs of sensors with similar
outputs. The sample flowrate, which is critical in these experiments, was controlled by a
small graphite-vane pump and needle valve, and the flowrate was measured with an
electronic bubble flowmeter (The Buck Mini-Calibrator, A.P. Buck, Inc., Orlando, FL). The
operation of the apparatus and the logging of data were controlled by a battery-operated
microcontroller (Onset Computer Corp., North Falmouth, MA) and suitable custom-made
interface circuits. Datafiles produced by the apparatus were processed using Lotus
spreadsheets.
Dilutions of alcohol, acetone, acetaldehyde, and acetoacetone were made by injecting
a measured amount of the liquid into a Tedlar gas sample bag and filling the bag with air.
Dilutions of carbon monoxide were made by mixing a certified mixture from a cylinder with
air using a pump, flowmeter, and stopwatch.
Figure 1. Schematic of the apparatus used for these experiments.
potentiostat were controlled by a miniature datalogging computer.
Solenoids and
Sample , j
exhaust
Sensors(2)
Sample
intet
Zero
solenoid
Zero filter
Potentiostat
Crossover
solenoids
and network
427
-------
Results
Dynamic Coulometry
The potential of dynamic coulometry has been previously demonstrated for carbon
monoxide (7). Operation was therefore tested by passing 40 ppm CO into the apparatus
at a flowrate of 10.4 mL/min. The zero-corrected results are shown in Figure 2. With the
gas movement in the forward direction, the output of sensor A is greater than that of sensor
B; when it is reversed, the output of A decreases, and that of B increases. This is the
pattern expected in a dynamic coulometry experiment. A substantial proportion of the CO
is removed from the sample by each sensor. In this particular experiment, the response of
sensor B was judged to be very slow, and the outputs of both sensors were very weak. This
judgment was based on the following calculation:
Carbon monoxide at 40 ppm contains 1.786.10"* mole/liter. At a flowrate of 10.4
mL/min, this is equivalent to 2.92.10"10 mole/sec. For the two-electron oxidation to carbon
dioxide, the current generated by 100% oxidation of ethanol supplied at this flow rate can
be calculated to be 56 microamperes. The maximum currents measured in this analysis are
8.86 and 8.70 microamps for sensors A and B, respectively. The corresponding decreases
in the signal due to depletion of the sample by the upstream sensor are 26.2% and 27.0%.
The calculated currents for 100% oxidation are therefore 8.86/.270 = 32.8 microamps and
8.70/.262 = 33.2 microamps. These values are low compared to the expected 56 microamps
and may be due to several causes, from a mistake in diluting the gases, to an invalid
assumption in the technique itself.
Both sensors were replaced by new ones for the remainder of these measurements,
and satisfactory results were obtained. Ethanol vapor was run at a flowrate of 11.9 mL/min
and concentrations of 12, 51 and 96 ppm. The plot of sensor output versus concentration
is shown in Figure 3. The calculations of dynamic coulometry are shown in Table I. The
two complex terms Ra and Rb are equivalent to the current expected from 100% reaction
of the analyte, and are derived from the theoretical treatment of dynamic coulometry
(Stetter and Zaromb, 1983):
X* = C.Ra/V, where Ra - L.f.L.r/ftr - L^f) (Eq. 1)
XB « C.Rb/V, where Rb = UW&f ' U) (Eq. 2)
C = RT/zFp (Eq. 3)
where X, = mole fraction of the vapor
V = the volumetric flowrate
Ii = observed currents in sensors A or B in the (f)orward or (r)everse directions of
gas flow
C is a composite of the gas constant R, Kelvin temperature T, number of electrons
in the reaction z, Faraday constant F, and atmospheric pressure p
428
-------
Figure 2. A test of the dynamic coulometry apparatus using CO. The scales of sensors A
and B have been changed so the lines do not overlap.
8 10 12 14 16 18 20
Fiigure 3. The maximum current outputs of the sensors are linear with concentration.
140
120
100
"5T
Q.
I 8°
o
o cn
'E °0
D)
CO
40
20
20 40 60 80
Ethanol Concentration (ppm)
100
429
-------
Table I. DCO calculations for ethanol vapor. 50 ppm of ethanol in
air was pumped through the instrument at a flowrate of 11.9 mL/min.
The terms Ra and Rb are explained in the text and correspond to the
expected current output of the sensors if the ethanol were 100%
consumed.
Concentration Sensor A Sensor B
(ppm) Net Effic- Ra Net Effic- Rb
Signal iency Signal iency
(ppm) (uA)
12.5 16.7
53 69.9
95 133
(%)
.179
.174
.170
(UA)
93.3
380
739
(u
6.
27
52
A)
75
.9
.4
(%)
.182
.186
.180
(UA)
37.1
160
308
The terms Ra and Rb should be identical for both sensors in a given experiment,
and they should be proportional to analyte concentration and inversely proportional to
flowrate. In this experiment, the values of Ra and Rb were plotted against concentration
in Figure 4 and compared to the expected current for 100% oxidation.
It is immediately clear that, although one sensor closely follows the theoretical
prediction for 100% oxidation, the remaining sensor yields a calculated value that is about
60% too high. In order to confirm the reality of this unexpected result, the experiment was
repeated using carbon monoxide, which has previously been shown to be a good analyte for
DCO (7). These results are shown in Figure 5, and lead to the same conclusion as the
experiment with ethanol.
At this time, we do not understand the reasons behind this discrepancy. Obvious
sources of error were checked: instrument calibration, flowrate, leaks, etc. We next
examined both the explicit and implicit assumptions made in deriving the DCO theory.
Violations of most of these assumptions would result in a negative error in the calculated
100% oxidation current. One implicit assumption remained as a possible cause of the
discrepancy: We have assumed that the efficiency of a sensor would be the same in both
the forward- and reverse-flow directions. If not, either positive or negative errors could
result. Testing of this assumption will require further investigation.
In spite of the departure from theoretical prediction, it should be pointed out that
the Ra and Rb were nevertheless linear with flow rate. This observation is sufficient to
demonstrate the validity of DCO in this application. An empirically-derived, or
coulometric, constant can be derived by calibration rather than by calculation from
Equation 3. Presumably, an instrument using. DCO should only need to be calibrated when
a new sensor is installed.
430
-------
Figure 4. The total oxidation current for ethanol is linear with flowrate in spite of the
change in sensor efficiency.
F
T3
1
IJ
15
O
800
1,200
D
^ 1,000
I
o
c
o
'•(3 600
;g
2 400
§
200
0
Ethanol 60.3 ppm
Sensor A
4-electron reaction
10 15
Flowrate (mLymin)
i—
20
25
30
Figure 5. Same as Figure 4, but for carbon monoxide.
500
c 400
Q>
Carbon monoxide 32.9 ppm
10
Sensor B
^^
2-electron reaction
20 30 40
Flowrate (mL/min)
50
60
431
-------
The success of the dynamic coulometry experiments led us to attempt a few brief
experiments with static coulometry, in which a measured volume of sample vapor is trapped
in the chamber of a single sensor. As the analyte is consumed by the sensor, the signal
should decrease in a logarithmic fashion (if the gas is well-mixed in the chamber). This
experiment was done by arranging the two solenoids on either side of a single sensor, to
direct the sample stream either through the sensor or around it. When the stream was
diverted, the fall of the signal was observed for a fixed time. The area under the curve (in
microampere-seconds, or microcoulombs) should represent the total ethanol in the volume
of sample.
Both the fall of the signal and its logarithmic nature were observed in these
preliminary experiments (Figure 6). An instrument employing static coulometry would be
simpler and perhaps just as effective as one using dynamic coulometry.
Responses of Different Sensors and Interferents
Potential interferences in measuring alcohol on the breath are acetaldehyde, acetone,
acetoacetone, and carbon monoxide. The last of these can be found in the breath of
smokers. The others are metabolites of alcohol, but can be found in the breath of some
individuals with metabolic diseases. Carbon monoxide has already been run; the other
three compounds were tested using the same program. The results are shown in Table II.
The responses of acetaldehyde and acetone are very weak, only 2.5-4% of that of ethanol.
Acetoacetone gives a strong but sluggish response, rising slowly over the course of the
experiment (Figure 7). Since the signal strength due to acetoacetone was similar to that
of ethanol, this would be a serious interference. Ethanol (50 ppm) and acetoacetone (27
ppm) were mixed in the same bag and run together (Figure 8). The ethanol signal is
clearly distinguishable over the acetoacetone signal, and could be separated in a computer
by suitable digital filtering of the signal.
Responses of the platinum sensor to ethanol and some interferents were measured
at several bias potentials. The results are shown in Table IV. The ethanol response does
not change over the bias range of -50 to 200 mV, and the change in the carbon monoxide
signal is less than twofold. Selectivity to these analytes therefore cannot be improved by
manipulating the bias potential.
A sensor with a high-surface-area gold working electrode was installed temporarily
in the apparatus and exposed to 221 ppm hydrogen sulfide (a positive control), 334 ppm
carbon monoxide, and 278 ppm ethanol. Neither carbon monoxide nor ethanol produced
any response on the gold electrode.
432
-------
70
Figure 6. Acetoacetone
elicits a sluggish
response from the sensor.
-ID -
• 10
Tin - iiln
14 10
Figure 7. The signal for
ethanol is clearly dis-
tinguishable from that of
acetoacetone.
I
3
-10
Figure 8. Static coulo-
metry of ethanol. The
descending line, when
appropriately plotted,
fits a semilogarithmic
relationship with time.
100
200
300
Time - sec
400
500
6C
433
-------
Table II. Responses of platinum- and gold-electrode sensors to
ethanol and its metabolites.
Analyte, concentration Outputs in microamps/ppm
and bias used Sensor A Sensor B
ethanol, +100 mV 1.52 0.86
60.3 ppm
carbon monoxide, 0.77 0.50
+100 mV, 33 ppm
acetone, - 100 mV 0.028 0.025
43 ppm
acetone, + 100 mV no signal no signal
43 ppm
acetoacetone, -100 mV (0.63) (0.48)
27 ppm slowly and continuously rising signal
acetoacetone, +100 mV (1.05) (1.08)
27 ppm slowly and continuously rising signal
acetaldehyde, +100 mV 0.038 0.023
48 ppm
Conclusions
Coulometry (either dynamic or static) is a potentially useful method for measuring
alcohol and carbon monoxide. The method may ultimately be used for other gases as well.
These methods offer the promise of instruments that will require only infrequent
calibration. Although the sensors did not produce the current predicted by the theoretical
development of the dynamic coulometry technique, the following relationships, derived from
the the theoretical development, were shown to hold:
* Peak sensor output was linear with concentration.
* The calculated 100% oxidation current extrapolated from the measured
signal was independent of the efficiency of the sensor.
In this work, we also demonstrated that the common metabolites of alcohol,
acetaldehyde, acetone, and acetoacetic acid^ produced signals that were either weak or
easily distinguished from that of alcohol. Carbon monoxide, however, is a serious
interference, and will have to be dealt with by special means.
434
-------
Acknowledgments
We wish to acknowledge the contributions of G.M Kocanda, Lombard, IL, i
instrument design and construction. Funding by provided by a Phase I grant from th
National Institutes of Health under the Small Business Innovative Research program.
Bibliography
1. M. Falkensson, W. Jones, B. Sorbo, "Bedside diagnosis of alcohol intoxication
with a pocket-size breath-alcohol device: sampling from unconscious subjects
and specificity for ethanol," Clin. Chem. 35/36: 918-921 (1989).
2. H.W. Bay, K.F. Blurton, H.C. Lieb, H.G. Oswin, "Electrochemical
measurements of blood alcohol levels," Nature 240: 52-53, 1972.
3. R.H. Cravey, N.C. Jain, "Current status of blood alcohol methods." J. Chrom.
Sci. 12: 209-213, 1974.
4. N.C. Jain, R.H. Cravey, "A review of breath alcohol methods," J. Chrom. Sci.
12: 214-218, 1974.
'!. W.T. Schaffer, K.R. Warren, "Analytical methods for the measurement of
tissue ethanol levels." Alcohol Health and Research World: Summer issue, p.
14ff, 1987.
6. J.R. Stetter, S. Zaromb, "A Dynamic Coulometric Technique for Gas
Analysis," J. Electroanal. and Interfacial Electrochem.. 148: 271-277 (1983).
7. S. Zaromb, J.R. Stetter, and D. O'Gorman, "Determination of Carbon
Monoxide in Air by Dynamic Coulometry," J. Electroanal. and Interfacial
Electrochem. 148: 279-287 (1983).
435
-------
ROUTES OF CHLOROFORN EXPOSURE FROM SHOWERING WITH CHLORINATED WATER
Clifford P. Weisel, Wan K. Jo* and Paul J. Lioy
Joint Graduate Teaching Program in Human Exposure,
Rutgers: The State University and UHDNJ - Robert Wood Johnson
Medical School and The Environmental and Occupational
Health Sciences Institute, Piscataway, NJ 08854
*current address: InJae University
Dept. of Environmental Science
A-Bang Dong, KimHae
Seoul, Korea
Abstract
Showering is the single, largest use of water in the home. This
activity results in exposure to volatile organic compounds (VOC) present in
the water that are released to the air and contact the skin. Inhalation and
dermal exposure increases the body burden when the compounds penetrate the
body barrier. The concentration of VOC in exhaled breath is related to the
body burden and was used to estimate the relative internal chloroform dose
from inhalation and dermal exposures during showering with chlorinated water.
Thirteen normal showers, which include inhalation and dermal exposure, and
thirteen inhalation only studies were analyzed using typical showering
conditions. The post-exposure breath chloroform concentrations ranged from
6.0 to 21 ng/m3 for normal showers and 2.4 to 10 ng/m* for inhalation only
exposure, while the pre-exposure concentrations were less than the minimum
detection limit of 0.86 ng/m*. The elevated chloroform concentration in
breath after exposure depended upon the water concentration, water
temperature, exposure duration, and the post-exposure delay time prior to
collecting the breath sample. To facilitate comparison of the body burden
all parameters, other than water concentration, were fixed for the study.
A statistically significant difference was found between the breath
concentration after showering (dermal plus inhalation exposure) and after
breathing air in a shower stall (inhalation only exposure). It was determined
that the inhalation and dermal exposures contributed approximately equally
to the elevated chloroform body burden resulting from showering.
Introduction
Showers have been proposed to be a major indoor air source of VOC (1,2)
since it uses large amounts of heated water which results in the release of
the VOC contained in the water. Chlorinated water contains *
-------
Methods
Sampling
Exhaled breath samples were collected from subjects by having them
breathe through a non-rebreathing two-way valve attached to a Tedlar sampling
bag. The subjects were supplied with purified, humidified air through the
valve from an inhalation bag. Breath from the collection bag was transferred
to a Tenax packed trap. Water samples were collected from the tap in the
same room as the shower using clean 50 ml vials, following EPA Method 502.1
(8). Ten minute breathing zone air samples were collected in the shower
stall at a flow rate between 750 and 1250 cc/min using Tenax traps.
Analysis
The breath and air samples were analyzed by thermally desorbing the
trap and transferring the compounds to a packed column gas chromatograph with
an electrolytic conductivity detector in the halogen-specific system mode.
The water was analyzed by purge and trap-gas chromatography following EPA
method 502 (8).
Quality Assurance
A blank and external standard was analyzed daily to monitor the
response of the GC. Typically, the blank concentrations were below the
detection limit, thus any response indicated contamination in the system. The
response of an external standard was compare to the value calculated from a
calibration equation. If the response differed by more than +20%, a new
calibration equation was determined. Ttre precision of the desorption and
water analytical systems for chloroform were 13% and 10%, respectively.
The minimum detection limit (MDL) of the desorption and water analytical
system were 13 nG and 0.65 uG/L, respectively.
Experiment I - Evaluation of Parameters
A model shower, constructed of stainless steel, was used to evaluate
the effect of water temperature and duration of inhalation exposure on the
breath concentration. Water was sprayed within the chamber using a standard
shower head at two temperatures, 34°C and 41°C, and the subject breathed the
air from chamber for 5, 10 or 15 minutes. A standard, full size shower was
use to evaluate the relationship between the chloroform air concentration
within the shower stall and the water concentration, with and without a
subject present. For this and all full size shower studies a shower duration
of 10 minutes and a water temperature of 40°C was used. All other
controllable parameters were fixed for all experiments including: the water
flowrate (8.7 L/min), shower head setting, ventilation system being off and
post exposure delay prior to collection of a breath sample (5 minutes). The
chloroform water concentration was measured whenever an air or breath sample
was collected.
Experiment II - Chloroform Exposure from Inhalation Only
Six subjects {4 males and 2 females) participated in thirteen
inhalation only exposure experiments using the same full-size shower, shower
parameters and protocols as in Experiment I. Each subject stood next to the
water stream within the shower stall for ten minutes, thereby exposing the
individual to chloroform vaporized from the water. The subject wore rubber
clothes and boots during the experiment to avoid dermal contact with the
shower water. Breath samples were collected from the subject prior to each
inhalation only exposure and five minutes after exposure. Air and water
samples were collected during each exposure.
437
-------
Experiment III - Chloroform Exposure from Normal Showers
Thirteen showers were taken by six subjects (5 males and 1 female) to
estimate total chloroform exposure from a typical shower. Breath samples
were collected from each subject prior to and five minutes after each shower.
Air and water samples were collected with each shower. All showering
conditions were set as indicated in Experiment I. The data obtained from
Experiment II and Experiment III were used to compare the chloroform body
burden resulting from a normal shower with that from an inhalation only
exposure. The comparison was conducted using a covariance analysis.
Results and Discussion
The air chloroform concentration was measured with and without a
showering individual present to determine whether the movement and splashing
by an individual altered the air concentration and could therefore affect the
inhalation exposure. As shown in figure 1 the air concentration/water
concentration relationship for the two conditions overlapped. The mean and
standard deviation of chloroform air concentrations without a showering
individual were 157 uG/m3 and 75.5 ng/m*, respectively. The mean and
standard deviation of air chloroform concentrations while an individual was
showering were 186 ^g/m3 and 76.0 (»g/m3, respectively. The concentrations
were compared using an F-test. The a F-value was 0.01 at a p=0.9294,
indicating the presence of an individual did not affect the chloroform air
concentration.^
The breath concentrations measured after exposure to air in the model
system demonstrated that both exposure duration and water temperature altered
the body burden (Table 1). The breath concentration was found to be
positively correlated with water temperature and exposure duration. Each
temperature and duration was found to be statistically significant by both
ANOVA and Duncan's Test.
The shower air concentration were observed to increase with water
concentration. The shower air concentration ranged between 69 and 330 ^g/m3
for a water concentration range of 12.9 and 40.0 ng/l (figure 1). Chloroform
was not detected in any of the air samples collected prior to the water being
turned on. The relationship between the air and water concentration 'is
described by the linear model:
'air'8'11 * C ' 39'2 *>r C >4.8
with an R2 = 0.87 at a P = 0.001
In addition to the air concentration increasing with water
concentration, the body burden, as measured by breath analysis, increased,
for both inhalation only exposure and for normal showers. Breath samples
were collected prior to and after each exposure type. In all cases,
chloroform was not detected in breath samples collected before exposure.
Thus, the pre-exposure breath chloroform concentration was always less than
the detection limit of 0.86 Mg/m3. The breath concentrations after a normal
shower, which includes both inhalation and dermal exposure, tended to
increase with the chloroform concentration for water in the range 5.3 to 36
^g/L (figure 2). The mean and standard deviation of the breath concentrations
were 13 ^g/m3 and 3.9 ng/m3, respectively. Minimum and maximum breath
concentrations were 6.0 and 21 jig/m3» respectively. Similarly, the breath
concentrations after inhalation only exposures tended to increase with
increasing water concentration for the range 10 to 37 j^g/L (figure 2). The
mean and standard deviation of breath concentrations measured after
438
-------
inhalation only exposure were 7.1 ^g/m and 2.5 ^g/m, respectively. Minimum
and maximum breath concentrations after inhalation only exposures were 2.4
and 10 Mg/m3, respectively. These two sets of breath samples were found to
be significantly different using the F-test at a probability of p=0.0001.
Two explanations for a difference in the breath concentrations after
the normal shower and inhalation only experiments are: 1) splashing of water
by the showering individual increases the volatilization of chloroform and
therefore the inhalation exposure and 2} dermal absorption of chloroform
directly from the water occurs. The first explanation was shown not to be
valid since the presence of an individual did not increase the air
concentration above that measured in an unoccupied shower stall. Dermal
absorption by non-polar hydrocarbons from dilute aqueous solutions has been
demonstrated for a number of compounds (9) and has been predicted to
contribute to the internal dose during showering based on theoretical
considerations using estimates of permeation of compounds through the skin
barrier and transfer rates within the body (6) and is the likely explanation
for the results observed in this study.
The relative contribution from inhalation and dermal exposures were
quantified based on the least mean squares (LSM) of chloroform breath
concentration for the two exposures, while controlling for variations in the
water concentration. The LSM of the breath concentrations resulting from a
normal shower, which is a composite of inhalation and dermal exposures, was
13 *«g/m3, while the LSM of the breath concentrations after inhalation
exposures only was 6.8 Mg/m3. The difference, 6.2 Mg/m3, is assumed to be
due to dermal exposure. The ratio of chloroform body burden increase from
dermal exposure to that from inhalation exposure is therefore 0.93. Hence,
both inhalation and dermal absorption contributed equally to the internal
chloroform dose.
Conclusions
An increase in the chloroform body burden resulting from inhalation
exposure and dermal exposure during a normal shower was observed in the
present study. The breath concentration after showering was approximately
twice as high as that after inhalation only exposure, indicating that the
contribution to the internal dose by dermal absorption was equivalent to
inhalation adsorption. The magnitude of the chloroform body burden increase
was positively correlated with water concentration, water temperature and
shower duration.
Acknowledqements
The research was funded by the New Jersey Department of Environmental
Protection, Division of Science and Research. Dr. Wan received Fellowship
support from the Environmental and Occupational Health Sciences Institute.
References
1. Andelman, J.B. (1985) Inhalation exposure in the home to volatile
organic contaminants of drinking water. Sci. Total Environ. 47:443-460.
2. Andelman, J.B. (1985) Human exposures to volatile halogenated organic
chemicals in indoor and outdoor air. Env. Health Persp. 62:313-318.
3. Krasner, S.W.; McGuire, M.J.; Jacangelo, J.G.; Patania, N.L.; Reagan,
K.M.; and Aieta, E. M. (1989) The occurrence of disinfection bv-products in
US drinking water. Journal American Water Works Assoc., p. 41-53.
439
-------
4. Cothern, C.R.; Coniglio, W.A.; Marcus, W.L. (1985) Techniques for the
assessment of carcinogenic risk to the U.S. population due to exposure from
selected volatile organic compounds from drinking water via the inqestion,
inhalation, and dermal routes. U.S. Environmental Protection Agency,
Office of Drinking Water, EPA 570/9-85-001.
5. McKone, T.E. (1987) Human exposure to volatile organic compounds in
household tap water: The indoor inhalation pathway. Env. Sci. & Tech.
21:1194-1201.
6. Wester, R.C. and Maiback H.I.; (1989) Human skin binding and absorption
of contaminants from ground and surface water during swimming and bathing.
J. of the Amer. Col. of Tox. 8:853-860.
7. Jo, W.K., Weisel, C.P.; Lioy, P.J. (1990) Routes of chloroform exposure
and body burden from showering with chlorinated tap water. Risk Analysis, in
press.
8. (April 1981) The determination of haloqenated chemicals in water by the
purge and trap method. Method 502.1; U.S. Environmental Protection Agency.
Environmental Monitoring and Support Laboratory, Cincinnati, Ohio 45268.
9. Scheuplein, R.J. and Blank, I.H. (1971) Permeability of the skin.
Physiological Reviews 5JU702-747.
Table 1
Breath concentration normalized to tap water concentration obtained after
inhalation exposure for two water temperatures and three inhalation durations
Replicate
Number
Warm (33.6°C)
5 min
10 min
15 min
Hot (40.8ol-J
5 min
10 min
15 min
1 0.16 0.22 0.37 0.21 0.36 0.41
2 0.16 0.29 0.42 0.25 0.37 0.52
3 0.20 0.30 0.48 0.26 0.39 0.53
4 0.21 0.34 0.48 0.29 0.41 0.63
5 0.24 0.37 0.59 0.30 0.44 0.65
440
-------
400
> 300 -
6
c
o 200 H
. iooH
o
.c
CO
01
a w/ individual
* w/o individual
10 20 30 40
Tap Water Conc.(ug/L)
50
Figure '. Chloroform shower air concentration with and without a showering
individual versus tap water concentrationfincludina 95% confidence interval).
30
CO
E
I! 20 -
o
c
O
O
a ioH
shower
inhalation
10 20 30
Tap water conc.(ug/L)
40
Figure Z. Chloroform breath concentration after normal showers
and inhalation onlv shower versus tao water concentration
441
-------
Biomarkers and Phannacokinetics
Jerry N. Blancato, Ph.D.
U.S. EPA
FOB 93478
Las Vegas, NV 89193-3478
Monitoring of biomarkers is a promising technique to help
assess both the exposure and dose to critical target sites of
toxic chemicals. Biomarkers can range from body burden to
antibodies able to detect, separate, and quantify specific
adducts formed when toxic chemicals interact with endogenous
proteins. To properly interpret the data so that assessments can
be made an understanding of the pharmacokinetics of the biomarker
is required. Several examples are presented here to illustrate
how monitoring and pharmacokinetic information might be
integrated to assess dose and exposure.
Acknowledgment: Although the research described in this paper
has been supported by the United States Environmental Protection
Agency, it has not been subjected to Agency review and therefore
does not necessarily reflect the views of the Agency and no
official endorsement should be inferred.
442
-------
INTRODUCTION
Biomarkers of exposure can be defined as any detectable
measure or evidence of previous or ongoing absorption of a
foreign substance into the body. Examples would include the body
burden of the chemical itself or one of its metabolites, any
products resulting from the interaction of the chemical and /or
its metabolites with endogenous molecules, and any alterations in
normal physiologic and biochemical processes. For the markers to
be of use in quantitative exposure assessments several necessary
criteria must be fulfilled. First, a suitable detection and
measurement method needs to be available. The method must allow
for accurate measurement at all the expected exposure regimens
and levels. Also, a quantitative relationship between exposure
levels and the levels of measured biomarker must exist. The
choice of the actual biomarker will depend upon many factors
including on the specific definition of dose used in the specific
case of interest.
In addition, several other characteristics are desirable,
and thus should also be sought when developing a marker for
exposure assessment. The detection and measurement techniques
should be at least as sensitive as existing methods used in
conventional exposure assessment methods. The monitoring methods
and devices should be relatively easy to transport and
inexpensive to use. The monitoring protocol should be designed
so as to insure minimal discomfort to the subject. The protocol
should facilitate the repeated monitoring of multiple subjects.
Methods to analyze and interpret the monitoring results are
essential for maximizing the value of information gained from
biomarker studies. Finally, the usefulness of a marker is greatly
enhanced if it can be used both as a marker of exposure and
effect.
When embarking on a biomarker program one must first
determine what enhancements to the exposure assessment process
will be made by their use. As such, it is very useful to
determine how many of the general utilities of biomarkers apply
to the specific case of interest. One of the most obvious of
these utilities is the ability to measure actual dose by using
biomarker based techniques. Conventional exposure assessment
methodologies only allow for the measurement or estimation of
concentration time profiles at the boundary between environment
and subject. Thus, from such data dose can be only be inferred.
Biomarker data on the other hand are measurements of the amount
of chemical having entered the body. From these measurements the
dose can be calculated rather than inferred.
Equally obvious is the fact that biomarker based assessments
result in a measure of the integrated dose, from all routes or
portals of entry into the body. Conventional exposure
assessments can only infer dose and even for this, considerable
knowledge regarding exposure by each portal of entry is required.
It should also be noted that in many cases, with some additional
443
-------
information it is possible to apportion the aforementioned
biomarker based integrated dose to the various portals of entry.
Thus, from biomarker data, both the integrated dose and the
specific route apportioned dose can be determined.
Still another important utility of the biomarker approach is
that, in many cases, dose, and thus previous exposure, can be
determined long after the exposure has ceased. In many cases a
biomarker may remain in the body, thereby giving evidence of
exposures, for weeks after the actual exposure episode has
ceased. Adducts formed by some xenobiotics and their metabolic
products to hemoglobin for example, will reside in the body for
the life of the erythrocyte which, for the human, is over 17
weeks. Such adducts can be monitored and thus provide evidence
of exposure and estimates of dose long after the last exposure
incident.
One more utility is that, with proper analysis, biomarker
information can be used to estimate the dose to critical organs
as well as estimating total body dose. Thorough pharmacokinetic
information and appropriate biomarker measurements an exposure
assessor can estimate the dose to critical tissues during and
long after the exposure period. In addition, critical temporal
parameters, such as the time necessary to rid the body or organ
of the toxic moiety, can be calculated.
Two cornerstones are necessary in order to realize the
fullest utility of the biomarker approach. First, there must, as
previously mentioned, be suitable detection and measurement
devices.and methods. Second, there needs to be an extensive
description of the pharmacokinetics of the biomarkers of
interest. The pharmacokinetics must be described and modeled in
a manner that will result in a maximum use of biomarker
measurement data for estimating dose and other related parameters
of exposure.
Physiologically based pharmacokinetic (PBPK) models, while
not the only approach, offer great promise as tools to accomplish
the required analysis of biomarker data. Properly formulated and
validated they can be used to first describe physiologic kinetics
of the biomarkers associated with a particular chemical. Then,
after this developmental stage, their output can serve as an
analytical tool for estimating the desired parameters from the
biomonitored data. It is their use in this second, or analysis
stage, that will be illustrated here.
METHODS
This exercise has been formulated solely for illustrative
purposes. As such, the biomarker and dose data were essentially
manufactured from past experiences and knowledge. Under actual
exposure conditions the formulation and validation of the PBPK
model could be quite involved and complex. Usually such models
are quite data intense. If models and appropriate data are
444
-------
lacking they must be generated simultaneously with, or even in
anticipation of, the biomarker development. As will be
illustrated, more than one biomarker and associated analysis tool
may be necessary to get accurate estimates for large exposure
ranges. If estimates of the exposure ranges are known apriori it
may be possible to reduce the number of necessary biomarkers
required.
Discussions of actual model formulation procedures will vary
according to the specific case and are far beyond the scope of
this paper. The same can be said for the mathematical procedures
for solving the mass balance equations that compose PBPK models.
Suffice it to say that, to date, complicated models are solved
numerically rather than analytically. The expanding capacity and
speed of computers has enabled the solution of models by what
were previously considered to be tedious numerical methods.
Model simulated output is then used with biomonitoring data
from subjects to estimate dose. Again, a variety of methods can
be used. The most complicated, but perhaps the method affording
the most accuracy and resolution between doses, uses the PBPK
model itself. Biomonitoring data is taken. The model is
repeatedly solved to determine an estimate for dose which results
in the best fit of model output to the monitored data. Numerical
methods are used to determine what possible dose level could have
resulted in the actual monitored values. Given proper numerical
methods and appropriate software, confidence values around the
estimated dose can be calculated. In this manner the more
reliable and .accurate monitoring data will result in greater
confidence in the final dose estimates.
A contrasting and less computationally intense method
involves using the PBPK model to give outputs at several
arbitrary dose ranges. The model outputs are then graphed and
the resultant graphs can be used as reference graphs from which
to estimate dose ranges after biomarker monitoring is performed.
A third approach is a comprise between the first two. For
this case model outputs at several dose ranges are put into a
computer data bank. An expert system, with access to that data
bank, is then employed to compare the monitored data to model
simulated outputs and thus give estimates of dose. Such a
computer based expert system could be formulated to perform
appropriate statistical analyses of multiple data points of the
monitored marker.
Each is advantageous under specific conditions. Following
is a simple illustration of the second, or graphical approach.
RESULTS
Figures 1 through 3 illustrate the use of pharmacokinetic
information to determine, dose in one particular exposure
scenario. Eight possible dose levels (1-8) are included in this
445
-------
illustration. Two things are assumed (with apriori knowledge)
about the scenario. First it is assumed or known that exposure
is still ongoing and second that the plateau or steady state has
been reached.
Three different biomarkers have been monitored. Figure 1
shows the profile of marker 1 for the 8 dose levels. The graph
shows the marker's concentration with time profile as determined
from the pharmacokinetic analyses for the various doses. If the
monitored marker level is less than 4.0 a dose range can be
established. For example, it can be observed that if the
concentration level (Y-axis) of marker 1 is 3.9 then the dose
must then be between level 2 and level 3 (value for marker at
dose level 2 = 3.0 and at dose level 3 or higher = 4.0). The
resolution could then be increased by increasing the number of
simulated model outputs within that range.
If the marker level were greater than 4 it would not be
possible to discern between dose levels 3 through 8. Figure 2 is
the analogous graph for marker 2. Inspection reveals that this
marker can discern between doses within the range of the first
four dose levels rather than just the first three.
One might wonder why use marker 1 at all when marker 2 is
adequate for a wider dose range? It may be that marker 1 is more
accurate at lower doses or that it is easier and cheaper to
obtain and analyze. Thus it may be the preferred marker for
screening or for determining the lower dose levels.
Figure 3 repeats the process but for marker 3 instead.
Inspection here reveals that this marker can discern between
doses within the range of levels 5 through 8. It also
demonstrates that this marker, due to its kinetic profile, is
incapable of being used to estimate the lower dose levels.
Next let us look at a case after exposure has ceased.
Figure 4 shows the concentration-time profile of a marker both
during and after the exposure period. Making the assumption that
monitoring is performed, as indicated, during the post exposure
period, a value of 1.0 concentration units for marker 1 can be
associated with three different doses of the parent chemical.
Resolving between doses can be accomplished by knowing the time
after exposure when the monitoring occurred.
Often the exact time that exposure ceased is not known.
Resolution between doses can be achieved in such cases by
measuring other markers as observed in Figure 5. The top panel
of Figure 5 shows the measured value of marker 2 (horizontal
line). Clearly this level is higher than any level that is
reached at dose level 1 and thus such a measured value would rule
out dose 1. However marker 2, at this value, cannot discern
between the two higher dose levels. The lower panel of Figure 5
shows the measured value of marker 3. This value can only be
reached by dose level 3. Thus, if monitoring of markers 1
446
-------
(Figure 4) and 2 (Figure 5, top) was not sufficient to discern
between the possible dose levels, marker 3 would be monitored.
Figure 6 depicts another series of marker determinations
that might be used to differentiate between two doses. Again
these measurements, as illustrated here, are made after exposure
has ceased. In this illustration one calculates the ratio of the
concentration of marker 1 to the concentration of marker 2. That
calculated ratio is then compared to ratios similarly calculated
for the standard curves (curves shown in Figure 6). It can
readily be observed from the example that the ratio of the
concentration of marker 1 to the concentration of marker 2 is
sjignificantly different for the two different doses depicted
here. For the case of dose l (upper panel of Figure 6) the ratio
of the concentration of Marker 1 to the concentration of marker 2
is approximately 2.0. For the case of dose 2 (bottom panel of
Figure 6) the value of the ratio is about 0.7.
CONCLUSIONS
Monitoring of biomarkers is a promising method to help
assess both exposure and dose. Biomarkers are evidence of dose
rather than just exposure and thus can significantly enhance the
conventional exposure assessment process. Also, some biomarkers,
clue to their long life-time within the body, can be used long
after exposure has ceased. Therefore, these markers would lend
themselves to monitoring persons whose exposure is expected on an
in-frequent basis.
The future use of biomarkers will depend upon several
factors. Only those markers which are most easily, accurately,
and cheaply applicable under field conditions will be practical.
Not all markers will be useful at all exposure concentrations.
Those that are only able to detect exposure at rather high
concentrations may only be useful for monitoring occupationally
exposed populations for example. Others, such as receptor based
methods will probably be more sensitive than existing analytical
methods and thus be applicable even at environmental exposure
levels. Some of these, however may lack selectivity and only be
able to indicate exposure to classes of compounds rather than
specific chemicals. In such cases some knowledge of the possible
chemicals that people have possibly been exposed to will greatly
help the assessment process. Other markers while both sensitive
and selective may be quite expensive to apply, and thus will be
used in only in cases of great impact or where other more
conventional monitoring techniques do not suffice. Perhaps the
single greatest promise for biomarker based exposure assessment
is for their utility in assessing exposure to a chemical found in
a mixture. Much remains to be done to determine how useful these
techniques will be in such cases, but many of the developing
markers are chemical, or at least, class specific and the thus
with proper development the impact of concominant exposure can be
quantified.
447
-------
fc
UL
O
u
Figure 1. Value of Marker 1 in the body for eight dose
levels of a chemical. Time: arbitrary units. Concen-
tration: arbitrary units. Note that the value of the
marker is.the same for dose levels 3-8.
in
\-
•z.
~z.
t— <
1
LU
Z
8
10
9
8
?
t
-
-
.
DOSE LEVEL 5-8
/
/ DOSE LEVEL 4
/ DOSE LEVEL 3
/^
/ DOSE LEVEL 2
DOSE LEVEL 1
/
^//
01 23456789 10
TIME
Figure 2. Value of Marker 2 in the body for eight levels
of a chemical. Time: arbitrary units. Concentration:
arbitrary units. Note that the value of the marker is the
same for dose levels 5-8.
448
-------
10
g
8
ui
5 n
•z. u
a
^H
£ 5
a:
^ ^
Ul
u
z
s 3
2
0
-
-
.
DOSE LEVEL B
/
X DOSE LEVEL 7
DOSE LEVEL 8
/
/ DOSE LEVEL S
x"
/ 1 1 1 1 1 1 1 1 1
01 23456789 10
TIME
Figure 3. Value of Marker 3 in .the body for eight levels
of a chemical. Time: arbitrary i^iiLs. Concentration:
arbitrary units.
LU
u
o
o
10
Q
8
7
6
5
4
3
2
MEASURED VALUE
OF MARKER 1
DOSE LEVEL 8
DOSE LEVEL 2
DOSE LEVEL 1
5 0
TIME
10
12
Figure 4. Concentration of Marker 1 at three dose levels
Measured value could be due to any of the three dose
levels. Time: arbitrary units. Concentration: arbitrary
units.
449
-------
DOSE LEVEL »
•"DOM LEVEL I
DOM IfVtL 1
MEASURED VALUE
OP HARKR E
23X567
TIME
a ID ii
10
D05E LEVEL 1
POgg LJVtL I
MEASURED VALUE
OF MARKER 1
01 ^ 3 4 5 6 7
TIME
9 10 I!
Figure 5
Measured Values of
Markers 2 and 3
Time: Arbitrary Units
Concentration: Arbitrary Units
450
-------
1
t
12 7
' ,
§
5 5
P .
1 *
U
1
DOSE LEVEL 1
HUSUKl
MARKED 1
/ X
/ HAIIKEH I
/ \
TiaE
ENT TIME
\
\
\\
DOSE LEVEL 2
0 I
12 13 M 15
Figure 6
Profiles of Two Biomarkers
at Two Dose Levels
Time: Arbitrary Units
Concentration: Arbitrary Units
451
-------
REVIEW OF THE PARTICLE TEAM 9 HOME FIELD STUDY
Wiener, R. W., Wallace, L. and Pahl, D., U. S. Environmental Protection
Agency, Atmospheric Research and Exposure Assessment Laboratory, Research
Triangle Park, NC 27711
Pellizzari, E., and Whittaker, D., Research Triangle Institute, Research
Triangle Park, NC 27709
Spengler, J., and Ozkaynak, H., Harvard School of Public Health, Boston, MA
02115.
The 9 home pilot study for the Particle Total Exposure Assessment
Methodology (PTEAM) Program has been completed and a population study is
being planned for fall 1990. The pilot study was a 9 home field test of
instruments, protocols, and sample analysis. The study site was the San
Gabriel Valley, CA.
The study logistics were generally successful. Shorter more precise
questionnaire material is being developed for a population study. The use
of portable microenvironmental samplers could be enhanced by using battery
powered samplers. Computer assisted interviewing and data collection was
useful for more rapid analysis and cross validation. Data analysis of
personal and microenvironmental aerosol samplers has indicated a possible
personal cloud (aerosol concentration gradient) effect. This effect is
currently being investigated. Elemental analyses have indicated adequate
material is collected for XRF.
The PTEAM study is being sponsored by the U.S. EPA and the California
Air Resources Board.
452
-------
Introduction
The U. S. Environmental Protection Agency (U.S. EPA) during the last
ten years has begun to assess the importance of human exposure to
pollutants as measured by personal and mlcroenviron- mental sampling. This
summary presents EPA's assessment of the information gathered through a
small scale study to assess the design of equipment and survey materials
for measuring human exposure to suspended particulate matter.
Classically, exposure to most pollutants was thought to occur when
either the individual went outside into a polluted atmosphere or the
polluted atmosphere infiltrated into the indoor environment. In the last
decade it has been realized that indoor pollutants may be substantially
different from outdoor and that indoor pollution may account for the
majority of population exposure to many pollutants. Improvement in outdoor
ambient air quality and changes in life style, building technology, and the
use of chemical agents (pesticides, cleaners, spray propellants,
deodorants, etc.) have increased the relative importance of indoor
exposure. The average adult spends at least 68-78% of time indoors.
U.S. EPA has conducted a series of field programs to measure
distribution of human exposure to volatile organic compounds (Total
Exposure Assessment Methodology Program, TEAM), pesticides
(Non-Occupational Pesticide Exposure Study, NOPES). The generic TEAM
concept seeks to answer fundamental questions regarding the number of
persons exposed, the sources of exposure-r transport of pollutant to the
population at risk, the effects of exposures, the meaning of sampling data
in terms of actual exposure, and the estimation of the level of exposure of
the actual population to the pollutant in question. The basic ingredients
of TEAM study are representative probability sampling, measurement of the
pollutant concentrations, measurement of body burden, and recording of each
person's daily activities. ''
This document discusses the initial field trial of the Particle TEAM
Program. The field study included nine homes and eighteen persons and was
concerned with the development and testing of measurement techniques to
determine human exposure to aerosols.
The Particle TEAM program seeks to answer questions relating the
measurement of aerosol concentration to human exposure. Health related
information collected in the Particle TEAM Program will be limited.
The Particle TEAM study objectives are four-fold. The first is to
estimate of the frequency distribution of human exposure to particles in
the PM^Qsize range for a metropolitan population. The second objective is
to discern the differences among the concentrations of particulate matter
measured by personal exposure monitoring, outdoor ambient air sampling, and
fixed site or microenvironmental monitoring. The third is the
identification of the major sources of exposure and the degree of exposure
of an urban population to provide exposure assessment and source
apportionment data. The fourth is to develop models for personal exposure
and source assessment.
453
-------
The PTEAM Program seeks to meet the objects through the use of survey
design statistics, the physical measurement of the aerosol concentrations
in microenvironments where individuals spend time, and by using personal
samplers that are carried by the subjects throughout the course of the
monitoring. Additionally questionnaires and diaries are used to assess
possible routes and sources of exposure to respirable and inhalable
particulate matter.
The Particle Total Exposure Assessment Methodology (PTEAM) Program has
been developed in three phases. In the first phase monitoring equipment
necessary for determining the level of human exposure to aerosols was
developed. This included both personal and microenvironmental sampling
instruments. The second phase consists of a pilot field program developed
to test these instruments in the field. The third and final phase consists
of a large field program where a probability design is used to select homes
from an urban population. This document concerns some of the results from
the first two phases. The third phase is to be implemented Fall of 1990.
From observations made in previous TEAM studies a number of objectives
were recognized as necessary in the design of the pre-pilot field program
(Phase 2). '^' First, adequate testing of methods, protocols, and
instruments is needed in the field environment. Questions about logistics
of implementation, sampler ruggedness, and suitability had to be answered
before the population study. The 9 home field test is the opportunity to
insure that the field personnel are adequately trained on all the sampling
instruments and protocols and are prepared to interact with the study
participants.
Experimental
The study site for the 9 home field test was the San Gabriel Valley,
CA. The area was selected for the high levels of both coarse and fine
particulate matter found year round in the ambient air.
The components of phase two included preparation of the work plan, site
selection, laboratory testing and calibration, development and
implementation of sampling design protocols. Two protocols were used in
the field test. First a temporal study was performed using five homes,
lasting seven days, and including two sets of microenvironmental aerosol
monitors within each home. Second, four additional homes were used in a
four day spatial study in which three sets of microenvironmental monitors
were placed inside each home. A total of eighteen persons, nine homes,
nine work places and two central sites were sampled. Methods included
microenvironmental and personal aerosol collection of both PM^Qand ^2.5-
Nicotine was measured using active and passive sampling and associated with
both personal and microenvironmental measurements. Radon samples were
taken at each of the homes. Air-exchange measurements were also made.
The aerosol sampling equipment used in this study consisted of two
types of devices, a personal exposure monitor (PEM) and a fixed location
microenvironmental monitor (MEM), The aerosol PEM consists of a flow
454
-------
controlled personal pump that is worn by the test subject as he/she
performs his/her normal daily activities and an inlet and filter pack which
are designed to collect particles near the breathing zone. New PEM inlets
were designed for Particle TEAM by the University of Minnesota for the
measurement of personal exposure to respirable PM2.yr inhalable PM^Q
particles. Each inlet consists of an impactor classifier to remove
particles larger than the predetermined cut size (050)and to provide a
sharp cut (Og) , and a filter to collect the remaining particles. The inlets
were designed to operate with 37mm filters and at 4 LPM, Both PM^gand
PM2_5-nlets were evaluated in the 9 home study. The MEM inlets were
designed by Marple, Spengler, and Turner to sample at 10 LPM with a sharp
cut and had been used in the Harvard Six City Study and others.
Survey questionnaires were developed to screen the population for
study selection and to relate activities and sources of aerosol emissions
to increased personal exposures. Information areas surveyed include:
demographic information (roster of participant household, participant
occupation, age, smoking status, sex hobbies, socioeconomic status, housing
type), sources of exposures to aerosols and chemical species of interest,
activities correlating with exposures, limited health effect or wellness
information, ventilation (air exchange rates,heating and air conditioning
sources), residence descriptives (e.g., multi-unit, attached),
transportation (commuting time, type of vehicle), occupation, and workplace
descriptives.
The information was obtained through the use of several different
questionnaires and forms administered to the participant and/or completed
by technical field personnel. The key survey instruments included a
household screening questionnaire, a participant questionnaire, and a 12
hour activity log. The screening questionnaire provides basic demographic
information necessary for stratification of the population being sampled.
The 9 home study was a purposeful study, participants were selected based
specific parameters, such as housing stock, type of employment, and
smoking. The participant questionnaire collects data on household
characteristics, personal characteristics, and workplace characteristics
from the subject. Morning and evening 12 hour activity logs composed of a
chronological log and a supplemental close-ended questionnaire were used to
provide detailed information on the participants daily activities during
monitoring.
Discussion
A series of questions were posed at the outset of the nine home
study. These questions were selected to test the ability of the pre-pilot
to meet the specific objectives listed above (i.e., testing of methods,
•protocols, and instruments, logistics, personnel training and interaction).
This discussion section presents EPA's assessment of the information
gathered through the pre-pilot study to answer some of these questions.
1. Is sufficient PM2 yr PMipmaterial being collected in a
455
-------
12 hr period for chemical and physical measurements?
• Do XRF, and/or FIXE have sufficient sensitivity to measure
selected elements on 12-hr PM2 ynd PM^QCollected fractions?
• . Are there serious chemical interferences precluding
quantitative analysis?
Data collected in the 9 Elemental analysis by XRF indicated that
sufficient mass should be obtained for any of the 12-hour sample types to
enable high percentage measurable values to be obtained for the most
important elements. The exception, however, is Cd. Using the a two-sigma
threshold as a quantifiable limit, the reported percent measurable data
drop to less than 10%, whether using NEA, Inc. Protocol 9 or 5. Thus, Cd
cannot be adequately measured by XRF on PEM and MEM samples in the Pilot
Study.
Among the 34 elements, the 13 of primary interest were Fe, Pb, Ni, K,
Si, Al, As, Cd, Mn, Se, V, Cr, and Sb. For many of these primary elements,
quantifiable concentrations for 100 percent of the analyzed filters were
found.
Of the 34 elements listed in these tables, six generally dominated in
terms of amount of particulate mass for which they accounted. These were:
Si, Ca, S, Fe, Al, and K. For instance, for PEM 10 /lm samples, these six
elements accounted, on average, for about 10 percent of the mass, while all
34 together accounted for about 11 percent.
Eight of the PEM filters analyzed by XRF (protocol 9) were analyzed
twice (three 2.5 /im and five 10 /im samples). For each element, standard
deviations and RSDs of the paired concentrations were calculated.
Precision was calculated only for those 17 elements exhibiting a high
percentage of quantifiable values. Though the sample sizes were obviously
small, the results for the different size cuts appeared similar, so the
results were combined. Thirteen of the elements had median RSDs between
one and ten percent, three elements had RSDs between 10-15%, and one
element (P) had a median RSD of 55.7%. Except for P, the precision for
the remaining elements (Al, Si, K, Cr, Mn, Fe, Ni, Pb, S, Cl, Ca, Ti, Cu,
Zn, Br and Sr) was of acceptable quality.
There appeared to be no serious chemical interferences precluding
quantitative analysis. However, the portion of MEM filters that had uneven
deposition patterns was examined for elemental non-uniformity. Repeated
XRF measurements of the same filter could not identify a problem. However,
the method did not allow sufficient room to analyze incongruent areas.
An additional examination was performed using duplicate MEM samples with
one or both exhibiting bleed. The filters were examined for variability
between the duplicate values. Comparing the RSD for 15 duplicates for 14
elements which had measurable concentrations (with those duplicate MEM
samples in which neither filter exhibited bleed) did not reveal any
obvious differences in variability.
Thirty PEM and MEM filters analyzed by XRF were also subjected to
PIXE analysis. In general, when the mean elemental concentration (e.g.
456
-------
>40 mg/m3) was well above the detection limits, good correlations were
observed between the two techniques (e.g., 0.70 to 0.98). For some of the
elements, especially for the PEM samples, there appeared to be a
difference in the average concentration levels of the two methods, with
the XRF usually being higher. This, coupled with the large number of
negative concentrations found by PIXE for some elements, suggests that an
over-correction of background from blank filters may have occurred for the
PIXE-determined concentrations. This interference would be a potential
problem in the Pilot study.
2. What Is the level of accuracy and precision attainable In PEM and MEM
collection for analysis (gravimetric methods')?
a. Precision of PEM Gravimetric Results
Paired PEM samplers carried by the field technicians during pilot
(refer to section 5.2) provided some insight into the precision of the PEM
method. The RSDs ranged from 1.1-15.0% (median 4.4%) and 4.2-24.8% (median
12.7%) for the PM^Q and PM2 5 samples , respectively. The precision was
deemed acceptable.
b. Precision of MEM Gravimetric Results
Twenty- four pairs of PM^Q and PM2 5 MEM indoor samples were collected
at the participants homes. The RSDs ranged from 0.6-7.4% {median 1.7%) and
0.6-24,4% (median 5.6%) for the 10 fJ.m and 2 , 5 flm samples, respectively.
Nine pairs of PM^Q outdoor MEM and eleven pairs of PM2t 5 outdoor MEM samples
were collected. The RSDs ranged from 0.8-33.3% (median 3.5%) and 0,6-12.5%
(median 2.8%) for the PM^g and PM2 5 samples, respectively.
It was concluded that the precision was acceptable.
c. Precision of PEM and MEM Filter Weighings
Duplicate weighings for a subset of PEM and MEM filters were
performed, thereby allowing an assessment of the precision associated with
this portion of the measurement process. The mean RSD for the 10 um and
2.5 urn PEM filter weighings was 1.5% (40 pairs) and 2.65% (19 pairs),
respectively. For the 10 /im and 2.5 Mm MEM filters, the mean RSD was 1.02%
(69 pairs) and 1.54% (62 pairs), respectively. The precision for PEM and
MEM filter weighings was felt to be acceptable.
d. Precision of SSI Gravimetric Results
The central site data included duplicate SSI PMiQConcentrations for
each of the 22 time periods of the Pre -pilot Study. The average
concentrations ranged from 23 to 149 mg/m3. The relative standard
deviations for the 22 pairs ranged from 0.1-23.0% (median 2.9%). This
precision was acceptable.
e. Accuracy of PEM and MEM Gravimetric Results Relative to SSI
457
-------
Two methods were used to estimate a relationships between the
central-site PEM and MEM PM^oconcentrations and the average of the two SSI
PMiQconcentrations. One was ordinary least squares (OLS), the other
maximum likelihood (ML) estimates. Positive intercepts, which appeared to
be of borderline statistical significance, indicated that the PEM-measured
concentrations may tend to run slightly higher than the MEM-measured
concentrations. On a percentage basis, the estimated difference is smaller
for higher concentrations, e.g. about 7% higher when the MEM concentration
was 130 mg/m3, and about 24% higher when MEM concentration was 30 mg/m3.
3. What are the difficulties associated with participants _
carrying the PEMs?
• What is the failure rate and performance record
for PEMs?
• What problems are voiced by the participants?
The failure rate for the PEMs was negligible in the field. There was
some problem with the repackaging and cracking of the pulsation damping
chamber. These problems can and were overcome in the pre-pilot and
additional PEM testing. The most significant problem with the PEMs had
been the accidental miss-cutting of the 2.5 /im orifice by MSP, Inc.
Twenty-two PM2.Camples were taken with effective cut sizes of 3.9 Jim.
These points were tagged in the data base. Debriefing of the participants
resulted in no serious complaints in regard to wearing the PEMs. The
overall conclusion was the PEMs could be successfully utilized in the
Pilot.
4. What are the difficulties associated with placing MEMs in homes?
• Inconveniences to participants?
• Adequately defining locations?
• What is the failure rate and performance record
for MEMs?
The majority of problems were encountered during the set-up visit.
The most notable problems were space limitations due to the amount of
equipment placed in the homes, borderline acceptable noise levels due to
the amount and type of equipment placed in the home and the lack of
properly grounded power outlets at both inside and outside sampling
locations. A combination of improved performance AC powered MEMs to
provide quiet indoor sampling and battery powered portable MEMs for outside
would solve these problems.
The failure rate for the MEM packages was negligible.
5. Does the Activity Pattern instrument provide usable information for
correlation vith monitoring data?
• Does the activity recall questionnaire yield adequate response
specificity to account/explain the monitoring data?
458
-------
Is it useful or necessary to know about minor sources and
short-term sources when evaluating 12 and 24 hour integrated
exposures.
Evaluation of survey instruments included the time activity diary,
and time activity diary supplement. The analysis of the time activity and
supporting questionnaire data indicated a number of ambiguities and
inconsistencies. It was concluded that the response categories in the
activity diary were too broad and highly variable to be useful in data
analysis and that time indications needed to be clarified. Exposure needs
to be better defined, especially in terms of active or passive exposure
node.
Higher MEM concentrations may be the result of home and personal
activities. Homes that had the highest indoor MEM2.5 and MEM10
concentrations were later identified to have potential sources of indoor
particles (for example, smoking and other dust generating activities, such
as cleaning, performing crafts projects). Outdoor activities associated
with elevated particulate levels included such activities as sweeping,
raking leaves, gardening, and working with horses. More rigorous analysis
of the activity data did not result in significant correlations with
measured concentrations.
The nicotine data indicated that levels were usually associated with
the presence of smokers. Discrepancies between high contact with smokers
and low recorded nicotine levels were also evaluated. These data indicated
that participants had a hard time distinguishing between smoker contact or
exposure and non-exposure. Certainly it appeared that participants could
tell when they had direct contact or were within the plume, but they had a
lesser ability to distinguish lower levels of contact. In addition, this
analysis reconfirmed the statistical analysis of nicotine/time activity and
particulate/time activity relationships, where time of reported smoker
contact indicated nicotine levels but not particulate levels.
While the field monitoring staff have knowledge which may be
beneficial for the accurate completion of the documents, the current survey
instrument requires too much time to administer and needs to be reduced in
scope and complexity. The time of administration must be reduced, as must
the burden of the respondent. Several comments from the respondent
debriefing were received concerning the level of detail requested in the
diary, and the difficulty in maintaining the document. It was found that
information obtained from many questions was not very useful for analysis
and such questions should be excluded. In summary, the survey instruments
will be redesigned and tested in a focus group setting prior to use in the
Pilot Study.
The survey data and analyses provide little hope for correlating
minor sources and short term sources when using integrated sampling of 12
and 24 hours. Real time personal sampling methods need to be developed
that can more directly tie activity to aerosol particle exposure.
459
-------
Summary
A number of questions have been posed to be addressed by the PTEAM
field study. They fall into four categories: physical sampling,
survey/questionnaire, data analysis, and illustrative. Most of the
questions have been answered successfully by the 9-home study. The
sampling instruments have been shown to work well in the field. Of the
total 2240 samples, 99.47% were captured. No significant difficulties were
associated with placing fixed location samplers or carrying personal
samplers. The level of accuracy and precision was largely resolved.
Sufficient Pi^.sor PM^o material was collected for gravimetric and
elemental analysis on almost all samples taken. Chemical analysis was
largely successful.
Disclaimer: This extended abstract does not necessarily reflect EPA policy.
References
1. Ott, W., Wallace, L. , et al., "The Environmental Protection
Agency's Research Program on Total Human Exposure." Environ.
Intern. 12:475-494 (1986).
2. Gammage, R.B. "Reality and Perception in Indoor Air Quality --an
overview." Presented at the Indoor Air Quality Symposium, Environ.
Health and Safety Div., Ga.Tech.Res. Inst., Ga. Inst.of Technol.,
Atlanta, GA 30332 (1986).
3. Newill, Vaun A, "The Role of Total Exposure Measurement in Risk
Management, Keynote Address" Proceedings of the 1987 EPA/APCA
Symposium on Measurement of Toxic and Related Air Pollutants, pp.
1-4. APCA, Pittsburg, PA. (1987).
4. Pellizzari, E.D., L.C. Michael, K. Perritt, D.J. Smith, T.D.
Hartwell, and J. Sebestik, "Comparison of Indoor Toxic Air Pollutant
Levels in Several Southern California Communities, Final Report," EPA
Contract No. 68-02-4544, (1988).
5. Pellizzari, E.D., K.W. Thomas, D.J. Smith, K. Perritt and M. Morgan,
"Total Exposure Assessment Methodology (TEAM): 1987 Study in New
Jersey, Final Report," EPA Contract No. 68-02-4544, (1988).
6. Pellizzari, E.D., T.D. Hartwell, H. Zelon, R. Perritt, J. Sebestik,
W. Williams, D.J. Smith, J. Keever, C.E. Decker, R.K.M. Jayanty, K.
Thomas, D.A. Whitaker and L.C. Michael, "Baltimore Total Exposure
Assessment Methodology (TEAM) Study, Final Report," EPA Contract No.
68-02-4406, (1988).
460
-------
PARTICLE TOTAL EXPOSURE ASSESSMENT METHODOLOGY (PTEAM):
STATISTICAL ANALYSIS OF SPATIAL AND TEMPORAL PATTERNS
C. Andrew Clayton, Research Triangle Institute
Research Triangle Park, NC, U.S.A. 27709
The Particle Total Exposure Assessment Methodology (PTEAM) Prepilot
Study was conducted in the Duarte/Azusa/Glendora, CA area during March of
1989. The study involved personal aerosol monitoring for two participants
in each of nine volunteer households, using personal exposure monitors, and
particulate monitoring in and near their homes, using microenvironmental
monitors. The study's primary purpose was to develop methodology for
personal exposure monitoring of particulates, a methodology that could be
applied later in a large-scale Pilot Study. In this paper we present some
of the statistical results from the Prepilot Study and indicate how such
results were used to aid in the design of the large-scale study.
461
-------
Introduction
In this paper, we examine the spatial and temporal variability of the
aerosol concentration data of the PTEAM Prepilot Study conducted in the
Duarte/Azusa/Glendora, CA area during March 1989. The study involved
personal aerosol monitoring for two participants in each of nine volunteer
households, using personal exposure monitors (PEMs), and particulate
monitoring in and near their homes, using microenvironmental monitors
(MEMs). Time periods of monitoring were of approximately 12 hours duration
starting in the evening and in the morning.
Study Design
Primary participants for the PTEAM Prepilot Study were selected from a
group of municipal workers who had indicated willingness to participate in
the study. Table I summarizes the participant characteristics. Except for
the primary participant in house 9, all of the participants were
nonsmokers. Table I also indicates the basic structure of the Prepilot
design. Monitoring for houses 1 through 5 began on the evening of March 6
and continued for 14 time periods. Starting on the evening of March 16, a
second group of homes (Houses 6-9) were monitored for eight periods. PEMs
were alternated daily between Wft and 2.5/1, while both size cuts were
simultaneously employed for the MEMs. The MEMs were placed outside of the
participants' homes, in the home's main living area (MLA), and in one or
two other locations:
House Main Living Area Area 1 Area 2
1
2
3
4
5
6
7
8
9
Den
Kitchen
Living Room
Living Room
Living Room
Den
Den
Living Room
Living Room
Master Bedroom
Master Bedroom
Master Bedroom
Master Bedroom
Bedroom
Master Bedroom
Master Bedroom
Master Bedroom
Master Bedroom
Kitchen
Kitchen
Study
Dining Room
In some cases, duplicate, collocated MEM samplers were used.
Results and Discussion
PEM Aerosol Concentration Data
The PEM data exhibited a substantial amount of variability, both from
person to person and among time periods for a given person. Examination of
the data clearly revealed that one major source of such variation was time
of day: higher PEM aerosol concentrations generally were observed for the
daytime periods. (A similar result holds for the outdoor ambient samples.)
Because of the day/night differences, separate statistical analyses were
performed for the day and the night periods.
This day/night difference is evident in the Table II results, for
instance. This table gives statistics that characterize the distributions
of the PEM gravimetric data -- by size cut, time of day, and time of week
(weekdays or weekend-days). The statistics reported are the following:
the sample size n, which is the number of observations over houses,
participants within a house, and time periods; the minimum, maximum, and
mean aerosol concentrations (in /*g/m3); the standard deviation (in /jg/m3);
and the coefficient of variation (CV), expressed as a percentage. The
462
-------
mean, standard deviation, and CV for the 17 nonsmoking participants are
also shown (the associated n value is one less than that shown for all
participants). A similar classification by work-day versus non-work-day,
rather than weekday versus weekend-day, was also generated. However,
because most of the daytime, weekday samples were associated with periods
when study participants went to work (30 of the 33 10/t samples, and 10 of
the 11 2.5/i samples), little difference between those results and those
depicted in Table II were evident.
Table III indicates the degree to which similar PEM aerosol
concentrations for primary and secondary participants were observed. For
most houses, the two participants had comparable means and exhibited
moderate to high positive correlations. The major exceptions were house 9
and house 6.
The main conclusions about the PEM data that can be drawn from
Tables II and III are shown below; analyses of variance (ANOVAs), performed
separately by size cut and time of day, tended to support these
conclusions.
• Daytime PEM aerosol concentrations, on average, were higher than
nighttime concentrations.
• For a given time of day (i.e., day or night), there was substantial vari-
ability in aerosol concentrations, both among participants and among time
periods within participants. The variability among the daytime aerosol
concentrations was generally larger than that among the nighttime
observations.
• On average, weekend/weekday (or workday/non-workday) differences in
aerosol concentrations, relative to the amount of variation among days
within these types, were not large.
• Participants from the same household often (but not always) tended to
have comparable levels and similar temporal patterns in their PEM aerosol
concentrations.
MEM Aerosol Concentration Data
Table IV shows a temporal summary of the MEM aerosol concentration data
from the main living areas of the nine participating households. The table
gives the sample size, minimum, maximum, mean, standard deviation, and CV,
by size cut, time of week, and time of day. Both daytime and nighttime
samples show extensive variation in the concentration levels. However,
analyses of variance performed on these data revealed that most of the
variation in the daytime samples arose from day-to-day variation within a
given house, while for the nighttime samples, when people tended to be at
home, there was more variation among houses and somewhat less among days
within houses.
A spatial summary of the indoor MEM aerosol concentration data is given
in Table V. The table shows means, by indoor location, and correlations
among the aerosol concentrations. Sample sizes were 13 or 14 for Houses 1-
5 and 7 or 8 for Houses 6-9. Comparable means were usually noted among the
various sampler locations within a house, and the correlations among the
concentrations from two locations were generally high (median correlation
of 0.89 for 10/i samples, and 0.94, for 2.5/* samples).
Table VI shows a temporal summary of the outdoor MEM aerosol
concentration data. The table reveals that lower levels tended to occur at
night, and that lower and less variable levels occurred on weekend-days,
when contrasted with weekdays.
463
-------
Table VII has two purposes: to provide a spatial summary of the
outdoor data, and to furnish a comparison of these data with the indoor MLA
data. For each of these two types of samples, the table shows, by house,
the mean aerosol concentration (/*g/m3) and the indoor/outdoor correlation.
Sample sizes equal 13 or 14 for Houses 1-5 and 7 or 8 for Houses 6-9.
ANOVAs applied to the outdoor aerosol data indicated that there was very
little house-to-house variability, relative to the day-to-day variation.
While the indoor/outdoor correlations were usually positive, they varied
widely among houses.
Conclusions
The results presented here demonstrate how the statistical examination
of temporal and spatial variability was useful in addressing some of the
design issues for the subsequent large-scale study. For example, these
analyses were used to draw the following implications for the PTEAM Pilot
Study, assuming that similar conditions will be encountered in the study
area and time frame chosen for that large-scale study:
• Person-days should be chosen as the basic sampling unit, because of the
large temporal variation. Stratification of days by weekend/weekday or
work-day/non-work-day is not likely to be very effective in improving
precision of estimated PEM gravimetric distributions.
• Because of correlations among PEM concentrations of persons in the same
household, the use of multiple persons per household, at the expense of
an overall smaller number of households, does not appear warranted.
•If the Pilot Study were to be implemented in the same study area and
season that were used in the Prepilot, then the Prepilot data would
suggest that the use of outdoor samplers at each house may not be
necessary. However, the extent to which similar conditions will prevail
in another area and time frame is unknown.
• For some homes and times (but not all), the outdoor aerosol level can be
expected to be useful for predicting indoor levels.
• For 12-hour samples, there appears to be little informational benefit to
be gained by using multiple MEM samplers placed at various locations
within a house.
Acknowledgements
This study was funded by the U.S. Environmental Protection Agency,
Contract No. 68-02-4544. It has not been subjected to the required peer
and administrative review: the views expressed do not necessarily reflect
the views of the Agency and no official endorsement should be inferred.
The author is grateful to the many individuals at RTI, Harvard University,
and Accurex for their contributions.
464
-------
TABLE I. SUMMARY OF PTEAM PREPILOT MONITORING DESIGN
AND PARTICIPANT CHARACTERISTICS
Participant^
Houseb Age Sex
1
2
3
4
5
6
7
8
gc
P 36
S 11
P 52
S 49
P 32
S 31
P 26
S 28
P 52
S 24
P 32
S 31
P 30
S 27
P 30
S 11
P 41
S 18
M
M
M
F
M
F
F
M
M
M
M
F
F
M
F
F
F
M
Number of
Monitoring
Periods
14
14
14
14
14
14
14
14
14
14
8
8
8
8
8
8
8
8
No. of
Periods
at Work
4
3
5
0
5
5
4
3
5
1
2
0
2
2
2
0
2
3
No. Periods
with
Exposure No. Periods with
to Smoking Smoking at Home
at Work Day Night
1
0
5
0
2
0
0
1
0
0
0
0
0
0
2
0
2
0
0
0
0
1
1
0
0
0
0
0
2
1
0
0
1
1
4
1
0
0
0
0
0
0
0
0
0
0
1
2
0
0
1
1
4
3
aP=primary, S=secondary.
bAll houses are single family units except numbers 4 and 6.
cprimary participant is a smoker.
TABLE II. TEMPORAL SUMMARY OF PEM AEROSOL CONCENTRATIONS 0*g/m3)
Size
Cut
IQp
2.5/1
Time
of
Day
Day
Night
Day
Night
Time
of
Week*
wd
we
wd
we
wd
wec
wd
we
All Participants
n
33
18
34
18
11
15
12
17
Min.
66.5
59.3
35.4
34.4
54.1
44.7
28.5
19.9
Max.
240.9
301.9
186.0
243.9
240.6
219.8
135.7
121.2
Mean
127.3
164.1
87.1
93.3
98.0
100.2
62.9
52.5
Std.
Dev.
43.7
70.6
34.0
49.9
55.5
52.7
37.1
27.4
CV
34
43
39
54
57
53
59
52
Omittinq Smokerb
Mean
126.1
156.0
86.6
84.4
98.9
91.6
56.2
48.2
Std.
Dev.
43.8
63.6
34.4
33.8
58.4
42.6
30.6
21.6
CV
35
41
40
40
59
46
54
45
awd=weekdays, we=weekend-days.
bvalues of n are one less than the n indicated for all participants.
cjwo "outliers" omitted for this row.
465
-------
TABLE III. SPATIAL SUMMARY OF PEM AEROSOL CONCENTRATIONS 0*g/m3)
House Number:
10^ Size Cut
n -Primary
Secondary
Mean-Primary
Secondary
Correlation
of Primary,
Secondary3
2.5^ Size Cut
n- Primary
Secondary
Mean -Primary
Secondary
Correlation
of Primary,
Secondary*
1
7
8
117
122
0.63
2
1
b
b
b
2
8
8
100
110
0.38
2
2
b
b
b
3
8
8
104
121
0.66
2
1
b
b
b
4
6
6
99
97
0.63
4
4
79
67
0.89
5
6
6
77
86
0.72
3
4
36
55
0.95
6
4
4
140
127
0.13
4
4
74
78
0.59
7
4
4
145
99
0.99
3
4
63
101
0.92
8
4
4
116
165
0.91
4
3
81
95
0.96
9
4
4
204
96
0.10
4
4
142
53
-0.33
Median
)
J
0.63
\
) 76
0.90
aCorrelations are based on the smaller n. Minimim and maximum correlations
are underlined.
blnsufficient sample size.
TABLE IV. TEMPORAL SUMMARY OF MEM MAIN-LIVING-AREA
AEROSOL CONCENTRATIONS (>tg/m3)
Size
Cut
10>
2.5/1
Time
of
Day
Day
Night
Day
Night
Time
of
Week
wd
we
wd
we
wd
we
wd
we
n
33
18
32
18
31
18
31
18
Min.
21.5
35.9
21.4
18.9
8.3
20.7
13.5
10.5
Max.
199.9
169.6
169.4
165.4
136.6
120.0
88.5
144.6
Mean
63.6
79.8
51.3
57.8
37.4
47.7
33.9
35.0
Std.
Dev.
33.7
37.0
30.1
45.7
23.6
23.4
19.0
36.0
CV
(%)
53
46
59
79
63
49
56
103
466
-------
TABLE V. SPATIAL SUMMARY OF INDOOR MEM AEROSOL CONCENTRATIONS (/(g/m3)
House
IQlt Si
Number:
ze Cut
Mean: MLA
Area 1
Area 2
Correlation:
MLA vs Areal
MLA vs Area2
Areal vs 2
2.5/i Size Cut
Mean:
Correl
MLA vs
MLA vs
Areal
MLA
Area 1
Area 2
ation:
Areal
Area2
vs 2
1
41
40
0.89
25
24
0.92
2
53
49
0.99
37
33
0.99
3
49
63
0.86
37
45
0.95
4
47
47
0.97
27
27
0.97
5
64
64
0.95
41
41
0.95
6
113
103
92
0.89
0.99
0.91
61
57
54
0.97
0.98
0.94
7
57
41
45
0.70
0.98
0.77
24
19
20
0.63
0.96
0.79
8
74
41
66
0.88
0.94
0.73
32
25
32
0.94
0.97
0.92
9
97
64
93
0.96
0.83
0.83
75
48
66
0.94
0.86
0.91
Median
60
0.89
35
0.94
TABLE VI. TEMPORAL SUMMARY OF OUTDOOR AEROSOL CONCENTRATIONS (/*g/m3)
Size
Cut
lO/i
2.5/t
Time
of
Day
Day
Night
Day
Night
Time
of
Week
wd
we
wd
we
wd
wea
wd
we
n
32
18
32
18
33
18
33
18
Min.
26.4
39.5
1.8
25.8
9.0
25.0
18.3
13.6
Max.
144.3
88.9
119.7
60.0
112.3
61.3
116.9
43.5
Mean
78.2
61.9
56.9
40.1
51.1
42.1
43.2
26.4
Std.
Dev.
31.6
13.8
29.1
10.3
29.1
12.6
24.5
10.9
CV
(%)
40
22
51
26
57
30
57
41
TABLE VII. SPATIAL COMPARISON OF OUTDOOR AND INDOOR
AEROSOL CONCENTRATIONS (/ig/m3)
House Number:
10/i Size Cut
Mean-Outdoor
MLA
Corr. Outdoor
& MLA
2.5ft Size Cut
Mean-Outdoor
MLA
Corr. Outdoor
& MLA
1
69
41
0.67
52
25
0.75
2
63
53
0.78
49
37
0.87
3
68
49
0.79
46
37
0.92
4
65
47
0.63
43
27
0.77
5
75
64
0.84
52
41
0.89
6
48
113
-0.64
25
61
-0.21
7
53
57
0.11
31
24
0.31
8
44
74
0.79
30
32
0.74
9
46
97
0.51
33
75
0.26
467
-------
MEASUREMENTS OF OZONE EXPOSURES
P. Koutrakis, and J.M. Wolfson
Harvard School of Public Health
665 Huntington Ave., Boston, MA 02115
J.L. Slater
University of Steubenville
Steubenville, OH 43952
J.D. Mulik
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
K. Kronmiller, and D.D. Williams
NSI Technology Services
Research Triangle Park, NC 27709
ABSTRACT
Recently, a personal sampling device was designed to measure
human exposures to ozone. This sampler, which is not sensitive to
other oxidants, is a filter pack containing a coated glass fiber
filter. The coating solution is a mixture of inorganic salts,
including nitrite ion. During sampling, ozone oxidizes nitrite to
form nitrate. Sampling flow can range between 0.1 and 1.5 Lmin'1.
Ozone concentration is determined from the amount of nitrate,
measured using ion chromatography. Results from laboratory and
field tests suggest a good agreement between measurements made with
the filter pack and with a UV-photometric ozone analyzer.
468
-------
I:HTRODDCTION
Ozone is an atmospheric oxidant formed through photochemical
reactions of volatile organic compounds and nitrogen oxides.
Although a great deal of effort has been made to decrease emissions
of ozone precursors, ambient concentrations of ozone have only
decreased approximately 10% over the last decade [1]. Daily
maximum l-hour ozone concentrations can range between 50 and 300
ppb, and often exceed the ozone National Ambient Air Quality
Standard of 120 ppb. For these ambient levels ozone can cause
respiratory health effects including changes in lung capacity, flow
resistance, and epithelial permeability [2]. While there is a
great deal of information about outdoor ozone concentrations, very
little is known about indoor concentrations. However, there is
some evidence that a large fraction of outdoor ozone penetrates
indoors [3]. The development of personal or microenvironmental
ozone samplers, can enhance our understanding of human exposures
to this harmful air pollutant. This paper discusses the
development and evaluation of such sampling methods.
OSSONE SAMPLER
The sampler consists of a filter pack and a small pump. The
filter pack contains a nitrite-coated glass fiber filter. The
coating solution is potassium and ^sodium salts of nitrite and
carbonate with glycerol. During sampling ozone oxidizes nitrite
to form nitrate:
NO2" + O3 > N03" + 02 [1]
Since the above reaction is pH dependent and its rate constant
increases with the pH, Na2CO3 is used to keep the collecting media
alkaline. Also, since the oxidation of nitrite by hydrogen
peroxide is fast only at low pH, this sampling technique is
insensitive to the presence of hydrogen peroxide which is also an
important oxidant.
Optimum collection efficiency is obtained when the nitrite and
carbonate in the coating have sodium as the cation for one and
potassium for the other. Results from collection efficiency tests
showed that ozone reacts with nitrite only when the nitrite and
carbonate come from salts of different metals. This may be
explained by the fact that the mixed potassium/sodium crystals
formed on glass fibers are more hygroscopic. Increasing the number
of water molecules at the surface of the crystals enhances the
oxidation reaction of nitrite by ozone. For this reason we also
include a hygroscopic compound, glycerol, in the coating solution.
Thus we speculate that the reaction between the ozone and the
coating material occurs through homogeneous aqueous reactions which
take place inside microscopic droplets.
469
-------
After sample collection, the glass fiber filter is transferred
to a 7.5-ml polyethylene bottle. Subsequently, 3ml of deionized
distilled water is added. The nitrate in this extract is measured
using ion chromatography. Since the number of moles of ozone
collected on the filter media is equal to the number of moles of
nitrate formed, see reaction [l], the air concentrations of ozone
are calculated from the nitrate concentrations. Of course nitric
acid gas and nitrate particles are collected simultaneously on the
alkaline filter media during ozone sampling. However, under usual
ambient conditions this positive interference represents less than
5% of the nitrate formed during the nitrite/ozone reaction.
Furthermore, reaction of ozone with organic aerosols collected on
the filter media can result in the underestimation of ozone
concentration. However, due to the amount of nitrite coating on
the filter relative to the organic aerosols, this interference is
also expected to be small.
The performance of the designed sampler was tested in
laboratory and field tests. For the laboratory tests known amounts
of ozone, ranging between 50 and 300 /ig, were generated, using an
ozone calibrator. Sampling periods varied between 1 and 12 hours.
Figure 1 compares the measurements made with the filter pack and
with a UV-photometric ozone analyzer. The ozone masses collected
by the filter pack were slightly lower (approximately 9%) , than
those generated by the ozone calibrator. However, the results of
these eight laboratory tests still show a good agreement between
the two methods. ' Considering that a number of other oxidant
species could interfere with the ozone measurements, a good
laboratory agreement does not necessarily qualify the designed
sampler for ambient monitoring. Therefore, the laboratory
experiments were followed by field tests. Two filter packs were
co-located with a UV-photometric ozone analyzer. Sampling
durations were approximately 24 hours. Figure 2 compares the
ozone concentrations obtained from the filter pack method and the
continuous ozone monitor. Again ozone measurements from the
continuous instrument are slightly higher, approximately 5%;
however this agreement is considered to be very satisfactory.
Further laboratory and outdoor experiments were conducted at
the U.S. EPA laboratories at Research Triangle Park, NC, during the
period of February 28 through March 28, 1990. The results are
given in Table 1. For the laboratory experiments, the exposure
time was 16 hours. Relative humidity varied between <10 and 60%.
For one of the chamber experiments, nitrogen dioxide was mixed with
ozone to investigate interferences of this oxidant with ozone
measurements. For all laboratory tests, as shown by Table 1, a
good agreement was found between the continuous U.V. analyzer and
the newly-developed ozone sampler. The relative difference between
these two methods was less than 10%. Furthermore, although only
one outdoor experiment was conducted, the results of Table 1
suggest again a good agreement between the two methods. The
observed difference was -9%.
470
-------
In conclusion, although the results from the above laboratory
and field comparison tests are limited, they strongly suggest that
the above technique can be used to determine ambient ozone
concentrations. More ambient tests must be conducted (different
locations and seasons) for a better evaluation of the method.
Finally, based on this limited number of data from the preliminary
tests we estimate an approximate precision of 9%. Most of this
variation was found to be associated with the blank variation. For
the future tests we will use filter coating with low nitrate
contamination.
DISCLAIMER: This paper has been reviewed in accordance with the
United States Environmental Protection Agency's peer review and
administrative review policies and approved for presentation and
publication.
REFERENCES
1. U.S. EPA "National Air Quality and Emissions Trends Report",
EPA-450/4-89-001, 1989.
2. Lippmann, M., "Health Effects of Ozone: A Critical Review",
Journal of Air and Waste Management Association. 39, 672-694,
1989.
3. Weschler C.J., Naik, D.V., and Shields, H.C., "Indoor Ozone
Exposures", Journal of Air and Waste Management Association.
JAPCA, 39, 1568-1567, 1989.
4. Schlesinger, R.B., Naumann, B.D., and Chen, L.C.,
"Physiological and Histological Alterations in the Bronchial
Mucociliary Clearance System of Rabbits Following Intermittent
Oral or Nasal Inhalation of sulfuric Acid Mist", Journal of
Toxicology Environmental Health. 12,441-465, 1983
471
-------
Table 1: Results from evaluation tests conducted at the U.S. EPA
Test
Type
chamber
Mil
Mil
M II
II II
II II
III!
Illl
II M
outdoor
Relative
Humidity
<10%
<10%
<10%*
35%
60%
<10%
<10%
<10%
<10%
outdoor
Exposure
Time (min)
960
960
960
960
960
960
960
960
960
1260
U.V. analyzer
ppb of 03
125.0
37.0
30.0
34.0
37.0
15.5
162.0
167.0
150.0
31.0
New method
ppb of 03
117.0
31.5
30.8
32.2
39.1
15.9
164.0
168.8
156.9
28.3
Difference
(%)
-7
-5
+ 1
-5
+6
+3
+1
+1
+5
-9
* In the presence of 76 ppb of NC>2
472
-------
D)
TD
OJ
-P
u
CD
O
O
300 i
250 -
200 -
150 -
cu 100
c
O
N
O
50
0
y-(0.913 +/- 0.027}x
0
50 100 150 200 250 300
Ozone Generated (ug)
Figure 1: Ozone Laboratory Experiment
473
-------
30
g
u
co
Q.
C_
CLJ
25 -
20 -
15
= (0.946 +/- 0.01B)x
15
20 25
Ozone Analyzer (ppb)
30
Figure 2: Ozone Ambient Experiment
474
-------
A PERSONAL EXPOSURE MONITOR FOR CO AND OZONE
William R. Penrose, Li Pan, Melvin W. Findlay,
G. Jordan Maclay, Joseph R. Stetter,
Transducer Research, Inc., 1228 Olympus Drive
Naperville, IL 60540, and
James D. Mulik and Keith G. Kronmiller
U.S. Environmental Protection Agency
Atmospheric Research and Exposure Assessment Laboratory
Research Triangle Park, NC 27711
A real-time personal exposure monitor/datalogger for nitrogen dioxide has beer
adapted to the measurement of ambient concentrations of carbon monoxide and ozone
Selective detection of CO at 0.1 ppm levels has been achieved by the careful selection o:
sensors, the use of an automatic zeroing technique, and the use of selective chemical filters
The sensitivity and stability of sensors for low CO concentrations has been empirically
related to the temperature coefficient of the baseline (no-sample) signal.
Ozone can be detected by the same sensor that is used for nitrogen dioxide
Selectivity is conferred by chemical filters which remove >95% of ozone and <5% o;
nitrogen dioxide. Critical to achieving sensitivity to ozone is the structure and individua
selection of sensors, the development of a selective chemical filter, and the partia
replacement of reactive plastics with Teflon. Depending on the sensor, detection limits o:
35-150 ppb have been achieved.
475
-------
Introduction
New discoveries in any field often follow the introduction of new techniques or
equipment. It is our purpose to develop instruments that facilitate the study of indoor air
pollution, in the hope that it will lead to new understanding of subtle pathological effects.
We have developed a real-time monitor and datalogger for very low concentrations of
nitrogen dioxide, and we have been working to adapt the same equipment to other injurious
gases (1). In this paper, we describe our progress toward the low-level measurement of
carbon monoxide and ozone.
Carbon monoxide is a difficult gas to measure in real time in the ambient (0 - 2
ppm) range. Infrared measurement is subject to interferences, and for all standard real-
time methods, the equipment is expensive and fragile, and requires expert operators. There
are excellent methods for measuring ozone at part-per-billion and even part-per-trilHon
levels, but again, the equipment is fragile and costly. Both gases are excellent analytes for
amperometric electrochemical detection, however.
There are many excellent reasons for choosing amperometric detection. These
sensors use virtually no power and are mechanically rugged, which suits them to portable
instruments. Since the materials of construction are inexpensive, there is no practical lower
limit to their cost; disposable sensors are a near-term possibility. We have already
successfully overcome many of the major technical obstacles to ppb-level detection by
amperometric methods in our earlier research on nitrogen dioxide. These obstacles, and our
responses to them, are outlined in Table I.
Table I. Practical and inherent problems with sub-ppm amperometric
detection, and how to deal with them.
Problem
Solutions
Output currents in nanoamps
Inherent noise
Baseline drift (temp.
coefficient of baseline)
Temperature coefficient
of span
Limited selectivity
High-performance op amps, metal
film resistors, etc.
Careful circuit layout
Active filtering of signal
Signal averaging
Very pure sensor materials and
clean construction practices
Autozero: frequent computer
controlled zeroing
Selection of sensor
Diffusion control of sample access
Post-measurement data processing
Chemically selective filters
476
-------
Ejcperimental Methods
Carbon monoxide dilutions were made from a 334 ppm standard supplied by
Matheson Co. A pump and flowmeter were used to dilute the concentrated standard with
"zero" air from a cylinder. The diluting air was scrubbed with Purafil, a platinum/alumina
ca.talyst, or potassium palladium sulfite, as described under Results. Purafil is a product of
Purafil, Inc., Doraville, GA. Platinum on alumina beads was obtained from Engelhard
Corporation, Iselin, NJ. Tubes containing potassium palladium sulfite were made by
Kitagawa, Inc. (distributed by Matheson Gas Products).
On-line dilutions of CO were generated by a custom-made diluter, which used Tylan
Corp. (Torrance, CA) Model FC-260 mass flow controllers under the control of a personal
computer. Any program of dilutions from 1 to 1000-fold could be set up and run on this
device.
Ozone was generated with a Thermo Environmental Corp. Model 565 Ozone
Calibrator. Clean air from a cylinder is passed through a UV radiation field at 2 L/min,
generating ozone concentrations which are adjustable from 0 - 0.5 ppm. Our instrument
was calibrated against a Thermo Environmental Corp. Model 560 ozone analyzer provided
by J. Woodring of Argonne National Laboratory.
All measurements were made using sensors manufactured by Transducer Research,
Inc., installed in a datalogging Personal Exposure Monitor (PEM) instrument developed by
TRI for the U.S. Environmental Protection Agency. The PEM is configured according to
Figure 1. The incoming sample stream is selected from one of two inlets by a low-power
solenoid. Different filters can be installed on the inlets to differentiate among analyte
gases. This instrument, and its performance for trace nitrogen dioxide, has been extensively
described in an earlier paper (Penrose et al., 1989).
Figure 1. Configuration of the Personal Exposure Monitor.
sanple lint
exhaust
cB5te niter
fitter
serial port
477
-------
Results
Carbon Monoxide
Because carbon monoxide was previously considered to be toxic only at high
concentrations, there has been little impetus to improve the sensitivity of amperometric CO
sensors. When first installed in the PEM, sensors that had passed QA for CO at
concentrations of 50-500 ppm were found to drift very badly in the 1-2 ppm range.
Reducing the specific surface area of the electrode by using vapor-deposited Pt instead of
Pt black is known to be effective in reducing drift. Unfortunately, it is also known to
abolish the CO signal (2). As an alternate approach, CO sensors were classified according
to the temperature coefficient of their baseline signal. This is a standard part of our testing
procedure. A sensor with a near-zero coefficient was installed in the PEM. The response
of the baseline output of this sensor with temperature is shown in Figure 2.
In Figure 3 is shown a calibration curve for CO at low concentrations using this
configuration, and in Figure 4, a picture of the behavior of the sensor output at the lowest
concentrations. The samples were diluted with zero air from a commercial cylinder, and
it is immediately obvious that there are several ppb of CO in the cylinder air.
Filters that remove CO are limited. Our standard CO filter is 1% platinum on
alumina beads. At room temperature, a column of this catalyst 1.5 cm diam x 12.5 cm long
was necessary to remove CO below the 100 ppb level at 400 mL/min. To generate air
that we assumed was completely stripped of CO, we passed cylinder air at 40 mL/min
through trace CO analysis tubes containing potassium palladium sulfite. Air thus prepared
was used to set the "true" zero of the PEM. For routine data collection, we
recommended a filter of activated charcoal and Purafil on the sample inlet. This
combination removes most organics but not CO. On the zero inlet, we use a 12.5 cm filter
of platinum/alumina to scrub CO.
Ozone
Results on ozone are preliminary but very promising. J. Spengler, Harvard
University, first alerted us to his observation that our nitrogen dioxide sensor also
responded to ozone. In order to selectively measure ozone or NO2 in the presence of each
other, it was necessary to find a chemical filter that efficiently removed one of these gases
but not the other. Ozone reacts with nearly all organic chemicals, but we eventually turned
up a mixture of organic acids and polyols that absorbed more than 95% of the ozone, while
absorbing less than 5% of the nitrogen dioxide. By placing this filter on the zero inlet of
the PEM, we can distinguish ozone from interfering gases.
The first experiment was done with the kind cooperation of Dave Halvorson of the
DuPont DMT Plant in Wilmington, North Carolina. In Figure 5 is shown the results of our
first attempt. At higher concentrations, the signal would not stabilize, but kept rising
apparently indefinitely. After the autozero phase, the signal did not start climbing all over
again, but took up again where it left off. We interpreted this as scrubbing of the interior
of the instrument by the ozone.
478
-------
Figure 2. For maximum performance, a sensor was chosen that had a minimal baseline shift
over the working temperature range.
500
1000 1500 2000
Time (sec)
2500
20
3000
Figure 3. Response of the calibrated instrument to increasing concentrations of CO from
an automatic dilution system. The autozero has been turned off.
2000
1500-
-o 1000
ra
50°
-500
10000
1250
840
540
270
0 10 20 30 40 50 60 70 80 90 100
Time (min)
479
-------
Figure 4. Responses of the PEM to carbon monoxide diluted into laboratory air and
cylinder (dry) air. The autozero function has been switched on. The spikes are electronic
noise from the switching of the autozero solenoid.
iOJUU-
1800-
1600-
1400-
E. 1200-
1,000.
800-
600-
400-
•wvt.
2.5
*-
*—
>pm/lab air
-**
w.
^™
1PF
lat
•
>ai
r
>m/
dry air
r
lat
*w.
jai
r
50 100 150 200 250
Time (min)
300
Figure 5. The signal from decreasing concentrations of ozone on a nitrogen dioxide sensor.
At high concentrations, an apparent "scrubbing" phenomenon is occurring. Note that the
baseline is different for the autozero and the sample measurements.
eo
2bU-
200-
150-
50-
-
en.
490 PPB /
/
/
/
/
•T-m
/
r
^
165 PPB
k^
u/l
r"
75 PPB
K . .
37 PPB
rn
OPPB
n
4
- X X t t
Autozero
20
40
60 80
Time (min)
160
480
-------
The gas path of the PEM was made of polyethylene, polypropylene, polyacetal, and
polyvinyl chloride; all of these react readily with ozone. Nevertheless, we were able to get
a fairly sensitive response to ozone, at least over the region of interest to industrial
hygienists. Note that the zero values through the two inlets are not the same. If this offset
is applied to the data, the points make a reasonably straight line (Figure 6). This was very
encouraging for an initial result.
An interesting feature of the ozone measurements using the autozeroing technique
is the very rapid return to zero when the autozero phase begins. This is probably due to
the very short lifetime of ozone in the instrument.
We have continued to characterize the ozone response. A variety of nitrogen dioxide
sensors have been tested, with variable results (Table H). We have compared the responses
of different sensors to NO2 and ozone concentrations of 500 ppb. The ratios of the ozone-
to-NO2 responses have varied from 0 to 1, that is, some sensors do not respond at all to
ozone. From a pragmatic point of view, we can segregate newly-made sensors into ozone
responders and non-responders, but it would be valuable to know what manufacturing
variables result in ozone-sensitive sensors. This is an area that we are actively pursuing.
Conclusions
(1) Carbon monoxide can be successfully detected^ concentrations down to 100 ppb
using amperometric sensors.
(2) Sensors with near-zero temperature coefficients for their baseline signal are
sufficiently free of drift to use as CO sensors in the ppb range.
(3) Ozone can be measured at concentrations down to 37 ppb, but the responses of
sensors are highly individualistic.
Acknowledgments
The majority of this work was funded by the U.S. Environmental Protection Agency,
Atmospheric Research and Exposure Assessment Laboratory, Research Triangle Park, NC.
Bibliography
1. W.R. Penrose, Li Pan, G J. Maclay, J.R. Stetter, J.D. Mulik, KXJ. Kronmiller,
"Personal exposure monitoring of nitrogen dioxide at part-per-billion levels:
Autozero and other corrections 1o electrochemical measurements," Proc. 1989
EPA/AWMA Symposium on Total Exposure Assessment Methodology: A
New Horizon, November 27-30, 1989, Las Vegas, NV.
2 J.R. Stetter, S. Zaromb, W.R. Penrose, T. Otagawa, A J. Sincali, J.O. Stull,
"Selective monitoring of hazardous chemicals in emergency situations, ttoc.
1984 JANNAF Safety and Environmental Subcommittee Meeting, Chemical
Propulsion Information Agency, Laurel, MD, 1984.
481
-------
Figure 6. The data from Figure 5, corrected for the zero offset.
250
160
200
Ozone (ppm)
Table II. Relative responses of sensors to nitrogen dioxide and
ozone.
Sensor
S/N
11-21-89-09
10337
11-21-89-05
3-7-90-02
3-7-90-03
11-21-89-04
Response to
nitrogen dioxide
( nanoamps/ppm)
46
55
50
60
35
not done
Response to
ozone
( nanoamps/ppm )
30
10
18
50
32
0
Ratio
O3/N02
0.65
0.18
0.36
0.83
0.91
__
482
-------
A HEALTH & WELFARE CANADA PROGRAM TO
DEVELOP PERSONAL EXPOSURE MONITORS
FOR AIRBORNE ORGANICS AT UG/M3
Rein Otson
Bureau of Chemical Hazards
Health and Welfare Canada
Rm.B-19, E.H.C., Tunney's Pasture
Ottawa, Ontario, Canada K1A OL2
The objectives, approach and progress in investigations conducted
by Health and Welfare Canada to provide suitable personal
monitoring methods are discussed. Existing and new, passive and
active, sampling procedures were examined for determination of
members of several classes of organics. Various sorbents, reagents
and devices for collection of the organics and analytical
procedures to allow selective and reliable determination over a
wide range of concentrations (0.1 ug/m3 to 500 mg/m3; 8 to 24 hr
exposures) were evaluated. Sampling tests were conducted over a
range of controlled test atmosphere conditions (e.g. temperature,
relative humidity, air velocity, co-contaminants) and in the
field. Procedures for determination of chlorinated, aliphatic and
simple aromatic hydrocarbons were developed and are now available.
Methods for selected aldehydes and polycyclic aromatics are at
various stages of evaluation and comparison.
Introduction
Current activities at the Environmental Health Directorate of the
Department of National Health and Welfare (Canada) include
investigations to determine the toxicity and potential human
exposure to substances such as polycyclic aromatic hydrocarbons
(PAH), aromatic amines (AA), aldehydes, methanol and chlorinated
and simple aromatic and aliphatic compounds. Exposure to these
organics may arise from their occurrence in indoor air due to
human activities and the use of certain materials in buildings or
through their presence during the development, production and use
of unconventional sources of energy, such as those derived from
coal and methanol. The principal objectives of the program are:
to assess the hazards to human health associated with indoor air
contaminants and those resulting from energy-related initiatives
483
-------
and developments; and, to make recommendations to minimize
potential health hazards.
In recognition of the need to define "acceptable air quality",
our Directorate, in collaboration with a Federal-Provincial
Working Group on Indoor Air Quality, recently published the
"Exposure Guidelines for Residential Air Quality". Also, a
protocol for investigation of indoor air quality (IAQ) problems
in offices and institutional buildings is in preparation. To
complement such guidelines and in response to and anticipation of
needs arising from energy-related initiatives, the Directorate is
conducting various research studies. The approach and recent
progress in the development of personal exposure monitoring
methods for several classes of organics are summarized here.
Experimental
Gas chromatographs (GCs) equipped with flame ionization detectors
(FIDs) were used for analyses of reagents and extracts from
chamber tests and were calibrated daily using standard solutions
over the full range of analyte concentrations. Capillary columns
(J&W Scientific Inc.) used were DB-WAX for volatile organic
compounds (VOCs) , aldehydes and methanol and DB-5 for PAH and AA.
Confirmatory analyses and analyses of field tests samples were
done with a Finnigan Ion Trap Detector (PAH and AA) and with a
Hewlett-Packard Mass Selective Detector (VOCs, aldehydes,
methanol) installed on Hewlett-Packard model 5890 GCs. Sampling
devices are briefly described in the appropriate sections.
Results and Discussion
Traditional integrative sampling methods for area and personal ex-
posure monitoring usually have been based on active sampling tech-
niques. Recently, passive sampling methods relying on diffusion
controlled analyte collection have gained some acceptance for use
in occupational monitoring of analytes, typically, at mg/m3 and
investigations in this laboratory1'2 have demonstrated their use-
fulness for determination of analytes at ug/m3. Although procedur-
es and availability of materials are often well established, some
disadvantages of active sampling methods are that they usually re-
quire equipment calibration and operation by experienced personnel
and are relatively cumbersome, inconvenient and expensive compared
to passive methods. There are few validated passive sampling me-
thods, sampling rates must be determined over a range of condi-
tions and the technique is not suitable for particle collection.
However, for most volatile organic compounds (VOCs) , the technique
provides sufficient sensitivity and is attractive due to its con-
venience and low cost. Therefore, our studies focussed, when
practical, on passive methods.
The selection of target compounds was based largely on reports of
their occurrence in the environments of concern, assessment of
their potential toxicity and consideration for their selective
determination together with other members of their group. In
addition, the compounds were selected to represent a range of
compound types within a class. The selected compounds were:
VOCs; n-hexane, n-decane, a-pinene, d-limonene, dichloromethane,
chloroform, 1,2-dichloroethane, trichloroethylene, tetrachloro-
ethylene, 1,1,2,2-tetrachloeoethane, penta- and hexachloroethanes,
benzene, toluene, ethylbenzene, o-,m-, and p-xylenes, styrene,
484
-------
p-cymene, naphthalene, 1,3,5- and 1,2,4-trimethylbenzenes,
m- and p-dichlorobenzenes, and 1,2,4-trichlorobenzene; methanol;
aldehydes; formaldehyde, acetaldehyde, acrolein, crotonaldehyde,
benzaldehyde; PAH; biphenyl, phenanthrene, benz(a)anthracene,
(naphthalene); AA; carbazole, 2-aminobiphenyl, quinoline.
Objectives and conditions for method development were:
Objectives; accuracy, within + 15 % of reference method; overall
precision, < 15 % RSD; collection and extraction efficiencies,
each > 75 %; storability of collected analyte on sorbent, 2 weeks
with < 10 % loss; passive device sampling rates, < ± 10 % over
range of evaluation conditions. An additional objective was to
use common materials and analytical procedures, where possible,
for the various classes of organics.
Evaluation conditions; exposure time-VOCs, 24 hr - aldehydes,
methanol, PAH & AA, 8 hr; method detection limit & quantitation
concentration range -VOCs & aldehydes, 10 & 50 to 5000 ug/m3 -
methanol, 0.5 & 2.5 to 500 mg/ms -PAH & AA, 0.1 & 0.5 to 50 ug/ms;
relative humidity, 10 to 95 %; temperature, 10 to 35 * C; air
velocity at passive sampling device, 0.05 to 1.8 m/sec. Effects
of light,pressure,etc. were not considered for these tests.
To determine the reliability of data on airborne contaminants and
human exposure, air sampling methods must be evaluated under a
wide range of atmospheric conditions. Effects of co-pollutants,
temperature, humidity and the physical state of target organics
are some of the influences which must be understood. Two test
atmosphere generating systems3'4 were designed and constructed to
provide accurate data on a wide range of controlled conditions
which are difficult to obtain in the "field" for sampling tests.
A system4 which allowed monitoring of artificially generated air-
borne analyte concentration by gravimetric and on-line GC methods
was used for most of the chamber studies on vapour phase organics.
VOCs
The home environment can contribute significantly to the exposure
of Canadians to airborne pollutants since they spend a large
proportion of their time indoors and apply various energy
conservation measures which can adversely effect indoor air
quality (IAQ). VOCs, such as chlorinated hydrocarbons, have been
identified among indoor, airborne substances which can cause
health concerns and require controlled exposure. However, there
is a lack of data on the occurrence of airborne VOCs in Canadian
homes, partly due to the lack of inexpensive and reliable
monitoring methods.
Although five commercially available devices were considered, only
the Pro-Tek G-AA (E.I. du Pont de Nemours Co.) and OVM 3500 (3M
Co.) monitors were evaluated extensively. The SIS monitor (Scien-
tific Instrumentation Specialists) was tested but was considered
too expensive for large surveys; the SKC monitor (SKC Inc.,catalog
no.530-01) design was judged unsuitable for study objectives; and
there was concern with the supply of Abcor GAS-BADGEs (Abcor
Development Corporation). Tests with 26 target compounds showed
recoveries > 75 % for most of the analytes extracted with carbon
clisulfide from the Pro-Tek and OVM sorbents spiked at levels
corresponding to the range of target concentrations. Although
included in subsequent tests, naphthalene showed variable and poor
recoveries (26 % overall) and its determination by use of these
charcoal based sorbents was considered semiquantitative. Difficul-
485
-------
ties with reliable quantitation was observed with some of the VOCs
at levels near the method detection limit (MDL). Problems with
reliable determination of styrene will be discussed elsewhere5.
Chamber tests were conducted with the simultaneous use of charcoal
tubes (SKC, 100mg/50mg) and exposure of both types of passive de-
vices at 0.01, 0.5 and 1.8 m/sec air (face) velocity. In summary,
no starvation effect was observed at 0.01 m/sec but the Pro-Tek
devices showed a significant increase (e.g. > 50 %) in sampling
rates (SRs) when the face velocity was increased from 0.5 to 1.8
m/sec and a slight increase (e.g. 10 %) in SRs with analyte con-
centration and temperature. The Pro-Tek device also showed greater
variability of SRs with changing relative humidity (RH) than the
OVM monitor. Since the effects of changing chamber conditions were
more pronounced with the Pro-Tek devices, only the OVM monitor was
evaluated for all target VOCs. Field tests with 10 of the target
VOCs showed excellent correlation (r > 0.95) between results ob-
tained with the OVM monitors and the charcoal tube (SKC , 400mg/
200mg) reference method, the passive method detection limits were
1 to 2 ug/m3 and the precision for replicate determinations (+ 7
to 10 %) was better than for the active method ( + 5 to 14 %) .
Results of studies with exposed monitors suggested that monitors
should be analyzed within two weeks since losses of about 20 % of
the collected anaytes could be expected after three weeks of
storage. Since the performance characteristics and reliability of
the method based on the OVM 3500 have been defined, a program to
monitor the occurrence of ca. 20 of the VOCs in Canadian homes is
scheduled to begin this year and after completion of a pilot study
to aid in establishment of the survey protocol. Meanwhile work
will continue on evaluation of a recently developed miniature
passive sampler using in situ extraction and analysis6.
Aldehydes and methanol
The development of personal exposure monitoring methods (PEM) for
airborne aldehydes and alcohols resulting from the use of vehicles
fueled by neat alcohol and alcohol blends was identified as a high
priority. As determined in reviews of the literature, there are
no suitable and/or validated methods for accurate determination
of personal exposure to airborne aldehydes7 and methanol at the
concentrations and conditions of concern.
Development of suitable active and passive methods has been slow
due to problems with standard atmosphere generation, method
blanks, recovery efficiencies, selectivity and other concerns.
Derivatization during collection was chosen since this potentially
provides better selectivity, sensitivity and storability than col-
lection of underivatized aldehydes. After evaluation of several
reagents, 2,4-dinitrophenylhydrazine (DNPH) was chosen for impreg-
nation of glass fibre filters subsequently placed into empty OVM
monitors (passive) and two stage, 37 mm filter cassettes (active)
and impregnation of XAD sorbent in Supelco ORBO-43 tubes (active)
or placed into OVM monitors (passive) and SKC monitors (passive).
A large number of chamber tests showed that the impregnated ORBO-
43 tubes provided results similar to those for the T05 reference
method8 adapted to GC analysis. Cassette samplers (active) demon-
strated analyte capacity limitations, reproducibility was poor for
the SKC samplers, there were discrepancies in calculated sampling
rates between passive samplers and acrolein and crotonaldehyde ex-
hibited low values for both active and passive samplers. Experi-
486
-------
irients were conducted, with some success, to find a suitable
wetting agent, e.g. glycerol, to cope with low recoveries at low
humidity. Presumably due to the reactive nature of aldehydes,
particularly the unsaturated compounds, it was difficult to
ascertain the true concentrations in the chamber by the gravi-
metric, on-line GC or T05 methods. Recently completed inves-
tigations suggest that the passive method using OVM monitors can
meet the stated objectives as effectively as the active methods.
However, the effects of co-pollutants, e.g. ozone, and other
factors on the collected analytes remain to be determined.
Initial evaluation of candidate sampling methods for methanol
showed that extraction with carbon disulfide gave low recoveries
(ca. 30 %) with Pro-Tek G-AA and OVM 3500 sorbents and that NIOSH
method no.2000 (150mg/150mg silica gel) was inadequate at the
target concentration range. Subsequently a variety of sorbents
(silica gel, Carbosieve, charcoal cloth, granular charcoal, OVM
sorbent) and extraction solvents (deionized water, butanol in
carbon disulfide, chloroform, etc.) were examined and preferred
candidates were chosen. Several chamber tests using large silica
gel tubes (700mg/150mg/150mg) for the reference method and OVM
3500 and charcoal cloth in empty OVM monitors for passive sampling
have been completed and show promising results.
PAH & AA
A critical literature review9 and a limited monitoring study10
emphasized the lack of suitable methods for reliable determination
of personal exposure to PAH and AA but provided guidance for
methods development and evaluation studies. Subsequently, active
sampling methods for particulate and vapour phase PAH and AA were
developed and evaluated11'12'13 and studies are continuing with use
of a special test atmosphere system3.
Development of a passive monitoring method for vapour phase com-
pounds has required extensive investigations to provide reliable
and accurately characterized test atmospheres and to find suitable
sorbents, extraction solvents and monitor design and to develop
sensitive analytical techniques. Current investigations suggest
that the use of ORBO-43 in OVM monitors may provide a suitable
sampling method. However, evaluation of candidate methods in test
atmospheres may not be completed due to a lack of resources.
Conclusions
Passive sampling methods for target VOCs, aldehydes and methanol
were characterized and the reliability determined for the range
of. conditions defined under study objectives. An initial objective
was achieved in part for the VOCs, aldehydes and methanol by the
common use of DB-WAX columns and the OVM 3500 monitor housing.
Instances of significant effects of atmospheric conditions were
identified. It is quite possible that some of the observed effects
are due to physical changes (e.g. dried sorbent, wet membrane) in
sampler structure.
Acknowledgement s
The contributions to research by P. Fellin and other personnel at
Concord Scientific Corp. and the Canadian Interdepartmental Panel
on Energy Research and Development are gratefully acknowledged.
487
-------
References
1. Otson,R.,Williams,D.T. ,Bothwell,P.O., Fabric protectors, Part
II. Propane, 1,1,1-trichloroethane and petroleum distillate
levels in air after application of fabric protectors, Amer.
Ind. Hyg. Assoc. J., 4J>. 28 (1984).
2. Otson,R.,Benoit,F.M., Surveys of selected organics in
residential air, in "Indoor air quality in cold climates.
Hazards and abatement measures", APCA, ed. D.S. Walkinshaw,
pp.224-233 (1986).
3. Fellin,P.,Brice,K.A.,Otson,R., Development of a test
atmosphere generation system for aerosols and vapours, in
Proceedings of APCA 80th Annual Meeting, New York, N.Y., June
21-26, 1987.
4. Fellin,P.,Otson,R.,Ernst, A versatile system for evaluation
of organic vapour monitoring methods, Proceedings of the 8th
World Clean Air Congress, 1989, The Hague, The Netherlands,
11-15 Sept., 1989, Elsevier Science Publishers B.V., Amsterdam,
eds. L.J.Brasser & W.C.Mulder, Vol.3, pp.675-680 (1989).
5. Tang,Y.Z.,Otson,R.,Fellin,P., Comparison of a charcoal tube
and a passive sampling device for determination of low concen-
trations of styrene in air, extended abstract submitted to
American Chemical Society, Symposium on Measurement of Air-
borne Compounds: Sampling, Analysis and Data Interpretation,
Aug. 26-31, 1990, Washington, D.C.
6. Otson,R., Miniature sampler using in situ extraction and anal-
ysis, J. Environ. Sci. Health - Part A, 21, 767-782 (1989).
7. Otson,R.,Fellin,P., A review of techniques for measurement of
airborne aldehydes, Sci. Tot. Environ., 77, 95-131 (1988).
8. U.S.E.P.A., 1984. Method for determination of aldehydes and
ketones in ambient air using high performance liquid
chromatography, U.S. EPA Method TO5, Revision 1.0.
9. Davis,C.S.,Fellin,P.,Otson,R., A review of sampling methods
for polyaromatic hydrocarbons in air, J.A.P.C.A., 37. 1397
(1987) .
10. Leach, J.M. ,Otson,R.,Armstrong,V., Airborne contaminants in two
small Canadian coal liquefaction pilot plants, Amer. Ind. Hyg.
ASSOC. J., 48, 693 (1987).
11. Otson,R.,Leach,J.M.,Chung,L.T.K., Sampling of polycyclic
aromatic hydrocarbons, Anal. Chem., 59. 1701 (1987).
12. Otson,R.,Hung,I.-F., Evaluation of a low-flow technique for
the determination of PNA in indoor air, in "Polynuclear
Aromatic Hydrocarbons: Mechanisms, Methods and Metabolism",
Batelle Press, Columbus, OH, eds. M. Cooke & A.J. Dennis,
pp.999-1012 (1985).
13. Otson,R.,Leach,J.M.,Chung,L.T.K., Sampling of airborne
aromatic amines, Anal. Chem., 59, 58 (1987).
488
-------
NEW DERMAL EXPOSURE SAMPLING TECHNIQUE
J. P. Hsu, David E. Camann, Herbert Schattenberg ffl, Bert Wheeler,
Kevin Villalobos, Michele Kyle and Shraddha Quarderer
Southwest Research Institute
San Antonio, Texas 78228-0510
Robert G. Lewis
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
A ring of polyurethane foam (PUF ring) on a stainless steel roller (PUF roller) is
designed to simulate a young child's dermal exposure as it is pulled over a sampling surface.
This is an economical and consistent sampling device which always applies constant weight on
the sampling surface. We found there is no statistically significant difference between the
recovery of any of 13 pesticide residues from aluminum foil by this sampling device as it rolls
over these residues and by dermal pressing on these residues with the human hand heel.
489
-------
INTRODUCTION
Young children may receive substantial exposure to pesticide residues on indoor surfaces
and in house dust, especially during the crawling and early walking stages. A sampling
instrument is needed which reliably reflects the exposure of young children to prevalent
pesticides on indoor surfaces. In our opinion, the sampling instrument should meet two
requirements. First, the absorbent used by the sampling instrument should have similar pesticide
absorption efficiency as the human skin. Secondly, the sampling instrument, while collecting
pesticides from the sampling surface, should apply constant pressure on the surface which is
similar to that of a toddler while crawling or walking. In addition, the dermal sampling area
should be easily measured.
The PUF roller sampler, which consists of a ring of polyurethane foam (PUF ring) on
a stainless steel roller (Figure 1), is designed to simulate the dermal exposure of a crawling or
walking toddler as it rolls over a sampling surface.
30cm
20cm
Figure 1. Schematic of PUF Roller Sampler
490
-------
The stainless steel roller provides constant pressure on a sampling surface through the PUF ring
as it rolls, because the pulling or pushing force of the person holding the PUF roller arm will
only be applied to the wheel, and not to the freely rotating roller axis. The sampling area rolled
over by the PUF ring is also easily measured. However, how does the PUF roller compare with
human skin in respect to collection of pesticides from the sampling surface? In this study, we
compare the PUF roller with the human heel by rolling the PUF ring over, or pressing hand heel
on, dry pesticide-spiked spots on an aluminum foil surface. The percentages of pesticides
collected by these two techniques will be compared and discussed. The usage of the aluminum
foil avoids the complexity of other sampling surfaces. The pesticides used in the experiment
are listed in Table I. These pesticides were selected because they are frequently found in
residential house dust and represent the major pesticide structural classes.
TABLE I. PESTICIDE EVAPORATIVE LOSS FROM ALUMINUM
FOIL SURFACE4
Spiked Pesticides
orf/w-Phenylphenol
Propoxur
Diazinon
Carbaryl
Heptachlor
Aldrin
Chlorpyrifos
y-Chlordane
a-Chlordane
p,p'-DDE
Dieldrin
Methoxychlor
Permethrin
Percent Recovery
Test 1
61
85
82
78
83
80
82
86
87
87
87
102
111
Test 2
101
136
115
217
98
98
102
103
103
105
104
105
106
Test 3
101
130
113
53
104
104
107
108
108
111
109
113
110
Mean Recovery
88
117
103
116
95
94
97
99
99
101
100
107
109
Last spiked residue spot allowed to dry
spiked foil squares are placed in Soxhlet
for 90 seconds before
extractor.
EXPERIMENTAL
Analytical Procedure
Four of the spiked pesticides (o-phenylphenol, propoxur, diazinon and carbaryl) arc
analyzed by gas chromatography/mass spectrometry (GC/MS). Gas chromatography with
electron capture detection (GC/ECD) and GC/MS are used as the primary and secondary
analytical methods, respectively, for the other spiked pesticides. Since 2-propanol is used to
wash the 13 target pesticides (Table I) from the human hand heel in this study, all injection
standards for both quantitation methods are prepared in 2-propanol. The internal standard
quantitation method is used for GC/MS analyses, while the external standard method is used for
GC/ECD analyses.1
491
-------
Aluminum Foil Sampling Surface Preparation
A new 120 cm x 30 cm dry acetone-washed aluminum foil strip is used for each test
involving the hand press, PUF roller, or surface pesticide evaporation loss. A total of 250 ^il
of a 400 jig/ml standard solution containing all the compounds in Table I is spiked onto the
aluminum foil surface, by using a 250-^1 syringe to uniformly transfer 25 \L\ of the solution to
the center of each of ten 10-cm sections of foil from 1 cm above the surface. Each application
takes approximately 5 sec. Using this procedure, each spike was observed to dry in a circular
disk (radius about 1.3 cm) on the aluminum foil surface in about 75 sec. All the tests involving
the PUF roller, hand press and pesticide evaporative loss from the aluminum foil surface
commence 90 sec after the last spike is applied and are conducted in the same laboratory
location (24.3-25.1°C at 50.5% relative humidity). This standarizes the amount of pesticides
left on the aluminum foil after the 90-sec drying period in each test.
Dermal Sampling by Hand Heel Press
Two subjects, one with dry hands and the other with moist hands, perform the hand
press recovery tests. The subject washes his hands with soap and water approximately 5 min
before each test. For the first hand press recovery test, the subject presses the heel of the hand
(beneath little finger to above wrist) over each dry circular spiked disk on the aluminum foil
surface for about 1 sec in a 90° rolling motion (from hand perpendicular to surface to hand flat
on surface) at a pressure of about 1 Ib/in1. After pressing the same hand heel over the remaining
nine spiked spots on the aluminum foil surface, the hand is immediately rinsed by squirting 20
ml of 2-propanol from a Teflon® squeeze bottle over the hand heel, with the rinsate passing
through a collection funnel into a vial. The collection runnel is rinsed with 10 ml of 2-propanol
into the vial. The 30-ml hand rinsate is then concentrated to 2 ml and split for GC/ECD and
GC/MS analysis. Each subject performed a second rinse of his hand heel with 2-propanol after
the first hand press test. The second rinsate is collected into a different vial, concentrated and
analyzed as before. The second rinsing monitors whether any pesticides remain on the hand
heel of each subject.
Two different hand motions were also studied. For the second hand press test, all
procedures are the same as the first test except the hand heel rests on each pesticide-spiked disk
for 5 sec. For the third hand press test, the only difference is to slide the hand heel over each
spiked disk.
PUF Roller Pesticide Recovery
The PUF roller sampler has an exchangable stainless steel roller. A PUF with 8.9 cm
OD and 8 cm length is cut in the center to form a ring with the dimensions shown in Figure 1.
After cleaning by consecutive Soxhlet extraction with acetone and with 6% ether/hexane, this
PUF ring is carefully slid onto the stainless steel roller with a dry hexane-washed Teflon® sheet
between the stainless steel roller and the PUF ring. The function of the Teflon® sheet is to ease
the PUF in sliding on and off the roller.
This study is designed to simulate the dermal exposure of a one-year old toddler to
pesticide residues on a hard surface. The stainless steel roller of the PUF roller used in this
experiment weighs 3.38 kg, which exerts a surface pressure of (3.38 kg x 9.8 m/sec2)/(0.080 m
x 0.057 m) = 7300 Pa while rolling over a surface. A crawling, 20-lb child will support his
weight with two hands (each approximately 2 in. by 3 in.) and two knees (each approximately
2 in. by 2 in.). The surface pressure when crawling is 20 lb/(2 x 6 in2 + 2 x 4 in4) = 1.0 lb/inz
= 6900 Pa. When walking, the child's weight is supported on two feet (each approximately
492
-------
2 in. x 4 in.), and the surface pressure is 20 lb/(2 x 8 in2) = 1.25 lb/in2 = 8,600 Pa. Hence the
PUF roller used in this study applies the same range of pressure as a one-year child when
crawling or walking.
To collect a sample, the PUF roller sampler is pulled once over a spiked aluminum foil
strip at a speed of 10 cm/sec. After a single pass, the PUF ring is carefully pulled away from
the roller using forceps. The forceps are then rinsed with 6% ethyl ether in hexane. The PUF
ring along with the rinsate is placed in a Soxhlet extractor. The PUF ring is then extracted with
6% ethyl ether in hexane for 18 h and the final extract concentrated to 2 ml for both GC/MS
and GC/ECD analysis.
Pesticide Evaporative Loss from Aluminum Foil Surface
The pesticides in hexane are spiked on the aluminum foil surface and the solvent
evaporated to leave pesticide residue spots. Experiments were performed to evaluate the loss
of pesticides due to evaporation during the 90-sec drying time.
Ten 4 cm x 4 cm dry acetone-cleaned squares of aluminum foil are placed on a 120 cm
x 30 cm aluminun foil strip and each aluminum square spiked from 1 cm above surface, with
25 {0,1 of a 400 u^g/ml standard solution containing all the compounds in Table I, allowing 5 sec
to apply each spike. The ten squares are collected in a Soxhlet extractor 90 sec after application
of the last spike and extracted with 6% ethyl ether in hexane for 18 h. The extract is
concentrated to 2 ml and split equally for GC/ECD and GC/MS analysis. This pesticide
evaporative loss experiment is performed three times in the same manner.
Extraction Efficiency of Pesticides from PUF Ring
Two clean PUF rings are each spiked with 250 ul of the 400 ng/ml standard solution
in hexane containing each compound in Table I. Each PUF ring is then Soxhlet extracted
separately with 6% ethyl ether in hexane for 18 h and each extract concentrated to 2 ml and
split equally for GC/ECD and GC/MS analysis.
RESULTS AND DISCUSSION
The analytical results of the pesticide evaporative loss from the aluminum foil surface
are shown in Table I. There is almost no loss of each pesticide from the aluminum foil surface
90 sec after the last spiking. Therefore, no correction is made for the evaporative loss of any
pesticide from the aluminum foil surface. Table n demonstrates good extraction recovery for
all the pesticides from the PUF ring. Thus no correction for extraction efficiency is made either.
After the first hand press test, each human subject rinses his hand heel twice. For each
pesticide, the ratio of the amount recovered in the first rinsing divided by the corresponding total
amount recovered in both rinsings is shown in Table ffl. It is clear that almost all pesticides
are recovered in the first rinsing.
493
-------
TABLE n. EXTRACTION EFFICIENCY OF PESTICIDE
FROM PUF RING
Spiked Pesticides
-------
double in this test. Therefore, sliding the PUF roller instead of just rolling over the pesticide
spiked spots will increase the pesticide recovery substantially.
TABLE IV. PUF ROLLER PESTICIDE RECOVERY1
Spiked Pesticides
ortho-Phenylphenol
Propoxur
Diazinon
Carbaryl
Heptachlor
Aldrin
Chlorpyrifos
y-Chlordane
a-Chlordane
p,p'-DDE
Dieldrin
Methoxychlor
Permethrin
Percent Recovery1"
Test 1
5.2
5.9
5.4
3.8
7.0
7.6
5.5
7.9
7.6
8.0
6.6
11
7.7
Test 2
7.8
9.7
6.9
5.9
8.0
7.6
6.9
6.8
6.9
7.5
6.6
6.6
6.7
TestS
10
7.9
7.2
6.4
8.5
8.5
8.8
8.7
8.4
9.4
8.4
8.9
9.2
Recovery
Mean
7.7
7.8
6.5
5.4
7.8
7.9
7.1
7.8
7.6
8.3
7.2
8.8
7.9
Std Dev
2.4
1.9
1.0
1.4
0.8
0.5
1.7
1.0
0.8
1.0
1.0
2.2
1.3
Recovery
Roller
Slide Test*
16
29
20
33
18
18
17
18
18
21
18
18
18
a The PUF roller exerted a pressure of 7300 Pa when rolling over the foil surface.
b The PUF roller rolled over 10 spots (total 100 ug for each pesticide) at 10 cm
apart at a speed of 10 cm per second.
c The roller accidently jammed after rolling over 7 of the 10 residue spots and slid
over the remaining 3 spots.
The hand heel press recovery of pesticides by both dry and moist hands using three
different hand motion over the ten dry pesticide residue spots are shown in Tables V and VI,
respectively. The mean pesticide recoveries were between 5 to 16%, with larger variation than
recoveries by the PUF roller. This larger variation may be due to the different types of hand
heel presses applied. The pesticide recoveries of the dry and moist hand heel roll presses with
1 sec duration have similar values. The pesticide recovery of the dry hand press for 5 sec has
similar recovery as that of the 1 sec roll press, but the dry hand slide collects more of the less
volatile pesticides than either of the other dry hand press motions. The moist hand press for 5
sec collects more of the less volatile pesticides than the I sec roll press by the same hand.
However, the moist hand has more friction while sliding over residues on the aluminum foil
surface which causes the moist hand to skip. The moist hand skipping may cause the lower
pesticide recovery than the dry hand obtained in the hand press slide. These data suggest that
the hand heel roll is more reproducible between subjects than other hand press motions.
495
-------
TABLE V. HAND HEEL PRESS RECOVERY OF PESTICIDES BY DRY
HAND USING THREE HAND MOTIONS OVER TEN DRIED SPIKED
PESTICIDE SPOTS
Spiked Pesticides
ort/w?-Phenylphenol
Propoxur
Diazinon
Carbaryl
Heptachlor
Aldrin
Chlorpyrifos
y-Chlordane
cc-Chlordane
p,p'-DDE
Dieldrin
Methoxychlor
Permethrin
Percent Recovery by Hand Heel Motion*
Is Heel
Press
8.4
5.4
6.4
9.3
9.5
9.5
7.7
8.6
8.6
8.6
8.8
6.7
6.4
5s Heel
Press
13
4.2
5.6
7.7
9.0
9.3
6.4
6.6
6.8
7.2
6.7
8.5
7.4
Heel
Slide
9.7
6.5
7.0
11
8.6
10.0
9.1
9.8
11
12
11
16
20
Subject 1
Mean
10.4
5.4
6.3
9.3
9.0
9.6
7.7
8.3
8.8
9.3
8.8
10
11.2
StdDev
2,4
1.2
0.7
1.7
0.5
0.4
1.4
1.6
2.1
2.5
2.2
4.9
7.6
All motions at hand heel pressure on foil of about 1 Ib/sq in. (7000 Pa).
TABLE VI. HAND HEEL PRESS RECOVERY OF PESTICIDE BY
MOIST HAND USING THREE HAND MOTIONS OVER TEN DRIED
SPIRED PESTICIDE SPOTS
Spiked Pesticides
-------
The null hypothesis of no difference in mean contact recovery between the PUF roller
and the hand press was evaluated against the two-sided alternative for each pesticide by the two-
sample t-test at the 0.05 signficiance level. All hand press recoveries of both subjects were
combined since mean dermal recovery did not differ significantly between the subjects for any
spiked pesticide. The results shown in Table VII indicate there is no statistically significant
difference in mean contact transfer recovery between the PUF roller and the hand press for any
of the 13 spiked pesticides.
TABLE VII. COMPARISON OF MEAN CONTACT
RECOVERY OF PESTICIDE RESIDUES BY PUF
ROLLER AND HAND HEEL PRESS
Spiked Pesticides
ortho-Phenylphenol
Propoxur
Diazinon
Carbaryl
Heptachlor
Aldrin
Chlorpyrifos
y-Chlordane
a-Chlordane
p,p'-DDE
Dieldrin
Methoxychlor
Permethrin
Mean Contact Recovery of
Residue from Aluminum
Foil, Percent
PUF Roller
7.7
7.8
6.5
5.4
7.8
7.9
7.1
7.8
7.6
8.3
7.2
8.8
7.9
Hand Press*
9.3
5.4
7.2
8.2
9.3
9.9
8.2
9.0
9.4
9.9
9.4
11.7
13.4
Statistically
Significant
Difference?"
(p-value)
No
No
(p=0.055)
No
No (p=0.08)
No
No
No
No
No
No
No
No
No
a For both subjects, since there was no statistically
significant difference in mean dermal contact recovery
between subjects for any spiked pesticide.
b Two sample two-sided t-test of difference in mean
recovery at o=0.05. p-values of borderline significant
differences in parentheses.
This finding suggests that the PUF roller may be an ideal dermal exposure sampling
technique. A follow-up study is evaluating the similarity of the PUF roller and hand press in
contact transfer recovery of pesticides in carpet-embedded house dust.
REFERENCE
1. J. P. Hsu, H. G. Wheeler, D. E. Camann, H. J. Schattenberg m, R. G. Lewis, A. E.
Bond, "Analytical methods for detection of nonoccupational exposure to pesticides," /.
Chromatogr Sci, 26: 181-189. (1988).
497
-------
A LOW COST SAMPLER FOR MONITORING WORKER EXPOSURE
TO HERBICIDE RESIDUES IN FOREST FIRE SMOKE
Charles K. McMahon, U.S. Forest Service,
Devall Dr, Auburn University, AL 36849
and
P. B. Bush, University of Georgia, Cooperative
Extension Service, Athens, GA 30605
ABSTRACT
Concerns have been raised about worker exposure to herbicide
residues in smoke from prescribed fires on herbicide-treated forest
sites. Modeling studies have shown that the risk to forest workers is
insignificant; however, the models had not been verified under actual
field conditions partly because detection devices needed to monitor
worker breathing zone smoke concentrations had been unavailable.
We developed a personal monitor and an area monitor which measure
concentrations of smoke particulate matter and airborne herbicide
residues in the breathing zone of workers who are engaged in normal
operations of prescribed forest fires. Both monitors employ 25 mm
diameter plastic disposable sampling cassettes which contain a glass
fiber filter and polyurethane foam as the collection media. The low cost
(approximately $1.50) of each loaded cassette permits a one-time use and
thereby eliminated cross contamination of samples and minimizes
logistical problems normally associated with sampler recycling.
The monitors were tested in the laboratory and successfully used on
operational prescribed fires at 14 sites in Georgia which had been
treated with herbicides at various time intervals prior to burning. No
airborne herbicide residues of hexazinone, triclopyr, imazypyr, or
picloram were detected (detection range of 0.1-4.0 ug/m3). Smoke
particulate matter concentrations up to 3700 ug/m3 were detected by the
personal monitors and 45,000 ug/m3 by the area monitors. These field
tests confirm earlier models which predict insignificant risk of forest
workers to herbicides in operational prescribed forest fires.
498
-------
INTRODUCTION
The application of herbicides to forest lands followed weeks later
by a prescribed fire is a widely used form of site preparation prior to
forest regeneration. In this process, the vegetation at a selected site
is often burned 45 to 180 days after herbicide treatment. This forest
management practice known as "brown and burn" has raised forest worker
and public concerns about possible exposure to herbicide residues in the
smoke from the fire.
Modeling assessments have shown that the risk to forest workers is
insignificant, even if the fire occurs immediately after herbicide
application (as might occur in a wildfire).*
Until recently, the only field measurements of smoke from herbicide
treated sites were made with single point samplers in limited unpublished
pilot studies in the South and West. The results from these studies
supported the model findings of no significant residues; however, no
direct correlation could be made to worker exposure since actual
breathing zone concentrations of residues in smoke were not measured
under field conditions. To provide that data and to validate the model,
we designed monitors to measure the concentrations of herbicides and
particulate matter in smoke from prescribed forest fires and measured
these concentrations in the breathing zones of workers on 14 operational
prescribed fires in Georgia.
METHODS AND RESULTS
SMOKE MONITORS. Two smoke-herbicide monitoring systems were developed
for use in this study. One system consisted of a personal monitor worn
by forest workers as they carried out their normal duties on the fires.
This would represent a realistic-operational scenario for exposure
assessment purposes. The second system was a portable area monitor which
could be deployed by research personnel and maintained in areas of high
smoke concentration. This would represent a worst-case operational
scenario. Both systems were designed to simultaneously trap breathing
zone concentrations of particulate matter and airborne herbicide residues
that might be present in the smoke.
Sampling strategy focused on recovery of parent herbicide compounds
which could volatilize in the heat of the prescribed fire. No attempt
was made to monitor or analyze for herbicide thermal decomposition
products. Under the operational and highly diluted field conditions,
the decomposition products would not be detectable or distinguishable
from the myriad of similar decomposition products which originate from
the burning of the dominant forest biomass fuels (leaves, needles, twigs,
etc.).
The sampling cartridge for both systems consisted of a 25 mm
diameter three-piece styrene acrylonitrile holder or cassette
(Nucleopore # 30015). This cassette is normally used for asbestos
sampling, and the 50 mm center section or extension cowl is designed to
prevent filter damage during open-face sampling. In our application, the
cowl was used to snugly hold a plug of polyurethane foam behind a glass
499
-------
fiber filter. Thus each cassette was loaded with a 25 mm binder-free
glass fiber filter (GFF) (Nucleopore Grade AA # 201618) in line with a 27
mm dia. x 35 mm long cylinder of polyurethane (PUF) foam (Analabs # RCS-
083). The GFF and PUF were supported by 25 mm porous plastic support
pads (Nucleopore # 220600) to prevent movement in the cassette (Fig. 1).
The cassette is held together by a compression fit of the three plastic
components. Insurance against leaks can be achieved by placing plastic
shrink seals or electrical tape around the two cassette joints.
To monitor the respirable range of smoke particles, the inlet to
each personal monitor cassette was fitted with a 10 mm Dorr-Oliver nylon
cyclone separator (MSA #456228) (Fig. 2). To make a secure connection
between the cyclone and the tapered cassette inlet, the plastic tip of
the cyclone vortex finder was replaced with a brass adapter (SKC #225-13-
3). Support for this connection was provided by a heavy duty rubber band
as shown in Fig. 2. The cyclone is normally operated at 1.7 L/min in
order to achieve a 50 percent collection efficiency for 3.5 urn particles
(aerodynamic diameter) in accordance with the American Conference of
Governmental Industrial Hygienists criteria for respirable particles. We
elected to operate the sampler at 4.0 L/min to improve our detection
limit for both particulate matter (via gravimetric methods) and
herbicide residues (via chromatographic methods). At 4.0 L/min, the
cyclone 50% collection efficiency is lowered to 2.3 urn according to the
findings of Blachman & Lippmann.2 To quantify the differences in smoke
particle collection efficiency for the two flow rates, a series of small
open burn experiments were conducted in a 85 nr greenhouse. Pine needle
and leaf litter were burned to generate smoke. Three pairs of personal
monitors w/cyclones were used to monitor the smoke on three separate
burns. The 1.7 L/min samplers collected 12 percent more particles than
the 4.0 L/min samplers (n=9, coefficient of variation 5.6 percent). This
is consistent with the particle collection efficiencies for the two flow
rates mentioned earlier.
Several investigators have reported the advantages and collection
efficiency of GFF for smoke particulate matter and PUF for pesticide,
especially when sampled at relatively high flow rates. When GFF and PUF
are combined in a single disposable cassette, they provide a means to
simultaneously sample smoke particulate matter and herbicide residues.
In our prior work,3'4 the GFF proved to be the primary collector for both
particulate matter and herbicide residues and the PUF simply provided a
back-up for possible herbicide breakthrough, especially if the sampler
was exposed to elevated temperatures from the combustion process. To
meet the objectives of this study, no pre-treatment or clean-up of PUF's
or GFF's was necessary.
Commercial samplers are available which employ the GFF/PUF concept;
however, they are too bulky for personal monitoring and for use in a
remote field environment. A small portable lightweight system was
needed for this study. The Nucleopore sampler was selected over other
commercially available monitors because of the fortuitous compression
fit of the 27 mm PUF into the 25 mm center section of the 3-piece
monitor. In many earlier studies, PUF specifications would normally
require tedious custom cutting and fitting of PUF cylinders out of a
larger sheet of foam. In this case, all materials were commercially
500
-------
available at a low cost and usable without further treatment or
modification. Total cost of each loaded cassette (without the cyclone)
was $1.58 which permitted a one-time use and thereby minimized
logistical problems associated with sampler preparation, storage,
transportation, and recycling. Cross contamination problems were also
eliminated. The cassette just described was adapted into a personal
monitoring system and an area monitoring system.
a. Personal Monitor: The components of the personal monitoring
system consisted of a small 1 kg battery operated pump (SKC Model 224)
and carrying case. The pump is equipped with a timer, flow disrupt
indicator, and internal flow controller to maintain constant flow
conditions. The cassette is connected to the pump by flexible tubing
which allows the cassette to be clipped near the workers breathing zone
while the pump is worn comfortably around the waist, allowing the
workers to carry out their normal duties (Fig. 3).
Our desired lower detection level for airborne smoke particulate
matter was 150 ug/m3, roughly twice the ambient particulate matter
background. For herbicide residues, our target airborne detection level
was 1 ug/m3. This level is well below the concentration range predicted
by the modelling method cited in the introduction. It is also a level
which is several thousand times below the Threshold Limit Values (TLV)
and Permissible Exposure Limits (PEL) for the herbicides 2,4-D (2,4-
dichlorophenoxy acetic acid) and picloram. Not all herbicides have
published TLV's or PEL's; however, the ILV values for 2,4-D and picloram
provide useful benchmarks for occupational risk assessment for the
herbicides used in this study. Even when compared to herbicides which do
have a high inhalation risk, the 1 ug/m3 detection level is still
conservative. For example, it is 100 times below the TLV for paraquat
which has the lowest TLV of all herbicides.
Our personal monitoring strategy anticipated that breathing zone
smoke concentrations could vary by a factor of ten depending on weather,
fire conditions, and worker assignment. In addition, smoke exposure
times in any given day would vary depending on the size and number of
tracts to be burned. In the case of a 4 hour exposure; a monitor with
flow rate of 4 L/min, coupled with a gravimetric analysis of the filter
to the nearest 0.01 ± 0.005 mg would yield a smoke particulate matter
detection level of 104 + 5 ug/m3. Those same conditions when coupled
with a herbicide chromatographic detection level in the lab of 1
ug/sample would yield a herbicide detection level of 1 ug/m3.
b. Area Monitor: To simulate worst case operational conditions on
each site without compromising normal worker behavior and operations, an
area monitoring system was also deployed on each fire by research
personnel. The need for rapid deployment and repositioning during the
fire called for a relatively lightweight, battery operated system which
could be carried by one person in relatively uneven forest terrain. In
addition, a system with higher flow rates than available on personnel
monitors was desirable to enhance our detection capability. The system
used on this study consisted of a small, 10.5 kg air sampler (Gilian
Model Aircon 520 DCT) powered by a small, 9.5 kg 12 V deep cycle marine
battery. The aircon 520 consists of a integral flow controller mounted
501
-------
on a 12 V dc pump and shock mounted into a compact enclosure. It also
includes a built-in rotameter, a flow control valve to adjust the
flowrate from 2 to 22 L/min, a suction/pressure gauge to indicate actual
load, an elapsed timer, and a built-in telescopic sampling mast to permit
"breathing zone" sampling at the 5 ft level. The sampling cassette used
on the area monitor was identical to the one used on the personnel
monitor minus the cyclone particle size separator. Operating the area
monitor without the cyclone permitted collection of both respirable smoke
particles generated by the combustion process as well as some of the
mechanically generated and inhalable soil/dust particles. Thus the area
sampler provided a means of collecting herbicide residues which could
volatilize and residues made airborne by mechanical disturbances of the
soil. Operation of the area monitor at 20 L/min (5 times the rate of the
personal monitor) would achieve the desired herbicide detection level of
1 ug/m^ in only 50 min of sampling.
Preliminary tests with the small marine battery demonstrated at
least one hour operating time at 20 L/min even under high smoke
concentrations. While larger batteries are available, the small battery
permitted one person to easily transport the entire system with the
weight balanced between both hands (Fig. 4). Backup batteries and
preloaded cassette samplers were available on site to monitor for longer
periods or to obtain time-sequenced samples as dictated by fire and smoke
conditions on the site.
c. Herbicide Recovery Experiments: The trapping efficiency and
laboratory recovery-of pesticides from GFF's and PUF's had been shown to
be effective in prior laboratory combustion studies.3'4 However, a
retention efficiency experiment under the highest sampler flow conditions
(20 L/min) was deemed necessary and performed for each of the herbicides
to be studied. Recoveries are shown in Table 1. Results are given in
conjunction with freezer storage periods which approximate the storage
period for test fire samples. All of the residues recovered from
hexazinone, triclopyr and picloram'were found in the glass fiber filters.
A small percentage of the overall recovery (11%) of imazypyr was found in
two of the four PUF samples analyzed. These filter/PUF recovery ratios
are consistent with results from our prior studies with pesticides and
with the results of Roberts and Ruby^ who found less than 0.1% of
semivolatile pesticides collected from dust particles to pass through the
filter to the PUF. These findings question the need to include the PUF
component as a part of the collection media when dealing with low
volatility chemicals such as the herbicides used in this study.
However, to be on the safe side, we included the PUF component in this
study as a backup collector, even though it required a separate analysis
and therefore increased analytical costs.
SMOKE MONITORING PROCEDURES. The forest workers conducting the burning
operation were outfitted with personal monitors. The pumps were
activated just prior to ignition and deactivated when the crew reassembled
to leave the fire site. If there was more than one fire in a day, the
process would be repeated with fresh cassettes for each fire. Workers
were encouraged to perform their duties using their normal operational
procedures.
502
-------
At each fire site, three area monitors were deployed to separate
locations and repositioned as necessary by research personnel to areas
with high breathing zone smoke concentrations. All of the sites had
numerous access areas (firelines, paths, roads) in and around the burn
area which allowed moving the area samplers without compromising personal
safety. After initial placement and activation of the area monitors, the
research personnel moved back to a relatively smoke free area and
periodically visited the area sampler site to check flow rates, move the
sampler and/or change cassettes as necessary. The time of operation for
the area sampler varied depending on smoke conditions and filter
loading. At the end of each sampling pericd and/or each fire, pump
operating time was recorded and flow rates were visually verified. The
labeled cassettes were capped and stored ir an ice chest for return to
the laboratory for gravimetric and chromatcgraphic analysis.
The monitors were used successfully or 14 operational prescribed
fires in Georgia. Tract size ranged from 6 to 380 acres. No herbicide
residues were detected in either the personal or area monitors (detection
range to 0.1 to 4.0 ug/m3). Worker respirable (2.3 urn particle cutpoint)
particulate matter concentrations ranged from 248 and 3,723 ug/m3 with a
mean of 1,429 ug/m3. Exposure times ranged from 1.2 to 6.3 hours with a
mean of 2.8 hours. Area monitor particulate concentrations (ranged from
2,000 and 45,000 ug/m3 with a mean of 8,215 ug/m3.
ACKNOWLEDGEMENT
This study was supported by the USDA National Agricultural Pesticide
Impact Assessment Program.
REFERENCES
1. Dost, F.N. 1982. Combustion of Herbicides. Unpublished report
prepared for Bonneville Power Administration. Available from author.
Dept. of Agricultural Chemistry, Oregon State Univ, Corvallis,
Oregon. 24pp.
2. Blachrnan, M.W.; Lippmann, M. 1974. Performance Characteristics of the
Multicyclone Aerosol Sampler. Am. Ind. Hvq. Assoc. J. 38:311-326.
3. McMahon, C.K.; Clements, H.B.; Bush, P.B.; Neary, U.G.; Taylor, J.W.
1985. Pesticides released from burning treated wood. In: Donoghue,
Linda R.; Martin, Robert E., eds. Proceedings of the eighth
conference on fire and forest meteorology. 1985 April 19-May 2;
Detroit. Bethesda, MD. Society of American Foresters: 145-152.
4. McMahon, Charles K. ; Bush, P.B.; 1986. Emissions from burning
herbicide treated forest fields -- a laboratory approach. I_n:
Proceedings, 79th Annual meeting of Air Pollution Control
Association; 1986 June 22-27; Minneapolis, MN; Pap No. 86-94 P. 8.
Pittsburgh, PA: Air Pollution Control Association. 2 p.
5. Roberts, J.W.; Ruby, M.G. 1989. Development of a high volume surface
sampler for pesticies in floor dust. EPA/600/S4-88/036, USEPA
Research Triangle Park, N.C. 6p.
503
-------
Table 1. Percent herbicide recovered from spiked sampling cassettes.3
Herbicide Without aspiration With aspiration13
percent
freezer storage period
at -10SC
weeks
Imazypyr
Hexazinone
Triclopyr
Picloram
73 + 20C
95+8
125 + 7
85 + 7
89 + 14
92+8
88+2
71 + 14
12
6
11
14
a20 ug spike all samples (n=4).
b20 L/min for 1 hr at 78°F and 55% RH.
cMean (± SO).
Figure 1. Smoke sampling cassette (upper), exploded
view (lower). A, end plugs; B, inlet; C, glass fiber
filter; D and G, porous plastic support pads; E,
polyurethane foam; F, 50 mm extension cowl; H, base.
504
-------
Figure 2. Smoke sampling cassette with 10 mm nylon
cyclone connected to inlet for sampling respirable
particles.
Figure 3. Personal monitoring
system for prescribed fire
smoke. Sampling cassette
with cyclone is worn at
workers breathing zone.
Figure 4. Area monitoring
"w^t.pm fnr nrp^rrihpd fi
lyurt: t. rtr ed muiiiLUi iny
system for prescribed fire
smoke. Sampling cassette
is located on telescopic
mast at the 5 ft breathing
zone level.
505
-------
DEVELOPMENT OF A TEST ATMOSPHERE GENERATION FACILITY
FOR PARTICLE BOUND ORGANIC COMPOUNDS
Philip Fellin , Rein Otson , Darel! L. Ernst"
"Concord Scientific Corporation
Downsview, Ontario, M3H 2V2
"Health and Welfare Canada
Tunney's Pasture, Ottawa
ABSTRACT
A test atmosphere generation chamber has been developed to generate particles
containing PAH and volatile organic compounds. Initially, work emphasized the
characterization of the apparatus and development of particle generation methods using
generators based on dust feeding or vibrating orifice aerosol generation techniques. Some
of this work has been reported elsewhere. In current work, further development has been
performed on generation methods and the apparatus has been used for its intended purpose
(i.e., to examine sampling methods for these compounds). Specifically, tests on a PAH
vapour condensation method were performed to develop a suitable protocol for generation
of artificial aerosols containing PAH. Also, different types of sampling trains suitable for
indoor air quality assessment and determination of airborne compounds in occupational
situations were examined. Included in the testing were investigations of aerosol generation
of particle bound organics and of performance of different sampling trains for airborne PAH.
Some features of the apparatus and preliminary data are presented demonstrating the utility
of the apparatus and outlining some comparative results.
INTRODUCTION
Measurement of ambient air in residential and occupational environments for volatile,
semi-volatile compounds {light PAH) and particle bound organics (heavy PAH), required the
development of suitable measurement methods. A prerequisite for this activity is
development of a laboratory test atmosphere generation system. Hence, the creation of
such a facility was established as a primary objective. The technical specifications of the
506
-------
various components were established from thu basic principles of test atmosphere
generation as set out by G.O. Nelson (1971) and on the basis of a literature review (Guerin
and Fellin, 1985). A preliminary design of a test atmosphere generation system (TAGS) was
prepared that included various generation options for particles and vapours. In subsequent
phases of the program the apparatus was assembled and subjected to preliminary tests to
evaluate its performance with a tracer gas, CO2. One of the particle generation options
identified, a dust generator instrument, was also adapted to the inlet (Particle Source -
Figure 1) and utilized to conduct preliminary characterization of the apparatus. These tests
were reported in Fellin et al., 1987 and Fellin et al., 1988. Further testing of the apparatus
using a vibrating orifice monodisperse aerosol generation method (VOAG) as outlined by
Berglund and Liu (1973) was performed. Coupling the VOAG and a PAH generator based
on diffusion tubes was attempted to generate particle-bound organics. Preliminary results
of these tests were reported in Fellin et al., 1988a. This paper reports on further testing and
optimization work which has been performed since these initial studies were completed and
evaluates the progress made toward establishing a test atmosphere generation facility
suitable for particle-bound and volatile organic compounds.
EXPERIMENTAL
J
The apparatus employed for testing is shown in Figure 1. The major components
were the particle generation instruments (Figure 1A aid 1B). These were the Wright Dust
Generator (WDF) Instrument (Mode! II, BGL Inc.) and a custom designed apparatus that
incorporated a vibrating orifice aerosol generation (VOAG) apparatus (Berglund - Liu type,
Model 3050, Thermo Systems Inc. (TSI)). Oust generation employed NIST standard
reference material no. 1649 that was certified for 5 PAH. Data on 14 other PAH compounds
was provided in NIST documentation. The diluent air source, chamber and exhaust system
nave been described previously (Fellin et al,,, 1988). Particle size distributions and
concentrations were measured using a piezoelectric qiartz micro balance 10 stage impactor
(QCM) (Model PC2 California Measurements Inc.), an Andersen 8 stage impactor (Model
0704, Andersen Inc.) and 37 mm polypropylene intecirative filter units incorporating glass
fibre filters (Type TX 40H120-WW, Pallflex Products). Vapour concentrations of PAH were
monitored using Orbo 43 sorbent (Supelco Inc.) Cyclones for some of the sampling trains
were provided by Levitt Safety Ltd. (cat. no. IV3799) and had a particle size cutoff of 50%
at 2.5 n,m when operated at a flow rate of 1.7 L/inin. Gravimetric analyses were performed
on a Sartorius (Model 2004 MP, Sartorius Balance) 5 place analytical balance. For vibrating
orifice aerosol characterization tests, filters were extracted with water and extracts were
analyzed on an ion exchange chromatograph (Model 10, Dionex Corp.). PAH were
extracted from filters and sorbents with methylene chloride and extracts were analyzed by
gas chromatography/mass spectrometry using an HP 5890 GC fitted with a 30 m DB-5
capillary column. The GC column was directly inserted into the source of an Ion Trap
Detector (Model 800, Finnegan Instruments). For some tests excess PAH vapours were
removed by using a hollow tube denuder placed n the transfer tubing between the
condensation chamber and the aerosol dilutor (Figure 1B). Charcoal impregnated GoreTex
(W.L Gore and Associates) inserted inside the ' inch I.D. x 12 in. long transfer line served
as the sorbent.
RESULTS
Previous tests had investigated the performance of the WDF in the TAGS and
established optimal operational settings. The particlo size distribution achieved with the
WDF (Figure 2) demonstrates significant bias toward particle sizes greater than 2.5 microns.
Since the dust generated was based on NIST standard reference material no. 1649 Urban
Particulate Dust which had nominal particle sizes or mass median diameters (mmd) below
2.5 microns when it was collected, the results sugciest aggregation of particles during
507
-------
generation steps. Homogeneity of particle distribution was examined by sampling at 9 points
inside the exposure zone of the chamber (Figure 1C) over 6 to 8 h with integrative filter
samplers. The results are featured in Table 1 for particle mass collected and Figure 2 for
particle size distribution. Similar results were evident when data were stratified vertically (Y
results) or horizontally (X results). Overall reproducibility using gravimetric analysis, as
defined by the precision estimates, approached ± 15%. It is important to recognize,
however, that mass collected on each filter element was relatively low (0.01 g or less)
because of the low flow rates employed (1-2 L/min) during the tests. Hence significant
imprecision in the gravimetric measurement step is introduced. The results suggest the
particle concentrations in different locations of the chamber were statistically
indistinguishable and sampling conducted in the exposure zone would not be biased by
sampler location.
The vibrating orifice aerosol generator was employed to assess several operational
aspects of the TAGS performance. The production of sub-micron sized aerosols was
establised as a primary objective. Several salt (NaCI) solutions of different concentrations
were tested as feed solutions for the VOAG to achieve this, but no success was achieved
until a TSI representative suggested that using impurities in the isopropanol solutions without
salt addition could lead to consistent production of sub-micron sized monodispersed
aerosols. Monodisperse aerosols were produced in the TAGS with a mass median diameter
of 0.8 microns by using the unmodified isopropanol, as determined with the QCM
microbalance. Comparison of salt impregnated solution and isopropanol solutions provided
in Figure 3 describes the size distributions achieved. When PAH was added to the particles
by the vapour condensation method shown in Figure 1B, a coating efficiency for
phenanthrene of 2-3% was found. The net impact of this was to allow much of the PAH to
enter the TAGS exposure zone in the vapour form, A diffusion denuder was created to
remove the vapour PAH by insertion of charcoal impregnated GoreTex in the transfer tubing
positioned ahead of the TAGS mixer (Figure 1). GoreTex was found to be a strong sorbent
for PAH in previous work. Calculations performed according to Gormley and Kennedy
(1949) to estimate denuder performance and Fuller et al. (1973) to evaluate PAH diffusivity
suggested a theoretical removal efficiency using a hollow tube configuration for the denuder
of 0.23 under one set of conditions. Tests indicated that the maximum removal achieved
was 0.20. This agreement is quite reasonable considering the denuder dimensions
employed and the uncertainties inherent in the estimation of the diffusivity of phenanthrene.
The relative success of these tests suggests it may be worthwhile to redesign the inlet to
incorporate an annular denuder to improve the removal efficiency for the vapour component
of the PAH to 95% or better.
The VOAG/PAH condensation technique was also employed to conduct preliminary
tests of sampling trains consisting of sorbent, filter-sorbent and cyclone-filter-sorbent
configurations. In these tests it was found that collection of phenanthrene by use of a
sorbent tube gave the highest results (280 ± 8% ng/sample) compared with 220 ± 8%
ng/sample for filter/sorbent tube combinations and 200 ± 6% ng/sample when a cyclone was
added. Al! results quoted were normalized to the same total volume of air sampled and
were adjusted for recovery efficiency. The results indicated differences among the different
sampling methods of up to 40% in the data for one PAH. These differences are difficult to
explain based on the random sampling positions employed for sampling and the careful
attention paid to monitoring and controlling the volume of air collected.
In conclusion, the TAGS has been established as a test facility. Its operation is
suitable when using dusts generated by means of WDF methods. However, this method
produces significant particle concentrations in the size ranges greater than 2.5 microns.
Aerosol generation via the VOAG has achieved particle sizes in the sub-micron range and
PAH precoated particles can be generated. However, the condensation process is relatively
508
-------
inefficient with phenanthrene, the PAH used for testing. Removal of excess vapour form of
PAH has also been demonstrated using a diffusion denuder. Results suggest that efficient
use of the denuder requires redesign of the inlet to achieve removal efficiencies of 95% or
better since the current inlet can only achieve removal of 20%.
REFERENCES
Berglund, R.N. and Liu, B.Y.H. (1973), "Generation of Monodisparse Aerosol Standards," Environ. Sci. Technol.,
Vol. 7, No. 2. pp. 147-153.
Fellin, P. and Brice, K.A. (1985), "A Report on the Construction and Testing of a Test Atmosphere Generation
System (TAGS)." A Report to Health and Welfare Canada. Concord Scientific Corporation, Report
J790, March 1986.
Feliin, P., Brice, K.A., Ernst, D. (1988), "Characterization of thi» Test Atmosphere Generation System (TAGS)
for Aerosol Generation Using the Wright Dust Foedar (WDF)." A Report to Health and Welfare Canada.
Concord Scientific Corporation, Report J1026-1, Marcn 1988.
Fellin, P., Brice, K.A., Ernst, 0. (1988a), "Testing of a Vapojr Condensation Technique for Generation of
Aerosols Coated with PAH." A Report to Health and Welfare Canada. Concord Scientific Corporation,
J1026-4, March 1988.
Fellin, P., Brice, K.A., Otson, R. (1987), "Development of a Test Atmosphere and Generation System for
Aerosols and Vapours." Proceedings of the 80th APCA Annual Meeting, New York, Naw York, June
21-26, 1987.
Fuller et al. in Perry, R.H. and Chilton, C.H. (1973), "Chemical Engineer's Handbook. 5th Ed., McGraw-Hill Book
Co., New York, New York, pp. 3-230 to 3-234,
Gormley, P.C. and Kennedy, M. (1949), "Diffusion from a Stream Flowing Through a Cylindrical Tube,"
Proceedings of the Royal Irish Academy 52A. pp. 163 169.
Guerin, S.G. and Fellin, P. (1985), "A Review and Design of Test Atmosphere Generation Systems for Aerosols
and Vapours." Volumes I and II, A Report to Headh and Welfare Canada, Concord Scientific
Corporation, Report J556, April 1985.
Nelson, G.O. (1971).."Controlled Test Atmospheres, Principles and Techniques", Ann Arbor Science Publishers,
Inc.
TABIE 1
Determination of Chamber Bias by Monitoring at Different
Sampling Points with intejjratlve FI(tera
» Longitudinal Precision (n - 1?)
X1 ±
Xg d:
Xg * 14%
~ - * • Vertical PredsiGn (n * 17}
V4 ± 11%
Y2 ± ?%
V 4- 7<&
Tg ± ^ /»
Precision ± 15%
Mo <*>n$isteiit vertical or hori2ant& bias was dotact.
509
-------
FIGURE 1
Test Apparatus
Aerosol Oilutor
(or
Inline Mixing
Air Source
TAGS
QCM
Particle Sources
1A Wright Du«t F*Ml (tenarator
1B Vapour Condensation Vibrating Orrflca Aaroal Generator
Cond«nM1ion Chwnbar
Airoul
Dilute
TAOSChwnbw
Vibrafng
Orrfici
Tub«FOm<»
Contaning
Difhjiion TubM
1 6 LTran Ni»og«n
250 L'mn ComprMud Air
2 L'mn Dtipwino Row
17LAiw>Ditu««Roi»
1C Tags Chamber
EXPOSURE CHAMBER
(Not to Scale)
310
160
-1190
denotes sampling port locations
many tests reported previously
All dimensions in cm.
EXHAUST
510
-------
FIGURES
Evaluation of WDF as Particle Feeder for TAGS
Typical Size Distribution Results
(NIST 1649 Dust)
I15
rF
ANDERSEN
Ti
0-02 O.OS 0,1 0,2 0.5 1 2 5 '0 20 50 100
parbcal diameter (urn)
MMD
1.9 um
1.6 um
FIGURE 3
Typical Particle Size Distribu
I 15
f **
1 0.5
1
r "
* 1.0
NaCI - 1 1 1 ppm
-
4 I I r i
Isopropyl Alcohol
-
-
0.02 005 0.1 0,2 05
~l
1
r
1 2
i i i
tion
CONCENTRATOR (u»'m3| MMO (um)
ACTUAL THEORY ACTUAL THEORY
26.4 32.7 2,15 1.33
8.2 0 0.8 0
5 10 23 50 100
partical diameter (um)
511
-------
Spectroscopic Identification of Organic Compounds by Library
Searching: Methods, Potentialities, Limitations.
J. T. Clerc
Dept. of Pharmacy, University of Berne
Baltzerstr. 5
CH-3012 Berne, Switzerland
1. Overview.
Library search for the structure elucidation
and identification of organic compounds
seems to be a very simple and straight forward
technique. First we record the spectra of the
sample at hand. The spectra of the sample are
then compared to all reference spectra in a
reference library. If there is a sufficiently close
match between the unknown sample's spectra
with one of the reference spectra, the chemical
structure of the respective reference com-
pound is believed to be similar to the sample's
chemical structure. This process is schemati-
cally represented in fig. 1. To gain insight into
the potentials and limitations of this seemingly
simple process a somewhat more detailed sys-
tem analysis is helpful.
2. The Basic Model,
Unknown
Reference
Library
R
R
R
R
R
R
LJZ-—^.
<=J
—
S f
s
s
s
s
s
, Hit-list of
best natches
i
? .-.
3, .„
4
Fig. 1, Library Search. The spectrum of the unknown
X is compared to all spectra of the reference com-
pounds R in the reference library, and the degree of
similarity S is evaluated. The reference compounds
most similar to the unknown are then put out.
The most important thing to realize when working with library search systems is that when we compare
spectra we really mean chemical structures. We tacitly assume that similar spectra imply similar chemical
structures. If in a library search system this hypothesis is not adequately met, the respective system will
be useless for practical applications. This is a required but not sufficient condition for a library search
system to be useful. Thus, in order to understand library search, one has to understand the comparison
process. The central hypothesis of library search is "If the spectra are similar then the chemical structures
are similar." This statement is often referred to as the library search hypothesis. Whether it is true or not
obviously depends on what is meant by "similar". We consider two objects to be similar if they are identi-
cal in many aspects believed to be relevant in the given context. Similarity thus requires that the objects
to be compared are characterized by many different features which can be either identical or different. A
simple measure of the similarity between two objects is the number of corresponding features in which
they are identical. A chemical compound and its spectra are thus characterized by a set of structural
descriptors and by a set of spectral descriptors. One may now set up two independent multidimensional
spaces for structures and spectra respectively, where one descriptor is assigned to each coordinate axis
from the respective set. Every conceivable chemical compound characterized by its spectra can now be
mapped into a point in each of these multidimensional spaces, using the values of the respective de-
scriptors as coordinate values. The two spaces are referred to as the structure space and the spectra
space respectively. A very
(/p^v simple example is given in
fig. 2.
Comparing the spectra of
two compounds amounts
to measuring the distance
between their respective
points in the spectra space.
The more close they are,
the more similar are the
two spectra. If the two
points coincide, the spec-
tra are considered as iden-
tical. The library search hy-
pothesis now requires that
[R barsd
]7DC Cn-l
spectra space
structure space
Fig. 2. Spectra space and structure space for 4 simple compounds.
Spectra and structures are characterized by selected features. The fea-
ture states are then used as coordinate values in the respective coordi-
nate system.
512
-------
points close together in the spectra space are al-
so close together in the structure space, i.e. the
two spaces should have the same (mathemati-
cal) structure (see fig. 3).
The spectral features defining the spectra space
and the distance measure to be used in this spa-
ce are an integral part of any library search sy-
stem. The goal in selecting the descriptors and
in defining the distance measure is to model in
the spectra space the (mathematical) structure
of the structure space as closely as possible. Si-
milarity in the structure space, however, is not
defined and depends on the problem to be sol-
ved and on the user and his preferences. Thus,
no library search system can work well for all
applications.
3. Feature Selection.
Sppc'tra Spec?
oo
Structure Space
Structure Space
Space
oo
0°
A
A
o
0
o
0
O A
o
a
o
o
Fig. 3. Relations between spectra space and struc-
ture space. Similarity between spectra and between
structures is related to distance between the re-
spective ]X)ints. If both spaces have globally the sa-
me (mathematical) structure, distances (and thus
similarities) may freely be transferred between the
two spaces (left). If the relation holds only locally,
only higi similarities approaching identity are
meaningful (centre). If the two spaces have diffe-
rent structures, the respective system is not useful.
The spectral features selected to describe the
spectra and to span the spectra space determine
how well the spectra space will conform to the
structure space. One wants the spectral features
to be sensitive to differences in the chemical
structure, but insensitive to technical and
instrumental conditions.
In some applications one may assume that the
unknown samples are represented in the refer-
ence library. This is often the case in environmental analysis and in government laboratories, where the
question is whether the sample is a compound specified in a (limited) list. Here, the degree of structural
sensitivity of the features selected is of no great concern. The system is expected to identify all com-
pounds in the list. If a sample is another compound, the answer "not identical to a reference compound"
is interpreted as 'not on the list" and is accepted as :;uch. Applications of this type are referred to as
identity search.
However, if the question is "What is it?", then feature selection becomes much more difficult. If the sam-
ple is not identical to one of the reference compounds, the user expects to be presented with a set of
model compounds similar to the unknown and uses the similarity measure (distance in the spectra space)
as an estimate of structural similarity (distance in the structure space). To get reasonable correlation
between the two similarity measures the spectral features h ive to be selected with care. This type of
application is referred to as similarity search.
4, Size and Contents of the Library.
All library search systems have one fundamental limitation. T ic set of answers they can provide the user
with is limited by the contents of the reference library. If, for a given unknown, no suitable reference
compound is part of the library, no useful answer can ever re< ult, independent of all other factors. Thus,
size and contents of the reference library are of fundamental importance for the performance of the sy-
stem.
For an identity search system it is obvious that the library has to include all compounds to be identified.
The larger the library, the more compounds we can sucxxssfully deal with. However, the other aspects of
the performance are not increased by additional compounds. On the contrary, search time and thus costs
will increase with the size of the library.
For similarity search systems, suitable reference compounds for all conceivable unknowns are required.
This seems to call for a comprehensive library containing the spectra of all known chemical compounds
or at least as many as one can get hold off. However, a moments reflection shows that this is not the
optimal solution. For each compound type one needs just one or maybe a few references. Large sets of
closely related reference compounds will, in a true similarity search system, not increase the performan-
ce. On the contrary, retrieving an excessive number of closely similar references for a given sample will
just increase the output volume without providing additional information. It is only the first reference
compound in a group of related compounds which matters. The other references, being very similar to
the first one, just repeat the same message over and over again.
What one really needs can be clearly stated within the space model (see fig. 4): The spectra space has to
be filled sparsely but evenly with reference compounds, to provide some close neighbors in every section
513
-------
Spectra Space
Spec"tra Space
Spectra Space
o 0uc
0
ok
o ojj DO'-'O oo^;
too dense
°0
=°Coi
Fig. 4. Point density and distribution in spectra space. The points
corresponding to the reference compounds most similar to the
unknown lie within a circle centered on the unknown, its radius
being given by the smallest useful similarity. If the density of ref-
erence points is too high, the number of close matches becomes
unmanageable (centre). An uneven distribution gives a biased re-
sult.
of the space. The correspondence
between spectra space and struc-
ture space insures that in this case
the structure space will also be
adequately populated. The popu-
lation density should reflect the
relevance of the respective space
sector to the user. In his fields of
interest the average number of
references may be higher than
elsewhere, in order to give a hig-
her resolution. Thus, an optimal
library will consist of two parts. A
general part, which provides a few
prototype compounds for every
compound class, and one or more
special parts, documenting the
user's field(s) of interest with en-
hanced resolution.
To ensure wide coverage of all fields and branches of chemistry, the general part of a reference library
will generally have to be bought from an outside source. The special parts, however, have to be assem-
bled by the user himself. All the spectra recorded in his laboratory are important in this context, even if
the respective compounds have no other relevance. The fact that the respective sample has been submit-
ted for analysis is a prove that it is relevant to the analyst. Thus, all spectra recorded in the user's lab are
candidates for inclusion in the reference library. However, to limit the size of the reference library only
the spectra of compounds from classes not yet adequately represented in the library should be added.
5. Evaluation of the Performance,
The performance of a library search system includes many different aspects. First of alL, we expect a
powerful library search system to be able to retrieve from the reference library compounds identical to
the samples, if such compounds exist in the library. If no compound identical to a given sample exists in
the library, the system should provide the user with reference compounds structurally similar to the un-
known. It should be able to do so regardless of the technical parameters used when recording the respec-
tive spectra. To each reference retrieved a reasonable measure of similarity to the unknown should be
put out. These similarity values have to inform the user whether a given reference compound can be
assumed to be structurally identical to the sample with high probability or whether it is similar only. In
the latter case the similarity measure should give a reasonable estimate of the structural similarity. Fur-
thermore, we expect the library search system to be fast, to be easy to use and to present the results in
easily interpretable form. Finally, there have to be program modules which allow for easy maintenance of
the system library. In particular it should be easy to add, delete and edit entries. Evaluating a library
search system amounts to assigning values (measured at least on an ordinal scale) to all these aspects of
performance. In the present lecture no attempt is made to discuss the aspects which depend to a large
extent of the user's personal taste, i.e the presentation of the results and the quality of the user interface.
The only topic to be discussed is how to estimate the quality of the results from a chemical point of view.
6. General considerations.
The most important qualifier for a library search system is how well it maps spectral similarity into struc-
tural similarity. The fundamental problem here is that presently no generally accepted similarity con-
cepts for chemical structures exist. The chemist's notion of structural similarity is strongly coined by tra-
ditional views (functional groups, skeletons, compound classes) and by the problem he currently studies.
Furthermore, different spectroscopic techniques focus on different structural entities and are thus inher-
ently biased towards certain aspects of structural similarity. Thus, no generally applicable procedure can
be given.
A second problem arises from the fact that the composition of the reference library is of paramount
importance. The set of possible answers is limited to the chemical structures represented in the library. A
system to be testes can produce a useful answer only if its library contains at least one suitable reference
compound for the test sample. Thus an unlucky choice of test samples can shift an otherwise excellent
system the bottom of the list.
514
-------
A A
X X Y AB AXXY
7. Test procedure.
Both these problems can be solved by using a standardized test library to be used by all systems to be
compared. The user should carefully select a set of about ten to twenty test compounds and record their
spectra under strictly routine conditions. It is important that user selects these compounds to be truly
representative for the compound classes he predominantly deals with in his practical work. This ensures
that the test is performed in the regions of the spectra anil structure space most important to the user.
The test compounds are grouped in subgroups (preferably triplets) containing at least one pair of (struc-
turally) highly similar compounds and one pair with moderite structural similarity, the degree of structu-
ral similarity being subjectively judged by the user in the coatext of his daily routine work. With this choi-
ce the user implicitly defines his concepts of structural similarity. Between the subgroups, however, the
structural similarity should be low. One or a few compounds should be duplicated, i.e. their spectra
should be recorded twice under different conditions (different matrix and/or different instrument set-
tings).
The test spectra are then assembled into a mini reference library to be used by all systems to be tested.
The same spectra will also be used as test samples. Each sample compound is submitted to the system as
an unknown, and for all references the similarity computed by the system is noted. The data are best
organized into a similarity matrix Each unknown is assigned to a row, the columns are assigned to the
references, and the members of subgroups of mutually similar compounds are kept together.
This procedure makes sure that for each test sample there is at least one structurally identical compound
in the library. Furthermore, for each sample there are one or more references which the user considers
to be structurally similar. The duplicates will give itiformation on how well the systems under test can
handle variations due to technical artifacts.
8. Data Interpretation.
From the similarity matrix a wealth of information re-
garding the behavior of the system under test may be
obtained (cf. fig. 4). The diagonal elements correspond to
comparing two identical spectra. Their values will corre-
spond to the highest possible similarity, namely to perfect
identity. If this is not the case, the mathematics of the sy-
stem are wrong, the system behaves irrationally and
further tests are unnecessary.
The rows and the columns for the duplicated samples
show how well the system can deal with instrumental and
technical variations. In the ideal case, the respective rows
and columns should be identical (However, a row is not
necessarily identical to its corresponding row, the simila-
rity matrix not necessarily being symmetrical). Any diffe-
rences between corresponding values are due to the fact
that the system misinterprets technical variations as being
caused by structural differences. In a real system this is
unavoidable. The rrns difference between corresponding
values is a measure for the amount of noise generated by
different registration parameters and/or different matri-
ces. Differences in similarity are meaningful only if they
are significantly greater than this value.
Off-diagonal elements for duplicates should exhibit si-
milarity values not significantly smaller than the diagonal
elements. Failure to do so indicates that the system, being
too sensitive for instrument settings and matrix effects, is
unable to reliably recognize the identity of compounds.
In the submatrices corresponding to subgroups of mu-
tually similar compounds all off-diagonal elements should
be significantly smaller than the values for duplicate com-
pounds, to indicate to the user that sample and reference
are not identical. If this condition is not met, then the sy-
stem has troubles in discriminating between true identity
and high similarity. The values should, however, be signifi-
cantly greater than the values for off-diagonal elements
not being in a submatrix for a subgroup, lo indicate thai
the respective compounds are definitely more similar as
CaXbXc)
(bXaXc)
CcXcXa)
CgXhXp^
X KdXdXdXbXaXcXgXhXe)
Y (cJ)(dXd)[(c)Cc)Ca)!
-------
two compounds picked at random. There are two ways to violate this requirement. There can be some
elements in the remainder of the matrix which are exceptionally high. As long as there are only a few
such entries, this is of no great concern. It is just an indication that the system's inherent similarity con-
cept for structures includes some aspects neither obvious nor relevant to the user. The other case, some
elements in the submatrix being too small however, has to be taken seriously. It indicates that the sy-
stem's similarity measure does not consider certain aspects definitely important to the user. Finally, the
sequence of similarity values should reflect the different degrees of structural similarity expected by the
user. However, only differences clearly above the noise level are meaningful for ranking.
The information extracted from the similarity matrix as stated above will give a reasonably clear indica-
tions as to whether the system under test uses basically the same similarity concepts as the user and whet-
her its sensitivity to structural variations and its insensitivity to instrumental and technical parameters is
low enough to give sufficient discrimination between the cases of identity, of high similarity and of no
similarity.
One important point has to be kept in mind, however. The above qualifiers are to a large extent subjecti-
ve. They measure the degree of correspondence between the user's and the system's concepts of structu-
ral similarity. The system's similarity concepts are defined implicitly by the designer, mainly by the spec-
tral features selected for comparing spectra. They define the overall mathematical structure of the spec-
tra space and its mapping into the structure space. By selecting groups of (in his opinion) structurally
similar test compounds the user defines (again implicitly) the local mathematical structure of the spectra
space. Thus, a given system not exhibiting top performance only proves that the system under test does
not use structural similarity concepts similar to the user's. This may be due to a poor design of the system
or due to the user's expectations being too specialized or irrational. Failure to perform as expected thus
disqualifies a system only within the context specified by the test compounds. For other applications the
respective system may well be the perfect choice.
Analysis of the similarity matrix can also supply information as to the search strategy implemented in the
system. Under the assumption that the test compounds have been well selected, the following procedure
reveals the respective information. First, all entries relating to identical spectra (diagonal elements only,
not duplicates) are removed from the matrix. The remaining entries are ranked in decreasing order of
similarity. Then one prepares a plot of similarity versus rank. This will give a curve which starts at rank 1
with a very high value for similarity and which subsequently drops to the lower similarity values for the
later ranking pairs. The general shape of this curve is related to the search strategy employed by the
system.
A similarity search system (as opposed to an identity search system) is expected to be able to retrieve
reference compounds structurally similar to an unknown sample. Furthermore, the similarity measure
should be able to discriminate between different degrees of similarity. Thus, the data set containing user
selected pairs of high but variable similarity, the first part of the curve should be a smoothly declining
curve, preferably almost linear. The length of the section to be considered is given by the number of
off-diagonal elements in submatrices corresponding to the groups of structurally similar compounds. The
shape of the remaining curve is of no concern (Nobody cares how useless a useless reference is). In an
identity search system one expects positive identification of identical pairs, but places no particular emp-
hasis on the similarity of non-identical pairs. The rank-vs-similarity curve thus starts with a short almost
horizontal section, whose length is given by the number of duplicate pairs in the test set. The curve
should then drop sharply and level off slowly. Of course, there is a gradual transition between the two
pure strategies, resulting in curve shapes somewhere between these extremes. However, a rough estimate
of the search strategy is generally possible.
An important part of the search strategy cannot be evaluated from the data in the similarity matrix. All
similarity measures employed in library search systems are in some way based on the number of elemen-
tary spectral attributes (spectral features) having identical attribute states (the same value) in the two
spectra to be compared. This number is set in relation to the total number of attributes considered. If all
attributes do have the same state, the system considers the two spectra to be identical. Thus, the degree
of similarity is measured as the number of attribute states identical between the two spectra compared,
relative to the number of attribute states identical between one of the two spectra compared to itself.
Whether the spectrum of the unknown or the spectrum of the reference is used as the base can lead to
large differences in the system's behavior. This is most easily explained using a strongly simplified mathe-
matical model.
Let the set of features present in the unknown sample X be GX and the corresponding set in the refer-
ence GR and let the intersection of GX with GR be the set GXR, the set of the features the two com-
pounds X and R have in common. The number of members in the three sets GX, GR, and GXR shall be
designated as TX, TR, and TXR respectively. A very crude measure for the similarity between X and R is
then given by TXR. To normalize this result it is divided either by TX or alternatively by TR. In the first
case, where the unknown sets the standard, the similarity becomes independent from the number of fea-
516
-------
tures present in the reference only. Thus, the system does not penalize excess features in the reference. It
just tries to make sure that the highest possible number of features present in X are also present in R.
There is thus a tendency to retrieve references which are "too big". Each reference tries to represent all
features of the unknown as completely as possible, the unknown tends to be a subset of the reference.
This approach is referred to as a "forward search". It is particularly appropriate for similarity search of
unknowns which can be assumed to be reasonably pure.
If the reference sets the standard, excess features in the unknown get along unpenalized. The system
prefers references which are "too small". Each reference tries to be completely represented in the un-
known, it tends to be a subset of the unknown. This strategy is referred to as "reverse search". It has its
main application with samples suspected or known to be mixtures, because it gives the user a chance to
identify components of the mixture.
Of course, forward and reverse search produce different results only if the spectra of the unknown and of
the references are treated differently. If this is the case, a non-symmetrical similarity matrix results. The
mathematical model given above is very much simplified, A more detailed analysis shows, that the resul-
ting search strategy further depends on whether the case of .1 given feature being present in both spectra
is valued differently from the case where the same feature is absent in both. If this is the case a non-sym-
metrical similarity matrix results even if the two spectra are not treated differently. At present, however,
there is no method known to determine from the similarity matrix which search strategy is used by a
given system. The only thing one can say is that systems producing a symmetrical similarity matrix either
treat unknown and reference alike or use a balanced search strategy exactly halfway between the two
extremes.
9. Outlook.
The philosophy behind any library search system for the interpretation of molecular spectra is based on
the hypothesis, that similar spectra imply similar chemical structures. A library search system is useful in
the real world of the analytical chemist only if it makes this hypothesis to become true. In order to build
better library search systems, one needs similarity measures for both spectra and structures. The de-
signer of a library search system decides upon the similarity tor spectra. The similarity measure for struc-
tures, however, is defined by the user and depends on the problems he has to solve. In some cases he may
be predominantly interested in references having the same functional group?, as his unknowns and does
not care a lot about the skeleton. In other applications, however, he may place main emphasis on the
skeleton. Consequently there will never be one single and universally applicable similarity measure for
chemical structures. Presently there is none at all. The designer of a library search system is therefore
faced with the impossible job of optimally mapping a similarity measure for spectra onto an unknown
and undefined similarity measure for structures. Currently he has no other choice than to rely on guess-
work. The development of even a poorly performing similarity measure for chemical structures would
allow for the application of formalized mathematical optimization methods and would almost immedi-
ately lead to better library search systems. Thus, future rese;irch in this field should concentrate on che-
mical structures rather than on spectra.
10. References.
J. T. Clerc, E. Pretsch, M. Zurcher: Performance Analysis of Infrared Library Search Systems. Mikro-
chim. Acta 1986II, 217-242 (1986).
J. T. Clerc: Automated Spectra Interpretation and Library Search Systems in: Computer-Enhanced An-
alytical Spectroscopy. Henk L. C. Meuzelaar and Thomas L. Isenhour, eds. Plenum Press, New York
1987 p 145-162.
J. T. Clerc: Computer Aided Interpretation of Spectra in: Research Instrumentation for the 21st Cen-
tury. Gary R. Beecher, ed. Martinus Nijhoff Publishers, Dordrecht 1988 p 403-418.
M. Zurcher, J. T. Clerc, M, Farkas and E. Pretsch: Genera Theory of Similarity Measures for Library
Search Systems. Analyt. Chim. Acata 206 161-172 (19?S)
517
-------
The Use of Principal Component Analysis to Display PAH
Concentration Patterns in Indoor Air
Somenath Mitra and Nancy K. Wilson
Atmospheric Research and Exposure Assesment Laboratory
Environmental Protection Agency, Research Triangle Park, NC 2711.
Abstract
Principal component analysis was used to display a data set
consisting of the concentrations of fifteen polynuclear aromatic
hydrocarbons (PAH) in indoor air. The data represented ten homes
having different emission sources such as gas utilities, wood-
burning fireplaces and cigarette smokers. The principal component
analysis enabled convenient displays of the multidimensional data
set from which characteristics of the PAH concentrations in the
indoor air could be understood. The PAH concentration patterns of
the indoor and the outdoor air were quite different even when air
exchange rates were fairly high. Among the homes, different PAH
concentration patterns emerged depending upon the emission
sources present. Of the sources, cigarette smoking appeared to
affect indoor air most adversely. The variable loading plots
along with the correlation matrix were used to identify the
interrelationships between the concentrations of PAH in indoor
air.
Introduction
Most environmental monitoring studies involve measuring
several pollutants simultaneously. Since the concentration of
each pollutant is a variable (or a dimension), the resulting
multidimensional data may be difficult to interpret. Multivariate
statistical analysis can be used to interpret multidimensional
data sets. Several such applications to a variety of
environmental problems have been reported in the past few years
(1-3). The objective of this study was to interpret a data set
of PAH concentrations in indoor air using principal component
analysis, which is an effective way of projecting
multidimensional data onto two or three dimensions while
preserving most of the variance in the data set (4).
Description of Data: The data were obtained from a study of
indoor and outdoor air at homes in Columbus, OH (5).
Stratification variables in the study were the following
combustion sources: cigarette smoking, use of woodburning
fireplaces, and heating and cooking by natural gas or
electricity. Details of the sampling and analysis procedures have
been published elsewhere (5). The fifteen PAH measured are listed
in Table 1.
518
-------
Data Analysis
The data matrix was formed by taking each PAH as a variable.
Each row represented a PAH and each column a room (kitchen,
living room or bedroom) in one of the: homes. Since we were doing
exploratory data analysis, none of the variables were rejected.
To eliminate the unduly large effect of those variables that had
large magnitudes, each variable was standardized so that it had a
;mean of zero and unit variance (4) . The PCs were rotated using
varimax rotation (4), to maximize the number of loadings that are
either high or near zero and minimizes the loadings that have
intermediate values (1). The univariate Pearson correlation
coefficients were also calculated. Fcr the data analysis, the
statistical package Statgraphics (6) was used.
Results and Discussion
Effects of combustion sources
The results of the PC analysis for all the homes is presented
in Figures 1A and IB. The first two PCs accounted for 73.7% of
the variance, of which PCI contributed 61.8%. Figure 1A shows the
variable loading plot, and Figure IB shows the corresponding
object score plot. Compounds 1,2 and 7-15 show high loadings
along PCI. Of these, PAH 7-15 are collected primarily from the
particulate phase. Some of the more volatile PAH 3,5 and 6, which
are more abundant in the vapor phase, show low loadings on PCI
but high loadings on PC2,
The Pearson correlation coefficents also indicated strong
correlations between the concentrations of PAH 3,5 and 6,
evidenced by correlation coefficients close to 0.9O. Fair
correlations existed between PAH 7-15, those PAH that correlated
strongly along PCI. Benzo[a]pyrene correlated strongly with PAH
9-14 (correlation coefficient close to 0.80). Concentrations of
benzo[g,h,i]perylene (PAH 14) and coronene (PAH 15) are well-
correlated.
In the object score plot Figure 13, we see clustering for
different types of homes. The all-electric homes (type E) are
characterized by low values on both axes, reflecting low
concentrations of all PAH. Clustering of homes with gas utilities
and woodburning fireplaces (type C), homes with gas utilities
only (type A) and the smokers' homes (type B and D) is also
evident. Despite the facts that the homes in this study were in
different neighborhoods, were of different ages and floor plans,
and that the sampling was done on different days, the class
separations for the homes with different stratification variables
are fairly good.
As seen in Figure IB, both gas utilities and fireplaces
affect the PAH concentration in indoor air. Homes A and C show
increased scores on both PC axes compcired to the all-electric
homes E. This implies a net increase of both types of PAH in
these homes. In Figure IB, the gas-heated homes that have
fireplaces (C) show relatively lower loadings on PCI than the
ones that do not have fireplaces (A). This lower PAH
concentration may be due to higher intrusion of relatively
519
-------
cleaner outdoor air when the fireplaces were in operation. An
increase in the concentration of pollutants correlated with PC2
is reasonable, because these more volatile PAH are generated in
greater quantities than are the larger PAH during combustion of
wood (7).
In Figure IB, the most dramatic effect is seen for homes with
smokers (type B and D). Smoking resulted in higher values for
both PCI and PC2, that is, it generated all PAH compounds and its
effects were more significant than those of the other
stratification variables. Other studies have also shown that
cigarette smoke produces a wide range of PAH (8).
The univariate correlation coefficents between the PAH in
nonsmokers' homes and the smokers' homes showed interesting
trends. The three-ring PAH 3, 5 and 6 were well correlated in
both cases. Notable were the correlations among PAH 7-15. In
smokers' homes, there was fairly good correlation between PAH 7-
15. PAH 12 had a correlation coefficent greater than 0.80 with
all PAH 7-13, and PAH 14 and 15 were strongly correlated. For the
nonsmokers' homes, the correlation between PAH 7-15 was much
smaller. PAH 12 has high correlation with PAH 13 and 14 only.
This absence of correlation among PAH 7-15 in nonsmokers1 homes
implied that they were produced from a variety of sources,
whereas in smokers' homes they came mainly from the cigarette
smoke. This indicated that cigarette smoke can be a major
contributor of PAH 7-15 in indoor air. Similar conclusions were
also drawn from the PC analysis of each subset.
Indoor/outdoor comparison: In the PC analysis of the outdoor and
the indoor PAH concentrations, the outdoor samples clustered
along PCI and indicated that the PAH concentration pattern
outdoors differed from that indoors. The indoor air showed much
higher values on PC2, indicating higher concentrations of the
smaller PAH (PAH 1, 3, 5r 6).
Migration of pollutants: When different rooms (kitchen, living
rooms and bedrooms) were projected on the object score plot, the
different rooms within a house had similar characteristics even
though a particular activity may not have taken place there. For
example, even if no cigarettes were smoked in the bedrooms in
smokers' homes, the bedroom units still clustered with rooms of
other smokers' homes. This implies that there is ample migration
of pollutants within a house. However, the kitchen, living room
and bedroom of a given home did not necessarily cluster together.
Regardless of varying air exchange rates, each type of home
maintained its PAH concentration pattern.
Acknowledgement: This work was done while the first author held a
National Research Council Research Associateship with the U. S.
Environmental Protection Agency.
520
-------
Disclaimer: Although the research described in this article has
been funded wholly or in part by the United States Environmental
Protection Agency through Cooperative Agreement CR 8154 18 with
the National Research Council, it has not been subjected to
Agency review and therefore does not necessarily reflect the
views of the agency, and no official (indorsement should be
inferred. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
References
1. Hopke P. K., Gladney E. S., Gordon G. E., Zoller W. H. and Jone
A. G. (1976) The use of multivariate analysis to identify sourc
of selective elements in the Boston urban aerosol. Atmos.
Environ. 10, 1015-1025.
2. Roscoe B. A., Hopke P. K., Dattner S. L. and Jenks J. M. (1982)
The use of principal component analysis to interpret particulat
composition data sets. J^ Air Pol_l_., Control Assoc. 32, 637.
3. Vogt N. B., Brakstad F., Thrane K., Nordenson S., Krane J., Aam
E., Kolset K., Esbenson K. and Steiner E. (1987) Polycyclic
aromatic hydrocarbons in soil and e.ir: statistical analysis and
classification by the SIMCA method. Envir. Sci. Technol. 21,
35-44.
4. Joliffe I. T., Principal component analysis (1986) Springer
Verlag, New York.
5. Chuang J. C., Mack G. A., Koetz J. R. and Peterson B. A. (1986)
Pilot study of sampling and analysis for polynuclear aromatic
compounds in indoor air. Report, U.S. EPA, No. EPA/600/4-86/036,
6,, STATGRAPHICS (1986), STSC INC. Rockville, MD (USA).
7. EPA (1989), Effect of burn rate, wood species moisture content
and weight of wood loaded on wood stove emissions. Report, U.S.
EPA, EPA/600/2-89/025.
8. Hoffman D., Wynder E. L. (1986) Chemical constituents and
bioactivity of tobacco smoke. Proceedings of international
meeting organized by International Agency of Research on Cancer.
IARC scientific publication No. 74. Lyon, France.
521
-------
06
02 -
02 -
-0.8 ~
— 1 1 1 1
3
-
-
_
_
-
2
"I , ,
i — i — i — i — i — i — i — i — " i ' i c
B •
4 _
t
11 —
10
12
1
7
13 —
14
-
1 I 1 , 1 1 1 1 1 ._! J L
2.1 ~
06
0.1 ~
-1.9 -
1.Z
-03 0.3 0.9
Principal Component 1
| I I 1 1 | 1 1 1 1 | 1 1 T 1 | ! 1 I I | I
0
I . DO °
' C D°D 0
D
c*
_ A •
f E * A
A
1 l r i i 1 i i i i 1 i i i i 1 i i , i 1 i
1 1 1 |_
-
_
_
~
-
-
i • i 1
PXH
1 Quinolin*
2 I«oquinolin«
3 Ph«n*nthr«n«
4 Anthrtc«n*
5 Fluortnthcn*
6 Pyr»n«
7 Cyclop«nt.«(c,d)pyr«ne
8 »«nLO(»)«nthr»c«n«
9 Chry»«n«
10 Itonzof luorant-haneB
11 B«n»o(«)pyr«n«
13 Ind«no( 1,2, 3-c,d)pyr«n«
14 B«nxo(
-------
Statistical Modelling of Ambient Air Toxics Impacts During Remedial
Investigations at a Landfill Site
Steven C, Mauch
Roy F. Weston, Inc.
West Chester, PA
Louis M. Militana
Roy F. Weston, Inc.
West: Chester, PA
ABSTRACT
At landfills or other waste disposal sites, the off-site impacts due to
air toxics generated by intrusive activities are a principal concern.
To assess these impacts, the multivariate statistical technique of
canonical correlation has been applied, to ambient air toxics sampling
data collected during a remedial investigation (RI) of a landfill in the
metropolitan area of Los Angeles, CA. Tie goal of the analysis is to
determine whether a site activity produces significant ambient air
toxics impacts in the area immediately downwind of the site.
Canonical correlation analysis of the delta collected at the downwind
site reveals that the primary physical process occurring is dilution of
contaminants by wind, with secondary sli.ght increases in contaminant
levels primarily due to boring activities. Although the canonical
models are not strong enough for quantitative predictions for this data
set, they do provide a realistic qualitative analysis of the physical
situation.
INIRCDUCTICN
This paper presents the results obtained from application of canonical
correlation analysis to ambient air toxics sampling data collected
downwind of a landfill site during RI activities. Canonical correlation
is a multivariate statistical technique that may be used to evaluate the
relationship between groups of variables, in this case, meteorological
conditions/ site activities, and ambient air toxics levels. The
technique is an extension of traditional multiple regression analysis,
which seeks to relate a single variable to a group of other variables.
Canonical correlation was chosen as an analytical tool because of its
ability to provide information beyonc. the scope of traditional
statistical comparison techniques, such a<; simple tests for equality of
means or multiple correlation. The use of multivariate methods allows
better resolution of the complex interactions between the atmosphere and
the variety of air toxics compounds that:, may be present due to intrusive
activities on a landfill site.
SITE DESCRIPTICN / DATA COLLECTION
The site was an urban landfill located in the Los Angeles area. The
area is located to the northeast of the intersection of two major
thoroughfares, with the upwind sampling site located near the
523
-------
intersection. The downwind site is located beyond the northeast corner
of the landfill area.
Ambient air samples were collected at the upwind and downwind locations
during the 3-week site investigation. Wind speed and direction data
were collected concurrently with the sampling periods. Due to the
consistent land-sea breeze circulation pattern at the site, daytime
winds were most frequently from the west -southwest . The upwind and
downwind sampling locations were chosen based on this wind pattern.
During the activity period, a total of 31 high-volume air samples and 33
volatile organics samples were collected. The compounds detected
included eight toxic volatile organic compounds (VOCs) , copper, lead,
zinc, and asbestos. The following eight VOCs were detected in at least
75% of the samples; acetone, benzene, ethylbenzene, styrene, toluene,
xylenes, tetrachloroethene, and 1,1,1-trichloroethane.
The VOCs were collected using passivated stainless steel canisters (EPA
Msthod TO-14), and the metals were analyzed from high-volume air samples
of particulate matter. Asbestos was determined using low-volume
personal pumps and filter cartridges.
Site activity was parameterized as the durations of the two principal
intrusive activities: boring {soil core samples) and drilling
(groundwater monitoring wells) . Activity durations were obtained from
the site log books.
UPWINVDCIWNWIM)
To assess the amount of contamination introduced into the ambient air by
site activities, a comparison of upwind and downwind means may be used.
Since normality is not a reasonable assumption for the air toxics data
being considered, the t-test for equality of means was not used in the
comparison. Instead, a nonparametric comparison of medians was
performed using the Wilcoxon two-sample test for independent samples.
None of the upwind/downwind pairs of medians were significantly
different at the 10-percent level.
The upwind/downwind sample sets were also compared graphically using
side-by-side box-and-whiskers plots, generated from descriptive
statistics for the activity period. Figure 1 shows the data for several
typical contaminants. There is a high degree of variability in the
samples, and a characteristic skewing towards lower values.
These plots effectively portray the difficulty in determining whether
the downwind contaminant levels are in any way distinctly greater than
the upwind levels. Comparing means or medians would lead to the
conclusion that site activities had no distinguishable impact on
contaminant levels.
CANONICAL CORRELATION
Canonical correlation extends the sample correlation concept from two
single variables to two sets of variables. The two sets are analogous
to the dependent and independent variables in traditional regression
524
-------
analysis. The canonical correlation procedure finds the most highly
correlated pairs of linear combinations of the variables in each set.
These linear combinations are known as -canonical variable scores, and
the sample correlation between a pair of scores is the canonical
correlation coefficient. The scores may Sue interpreted by examining the
conponent variables' sample correlations with the resultant score.
The canonical correlation procedure was performed based on the
correlation matrix for the contaminant variables combined with the wind
vector components, and the activity durations. Solid and volatile
contaminants were treated separately. The northerly and easterly wind
speed components were mean values covering the period from 0700-1700 L
each day.
Solid Contaminants
A summary of the results of the canonical correlation analysis for the
solid contaminants at the downwind site i.s shown on Table 1. The first
two pairs of canonical variates are significant at the 10-percent level.
This and other coefficients significant at the 10-percent level are
underlined. Table 2 shows the correlations of these pairs of scores
with their conponent variables.
The general relationship expressed by the correlations of the first pair
of scores is lower contaminant levels and more southerly winds. Based
on site geography, the southerly (i.e., positive northerly) component
would contribute to transport away from the downwind site (dilution) .
Therefore, the first set of canonical variates appears to represent the
general reduction of contaminant levels at the downwind site by
dilution.
The second pair of variates reflects higher contaminant levels and
longer boring periods, based on the correlations. There is also a
relatively high correlation in the ac±ivity/wind score with winds
having relatively lower easterly conponents. The mean wind vector
direction from the 13 activity days is vest-southwesterly, with a mean
easterly component roughly 1.5 times the mean northerly component.
Winds having lower easterly components would be more from the southwest.
Southwest winds produce the greatest over-site fetch at the downind
sampling location. Therefore, the second pair of variates may be
interpreted as reflecting a general elevation of contaminant levels at
the downwind location during site activity with more across-site winds.
The logical extension of this analysis would be to attempt to predict
the Quantitative effects of varying levels of site activity on
contaminant levels. Constructing such a model would require
establishing a solid relationship between the variables and the scores.
Unfortunately, the correlations are too weak to be of predictive value.
However, the canonical correlation analysis does indicate that elevated
contaminant levels are at least qualitatively associated with increased
boring activity.
525
-------
Volatile Organic Contaminants
The canonical correlation analysis of the VOCs indicated one
significant pair of variates. However, no contaminants were
significantly correlated with the contaminant score.
This type of ambiguity occurs in canonical correlation analyses
whenever there are strong correlations between many variables in either
group. Such a high degree of correlation does exist amongst many pairs
of VOCs, principally due to the influence of nearby traffic emissions.
The use of highly correlated predictor variables in linear regression
produces an analogous effect. More definitive results may be possible
if some of the highly correlated variables were eliminated.
OCNOJUSICNS
Ambient air toxics data collected upwind and downwind of a landfill
during intrusive activities were evaluated using basic satitistical
comparisons. Both side-by-side box plots and Wilcoxon two-sample tests
for equal medians showed no significant differences in contaminant
levels at the two sampling sites.
Canonical correlation analysis of the solid contaminant levels and the
activity/wind variables at the downwind site shows that: 1) there is
primarily dispersion of contaminants across the normal sea-breeze wind
direction (southwest), and 2) boring duration and metals levels are
positively related. These canonical relationships are not strong enough
for quantitative use.
Canonical correlation analysis of the VOC data at the downwind site are
rendered indeterminant due to a high degree of inter-correlation among
the volatile contaminants. These interrelationships are mainly due to
the source signature of traffic on the thoroughfares bordering the site,
which likely obscures any relationships between VOC levels and site
activity.
The use of canonical correlation to analyze data from air sampling
efforts provided useful results in this case, beyond the information
revealed by simple statistical comparisons. The results obtained were
consistent with the expected physical process of dispersion. Although
no quantitative judgments could be made from this data set, the method
has the potential to provide quantitative results as well.
526
-------
700
600
500
S
E
0 400
3
I
E
| 300
a
b
E
100
0
i
1
'
]
1
UP
1
I
1
A
T
|
.
I
1
•
•
.
i
350
300
250
3
-------
TABLE 1. Canonical Correlation Results: Downwind, Solid Contaminants
COMPONENT
NUMBER
1
2
3
4
CANONICAL
CORRELATION
0.9715
0,9027
0.7421
0.4159
SIGNIFICANCE
LEVEL
0.0069
0.0951
0.3274
0.5145
TABLE 2. Correlations of Scores with Original Variables
CONTAMINANT SCORE
VARIABLE
Asbestos
TSP
Copper
Lead
Zinc
COMPONENT
1
-0.427
-0.170
-0.51 1
-0.199
0.122
COMPONENT
2
-0.048
0.228
-0.190
0.669
0.467
WIND/ACTIVITY SCORE
VARIABLE
Northerly
Easterly
Boring
Drilling
COMPONENT
1
Q.7Q4
-0.061
-0.209
0.383
COMPONENT
2
0.248
-0.427
Q.939
-0.150
528
-------
COMPARISON OF THE SOURCE LOCATIONS
AND THEIR SEASONAL PATTERNS FOR
SULFUR SPECIES IN PRECIPITATION
AND AMBIENT PARTICLES IN ONTARIO, CANADA
Yousheng Zeng and Philip K. Hopke
Department of Chemistry
Clarkson University
Potsdam, NY 13699-5810
The Potential Source Contribution Function (PSCF) is a probability function based on
the air parcel trajectory data coupled with information regarding the nature of the contami-
nants measured in that air parcel. PSCF is the ratio of the probability of a contaminated air
parcel having traversed a 1° latitude by 1° longitude area to the probability that any air parcel
traversed that area. Regions with high PSCF values thus have a higher probability of
contributing pollutants to the measured concentrations at the receptor site. The PSCF
analysis has been applied to sulfur species in both precipitation samples and particulate
samples collected by Acidic Precipitation in Ontario Study (APIOS) network. Analysis has
been performed separately for the winter and summer seasons so that the comparison
between particles and precipitation and/or winter and summer can be made. The results show
that the U.S. midwest, east coast, and nearby Atlantic Ocean region are source areas for
sulfur species, both in precipitation and in ambient particles, at Dorset, Ontario, Canada. The
influence of these regions on SO^" level is stronger for the particles than for the precipitation,
and much stronger in summer than in winter. However, the influence on SO2 is much greater
in winter due to its longer life time in winter. Ocean emissions play a significant role in
summer owing to more biological and photochemical activities.
529
-------
Introduction
The Potential Source Contribution Function (PSCF) has been introduced to indicate
those geographical areas that have a high probability of being source areas of pollution events
at a specific receptor site1. The important feature of a PSCF analysis is the capability of
geographically locating pollution sources whereas most commonly used receptor models2
identify pollution sources by their chemical characteristics rather than their locations. It is
obvious that the locations of airborne pollutant sources are important information for air
quality management.
The PSCF method has been described by the authors3. A brief review is given here.
In a PSCF analysis, both sample chemistry data and related meteorological data are needed.
From meteorological data, air parcel back trajectories ending at a receptor site can be
calculated with a trajectory model4'5. The trajectory segment endpoint at particular time
relative to a sample period starting time is given by longitude and latitude. To calculate
PSCF, the whole geographic region covered by the trajectories is divided into 1° longitude by
1° latitude cells so that PSCF will be a function of longitude and latitude. The PSCF value
for the ij-th cell is calculated by:
- ^il (1)
where m,y represents number of endpoints that correspond to the trajectories that arrived at a
receptor site with pollutant concentrations higher than some pre-specified value, and n^ stand
for the total number of endpoints that fall into the ij-lh cell.
PSCFtj is the probability that an air mass with specified pollutant concentrations arrives
at a receptor site after having been observed to reside in the ij-lh cell. Cells containing
pollutant sources will have high probabilities. Therefore, the PSCF will identify those source
areas that have a potential to contribute to the high concentrations of contaminants observed
at the receptor site1.
Previously, the PSCF method was applied to the acidic precipitation chemistry data
obtained from Acidic Precipitation in Ontario Study (APIOS) network. The source locations
of the precipitation constituents were investigated in that study3. This paper will present a
PSCF analysis of sulfur species in both precipitation and ambient particles collected by the
APIOS network. The source locations of the species in precipitation and ambient particles
will be compared. The seasonal patterns of such locations will be demonstrated. A better -
understanding of atmospheric processes regarding acidic deposition is achieved by these
comparisons.
Data Description
The chemistry data of daily ambient paniculate samples and event precipitation
samples are taken from APIOS network data base6. This study only focuses on sulfur species
(SO^" and SO2) at one site in the network, Dorset (station code 3011, see Figure ]). Two
measurements for ambient particles were used: sulfate in jug SO^'/m3 and sulfur dioxide in ng
SOVm3. For precipitation, sulfate data were analyzed. Before calculating PSCF, the original
SOl concentration (in mg SOf/l) was converted to SOf wet deposition (mg SO;;") in a unit
area by multiplying the concentration by the precipitation sample volume in liter. The
purpose of this transformation is to eliminate the dilution effect caused by the differences in
precipitation volume.
The samples used in this study were those collected from 1984 to 1986. After the data
designated as unreliable were eliminated, the number of valid data points for precipitation
530
-------
l', paniculate SOf, and particulale SO2 were 401, 992, and 986, respectively, and the
average values of these variables are 0.993 mg SOj, 3.14 ng SO^'/rn3, and 3.67 ^g SO2/m3,
respectively. These average values were used as the cut-off values in PSCF computation of
corresponding species.
In order to investigate seasonal variation, ths data set for each variable was divided
into two subsets, one for winter (from November through April), another for summer (from
May through October).
Air parcel back trajectories ending at Dorse, have been calculated by Ontario Ministry
of the Environment4. The trajectory data were provided in the form of a list of time intervals
and coordinates of the trajectory segment endpoints for each trajectory. Trajectories using
surface level data were calculated each day at 0:00, 5:00, 12:00, and 18:00 using 3 hour time
intervals for 48 hours backward in time. In order to increase the resolution, a linear interpo-
lation was performed to obtain 1 hour time interval trajectories. Thus for each 24 hour
sample, 192 trajectory endpoints arc available for the PSCF analysis.
Results and Discussion
Samples with a species deposition (for precipitation) or concentration (for particles)
higher than the average value of this species arc designated as polluted samples. The PSCF
values can be calculated for each cell based on equation (1) where mi; is the number of
endpoints associated to these designated polluted samples and n- is the number of endpoints
for all samples. The PSCF values then can be plotted on the map (Figures 2-4).
Before interpreting these probability plots, two facts should be brought into notice: 1)
The cells near boundary of a figure usually have far less total number of endpoints. There-
fore, the confidence level for the PSCF results in th^se cells is much lower than the confi-
dence level in ceils in interior. To be specific the cells in interior usually have endpoints of
50-5,000 whereas the cells near the boundary, particularly on the west side, have 20 or fewer
endpoints. The results representing the interior of these figures are statistically more
significant. 2) Errors in a trajectory increase as the ;aiculation proceeds back in time from a
starting point (receptor site). In this study, the west side of the region is far from the Dorset
site (Figure 1) so that the trajectories over these areas probably have greater errors.
Precipitation Data
Figure 2 is the PSCF plots based on wet deposition of SO^ It shows winter (a) and
summer (b) patterns. For high SO^" wet deposition events, the areas from Tennessee,
Kentucky, and Indiana to the east coast and some aieas in the Atlantic Ocean are the source
areas. However, the influence of these areas in winter is much weaker than in summer
[compare Figure 2 (a) and (b)]. In summer, the most of these areas have PSCF values higher
than 0.6. Some cells in this region, particularly along the cast coasl and nearby ocean areas
have PSCF values in the range of 0.8-1. This result indicates that the wide region of U.S.
midwest and the east coast and some ocean areas have high probability of being the source
areas for summer high SO:;' wet deposition at Dorse . Some areas around Missouri also show
a strong effect. Considering the available emission inventory information and possible greater
errors in trajectories over this region (as discussed in previous paragraph), we believe that
these high PSCF value spots have been somewhat shifted from the areas between Missouri
and Illinois (around St. Louis area) to the location shown in the figures.
It has been observed that the cells in the Atlantic Ocean off the east coast of the
United States have high PSCF values in summer. It appears that the ocean becomes a
significant source in summer when biological processes are very active and high temperature is
favorable for these species to escape from the water. A important species transferred from
531
-------
ocean to atmosphere is dimethylsulfide (DMS). The DMS emission from ocean is comparable
in magnitude to the SO2 emission from fossil-fuel burning7. DMS will react in atmosphere
and produce SO2 and other products. Reactions of DMS and OH radical significantly
consume DMS and produce SO2 8'9, especially during the summer when photochemical activity
is high. Thus the combination of biogenic sulfur emissions and photochemical activity may be
the cause for the summer high PSCF values. According to these figures, the effect from the
ocean is negligible in winter. Another factor that could possibly cause this seasonal difference
is difference of transport pathways in winter and summer. In order to investigate the
pathways, all the trajectories over this region were reviewed. It was observed that even more
trajectories passed through this ocean region in winter than in summer. Therefore, the
difference appears to be due to the source intensity rather than lack of the transport pathway
in winter.
Ambient Paniculate Data
Two sulfur species have different seasonal variations (Figure 3 and 4). In summer, the
midwest region, especially the Ohio River Valley, is very strong source areas for SO;j at
Dorset, However, the SO2 level at Dorset is only slightly affected by the midwest region, and
strong influence is from the Atlantic Ocean. Although sulfur compounds' sources (SO2
emission) in the Midwest are similar in winter and summer, photochemical reactions are far
more important in summer. SO2 will be easily transformed to SOj so that the influence of
the midwest sources is observed in form of SO^. Only limited amount of SO2 remains in air
when the air parcel reaches the receptor site. For similar reasons discussed in the previous
subsection of precipitation, the ocean source can be seen in both SOj" and SO2 form and its
influence on SO2 is distinct. This result also suggests that SO2 produced by ocean has longer
life time than SO2 emitted from industries. More oxidants and radicals in continental
atmosphere and high temperature make SO2 to easily transform whereas marine atmosphere is
cleaner.
In winter, the midwest region has less influence on SOj" (but it is still a strong source
region) and a stronger influence on SO2 because SO2 - SOj transformation process is much
slower in winter. The effect of the ocean is very limited in winter. It is also noticed that
Sudbury area (see Figure 1), one of the largest SO2 sources in North America, has some
influence on the SO2 level at Dorset in the winter [see Figure 4(a)]. This effect is not
observed in summer because of the prevailing wind directions. No effect of Sudbury on SO^
(particles or precipitation, Figures 3 and 2) at Dorset is observed. The stack emitting SO2 at
Sudbury is very high and the distance between Sudbury and Dorset is short (about 225km).
There is only a limited probability that SO2 from Sudbury can reach the ground at Dorset and
there is not sufficient time for SO2 to transform to SO^" within that relatively short time
period.
Comparison between Precipitation and Particles
The comparison between precipitation and particles can only be made for SO^"
(Figures 2 and 3). In general, U.S. midwest region has a greater influence on particles than
on precipitation, especially in winter. In summer, the source areas for particles are concen-
trated in the Ohio River Valley region while more source areas for precipitation tend to be in
the east coast and the region around Missouri and Illinois.
According to this comparison, the sulfate coming from the midwest tends to remain in
particle phase. It does not efficiently participate in precipitation formation process and it is
not efficiently removed by the precipitation because the sulfate particles are too small to be
efficiently washed out. On the other hand, the species from ocean and the east coast seem
more efficiently involved the precipitation process. However, we have difficulty to interpret
the influence of western region (around Missouri) to the SO^" in precipitation. For further
532
-------
interpretation of this phenomenon, studies on meteorology and mechanism of cloud and
precipitation are needed.
The trajectories used in this work are surface trajectories. When multilayer trajecto-
ries become available in future studies, we could expect more information that may be helpful
in interpreting these results.
CONCLUSION
An area with large emission rate is not necessarily a source area of a pollutant at a
specific receptor site because factors, such as meteorological condition, atmospheric chemical
and physical processes, will determine the transport of the pollutant. With PSCF analysis,
pollution source locations can be easily identified in sense of probabilities. The PSCF analysis
has been applied to precipitation and particulate samp es that is collected by APIOS network
during 1984-86. The PSCF plots indicate the U.S. midwest, east coast, and nearby Atlantic
Ocean region as source areas for sulfur species, bolh in precipitation and in ambient particles,
at Dorset, Ontario, Canada. The SOj" level in particles is affected by these regions more
greatly than in precipitation. The influence on SO^ is much weaker in winter than in
summer. For SO7, it is much stronger in winter due to its longer life time in winter. The east
coast, the ocean region, and the region around St. Louis are more responsible for the acidic
species in precipitation whereas Ohio River Valley region is more responsible for the acidic
species in the particles. Ocean emissions play a significant role in summer owing to more
biological and photochemical activities. Sudbury, one of the most important source area in
North America, only has significant impact to SO2 level at Dorset in winter. These results
should be taken into consideration when developing control strategies.
This report was prepared for the Ontario Ministry of the Environment as part of a Ministry
funded project (No. 311 PL). The views and ideas expressed in this report are those of the author and
do not necessarily reflect the views and policies of the Ministry of the Environment, nor does mention
of trade names or commercial products constitute endorsement or recommendation for use. This work
was also supported in part by the U.S. National Science Fo mdation under grant ATM 89-96203.
References
1. W.C. Malm, C.E. Johnson, J.F. Bresch, In Receptoi Methods for Source Apportionment;
T.G. Pace Ed.; Publication TR-5, Air Pollution Control Association, Pittsburgh, PA, 1986,
pp. 127-148.
2. P.K. Hopke, Receptor Modeling in Environmenial Chemistry; John Wiley & Sons, New
York, 1985.
3. Y. Zeng, P.K. Hopke, Atoms. Environ.. 23: 1499-1509 (1989).
4. M.P. Olson, K.K. Oikawa, A.W. Macafee, "A trajectory model applied to the long-range
transport of air pollutants"; Report of Atmos. Environ. Service, Downsview, Ontario, 1978.
5. J.E. Bresch, L.L. Ashbaugh, T. Henmi, E.R. Reiter, "Comparison of a single-layer and a
multilayer transport model for residence time analysis"; 77th Annual Air Pollution Control
Assoc. Meeting, San Francisco, CA, 1984.
6. W.H. Chan, D.B. Orr, W.S. Bardswick,; R.J. Vet, "Acidic Precipitation in Ontario Study
(APIOS); An Overview: The Event Wet/Dry Deposition Network (1st revised edition)";
Ontario Ministry of the Environment, Report # ARB-142-85-AQM, APIOS-025-85, 1985.
7. C.F. Cullis, M.M. Hirschler, Atmos. Environ.. 14: 1263-1278 (1980).
8. D. Grosjean, Environ. Sci. Technol.. 18: 460-468 (1984).
9. R.J. Ferek, R,B. Chatfield, M.O. Andreae, Nature, !520: 514-516 (1986).
533
-------
Figure 1 Map of studied region, a) Dorset, the receptor site; b)
Sudbury, the bigest S02 point source in Canada.
a. Winter.
b. Summer.
Figure 2 PSCF based on SO^ wet deposition.
.a - 1
.6 - .8
.4 - ,6
.2 - .4
0 - .2
534
-------
a. Winter.
,8 - 1
.6 - .8
,4 - .6
.2 - .4
0 - .2
b. Summer.
Figure 3 PSCF based on ambient SO^" concentration.
a. Winter.
b. Summer.
Figure 4 PSCF based on ambient SO2 c concentration.
535
-------
MONITORING TOXIC VOCS IN URBAN AIR IN ILLINOIS
Clyde W. Sweet and S. J. Vermette
Atmospheric Chemistry Section
Illinois State Water Survey
2204 Griffith Drive
Champaign, IL 61820
Abstract
Toxic VOCs have been monitored at several sites in southeast Chicago and in the East St. Louis, IL
metro area since 1987 using a canister-based sampling system. In both of these areas, toxic VOCs are emitted
from a variety of area and point sources. Using wind trajectory analysis and chemical mass balance statistical
methodSj an attempt has been made to evaluate the effect of various types of emissions on overall levels of
toxic VOCs in urban air. The data have also been analyzed to determine the influence of wind direction on
concentrations of toxic VOCs in ambient air. In both study areas, major industrial point sources are
responsible for occasional episodes with elevated concentrations of toxic VOCs. Average ambient
concentrations, on the other hand, are influenced more by area sources that also affect regional background
levels.
Introduction
Toxic volatile organic chemicals (VOCs) are recognized as important cancer risk factors in urban air.1
Because these pollutants are emitted from such a wide variety of sources, identification and apportionment
of individual chemicals among major sources is a complex problem. Other researchers have identified and
characterized important sources of VOCs in urban areas.2 VOCs can be apportioned among these sources
using ambient air data with receptor modeling statistics3 or emissions inventory data and dispersion
modeling.4 However, there is a high degree of uncertainty associated with these estimations due to
uncertainties in emissions inventories, source profiles, meteorological variables and in the ambient air
databases.
For the past three years, we have been monitoring ambient air in three industrialized urban areas and
one rural background site in Illinois. The target compounds in this monitoring effort are a selected group of
toxic aromatic and polychlorinated hydrocarbons that can be readily quantified in ambient air. To reduce the
uncertainties associated with source apportionment by receptor modeling, we have used wind trajectory
analysis, characterized site specific source profiles and incorporated regional background data. This approach
is illustrated here using specific examples from our VOC monitoring data.
Methods
Samples and meteorological data were collected in East St. Louis and Granite City (part of the St. Louis
metropolitan area) and in southeast Chicago. Ail three locations have typical urban sources as well as heavy
industry (steel, chemicals, refineries and smelters). These areas have the worst air quality in Illinois in terms
of criteria pollutants. Background samples were collected at a rural site near Champaign chosen to be
representative of regional air quality.
536
-------
The sampling and analytical methods used in thi;; work and the results of quality assurance
experiments have been described previously.5 Briefly, sample.1, were automatically collected in stainless steel
canisters using a commercially available sampler (SIS, Moscow, ID). VOCs are then analyzed using cryogenic
pTeconcentration followed by capillary gas chromatography with flame ionization and electron capture
detection. Complete meteorological data including wind speed and direction were collected with all samples.
EPA software6 was employed for the chemical mass be lance (CMB) analysis. The chemicals analyzed
were the toxic VOCs together with propane, n-butane and isopentane which were used as additional fitting
species. Except where noted, source profiles are from Scheff et al.2
Flesults and Discussion
A data summary from approximately 250 analyses is presented in Table 1 for the 11 toxic compounds
that were quantifiable in most samples. Even though the aieas studied are heavily impacted by industrial
sources, the average concentrations of toxic VOCs are similar to average urban values in the U.S.7 and are
only 2 to 5 times higher than rural background levels in Illinois. The urban data are more variable than the
rural data with maximum concentrations often more than ten times higher than the maximum levels seen at
our rural background site. Typically the concentrations of to;cic VOCs in urban areas approach rural levels
except when the wind is blowing from a nearby major source.
This dependency on wind direction is shown in the diita from one of our sites (Sauget) in the E. St.
Louis area. This site is southeast of a large chemical manufacturing complex. This source affects the
concentrations of several toxic VOCs in samples taken of this site during periods of steady (standard deviation
of the wind direction < 20 ) northwesterly winds (Table 2). A similar relationship is seen with benzene at
the Chicago sites due to the influence of nearby coking operations. Used in this way, wind trajectory analysis
(WTA) can verify the influence of inventoried point sources on air quality and confirm the presence and
impact of uninventoried point sources.
Chemical mass balance (CMB) statistics were applied to a number of individual samples from the
urban and rural background sites. A composite of VOC measurements taken at our rural site was developed
as a regional profile. For East St. Louis, a chemical pliint profile was developed by subtraction of
simultaneous upwind from downwind fenceline measurements.5
Prior to a detailed breakdown of the sources of individual compounds in the urban samples, some
guneral findings are worth noting. CMB statistics for ttie rural samples are almost entirely accounted for by
tailpipe emissions (67%) and gasoline vapors (19%) with a variety of other sources making up the remaining
1.5%. The urban samples tend to be highly variable even when comparing samples taken during periods of
similar wind direction. For example, two East St. Louis samples were taken with winds coming from the
chemical plant area. In one case, CMB indicated that the chemical plants accounted for 93% of the benzene;
while in the other case, tailpipe emissions explained 100% of the benzene.
In Table 3 results are given for two CMB runs on a Chicago sample that reflected more or less
average conditions at the site. The first indicates that a subs:antial portion of the chlorinated solvents are
emitted from wastewater treatment. Although there is a sewage treatment plant in the study area, we have
found no evidence from the emissions inventory, WTA or fenceline sampling that this plant is a significant
point source. Secondly, chloroform, an important indicator of wastewater treatment emissions, is no higher
ir this sample than in an average rural sample. In the second CMB run, a regional background signature is
incorporated. This "source" accounts for most of the emissions. We interpret this to mean that most of the
toxic VOCs in this sample come from the same sources as those contributing to the regional background. The
additional compounds contributed by specific sources in the stidy area such as the coke ovens are highlighted
in this example.
Table 4 gives the results of CMB analysis of a sample taken in East St. Louis that was strongly
537
-------
influenced by chemical plant emissions. CMS accurately attributes some compounds (benzene for example)
to the chemical plants. Others such as trichloroethylene (TCE) are attributed to other sources even though
we know from WTA and fenceline sampling that the chemical plants are the dominant sources of TCE in this
area. The overall statistics for this CMB run are not very good. This is probably due to the fact that the
chemical plant profile is based on a limited number of grab samples. An improved profile would presumably
give more accurate source attribution. Another point of interest in this CMB analysis is that gasoline vapors
make up only a minor portion of toxic VOC emissions at this site even though petroleum product storage and
transfer station emissions make up 70% of the point source hydrocarbon emissions inventory. This result
agrees with WTA at the site (Table 2) in that levels of benzene, toluene and xylene are not elevated when the
sampler is downwind of these, sources.
Conclusions
1, The levels of airborne toxic VOCs in the study areas are strongly influenced by major point sources
downwind of the sampler.
2. CMB analysis of VOC monitoring data can give misleading results due to uncertainties in source
profiles and emissions inventories.
3. Wind trajectory analysis, specific source sampling and knowledge of regional background levels can
provide important information for interpreting CMB results.
References
1. V.E. Thompson, A. Jones, E. Haemisegger, B. Steigerwald, "The Air Toxics Problem in the United
States: An Analysis of Cancer Risks Posed by Selected Air Pollutants." JAPCA 35 535. (1985)
2. P.A. Scheff, R.A. Wadden, B.A. Bates and P.P. Aronian, "Source Fingerprints for Receptor Modeling
of Volatile Organics." JAPCA 39 469-478. (1989)
3. W.J. O'Shea and P.A Scheff, "A Chemical Mass Balance for Volatile Organics in Chicago." JAPCA
38 1020-1026. (1988)
4. D.D. Lane, R.E, Carter and G,A. Marotz, "Sampling and Preliminary Modeling of Ambient Air VOC
Content Using a Bulk Air Technique." Paper 86-20.5 presented at the 79th Annual Meeting of the Air
Pollution Control Association, Pittsburgh, PA. (1986)
5. C.W. Sweet and M.B. Willett, "Monitoring Toxic VOCs in Urban Air Using a Canister-Based Sampling
System." Proceedings of the 1989 EPA/AWMA Symposium on Measurement of Toxic and Related Air
Pollutants. Publication VIP-13 pp. 55-59. Air and Waste Management Association, Pittsburgh, PA.
(1989)
6. 1C Axetell and J.G. Watson, "Receptor Model Technical Series, Volume III (Revised): CMB Users
Manual (Version 6.0)." EPA-450/4-83-01R. Office of Air Quality Planning and Standards, U.S. EPA,
Research Triangle Park, NC (1987)
7. S.A. Edgerton, M. W. Holdren, D.L. Smith and J.J. Shah, "Inter-Urban Comparison of Ambient Volatile
Organic Compound Concentrations in U.S. Cities." JAPCA 39 729-732. (1989)
538
-------
Table 1. Toxic VOC concentrations in urban industrial areas3
Compound
benzene
toluene
m,p-xylene
ethylbenzene
o-xylene
chloroform
1,1, l,trichoroethane
carbon tetrachloride
trichloroelhylene
tetrachloroethylene
chlorobenzene
a Concentrations in /ig/m3 ± standard deviation, maximum value in parenthesis
Table 2. Wind trajectory analysis of Sauget/E. St. Louis VOC Data"
Compound Northwest Winds Other Winds Average
benzene 30 ± 8.0 2.2 ± 2.2* 14 ± 21
toluene 6±6 9±9 8±8
m, p-xylene 52 ± 73 2.6 ± 2.2* 21 ± 50
trichloroethylene 5.9 ± 10 0.5 ± 0.5* 2.7 ± 7.0
chlorobenzene 5.9 ± 6.8 0.5 ± 0.4* 3.1 ± 5.4
p-dichlorobenzene 58 ± 61 3.0 ± 1.8* 38 ± 54
a Concentrations are in jig/m3± standard deviation, n=10 for each wind sector, * indicates values are
significantly lower (p>.95) than other values in the same row.
Chicago
(n = 99)
4.2 i 4.2
(26)
9.1 ± 9.3
(56)
3.9 ± 3.4
(21)
1.4 ± 1.3
(7.6)
3.0 ± 5.7
(44)
0.3 ± 0.2
(1.6)
3.3 ± 3.6
(25)
0.8 ± 0.3
(1.7)
1.0 ± 1.0
(5.9)
1.8 ± 1.6
(9.1)
0.3 ± 0.2
(1.5)
E. St. Louis
(n = 81)
10.9 ± 17.5
(102)
8.6 ± 9.5
(45)
17 ± 43
(312)
7.2 ± 17
(110)
3.4 ± 8.8
(55)
0.5 ± 0.9
(6.6)
4.0 ± 6.1
(31)
1.0 ± 0.4
(1.9)
2.1 ± 5.8
(43)
1.4 ± 1.3
(6.1)
3.1 ± 6.4
(36)
Granite City
(n = 24)
2.3 ± 1.8
(7.5)
6.0 ± 7.7
(40)
4.2 ± 4.8
(24)
1.6 ± 1.8
(7.4)
3.3 ± 3.8
(13)
0.3 ± 0.1
(0.5)
1.5 ± 1.6
(7.5)
0.7 ± 0.4
(1.8)
0.6 ± 0.5
(2.1)
0.6 ± 0.7
(3.3)
0.4 ± 0.2
(0.7)
Rural Site
(n=19)
1.3 ± .5
(2.4)
2.7 ± 2.0
(9.4)
1.2 ± 0.8
(3.9)
.5 ± .4
(1.6)
1.1 ± 1.0
(4.3)
0.2 ± 0.1
(0.4)
1.2 0.4
(1.9)
0.8 ± 0.3
(1.5)
0.4 ± 0.3
(1.5)
0.4 ± 0.3
(1.2)
0.2 ± 0.1
(0.5)
539
-------
Table 3. CMB results from Chicago (Washington School)
Filling statistics
Date = 88/07/10
Wind Direction = 225°
Wind Speed = 13 mph
VOC
Benzene
Toluene
Ethylbenzene
m,p-Xylene
o-Xylene
Chloroform
Melhylchloroform
Trichloroethylene
Perchioroethylene
Date = 88/07/10
Wind Direction - 225°
Wind Speed = 13 mph
VOC
Benzene
Toluene
Ethylbenzene
m,p-Xylene
o-Xylene
Chloroform
Mcthylchlorofonn
Trichloroethylene
Perchioroethylene
Calculated
Measured
0.84
0.96
1.05
1.03
0.67
1.07
0.75
1.19
1.00
Calculated
Measured
0.98
1.02
0.%
0.99
0.98
1.36
0.81
0.84
1.00
R2 = 0.97 DF = 7
X2 = 3.45 Predicted Total Concentration
of Fitting Elements - 96%
Tailpipe Gasoline Wastewater Vapor Dry
Emissions Vapors Treatment Degreasing Cleaning
81% 1% 11% 0% 0%
71% 4% 26% 0% 0%
82% 2% 16% 0% 0%
96% 5% n.a. 0% 0%
98% 2% n.a. 0% 0%
n.a. n.a. 100% 0% 0%
n.a. n.a. 38% 62% 0%
n.a. n.a. 27% 73% 0%
n,a. n.a. 37% 19% 44%
Fitting statistics
R2 = 0.99 DF = 7
X2 = 1.28 Predicted Total Concentration
of Fitting Elements = 98%
Coke Gasoline Architectural Dry Regional
Ovens Vapors Coatings Cleaning Background
38% 5% 0% 0% 56%
0% 2% 44% 0% 53%
4% 2% 7% 0% 86%
2% 4% 15% 0% 79%
0% 1% 18% 0% 81%
n.a. n.a. n.a. 0% 100%
n.a. n.a. n.a. 0% 100%
n.a. n.a. n.a. 0% 100%
n.a. n.a. n.a. 48% 52%
n.a. not available
540
-------
Table 4. CMB results from East Si. Louis (Sauget)
Dale = 88/11/06
Wind Direction = 270°
Filling statistics
*2 - 0.90 DF = 5
X2 = 10.78 Predicted Total Concentration
of Fitting Elements - 101%
VOC
Benzene
Toluene
Bthylbenzene
m,p-Xylene
o-Xylene
Chloroform
Melhylchloroform
Trichloroethylene
Perchloroelhylene
Chlorobenzene
Calculated
Measured
0.37
1.23
1.31
1.17
0.69
0.84
0.47
1.13
1.15
1.18
Tailpipe Gasoline
Emissions Vapors
8% 2%
74% 8%
40% 2%
41% 4%
96% 4%
n.a. n.a
n.a. n.a
n.a. n.a
n.a. n.a
n.a. n.a,
Wastewater
Treatment
1%
15%
4%
n.a.
n.a.
40%
5%
4%
16%
0%
Vapor
Degrea&ing
0%
0%
0%
0%
0%
0%
52%
72%
49%
0%
Chemical
Plant
91%
3%
57%
55%
0%
59%
43%
23%
36%
100%
n.a. not available
541
-------
CABIN AIR QUALITY: COTININE AS A BIOMARKER OF
ENVIRONMENTAL TOBACCO SMOKE IN COMMERCIAL
AIRCRAFT
Delbert J. Eatough, Fern M. Caka, John
Crawford, Scott Braithwaite, Lee D. Hansen and
Edwin A. Lewis
Department of Chemistry, Brigham Young
University, Provo, Utah 84602, U.S.A.
The use of cotinine as a biomarker for evaluating air quality with
respect to environmental tobacco smoke in commercial aircraft has been
studied by determining the variation in concentration of nicotine and other
environmental tobacco smoke pollutants in both smoking and nonsmoking cabin
sections. Four never-smoker volunteers were exposed during commercial
passenger flights and atmospheric samples were collected to determine
exposure of the personnel to nicotine for at least the 24 hour period before
and 48 hour period after the flight. Total urine samples were obtained from
the personnel collecting the atmospheric samples for the time period
extending from 24 hours before the flight to 48 hours after the flight.
Exposure in airport terminals can be as significant as exposure for persons
sitting a few rows in front of the smoking section during a flight. Urine
cotinine concentrations were correllated with exposure to nicotine but not
with exposure to many other constituents of environmental tobacco smoke in a
series of DC-10 flights.
542
-------
Introduction
In recent years there has been an Increased interest in determining the
concentrations of ETS in commercial aircraft passenger cabins in order to
quantify the concentrations of pollutants afssociated with ETS which may be
present, determine the factors which control the concentrations of ETS
present in nonsmoking sections of passenger cabins, and develop models for
predicting personal exposure in commercial aircraft. A National Academy of
Sciences report (1) recommended banning smcking on all commercial flights
for the following reasons: minimization of irritation, reduction of health
risks and fire hazards, and to bring levels of pollutants in cabin air in
line with those in other indoor environments. In April 1988, the U.S.
Congress enacted a temporary law banning smoking on all flights of two hours
or less. In February 1990, a new law went into effect which banned smoking
on domestic U.S. airline flights. Similar legislation is in effect in
Canada. Most flights in other countries do not presently ban smoking.
Several studies have determined the concentration of ETS components
present in commercial aircraft cabins. Data have been reported on the
concentrations of nicotine present in the cabin environment for a number of
commercial aircraft flights (2-6). The exposure of airline flight
attendants (4,7) and passengers (4) to environmental tobacco smoke has been
estimated from measurements of nicotine and ^otinine in urine. Oldaker et.
£.1 (2) have reported the determination of the concentrations of nicotine,
P.SP and UV-PM on several long commercial flights using a portable air
sampling system. A similar sampling system was used to determine the
concentrations of nicotine, CO and RSP at four locations in the passenger
cabin of flights on MD-80 aircraft (5). The latter two studies are the
only studies reported to date which bavts attempted to correlate the
concentrations of nicotine in the passenger cabin of commercial aircraft
with the concentrations of other constituents of environmental tobacco
smoke. Most of the studies have used nicotine as the tracer to quantify
exposure. However, exposure calculations based only on nicotine will
underestimate total exposure to ETS since nicotine is removed from indoor
environments at rates faster than other species associated with ETS (8,9).
We have conducted a study (10) to measure a variety of compounds
associated with ETS as well as several non-unique species (such as RSP and
CO) in both smoking and nonsmoking sections of aircraft cabins. The
spectrum of species and aircraft sampled provided a data base for the
development of models for the prediction of ETS concentrations in aircraft
cabins under a variety of conditions. As part of that study, the concen-
trations of cotinine in the urine of passengers with known exposure to
nicotine from environmental tobacco smoke before, during and after
commercial flights was determined. This paper presents the results obtained
from a series of DC-10 flights.
Methods
Sanpling Equipment and Analysis Methods
Data on the aircraft were collected by ::our volunteers using Briefcase
Automated Sampling Systems (BASS) (11). Each BASS contained various
components designed to sample for specific compounds associated with ETS and
other atmospheric pollutants and variables. Compounds measured during a
flight included gas and particulate phase nicotine, 3-ethenylpyridine, fine
(>2.5pm) particulate matter, UV-PM (2), NOX and CO. The concentrations of
gas and particulate phase nicotine were determined using two mini-annular
denuder sections coated with benzenesulfonic acid (BSA) for collection of
gas phase nicotine and 3-ethenylpyridine followed by a 1 micron Teflon
543
-------
filter (Zefluor, Gelman Sciences) for collection and determination of
nicotine (12). Following the Teflon filter was a BSA saturated filter for
the collection of any nicotine lost from particles during sampling (12) .
Air was drawn through the denuder system at a rate of 2 sLpm. Details on
the collection device used to measure the other chemical species have been
given (10,11). The concentrations of nicotine to which the four volunteers
were exposed in airports prior to and after each flight were determined
using a filter pack with a 1 micron Teflon filter (Zefluor, Gelman Sciences)
and a BSA saturated filter sampling at 4 sLpm for the collection of
particulate and gas phase nicotine (12). The concentrations of nicotine
present in the environment other than in the airport terminals and during
the flight to which the volunteers were exposed was determined using a BSA
saturated filter in a passive monitor of the design described by Hammond et
al. (13).
A measured fraction of each urine void was collected by each of the
four volunteers for the 24 hour period prior to and 48 hour period following
each flight. Aliquots of the various samples were combined to give three
composite 24 hour samples. After collection the individual urine samples
were frozen and kept frozen at -80°C until combined to the 24 hour samples
and analyzed.
Annular denuder and filter pack samples were extracted with water and
analyzed by ion chromatography for nicotine and 3-ethenylpyridine (14) with
the exception of the Teflon filter. The Teflon filter was extracted with
methanol, with half of the extract analyzed for UV absorbance using a
spectrophotometer to determine UV-PM (15), and the other half was analyzed
for nicotine by ion chromatography (16). The concentrations of nicotine
collected by the passive samplers was determined by gas chromatography using
an NPD detector (12). Cotinine in the collected urine samples was
determined by gas chromatography using an NPD detector and internal standard
as previously described (14).
Sampling Protocol
Four different volunteer non-smokers participated in four DC-10
flights. Each subject carried a BASS and was seated in the rear passenger
cabin which contained the economy class smoking section at the back of the
aircraft. The location in the passenger cabin of the volunteers during each
flight is given in Table I. Sampling was begun after takeoff when the no-
smoking sign was turned off. Sampling was concluded when the no-smoking
sign was turned on prior to landing. Flights 1 and 3 and flights 2 and 4
were the same origination and destination, however, a different DC-10
aircraft was flown for each flight. The four volunteers wore the personal
passive sampler for a 24 hour period prior to the flight when not in an
airport and were at the location of the filter pack sampling system when in
an airport. After the flight, the volunteers were again in the area of the
filter pack sampling system when in the airport and wore a passive sampler
for the 48 hour period after leaving the airport. All volunteers stayed in
smoke free residences and avoided any locations where smoking was present
before and after each flight.
Results and Discussion
The concentrations of nicotine and the time duration of exposure for
the four volunteers for the periods prior to, during, and after each of the
four DC-10 flights are given in Table I. The concentrations given are for
total nicotine determined using the annular denuder or filter pack sampling
system, or of gaseous nicotine determined using the passive sampler. In all
cases where both gas and particulate phase nicotine were determined, <95% of
544
-------
the nicotine was in the gas phase.
The concentrations of nicotine and other selected environmental tobacco
smoke constituents as a function of seat location in a flight with a high
and moderate concentration of environmental tobacco smoke are given in
Figure 1. Complete data for the four flights are available (10,17). The
rate of penetration of environmental tobacco smoke constituents from the
smoking section into the nonsmoking section follows a first order mechanism
(10). The rate of penetration was the same for the various DC-10 aircraft
flown in this study. The expected rate of decrease in the concentration of
various constituents with distance into the nonsmoking section can be
altered by selective removal of compounds by cabin surfaces (e.g. nicotine,
Figure 1) or by the presence of non-ETS sources of some species in the
nonsmoking section, e.g. CO, RSP or NOX (10).
In some cases exposure of the four volunteers to environmental tobacco
smoke in the airport terminal was significant, Table I. In all cases,
exposure to environmental tobacco smoke other than in the aircraft cabin or
in the airport terminals was insignificant. The airport terminal exposure
concentrations are very low for the after flight data for Flights 1 and 3
and the before flight data for Flights 2 and 4 because of minimal smoking
and the open air nature of the airport terminal. Because of the significant
concentration present in the airport terminal and the longer waiting period
before Flight 1, Subjects C and D on this flight were exposed to more
nicotine in the airport terminal than during the flight even though they
tried to avoid cigarette smoke in the terminal, Table I.
A total dose exposure to nicotine, Table II, was calculated for each of
the subjects on each of the flights from the measured concentrations of
nicotine in the airport terminals and during the flights, and the time of
exposure at each location. An average tidal volume of 8.5 L/min (17) was
used in these calculations. The total cotinine excreted in the 24 hour
period before each flight and in the first: two 24 hour periods after each
flight are given in Table II. The calculated dose is compared to the 48 hr
excreted cotinine in Figure 2.
The amount of cotinine excreted during the first (X;[) and second (X2)
24-hr periods after each flight can be used to calculate the fraction of the
total to be excreted present in the first 24-hr sample. This number was
constant for all volunteers where X2 was measurable, 0.8410.07, except
subject A in Flight 1. The cotinine elimination half time of 9±2 hr agrees
with the results of controlled exposure studies (14). Linear regression
analysis of the data given in Figure 2 gives r-0.78 with a calculated slope
of 0.13±0.03 mol cotinine/mol nicotine. The slope is consistent with the
expected conversion of about 10% of the inhaled nicotine to excreted
cotinine (14). The large uncertainty in the slope apparently results from
individual variations in nicotine metabolism, Table II. These variations
result in an uncertainty of a factor of 2 in the use of urinary cotinine to
predict exposure to nicotine. However, the more rapid removal of nicotine
as compared to other constitutents of environmental tobacco smoke in the
cabin environment, Figure 1, results in larger errors in the use of cotinine
to estimate exposure to these constituents:. Such estimates are low by a
factor of from 2-6 (10,17) for the data reported here.
Acknowledgement
This research was supported by a granc from the Center for Indoor Air
Research. Appreciation is expressed to An~on Jensen, Laura Lewis and John
D. Lamb for technical assistance and to S. Katherine Hammond for providing
passive samplers for the study.
545
-------
References
1. NAS "The Airliner Cabin Environment: Air Quality and Safety." National
Research Council. National Academy Press, Washington, B.C., 303 pp (1986).
2. G.B. Oldaker III, M.W. Stancill, F.C. Conrad Jr., B.B. Collie, R.A.
Fenner, S.O. Lephardt, P.G. Baker and S. Lyons-Hart Indoor Air Quality and
Ventilation. Lanau F. and Reynolds G.L., eds, Selper Ltd, pp 442-454 (1990).
3. G.B. Oldaker III and F.C. Conrad Jr. Environ. Sci. Technol. 21: 994-999
(1987).
4. M.E. Mattson, G. Boyd, D. Byar, C. Brown, J.F. Callahan, D. Corle, J.W.
Cullen, J. Greenblatt, N.J. Haley, S.K. Hammond, J. Lewtas and W. Reeves
JAMA 261: 867-872 (1989).
5. T. Malmfors, D. Thorburn and A. Westlin A. Environmental Technology
Letters 10: 613-628 (1989).
6. M. Muramatsu, S. Umemua, J, Fukui, T. Arai and S. Kira Int. Arch.
Occup. Environ. Health 59: 545-550 (1987).
7. D. Foliart, N.L. Benowitz and C.E. Becker N. Engl. J. Med. 308: 1105
(1983).
8. D.J. Eatough, L.D. Hansen and E.A. Lewis Environmental Tobacco Smoke.
Proceedings of the International Symposium at McGill University 1989. D. J.
Ecobichon and J. M. Wu, Eds., Lexington Books, 1990, pp 3-39.
9. H. Tang, D.J. Eatough, E.A. Lewis, L.D. Hansen, K. Gunther, D. Belnap
and J. Crawford, Measurement of Toxic and Related Air Pollutants. Air and
Waste Management Association, pp 596-605 (1989).
10. D.J. Eatough, P.M. Caka, J. Crawford, S.K. Braithwaite, L.D. Hansen and
E.A. Lewis. "Environmental Tobacco Smoke in Commercial
Aircraft."Proceedings. Indoor Air '90. submitted (1990).
11. D.J. Eatough, F.M. Caka, K. Wall, J. Crawford, L.D. Hansen and E.A.
Lewis Measurement of Toxic and Related Air Pollutants. Air and Waste
Management Association, pp 565-576 (1989).
12. F.M. Caka, D.J. Eatough, E.A. Lewis, H. Tang, S.K, Hammond, B.P.
Leaderer, P. Koutrakis, J.D. Spengler, A. Fasano, M.W. Ogden and J. Lewtas
"An Intercomparison of Sampling Techniques for Nicotine in Indoor
Environments", Eny. Sci. and Tech.. in press (1990).
13. S.K. Hammond and B.P. Leaderer Environ. Sci. Tech. 21: 494-497 (1987).
14. E.A. Lewis, H. Tang, W. Winiwarter, K. Gunther, D. Belnap, A. Jensen,
L.D. Hansen, D.J. Eatough, N.J. Baiter and S.L. Schwartz. "Use of Urine
Nicotine and Cotinine Measurements to Determine Exposure of Nonsmokers to
Sidestream Tobacco Smoke," Proceedings. Indoor Air '90. submitted (1990b).
15. J.R. Carson and C.A. Erikson Environ. Tech. Letters 9: 501-508 (1988).
16. L.J. Lewis, J.D. Lamb, D.J. Eatough, L.D. Hansen and E.A. Lewis J. of
Chrom. Sci. 28. 200-203 (1990a).
17. D.J. Eatough, F.M. Caka, J. Crawford, L.D. Hansen and E.A. Lewis,
Report to the Center for Indoor Air Research (1990b).
546
-------
Table I. Concentrations of Total Nicotine from Environmental Tobacco Smoke
to Which the Volunteers were Exposed Before, During and After Four DC-10
Flights. Rows
Before Total Nicotine, runol/nr (Exposure Time, hr)
Smoking In Airport In Airport
Flight Subject Seat Section Before Flighta During Flight After Flight3
319 (4.87)b >1 (1.20)a
208
15.4
0.1
84 (4.30)b >1 (1.0)a
78
39
0.1
475 (4.83)b >1
304
127
20.4
71 (4.55)b 13
36
12.2
1.2
aBoth concentration and time were the same for all four volunteers.
Time was the same for all four volunteers.
1
2
3
4
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
35D
34G
32J
17D
34C
33C
33A
24C
32F
31F
30D
26F
35D
33D
29C
24F
0
0
1
16
0
1
1
10
0
1
2
6
0
2
6
11
58 (4.50)
>1 (1.0) a
39 (1.87)
>1 (1.3)a
(0.5)'
"able II. Nicotine Exposure and Cotinine in Urine for Each of the
Volunteers Involved in the DC-10 Flights.
Flight Subject
1 A
B
C
D
2 A(C)C
B(B)
C(A)
D(D)
3 A
B
C
D
4 A(C)
B(A)
C(B)
D(D)
Inhaled
Nicotine
nmol
925
650
171
133
184
171
86
0.2
1207
786
350
87
162
88
32
7
% of
Exposure
During
Flight
86
80
22
0.2
100
100
100
100
97
95
89
57
98
95
87
41
Total Urinary Cotinine, nmol
24 Hr
Before
Flight
<3
152a
NAb
<3
NAb
11
<3
<3
<3
<3
4
<3
0-24 Hr
After
Flight
144
107a
NAb
16
NAb
88
35
20
9
<3
3
24-48 Hr
After
Flight
66
19a
NAb
<3
NAb
N^L
23
4
3
<3
<3
<3
% of
Inhaled
Nicotine
22.7
19.4
NA
11.9
NA
NA
9.2
5.0
6.6
9.9
- -
e-Preflight exposure to nicotine resulted from work in the analytical lab and
not from ETS. Subsequent data have been corrected for this exposure
6.ssuming a constant elimination half life.
Cotinine could not be determined due to nitrogen containing interfering
compounds.
cThe letter in () is the identification for this volunteer in the previous
flight.
547
-------
o
o
d
o
-I
-1.50
-2.0O
6
O
*>»
O
Q
O
12
Rows from Smoking Section
Figure 1. The log(C/C0) , where C0 is the concentration in the smoking
section, versus number of rows from the nonsmoking section into
the smoking section for Flights 3 (high exposure) and 4 (moderate
exposure).
548
-------
o
o
o
o
CJ
o
o
o
T- ^
o
o
00
o
o
(0
o
o
o
o
CM
o
o
z
"o
c
0)
CO
o
Q
s-l
Hi
4-1
lw
tfl
10
3
o
CO
Hi
o
•H
4J
0)
(D
c
•H
4J
o
u
o
p.
X
0)
c
• H
4J
O jj
0 X
•d to
C 'r^
r-l
T-l U-l
cO
4-> Q)
o x
H 4J
CM
4!
O O O
to o v>
CM CM i-
O O O
o in
549
-------
PROBLEMS WITH THE USE OF NICOTINE
AS A PREDICTIVE ENVIRONMENTAL
TOBACCO SMOKE MARKER
P.R. Nelson, D.L. Heavner, G.B. Oldaker III
R.J. Reynolds Tobacco Company
Bowman Gray Technical Center
Winston-Salem, NC 27102 USA
A series of experiments was performed to evaluate the utility of nicotine as an
environmental tobacco smoke (ETS) marker. Two University of Kentucky reference
cigarettes (1R4F) were smoked in an 18-m environmental chamber. Air exchange
rates within the chamber were varied from 0-4 air changes per hour, and the
concentrations of numerous ETS components were monitored for up to six hours after
smoking. Under most ventilation conditions, nicotine initially decayed more rapidly
than other ETS constituents; however, as sampling time was extended, nicotine decayed
more slowly. The change in nicotine decay rate can lead to overestimation of ETS
exposure when nicotine is used as the sole ETS marker. Confirmatory results obtained
from other field and chamber studies are also presented.
INTRODUCTION
The utility of nicotine as an environmental tobacco smoke (ETS) marker has
been questioned by a number of investigators (1-3). Some have simply stated that
it is a poor marker (1), while others have suggested that nicotine may underestimate
ETS exposure (2,3) by as much as an order of magnitude.
The National Research Council (NRC) recommended that any chosen ETS
marker should be present in a consistent ratio to ETS components of interest (4).
To attain this criterion, the marker should possess the same decay characteristics as
the component of interest.
If an ETS component is not generated or eliminated by chemical reaction, or
does not interact with environmental surfaces, then it should demonstrate a first order
decay with a rate constant proportional to the air exchange rate in a
microenvironment. Previous studies have shown that nicotine does not undergo first
order decay in microenvironments (2,3).
550
-------
This present work was performed to systematically evaluate the effect of
nicotine's unique decay on its ratio to other ETS components. Due to nicotine's
different decay rate, both ventilation rate and sampling time would be expected to
exert an effect on the ratio of nicotine to other ETS constituents. Therefore, ratios
were determined at air exchange rates of 0, 0.5, 1, 2 and 4 air changes per hour
(ACH), and time-weighted-average concentrations of selected analytes were determined
for 0-30, 0-60, 0-120, 0-240, and 0-360 minutes fol owing the smoking of 2 University
of Kentucky reference cigarettes (1R4F). The results obtained in the chamber were
then related to results obtained previously in fielc and chamber studies.
EXPERIMENTAL
Chamber Studies
<}
All ETS decay experiments were performed in an 18-m environmental
chamber described elsewhere (5). Three to five replicate experiments were performed
at 0, 0.5, 1, 2, and 4 ACH. Real-time nicotine concentrations were monitored with
a SCIEX TAGA 6000 tandem mass spectrometer (6). Real-time concentrations of
carbon monoxide, nitrogen oxides, volatile organic compounds (estimated by FID
response), and particle mass concentration were obtained with commercial analyzers
described elsewhere (7). Vapor phase nicotine and 3-ethenyIpyridine were collected
using XAD-4 sorbent tubes and analyzed by gas chromatography with nitrogen
phosphorus detection (8), Solanesol, gravimetric respirable suspended particles (RSP),
ultraviolet particulate matter (UVPM), and fluorescent particulale matter (FPM) were
collected on Fluoropore filters and analyzed as described elsewhere (9,10). Duplicate
nicotine and particulate samples were collected over the periods 0-30, 30-60, 60-120,
120-240, and 240-360 minutes, and integrated average concentrations were determined.
Each run during the decay rate studies lasted a total of 384 minutes. The first
twelve minutes of the run were used to measure Background concentrations of ETS
constituents. A smoker then entered the chamber and smoked two University of
Kentucky 1R4F cigarettes in 10'/2 minutes. The two cigarettes were lit at 30-second
intervals, and the smoker took one puff on alternating cigarettes at one-minute
intervals. Each cigarette lasted for an average of ten puffs. At 24 minutes, the
smoker exited the chamber which was subsequently resealed for the final 360 minutes
of the experiment.
The effect of residua] nicotine on a smokers' clothing was determined from the
a.verage of 15 runs performed at various times over a three-month period in the
environmental chamber which was operated at 0 ACH. After a twelve-minute
background measurement, the smoker entered the chamber and stayed for twelve
minutes. The smoker then exited the chamber, and nicotine concentration was
monitored for an additional 36 minutes.
Field Studies
Nicotine samples were collected on XAD-4 sorbent tubes and analyzed by the
method of Ogden et al. (8). Six-hour samples were collected in a smoker's van and
the den of a non-smoker's house. Eight one-houi samples were collected overnight
in a B767 aircraft which had completed a flight on which smoking was allowed.
551
-------
Particulate samples were collected and analyzed for the aircraft study by the same
methods used in the chamber study.
RESULTS AND DISCUSSION
Nicotine Decay
MO
At each of the air exchange rates studied, nicotine initially decayed more
rapidly than the other ETS constituents measured. However, at longer times the
decay of nicotine slowed, and typically achieved a near-steady state concentration
which was higher than the initial background level. On the other hand, constituents
such as CO, volatile organic compounds, nitrogen oxides, and particle mass
concentration decayed according to first order kinetics with decay rate constants which
were proportional to the air exchange rate. The other constituents did decay to initial
background levels in a time-scale consistent with normal first order decay.
A plot of nicotine concentration vs. time obtained from the average of five runs
performed at 2 ACH is shown in Figure 1. The solid line in this figure is the average
nicotine concentration measured in the chamber. The dashed line in the same figure
is the nicotine concentration profile which would be predicted by first order decay.
The theoretical curve in Figure 1 is
representative of the behavior of
the other measured ETS
constituents. Time-weighted-
average (TWA) concentration ratios
between nicotine and the other
constituents are proportional to the
ratio of the areas under the two
curves. For a sampling period
extending from 0-30 minutes, the
ratio of nicotine to analyte will be
lower than that predicted by first
order kinetics. At about 30
minutes, nicotine decays less rapidly
than the other constituents, and for
sampling times greater than 60
minutes, the ratio of nicotine to
0.0
1.0
2.0 3.0 4.0
TIME (hours)
8.0
Figure 1. Average of five real-time nicotine
concentrations (solid) measured in a controlled
environment chamber operated at two ACH. The
dashed line represents the concentration predicted by
a first order decay mechanism.
other analytes will become increasingly larger than those predicted by first order decay
of constituents. The magnitude of overestimation becomes larger if measurements are
started at times long after smoking has occurred. For the example illustrated in
Figure 1, the other constituents would have decayed to background at about 180
minutes; but significant concentrations of nicotine are still present in the chamber.
In this case, a person exposed to the atmosphere in the chamber would not be
exposed to measurable ETS particulate, CO, or volatile organic compounds, but they
would have measurable exposure to nicotine.
Ratios Involving Nicotine
The actual variations of nicotine to RSP and nicotine to FID response ratios
as a function of both sampling time and air exchange rate are shown in Figures 2a
552
-------
and 2b. Ratios presented in these figures have been normalized to the values
obtained for the 0-30 minute sample at 0 ACH. Sampling for increasing periods of
time or sampling for a constant period of time at different air exchange rates has a
dramatic influence on the observed ratio of nicotine to either vapor or paniculate
phase ETS components. Furthermore, these figures demonstrate that ratios of
nicotine to other ETS components which are determined in chambers operated in
static modes are not applicable to other environments because of the large
overestimation of ETS exposure which they would predict.
Figures 2a & 2b. Normalized ratios between nicotine and gravimetric RSP (2a) and FID
response (2b) as a function of both air exchange rate and sampling time. All values in figure
2a and 2b have been normalized to the ratio at 0 ACH and 30 min (0.108 in 2a & 74.1 in
2b). Lines of constant ratio are drawn at intervals of 0.333 beginning at a ratio of 0,667.
Field Measurements
Results consistent with those obtained in the chamber also have been observed
in the field. Background nicotine concentrations in the absence of smoking have been
measured in homes, automobiles and aircraft. In addition, nicotine desorption from
the clothes and person of a smoker has been observed in the environmental chamber.
Results of these studies suggest that the initial rapid decay of nicotine is due to its
adsorption on clothing and other surfaces. As atmospheric nicotine is depleted, due
to the effects of adsorption and dilution by fresh air, adsorbed nicotine then desorbs,
and leads to measurable nicotine concentrations in the absence of smoking.
Figure 3 shows that smokers can be a source of nicotine contributed to the
environment even when they are not smoking. This figure shows the average increase
in nicotine concentration measured during 15 replicate experiments performed over
a three-month period in which a smoker entered the environmental chamber for
twelve minutes. The increased nicotine concentration is presumably due to the
evolution of nicotine from the smoker's clothes. When the smoker exited the
chamber, the nicotine concentration ceased its increase.
A nicotine concentration of 0.09 jug/m w,as measured in the den (=45 m )
of a non-smoker's house. The sample was obtained two days after smoking had
occurred in the room. This background level corresponds to small amounts of
residual nicotine desorbing from room furnishings over a long period of time.
Nicotine samples were collected in a van (=5 m ) in which smoking regularly
553
-------
occurred. The sample was
taken overnight and at least
four hours after smoking had
occurred in the vehicle. The
sample was obtained at
dashboard level =40 cm above
an open ashtray. A background
concentration of 0.126 /xg/nr
was measured in the vehicle.
Once again, this background is
not due to the presence of
ETS, but instead it comes from
nicotine desorbing from the
interior of the van and from
cigarette butts present in the
ashtray.
0.75
o> oao
C
1X00
—oja
15 30 45
TIME (mln)
60
Figure 3. Average increase in nicotine concentration
due to the nicotine desorption from a smoker's
clothing in a controlled environment test chamber
operated at 0 ACH. The smoker was in the
chamber for the period 12-24 minutes.
Samples were also collected in a B767 aircraft in which smoking was permitted.
The aircraft had returned from a regularly scheduled 4 hour 20 minute flight from
Los Angeles, CA to Charlotte, NC. Samples were collected when the plane was on
the ground and parked at the gate after all passengers and crew had left the plane.
While the samples were collected, the aircraft was served by an auxiliary heating,
ventilating and air conditioning system which provided fresh air at 13-26 ACH (11).
Smoking was not permitted while the aircraft was at the gate, and none was observed.
Samples were obtained in the coach smoking section and seats in the non-smoking
border section. A total of 64 cigarette butts were counted in the ashtrays prior to
sampling. The results of this investigation are presented in Table I. During time
period 4, the crew
collected trash, emptied Table I. Background nicotine concentrations (/ig/m3)
measured overnight on a B767 aircraft with auxiliary
ventilation. Smoking samples (S.) were obtained in the
coach smoking section. Non-smoking samples (N.S.) were
obtained in coach smoking/nonsmoking border seats.
Nicotine
Time
S.
N.S.
ashtrays during period 5,
and vacuumed the cabin
during period 7. UVPM
measurements obtained
from samples taken over
the entire 8-hour period
showed less than 1.5
T
^ig/m particulate matter
was present which could
possibly be attributed to
ETS. Once again, the
only source of nicotine in
the cabin would appear
to be nicotine desorbing
from interior surfaces
and cigarette butts. Measurable nicotine exposure could be expected in the non-
smoking boundary section without concurrent exposure to ETS gas or particulate
phase material.
1
2
3
4
5
6
7
8
22:20-23:20
23:20-00:20
00:20-01:20
01:20-02:20
02:20-03:20
03:20-04:20
04:20-05:20
05:20-06:20
5.2
4.9
2.8
8.7
8.0
4.7
5.2
8.2
1.7
1.3
1.1
1.8
1.6
1.3
2.7
2.5
554
-------
CONCLUSIONS
Nicotine does not fit the NRC criteria for an ETS marker. The ratio of
nicotine to other ETS constituents such as RSP are highly variable and dependent on
both the air exchange rate at the sampling site and sampling time. Desorption of
nicotine from clothing, interior surfaces, and cigarette butts leads to measurable
nicotine exposure in the absence of ETS. The findings reported here show that when
nicotine is used as the sole marker, it may greatly overestimate ETS exposure.
ACKNOWLEDGEMENTS
The hard work of Ms. Barbara Collie, Katherine Maiolo, Patricia
DeLuca and Mr. Fred Conrad in collection and analysis of samples and the
helpful suggestions of Dr. Mike Ogden are gratefully acknowledged.
REFERENCES
1. C.W. Bayer and M.S. Black, "Passive Smoking: Survey Analysis of Office Smoking
Areas vs. Environmental Chamber Studies," Proceedings of the ASHRAE Conference
IAQ '86, 1986, pp 281-291.
2. R.R. Baker, P.D. Case, and N.D. Warren, "The Build-up and Decay of Environmental
Tobacco Smoke Constituents as a Function of Room Condition," in Indoor and
Ambient Air Quality, Selper, London, 1988, pp 121-130.
3. NJ. Baiter, D.J. Eatough, and S.L. Schwartz, "Application of Pharmacokinetic Modeling
to the Design of Human Exposure Studies with Environmental Tobacco Smoke," in
Indoor and Ambient Air Quality. Selper, London, 1988, pp. 179-188.
4. Environmental T°bacc° Smoke, National Research Council, National Academy Press,
Washington, D.C. (1986).
5. D.L. Heavner, et al, "A Test Chamber and Instrumentation for the Analysis of
Selected Environmental Tobacco Smoke (ETS"< Components," Proc. 79 Annual APCA
meeting (1986) 86-37.9.
6. P.R. Nelson, D.L. Heavner, B.B. Collie, "Characterization of the Environmental
Tobacco Smoke Generated by Different Cigarettes," in Present and Future of Indoor
Air Quality. Excerpta Medica Int. Cone. Ser. 860:277 (1989).
7. F.A. Thome, et ai, "Environmental Tobacco Smoke Monitoring with an Atmospheric
Pressure Chemical lonization Mass Spectrometer Coupled to a Test Chamber," Proc.
79ffi Annual APCA Meeting (1986) 86-37,6.
8. M.W. Ogden, et al, "Improved Gas Chroma tographic Determination of Nicotine in
Environmental Tobacco Smoke," Analyst 114:1005 (1989).
9. M.W. Ogden and K.C. Maiolo, "Collection anc Determination of Solanesol as a Tracer
of Environmental Tobacco Smoke in Indoor Air," Environ. Sci. Technol. 23:1148.
(1989).
10. J.M. Connor, G.B. Oldaker, and JJ. Murphy, 'Method for Estimating the Contribution
of Environmental Tobacco Smoke to Respirable Suspended Particles," Environ.
Technol. 11:189 (1990).
11. John W. Drake, Transportation Consultant, W. Lafayette, IN, private communication
(1990).
555
-------
POLYCYCLIC AROMATIC COMPOUND CONCENTRATIONS IN
RESIDENTIAL AIR ASSOCIATED WITH CIGARETTE SMOKING
AND GAS OR ELECTRIC HEATING AND COOKING SYSTEMS
Jane C. Chuang, Gregory A. Mack, and Michael R. Kuhlman
Battelle
Columbus, Ohio
Nancy K. Wilson
U.S. EPA, Atmospheric Research and Exposure
Assessment Laboratory
Research Triangle Park, North Carolina
A small field study was conducted in Columbus, Ohio, during the winter
of 1986-1987. Eight homes were selected for sampling on the basis of the
following characteristics: electric/gas heating system, electric/gas cooking
appliances, and absence/presence of environmental tobacco smoke (ETS). A 224
L/min sampler developed by Battelle was used indoors, and a PS-1 sampler was
used outdoors. Sampler modules used both indoors and outdoors consisted of a
quartz fiber filter in series with an XAD-4 trap to collect particles and
semivolatile organic compounds. We measured 28 polycyclic aromatic hydrocar-
bons (PAH) and derivatives. The most abundant PAH found indoors and outdoors
was naphthalene; the least abundant PAH was cyclopenta[c,d]pyrene. Higher
average indoor concentrations of all but three target compounds were found
when compared to outdoor levels. The presence of ETS in residential indoor
air was identified as the most significant influence on indoor PAH and PAH
derivative levels. Quinoline and isoquinoline concentrations correlated very
well with nicotine concentrations and can be used as markers for indoor
levels of ETS.
556
-------
I. INTRODUCTION
Environmental tobacco smoke (ETS), which consists of sidestream smoke
and exhaled mainstream smoke, has been considered as an important component
of indoor air pollution. Many studies have shown that ETS can substantially
increase the levels of mutagenicity, particulate matter, and polycyclic
aromatic hydrocarbons (PAH) in indoor air (1-4). In most of the studies,
indoor PAH concentrations were measured in test rooms with either smoking
machines or smokers; the effects of ETS on indoor PAH levels in typical
residences have not been fully investigated. The effects of ETS on residen-
tial indoor levels of PAH derivatives including nitro-PAH (N02-PAH),
oxygenated PAH (OXY-PAH), and nitrogen heterocyclic compounds have not been
reported. Among the PAH and their derivatives found in air, many compounds
are mutagens or carcinogens (2,3,5,6). Human exposure to these pollutants in
indoor air is of increasing concern because we spend approximately 80 percent
of our time indoors (7).
A small pilot field study was conducted in eight homes in Columbus,
Ohio, during the winter heating season of 1986/87 using the sampling and
analysis methodology developed in our previous studies (9-19). The objective
was to investigate the influence of several indoor pollution sources on the
indoor pollutant levels. The indoor pollution sources included ETS, gas or
electric furnaces, and gas or electric stoves. We measured 15 PAH, five N02-
PAH, five OXY-PAH, and three nitrogen heterocyclic compounds.
II. EXPERIMENTAL METHODS
Sampling Methods
The homes were selected to give a large range of the number of ciga-
rettes smoked. In addition, homes were selected so that nearly equal numbers
of samples could be taken from each available combination of heating system
and cooking appliance. At each home, two sequential 8-hr indoor samples were
taken in the kitchen and living room using the quiet sampler (17). Single
16-hr outdoor samples were taken concurrently using a PS-1 sampler (General
Metal Works, Cleves OH). Questionnaires describing the house characteris-
tics, the ongoing activities in the house, and the sampling conditions were
completed for each house by the field technician and the resident.
The sampling module consisted of a quartz fiber filter (104-mm QAST,
Pall flex, Putnam CT) and XAD-4 (Supelco, Bellefonte PA) in series to collect
both particle-bound and vapor-phase target compounds. We used XAD-4 in
^reference to XAD-2 in this study because of its better collection efficiency
for nicotine (16). The detailed sampling procedures are described elsewhere
(12). The air exchange rate of each house was determined from the gas
chromatography/electron capture detector (GC/ECD) time profile of injected
sulfur hexafluoride.
Analytical Method
The corresponding XAD-4 and filter samples were combined and extracted
with dichloromethane (DCM) for 16 hr, then extracted with ethyl acetate (EAC)
557
-------
for an additional 8 hr. The DCM and EAC extracts were combined and con-
centrated for subsequent chemical analysis.
Sample extracts were analyzed by gas chromatography/mass spectrometry
(GC/MS) with positive chemical ionization (PCI) to determine PAH, quinoline,
isoquinoline, and nicotine and by GC/MS with negative chemical ionization
(NCI) to determine N02-PAH and OXY-PAH. A Finnigan 4500 quadrupole GC/MS
system and an INCOS 2300 data system were employed. Methane was the reagent
gas for both PCI and NCI, and compounds were ionized by a 150 eV electron
beam. Peaks of target compounds monitored were the protonated molecular ion
peaks for the PCI method and molecular ion peaks for the NCI method (12).
III. Results
The concentrations of the individual PAH measured ranged from 0.12
ng/m3 to 4,200 ng/m3. The most abundant PAH found in both indoor and outdoor
air was naphthalene, and the least abundant PAH was cyclopenta[c,d]pyrene.
The levels of N02-PAH and OXY-PAH were lower than those of their parent PAH.
The highest indoor concentrations for all PAH and most of their derivatives
were from home 5 during the living room sampling period (during which the
highest number of cigarettes, a total of 20, was smoked).
When comparing the average concentrations in the three types of homes,
we see that homes occupied by smokers had higher indoor concentrations of
most target compounds than those from the same type of homes occupied by non-
smokers. Thus, the presence of ETS appears to increase indoor levels of PAH
and other pollutants.
Comparing the indoor concentrations over three types of homes occupied
by nonsmokers, the homes having gas heating and cooking appliances had the
highest average concentrations of most PAH compounds, followed by homes
having gas heating and electric cooking appliances. Homes having electric
heating and cooking appliances had the lowest concentrations of most PAH.
However, the concentrations of PAH derivatives did not follow the same trend
as those of PAH.
Higher average indoor levels were observed compared to the average
outdoor levels for all but three PAH or derivatives (naphthalene, pyrene
dicarboxylic acid anhydrides, and 2-nitrofluoranthene). For 2-nitrofluor-
anthene, the average outdoor level (0.06 ng/m3) was only slightly higher than
the indoor level (0.05 ng/m3); for the other two PAH dicarboxylic acid
anhydrides, the outdoor levels were approximately twice the indoor levels.
In another study (13), we demonstrated that naphthalene and pyrene
dicarboxylic acid anhydrides can be formed through the oxidation reactions of
acenaphthylene and cyclopenta[c,d]pyrene with ultraviolet radiation or ozone.
Therefore, higher outdoor concentrations of these compounds may arise from
atmospheric reactions since higher ozone concentrations and ultraviolet
radiation are expected outdoors. For the remaining target compounds, rela-
tively lower outdoor concentrations are expected because there were no
stationary contamination sources nearby, and local traffic was low in these
residential areas.
558
-------
Nicotine is a major component in ETS and has been used as a marker for
[ITS (20,21). Due to the large concentration ranges of nicotine among
smokers' and nonsmokers1 homes, we had to dilute some sample extracts and
reanalyze them. We investigated the relationship among nicotine, quinoline,
and isoquinoline to determine whether or not quinoline or isoquinoline can be
used to monitor ETS more easily. Both quinoline and nicotine are found in
cigarette smoke condensate. Quinoline has been reported to be the most
abundant aza-arene in cigarette smoke and to be present at higher levels in
sidestream smoke than in mainstream smoke (22). We used all the measured
concentration levels across eight homes and three sampling locations to
obtain a matrix of 75 Pearson correlation coefficients. The estimated
correlation coefficients between nicotine and quinoline as well as nicotine
and isoquinoline are 0.96 and 0.97, respectively. These results demonstrated
that indoor concentrations of quinoline and isoquinoline correlate very well
with those of nicotine. Based on this finding, we recommend the use of
quinoline or isoquinoline instead of nicotine as an ETS marker for future
field studies.
IV. Conclusions
The following conclusions can be drawn "rorn this study:
(1) The presence of ETS was the most significant contributor
to indoor levels of PAH and most PAH derivatives.
(2) Homes with gas heating systems; were associated with
higher pollutant levels, but the effect was not as
important as the effect of ETS.
(3) Quinoline and isoquinoline can be used as a marker for
nicotine to measure residential indoor ETS levels.
(1)
(2)
(3)
(4)
References
Grimmer, G., Boehnke, H., and Harke, H, P. (1977), Passive Smoking
Intake of Polynuclear Aromatic Hydrocarbons by Breathing of Cigarette
Smoke Containing Air, Int. Arch. Occup.. Environ. Health, 40, 93-100.
Lewtas, J., Goto, S., Williams, K., Chuang, J. C., Petersen, B. A., and
Wilson, N. K. (1987), The Mutagenicity of Indoor Air in a Residential
Pilot Field Study: Application ar-d Evaluation of New Methodologies,
Atmos. Environ.. 21 (2), 443-449.
Salomaa, S., Tuominen, J., and Skytta, E. (1988), Genotoxicity and PAC
Analysis of Particulate and Vapor Phaseis of Environmental Tobacco
Smoke, Mutat. Res.. 204 (2), 173-183.
Chuang, J. C., Mack, G. A., Koetz, J. F:., and Petersen, B. A. (1985),
Pilot Study of Sampling and Analysis for Polynuclear Aromatic Compounds
in Indoor Air, N. K. Wilson, Project Officer, Report, U.S. Environ-
mental Protection Agency, Research Triangle Park, NC, EPA/600/4-86/036.
559
-------
(5) Arey, J., Zielinska, B., Atkinson, R., and Winer, A. M. (1987), Poly-
cyclic Aromatic Hydrocarbons and Nitroarene Concentrations in Ambient
Air During a Wintertime High N02 Episode in the Los Angeles Basin,
Atmos. Environ.. 21 (6), 1437-1444.
(6) Motykiewicz, G., Michalska, J., Szeliga, J., and Cimander, B. (1988),
Mutagenic and Clastogenic Activity of Direct-Acting Components from Air
Pollutants of Silesian Industrial Region, Mutat. Res.. 204 (2),
289-296.
(7) National Research Council (1981), Indoor Pollutants, National Academy
Press, Washington DC.
(8) Chuang, J. C., Mack, G. A., Mondron, P. J., and Petersen, B. A. (1984),
Development of Sampling and Analytical Methodology for Polynuclear
Aromatic Compounds in Air, N. K. Wilson, Project Officer. Report, U.S.
Environmental Protection Agency, Research Triangle Park, NC,
EPA/600/4-85/065.
(9) Wilson, N. K. and Petersen, B. A. (1984), Evaluation of Sampling and
Analysis Methodology for Polynuclear Aromatic Compounds in Indoor Air,
presented at the 1984 International Chemical Congress of Pacific Basin
Societies, Honolulu, HI.
(10) Chuang, J. C., Mack, G. A., Petersen, B. A., and Wilson, N. K. (1986),
Identification and Quantification of Nitro Polynuclear Aromatic Hydro-
carbons in Ambient and Indoor Air Particulate Samples, in Polynuclear.
Aromatic Hydrocarbons: Chemistry. Characterization, and Carcinggene-
sis. Battelle Press, 155-171, Columbus, OH.
(11) Chuang, J. C., Hannan, S. W., and Wilson, N. K. (1987), Field Com-
parison of Polyurethane Foam and XAD-2 Resin for Air Sampling for
Polynuclear Aromatic Hydrocarbons, Environ. Sci. Technol., 2_1, 798-804.
(12) Chuang, J. C., Kuhlman, M. R., Hannan, S. W., and Bridges, C. (1987),
Evaluation of Sampling and Analysis Methodology for Nicotine and
Polynuclear Aromatic Hydrocarbons in Indoor Air, N. K. Wilson, Project
Officer, Report, U.S. Environmental Protection Agency, Research
Triangle Park, NC, EPA/600/4-87-031.
(13) Chuang, J. C., Hannan, S. W., and Slivon, L. E. (1988), Chemical
Characterization of Polynuclear Aromatic Hydrocarbon Degradation
Products from Sampling Artifacts, N. K. Wilson, Project Officer,
Report, U.S. Environmental Protection Agency, Research Triangle Park,
NC, EPA/600/5-487/039.
(14) Chuang, J. C.( Mack, G. A., Stockrahm, J. W., Hannan, S. W., Bridges,
C., and Kuhlman, M. R. (1988), Field Evaluation of Sampling and
Analysis for Organic Pollutants in Indoor Air, N. K. Wilson, Project
Officer, Report, U.S. Environmental Protection Agency, Research
Triangle Park, NC, EPA/600/4-88/028.
560
-------
(15) Chuang, J. C., Holdren, M. W., Kuhlman, M. R.r and Wilson, N. K.
(1989), Methodology of Indoor Air Monitoring for Polynuclear Aromatic
Hydrocarbons and Related Compounds, Proceedings of the 1989 Inter-
national Symposium on Measurement of Toxic and Related Air Pollutants,
Pub. VIP-13, AWMA, Pittsburgh, PA, pp 495-502.
(16) Chuang, J. C., Kuhlman, M. R., and Wilson, N. K. (1990), Evaluation of
Methods for Simultaneous Collection and Determination of Nicotine and
Polynuclear Aromatic Hydrocarbons in Indoor Air, Environ. Sci.
Techno!.. 24, 661-665.
(17) Wilson, N. K., Kuhlman, M. R., Chuang, J. C., Mack, G. A., and Howes,
J. E., Jr. (1989), A Quiet Sampler for the Collection of Semivolatile
Organic Pollutants in Indoor Air, Environ. Sci. Techno!.. 23,
1112-1116.
(18) Wilson, N. K. and Chuang, J. C. (1990), Indoor Levels of PAH and
Related Compounds in an Eight-Home Pilot Study, in Polynuclear Aromatic
Hydrocarbons. Battelle Press, Columbus, OH, in press.
(19) Wilson, N. K., Chuang, J. C., Kuhlman, M. R., and Mack, G. A. (1990),
Measurement of Polycyclic Aromatic Hydrocarbons and Other Semivolatile
Organic Compounds in Indoor Air, Proceedings of the EPA/AWMA Inter-
national Symposium on Total Exposure Assessment Methodology, AWMA,
Pittsburgh, PA, in press.
(20) Eatough, D. J.f Benner, C. L., Bayona, J. M., Caka, F. M., Mooney,
R. L., Lamb, J. D., Lee, M. L., Lewis, E. A., Hansen, L. D., and
Eatough, N. L. (1987), Identification of Conservative Tracers of
Environmental Tobacco Smoke, Proceedings of the 4th International
Conference on Indoor Air Quality and Climate, Berlin, West Germany, pp
3-7.
(21) Muramatsu, M., Umemura, S., Okada, T., and Tomita, H. (1984), Estima-
tion of Personal Exposure to Tobacco Smoke with a Newly Developed
Nicotine Personal Monitor, Environ. Re;.., 35, 218-227.
(22) Dong, M., Schmeltz, I., Jacobs, E., and Hoffman, D. (1978), Aza-Arenes
in Tobacco Smoke, J. Anal. Toxicol.. 2 (1), 21-25.
561
-------
COMPARISON OF AREA AND PERSONAL SAMPLING METHODS FOR
DETERMINING NICOTINE IN ENVIRONMENTAL TOBACCO SMOKE
William E. Grouse
Lorillard Tobacco Company, Research Center,
420 English Street, Greensboro, NC 27405 U.S.A.
Guy B. Oldaker III
R.J. Reynolds Tobacco Company, Bowman Gray Technical
Center, Winston-Salem, NC 27102 U.S.A.
Nicotine in environmental tobacco smoke (ETS) was determined with
portable air sampling systems (PASS's) and personal sampling devices to assess
whether the two different methods give comparable estimates of exposure.
Samples were collected with each method in 21 restaurants. For both methods
nicotine is collected on XAD-4 resin and determined by gas chroinatography with
nitrogen selective detection. The mean nicotine concentrations determined by
the PASS'S and personal sampling devices were 6.3 and 4.3 jLtg/in , respectively.
Results from a paired t-test and the Wilcoxon Signed Rank Test show that the
PASS values are significantly higher than those of the personal sampling
device: P = 0,02 and P = 0.01, respectively.
Introduction
The assessment of exposure to environmental tobacco smoke (ETS) is
constrained by the absence of an easily determined, reliable indicator and by
the need to collect samples unobtrusively. The portable air sampling system
(PASS) represents one methodology used for assessing exposure to ETS (1). The
PASS is used unobtrusively to determine both particle and gas phase indicators
of ETS, the latter including nicotine. The PASS has been used in numerous
surveys of ETS in several environmental categories including, for example,
offices (2), restaurants (1, 2), and passenger cabins of commercial aircraft
(3).
Because of obvious practical constraints, the PASS is applied as an area
sampling method. Results from the PASS, therefore, might not be
representative of personal exposures, which would be determined by use of
personal sampling methods. The present study focuses on the determination of
exposure to ETS nicotine. The objective of the study was to evaluate the
comparability of results from determinations of ETS nicotine by the PASS and
by personal sampling.
Experimental
Environmental Category
Samples were collected during September through December of 1989 in
restaurants located in Greensboro, North Carolina. The restaurants represent
a subset of a stratified, random population. The original population,
selected for a survey performed in 1988 (1), was identified from the Dun and
Bradstreet's Electronic Yellow Pages data base. Stratification was performed
on the basis of restaurant styles including: "American Style, Fast Food,
562
-------
Ethnic, and Other." This last designation includes, for example, "Barbecue
Restaurants, Soda Shops, and Pizza Establishments," Two criteria observed for
selecting restaurants were that seating be available for at least 50 diners
and that there be no restrictions on smoking,. The sample population for the
1989 survey had fewer restaurants than that of the 1988 survey because some
restaurants went out of business and because some implemented smoking
restrictions.
Sampling Locations
Sampling locations within restaurants were selected to conform as
closely as possible to the guidelines described by Nagda and Rector (4) with
the added constraint that samples be collected as close as possible to the
same locations as for the 1988 survey. Samples were obtained with the PASS
from locations between 0.6 and 1.8 ra above the floor and at least 0,6m from
walls and other surfaces.
Sampling Procedures
All samples were collected for a minimum of one hour during the expected
peak times for lunch. These times ranged from 11:30 a.m. to 1:00 p.m. and
corresponded to those times when samples were collected for the 1988 survey.
Area samples for nicotine were obtained with the PASS. The nicotine
collection system includes a constant flow sampling pump (SKC Inc., Eighty
Four, PA) operated at a flow rate of 1 L/min, that is connected with rubber
tubing to a sorbent tube containing XAD-4 resin (SKC Inc.). The inlet of the
sorbent tube projects approximately 2 cm bey2nd the surface of the briefcase.
Personal samples for nicotine were collected with essentially the same
equipment as used by the PASS. A constant flow sampling pump operating at 1
L/min was worn on the belt of the 5;ampl Lng team member, a rubber tube
connected to the sorbent tube holder, and the XAD-4 sorbent tube was
positioned within the breathing zone. With the exception of the XAD-4 sorbent
tube and its holder, all other components of the personal sampling device were
concealed under clothing.
In each restaurant, duplicate samples were collected concurrently with
PASS'S and personal sampling devices. During, sample collection, sampling team
members ate lunch and refrained from smoking.
Sampling pumps were calibrated with a film flow meter (Gilibrator,
Gilian Inc.) before and after samples were collected. The mean flow rate was
used to compute air volumes for those samples where calibration results showed
a deviation less than 5%. For five s;amples, deviations greater than this
value were observed; for these results the lower flow rate was used to compute
sample volume, thus, giving an upper limit for the nicotine concentration,
Analytical Procedure
Nicotine was analyzed by the method described by Ogden et al. (5). XAD-
4 sorbent tubes were extracted with ethyl acetate containing 0.01% (v/v)
triethylamine. Analysis was performed by gas chromatography with nitrogen
selective detection. Front and back sections of the sorbent tubes were
analyzed separately to confirm the absence of breakthrough, of which none was
observed.
563
-------
RESULTS AND DISCUSSION
Table I presents the results from the determinations of nicotine with
area sampling devices and personal sampling devices, which are identified
"PASS" and "Personal," respectively. In addition, area and personal sampling
results are identified by either "A" or "B." These designations show the
pairing of sampling devices with each team member. Concentrations are in
units of jUg/m . Eight field blanks were analyzed with the PASS and personal
samples; nicotine was not detected. One sample, associated with restaurant
number 12 and PASS B, was lost.
Table II presents descriptive statistics for the nicotine concentration
data. These statistics include arithmetic means, medians, minimum and maximum
values, and number of samples. The arithmetic mean concentrations determined
by the PASS's are greater than the corresponding concentrations determined by
the personal sampling devices. With results pooled by sampling device, the
mean concentration associated with the PASS'S is 2.0 Mg/m greater (46%) than
that of the personal sampling devices.
Median nicotine concentrations are less than corresponding arithmetic
mean concentrations, a relationship indicating that the distribution of the
data is skewed toward higher concentrations. The median concentration results
presented in Table II show a relationship between the PASS and the personal
sampling devices that is quantitatively similar to that shown by the
arithmetic mean concentration data. Thus, the median concentration determined
•3
by the PASS's is 1.3 )lig/m greater (45%) than the median of the personal
sampling devices.
The concentration ranges determined by the PASS's and personal sampling
devices are similar. Concentration values span three orders of magnitude
ranging from 0.3 /ig/m , just above the limit of detection, to approximately
25 /ig/ro3.
Statistical tests were performed to assess whether differences between
concentrations determined by the PASS and personal sampling methods are
significant. Data for each method and restaurant were averaged, because the
sampling methods could not reasonably be assumed to be independent. (This
approach is statistically more conservative with respect to detecting
differences between the sampling methods.) Because environmental data are
typically distributed log normally (6) , Shapiro-Wilkes tests were done on raw
and log transformed concentration data from the PASS and personal sampling
devices. Results of these tests support a log normal distribution: for the
PASS raw data, P < 0.01; for the personal raw data, P < 0.01; for the log
transformed PASS data, P - 0.41; and for the log transformed personal data,
P - 0.97.
A paired t-test was applied to the log transformed data to test for
differences between the PASS and personal sampling methods. Results show that
mean nicotine concentrations determined by PASS's are significantly greater
than mean concentrations determined by personal sampling devices (P = 0,02).
The same overall result is obtained when the data are evaluated with the
Wilcoxon Signed Rank Test, a nonparametric statistical test. Thus, the
results of this test show a statistically significant difference between mean
concentrations determined by the PASS and personal sampling devices
(P - 0.01).
Currently, we can only speculate on why there is a difference between
the concentrations measured by the PASS and the personal sampling devices.
We hypothesize that this difference reflects depletion of ETS nicotine at the
564
-------
breathing zone relative to nicotine in the general air space sampled by the
PASS. Adsorption of nicotine by fabrics and. other surfaces at the breathing
zone, as well as removal by human respiration, might account for such
depletion,
CONCLUSIONS
Results from this investigation show that concentrations of ETS nicotine
determined with the PASS, an area sampling device, are significantly greater
than concentrations determined with personal sampling devices. This observed
relationship indicates that results provided by the PASS in ETS surveys are
conservative in their estimation of personal exposure to ETS nicotine.
Additional investigations need to be done to assess the relationships
between PASS and personal determinations of other ETS indicators.
ACKNOWLEDGMENTS
The authors thank L. L. Van Meter, C. S. Williard, G. D. Brown, D. A,
Williams, L. H. Gains, R. M. Striegel, Jr., V. G. Garrard, S. S. Miller, and
M. M. Dozier for collecting samples and Fred W. Conrad, Jr. for performing the
analyses of nicotine. In addition, we are grateful to Ross M. DePinto, Walter
T. Morgan, and P. 0. DeLuca for their efforts in connection with the
statistical analysis of the data.
REFERENCES
1. W. E. Grouse, M. S. Ireland, J. M. Johnson, R. M. Striegel, Jr., C. S.
Williard, R. M. DePinto, G. B, Oldaker III, R. L. McBride, "Results
from a Survey of Environmental Tobacco Smoke (ETS) in Restaurants,"
Combustion Processes and the Quality of the Indoor Environment, J. P.
Harper, Ed., Air & Waste Management Association, Pittsburgh, PA, 1989,
pp. 214-222.
2. G. B. Oldaker III, P. F. Perfetti, F. W. Conrad, Jr., J. M. Conner,
R. L. McBride, "Results from Surveys of Environmental Tobacco Smoke in
Offices and Restaurants," Indoor Air Quality (Int. Arch. Occup.
Environ. Health Suppl.). H. Kasuga, Ed., Springer-Verlag, Berlin, 1990,
pp. 99-104.
3. G. B. Oldaker III and F. W. Conrad, Jr., "Estimation of the Effect of
Environmental Tobacco Smoke on Air Quality in Passenger Cabins of
Commercial Aircraft," Environ. Sci. Technol. 21: 994 (1987).
4. N. L. Nagda and H. E. Rector, Guidelines for Monitoring Indoor Air
Quality. EPA-600/4-83-046, Sept. 1983.
5. M. W. Ogden, "Gas Chromatographic Determination of Nicotine in
Environmental Tobacco Smoke: Collabora.tive Study," J. Assoc. Off. Anal.
Chem. 72: 1002 (1989).
6. R. 0. Gilbert, Statistical Methods for Environmental Pollution
Monitoring. Van Nostrand, NY, 1987.
565
-------
Table I.
Restaurant
Number
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
•A
Sample lost.
Results from Determinations of Nicotine with Area and
Personal Sampling Devices in Restaurants
•D
Nicotine Concentration. Ug/m
PASS
3
4
2
3
5
13
17
5
7
3
0
2
4
5
7
1
3
10
1
2
15
A
.7
.5
.3
.6
.9
.2
.8
.6
.0
.8
,4
.8
.2
.9
.6
.4
.8
.8
.2
.9
.9
1
4
1
3
5
16
17
2
24
3
0
-
4
6
9
1
4
7
1
2
13
B
.7
.5
.3
.3
.3
,5
.5
.2
.8
.8
.3
*
.8
.2
.8
.7
.0
.3
.3
.7
.8
4
1
1
0
2
9
8
5
24
1
2
0
4
1
5
1
3
6
1
2
13
PERSONAL
A
.4
.4
.0
.3
.9
.7
.6
.2
.0
.2
.1
.3
.2
.8
.6
.5
.0
.1
.3
.6
.4
2
3
0
0
4
9
9
5
13
2
0
0
1
1
7
2
3
4
1
1
4
B
.5
.0
.5
.5
.1
.3
.3
.6
.0
.9
.8
.9
.9
.2
.6
.9
.4
.1
.3
.9
.2
Table II.
Mean
Median
Minimum
Maximum
Descriptive Statistics for Area and Personal Sampling
Results for Nicotine.
Nicotine Concentration. UE/m
PASS
A B Pooled
5.9 6.6 6.3
4.2 4.2 4.2
0.4 0.3 0.3
17.8 24.8 24.8
Personal
A B Pooled
4.8 3.9 4.3
2.9 2.9 2.9
0.3 0.5 0.3
24.0 13.0 24.0
N
21
20
41
21
21
42
566
-------
THE IMPACT OF CIGARETTE SMOKING ON
INDOOR AEROSOL MASS AND ELEMENTAL CONCENTRATIONS
B.P. Leaderer
Yale University, John B. Pierce Foundation
New Haven, CT 06519
P. Koutrakis, and S.L.K. Briggs
Harvard University, School of Public Health
Boston, MA 02115
J. Rizzuto
New York State Energy Research
and Development Authority
Albany, NY 12223
An indoor air quality study was conducted in two New York
counties during the period of January 6 through April 15,1986.
Suffolk County is located on eastern :Long Island, east of New
York City. Onondaga County is situated in northwestern New York
State and includes the city of Syracuse. Week long fine particle
mass samples were collected indoors and outdoors of 394 homes.
The homes were selected according to their potential indoor
aerosol sources such as cigarette smoke, gas stoves, and heating
sources. In this paper two home groups were selected to examine
the impacts of cigarette smoke on the fine aerosol mass and
elemental concentrations. These two data sets were created by
removing all samples taken in homes with one or more indoor
sources such as kerosene heaters, wood stoves, or fire places.
For each of the counties, the remaining homes were divided into
two groups: homes with smokers and homes without smokers. A
simple comparison of average mass and elemental concentrations
between smoking and non-smoking homes in each county enabled us
to determine the elemental profile of cigarette smoke. Finally,
the room to room variations in indoor concentrations were
studied. The results of this analysis showed that all elements,
including those associated with cigarette smoke, presented the
same concentrations in the kitchen, living room, or other living
areas.
567
-------
Introduction
During the period of January 6 and April 15, 1986, an
extensive indoor air quality program was conducted in Onondaga
County and Suffolk County, in New York State. Suffolk County is
located on eastern Long Island, east of New York City. Onondaga
County is situated in northwestern New York State and includes
the city of Syracuse. Week long fine particle mass samples were
collected indoors in a total of 394 homes. The homes were
selected according to their potential indoor aerosol sources such
as cigarette smokers, gas stoves, wood burning, and kerosene
heaters. Comparisons of indoor aerosol mass and elemental
concentrations between homes with and homes without smokers will
allow us to determine the elemental profiles of cigarette smoke
emissions and determine their relative importance. Also in this
paper we examine whether room to room differences in elemental
concentrations exist. These comparisons included elements
associated with cigarette smoke and elements associated with
other indoor and outdoor sources.
Sampling and Analysis
The fine particle mass measurements were obtained using the
Harvard Impactor which collects fine particles with aerodynamic
diameters < 2.5nm at a flow rate of 41pm 3.. A silent pumping
unit was specially designed for indoor use. For single sample
homes, the optimal sampling duration was 168 hours, or seven
days. For homes with more than one sampling location, a time
share unit with a solenoid system was installed on the pumping
system to sample alternately for a fifteen minute period at each
home location. Thus, these filters sampled for a total of 84
hours, approximately.
The fine particle mass collected on Teflon filters was
gravimetrically determined using a Cahn balance. Subsequently,
concentrations of elements associated with fine mass were
determined by X-ray fluorescence (XRF): Al, Si, S, Cl, K, Ca, Ti,
V, Cr, Mn, Fe, Ni, Cu, Zn,.As, Se, Br, Sn, Sb, Ba, La, Cd, Pb.
Among the above elements, Al, Ti, Cr, Sn, Sb, Ba, and La,
presented a high percentage of concentrations below the detection
limits; therefore they will not be examined.
In suburban Suffolk County, sampling was conducted form
January 6 to February 24, 1986 in 194 residences resulting in 260
indoor and 20 outdoor samples. Indoor monitoring occurred in the
living area of 187 homes. Monitoring commenced in approximately
five different homes each day for an optimal duration of seven
consecutive days throughout the study period. A subgroup of 73
homes had additional samplers placed in several rooms to examine
568
-------
the room to room variation in aerosol mass and elemental
concentrations. Onondaga County, a suburban-rural area, was
sampled between February 24 and April 15,1986. A total of 279
indoor and 37 outdoor samples were taken under the same sampling
clesignin 200 residences. A subgroup of 83 homes had samples
measured in rooms other that the living area.
Telephone interviews were conducted to poll prospective
participants for the study. Questions concerning the presence
a.nd typical usage of indoor air pollution sources such as gas
stoves, wood stoves or fireplaces, kerosene heaters, and smokers
allowed for a preliminary stratification of the homes using a
predetermined minimal source usage level. Information was also
gathered on housing characteristics such as volume and air
exchange rates. A diary of actual source usage such as minutes
of gas stove use, total cigarettes smoked, and hours of kerosene
burner, wood stove or fireplace use was kept by the participants
in order to reclassify the homes based on actual source usage,
again using minimal source usage rates;. A comparison of the
original strata and the actual strata suggests 49% of the Suffolk
homes and 55% of the Onondaga homes we're misclassif ied when
relying solely on the telephone interview. This shows the
importance of gathering actual source usage information for the
specific sampling period. Although a predetermined minimal
source usage level is appropriate for screening participants,
when describing and assessing the impact of a source on the
indoor environment an absolute criteria is necessary. Thus, for
the purpose of this paper, any use of a source requires inclusion
in that source category. Table 1 gives the mean and standard
deviation of these source usage variables for Suffolk and
Onondaga Counties. The table shows that more cigarettes were
smoked in Suffolk, yet cigars and pipes were smoked more in
Onondaga. Table 1 also gives a summary of the air exchange rates
and house volumes for all homes. The average air exchange rates
were similar for both counties however Suffolk homes showed more
variability, cf=.42 versus o=.21 per hour. House volumes were
comparable on average yet the standard deviations of these
measurements showed Onondaga almost twice as variable as Suffolk,
suggesting varied housing stock.
Most sampling occurred in the living area of each home,
defined as the room where the families spent most of their time.
In addition, kitchens were monitored in 58 Suffolk homes and 62
Onondaga homes. In some instances, the family room, living room,
or some other room was also sampled, however, these homes also
had samples collected in the main living area. Lastly 19 valid
outdoor samples were measured in Suffolk and 36 in Onondaga.
These data were analyzed for room to room variation in aerosol
concentrations and presented below. However, in describing the
569
-------
distribution of elemental concentrations for various source
categories, only the living area samples were included so as not
to bias the estimates toward the multiple sample homes, and also
because the room to room variation analysis supported this
approach as shown below.
Results and Discussion
To investigate differences in fine aerosol mass and
elemental concentrations between homes with and without smokers,
a subset of homes was selected. This subset includes homes, from
both counties, in which major indoor sources such as kerosene
heaters, wood stoves and fireplaces are not present.
Subsequently, these two data sets, one per county, were further
separated into groups. The first group includes homes without
smokers, and the second includes homes with smokers. Note that
both smoking and non-smoking home groups include gas stove use,
since previous analysis of this data has shown that gas stove use
did not contribute to indoor aerosol mass and elemental
concentrations. Thus this source was not considered for the
above home classification. Tables 2a and 2b depict the
calculated geometric mean and standard deviation of mass and
elemental concentrations for non-smoking homes, smoking homes,
and outdoors, for Suffolk and Onondaga Counties, respectively.
For both counties, the average indoor fine mass
concentration in no smoking homes and the outdoor concentration
exhibit similar levels. Since not all outdoor soil particles
penetrate the indoor environment, these results suggest the
presence of indoor sources. In fact, some activities such as
showering, vacuuming, cooking, or biological aerosols generated
indoors could contribute to indoor fine mass. Furthermore, for
both counties, fine mass concentrations were considerably higher
in homes with smokers than homes with no smokers. For Suffolk
and Onondaga the respective concentrations were 17.3 versus 49.3
Hg/m3 and 14.4 versus 36.5 pg/m3, respectively. Thus, homes with
smokers present mass concentrations which are approximately three
times higher than homes with no smokers.
The results of Tables 2a and 2b suggest that chloride and
potassium concentrations in homes with no smokers are associated
with indoor as well as outdoor sources. For non-smoking homes,
chloride levels are approximately two times higher than the
outdoor concentration. High chloride concentrations in
non-smoking homes have been observed in previous indoor air
quality studies and were attributed to the use of consumer
products 2'3. For homes with smokers, the chloride levels were
much higher, five to eight times the outdoor concentration.
Moreover, for non-smoking homes, indoor potassium concentrations
570
-------
were found to exhibit similar levels to those observed outdoors
in both Suffolk and Onondaga County, as shown by Tables 2a and
2b, respectivley. Potassium is mostly associated with soil
particles in the non-smoking homes. However, for the case of
smoking homes, potassium levels can te five to six times the
outdoor concentration. These results are in good agreement with
findings reported by Santanam and Spengler 4.
Cadmium and to a lesser extent bromine are also associated
with cigarette smoke emissions. Indoor cadmium concentrations
were similar to those observed outdoors, suggesting the presence
of a minor indoor source. However, there is no information about
the existance of indoor cadmium sources in exception to cigarette
smoking 5. In fact, homes with smokers presented concentrations
which are two to three times higher than outdoors, suggesting
that cigarette emissions contain cadmium. Of course, the
contribution of smoking to indoor cadmium concentrations will
depend on, among other things, the number of cigarettes smoked
and the type of tobacco used. Similarly, it has been found that
in homes with no smokers, bromine concentrations were lower than
outdoors suggesting that this element is mostly related to
outdoor sources; that is, automobile emissions. However, for
homes with smokers, bromine showed slightly higher concentrations
than outdoors providing some evidence that indoor bromine is
associated with this source.
Finally, Tables 2a and 2b suggest that the rest of the
analyzed elements are not associated with cigarette smoke in
Suffolk and Onondaga Counties, respectively. Among them S, V,
Mn, Fe, Ni, Se, and Pb, and to a lesser extent Si, Zn, and As are
mostly related with outdoor sources. The rest of the elements,
Ca and Cu have significant indoor sources.
Furthermore, we investigated room to room variation of fine
particle mass and elemental concentrations. More precisely, mass
and elemental concentrations obtained from samples collected at
locations other than the living area, such as kitchen, family
room, or other living space were collocated with samples from the
living area. Since some of the analyzed elements are cigarette
smoke tracers, such as Cl and K, as shown above, the room to room
variation provides us with great insight into the source emission
distribution.
The sample design dictated that the analysis be separated
into kitchen versus living area, and other living space versus
living area with other living space defined as any sample
location other than the kitchen. Mass and elemental
concentrations measured in these rooms were regressed on the
corresponding living area concentrations. The regression results
of mass and all elements are shown in Table 3 for the kitchen and
571
-------
living area analysis. This analysis includes homes with no
smokers for both Suffolk and Onondaga counties. Coefficients
which are not statistically significant are marked with an
asterisk. The regression results suggest that the mass, Si, S,
Cl, K, Ca, V, Mn, Fe, Ni, Cu, Zn, Br and Pb concentrations in the
kitchen can be predicted satisfactorily from those measured in
the living room. For Cd, As, and to a lesser extent, Se, the
regression results are not satisfactory. This is due to the fact
that X-ray fluorescence is unable to determine precisly the
concentrations of these elements in the nanogram level. As can
be seen by these results, fine aerosols are well mixed in the
indoor environment, as shown previously by Ju and Spengler &. Of
course, this happens because the characteristic time of fine
particle mixing within the home is significantly shorter than the
sampling period. Room to room elemental aerosol profile
variations were also examined by comparing living area
concentrations to other living space concentrations. The results
of the regression analysis suggested that elemental
concentrations in other living spaces can be estimated using
those measured in the living area. Again for Cd, Se, and As and
to a lesser extent, Ni and Cu, the correlation coefficients were
very low, reflecting the high analytical uncertainties. These
relationships are similar to those obtained for the
kitchen/living area comparison. Finally, a similar analysis was
conducted for only homes with smokers. The results of these
analyses, shown in Table 4, reveal the same relationship for all
location comparisons, suggesting no need for further analysis by
home category.
Using a nonparametric analysis of variance we again examined
the assumption that fine aerosols are well mixed within a home.
For this test the hypothesis is that the concentrations in the
kitchen are similar to those in the living area. Concentrations
of fine mass and all elements are not significantly different in
the kitchen or living area, at the p=.05 level. The same
hypothesis was posed for living area versus other living space
locations. The results also confirm no significant difference
between these rooms for any element except Se. In conclusion,
the above comparisons reveal that a single sampling location is
enough to assess indoor human exposure to fine aerosols.
Acknowledgements
This work is supported by EPA cooperative agreement No.
CR-814150. The field portion of this study was conducted by the
Research Triangle Institute (RTI) for the New York State Energy
Research Development Authority (NYSERDA). We thank Drs. L.
Sheldon and T. Hartwell of RTI for their considerable cooperation
in providing rapid access to the particulate filter samples and
collected field data.
572
-------
References
1. V.A. Marple, K.L. Rubow, W. Turner, J.D. Spengler, "Low flow
rate sharp cut impactors for indoor air sampling: design and
calibration." Journal of Air Pollution Control Association.
37:1303-1307 (1987).
2. S.D. Colome, J.D. Spengler, S. McCarthy, "Comparisons of
elements and inorganic compounds inside and outside of
residences." Environment Internationa1, 8:197-212 (1982).
3. P. Koutrakis, J.D. Spengler, B.H. Chang, and H. Ozkaynak,
"Characterizing sources of indoor and outdoor aerosols using
PIXE." Nuclear Instruments and Methods in Physics Research B22:
331-336 (1987).
4. S. Santanam, J.D. Spengler, "Source apportionment of indoor
aerosols: a study at two United States cities." 82nd Annual
Meeting of AWMA, Anaheim, CA, June 25-30, 1989.
5. E. Lebret, J. McCarthy, J.D. Spengler, B.H. Chang, "Elemental
composition of indoor fine particles."1 The Fourth International
Conference on Indoor Air Quality and Climate, West Berlin,
GJermany, August 17-21, 1987.
6. C. Ju, and J.D. Spengler, "Room-to-room variations in
concentration of respirable particles in residences." EST 15:592-
(1981).
Table 1 Summary of source usage data and home characteristics.
Variable
Cigarettes
(# smoked)
Cigars
(# smoked)
Pipefuls
(# smoked)
Air Changes
(per hr)
House Volume
N
130
2
4
189
193
SUFFOLK
Mean
113.4
10.5
5.5
0.58
12441
Std
136.1
13.4
5.1
0.42
5852
N
120
3
10
197
200
ONONDAGA
Mean
85.2
20.0
10.1
0.50
12470
Std
90.9
24.3
12.6
0.27
8954
573
-------
Table 2a Geometric means and standard deviations of mass
(pg/m3) and elemental (ng/m3) concentrations in Suffolk County.
Non-Smoking
Element
Mass
Si
S
Cl
K
Ca
V
Mn
Fe
Ni
Cu
Zn
As
Se
Br
Cd
Pb
Mean
17.3
82.8
1224.1
37.4
86.6
39.5
7.2
2.6
33.1
4.5
6.6
27.4
1.1
0.7
10.6
0.7
46.9
N=30
a
1.7
2.1
1.6
2.6
2.2
2.9
1.8
1.6
1.9
1.7
2.1
2.4
2.0
2.1
2.3
2.4
2.2
Smoking
Mean
49.3
90.3
1161.1
197.2
431.5
55.5
7.5
2.4
35.6
3.5
6.8
21.3
1.0
0.5
16.8
1.2
35.3
N=61
a
1.8
3.3
1.7
2.9
2.4
3.5
1.8
1.6
1.6
1.9
2.5
1.9
2.0
2.0
1.8
2.5
2.2
Outdoors
Mean
16.9
88.4
1779.1
21.3
71.3
20.3
12.2
3.5
40.5
7.0
3.2
29.4
1.1
1.4
14.2
0.6
73.2
N=19
a
1.3
1.3
1.4
1.4
1.5
1.5
1.5
1.4
1.4
1.4
1.4
1.6
2.0
1.5
1.8
1.9
1.6
Table 2b Geometric means and standard deviations of mass
3) and elemental (ng/m3) concentrations in Onondaga County,
Non-Smoking
Element
Mass
Si
S
Cl
K
Ca
V
Mn
Fe
Ni
Cu
Zn
As
Se
Br
Cd
Pb
Mean
14.4
117.1
874.5
31.0
64.5
103.1
2.2
2.9
37.1
1.0
5.6
12.0
0.7
0.5
4.0
0.5
15.2
N=45
a
1.7
1.8
1.6
4.1
2.2
3.5
2.2
1.9
1.8
1.8
2.9
1.7
1.8
1.9
1.7
2.9
1.9
Smoking
Mean
36.5
83.4
909.5
117.0
296.2
98.8
3.5
3.2
45.8
0.9
5.0
15.1
0.9
0.4
9.4
1.5
17.2
N=80
a
2.4
2.4
2.7
3.4
3.6
2.3
1.9
1.5
1.5
1.7
2.6
1.4
2.1
2.1
2.0
2.2
1.7
Outdoors
Mean
15.8
122.5
1420.5
19.4
68.7
77.4
3.3
4.6
53.9
1.4
2.0
18.8
1.2
1.4
8.0
0.6
30.4
N=36
a
1.5
1.5
1.4
1.7
1.7
1.9
1.8
1.6
1.6
1.7
1.5
1.5
1.8
1.7
1.8
2.3
1.7
574
-------
Table 3 Regression coefficients of kitchen concentrations
versus living area concentrations (N=98).
ELEMENT
Mass
Si
S
Cl
K
Ca
V
Mn
Fe
Ni
Cu
Zn
As
Se
Br
Cd
Pb
SLOPE
.92
.96
1.02
.84
1.02
.83
.84
.76
.84
1.01
.87
1.14
0.18*
.59
1.08
.11*
1.09
INTERCEPT
6.7
7.6*
30.4*
38.8
28.0*
18.8
1.3
0.9
8.4
0.0*
1.6
-1.9
0.9
-0-3
-0.6*
1.1
-2.3
R2
.79
.94
.95
.76
.89
.98
.74
.78
.87
.97
.91
.94
.04
.38
.94
.01
.97
Table 4 Regression coefficients of kitchen concentrations
versus living area concentrations for smoking homes (N=50).
ELEMENT
Mass
Si
S
Cl
K
Ca
V
Mn
Fe
Ni
Cu
Zn
As
Se
Br
Cd
Pb
SLOPE
.97
.96
.98
.82
.95
.84
.94
.92
.91
1.03
.87
1.01
.13*
.62
1.08
.04*
1.06
INTERCEPT
5.2*
3.8*
52.0
50.5
70.0*
20.4
0.5*
0.3*
4.1
0.0*
1.0
1.0*
1.1
0.2
-0.3*
1.2
-1.2*
R2
.91
.98
.99
.85
.88
.99
.81
.88
.96
.98
.96
.87
.01
.47
.96
.00
.98
*not significant at the p=.05 level
575
-------
CHROMIUM SAMPLING METHOD
By Frank R. Clay
Emission Measurement Branch
Technical Support Division
The method that I am going to describe today is a method that we have
developed for the chromium electroplating industry. It has been called a
"screening method"; however, our present plans are to use it for compliance.
The method is a simplified version of Method 5 that uses a fixed sampling rate
and obtains a proportional sample. It is made from simple and readily
available inexpensive components and can be used by plant personnel.
There are over five thousand plating and anodizing facilities in the
United States, and many are small, family-owned concerns. Many of these firms
are too small to afford conventional testing. A conservative estimate of the
cost of three isokinetic sampling runs is from $3,500 to $5,000 plus travel
expenses. Still there exists a need to quantify chromium emissions from these
sources.
The original goal of this project was to develop an inexpensive
screening method that was ±50 percent accurate. This method was to be
performed at the plant for around $500 or less, and the results could be used
to determine if a conventional test was needed. However, if the screening
results indicated that conventional testing was necessary, additional costs
would be incurred by the plater, and the combined costs of screening plus
conventional testing would place many platers under a financial hardship.
What was really needed was an accurate, simple and inexpensive method that
would determine chromium emissions from electroplaters.
The sampling train that we have developed costs less than $600 to
fabricate and about $350 to perform three sample runs. It samples at a
constant rate (about 0.75 dscfm) and the sampling time at each point is varied
in order to obtain a proportional sample.
This is a slide of the assembled train as used in the field. Stack gas
enters the nozzle and flows to the impingers by way of clear plastic tubing.
From the last impinger, the gas flows to a critical orifice that fixes the
sample rate. After the orifice, the gas goes to the pump and through the dry
gas meter to the atmosphere.
I will now describe the individual components of the train, and we will
begin with the nozzle, liner, and sheath. The nozzle and liner are combined
into one piece and is simply a 1/4-inch I.D. glass or plastic tube with a 90
degree bend. The nozzle end is tapered, which in the case of the plastic tube
was done with a pencil sharpener. The sheath is a piece of 1/2-inch diameter
tubing {PVC or copper pipe will work just as well) that is used to keep the
glass liner from being broken, or in the case of the plastic tubing, to give
rigidity to the apparatus.
576
-------
From the probe assembly, stack gas flows to the impingers. The train
that was used to obtain field samples was assembled using expensive Greenburg-
Smith impingers. The impingers, clamps and connecting glassware cost about
$550. In order to reduce costs, this glassware was replaced with the
impingers shown in the slide. These are made of Mason jars, plastic or glass
tubing, polyester body putty, and silicone rubber cement. The cost of an
impinger using the plastic tubing is about $1.45 for materials while the glass
tubing impinger costs $1 more. It takes about an hour to make each impinger.
If labor cost is $10 per hour, two impingers with plastic tubing would cost
$22.90. It is possible to do the same job with the Mason jar impingers as it
is with the Greenberg-Smith impingers, but for $527 less.
From the impingers, the gas flows through a critical orifice. If a
0.47-atmosphere vacuum is placed on the downstream side of a critical orifice,
flow through the orifice will be at the speed of sound and a constant flow
rate would be obtained. Knowing this, the initial idea of using the critical
orifice was to determine the sample volume without using a dry gas meter.
Laboratory work showed that this concept was not practical, but the orifice
did work well to control the flow through the train. As a result, the
critical orifice is simply a flow control device and sets the sampling rate at
about 0.75 dscfm which is the same sampling rate as our standard isokinetic
sampling train.
The orifice is made of brass tubing that is available in any hobby shop.
A 1/8-inch I.D. piece of tubing 3 inches lorg is cemented (epoxy) into two
larger pieces of brass tubing to make the assembly. Materials cost about
$0.96, and 1/2 hour of labor is required to produce it.
The gas goes from the orifice to a rotary vane pump. The pump is the
most expensive component of the train and ccsts $233.
The next item in the train is the dry gas meter which is used to
determine sample volume. Water displacement and calibrated orifices were also
considered for volume determination, but the dry gas meter seems to be the
easiest and least error-prone method to use. The cost of a dry gas meter is
$180.
One additional item that was going to be used in this method was an
HP-41 CV calculator with a special plug-in module. The module would contain
all the programs necessary to perform the testing and to determine emission
rates. The calculator/module combination was inexpensive and easy to use.
Unfortunately, Hewlett-Packard has discontinued production of the HP-41
calculator, but there are other Hewlett-Packard calculators such as the HP-42S
or the HP-48SX that may be suitable. Since many firms have computers, there
is also the possibility of writing a computer program that will simplify
calculations. While a substitute for the HP-41 has not yet been found, it is
not a major problem and rather than specify a single approach to performing
calculations, several options will be offered instead.
The procedure for taking a sample is pretty simple and should be
relatively easy for the plating facility to perform. This will be facilitated
by a video tape that we are planning to produce.
To perform a test, a site is located arid ports are installed as
specified in Reference Method 1 in Part 60 of the Code of Federal Regulations.
Twenty-four sampling points are located, and a velocity traverse is performed
using Method 2 of Part 60. These data are used to determine the specific
sampling time for each point ffooint velocity -. average velocity) x base
577
-------
time). When this has been done, all the operator has to do to obtain a sample
is to turn on the pump, place the probe at the appropriate point, and sample
for the specified number of minutes and seconds required at that point. This
procedure is repeated until all points have been sampled.
When the sampling has been completed, the sample is recovered by rinsing
the probe and connecting tubing into the first impinger jar. A piece of
plastic wrap is placed over the mouth of the jar, and a standard Mason jar lid
and band are used to seal the sample. The impinger jar now becomes the sample
jar which can be sent to a laboratory for analysis. The laboratory will
analyze the sample for hexavalent chromium using the diphenylcarbazide
colorimetric method. The estimated cost for the analysis is $50.00 per sample
or $150.00 for three sample runs.
Once the laboratory has given the chrome plater the total amount of
hexavalent chromium, concentration and mass emission rates may be determined.
The reasons that make the method work are also its limitations. It
requires ambient temperature, ambient moisture, ambient air, and a small
particle size. With ambient temperature, there is no need to heat the probe,
and the temperatures encountered will be from 60 to 120 degrees F. Ambient
moisture eliminates the need for a Method 4 run, and moisture can be
determined from a wet-bulb dry-bulb thermometer or from a relative humidity
indicator. The molecular weight of ambient air is always 28.95, so the need
for an ORSAT (Method 3) sample is eliminated. Lastly, the particle size of
the sample must be less than 10 microns in diameter. The small particles
found in chromic acid gas streams act more like a gas than a particle so that
isokinetic sampling is not critical.
During field sampling with this method, it was found that locations
close to the plating tank did not produce good results. The mist coming off
the plating tank bath tends to collect on the duct work close to the tank and
then become reentrained in the gas stream. This means that there is a varying
particle size distribution in this area and isokinetic sampling is necessary
to obtain a more representative sample. As the distance from the plating tank
is increased, the larger particles in the gas stream collect on the walls of
the duct and are not reentrained; thus, particle size becomes more uniform.
The method works after a control device, or far enough downstream from the
plating tank that all the particles are 10 microns or less in diameter.
In the future, we plan to produce a video tape showing how to use the
method as well as how to fabricate the sampling train. We will also write an
instruction booklet that will make calculations as simple as possible.
The initial goal of this project was to produce a screening method that
was ±50 percent accurate and costs $500 or less to perform. At the present
time, we feel that our method is as accurate as our isokinetic method at
outlet locations, is suitable for compliance, and costs $350 to perform.
578
-------
DEVELOPMENT OF A SOURCE TEST METHOD FOR HEXAVALENT CHROMIUM
by
J.E. Knoll and M.R. Midgett, U.S. EPA, Research Triangle Park, NC
A.C. Carver, S.C. Steinsberger ctnd W.G. DeWees, Entropy
Environmentalists, Inc., Research Triangle Park, NC
Introduction
Chromate is a strong oxidizing agent, and has a wid«
industrial application as such, in electroplating and as ar
intermediate in the production of chromium salts from chromiun
ores. The latter are employed in pigments, tanning and a great
variety of other processes. There is concern regarding the
nature of the chromium species in the emissions from combustior
sources. Hexavalent chromium, Cr(Vl), is a known carcinogen,
while trivalent chromium, Cr(III), is not so considered. ?.
comprehensive document has been released by EPA on the health
assessment of chromium (1) . The trivalent oxide is the most
stable of the chromium oxides and is formed by heating the metal
or Cr(VI) oxide in air (2). However, when fused with alkaline
substances, air oxidation of Cr(III) to Cr(VI) may take place
(3) . On the other hand, hexavalent chromium may react with any
reducing agents present and be converted to Cr(III). Thus,
because of these competing processes and the importance of
determining the risk that chromiun emissions present to the
public, a study was undertaken to develop a source test method to
measure chromium species in emissions from combustion sources.
579
-------
However, significant problems may result in Cr(VI) sampling
and analysis. That species often occurs at relatively low
concentrations, but which are nonetheless environmentally
significant. Low concentrations require lengthy sampling times
to concentrate the sample, but such procedures also concentrate
interfering substances that alter the valence states of chromium.
Thus, techniques that preserve chromium valence states must be an
important feature of chromium speciation methodology. Several
studies (4) , (5) have been done to develop source test methods
for chromium speciation. However, these methods have addressed
emission sources from which interfering substances were largely
absent. The present study deals with a sampling system that
collects chromium emissions in alkaline solution, in which the
species are quite stable. The sampling train utilized a
recirculating feature that continuously rinsed the sample probe
with the collecting reagent. Samples were analyzed for Cr(VI)
using an ion chromatograph with a post column reactor that
employed the diphenyl carbazide spectrophotometric procedure.
Valence stability was assessed using the radio-active isotope,
Cr51, in the chromate form.
Experimental
A multiple hearth incinerator using lime and ferric chloride
for sludge conditioning and having a high concentration of
chromium in the sludge was tested. Multiple, independent
sampling trains were employed. A schematic of the recirculatory
sampling train is shown in Figure 1. All portions of the train
that came in contact with the sample were Teflon or glass. A
Teflon aspirator capable of recirculating absorbing reagent at 50
580
-------
ml/min while operating at 0.75 cfm was employed. A Teflon union-
T was connected behind the nozzle to provide the absorbing
reagent/sample gas mix. The absorbing solution was 0.2 molar
NaOH with additions of hexavalent 51Cr to determine the extent of
conversion of Cr(VI) to Cr(III). Sampling was conducted for from
2-4 hours, and from 2-5 cubic meters of gas were collected.
Following sampling, the trains were flushed with nitrogen gas and
the samples were recovered and filtered using an all-Teflon
system with a 0.45 micron filter. The samples were analyzed for
Cr(VI) using a model 2110i Dionex Ion Chromatograph equipped with
a post column reactor that employed the diphenyl carbazide
process and a UV/Vis detector. Conversion of Cr(VI) was
determined by measuring the 51Cr eluted with the Cr{VI) peak.
Results
The results of measurements made at a sewage sludge
incinerator are shown in Table 1. Two problems were encountered:
the reduction of Cr(VI) by sulfur dioxide, and the slow oxidation
of Cr(III) in the alkaline collection medium. A 30-minute
nitrogen purge of the train at 20 liters/minute followed by
pressure filtration eliminated these problems. Conversion was
almost completely eliminated, as shown in Table 2.
Ratios of CO/CO2 were also measured at the incinerator and
are plotted vs. the Cr(VI)/Cr(III) ratios in Figure 2. A
relationship between good combustion and higher Cr(VI) is
evident. At low CO levels (good combustion) the ratio of Cr(VI)
to Cr(III) is highest, with approximately 10% of the total
chromium in the hexavalent form. At high CO levels (poor
combustion), the hexavalent chromium is significantly reduced.
581
-------
Mechanism of Chromate Reduction
The procedures that improved the stability of the chromate
samples, flushing with nitrogen to remove dissolved gases and
filtering to remove insoluble Cr(III) hydroxide, suggested the
importance of obtaining some information about the interfering
processes. In sampling combustion sources, the most important of
these is the reduction of chromate by sulfur dioxide. Although
reduction to the trivalent chromic state by sulfur dioxide in
acid solution is the usual method for the treatment of chromate-
containing waste waters, this reaction has received little
attention. Only one kinetic study has been reported (7) , which
was carried out in acid solutions, and at relatively high
concentrations where dichromate was the predominant species.
The present study was carried out in alkaline solutions and
at low concentrations where the chromate ion predominated.
Solutions were prepared containing chromate in the parts-per-
billion range and sodium sulfite at higher concentrations (~.02M)
which approximated the values produced by sampling a gas stream
containing 100 ppm S02. The pH was set using sulfuric acid and
the ionic strength was held constant at 0.2M by addition of
sodium sulfate. This ionic strength value was selected because
it was the value of the collecting solution in the recirculating
sampling train. The pH of the reaction mixtures was measured
with a Model 701A Orion Research pH meter; it remained relatively
unchanged during the course of the kinetic reactions.
Because the sulfite ion concentration greatly exceeded the
chromate concentration, pseudo-first order kinetics were
observed. However, the reaction rate was shown to be
582
-------
proportional to both reactants:
drcr04=1 = k[Cr04=][S02*]
dt
In this equation, the unionized S(IV) species is assumed to be
the reactant, because of the low probability of reaction between
similarly charged ions, and because the pseudo-first order rate
constant varies with pH approximately as does [SO2*]. (The
designation SO2* indicates the combined sulfur dioxide and
sulfurous acid concentrations.) These concentrations were
calculated using the concentrations of added sulfite, the pHs,
published values of the acid dissociation constants (8) , and
standard acid-base equations.
Calculation of second-order rate constants (Table 3)
indicated that the reaction was controlled by a general acid
catalysis, and multiple regression analysis was performed to
determine the contributions by hydronium ion and bisulfite ion to
the reaction rate. The results (Table 4) indicated that both
species produced significant contributions. Evidence for
general acid catalysis received further support when reactions
were run in phosphate buffer. Under those conditions, the
reaction rate increased markedly (Table 5). The contribution of
phosphate to the rate constant was determined by estimating the
hydronium and bisulfite ion contributions, and subtracting those
values from the measured rate constants. Further evidence that
unionized S(IV) species were the principal reactants was provided
by the phosphate buffer studies, where the effects of hydronium
and bisulfite ions were swamped out. Plots of the log of the
pseudo-first-order rate constants vs. pH yielded a slope very
583
-------
close to 2 (Figure 4) , indicating that the rate constant varied
in the same manner that [SO2*3 varies in the pH range of the
investigation.
The presence of acid catalysis provides additional
information about the chromate-sulfur dioxide reaction. Cr(VI)
is known to transfer its electrons in a series of steps, in which
the conversion from the Cr(V) valence state to Cr(IV) is rate
determining, because a change in coordination number from 4 to 6
occurs (9). General acid catalysis indicates that the transfer
of a hydrogen ion facilitates this changeover in the chromate-
sulfur dioxide reactive intermediate.
These mechanistic studies also have practical significance,
and indicate the importance of collecting Cr(VI) at pH's above 8
and avoiding the use of buffers that catalyze the reaction with
sulfur dioxide.
REFERENCES
(1) "Health Assessment Document for Chromium".
1983, EPA-600/83-014A
(2) "Reference Book of Inorganic Chemistry", W.M. Latimer &
J. H. Hildebrand, The Macmillan co., New York, N.Y., 1940
pp 353-354.
(3) "Effect of Lime and Other Precipitants or Sludge
Conditioners on Conversion of Chromium to the Hexavalent
State when Sludge is Incinerated". Final Report EPA
Contract No. 68-03-3346, 1988.
(4) Cox, X.B; Linton, R.W.; and Butler, F.E.; "Determination of
584
-------
Chromium speciation in Environmental Particles.
Multitechnique Study of Ferrochrome Smelter Dust". Environ.
Sci. Technol., 1985, 19: 345-352.
(5) Butler, F.E.; Knoll, J.E.; and Midgett, M.R. "Chromium
Analysis at a Ferrochrome Smelter, a Chemical Plant and a
Refractory Brick Plant". JAPCA 1986, 16: 581-584.
(6) Zatka, V.J. "Speciation of Hexavalent chromium in Welding
Fumes. Interference by Air Oxidation of Chromium", Am. Ind.
Hyg. Assoc. J. 1985, 46; 327.
(7) Mapstone, G. E.; "The Reduction of Dichromate by Sulfur
Dioxide", Chem. in Australia, 1981, 48; 95-97.
(8) Millero, F. J., Hershey, J.P., Johnson, G., and Zhang, J.Z.
"The Solubility of SO2 and the Dissociation of H2SO3 in Nad
Solutions", J. Atm. Chem. 1989, 8: 377-389.
(9) "Mechanisms of Inorganic Reactions", F. Basolo and R.G.
Pearson. John Wiley and Sons, Inc. New York, N.Y. 1968,
pp 498-499.
585
-------
TABLE 1
SAMPLING RESULTS FOR HEXAVALENT CHROMIUM EMISSIONS
FROM A SEWAGE SLUDGE INCINERATOR
Conversion of
Run hexavalent chromium Cr{VI) Relative
No. during sampling micrograms/dscm error
3 1.2 0.15
1.1 0.06
24.6 0.03
12.7 0.01
7 13.1 0.17
12.6 0.15
15.3 0.16
9.6 0.14 21%
9 6.1 0.18
3.3 0.29 33%
11 5.4 0.31
2.7 0.38 14%
13
6.0
3.2
0.04
0.02
47%
586
-------
TABLE 2
STABILITY STUDY OF SEWAGE SLUDGE
INCINERATOR EMISSION SAMPLES
Hexavalent Chromium Concentration, ppb
00
Run No. 10/25/89 12/12/89 difference
9
11
0.78
0.46
0.57
0.52
0.80
0.46
0.59
0.57
0.02
0.00
0.02
0.05
13 0.25 0.29 0.04
-------
TABLE 3
00
00
VARIATION OF RATE CONSTANT
WITH HYDROGEN ION AND
BISULFITE ION CONCENTRATIONS
k, L/M*hrs-1*10-6 [H+], M/L*107 [HSO3-], M/L
2.68 0.43 0.0078
6.78 1.02 0.0120
7.78 1.01 0.0240
2.75 0.28 0.0205
0.67 0.16 0.0132
6.45 0.49 0.0295
-------
TABLE 4
DEPENDENCE OF RATE CONSTANT
ON HYDROGEN ION AND
BISULFITE ION CONCENTRATIONS
Cn
00
CO
k=k0 + k^H*] + k2[HS03-]
Rate Constant Std. Dev.
k0 -8016 4812
k, 5.60E13 1.22E13
k2 8.66E7 4.15E7
-------
TABLE 5
EFFECT OF THE PRESENCE OF PHOSPHATE BUFFER
ON THE CHROMATE-SULFUR DIOXIDE REACTION RATE
k, L/M*hrs 1*IO~7
7.
7.
7 .
7.
7.
7.
pH
030
311
330
460
746
921
^p ks ^p ^s
2.
1.
2.
0.
1.
1.
71
26
22
89
75
64
0.
0.
0.
0.
0.
0.
72
40
39
30
18
13
1
0
1
0
1
1
.99
.86
.83
.59
.58
.51
[H
0
0
0
0
0
0
std
2P04-3
.0238
.0139
.0256
.0124
.0275
.0280
mean
. dev
k
8
6
7
4
5
5
6
1
3*10
.34
.20
.16
.74
.74
.41
.27
. 2
k : measured rate constant with phosphate buffer present.
k£: calculated rate constant with only bisulfite present.
k3: general acid catalysis contribution by phosphate,
k = k0 + k1[H+] + k2[HS03-] +k3[H2P04-]
590
-------
GLASS
IMPINGERS •
TEFLON
IMPINGERS-
CO
RECIRCULAT1NG
LIQUID
TO
METHOD 5-TYPE
METERBOX
Figure Cr*6 -1. Schematic of recirculatory impinger train.
-------
CO
ro
I
£
6
J!
2
I
8-
6-
4-
2-
Hex Cr/Total Cr vs CO/CO2 Ratios
12
10-
Slope = -0.146, R*2 = 0.96
40
60
80
100
120
CO to CO2 Ratio (ppm to %)
-------
Log of Psuedo-First Order Rate Constant vs pH
2.0
1.5-
Slope = -2.79
i.o-
cn
CD
0.5-
o.o-
-0.5-
-1.0
6.8
7.0
7.2
7.4
7.6
7.8
I«
-------
Log (k) vs pH in Phosphate Buffer
2.0
CO
1.5 -
i.o-
0.5 '
o.o-
Slope = -2.01
-0.5
7.0 7.2 7.4 7.6 7.8 8.0
pH
-------
DEVELOPMENT AND FIELD VALIDATION OF A SAMPLING AND
ANALYTICAL METHOD FOR AIRBORNE HEXAVALENT CHROMIUM
P. Sheehan, R. Ricks, G. Brorby,
S. Flack and D. Paustenbach
ChemRisk
1135 Atlantic Avenue
Alameda, California
Hexavalent chromium [Cr(VI)] is classified as a human respiratory
carcinogen by the U.S. Environmental Protection Agency (USEPA). Airborne
Cr(VI) emissions are associated with a number of industrial sources
including metal plating, tanning, chromite ore processing and spray painting
operations; combustion sources such as automobiles and incinerators; and
fugitive dusts from contaminated soil. To date, only a workplace sampling
and analytical method has been developed and validated for measuring
airborne Cr(VI) . The method can detect concentrations as low as 0.5 /ig/m3;
however, environmental concentrations of airborne Cr(VI) are about 1000-
fold lower than this limit of detection. No sampling and analytical method
specific to Cr(VI) at environmental concentra.tions has been previously field
tested. This paper describes a sampling and analytical method for the
quantitation of airborne Cr(VI) at concentrations as low as 0.1 ng/m . The
collection method uses three 500-ml glass impingers In series, each fitted
with Greenberg-Smith impactors operated at 15 liters per minute (1pm) for
24 hours, and containing sodium bicarbonate buffer solution. The average
collection efficiency in the first two of three impingers was 81% of the
mass for airborne Cr(VI) particulate concentrations of 3 to 10 ng/m . It
was found that both Cr(VI) and Cr(III) were stable in the collection medium.
The described method was used to measure ambient levels of hexavalent
chromium at sites having chromium-contaminated soil.
595
-------
Introduction
Airborne hexavalent chromium [Cr(VI)] emissions are associated with a
number of industrial operations. The focus of this study is a Cr(VI) source
of more recent concern, the release of fugitive dusts from contaminated
soil. Although chromium may exist in several valence states, trivalent
chromium [Cr(III)] and Cr(VI) are the two valences of interest for human
health. Cr(III), which is relatively non-toxic, is the predominant
naturally occurring environmental form of chromium, and an essential human
dietary mineral. In contrast, Cr(VJ) is primarily man-made and has been
shown to be carcinogenic to humans exposed to relatively high levels in the
workplace air. Therefore, the differentiation between airborne
concentrations of Cr(VI) and Cr(III) is important in the assessment of
potential human hazards from chromium.
At present there is no standardized method with the necessary specificity
and sensitivity for measuring environmental levels of airborne Cr(VI),
particularly from contaminated soil. The NIOSH Method 7600, developed in
1984, has been used to measure Cr(VI) concentrations in the occupational
setting. This method uses a 37-mm diameter, 5.0 /An pore size PVC membrane
filter for collection, and Visible Absorption Spectrophotometry (VAS) for
quantitation of Cr(VI). However, the detection limit of the NIOSH method
is 500 ng/m3'1', approximately 25 to 50 times higher than typical
environmental concentrations of airborne total chromium'2'3'. Refinements
of Method 7600 for environmental sampling applications have achieved some
increase in analytic sensitivity, although, in some cases, at the cost of
specificity for Cr(VI)(4' 5- 6' 7- 8- 9). The increased sensitivity that these
refinements produce still does not meet the needs of environmental sampling.
In 1987, the California Air Resources Board (GARB) developed a sensitive
sampling and analytical procedure to measure Cr(VI) downwind of industrial
point sources'10'. The GARB method, which relies on the water solubility of
airborne chromium-containing particulates for collection, is a modification
of EPA Method 5 that was originally proposed by the Research Triangle
Institute'11'. This method is currently being evaluated by California's
South Coast Air Quality Management District (SCAQMD)'12'. A typical impinger
train consists of three Greenberg-Smith impingers in series. The first and
second impingers are filled with 100 ml of a 0.02 M sodium bicarbonate
solution (pH 8-9), and the third impinger is empty. The buffer solution
prevents the reduction of Cr(VI) also ensuring stability during preparation
and analysis. The GARB method makes use of an ion chromatography (1C)
column to separate the Cr(VI) from the impinger solution for quantitation.
Following chromatographic separation, the Cr(VI) is complexed with
diphenylcarbazide and measured spectrophotometrically at 520 nm. The limit-
of-detection (LOD) of the GARB method for a 20 m3, air sample is
approximately 1 to 0.1 ng/m3 depending on whether a preconcentration step
is used in sample preparation.
After reviewing the available methods, it was concluded that the GARB
method offered the specificity and stability, as well as the analytical
sensitivity, required to assess environmental concentrations of airborne
hexavalent chromium. Although the GARB method has been used for regulatory
compliance, field validation of the method has not been conducted. Further,
the applicability of the method to the sampling of airborne soil particles
containing Cr(VI) has not been demonstrated'11'. This paper describes the
development and field validation of an impinger/IC method that can
accurately quantitate the airborne concentration of Cr(VI) over a fairly
596
-------
wide range (0.5 to 50 ng/m3) . The field study was restricted to validating
the method at sites of Cr(VI) contaminated soil, although this method should
be applicable to Cr(VI) in a mist or fume.
Experimental Materials and Methods
The Cr(VI) sampling train consisted of three glass 500-ml impingers in
series filled with 0.02 M sodium bicarbonate solution followed by a 37 mm
diameter, 0.5 /Im glass-fiber backup filter. The impinger series was placed
in an ice bath. The impinger train was connected to a Dwyer pump, and was
calibrated using a Gllian digital flow meter. A Gilmont compact rotameter
was placed in-line between the cassette and the pump for the purpose of
identifying any changes in the flow rate.
Air was drawn through the sampling train at a rate of approximately 15
liters per minute. Flow rate calibration was performed at the beginning
and at the end of the 24-hour sampling period. Sampling apparatus
parameters were monitored every 4-6 hours. The impinger solution volume
and pH were recorded at the beginning and end of the sampling period. All
samples were transported on ice to the analytical laboratory within 24 hours
after collection. Strict adherence to a decontamination protocol was
necessary to prevent soil contamination of the sampling equipment. Prior
to sampling apparatus assembly, impingers and tubing were soaked in a dilute
nitric acid (1%) bath, rinsed three times with distilled water, and
"charged" once with buffer solution. Field blanks were included for every
sampling period.
This collection method differs from that used by GARB in two ways: 1)
the volume of buffer in each impinger is 200 ml, rather than 100 ml, and 2)
the third impinger contains fluid rather than being empty. These
modifications were made to enhance the particulate Cr(VI) collection
efficiency of the sampling train by providing an additional impaction step
(the 3rd impinger} and an increased contact time (additional impinger
solution).
Cr(VI) collected in the buffer solution was separated using 1C. The 1C
column was packed with lonPac AS7 (manufactured by Dionex). The eluent
consisted of 0.25M ammonium sulfate and 0.1M ammonium hydroxide at a pH of
7.0. The eluent was passed through the coltunn at a flow rate of 1.5 ml/min.
The chromate ion eluted at 3.5 to 4.0 minutes and was then complexed with
0.002 M diphenylcarbazide. The red chromium complex was spectrophoto-
metrically quantified at 520 nm. Based on a total sample volume of 20 m ,
the approximate LOD of Cr(VI) in air for this method is 0.1 ng/m3.
Study Design and Results
Breakthrough
The breakthrough of Cr(VI) in solution, i.e., breakthrough from one
impinger to the next, was assessed by drawing air through four different
sampling trains at 15 1pm for a 24 hour period. Trains 1 and 2 contained
two impingers in series with the first impinger spiked with 10 ng/L Cr(VI).
Trains 3 and 4 contained unspiked buffer. A 37 mm diameter, 0.5 Jim pore
size Teflon™ filter was placed ahead of the impingers in all the sampling
trains to prevent the collection of any chromium contaminated particulates.
Trains 1 and 2 were used to assess the potential movement of Cr(VI) in
solution from the first to the second impinger. Trains 3 and 4 were used
597
-------
to demonstrate that detectable levels of ambient chromium were not collected
by the sampling trains during these tests.
No detectable amount of Cr(VI) was transported from the first impinger to
the second and no ambient chromium was collected.
Collection Efficiency
The impinger collection mechanisms for chromium contaminated particulates
likely involve a combination of impaction and solubility. The sampling
flow rate is a key variable influencing particle collection efficiency for
both of these collection mechanisms. In evaluating the effect of flow rate,
collection efficiency is defined as the percentage of the total amount of
Cr(VI) collected by the sampling train that was found In the first impinger.
No statistically significant difference in the concentration of Cr(VI) in
the first impinger was found for flow rates of 5, 10, and 15 L/min. Samples
where the total Cr(VI) concentration was at least ten times the limit of
detection, most of the chromium was collected in the first two impingers in
series. In approximately 73% of these sampling trains less than 15% of the
total Cr(VI) mass was in the third impinger indicating good relative
collection efficiency at environmental relevant Cr(VI) concentrations.
Cr(VI) Stability
To assess the stability of Cr(VI) in the buffer solution, a 10 Hg/L field
spike sample was analyzed on days 2, 6, 8, 10, 13, 15, and 20 after spiking.
The Cr(VI) concentrations were not significantly different over the twenty
day storage period, indicating that samples are stable and do not require
immediate analysis following collection.
Cr(III) Stability
Under environmental conditions, Cr(III) is expected to make up a large
portion of total airborne chromium. The oxidation of Cr(III) to Cr(VI) is
therefore a potential concern to be addressed in sampling for Cr(VI). Under
reduced temperature conditions approximately 0.2% of a 1000 Mg/L Cr(III)
spike was converted to Cr(VI). The portion converted under ambient
temperature conditions was 0.7%. Based on the results of a t-test, there
was no statistically significant difference between the groups at the 95%
confidence limit. This conversion is not expected to have a measurable
effect on Cr(VI) concentrations at the concentrations of total airborne
chromium typically sampled in ambient urban air since they are 10 to 100
times less than the Cr(III) spike level used in this experiment.
Precision and Spike Recovery
Triplicate co-located impinger sampling trains, designated 1, 2, and 3,
were used to assess the combined sampling and analytic precision. As
presented in Table I, the pooled coefficient of variation between replicates
was 20%. These data indicate that the method precision is within the range
specified by the USEPA - Contract Laboratory Program for quantification of
trace metals(13). This level of precision is also well within the range
expected for industrial hygiene sampling procedures^145.
True accuracy of the sampling and analytical methods could not be
determined since a Cr(VI) reference standard was not available. As a
surrogate measure, analytic accuracy was assessed as a function of recovery
598
-------
of a 1 M8/L Cr(VI) field spike in the validation study program.
Additionally, 10 Mg/L Cr(VI) field spikes were collected throughout the
sampling program. As presented in Table IX, under both reduced (ice bath)
and ambient temperature conditions, spike recovery of the 1 Mg/L Cr(VI)
spikes ranged from 87 to 99%. The mean percent recovery was approximately
92%, Recovery of the 10 Jig/L Cr(VI) spikes analyzed during development of
the sampling method ranged from 65% to 118%, with a mean percent recovery
of 98%. Based on this data, method accuracy for the validated sampling
program is within USEPA-CLP accuracy criteria (± 25%) and NIOSH accuracy
criteria (i.e., greater than 90%)(14). Additionally, there was no
statistically significant difference between the two groups at the 95%
confidence limit. These results suggest that Cr(VI) at environmentally
relevant concentrations is stable in the buffer solution under both reduced
(ice bath) and ambient temperature conditions.
Field Experience
Three hundred samples were collected between August and October 1989 at
17 different Northern New Jersey locations known to have fill material
containing chromium residue. The sampling period Is believed to represent
the time of year during which the potential for dust generation is the
greatest. Sampling for total chromium and Cr(VI) was performed at each
sampling location. Fourteen of the seventeen sites had indoor locations.
At most sites, two outdoor locations, one upwind and one downwind, and two
indoor locations were sampled. A summary of the air sampling data is
presented in Table III.
Conclusions
The sampling and analytical methods evaluated in this study are
sufficiently sensitive and specific for field sampling of airborne
hexavalent chromium particulates. Recovery rates are excellent regardless
of whether or not an ice bath is utilized during sampling. The combined
sampling and analytical method meets the requirements for both spike
recovery and precision as specified by EPA criteria. The results of this
method development and field validation indicate that the selected sampling
and analytical methods fulfill the need for accurate quantitation of ambient
concentrations of Cr(VI) in air.
OA02ALR1
599
-------
REFERENCES
1. National Institute of Occupational Safety and Health: NIOSH Manual of
Analytical Methods. 3rd ed. , Method #7600, by D. Molina and M.T. Abell
(NIOSH Pub. No. 84-100). Cincinnati, Ohio, 1984.
2. J. 0. Nriagu, J. M. Pacyna, J. B. Milford, and C. I. Davison:
Distribution and Characteristic Features of Chromium in the Atmosphere
In: Chromium in the Natural and Human Environment. pp. 125-172
(1988).
3. L. Fishbein: Sources, Transport, and Alterations of Metal Compounds:
An overview. I. Arsenic, Beryllium, Cadmium, Chromium and Nickel.
Environ. Health Perspect. 40:43-64 (1981).
4. U.S. Environmental Protection Agency (USEPA): Modified EPA Method
218.1. Atomic Absorption, Flame Technique. USEPA Contract Laboratory
Program, 1987.
5. National Institute for Occupational Safety and Health: Method No.:
P&CAM 173. NIOSH Manual of Analytical Methods. General Procedure for
Metals, Atomic Absorption, 1974,
6. National Institute of Occupational Safety and Health: Method #7024.
NIOSH Manual of Analytical Methods. 3rd ed., M. Millson and R. Delon
Hull. Cincinnati, Ohio, 1984.
7. U.S. Environmental Protection Agency (USEPA): Test Method for
Chromium, Dissolved Hexavalent (Atomic Absorption, Furnace Technique) -
Method 218.5. Environmental Monitoring and Support Laboratory,
Cincinnati, OH, 1983.
8. U.S. Environmental Protection Agency (USEPA): Test Method for
Inductively Coupled Plasma-Atomic Emission Spectrometrlc; Trace Element
Analysis of Water and Wastes -Method 200.7. Environmental Monitoring
and Support Laboratory, Cincinnati, OH, 1982.
9. National Institute of Occupational Safety and Health: Method #7300.
NIOSH Manual of Analytical Methods. 3rd ed., M. Millson and R. Delon
Hull. Cincinnati, Ohio, 1984.
10. California Air Resources Board (GARB): Memorandum for the Air
Resources Board to G. Murchison dated November 2, 1987.
11. Research Triangle Institute (RTI): The Fate of Hexavalent Chromium in
the Atmosphere. RTI/3798/00-01F. Prepared for the California Air
Resources Board, 1988.
12. Southern California Air Quality Management District (SCAQMD): Mr. Rudy
Eden, personal communication. 1989.
13. U.S. Environmental Protection Agency (USEPA): Laboratory Data
Validation Functional Guidelines for Evaluating Inorganic Analyses.
Office of Emergency and Remedial Response. Washington D.C., 1987.
14. National Institute for Occupational Safety and Health: P&CAM 319 NIOSH
Manual of Analytical Methods. 2nd Ed., Vol. 6, (NIOSH Pub. No. 80-
125). Cincinnati, Ohio,1980.
600
-------
TABLE I: METHOD PRECISION
Study
# Replicates
Coefficient of Variation (CV)
Pool CV
7.4X
6.8X
8.9%
20. OX
10.8X
TABLE II: RECOVERY OF FIELD SPIKES
Study
1
2
3
X>
Si-
1 ppb
X Recovery
91 X
87X
99X
92X
6X
Study
1
2
3
4
5
6
X~2=
4=
10 ppb
X Recovery
103X
118%
90X
95X
117X
97*
103X
12X
TABLE III: AIR SAMPLING DATA SUMMARY
Site
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Summary Statistics
Mean (standard deviation)
Geometric Mean (GSD)
Average
Indoor Cr(vi)
ng/m3
3.2
No building on site
9.6
5.2
No building on site
6.0
3.3
1.5
2.4
1.8
0.2
3.5
3.3
2.3
1.7
1.4
No building on site
3.2 (2.4)
2.2 (2.7)
Average
Outdoor IJ-(VI)
ng/m
7.9
1.1
4.4
7.8
6.5
9.6
6.9
6.9
6.9
4.3
2.3
2.7
0.5
1.7
3.0
3.3
2.1
4.6 (2.8)
3.1 (3)
0402ALR1
601
-------
Determination of Average Ambient PCDDs/PCDFs Concentrations in the Vicinity
of Pre-Operational Resource Recovery Facilities in Connecticut
Bruce E. Maisel
ENSR Consulting and Engineering
35 Nagog Park
Acton, MA 01720
ABSTRACT
Recent regulatory statutes issued by the State of Connecticut require that
ambient monitoring for 2,3,7,8-substituted PCDDs/PCDFs be conducted on both
a pre-operatlonal and post-operational basis. In response to this requisite, pre-
operational monitoring programs designed to determine background levels of
PCDDs/PCDFs in ambient air have been completed in the vicinity of four resource
recovery facilities located in Bridgeport, Bristol, Hartford and Wallingford,
Connecticut. Sampling and analytical methodology involved the use of high
volume sorbent samplers in conjunction with high resolution (magnetic sector)
mass spectrometry to determine background ambient PCDDs/PCDFs
concentrations in the 0.01-0.1 pg/m3 range.
This paper presents average ambient PCDDs/PCDFs levels determined for
each of the four pre-operational monitoring networks listed above. Congener
profiles established on a region-specific basis will be compared to identify
variation in average ambient PCDDs/PCDFs burdens found to occur among the
four study areas. In addition, similarities between ambient and combustion
source profiles will be examined as will the noted predominance, on a region-
specific basis, of certain 2,3,7,8-substituted PCDDs/PCDFs species. Toxicity
assessments for each of the four regions will also be presented and compared
through application of the EPA Toxic Equivalency Factor (TEF) model to the
Connecticut ambient PCDDs/PCDFs database.
602
-------
INTRODUCTION
The State of Connecticut requires that ambient monitoring for
PCDDs/PCDFs be conducted in the vicinity of resource recovery facilities
constructed in Connecticut on both a pre-operational and post-operational basis
[1]. ENSR Consulting and Engineering has completed such pre-operational
studies around four resource recovery facilities located in Bristol, Hartford,
Bridgeport and Waliingford, CT. The Hartford and Bristol programs were
conducted concurrently from July through September, 1987; the Bridgeport
program was conducted from November 1987 to January 1988, and the
Wallingford program was conducted from August to October, 1988.
This paper presents average ambient PCDDs/PCDFs levels as determined
from conduct of the aforementioned programs. Average ambient PCDDs/PCDFs
burdens for the four study regions are presented and compared along with
profiles for the tetra through octa PCDDs/PCDFs congener classes and individual
2,3,7,8-substituted PCDDs/PCDFs. In addition, the Connecticut PCDDs/PCDFs
average ambient database is applied to the EPA Toxic Equivalency Factor (TEF)
model to provide average atmospheric PCDDs/PCDFs burdens expressed as
2,3,7,8-TCDD toxic equivalents on a network-specific basis. These values are
then compared to the Connecticut ambient PCDDs/PCDFs standard of 1.0
pg/m3.
SAMPLING AND ANALYSIS METHODOLOGY
General Metal Works Polyurethane Foam (PUF) PS-1 samplers were used
to collect the PCDDs/PCDFs isomers listed in Table 1. The samplers are
essentially modified high volume air samplers employing a glass fiber filter in
tandem with a sorbent trap to collect particulate-associated and vapor-phase
PCDDs/PCDFs, respectively. Air flow rates between 140 and 220 Ipm were
utilized, in conjunction with 24 to 96 hour sample sessions to produce sample
volumes between 350 m3 and 950 m3. All PS-1 samplers were calibrated prior to
and at the conclusion of each sampling session using an NBS traceable
calibrated orifice. Quality Assurance/Quality Control elements implemented for
these programs included field blanks, method blanks, field surrogates, internal
standards and collocated samples [2].
All program samples selected for analysis were prepared and analyzed
based on the protocol outlined in EPA Methods 8280 and 8290. Native dioxins
and furans collected from the ambient air were quantified against isotopically
labelled internal standards added to each sample prior to extraction with toluene.
Extracts were cleaned by column chromatography and subjected to complete
PCDDs/PCDFs analyses by high resolution gas chromatography/high resolution
mass spectrometry (HRGC/HRMS). Detection limits of 10 to 50 fg/m3 were
achieved.
603
-------
RESULTS AND DISCUSSION
Ambient air samples were collected as described above in the vicinity of
the Bristol (n = 23), Hartford (n = 31), Bridgeport (n = 22), and Wallingford (n = 24),
Connecticut Resource Recovery Facilities for the target parameters listed in Table
1. An average concentration for each target parameter was calculated for the four
study areas listed above, with non-detected values included into the database as
one-half the reported detection limit. This treatment of non-detected observations
has been discussed by others in the open literature [3,4].
Table 2 provides average ambient concentrations of PCDDs/PCDFs (total
CI4 through Clg) for each monitoring network, which ranged from 1.8 pg/m3
(Bristol) to 7,1 pg/m3 (Bridgeport). These values are comparable to those
measured by others from similar ambient monitoring studies [5,6,7]. The
PCDDs burden for aii networks included a significant contribution from OCDD,
which was the predominant species of PCDDs/PCDFs measured at each network
during the Connecticut programs.
Seasonal influences and regional transport phenomena may contribute to
the variability in average ambient PCDDs/PCDFs burden found to exist among the
four monitoring networks. For example, additional combustion source
aggregates present during wintertime, may have contributed to the higher
PCDDs/PCDFs levels measured at the Bridgeport network. These additional
wintertime combustion sources would not have impacted the Mid-Connecticut,
Bristol and Wallingford monitoring networks, which were operated during the
summer and early fall seasons. It has also been shown that the majority of
PCDDs/PCDFs are particulate-associated, particularly at lower ambient
temperatures, such as encountered during wintertime [8,9]. For this reason,
regional and/or long-range transport of combustion source particulate-associated
PCDDs/PCDFs may also contribute to the seasonal variability of PCDDs/PCDFs
levels by region. This is particularly likely for the Bridgeport network which may
have been impacted by particulate-associated PCDDs/PCDFs originating from
combustion sources located in the Metropolitan New York City area [9].
Figures 1 and 2 show the average ambient concentration of the CI4
through Cla PCDDs and PCDFs congener classes, respectively, for each of the
four monitoring networks. These congener profiles resemble those seen for
combustion sources, characterized by higher atmospheric levels of PCDDs as
chlorination increases (TCDD
-------
however, is the predominance of the 1,2,3,4,6,7,8-HpCDF congener in all four
monitoring networks.
Ambient PCDDs/PCDFs data gathered from these programs are presented
in terms of 2,3,7,8-TCDD toxic equivalents. This is accomplished by applying the
EPA Toxic Equivalency Factor model [12], contained in Table 3, to the ambient
PCDDs/PCDFs database established through this study. Average toxic
equivalents resulting from the EPA TEF model for each of the four ambient
monitoring programs are contained in Table 4. The State of Connecticut has
issued a standard for ambient PCDDs/PCDFs levels of 1.0 pg/m3 expressed as
EPA 2,3,7,8-TCDD toxic equivalents [1]. As noted in Table 4, none of the four
regions studied resulted in a toxic equivalents sum which exceeded this standard.
The average ambient PCDDs/PCDFs levels expressed as EPA toxic equivalents
ranged from 0.012 pg/m3 (Bristol) to 0.049 pg/m3 (Bridgeport). It should be
noted that numerous other Toxic Equivalency Factor models exist (Nordic,
Ontario, New York, California, International, etc.}, each of which produces varying
toxicity assessments upon application to the same PCDDs/PCDFs database [13].
SUMMARY AND CONCLUSIONS
Ambient monitoring for PCDDs/PCDFs in the vicinity of pre-operational
recovery facilities in Connecticut showed that average burdens of PCDDs/PCDFs
vary by region. This regional variation may be explained by seasonal influences
and atmospheric transport phenomena. The ambient congener profiles mirror
those of combustion sources as seen by the increased predominance of the
higher homolog PCDDs congener classes (CI4
-------
REFERENCES
1. Bruckman, L, "An Overview of Connecticut's Air Pollution Control Program
for Dioxin and Furan Emissions," Chemosphere (1990). (in Press)
2. Maisel, B. and G. Hunt, "The Role of Quality Assurance/Quality Control in
the Interpretation of Ambient PCDDs/PCDFs Data." Proceedings of the
1989 International Symposium on the Measurement of Toxic and Related
Air Pollutants, Raleigh, NC (1989).
3. Nehls, G. and G. Akland, "Procedures for Handling Aerometric Data,"
JAPCA, 23:180-184 (1973).
4. Kushner, E., "On Determining the Statistical Parameters for Pollution
Concentrations from a Truncated Data Set," Atmospheric Environment,
10:975 (1976).
5. Tiernan, T., D. Wagel, G. Vanness, J. Garrett, J. Solch, and L. Harden,
"PCDD/PCDF in the Ambient Air of a Metropolitan Area in the U.S.,"
Chemosphere, Vol. 19, 541-546 (1989).
6. Nakano, T., M. Tsuji and T.Okuno, "Level of Chlorinated Organic
Compounds in the Atmosphere," Chemosphere, Vol. 16, Nos. 8/9, 1781-
1786 (1987).
7. Hunt, G., "Measurement of PCDDs/PCDFs in Ambient Air," JAPCA, Vol.
39, No. 3 (1989).
8. Eitzer, B. and R. Hites, "Polychlorinated Dibenzo-p-dioxins and
Dibenzofurans in the Ambient Atmosphere of Bloomington, Indiana,'
Environ. Sci. Technol., Vol. 23, No. 11 (1989).
9. Hunt, G. and B. Maisel, "Atmospheric PCDDs/PCDFs During Wintertime
in a Northeastern U.S. Urban Coastal Environment," Chemosphere (1990).
(In Press)
10. Rappe, C., S. Marklund and L. Kjeller, "Long-Range Transport of PCDDs
and PCDFs in Airborne Particles," Chemosphere, Vol. 18, Nos. 1-6, 1283-
1290 (1989).
11. Czuczwa, J. and R. Hites, "Airborne Dioxins and Dibenzofurans: Sources
and Fates," Environmental Science & Technology, 20 (2), 195-200 (1986).
12. US EPA, "Interim Procedures for Estimating Risks Associated with
Exposures to Mixtures of Chlorinated Dibenzo-p-dioxins and
Dibenzofurans (CDDs and CDFs)," EPA/625/3-87/012 (1987).
13. Maisel, B. and G. Hunt, "Background Concentrations of PCDDs/PCDFs in
Ambient Air -A Comparison of Toxic Equivalency Factor (TEF) Models,"
Chemosphere (1990). (In Press)
14. Hunt, G., B. Maisel and M. Hoyt, "Ambient Concentrations of
PCDDs/PCDFs in the South Coast Air Basin," NTIS Document No. PB90-
169970/WEP (1990).
606
-------
Table 1. Target Parameter List.
PCDDs
2,3,7,8 - TCDD
1,2,3,7,8-PeCDD
1,2,3,4,7,8-HxCDD
1,2,3,6,7,8-HxCDD
1,2,3,7,8,9-HxCDD
1,2,3,4,6,7,8-HpCDD
OCDD
Key
PCDFs
1
2
3
4
5
6
(Also tetra through hepta
PCDDs/PCDFs congener class totals)
2,3,7,8-TCDF
1,2,3,7,8-PeCDF
2,3,4,7,8 - PeCDF
1,2,3,4,7,8-HxCDF
1,2,3,6,7,8-HxCDF
2,3,4,6,7.8 - HxCDF
1,2,3,7,8,9-HxCDF
1,2,3,4,6,7,8-HpCDF
1,2,3,4,7,8,9-HpCDF
OCDF
Key
7
8
9
10
11
12
13
14
15
Table 2. Average Ambient Levels of PCDDs/PCDFs (tetra through octa) for
the Connecticut Study Areas
Monitoring
Network
Location
Bristol
Hartford
Bridgeport
Wallingford
Season
Summer
Summer
Winter
Summer/Fall
Average PCDDs
Concentration
(pg/m3)
1.2
2.6
4.4
3.8
Average PCDFs
Concentration
(pg/m3)
0.6
0.8
2.7
1.7
Total
Burden
(pg/m3)
1.8
3.4
7.1
5.5
Number of
Samples
(n)
23
31
22
24
607
-------
Table 3. EPA Toxic Equivalency Factor (TEF) Model.
PCDDs
2,3,7,8 - TCDD
OTHER TCDD
1,2,3,7,8 - PeCDD
OTHER PeCDD
1,2,3,4,7,8-HxCDD
1,2,3,6,7,8 -HxCDD
1,2,3,7,8,9-HxCDD
OTHER HxCDD
1,2,3,4,6,7,8-HpCDD
OTHER HpCDD
OCDD
Toxic
Equivalency
Factor
1
0.01
0.5
0.005
0.04
0.04
0.04
0.0004
0.001
0,00001
0
PCDFs
2,3,7,8 - TCDF
OTHER TCDF
1,2,3,7,8 -PeCDF
2,3,4,7,8 - PeCDF
OTHER PeCDF
1,2,3,4,7,8-HxCDF
1,2,3,6,7,8-HxCDF
2,3,4,6,7,8- HxCDF
1,2,3,7,8,9-HxCDF
OTHER HxCDF
1,2,3,4,6,7,8-HpCDF
1,2,3,4,7,8,9-HpCDF
OTHER HpCDF
OCDF
Toxic
Equivalency
Factor
oT
0.001
0.1
0.1
0.001
0.01
0.01
0.01
0.01
0.0001
0.001
0.001
0.00001
0
Table 4. Average Toxic Equivalents Using EPA TEF Model for Connecticut Study Regions.
Monitoring
Network
Average Toxic
Equivalents (EPA)
(P9/m3)
Bristol
Hartford
Bridgeport
Wallingford
0.012
0.027
0.049
0.014
Connecticut ambient PCDDs/PC&F$ standard /$ I.0pg/m3(anrtuafizsd, £PA Model}
expnsssatf as 2,3,7,8-TCDD equivalents.
608
-------
Figure 1
Average Ambient Concentrations of F'CDDs
Congener Class Sums (CI4 through C18)
3.0
TCDD PeCDD HxCDD HpCDD OCDD
• BRISTOL
II HARTFORD
• BRIDGEPORT
M WALLINGFORD
Figure 2
Average Ambient Concentrations of PCDFs
Congener Class Sums {CI4 through Cle )
3.0
2.5
2.0
ra 1.5
£3
c
01
o
8 1.0
0.5
TCDF
PeCDF HxCDF HpCDF OCDF
BRISTOL
HARTFORD
BRIDGEPORT
WALLINGFORD
609
-------
Figure 3
Average Ambient Concentrations of
2,3,7,8 - Substituted PCDDs
(470)
300 _
"o>
•I
150
c
100
DO
tetra
v1
-HI,
penta
2
j-mr,
hexa
3
I
hexa
4
1
I
d-V
hexa
5
1
n
r
hepta
6 ,
V
(3»e t»Ws 1 lor kty)
BRISTOL
HARTFORD
BRIDGEPORT
WALLINGFORD
Figure 4
Average Ambient Concentrations of
2,3,7,8 - Substituted PCDFs
300
250
E 200
c
o
I 15°
c
100
50
tetra ' penta penta hexa hexa ' hexa
78 9 10 11 12
i
hexa hepta hepta
13 14 15
610
-------
DEVELOPMENT AND EVALUATION OF METHODS
TO DETERMINE PATHOGENS IN AIR EMISSIONS
FROM MEDICAL WASTE INCINERATORS
R. R. Segall, W. G. DeWees, and G. C. Blanschan
Entropy Environmentalists, Inc.
Research Triangle Park, North Carolina 27709
F. Curtis and R. T. Shigehara
Emission Measurement Branch
U. S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
K. M. Hendry and K. Leese
Research Triangle Institute
Research Triangle Park, North Carolina 27709
The U. S, Environmental Protection Agency has determined that medical
waste incinerator emissions may reasonably be expected to contribute to
the endangerment of public health and welfare. A study has been initiated
to determine which air pollutants may need to be regulated under the New
Source Performance Standards. One pollutant of potential concern is
pathogenic microorganisms or pathogens. However, there is currently no
accepted method for measurement of these emissions. An EPA-sponsored
study was conducted to develop and evaluate such a method. A technique
relying on measurement in the emissions of an indicator organism (the heat
resistant spores of the bacteria, Bacillus stearothermophilus) spiked into
the incinerated waste was selected for evaluation. Laboratory studies
evaluated the precision of the proposed analytical technique, the
suitability of the sampling buffer, and sample storage constraints. Field
evaluation testing at a typical medical waste incinerator was used to
assess the effects of sample gas components on sample recovery and the
accuracy of the combined sampling and analytical procedures. This paper
summarizes the results of these studies.
611
-------
Introduction
The U. S. Environmental Protection Agency (EPA) has determined that
medical waste incinerator emissions may reasonably be expected to
contribute to the endangerment of public health and welfare.
Consequently, it is anticipated that new source performance standards
(NSPS) for new medical waste incinerators will be developed under Section
lll(b) of the Clean Air Act. The Office of Air Quality Planning and
Standards of the EPA has initiated a study to determine which air
pollutants may need to be regulated. One pollutant of potential concern
is pathogenic microorganisms (pathogens). As part of this study, the EPA
will test a series of medical waste incinerators to determine the
destruction efficiencies for pathogens under different operating
conditions. A test method was required to make these determinations.
The small amount of research conducted to date has demonstrated two
general approaches for measuring pathogen emissions from incinerators.1'6
The first is to collect and culture the native species present in the
stack gas and identify and quantify the same.^' The second is to spike
the incinerator waste feed with an "indicator" organism (typically highly
heat resistant bacterial spores), collect the "indicator spores," and then
selectively culture and quantify them,1'3'5'6 The second approach was
selected by the EPA to provide the basis for a first draft sampling and
analytical protocol because (1) the incinerator is challenged with heat
resistent spores simulating worst case conditions, (2) enough "indicator
spores" can be spiked to ensure that the destruction efficiency of .the
incinerator can be demonstrated to 99-9999$ efficiency, (3) the sample
recovery and handling requirements are predictable, (4) the analysis of
the indicator spores requires specific culturing procedures which serves
to protect the analyst and minimizes sample contamination, and (5) the
analysis is greatly simplified and less costly because it does not
require identification of multiple species.
To summarize the first draft protocol, the incinerator is challenged
with wastes spiked with a known quantity of Bacillus stearothermophilus
spores. Emission samples are collected isokinetically using an EPA Method
5~type train containing a buffered impinger solution and a backup filter.
The recovered sample is heat-shocked to kill non-spore microorganisms,
selectively cultured for the indicator spores, and the indicator spores
are quantified using a plate counting technique. Entropy
Environmentalists, Inc. (Entropy) and Research Triangle Institute (RTI)
conducted the laboratory and field evaluations necessary to refine the
protocol and to assess its collection efficiency, accuracy, and precision.
Laboratory Evaluation
The laboratory evaluations involved confirming the selection of the
"indicator" organism, determination of the analytical precision,
identification of a suitable sampling buffer, assessment of sample
storage constraints, assessment of the potential for background
contamination, simulated sampling to estimate the accuracy of sample
recovery, and design and construction of a water-cooled probe.
Bacillus stearothermophilus was selected as the "indicator spore"
species for the following reasons: (1) its high relative inherent heat
resistance, (2) it is commonly used in assessing sterilization processes,
(3) its high culturing temperature (65°C) which excludes most other
microorganisms, (4) the relative ease of culturing for counting purposes,
and (5) it is not pathogenic.
612
-------
Cultures of B. stearothermophilus were used to evaluate and refine
the analytical techniques in the draft protocol and, in particular, to
determine the precision for enumeration (counting) of the spores.
Initially, an agar plate streaking technique was selected for enumeration,
Briefly, this techniques involves: suspending bacteria in a suitable
diluent; streaking nutrient agar petri plates using a glass rod with known
quantities of this suspension; counting the number of colonies which
develop after incubation of the plates; and multiplying the number
obtained by the dilution factor. However, a number of experiments
conducted to assess the analytical precision of the techniques yielded
relatively poor precisions [relative standard deviation (RSD) = 30$ to
120JI. ]. It was suspected that cell clumping and adhesion to the glass rod
used for dispersal of the sample might be causing variability. The
plating protocol was then modified slightly to use a solution of soft agar
to disperse the sample on the plate rather than the glass rod. This
technique was found unsuitable for the EK _ sitearothermophilus because the
agar melted and dripped at the high incubation temperatures. Finally,
based on techniques successfully used by the Ontario Ministry of the
Environment,7'8 an experiment was conducted to assess the precision of
enumeration using microbiological filters. For this technique, known
quantities of a bacterial suspension are VEicuum filtered through a
cellulose nitrate filter (e.g., NalgeneR 0.2 urn disposable sterile filter
units) ; the filter is removed from the unit and placed on an agar plate;
and the plate is incubated and counted as previously described. Counts of
ten replicate filters yielded precisions in the range (RSD = W% to 25%)
expected for this type of measurement (RSD = 18#) . Relative standard
deviations from laboratory experiments conducted subsequently have ranged
from 9% to
Because of the potential for high acid emissions (up to 2000 ppm)
from this type of incinerator and the intolerance of the living spores for
large extremes in pH, the sample collection reagent must be buffered.
Work previously conducted by the Ontario Ministry of the Environment
indicated that phosphate buffer could be a suitable sample collection
media. It was necessary to consider sample exposure to the candidate
buffer and acid catch during testing and storage of the sample in the
chosen buffer solution. An experiment was designed and conducted to
assess spore viability with exposure to phosphate buffers of different
ionic strength and decreased pH. Each of four buffer solutions was spiked
with approximately 10^ spores of B. stearothermophilus: (1) 2.0 M
phosphate buffer, pH = 6.9, (2) 2.0 M phosphate buffer, adjusted with HC1
to pH = 5-7 (HC1 roughly equivalent to amount collected when sampling a
1000 ppm stack for two h) , (3) 0-5 M phosphate buffer, pH = 6.9, and (4)
0.5 M phosphate buffer, adjusted with HC1 to pH = 5.7. Portions of the
spike solutions were appropriately diluted, heat shocked, filtered,
plated, and incubated to obtain a count for the spike. The spiked buffer
solutions were left at room temperature for two h. Then representative
aliquots were removed from each solution, heat shocked, filtered, plated,
incubated, and counted. Filtration and plating of each buffer solution
was performed in quadruplicate, the plates incubated overnight at 65° C,
and the colonies enumerated the next day. Following removal of the first
set of aliquots, the solutions were stored at room temperature. The
aliquotting and analytical procedures were repeated at intervals of one,
three, and seven days after spiking. There was no significant change in
spore survival over a seven-day period except in the 0.5 M phosphate
buffer with acid added. Therefore, the 2.0 M phosphate buffer was
selected to be the sample collection reagent.
To assess the potential for contamination of emission samples from
613
-------
spores in the ambient air, samples of the ambient air entering the
combustion chamber were collected at two incinerator sites using midget
impingers and personnel sampling pumps. For each site, three separate
midget impinger trains were charged as follows: Trains 1 and 2 - 0.05 M
phosphate buffer and Train 3 - 0.05 M phosphate buffer spiked with ~103
indicator spores. Each train collected nominally 0.2 m^ of sample during
a 2-h sampling period. The impingers were sterilized by autoclaving or
rinsing with 90% ethanol (EtOH) to see if a difference in the two
techniques could be detected. Recovered samples were heat shocked,
aliquotted, plated, incubated at 65°C, and counted after about 18 h. The
results showed that there was no background contamination of any type and
the use of a 90% EtOH rinse instead of autoclaving for sampling equipment
sterilization did not affect the background contamination. It was
concluded that background contamination is not a significant concern,
however, the final draft protocol specifies collection of one background
sample with each series of emission samples.
A problem anticipated with sampling in high temperature (1000°F to
25000F) incinerator stacks using a Method 5~type impinger train is the
collection of some of the spores on the walls of the probe where they
would be exposed to high temperatures during the remainder of the sampling
run. To resolve this problem, a water-cooled probe assembly was designed
and constructed (see Figure 1). In this design, the probe liner along
with an S-type pitot and a thermocouple conduit are completely jacketed by
a stainless steel cylinder to allow water cooling by recirculating ice
water or flow-through tap water.
Field Test Evaluation
Field evaluation testing of the candidate protocol was conducted to
evaluate it under field conditions and assess the accuracy of the sampling
phase. Early field evaluation testing revealed several problems with the
protocol. This section begins with a description of the field evaluation
test site, followed by the highlights of the problems encountered and
their resolution, and is completed by a detailed description of and
results from the latest field evaluation test.
A medical waste incinerator at a North Carolina hospital was
selected for the method evaluation testing because (1) it is a state-of-
the-art incinerator for its size, (2) of cooperation of facility
personnel, (3) of its proximity to EPA/Entropy/RTI, which facilitated test
program modifications, (4) it is representative of smaller hospital
incinerators, and (5) it was the site of previous background contamination
sampling. The unit is a Cleaver-Brooks pyrolytic incinerator with a
single combustion chamber and a high-temperature (2000°F) retention
chamber to increase residence time of the gases.
The samples from an initial field evaluation test revealed two
problems with the sampling methodology which required correction.
Immediately following the first test run conducted over a two-h period,
the pH of the sampling buffer in the first impinger of each of the four
trains was measured; all were pH 3 or less. Based on the buffering
capacity of the sampling buffer and the volume of gas sampled, the HC1
concentration in the sample gas approached 2000 ppm. The remaining
sampling runs were reduced to one h to ensure that the sampling buffer
could maintain the sample pH between 5-5 and 7-5; for all but one of the
remaining 12 samples, it was between 6 and 7• A second problem,
encountered in the course of filtering sample aliquots for analysis, was
build-up of a shiny green/gray film on the filter, most likely a
614
-------
by-product of the interaction of HC1 in the sample gas with the stainless
steel nozzle and probe liner. When a second species of spores spiked into
some of the trains to assess sample recovery could not be detected on the
filters, it was determined that the protocol should be modified to utilize
non-metal components for all sample-exposed surfaces. A second field
evaluation test revealed that if a spore spike was to be used to assess
the accuracy of the sampling protocol, it would have to be introduced at
the probe nozzle. Subsequent testing utilizing this spore spiking
technique indicated that better spike recoveries might be possible by
improving the environment in the sampling probe. This prompted
development of the probe gas buffering system described below.
Sampling for the final field evaluation test involved six 3
runs conducted over two days. The sampling train was modified to allow
introduction of sampling buffer into the quartz liner of the water-
cooled probe at the point of introduction of the spore spiking solution
(see Figure 1). The buffer was supplied from a reservoir external to the
train at a rate of approximately 20 to 30 ml/'min. The volume of the first
impinger of the sampling train was increased to two liters to provide
capacity for the buffer introduced at the probe. Approximately 2.5 x 101*
spores of B. stearothermophilus were injected into the probe through a
septum over the course of each sampling run.
Prior to use, sampling equipment was rinsed with $0% EtOH. Sampling
was conducted using Method 5 procedures except that it was conducted at a
single point and at a constant rate based on stack conditions from
previous field evaluation testing. A temperature traverse of the inside
of the water-cooled probe was performed under sampling conditions. An
extra impinger containing 100 ml of sampling buffer was placed prior to
the silica gel impinger for two of the six runs. The contents of this
impinger were recovered and analyzed separately to determine the sampling
train collection efficiency. Analysis of each sample involved (1) heat
shocking to eliminate non-spore microorganisms, (2) aliquotting of the
sample into 1-, 10-, and ~100-ml portions, (;>) vacuum filtration through
0.2 urn cellulose nitrate filters, (4) agar plating of filters, (5)
incubation of filters at 65°C, and (6) enumeration of the colonies on the
filters. The recovery of the spiked spores was calculated as a percentage
of the original spike amount indicated by thf> control sample. The
precision of the sample analysis was expressed as a relative standard
deviation between the results for replicate aliquots.
The spore spike recoveries (or method accuracy) ranged from 37# to 133$
and centered around 60% . The RSDs for the 10-ml aliquots ranged from 3% to
9D# and were typically 5% to 30%; the RSDs for the 1-ml and nominally 100-ml
aliquots were significantly higher. The nominally 100-ml aliquots (~75 ml for
the probe rinse samples and ~600 ml for the impinger samples), yielded colonies
which were more irregular, harder to see, more difficult to count, and often
more prevalent away from the ash particles ce.ught on the filter. They also
showed reduced sample recoveries compared to the 10-ml aliquots, possibly as a
result of growth inhibition caused by the ash. The 1-ml aliquots showed more
variability as would be expected, because the spore counts were closer to the
detection limit of the analytical technique. No colonies and only two colonies
were present on the filters from the extra iirpinger samples. This represented
significantly less than one percent of the tctal spore sample indicating good
collection efficiency and justifying elimination of the backup filter from the
candidate train. The temperature traverse demonstrated a rapid reduction in
sample gas temperature from 1850 °F at the entrance to the probe to 70° F at the
first impinger. The probe buffering system was shown to be of benefit in
improving the spike recoveries.
615
-------
Conclusions
Considering that the candidate protocol will be used to calculate a
destruction efficiency which is an order of magnitude measurement, the method
accuracy of approximately 60% is adequate for the purposes of the method. The
final draft method reflects the results of the laboratory and field evaluation
testing and incorporates these key elements: a water-cooled quartz-glass probe
with buttonhook nozzle and probe gas buffering system; a large first impinger
to provide a reservoir for the introduced buffer; measurement of the first
impinger pH; an incinerator waste spike of at least 1012 indicator spores;
disinfection/sterilization of the sampling train components using ethanol; use
of a 2.0 M phosphate buffer impinger solution; use of the filtering technique
for spore enumeration; collection of a background sample; and aliquotting of
the sample to minimize matrix interference, provide measurements in a suitable
analytical range, and provide an indication of analytical precision.
References
1. M. S. Barbeito, L. A. Taylor, and R. W. Seiders, Applied Microbiology, Vol.
16, March 1968, p. 490.
2. M.S. Barbeito and G. G. Gremillion, Applied Microbiology, Vol. 16, No. 2,
February 1968, p. 291.
3. M. S. Barbeito and M. Shapiro, J. Pled. Prtmatol, Vol. 6, 1977, P- 264.
4. N. A. Kelly, Masters Thesis submitted to University of Illinois, Health
Sciences Center, 1982.
5- B. L. Jackson and S. Ziengenfuss, Paper presented at the MASS-APCA
Technical Conference, Atlantic City, New Jersey, November 1987-
6. R. J. Allen, G. R. Brenniman, R. R. Logue, and V. A. Strand, JAPCA, Vol.
39, No. 2:164-168, 1989.
7. "Laboratory Procedures for Microbial Analysis," from "Biomedical Waste
Incinerator Study," Ontario Ministry of the Environment, Laboratory
Services Branch, Rexdale, Ontario, 1989.
8. Personal communication with G. Horsnell, Ontario Ministry of the
Environment, Laboratory Services Branch, Rexdale, Ontario, March 31.
Quartz-Glass
Probe Liner /
Button-Hook Nozzle
1 -Liter Reservoir
of 2.0 M
Phosphate
Buffer
1/8'Teflon Lirw
(Reaching to rearol
button-hook nozzle)
Z Pilot
Manometer
....jr-Coolad or
Probe Differential
Pressure
Gaugels)
Reverse-Type
Pilot Tube
100mL Empty *» Silica »
Gel
Phosphate Phosphate ^*»*#»«'»,
Butler >•*+»»*»***•
;ure 1. Schematic of sampling train for determining pathogen destruction efficien'
616
-------
PRIORITY POLLUTANT METALS IN RESPIRABLE AND
INHALABLE (PM10) PARTICLES
James E. Houck
OMNI Environmental Services, Inc.
Beaverton, Oregon
Lyle C. Pritchett, John G. Watson, and
Judith C. Chow
Desert Research Institute
Reno, Nevada
Respirable (<2.5p.) and inhalable (<10pi) particles have been collected
from over eighty area and point sources in California, Colorado, and
Montana. These sources have included a variety of industrial stacks,
windblown and road dusts, vehicular exhausts, residential wood combustion,
and emissions from agricultural activities. Custom sampling procedures
have been developed for the collection of representative samples from each
of the sources. Dilution tunnel sampling, ground-based fugitive emission
sampling, and bulk sampling/resuspension protocols have been established.
Multi-element analysis has been conducted for thirty-eight elements by
x-ray fluorescence spectrometry. Among the elements measured, the priority
pollutant metals of antimony, arsenic, cadmium, chromium, copper, lead,
mercury, nickel, selenium, silver, thallium, and zinc were quantified. The
sampling and analytical methods which have been developed are discussed and
a comparison of priority pollutant metal concentrations for selected
sources is presented.
617
-------
Introduction
Priority pollutants are listed in Section 307(a)(l) of the 1977 Clean
Water Act. Among the 129 priority pollutants listed in the Act are 13
metals. These priority pollutant metals are antimony, arsenic, beryllium,
cadmium, chromium, copper, lead, mercury, nickel, selenium, silver,
thallium, and zinc. Subsequent to their listing in the Clean Water Act,
they have been used as target metals to assess pollution in all media and
are, for example, listed as hazardous substances in response to CERCLA (40
CFR Section 302.4}.
Emission rates of priority pollutant metals from point, area, and
fugitive aerosol sources are frequently of concern. Similarly, the
documentation of metal concentrations in airborne particles in indoor and
outdoor environments has been the subject of numerous investigations.
Assessments of environmental impacts, community-wide health risks, indus-
trial exposure and regulatory compliance often require their measurement.
During the last decade, chemical mass balance (CMB) modeling for air
pollutant source identification and apportionment has gained wide
acceptance. To support this air quality model, specialized sampling
instrumentation, protocols, and analytical procedures have been developed
to accurately quantify multiple pollutants emitted from aerosol sources.
Among the 13 priority pollutant metals, 12 have been routinely and
accurately measured as part of the process. (Only beryllium is not
routinely measured.) Numerous CMB source profile libraries have been
compiled listing, among the measured constituent, the 12 metals.
A description of the sampling instrumentation, protocols, and
analytical procedures appropriate for accurate priority pollutant metal
quantification in PM2.s and PM10 particles from a variety of source types is
presented in this paper. In addition, some representative data contained
in CMB source profile libraries prepared from work conducted in California,
Colorado, and Montana are presented.
Methodology
The objective of aerosol sampling and analysis for the development of
source profiles for use in CMB modeling is to obtain an accurate and
representative chemical composition. A streamlined set of sampling,
analysis, and data reduction protocols has been developed to achieve this
objective.1 Since the physical and chemical environments encountered
during source sampling differ from source to source, several source
sampling instrument systems and protocols have been developed. In terms of
sampling requirements, sources can be grouped into three major categories.
These are: (1) point and combustion sources; (2) process fugitive sources;
and (3) passive fugitive sources.
Point sources (including combustion sources such as residential wood
combustion [RWC] and vehicular exhaust) represent a special problem for
source sampling. The alteration in particulate chemistry and size
distribution which occurs when high-temperature emissions cool and mix with
ambient air requires that a dilution/cooling tunnel be used prior to
aerosol sample collection. Condensation, agglomeration, volatilization,
and secondary chemical reactions can all modify the character of source
particles. A number of dilution source sampling instruments have been
developed2. Factors taken into consideration in dilution source sampler
designs include variable dilution ratios, interfacing with size-
categorizing equipment, portability, inert construction materials,
618
-------
isokinetic sampling, and the ability to be assembled in a variety of
geometric configurations to conform to real-world space constraints. A
detailed description of a dilution source sampling system which has been
used to sample numerous sources is presented in reference 3.
Process fugitive sources have been sampled in two general ways.
These are (1) custom siting of ground-based sampling equipment or (2) plume
sampling with balloons or aircraft. Unducted emissions from industrial
complexes, intermittent emissions from slag, pouring, and dust generated by
heavy industrial equipment, are examples of process fugitive sources.
Instrumenting requirements can be as simple as using commercial ambient
samplers, perhaps placed on a platform or tower, or as sophisticated as
custom-designed, ultra-lightweight samplers for use with helium balloons.''
A key criterion for sampling process fugitive emissions using either
ground-based or airborne samplers is that the source of interest dominates
the particle concentration during sample collection.
The passive fugitive dust category fcr the purposes of source
sampling strategies can be either wind-blown dust from open areas or
suspension of particles by traffic or heavy equipment. Detailed standard
protocols have been developed for the sampling and analysis of such dust.5
Sampling protocols include procedures for (1) sampling paved roads; (2)
sampling unpaved areas which have a surface layer with a distinct chemical
character due to anthropogenic impact (e.g., unpaved roads and parking
lots); and (3) sampling of dust sources with a relatively homogeneous near-
surface chemical composition (e.g., tilled agricultural soils, native
soils, and bulk storage piles). While the actual physical collection of
the dust samples is relatively simple, ensuring representative samples is
not. Compositing samples is a useful technique to ensure that
representative chemical source profiles are produced. Collection of sub-
samples at regular intervals along a roadway or at various points in an
agricultural field or from different storage piles is a reasonable approach
to compositing. Once bulk samples area collected, laboratory drying at low
temperatures, sieving, resuspension, and particulate collection with size-
segregating samplers can be conducted.3
A variety of analytical procedures have been developed to chemically
characterize the particulate material contained in source samples. Most
useful for priority pollutant metals is x-ray fluorescence spectrometry
(XRF), which can quantify all priority pollutant metals except beryllium.
While the detection limit of the technique varies from element to element
and depends on the mass loading as well as the overall chemical matrix, the
detection limits are typically at the hundredth-of-a-percent level.
Results and Discussions
Table I lists priority pollution metals in PMJO particles from a
number of sources compiled as part of CUB source profile libraries.3'6'7
Uncertainties and detection limits listed in Table I are not only due to
the analytical techniques, but represent sample variability in replicate
samples or sequential sampling runs. In addition, some of the values,
propagated uncertainties, and detection limits were produced by "blending"
results from individual sub-classes of emissions. For example, the
vehicular exhaust data are a blend of unleaded, leaded, and diesel
emissions (30Z, 202, and 502, respectively) and the RWC data are a blend of
emissions from a number of woodstoves and f-.replaces.
Upon review of the data that has been compiled in the various source
libraries,3'6'7 two key observations can be made. One is that many indiv-
619
-------
idual sources can have specific priority pollutant metals in PM10 particles
at the percent level. For example, as illustrated in Table I, mercury from
municipal garbage incineration, zinc from combustion of wood in hog fuel
boilers, and nickel from crude oil combustion can all be above the percent
level in PM10 particles. The other observation is that some of the priority
pollutant metals occur at above detection limits in ubiquitous area sources
such as road dust, agricultural dust, and vehicular exhaust. The metals of
copper, chromium, and nickel are at low but measurable concentrations in
dusts, which can be due simply to their crustal abundance. Lead and zinc
in urban road dust are, on the other hand, clearly anthropogenic in origin,
and are from the use of leaded gasoline and tire wear. Interestingly, and
notable in terms of human health impact, lead and zinc collected in urban
road dust is most highly enriched in respirable (PM2.5) or finer particles
(Figures 1 and 2).
In summary, source sampling methods and analytical procedures have
been developed in the past ten years to accurately measure priority
pollutant metals in PM10 as well as other size particles from particulate
sources. While the sampling and analytical activities have been primarily
to support CMS modeling, the same techniques can be used for measuring
priority pollutant metals for the evaluation of health and environmental
impacts. In addition, published CMS source profile libraries can provide
an existing and useful data base for such evaluations.
References
1. J. Core and J.E. Houck, (eds.), Pacific Northwest Source Profile
Library. Sampling and Analytical Protocols, final report, Oregon
Department of Environmental Quality, Portland, Oregon, 1987.
2. J. Shah, R. Johnson, and J.E. Houck, Source Characterization Using
Tethered Balloons, Transactions of Receptor Models in Air Resources
Management, APCA Specialty Conference, p. 334-345, 1989.
3. J.E. Houck, J.C. Chow, J.G. Watson, C.A. Simons, L.C. Pritchett, J.M.
Goulet, and C.A. Frazier, Determination of Particle Size Distribution
and Chemical Composition of Particulate Matter From Selected Sources
in California, final report, California Air Resources Board,
Agreement No. A6-175-32,1989.
4. J.E. Houck, J.A. Cooper, and E.R. Lars.on, Dilutions Sampling for
Chemical Receptor Source Fingerprinting, Proceedings 75th Annual Air
Pollution Control Association meeting, paper 82-61M.2, 1982.
5. J.E. Houck, J.C. Chow, and M.S. Ahuja, The Chemical and Size
Characterization of Particulate Material Originating from Geological
Sources in California, in Transactions Receptor Models in Air
Resource Management, p. 322-333, 1988.
6. J.G. Watson, J.C. Chow, L.W. Richards, W.D. Neff, S.R. Anderson, D.L.
Dietrich, J.E. Houck, and I. Olmez, The 1987-88 Metro Denver Brown
Cloud Study, final report. The 1987-88 Metro Denver Brown Cloud
Study, Inc., Denver, Colorado, 1988.
7. J.M. Goulet, J.E. Houck, H.W. Robbins, J.J. Olsen, L.C. Pritchett,
and C.A. Frazier, Champion International Corporation Chemical Mass
Balance Source Sampling Report, final report, Champion International
Corporation, Libby, Montana, 1989.
620
-------
Table I. The priority pollutant metal content of PM10 particles from selected sources1.
Priority
Pollutant
Metal
Antimony
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Silver
Thallium
Zinc
Area Sources
Urban
Street Dust
(Fresno,
CA)
<0.03
<0.04
<0.02
0.03010.003
0.02010.002
0.26±0.03
<0.008
0.01110.001
<0.002
<0.02
ND
0.17±0.02
Agricultural
Soil Dust
(Visalia,
CA)
<0.01
<0.003
<0.008
0.03010.003
0.00610.001
O.OlOiO.OOl
<0.003
0.00610.001
<0.001
<0.008
KD
0.02610.002
Vehicular Exhaust2
(301 unleaded, 201
leaded, 501 diesel]
Denver, CO)
<0.08
<0.21
<0.04
«0. 01
sO. 01
1.3±O.A
<0.02
<0.07
<0.01
<0.04
ND
0.11±0.06
Residential
Uood
Combustion
(Denver, CO)
<0.01
<0.002
<0.008
<0.001
<0.001
<0.003
<0.003
<0.005
<0.001
<0.007
ND
<0.04
Point Sources
Garbage
Incinerator
(CA)
<0.1
<0.03
<0.06
0.0410.02
0.0410.01
0.1610.04
1.111.0
0.01410.006
<0.007
<0.06
ND
0.510.2
Hog Fuel
Boiler (MT)
<0.01
0.02910.004
<0.009
0.00410,001
0.04210.003
0.07110.006
<0.002
0.001310-0003
0.002610.0007
<0 . 008
<0.003
1.2210.09
Crude Oil
Combustion (Steam
Injection, Kern
R. Oilfield)
<0.01
0.01910.002
<0.005
<0.03
sO. 01
<0.003
<0.002
2.6±0.3
0.02710.004
<0.004
ND
0.0210.01
Coal
Combustion'
(Baghouse;
Denver, CO)
<0.08
<0.02
<0.05
0.015*0.004
0.02±0.01
0.10±0.02
<0.02
0.00810.002
0.00710.003
<0.04
ND
0.0810.05
OS
1. All values are weight percent; each source profile is a mean of numerous samples or a "blend" of
various individual profiles. Uncertainties are propagated values, not analytical uncertainties.
2.
The vehicle exhaust and coal combustion values are for PM2.5.
-------
w
E
P
E
R
C
E
N
T
0.5
0.4
G
H 0.3
T
0.2
0.1
<1 1-2.5 2.5-10 >10
AERODYNAMIC DIAMETER (microns)
Figure 1 Weight percent lead vs. size categories for urban road dust.
w
E
I
G
H
T
P
E
R
C
E
N
T
0.3
0.25
0.2
0.15
0.1
0.05
<1 1-2.5 2.5-10 >10
AERODYNAMIC DIAMETER (microns)
Figure 2 Weight percent zinc vs. size categories for urban road dust.
622
-------
ATMOSPHERIC TRANSPORT OF TOXIC CHEMICALS AND THEIR POTENTIAL
IMPACTS ON TERRESTRIAL VEGETATION: AN OVERVIEW
Thomas J. Moser1, Jerry R. Barker1, and David T. Tingey2
NSI Technology Services Corporation1
U.S. Environmental Protection Agency2
U.S. EPA Environmental Research Laboratory
Corvallis, Oregon
Numerous anthropogenic sources emit a large variety and
quantity of toxic chemicals into the atmosphere. Through the
processes of atmospheric transport and deposition, toxic
chemicals have found their way to remote environments far from
emission sources. Recent data strongly suggests that the
enriched concentrations of several contaminants detected in
the air, water, soil and biota of rural and remote
environments are the result of long-range atmospheric
transport from sources located in temperate and sub-tropical
latitudes of North America and Eurasia. Many of these
chemials are persistent, bioaccumulate and remain biologically
harmful for long periods of time. Although air toxics have
been primarily considered an urban health problem, there is
increasing concern among scientists that adverse ecological
impacts may result from their deposition into terrestrial
ecosystems and their subsequent exposure of plants. The
chronic exposure of vegetation to low concentrations of air
toxics may result in sublethal effects such as decreased plant
productivity and vigor, which may culminate in changes in
plant communities and ecosystem structure, composition and
function. This paper will present an overview of air toxics
emissions, atmospheric transport and deposition, their
potential effects on terrestrial vegetation, and recommend
research needs.
623
-------
Introduction
Airborne pollutants can be broadly defined as any
chemical occurring in the atmosphere that may pose a threat to
human health or the environment. This broad definition
includes an array of chemicals ranging from the well-studied
criteria pollutants to the less-studied radiatively-important
trace gases. However, to provide a more focused definition
for this paper, air toxics refers to the following groups of
airborne substances: (1) synthetic industrial organics, (2)
agricultural pesticides, (3) trace metals, (4) organometallic
compounds, and (5) non-metallic inorganics.
Since the release of the Environmental Protection
Agency's Toxic Release Inventory estimates for 19871, there
has been a heightened concern over the nation's air quality.
Primarily this concern has been directed at human health
effects in industrial-urban areas. The fact that many
airborne chemicals pose hazards to human health is only one
aspect of the problem. The continued deposition of airborne
toxic chemicals on a regional to global scale will impact
public welfare if it results in adverse impacts to the
structure and function of sensitive terrestrial and aquatic
ecosystems. Although airborne toxic chemicals pose threats to
both terrestrial and aquatic ecosystems, this paper will limit
its discussion to terrestrial vegetation.
Emission Sources
Airborne chemicals are emitted into the atmosphere from a
large number and variety of point- and area-sources.
Anthropogenic emissions emanate from industrial, urban and
agricultural sources such as chemical, metal, plastic, and
paper/pulp industries; fossil fuel processing plants; motor
vehicles and aircraft; municipal waste incinerators;
agricultural practices such as pesticide usage and field
burning; and small businesses such as dry cleaners. Emissions
of toxic chemicals into the atmosphere may occur directly by
the deliberate or inadvertent releases from industrial or
urban sources, or indirectly through volatilization following
the deliberate or accidental discharge of chemicals into water
or soil resources. Also, considerable amounts of toxics enter
the atmosphere from wind drift and volatilization after
agricultural chemical applications.
Industry is probably the major anthropogenic source of
airborne toxic chemicals. Approximately 65,000 chemicals are
used worldwide for industrial purposes. Many of these
chemicals eventually are emitted into the atmosphere. The
1987 Toxic Release Inventory (TRI) reported that industries
within the United States emitted over 1.2 billion kilograms
of toxic chemicals into the atmosphere.1 The TRI
underestimated the actual air emissions as it did not include
air emissions from numerous area sources (e.g., agriculture,
households, motor vehicles), industrial categories such as
petroleum tank farms, companies with less than 10 employees,
or urban businesses (e.g., dry cleaners).
624
-------
A secondary source of airborne toxics is the application
of pesticides. Atmospheric loads of pesticide residues
resulting from wind drift, as well as volatilization from soil
and plant surfaces after their application to agricultural,
forest, industrial and household lands are likely significant.
Over 455 million kilograms of pesticide active ingredients are
used annually on 16% of the total land area of the United
States.3 Agricultural lands account: for approximately 75% of
the pesticide usage in the United States. The magnitude of
pesticide active ingredients entering the atmosphere during
and following application are poorly known, but are estimated
to range between 30-55%.4
Atmospheric Processes
Once chemicals are airborne, they are subjected to the
prevailing atmospheric conditions. Wind, precipitation,
humidity, clouds, fog, solar irradiation, and temperature
influence the environmental fate of air toxics.2'5 The
complex reactions within the atmosphere that are driven by
chemical processes such as hydroxyl scavenging or solar
irradience may result in the formation of products that can be
as toxic or more so than the parent compounds. On the other
hand, transformation reactions may also render a toxic
substance harmless.
The atmosphere is a major pathway for the transport and
deposition of the air toxics from emission sources to the
terrestrial ecosystem receptors - vegetation and soil. The
prevailing meteorological conditions; and the physio-chemical
properties of the chemicals will dictate atmospheric residence
times and deposition velocities to the receptors.2
Atmospheric residence times depend on such characteristics as
mode and rate of emission, atmospheric transformations,
physical state (gas, solid, liquid), particle size and
chemical reactivity.2 Thus, airborne pollutants may be
deposited close to their sources or be carried great distances
before being deposited into remote ecosystems.
The movement of airborne chemicals downwind from point
sources has received a great amount of attention since the
early 1900's due to the damaging effects on vegetation that
occurred within the plumes.6 During the last 10 to 20 years,
however, the phenomenon of long-range atmospheric transport
has been implicated in the wide distribution of anthropogenic
contaminants on regional and global scales. Industrial
organic compounds, trace metals and pesticide residues have
been detected in the vegetation of remote terrestrial
ecosystems such as the Arctic,7 Antarctic,8 forests,9 and
peatlands.10
Environmental Partitioning and Vegetation Exposure
Terrestrial plants are exposed to toxic chemicals through
the environmental media of air, water and soil. The
atmosphere, however, is of prime importance due to its
potential for pollutant dispersal on a regional to global
625
-------
scale, its ability to move pollutants rapidly, and its dynamic
nature. After pollutant deposition to terrestrial ecosystems,
the fate of the toxic compounds depends on their partitioning
coefficients.
The environmental partitioning of pollutants within
ecosystems will dictate their potential ecological impact to
vegetation and other biota.11 For example, trace metals tend
to accumulate on soil surfaces via their adsorbtion to organic
matter. Trace metal accumulation may reduce plant growth and
vigor through the disruption of nutrient uptake by the roots
and decreased organic matter decomposition. Gaseous chemicals
reside in the atmosphere with the potential to disrupt plant-
leaf biochemical processes after absorption through the
stomata or cuticle. Because of the lipophyllic nature of many
synthetic organics, the waxy cuticle of vegetation may
accumulate high levels of these substances. Transfer of toxic
chemicals among ecosystem compartments will occur. Trace
metals may be absorbed by plant roots or deposited onto the
leaves and then transferred to the soil through deciduous
tissue loss and decay. Contaminants may be passed along food
chains through herbivory with the potential for
biomagnification. The deposition of airborne toxic chemicals
into agricultural ecosystems has the potential to contaminate
human food resources.
Impacts on Terrestrial Vegetation
Scientists have recognized that airborne pollutants can
adversely impact agricultural and natural plant communities by
reducing plant production and altering successional
pathways. 1'12 Emissions of sulfur dioxide, hydrogen
fluoride, trace metals and other toxics from pulp and paper
mills, ore smelters, and power plants have severely reduced
vegetation cover, biodiversity and ecosystem integrity
downwind from point sources.6'12 In addition to local plume
effects, atmospheric pollutants can also cause regional damage
to plant communities through exposure to chemical oxidants
such as ozone and peroxyacetyl nitrates or acid
precipitation.12' 1-J
The potential biological effects of air toxics on
terrestrial vegetation are numerous and mediated through
individual plants to the community and ecosystem.11 The type
and magnitude of these effects will depend on the pattern of
exposure (e.g., duration, concentration, frequency, season)
individual plants receive, their sensitivity to the polluant,
and the phytotoxicity of the chemical. When an airborne toxic
chemical is introduced into a plant community some plants will
be more affected than others depending on individual
tolerances endowed by their genotype, as well as their
phenology and various modifying microclimatic variables. The
sensitive plants or species are no longer able to compete
adequately with the tolerant plants or species and will be
partially or completely replaced. Those plants that survive
and persist in contaminated habitats are the result of their
tolerance or microhabitat protection.
626
-------
After pollutant absorption by the plants through the
leaves or roots, biochemical processes are the first site of
action. If enzymatic degradation detoxifies the pollutant
then no injury will occur. However, if enzymatic action
cannot render the pollutant or its metabolites harmless, then
alterations in plant metabolism may result in foliar injury,
altered carbohydrate and nutrient allocation, reduced growth
and reproductive capability13. The degree of impact to the
plant will depend on the toxicity of the pollutant and its
exposure pattern. Acute exposures usually cause observable
morphological damage, such as leaf lesions, stunted growth or
even death. Plant death resulting from acute exposure is
usually localized when it does occur; resulting from an
inordinate amount of toxic chemical exposure via an accidental
release or pesticide wind drift.
However, chronic, sub-lethal exposures may not induce
observable morphological damage, but rather alter biochemical
pathways which can result in decreased vigor and productivity,
altered phenology, loss of tissue or reduced reproductive
potential. Altered physiological processes will cause a loss
of vigor and render the plant more susceptible to insect
damage or disease. Decreased reproduction will impact the
population through the loss of new recruitments to the plant
community. With continual exposure, even at sublethal
concentrations, sensitive plant populations may decrease in
numbers allowing tolerant species to become dominant. Thus,
shifts in plant community structure and composition could
result in decreased biological diversity and altered ecosystem
functions.
Plant damage resulting from acute air toxic exposures are
usually limited in time and space as a result of control
technology and legislation. However, sublethal, long-term
plant exposure to airborne pollutants may predispose
vegetation to other natural stressors and induce damage or
mortality. Even though air toxic damage may not cause
permanent functional loss, the diversion of biochemical
resources to repair the injury will inhibit normal plant
functions. Thus, air toxic induced physiological stress may
predispose a plant to other stressors such as frost, drought,
insects, or disease. Some scientist propose that the
widespread forest tree decline is not the result of a single
agent but an interaction among chronic exposures of air
pollutants and natural stresses.
Air toxics may also pose indirect effects on vegetation
by directly affecting other organisms which are critically
associated with the plants. Soil microorganisms and
invertebrates are critical in ecosystems for litter
decomposition and nutrient cycling. An accumulation of trace
metals within the "0" horizon of the soil may limit organic
matter decomposition and nutrient availability to plants.
Many plants rely on insects for pollination. Airborne
pollutants from the use of insecticides can reduce non-target
insect populations, resulting in inadequate flower pollination
and subsequent seed set.
627
-------
The effects caused by air toxics on vegetation can be
extended to the animals within the ecosystem. Reduced plant
cover and habitat quality will result in animals being more
susceptible to predation and disease. Adequate birthing sites
may be reduced because of changes in vegetation structure and
cover. Animal forage that is contaminated may result in
decreased population size or starvation. Toxic chemicals may
be passed along food chains with the potential to reduce the
health of herbivores or bioaccumulate within predators.
Animal populations will respond to such habitat changes
through decreased reproduction, emigration or mortality. Even
though the populations of certain species may increase because
of habitat changes that are favorable to them, the biological
diversity of the overall ecosystem will still decrease.
Research Needs
Air toxic chemicals introduced into plant communities can
cause effects ranging from the biochemical level to changes in
plant community structure and composition. The effects of
acute exposure to plants are well documented. However, a
paucity of information exists on the effects of long-term,
chronic exposures of air toxics to vegetation. Several areas
of research would provide data to quantify vegetation
responses to chronic air toxic exposure:
1. Given the large number of toxic chemicals emitted into
the atmosphere, research is required to identify and
prioritize the most critical airborne contaminants
and sensitive ecosystems. A comprehensive computer-
based system would be useful for conducting preliminary
risk assessments of the numerous airborne toxic
chemicals and their effects on vegetation and provide
research guidance.
2. Quantify and model the exposure, deposition velocity and
absorption of air toxics to plants.
3. Determine the biochemical and physiological responses of
plants to chronic exposures to air toxic chemicals and
develop exposure-response functions. This research then
could be extended to quantify the response of plant
populations and simulated plant communities.
4. Initiate long-term studies to determine sensitive
elements of plant community structure and function that
would lead to significant change and degradation from
air toxic exposure. This research would identify
unacceptable change in plant communities and identify
early warning signals.
References
1. "The Toxic Release Inventory: A National Perspective"
U.S. Environmental Protection Agency, EPA560/4-89-005,
Washington, D.C., 1989.
628
-------
2. W.H. Schroeder, D.A. Lane, "The fate of toxic airborne
pollutants," Environ. Sci. Technol. 22: 240 (1988).
3. D. Pimentel, L. Levitan, "Pesticides: amounts applied and
amounts reaching pests," BioScience 36: 86 (1986).
4. T.E. Waddel, B.T. Bower, "Managing agricultural chemicals
in the environment: the case for a multimedia approach,"
The Conservation Foundation, Washington, D.C. 1988, pp 42-
43.
5. T.E. Bidleman, "Atmospheric processes," Environ. Sci.
Technol. 22: 361 (1988).
6. A.G. Gordon, E. Gorham, "Ecological effects of air pollu-
tion from an iron-sintering plant at Wawa, Ontario," Can.
J. Bot. 41: 1063. (1963).
7. W. Thomas. "Accumulation of airborne tr4e pollutants by
Arctic plants and soil," Wat. Sci. Tech. 18: 47 (1986).
8. E. Bacci, D. Calamari, C. Gaggi, R. Fanelli, S. Focardi,
M. Morosini, "Chlorinated hydrocarbons in lichen and moss
samples from the Antarctic Peninsula," Chemosphere 15:
747. (1986).
9. G. Eriksson, S. Jensen, H. Kylin, W. Strachan, "The pine
needle as a monitor of atmospheric pollution," Nature
341: 42 (1989).
10. R.A. Rapaport, S.J. Eisenreich, "Historical atmospheric
inputs of high molecular weight chlorinated hydrocarbons
to eastern North America," Environ. Sci. Technol. 22:
931. (1988).
11. D.A. Weinstein, E.M. Birk, "The effects of chemicals on
the structure of terrestrial ecosystems: Mechanisms and
patters of change", in S.A. Levin, M,A, Harwell, J.R.
Kelly, K.D. Kimball (eds.), Ecotoxicology: Problems and
Approaches. Springer-Verlag, New York. 1989, pp. 181-212.
12. J.J. MacKenzie, M.T. El-Ashry, Air Pollution1s Toll on
Forests and Crops, Yale University Press, New Haven.
1989, pp. 1-21.
13. R. Guderian, Air Pollution by Photochemical OxIdants:
Formation, Transport. Control. and Effects on Plants,
Springer Verlag, New York. 1985, pp. 129-169.
629
-------
FLUORIDE PHYTOTOXICITY: PAST, PRESENT AND FUTURE
David C. MacLean
Boyce Thompson Institute, Cornell University
Ithaca, New York 14853 USA
Fluoride is markedly more phytotoxic than any of the major air pollutants and its
deleterious effects on crops, forests and other vegetation have been chronicled for more
than a century. Plants can also serve as vectors for the transfer of fluoride from the
atmosphere to animals. The accumulation of high concentrations of fluoride in foliar
tissues may be hazardous if ingested by herbivores. During the past two decades,
improvements in emission control technology have dramatically reduced the incidence
and severity of fluoride-induced plant damage. These reductions have shifted the
emphasis of environmental concerns from the direct to the indirect and subtle effects of
fluoride on vegetation, plant communities and even ecosystems. This review, in addition
to providing a summary of the direct and indirect effects of fluoride on plants, also
includes information on the factors that alter the plant's responses to exposure. The aim
of future research will be to provide the information necessary to establish realistic air
quality guidelines or standards for atmospheric fluoride that will allow industry and
agriculture/forestry to coexist.
630
-------
INTRODUCTION
Of the major air pollutants, fluoride is the most toxic to vegetation and its
deleterious effects have been the subject of controversy and concern for more than a
century. The extent and severity of the problems associated with atmospheric fluoride
increased with industrialization, peaked during and after World War II, and then, with
the imposition of governmental regulations and improvements in emission control
technologies, diminished. Research efforts closely paralleled this progression and were
largely responsible for the actions taken to cope with it.
These investigations included: (i) vegetation surveys around sources where plant
injury occurred; (ii) experimental exposures to atmospheric fluoride to reproduce and
validate the symptoms observed; (iii) development of methods to monitor fluoride in the
air and measure its accumulation in plants; and (iv) controlled exposures to determine
the relative susceptibility of various plant species, dose-response relationships, and the
factors that can alter or modify the plant's response to fluoride.1
A synthesis of the historical aspects of the .growth and diminution of the impacts
of atmospheric fluoride on vegetation provides a useful example of how a major
environmental problem can be dealt with. Thus a review of this kind serves two
purposes. First, it sheds light on the fluoride-vegetation problem per se; the sequential
events leading to plant responses and the consequences of these effects with respect to
impacts on growth, quality and yield. Second, it illustrates how the joint contributions
of science, technology, and policy can contribute towards the solution, at least in pan,
of a major environmental problem.
THE PROBLEM
Toxicity
Air pollutants such as ozone and other oxidants, sulfur dioxide, and oxides of
nitrogen are more important than airborne fluoride with respect to their total economic
effects on crops, forests and indigenous vegetation. Although fluoride is ranked lower
in terms of economic losses, on a concentration basis it is one to three orders of
magnitude more phytotoxic than any of those air pollutants.2
Plants are considerably more susceptible to fluoride than man or other mammals.
For example, the OSHA standard for man in the workplace environment is 2.5 mg F m"
3 for eight hours daily, five days per week over the working lifetime of a healthy adult.
At this level of exposure it is presumed that no deleterious effects on human health will
occur. For plants, however, the limits are much lower and air quality standards or
guidelines to protect vegetation have been adopted by several states in the U. S. and
provinces in Canada. The units of measure for these are not as 'mg', but as !(ig', and
range from limits of less than 4 ^.g F m"3 for 12 hours down to a mean concentration of
less than 1 ^ig F m"3 over 30 days.
In addition to the direct phytotoxicity of fluoride, vegetation can also serve as the
means for transferring fluoride from the atmosphere to animals. Plant foliage may
accumulate atmospheric fluoride to concentrations that may be injurious to livestock that
consume it.3
For these reasons, regulations to limit the harmful effects of fluoride usually
consist of one or more of three approaches: (i) restrictions of emissions at the source,
(ii) ceilings on the maximum allowable air concentrations or doses, and (iii) limits for
::he concentrations of fluoride in vegetatio.., especially forages.
631
-------
Sources
Sources of fluoride for plant uptake are soils, water and air, and fluoride
phytotoxicity is usually ascribed to the airborne fraction.4 Soils and waters naturally
contain fluoride to some degree, and this accounts for the concentrations found in plants
grown where there are no atmospheric fluorides. Background concentrations are usually
low (<20 ppm), but some plants are known to be "accumulators". Species of Camellia,
Deutzia, and to a lesser degree Gossypium, Carya and Cornus5'6 normally accumulate
from the soil high concentrations of fluoride in their leaves (60-1370 ppm) while other
plants in the same environment take up very little. These "natural" accumulations of
fluoride rarely, if ever, affect the plant per se, but if sufficiently high they may be
harmful when consumed by herbivores. Phytotoxic responses to more susceptible plants
do occur, however, when elevated concentrations of fluoride in foliage are derived from
exposures to fluoride in the atmosphere.4
Natural sources of gaseous and paniculate fluorides in the atmosphere include
volcanoes and fumaroles7, and these may cause serious damage to plants, animals and
ecosystems. This review, however, will concentrate on airborne fluorides of
anthropogenic origin. Essentially all industrial processes that utilize fluoride-containing
substances as raw materials or as fluxes, catalysts, etc. are potential sources of emissions
of fluoride to the atmosphere. Some examples are aluminum smelting, coal combustion,
the manufacture of phosphoric acid, phosphate fertilizers and feeds, steel, glass and frit,
and brick and ceramic products.4 The amount and composition of fluoride emissions
from a source will depend on the kinds and amounts of material manufactured,
processed or consumed and the efficiency of effluent controls.
Historical
The earliest reports of effects on vegetation, that have subsequently been ascribed
to fluorides, were associated with volcanic activity in Iceland more than 1,000 years
ago. More recent records from eruptions of Mt. Hekla in the 17th, 18th and 19th
centuries describe catastrophic effects on plants, animals and entire ecosystems from the
immediate effects of gaseous hydrogen fluoride (HF) and the lingering effects of fallout
of fluoride-containing ash.7
The main focus of this paper, however, is fluoride of anthropogenic origin.
Confirmed cases of injury to vegetation from industrial emissions began to appear in
Europe in the late 1800s. Copper smelters, superphosphate factories and glass works
were the sources of these early incidents and the involvement of fluoride was confirmed
initially by leaf analysis and later by experimental exposures to HF.8 In the United
States significant problems were first detected during World War II with the rapid
expansion of aluminum production to supply aircraft and other wartime needs; pollution
control had low priority. Post-war growth of aluminum smelting and of the production
of phosphate and superphosphate fertilizers increased fluoride emissions to the
atmosphere; but the increases were not proportional to production as the installation of
scrubbing equipment was beginning to take hold. Nevertheless, the post-war period saw
many complaints for damages to vegetation that resulted in significant and extended
lawsuits. Evidence presented in the trials revealed that there was little reliable
information about the effects of fluoride that either side could rely on. Both
government and industry began to support investigations to meet these needs and by the
late 1950s research programs were in place at several universities and institutions.9
Several of these programs are still in existence, and some of the concepts revealed by
research are summarized below.
632
-------
DIRECT EFFECTS ON VEGETATION
Uptake
All plant surfaces are capable of intercepting and adsorbing both gaseous and
paniculate fluorides. Gaseous forms readily enter che plant by diffusion through
stomata. Forms resident on leaf surfaces may penetrate the cuticle, but this is a
relatively slow process and its rate is dependent on the solubility of the adsorbed
fluoride and the physical and chemical characteristics of the foliar surface.5-10 A
significant proportion of fluoride-containing paniculate material deposited on leaves does
not penetrate.
Once fluoride has entered the leaf it dissolves in plant fluids and is translocated
acropetally via the transpiration stream to the margins and tips of leaves where it
accumulates and, as water is lost, concentrates. When toxic concentrations are achieved,
injury occurs.5
Progression of Effects
The initial fluoride-induced changes that occur within cells include altered
metabolism, distortions and disruptions of organelles, and breakdown of the tonoplast
and other membranes." Effects at each level of organization may affect the next higher
level. For example, cellular changes lead to symptoms and reduced assimilatory
capacity of leaves, and foliar injuries can result in effects on growth or reproduction of
the entire plant.4
Symptoms
The most obvious effect of fluoride on the plant is foliar injury, which has long
been used for diagnosis and evaluation in the field. Chlorosis from the direct effects of
fluoride on chlorophyll synthesis, and necrosis from the death of groups of cells are the
most common symptoms;11 but deformations from uneven leaf growth are also associated
with fluoride injury. The degree and pattern of leaf injuries are dependent on the
relative susceptibility of the plant and how the fluoride was accumulated. Symptoms
from chronic (long-term, low concentration) or recurrent exposures are quite different
than those induced by acute (short-term, high concentration) exposures. For example, in
chronic exposures where uptake and translocation are in balance, chlorosis and necrosis
first appear in apical and marginal areas of young, expanding leaves. During acute
exposures, which rarely occur in the field, uptake exceeds transport resulting in toxic
concentrations in localized areas of the leaf where irregularly-shaped necrotic lesions
develop.1
Growth and Yield
Fluoride-induced chlorosis may result in altered growth through reductions in the
rate or efficiency of photosynthesis. Foliar necrosis reduces the area of
photosynthetically active tissue and consequently may affect growth and yield.11'12
However, reductions in foliar surface area from exposure to fluoride or even from
clipping do not always affect growth and reductions in growth can occur without the
concomitant appearance of foliar symptoms.
The effects of fluoride on yield and on foliar injury are for some plants
independent. Exposure to HF reduced fruiting of pole bean13 and the yield of wheat14
and snap bean15 in plants that did not develop leaf symptoms. Disruptions in the
processes of pollination or fertilization have been suggested to explain these
633
-------
observations, but other factors may be involved as well (see below).
The direct effects of fluoride on vegetation are considered to be damaging if the
intended use of the plant is impaired. Foliar injuries or growth reductions that do not
affect the plant's aesthetic, economic or ecologic value are often of little concern.
INDIRECT EFFECTS ON VEGETATION
Plants as Vectors
Vegetation can serve as a vector for the transfer of fluoride from the atmosphere
to other components of an ecosystem. Much of the fluoride accumulated in leaves
subsequently enters the upper horizons of soil through litter decomposition. Foliar
fluoride may also be transported to insects, livestock and other herbivores. Forage crops
are of particular concern. The concentration of fluoride both on and in plant tissues can
be important even when there are no effects on the plant per se. Consumption of
forages by livestock containing elevated concentrations of fluoride may cause dental
fluorosis or osteofluorosis.3
Insects and Diseases
Alterations by fluoride of plant-insect and plant-pathogen relationships are
examples of indirect effects that can affect yield or increase the costs of production.
Research in these areas is limited, but has revealed a broad range of responses. For
example, fluoride has stimulated the development of some diseases of plants and
inhibited others through direct effects on the disease organism or indirectly by altering
the host plant.16 The effect of fluoride on plant-insect relationships may be through
changes in the nutritive value of plants, modifications of chemical cues insects use to
find food, or alterations in plant defense mechanisms against insect attack.17
Compared to direct effects, relatively little is known about these indirect effects
of fluoride. It is likely, however, that they will become more important as the
frequency and severity of direct effects diminishes. Attention will then be directed
towards the more subtle, indirect effects.
FACTORS THAT MODIFY PLANT RESPONSE
There is a sequence of events that must occur before fluoride effects occur in
plants: uptake, translocation, and accumulation to a toxic concentration. Any external or
internal factor that affects one or more of these events, directly or indirectly, can
determine the kind of response, its magnitude, or even whether or not an effect will
occur.
Although these factors are interdependent, they can be separated into three
groups: exposure properties, plant characteristics and environmental influences. Exposure
properties include the chemical and physical forms of the pollutant, its concentration, the
duration of exposure and whether it is continuous or intermittent, and the presence of
other air pollutants in the atmosphere with fluoride.18
The biologic factors include genetic influences on the differential sensitivity of
different species or even varieties of the same species. The age of tissues or organs and
the developmental stage of the plant when exposed to fluoride are also biologic
influences that alter plant responses. Environmental factors such as temperature, relative
humidity, rainfall, soil moisture, and the quality or intensity of light exert their influence
on the plant which in turn determines the degree of response to fluoride exposures.
Examples of altered responses for all of these internal and external factors have been
634
-------
reviewed elsewhere.18
FUTURE OUTLOOK
Although relatively efficient controls for fluoride emissions to the atmosphere are
available and regulatory safeguards are in place at most locations in the US, Canada and
western Europe, there are still many uncontrolled or poorly controlled stationary sources,
especially in the developing world. International cooperation (and possibly subsidies)
will be necessary to resolve this gap and reduce the impact of fluoride-emitting
industries on local ecosystems. In those areas where: controls are in place emissions
have been reduced but not eliminated, and the occurrence and severity of obvious
'typical' fluoride injuries to plants have all but disappeared. As a result, greater
attention is currendy focused on the subtle or indirect effects that are likely to occur
from chronic or intermittent exposures to relatively low concentrations of fluoride in the
atmosphere. Future research needs to be directed at identifying those subtle effects and
the dose-response relationships required to induce them. Results from these kinds of
investigations will be necessary to establish realistic and fair air quality standards or
guidelines that will permit industry to coexist with ajjriculture and forestry.
REFERENCES
1. D. C. MacLean, D. C. McCune, J. A. Laurence, L. H. Weinstein, "Advances in
understanding effects of atmospheric fluoride on vegetation," In: Light Metals
1986, (R. E. Miller, ed.), The Metallurgical Society, Warrendale, Pennsylvania.
1986, pp.933-938.
2. L. H. Weinstein, "Fluoride and plant life," J. Occup. Med. 19: 49. (1977).
3. J. W. Suttie, "Effects of fluoride on livestock," J. Occup. Med. 19: 40. (1977).
4. National Academy of Sciences, Fluorides, National Research Council,
Washington, D.C. 1971, pp. 77-132.
5. A. W. Davison, "Uptake, transport and accumulation of soil and airborne
fluorides by vegetation," In: Fluorides: Effects on Vegetation, Animals and
Humans. (J. L. Shupe, H. B. Peterson, N. C. I^eone, eds.), Paragon Press, Salt
Lake City, Utah. 1983, pp. 61-82.
6. L. H. Weinstein, R. Alsher-Herman, "Physiological Responses of plants to
fluorine," In: Effects of Gaseous Air Pollution in Agriculture and Horticulture,
(M. H. Unsworth, D. P. Ormrod, eds.), Butterworth Scientific. London, 1982, pp.
139-167.
7. S. Fridriksson, "Fluoride problems following volcanic eruptions," In: Fluorides:
Effects on Vegetation. Animals and Humans, (J. L. Shupe, H. B. Peterson, N. C.
Leone, eds.), Paragon Press, Salt Lake City, Utah. 1983. pp. 339-344.
8. M. D. Thomas, E. W. Alther, "The effect of fluoride on plants,: In: Handbook
of Experimental Pharmacology, Part 1. Springe]--Verlag, New York. 1966, pp.
231-306.
9. F. L. Seamans, "Historical, economic and legal aspects of fluoride," In:
Fluorides: Effects on Vegetation. Animals and Humans, (J. L. Shupe, H. B.
Peterson, N. C. Leone, eds.), Paragon Press, Salt Lake City, Utah. 1983, pp. 5-6.
635
-------
10. D. C. McCune, D. H. Silberman, L. H. Weinstein, "Effects of relative humidity
and free water on the phytotoxicity of hydrogen fluoride and cryolite," Proc.
Intern. Clean Air Congr. 4: 116. (1977).
11. G, W. Miller, Ming-Ho Yu, J. C. Pushnik, "Basic metabolism and physiologic
effects of fluorides," In: Fluorides: Effects on Vegetation, Animals and Humans.
(J. L. Shupe, H. B. Peterson, N. C. Leone, eds.), Paragon Press, Salt Lake City,
Utah. 1983, pp. 83-104.
12. D. Doley, Plant-Fluoride Relationships-An Analysis With Particular Reference to
Australian Vegetation, Inkata Press, Melbourne. 1986, pp. 1-43.
13. A. C. Hill, M. R. Pack, "Effects of atmospheric fluoride on plant growth," In:
Fluorides: Effects on Vegetation, Animals and Humans. (J. L. Shupe, H. B.
Peterson, N. C. Leone, eds.), Paragon Press, Salt Lake City, Utah. 1983, pp. 105-
120.
14. D. C. MacLean, R. E. Schneider, "Effects of gaseous hydrogen fluoride on the
yield of field-grown wheat," Environ. Pollut. 24: 39. (1981).
15. D. C. MacLean, R. E. Schneider, D. C. McCune, "Effects of chronic exposure to
gaseous fluoride on yield of field-grown bean and tomato plants," J. Amer. Soc.
Hort. Sci. 102: 297. (1977).
16. J. A. Laurence, "Effects of air pollutants on plant-pathogen interactions. L
Pflanzenkr. Pflanzenschutz 88: 156. (1981).
17. D. N. Alstad, G. F. Edmunds, Jr., L. H. Weinstein, "The effects of air pollutants
on insect populations," Ann. Rev. Entomol. 27: 369 (1982).
18. D. C. MacLean, "Factors that modify the response of plants to fluoride," In:
Fluorides: Effects on Vegetation, Animals and Humans, (J. L. Shupe, H. B.
Peterson, N. C. Leone, eds.), Paragon Press, Salt Lake City, Utah. 1983, pp. 135-
144.
636
-------
BIOGEOCHEMISTRY OF TRACE METALS AT THE
HUBBARD BROOK EXPERIMENTAL FOREST, NH
Charles T. Driscoll
Chris E. Johnson
Department of Civil and
Environmental Engineering
Syracuse University
Syracuse, NY 13244
Thomas G. Siccama
Yale School of Forestry
Yale University
New Haven, CT 06511
Gene E. Likens
Institute of Ecosystem Studies
Millbrook, NY 12454
The biogeochemical cycles of selected trace metals {Al, Cd, Cu, Fer
Mn, Ni, Pb, Zn) were studied at the Hubbard Brook Experimental Forest, NH.
Mass balances and soil solution chemistry showed contrasting patterns of
metal behavior. Atmospheric deposition of Cd, Cu and Pb greatly exceeded
stream outflow, suggesting that the forest is a sink for these metals.
Detailed studies of Pb have indicated that inputs of this metal largely
accumulate in the forest floor. Soil solution concentrations of Pb
declined with depth in the mineral soil, resulting in low concentrations in
streamwater. There was a large net loss of Al and Mn from the ecosystem,
indicating that the forest is a source of these elements. Soil solutions
also exhibited elevated concentrations of potentially toxic inorganic
monomeric Al. Atmospheric inputs of other metals were nearly balanced by
outputs {Fe, Ni, Zn). The chemistry and transport of these metals is
discussed.
637
-------
Introduction
There is concern over the effects of atmospheric deposition on the
biogeochemistry of trace metals in forest ecosystems. Inputs of acidic
atmospheric deposition to acid-sensitive (low base saturation) watersheds
may result in the acidification of soil and drainage water1'^. This
process can facilitate leaching of trace metals and may have adverse
consequences on interconnected aquatic ecosytems^'^. Atmospheric
deposition may also serve as an important pathway of trace metal influx to
forest ecosystems5'6.
A variety of biotic and abiotic processes can alter trace metal
cycling within forest ecosystems. Trace metals can be immobilized within
the forest floor by adsorption''**'^. Conversely, soluble organic acids
from leaf leachate or mineralization of soil organic matter may form
aqueous complexes with trace metals, facilitating transport'to the lower
mineral soil or to surface water1^'11. The mineral soil may serve as a
trace metal sink through adsorption and precipitation reactions or as a
trace metal source through desorption and dissolution reactions1^'^.
Finally, forest vegetation may assimilate trace metals and facilitate
cycling through canopy leaching, or decomposition of leaf and root
litter5'15'16.
Mass balances are an effective tool in understanding the transport of
trace metals to and within forest ecosystems, and their effects on surface
waters. The Hubbard Brook Experimental Forest in New Hampshire has been a
site of element cycling studies since 1963. By monitoring precipitation
inputs and stream outputs from small watersheds that are essentially free
of deep seepage, it is possible to construct accurate element balances.
The precipitation and stream monitoring program at Hubbard Brook has been
supplemented by detailed studies of soil and soil solution chemistry, and
forest floor and vegetation dynamics. The objective in this study was to
compile and summarize available information on the biogeochemistry of
selected trace metals for the Hubbard Brook Experimental Forest.
Experimental Methods
Study Site
The Hubbard Brook Experimental Forest (HBEF) is located in the White
Mountains of New Hampshire (43°56'Nr71°45'W). This investigation was
conducted in watershed 6 (w6), the biogeochemical reference watershed at
Hubbard Brook (13.23 ha, elevation 546-791 m: Figure 1). The watershed has
a southeasterly aspect and a slope of 20-30% , Soils at the site are
acidic, well-drained Spodosols (Haplorthods and Fragiothords) with a well
drained organic layer (3-15 cm}. They are underlain with variable depths of
glacial till and impervious bedrock (Littleton formation; schist)16. Soils
are shallow at high elevations and increase in depth with decreasing
elevation1''.
Climate at the HBEF is cool temperate, humid continental with mean
July and January temperatures of 19°C and -9°C, respectively (at 450 m
elevation). Mean annual precipitation is approximately 130 cm, with 25-33%
of the total occurring as snow15' . Mean annual streamflow from w6 is 80
cm16.
638
-------
Northern hardwood vegetation dominates most of w6, consisting of
American beech (Fagus grandifolia Ehrh.), yellow birch (Betula
cilleghaniensis Britt.) and sugar maple (Acer saccharum Marsh.), from 500-
730 m. Coniferous vegetation, consisting primarily of red spruce (Picea
rubens Sarg.} and balsam fir (Abies balsamea (L.) Hill) dominates
elevations above 730 m. The HBEF was logged between 1909-1917, and there
is no evidence of recent fire^.
Data Analysis
This paper is a summary of many trace metal studies that have been
conducted at the HBEF. Data collected include fluxes and concentrations of
trace metals in bulk precipitation, streams^''^, soil solutions-^'^0,
forest floorer 21^ mineral soil^l and vegetation!-". Details of sampling
and analytical methods for these studies are provided elsewhere. Trace
rretal biogeochemistry from these studies are summarized here through
element budgets (Figure 2). Ecosystem elemer.t pools include: above and
below ground biomass, the total forest floor content, exchangeable metals
in E+Bh horizons, Bsl horizon and Bs2 horizor. soil, and the total mineral
soil pool {< 2mm size fraction). Ecosystem fluxes include bulk
precipitation inputs, uptake by biomass, solution fluxes through the Oa, Bh
and Bs horizons and stream outflow. Weathering fluxes are not determined
directly but are calculated as differences in the mass balance. This
calculation assumes that exchangeable element pools are at steady-state.
Results and Discussion
Element balances for the trace metals studied show distinct patterns
{Table 1). Values of atmospheric deposition of trace metals at Hubbard
Brook are high for a remote area, but are consistent with other studies of
r C ')')
trace metal deposition in the northeastern United States-5'0' .
Streamwater outputs are dominated by dissolved or fine particulate forms.
This pattern is consistent with the low sediment transport from this
undisturbed forest. Exceptions to this are evident for Al and Fe. These
metals show 41% and 30% of their efflux associated with particulate matter.
Mass balance calculations show three general classes of trace metals: 1)
metals that are strongly retained within the ecosystem (Cd, Cu, Pb); 2}
metals that are strongly leached from the ecosystem (Al, Mn); and 3) metals
that approximately balance between atmospheric deposition and stream losses
(Fe, Ni, Zn). To illustrate differences in trace metal chemistry we
present detailed budgets for Pb and Al.
Lead Budget for the HBEF
Smith and Siccama^ previously developed a Pb budget for Hubbard
Brook. With additional data on soil^l and soil solution^ chemistry we are
able to expand this budget (Figure 3a). Atmospheric deposition of Pb has
been declining since Pb was added to the precipitation monitoring program
in 1975^' -^. This decline is consistent with the decreased use of leaded
gasoline over the same period-^'23_
The mineral soil and forest floor are the major pools of Pb in the
ecosystem (Figure 3a). Mineral soil pools are generally the largest
element pools for the HBEF^, however this includes relatively unreactive
soil minerals. Deposition and accumulation of Pb in the forest floor has
been the focus of a number of investigations^ • ^' ^' ^. At Hubbard Brook
much of the Pb entering the ecosystem from the atmosphere appears to be
retained in the forest floor. Concentrations and fluxes of Pb in bulk
639
-------
precipitation are much greater than in Oa horizon leachate. Concentrations
and fluxes of Pb decrease through the soil profile13 and losses in
strearawater are low. Driscoll et al.13 noted that there was a strong
correlation between concentrations of Pb and dissolved organic carbon (DOC)
in soil solutions and streamwater at Hubbard Brook. Moreover Smith and
Siccama1* showed that acid extractable Pb concentrations were elevated in
the forest floor (0.43 mmol/kg), low in the E horizon (0.025mmol/kg),
intermediate in the B horizon (0.058 mmol/kg) and low in the C horizon
(0.042 mmol/kg). This pattern is consistent with decomposition and release
of DOC from the forest floor, transport through the E horizon and
deposition in the B horizon which occurs as part of soil development (the
podzolization process). Mobilization and immobilization of Pb at Hubbard
Brook appears to be associated with the dynamics of soil organic matter.
Vegetation pools and uptake of Pb at Hubbard Brook are small. Lead
is not a plant nutrient^ an(j therefore it is not surprising that
assimilation is low. The calculated weathering input for Pb at Hubbard
Brook is negative (-0.84 mol/ha-yr). This is likely due to changes in
mineral soil Pb pools over the study period. Siccama" has observed marked
declines in the Pb content of the forest floor. This pattern is probably
due to the recent decreases in precipitation inputs of Pb.
There has been some concern over the effects of elevated inputs of Pb
to forest ecosystems^'6'9'18. These potential effects include leaching of
Pb to surface waters^ and mineralization of Pb in the forest floor and
release following clearcutting disturbance1". Results from long-term
monitoring at Hubbard Brook suggests that streamwater concentrations are
very low and not a water quality concern13'18. In addition, a study of Pb
in soil solutions and streamwater following a commercial whole-tree harvest
at Hubbard Brook showed that Pb was not released to drainage waters from
clearcutting activities^.
Aluminum Budget for the HBEF
The Al cycle is characterized by large soil pools and relatively
small fluxes (Figure 3b). The Hubbard Brook Al budget shows that the
watershed exhibits a large net release of Al, originating from dissolution
of soil minerals. Studies from the HBEF have shown that soil and
streamwaters are in apparent equilibrium with A1(OH)3 solubility1'''°'^".
The solubility of aluminosilicate minerals is pH dependent and leaching of
Al is elevated under acidic conditions1'^. At Hubbard Brook drainage
waters are acidic (pH < 5.0) due to elevated inputs of SO^" and limited
release of basic cations (Ca^"1", Mg2+, Na+, K+) . As a result leaching
losses of Al are high relative to watersheds which are not impacted by
strong acid inputs3".
An understanding of the biogeochemistry of Al at Hubbard Brook has
been facilitated by examination of the speciation of aqueous Al. Solution
Al is present as nonlabile monomeric Al, which is an estimate of alumino-
organic complexes, and labile monomeric Al, which represents inorganic
species of Al. The fractionation of Al can be used to assess the cycling
of Al in forest soils as well as the potential ecological effects of
elevated Al concentrations. Inorganic forms of Al are thought to be toxic
to vegetation31 and aquatic organisms3^. The speciation of monomeric Al in
soil solutions draining three horizons (Oa, Bh, Bs) in three elevational
zones (spruce-fir, 750 m; high elevation hardwood, 730 m; low elevation
hardwood, 600 m; Figure 1) is shown in Figure 4. Organic horizon (Oa)
leachate (which is characterized by elevated concentrations of DOC13) has
high concentrations of Al which is largely in an organic form. Deeper in
640
-------
the mineral soil, DOC concentrations decrease and Al shifts from
predominantly an organic form to an inorganic form. Aluminum losses are
most pronounced at high elevation sites wi :h shallow acidic soils. At
lower elevations the thickness and ba:se saturation of the mineral soil
increases, which coincides with higher soil solution pH and lower
concentrations of solution Al.
There is concern over the ecological effects of elevated
concentrations of inorganic monomeric Al. For example, surface water
concentrations in excess of 7 umol/L are thought to be toxic to fish-^.
Also, high concentrations of Al are though1: to contribute to red spruce
decline in the northeastern U.S.3^ Laboratory experiments suggest that
100 umol/L inorganic monomeric Al is a toxic threshold for red spruce .
In addition, laboratory experiments have indicated that an Al/Ca molar
ratio above 1 may impair red spruce growth . Soil solutions from the
spruce fir zone at Hubbard Brook are well below the concentration threshold
{average inorganic raonomeric Al concentrations were 1 umol/L in Oa soil
solutions and 22 umol/L in Bs soil solutions). The Al/Ca in Oa horizon
solutions (0.04) is well below the critical value of 1, however the Al/Ca
in the Bs horizon (1.6) is somewhat above ':his level. While there is no
evidence of spruce decline at Hubbard Brook, the effects of elevated
concentrations of Al warrant further ;>tudy,
Acknowledgement
This manuscript is a contribution of the Hubbard Brook Ecosystem
Study. We thank C. Wayne Martin, D. Buso, B. Pierce and F.H. Bormann for
their help and support. The Hubbard Brook Experimental Forest is operated
by the USDA Forest Service, Broomall, Pennsylvania.
References
1. N. van Breemen, C.T. Driscoll and J. Mulder, "The role of acidic
deposition and internal proton sources in the acidification of soil
and water," Nature 307: 599. (1984).
2. C.T. Driscoll and R.M. Newton, "Chem:,cal characteristics of acid-
sensitive lakes in the Adirondack region of New York," Environ. Sci.
Technol. 19: 1182. (1985).
3. H. Heinrichs and R. Mayer, "Distribution and cycling of major and
trace elements in two central European forest ecosystems," J^
Environ. Qual. 6: 402. (1977).
4. W.J. Miller, W.W. McFee and J.M. Kelly, "Mobility and retention of
heavy metals in sandy soils," J_._^nv_:.xon_._,Qua_K_12^: 579. (1983).
5. W.A. Reiners, R.H. Marks and P.M. Vitousek, "Heavy metals in
subalpine soils of New Hampshire," 0:.kos 26: 264. (1975).
6. A.M. Johnson, T.G. Siccama and A.J. I'riedland, "Spatial and temporal
patterns of lead accumulation in the forest floor in the northeastern
United States," J. Environ. Qual. 11: 577. (1982).
7. J.R. Gosz, G.E. Likens and F.H. Bormann, "Organic matter and nutrient
dynamics of the forest and forest floor in the Hubbard Brook Forest,"
Oecologia (Berl.) 22: 305. (1976).
641
-------
8. K.A. Swanson and A.H. Johnson, "Trace metal budgets for a forested
watershed in the New Jersey Pine Barrens," Water Resour. Res. 16:
373. (1980).
9. A.J. Friedland and A.H. Johnson, "Lead distribution and fluxes in a
high-elevation forest in northern Vermont," J. Environ. Qual. 14:
332. (1985).
10. E.M. Perdue, K.C. Beck and J.H. Reuter, "Organic complexes of iron
and aluminum in natural waters," Nature 260: 418. (1976).
11. R.S. Turner, D.W. Wang, and A.H. Johnson, "Biogeochemistry of lead in
McDonald's Branch Watershed, New Jersey Pine Barrens," J. Environ.
Qual. 14: 305. (1985).
12. W.R. Boggess, "Summary and conclusions," In: Lead in the Environment,
National Science Foundation, Washington, D.C. 1977, pp. 267-272.
13. C.T. Driscoll, R.D. Fuller and D.M. Simone, "Longitudinal variations
in trace metal concentrations in a northern forested ecosystem," J_._
Environ. Qual. 17: 101. (1988).
14. J.R. Gosz, G.E. Likens, and F.H. Bormann, "Nutrient release from
decomposing leaf litter in the Hubbard Brook Forest, New Hampshire,"
Ecol. Monogr. 43: 173. (1973).
15. R.H. Whittaker, G.E. Likens, F.H. Bormann, J.S. Eaton, and T.G.
Siccama, "The Hubbard Brook ecosystem study: forest nutrient cycling
and element behavior," Ecology 60: 203. (1979).
16. G.E. Likens, F.H. Bormann, R.S. Pierce J.S. Eaton, and N.M. Johnson,
Biogeochemistry of a forested ecosystem. Springer-Verlag, New York.
1977.
17. G.B. Lawrence, R.D. Fuller, and C.T. Driscoll, "Spatial relationships
of aluminum chemistry in streams of the Hubbard Brook Experimental
Forest, New Hampshire," Biogeochemistry 2: 115. (1986).
18. W.H. Smith and T.G. Siccama, "The Hubbard Brook ecosystem study:
biogeochemistry of lead in a northern hardwood forest," J.Environ.
Qual. 10: 323. (1981).
19. W.H. Smith, T.G. Siccama, and S. Clark, "Atmospheric deposition of
heavy metals and forest health: an overview including a ten year
budget for the input/output of seven heavy metals to a northern
hardwood forest," FWS-87-02, Virginia Polytechnic Institute and State
Univ., Blackburg, VA. (1986).
20. R.D. Fuller, C.T. Driscoll, G.B. Lawrence and S.C. Nodvin, "Processes
regulating sulfate flux after whole-tree harvesting," Nature 325:
707. (1987).
21. C.E. Johnson, "The Chemical and Physical Properties of a Northern
Hardwood Forest Soil: Harvesting Effects, Soil-Tree Relations and
Sample Size Determination," Ph.D Dissertation, University of
Pennsylvania, Philadelphia, PA. (1989).
642
-------
22. A.J. Friedland, A.H. Johnson, T.G. Siccama, and D.L. Mader, "Trace
metal profiles in the forest floor of New England," Soil Sci. Soc.
Am. J. 48: 422. (1984).
23. U.S. Environmental Protection Agency, "National Air Quality and
Emission Trends Report 1982," EPA-450/4-84-002. Research Triangle
Park, NC. (1984) .
24. H. Bohn, B. McNeal, and G. O'Connor, Soil Chemistry, Wiley-
Interscience, New York. (1979).
25. T.G. Siccama, Yale University, New Haven, CT, unpublished data,
(1990).
26. J.R. White and C.T. Driscoll, "Lead cycling in an acidic Adirondack
lake," Environ. Sci. Technol. 19: 1182. (1985).
27. R.D. Fuller, D.M. Simone, and C.T. Driscoll, "Forest clearcutting and
effects on trace metal concentrations: spatial patterns in soil
solutions and streams," Water, Air, Soil Pollut. 40: 185. (1988).
28. N.M. Johnson, C.T. Driscoll, J.S. Eaton, G.E. Likens and W.H.
McDowell, "Acid rain, dissolved aluminum and chemical weathering at
the Hubbard Brook Experimental Forest, New Hampshire," Geochim^
Cosmochim. Acta. 45: 1421. (1981).
29. C.T. Driscoll, N. van Breemen, and J. Mulder, "Aluminum chemistry in
a forested Spodosol," Soil Sci. Soc. Am. J. 49: 437. (1985).
30. C.T. Driscoll, N.M. Johnson, G.S. Likens, and M.C. Feller, "The
effects of acidic deposition on stream water chemistry: a comparison
between Hubbard Brook, New Hampshire and Jamieson Creek, British
Columbia," Water Resour. Res. 24: 195. (1988).
31. C.T. Driscoll, J.P. Baker, J.J. Bisogni, and C.L. Schofield,
"Aluminum speciation and its effect on fish in dilute acidified
waters," Nature 284: 161. (1980).
32. D.R. Parker, L.W. Zelazny, and T.B. Kinraide, "Chemical speciation
and plant toxicity of aqueous aluminum," In: T.Lewis (ed.),
Environmental Chemistry and Toxicology of Aluminum, Lewis Publishers,
Chelsea, MI. 1989. pp. 117-146. '
33. D.R. Mount and M.D. Marcus (eds.), "Physiologic, Toxicologic and
Population Responses of Brook Trout ~o Acidification," EPRI En-6238,
Electric Power Research Institute, Palo Alto, CA. (1989).
34. W. Shortle and K. Smith, "Aluminum induced calcium deficiency
syndrome in declining red spruce," Science 240: 1017. (1988).
35. J.D. Joslin and M.H. Wolfe, "Responses of red spruce seedings to
changes in soil Al in six amended forest soil horizons," Can J. For.
Res. 18; 1614. (1988).
36. W.H. Schroder, J. Bauch, and R. Endeward, "Microprobe analysis of
calcium exchange and uptake in the fine roots of spruce: influence
of pH and aluminum," Trees_2: 96. (1988).
643
-------
Table 1. Trace Metal Input/Output Budgets for w6 at the HBEF (mol/ha-yr)
Input Output Net
dissolved particulate
Aluminum (Al) 8 74 51 -117
Cadmium (Cd) 0.11 0.013 9xlO~5 +0.099
Copper (Cu) 0.25 0.077 0.003 +0.17
Iron (Fe) 8.3 7.3 3.2 -2.2
Lead
-------
Watershed 6
Transition
zone
660 m
Spruce-fir
zone
Lower
hardwood
zone
540 m
Percent basal area (m2/ha)
of spruce plus fir trees
>0
25>
50 >
>25
>50
Figure 1. Map of HBEF watershed 6. Lysiraeter and
stream sampling stations are indicated. The
percent basal area of red spruce; and balsam fir
vegetation are shown.
645
-------
Bulk
PreclpiUtion
Above-Ground
Biomass
Stream
Output
Figure 2. Schematic drawing of the biogeochemical cycle
in a forest ecosystem.
646
-------
155
Lead
Units
Pool* — mol
St»tm Output
0.025
(b)
Aluminum
Units
Pool* - mol ha"1
Fluxes — mol ha-1yr~l
Stream Output
125
Figure 3. Biogeochemical cycles of Pb (a) and Al (b) at Hubbard Brook.
-------
Monomeric Al Fractionation
0
Non — labile
Labile
10 20 30 40
Al
Figure 4. Concentrations of nonlabile (organic) and labile
(inorganic) monomeric Al in soil solutions draining Oa, Bh,
and Bs horizons at three vegetation zones. The lysimeter
sampling stations are shown in Figure 1.
648
-------
USE OF THE PHYTOTOX DATABASE TO ESTIMATE THE INFLUENCE OF
HERBICIDE DRIFT ON NATURAL HABITATS IN AGROECOSYSTEMS
James E. Nellessen and John S. Fletcher
Dept. of Botany and Microbiology
Univ. of Oklahoma
Norman, OK 73019-0245
It was shown that if the maximum amounts of trifluralin
and alachlor known to volatilize from plowed fields are
assumed to be wind blown onto adjacent vegetation in an oak-
hickory plant community the growth of 8 different genera
could be influenced. The most sensitive plant genera were
Urtica and Cassia. The PHYTOTOX database provides a tool
for estimating the influence of herbicide drift on the
productivity and composition of natural plant communities.
Introduction
In agroecosystems, there exists a patchwork of row crops
intermixed with pasture and natura.l plant communities. The
extensive use of herbicides in various agroecosystems across
the U.S. poses a potential threat to vegetation growing on
adjacent land if chemicals applied to cropland inadvertently
drift onto nontarget areas. Adverse consequences of such an
event could be: reduced production of plant biomass (yield),
and/or change in the species compcsition (diversity) of the
nontarget plant community. Whether or not a plant community
is effected in either or both of these ways is dependent on
two factors: the nature of the exposure and the
sensitivities of the plant species within the nontarget
community.
649
-------
Estimating chemical exposure of nontarget plants depends
on several variables: the chemical gradient (concentration
vs. distance) extending across the nontarget zone, duration
of exposure, and frequency of exposure. The magnitude of
each of these variables depends in turn on the chemical and
physical properties of the compound, mode of application,
and prevailing weather conditions. Since all of these
parameters can be quantified the process of predicting the
amount of drift and subsequent exposure of nontarget
vegetation lends itself to mathematical modeling. A variety
of distinctive models have been developed by various
investigators to deal with different combinations of
pesticide application (plane vs. tractor), chemical features
(liquid vs. dust), and mode of drift (direct from applicator
or indirect following volatilization from field).
The response of nontarget vegetation to pesticide drift
has received less attention than development of the drift
models. Some models appear to have never been validated by
biomonitoring1, and others have only been validated by
examining cultivated crops2. There are only a few isolated
reports on how herbicides influenced the productivity or
composition of native plant communities3'4' '6'7. Only one of
the reports was done in connection with drift modeling, one
dealt with a tree community, and none considered the
influence of indirect (field volatilization) drift.
In the absence of such studies, an alternative is to use
dose-response data taken from the literature for individual
plant species which are known to be present in natural plant
communities frequently found in pesticide treated
agroecosystems. The dose-response data compiled in the
PHYTOTOX database8 is ideally suited for predicting
potential hazards posed by pesticide drift to nontarget
vegetation. In this pilot study we used a portion of
PHYTOTOX to predict the potential hazard posed by the
volatilization and drift of trifluralin and alachlor on
forest communities in Illinois.
Materials and Methods
The PHYTOTOX database is a computerized information
resource that permits the rapid retrieval and comparison of
data pertaining to the response of terrestrial plants to the
application of organic chemicals8. As of January 1, 1990,
PHYTOTOX possessed information on approximately 8,000
different chemicals, 2,000 species, and 50 plant responses.
The data has been compiled from over 3,500 articles
published between 1926 to 1988. The database has two files:
a Bibliographic File and an Effects File. The Bibliographic
File possesses information on each paper which has been used
as a source of data for compiling effects records. The
Effects File contains approximately 100,000 records. The
Effects File differs from most biological databases, because
it contains quantitative numerical data pertaining to
chemical doses, plant responses and experimental parameters.
Each record in the Effects File contains information
concerning the effect(s) of one dose of a single chemical
650
-------
applied to a particular plant specj.es as reported in one
publication. The information associated with each record is
organized under labels that may be sorted separately during
computer searches.
Trifluralin and alachlor were examined in this study,
because of their high usage on corr and soybeans in
Illinois9, and their strong tendency to volatilize from the
soil10'11. Attention was focused on the oak-hickory community
since small wood lots meeting this description are
interdispersed among corn and soybean fields in Illinois .
A hierarchal search of PHYTOTOX was conducted to recover
data pertaining to oak-hickory community plants treated with
comparative doses (kg/ha) of either trifluralin or alachlor.
Results
PHYTOTOX possessed information on 100 species which occur
in oak-hickory communities. This number was reduced to 8
when only records pertaining to trifluralin and alachlor
were considered (Table I). The plant list was expanded to
include data on an additional 1.0 species which are in the
same genera as plants found in oak-hickory communities.
Justification for expanding the list in this manner comes
from previous analyses where it was shown that species
within the same genus had a high correlation of response to
the same chemical13.
Examination of the data in Table I shows that different
species vary widely in their sensitivities, and a single
species will respond differently to different herbicides.
For example, while Senecio vulqaris experienced 100% control
(kill) when treated with 1.9 kg/ha alachlor, it was not
affected at all by a somewhat lower dose of trifluralin.
This variable response by different plant species to
chemicals has been capitalized on in selectively killing
unwanted plants in cultivated fields. This feature also has
the potential of eliminating biodiversity in nontarget areas
if the amounts of drifting chemicals reach the inhibitory
level for sensitive plants.
Trifluralin is an extremely volatile herbicide.
Glotfelty et al. have shown that 90% of the trifluralin
applied to moist soil will be lost to the air in 2-7 days
following application10. These investigators established
the maximum rate of volatilization to be 195 g/ha/h and
measured air concentrations as high as 40 ng/m (0.003 ppm)
at 50 cm above the soil surface. Based on these data it can
be hypothesized that if 2.8 kg/ha is applied to a cultivated
field it is possible that moving air could transport 2.5
kg/ha to adjacent nontarget plants aver a 2-7 day period.
When this level of exposure is compared to dose-response
data in Table I it was found that 8 different genera (Acer,
Cassia, Dioscorea, Ilex, RhododendrDn, Solanum, Thuja, and
Urtica) had sensitivities low enough to be affected. Among
these genera the species Urtica chanaedrvoides is on the
Illinois endangered and threatened species list.
651
-------
Alachlor has a lower field volatility than Trifluralin,
but even at the reduced rate of 8.1 g/ha/h it has been shown
that 19% of applied doses {420 g/ha) were lost in 21 days11.
If weather conditions caused this amount of volatilized
alachlor to be redeposited on vegetation adjacent to an
applied field it could influence the growth of 3 genera
listed in Table I. This level of alachlor exposure
represents 25% of a dose giving 23% control (kill) of Cassia
obtusifolia and 22% of a dose giving 100% control of Urtica
urens and Senecio vulgaris (Table I). Two species of
Cassia, 3 of Senecio, and the aforementioned U.
chamaedryoides growing in Illinois woodlots could be
affected.
Conclusions
A comparison between maximum drift values estimated from
the literature with dose-response data taken from PHYTOTOX
indicated that some nontarget species growing in oak-hickory
communities could be influenced by the drift of trifluralin
and/or alachlor. The most sensitive plants were Cassia and
Urtica. Although we are not aware of any field measurements
or biomonitoring data collected in native plant communities
which would substantiate this prediction, there is a report
by Behrens and Lueschen where 0.28 kg/ha of dicamba applied
to a corn field affected soybean growth 60 m downwind in an
adjacent field14. Such data certainly is cause for concern
and suggests that herbicide drift may have a profound
influence on the productivity and composition of natural
plant communities.
The biota of the U.S. has been described as having.
approximately 116 different native plant communities 2. The
geographical location and species composition of each of
these communities is known. Therefore it is possible to
identify natural plant communities which are in association
with various crops in different agroecosystems located
throughout the U.S. In this pilot study with PHYTOTOX we
considered only one plant community (the oak-hickory forest)
and only two herbicides (trifluralin and alachlor). Similar
analyses could be conducted for all 116 plant communities in
the U.S. and for an extended list of herbicides.
References
1. N. Thompson, "Diffusion and uptake of chemical vapour
volatilizing from a sprayed target area," Pestic. Sci.
14:33. (1983).
2.W. E. Yates, N. B. Akesson, D. E. Bayer, "Drift of
glyphosate sprays applied with aerial and ground
equipment," Weed Sci. 26:597. (1978)
3.R. H. Marrs, "The effects of potential bracken and scrub
control herbicides on lowland Calluna and grass heath
communities in East Anglia, UK," Biol. Conserv. 32:13.
(1985).
652
-------
4.R. H. Marrs, C. T. Williams, A. J. Frost, R. A. Plant,
"Assessment of the effects of herbicide spray drift on a
range of plant species of conservation interest,"
Environ. Pollut. 59:71. (1989).
5. C. R. Malone, "Effects of a nonsc'lective arsenical
herbicide on plant biomass and community structure in a
fescue meadow," Ecology 53:507. (1972).
6.R. L. Gillen, D. R. Rollins, J. F. Stritzke, "Atrazine,
spring burning, and nitrogen for improvement of tallgrass
prairie," J. Range Manage. 40:444. (1987).
7.B. F. Swindell, J. E. Smith, D. G. Neary, N. B.
Comerford, "Recent research indicates plant community
responses to intensive treatment including chemical
amendments," South. J. Appl. For. 13:152. (1989).
8.C. L. Royce, J. S. Fletcher, P. R. Risser, J. C.
McFarlane, F. E. Benenati, "PHYTOTOX: A database dealing
with the effect of organic chemicals on terrestrial
vascular plants," J. Chem. Inf. Comput. Sci. 24:7.
(1984).
9. L. P. Gianessi, C. A. Puffen, Use of Selected Pesticides
in Agricultural Crop Production by State, Quality of the
Environment Division, Resources for the Future. (1988).
10. D. E. Glotfelty, A. W. Taylor, B. C. Turner, W. H.
Zoller, "Volatilization of surface applied pesticides
from fallow soil," J. Agric. Food Chem. 32:638. (1984).
11. D. E. Glotfelty, M. M. Leech, J. Jersey, A. W. Taylor,
"Volatilization and wind erosion of soil surface applied
atrazine, simazine, alachlor, and toxaphene," J. Aaric.
Food Chem.. 37:546. (1989).
12. A. W. Kuchler, Potential Natural Vegetation of the
Conterminous United States. American Geographical
Society Pub. 36, Princeton Polychrome Press. 1964.
13. J. S. Fletcher, F. L. Johnson, J. C. McFarlane,
"Influence of greenhouse versus field testing and
taxonomic differences on plant sensitivity to chemical
treatment," Environ. Toxicol. Chem. 9:769. (1990).
14. R. Behrens, W. E. Lueschen," Dicamba volatility," Weed
Sci. 27:486. (1979).
653
-------
Table I. Dose-response data from PHYTOTOX for two herbicides commonly used in Illinois and taxa from
genera known to occur in Illinois wood lots.
Herbicide
Plant
Species
Acer palmatum3
Cassia obtusifolia3
Dioscorea sp.
Euonymus fortune!8
tlex cornuta8
Ilex crenata8
Juniperus horizontalis
Pinus echinata
Pinus strobus
Rhododendron obtusuma
Sedum brevifolium8
Senecio vulgaris3
Solanum nigrum
Solanum sp.
Thuja occidentalis
Urtica urensa
Trifluralin
Common Name
Japanese maple
Sickle-pod
Wild yam
Wintercreeper
Chinese holly
Japanese holly
Creeping cedar
Short-leaf pine
White pine
Kirishima azalea
Stonecrop
Common ragwort
Black nightshade
Hairy nightshade9
White cedar
Burning nettle
Dose
kg/ha
3.1
0.9
0.2
8.9
2.5
1.1
4.5
2.5
1.1
0.6
0.6
0.6
2.5
Responseb
%
20 injury
81 control
3DMD
16 RT FMD
18 LF CHL
None
None
35TRD
None
15 control
35 control
3TRD
100 kill
Alachlor
Dose
kg/ha
6.2
1.7
20.0
12.5
13.6
6.0
2.5
1.9
13.6
1.9
Response13
%
Injury
23 control
49DMI
6SZ1
None
9 injury
15 injury
100 control
None
100 control
which do not occur in Illinois woodlots but are members of genera that can be found in Illinois wood lots.
bAII responses refer to whole plants unless indicated otherwise. Control - similar in meaning to plant kill, DMD -
dry mass decrease, DMI-dry mass increase, LF CHL-leaf chlorosis, RT FMD - root fresh mass decrease, SZl-size
increase, TRD-transpiration decrease.
-------
DETECTING EFFECTS OF AIR TOXICS USING WILDLIFE
R. Kent Schreiber and James R. Newman
U.S. Fish and Wildlife Service KBN Engineering & Applied Sciences
National Fisheries Research Center 1034 Northwest 57th St.
Box 700 Gainesville, FL 32605
Keameysville, WV 25430
Biologists and resource managers are interested in the impacts of air toxics on wildlife
from two perspectives. First, animals can provide early detection of the presence of air toxics
in the ecosystem. Their response can contribute to OUT knowledge of biological damage and
of both human and animal health effects, as well as assist with the establishment of emission
standards. Second, effective protection and management of wildlife requires information
about any aspect of their environment that can diredly or indirectly impact their status and
function in the ecosystem. Once an air toxic is detected in a species it must also be
interpreted in terms of consequences to the indiv:.dual (e.g., mortality, reduced vigor),
population (e.g., genetic loss), and ecosystem (e.g., change in energy transfer). Information
about ecological effects exists for only a small number of the 308 chemicals classified as air
toxics. Animals have an important role in the development of the risk assessment
methodology associated with airborne contaminants. Many chaDenges remain in developing
effective biomonitoring programs. No standards currently exist for most air toxics and few
protocols have been established for standardizing data collection and analyses.
655
-------
INTRODUCTION
The air resource provides the exchange of gases basic to life. It also serves as the
pathway and transportation medium for a diversity of materials, including contaminants.
Animals have long served as biological indicators and monitors of harmful chemicals in the
air environment1'2. The task of regulating these chemicals and their potential impacts to total
ecosystems has been hindered, however, by inadequate development of concepts and methods
that translate individual effects to populations and ultimately to the ecosystem level3.
Ecosystems may modify the transport, fate, and effect of the chemical (positive/negative
interaction) or accumulate and magnify the effects (negative interaction). New ecosystem
threats, including global warming, climate change, and increasing levels of volatile organics,
offer additional challenges to both detection and measurement of effects. Animals have an
important role in the development of associated risk assessment methodology.
Definition of Air Toxics
Air toxics are classified as all non-criteria pollutants as defined by Section 313 of the
Superfund Amendments and Reauthorization Act (SARA) of 1986. They include 308 chemicals
and 20 chemical categories. The chemicals include a large number of organics (e.g., benzene,
ethylene, vinyl chloride) as well as pesticides (e.g., lindane and aldrin, and 2,4-D); inorganics
(e.g., nitric acid, sulfuric acid); and metals (e.g., arsenic, mercury, cadmium). The 20 chemical
categories are primarily metallic compounds along with PCBs, chlorophenols, and glycol
ethers4. EPA has published a list of industrial processes and sources that emit air toxics
including chemical plants, chrome plating facilities, coal-fired power plants, dry cleaners, and
non-point sources such as landfills, mines, and residential wood combustion areas5.
Legislation
The laws currently regulating toxic substances include: (1) Clean Air Act (CAA) of
1970, as amended 1977; (2) Federal Water Pollution Control Act (FWPCA) of 1972, as
amended 1977 (also known as the Clean Water Act); (3) Federal Insecticide, Fungicide, and
Rodenticide Act (FIFRA) of 1972, as amended in 1975 and the Federal Pesticide Act of 1978;
(4) Safe Drinking Water Act (SDWA) of 1974; (5) Marine Protection Research and Sanctuaries
Act of 1972 (Ocean Dumping Act); (6) Resource Conservation and Recovery Act of 1976
(RCRA); (7) Toxic Substances Control Act (TSCA) of 1976; and (8) the Comprehensive
Environmental Response, Compensation, and Liability Act ("Superfund") of 19806. These laws
are designed to safeguard human health and the environment, and animals are an important
link in detecting, monitoring, and evaluating impacts.
WILDLIFE AS INDICATORS
Biological Considerations
Biologists and resource managers are interested in the impacts of air toxics on wildlife
from two perspectives. First, animals can provide early detection of air toxics in the ecosystem
and their response can contribute to our knowledge of health effects and assist with the
establishment of emission standards. Wildlife indicators may provide conservative estimates
of pollutant effects on human populations. Second, effective protection and management of
wildlife requires information on any aspect of their environment that can directly or indirectly
impact their status and function in the ecosystem. Air toxics represent one of the major
external threats to many of these natural systems7. Once an air toxic is detected in a species
it must additionally be interpreted in terms of consequences to the individual (e.g., mortality,
656
-------
reduced vigor), population (e.g., genetic loss), and ecosystem (e.g., change in energy transfer).
Wildlife in sub-optimal habitats may be more sensitive to air emissions than portions of the
same population in more suitable parts of their habitats. This difference has implications
especially for endangered and threatened species that may already be restricted to limited
habitats. Federal land managers responsible for wilderness areas and Class I (clean air) lands
are particularly interested in developing biomonitormg that can be implemented in remote
areas with restricted access and with limitations for collecting air toxics information by
conventional methods and equipment*.
There is also a need to understand the synergism and antagonism that may exist with
air toxics. When one air contaminant is a problem, invariably other contaminants occur with
varying and conceivable opposing effects. The toxicity of airborne pollutants is influenced by
a number of variables, including type and concentration of pollutant, prevailing winds,
temperature, humidity, precipitation, topography, season, species, age, management and
activity of animals, length of exposure, nutrition, genetics, and physiology9. Toxics that affect
a diversity of animals, persist for extended periods :in the environment, are mobile, have a
high solubility in fat, and have high potential for bioaccumulation are of particular concern.
Unfortunately information about ecological effects exists for only a small number of
the 308 chemicals classified as air toxics. Substantial information exists on the ecological
effects for most metals classified as air toxics. For some of these metals, such as arsenic,
ecological effects have been reported for over 100 years1. Except for pesticides classified as
air toxics, very little information exists on the effect of most organic air toxics on free-living
animals.
Wildlife species have good potential as biological indicators of air toxics for several
reasons: (1) they integrate all environmental conditions, both natural and man-made, in their
responses; (2) they can show pathway points of accumulation and stress in natural and man-
influenced ecosystems; and (3) they can be used as biological standards to verify the physical
and chemical standards of air quality.
Wildlife may be used to assess air toxics in three main ways. The most commonly
used method is to monitor residues or tissue concentrations of chemicals that are not readily
metabolized and excreted. This group of compounds includes most metals, fluoride, and lipid
soluble organics. Second, sublethal responses of animals to air pollutants may be monitored.
This bioindicator approach has been applied frequently in recent years, both for air and other
sources of pollutants. Physiological responses, especially enzyme activities, have proven useful
for some chemicals10. However, many physiological responses are not specific for individual
chemicals, and hence, is one of the drawbacks of this approach. Despite almost three decades
of interest and research in wildlife toxicology, interpretive information for chemical residue
data and physiological responses extends to only a few substances, and is further limited by
species sensitivities and substrates (bone vs. blood, for example). Therefore, a third approach,
assessment of effects on populations (although in reality it is often individuals), is being
welcomed into the field. This approach, which generally examines recruitment and
survivorship, is the most difficult, but perhaps the most meaningful, for wildlife populations.
The routes of exposure of animals to air toxics can be either direct or indirect.
Animals are directly influenced by inhalation of toxic gases, particulates, and aerosols.
Indirectly, animals may be affected by ingestion of contaminated food and water. Herbivores
are exposed when toxics are accumulated by vegetation from direct deposition on plant
surfaces or through the soil. Nutritional value of the vegetation may also be reduced.
Carnivores can be similarly affected by consumption of prey species that have bioaccumulated
toxics. Indirect effects also include the loss of potential prey and habitat degradation from
657
-------
the effects of the toxics.
Although animals are routinely used in laboratory testing of new chemicals and to
develop standards for human health, extrapolation of laboratory results to natural populations
and ecosystems rarely occurs. In the natural environment, communities may be less
susceptible than individuals to a particular stress because the more tolerant individuals may
replace the most susceptible ones, providing continuity in the community trophic dynamics.
Alternatively, communities may be more susceptible than individuals because of the loss of
species (or individuals) causing a change in the functioning of the ecosystem, resulting in a
pyramid of effects".
Bioindicator Selection
The use of animals as indicators of air pollutants, including air toxics, has been
reviewed in a number of recent reports12'13'14. Factors to be considered in selecting wildlife
indicators are trophic level, food habits, sensitivity to chemicals, availability for sampling,
population age structure (many substances, cadmium for example, show a strong tendency to
accumulate with age), and species mobility (which correlates with time potentially spent in
contaminated environments). For situations in which the air toxic accumulates on plant
surfaces or is incorporated in plant material, herbivorous animals would be best suited for a
monitoring program because larger quantities of an air toxic per unit body weight enter the
body via ingestion as compared to inhalation. Capture, collection, and sampling methods are
also important. Proper instruments and handling techniques need to be followed to avoid
contamination of samples after they are collected.
DISCUSSION
Many challenges remain in developing effective wildlife biomonitoring programs. No
standards exist for most of the air toxics and there is a critical need for information about
biological and ecological effects. There is a general lack of ambient data and few protocols
have been established for standardizing data collection and analyses.
The ecological impact of atmospheric pollutants emitted from point-sources depends
upon both meteorological and biological factors. Modeling may provide some help in
assessing and interpreting these interactions. Dixon and Murphy proposed a discrete-event
approach to predicting the effects of atmospheric pollutants on wildlife populations15. This
approach provides recurrent exposure to short-term pulses of high concentration and can be
correlated to seasonal variation in biological activity. Initial results of this type of model
suggest that the biological response to varying concentrations of pollutants may differ from
responses to chronic low-level exposures. Further work in this area is needed.
National biomonitoring programs have been used to some extent and new initiatives
are underway. The National Contaminant Biomonitoring Program has produced some broadly-
scoped information on the relations between biological populations and certain environmental
contaminants16. This program is currently restricted to collecting information on persistent
organochlorines and some inorganic contaminants including arsenic, cadmium, copper, lead,
mercury, selenium, and zinc. The Environmental Monitoring and Assessment Program (EMAP)
proposed by the Environmental Protection Agency will establish standard procedures and an
interagency network of ecological monitoring for a variety of materials (D. McKenzie, pers.
comm.). Efforts are also underway by the Fish and Wildlife Service to develop a National
Wildlife Refuge Monitoring Program focused on contaminants and wildlife. These types of
efforts will be further assisted by research conducted by facilities such as the Institute of
658
-------
Wildlife and Environmental Toxicology, which studies the use of chemicals as they affect
wildlife10.
Ecotoxicology is the emerging multidiseiplinary field of science that mixes toxicology,
which is concerned with effects at the level of the individual organisms, with environmental
chemistry, which measures occurrence of chemicals in the environment and analyzes processes
related to their distribution, and with ecology, which deals with the relations among species
and their abiotic environment17. Wildlife, as indicators and monitors, have an important
function in the linkage of these fields.
As new approaches are developed for detecting and monitoring air toxics, continuing
attention should be given to those wildlife species that best serve as biological filters for
selected toxics and for those sensitive species that occ upy key positions in regulating important
ecosystem functions.
ACKNOWLEDGMENTS
We thank Jim Fleming for his comments and input on discussions on this topic and
Rita Villella, David Lemarie and Judy Scherpetz for their reviews of early drafts.
REFERENCES
'J.R. Newman, "Effects of industrial air pollution on wildlife," Biol. Conserv. 15:181-190.
(1979).
2National Research Council, Animals as monitors of environmental pollutants. National
Academy of Sciences, Washington. 1979, 421 pp.
3S.A. Levin, M.A. Harwell, J.R. Kelly, K.D. Kimball (eds.), Ecotoxicology: Problems and
Approaches. Springer-Verlag, New York. 1989, 547 pp.
4Environmental Protection Agency, Washington, DC, EPA 560/4-88-005 (Jan. 1989).
SA.A. Pope, P.A. Cruse, C.C. Most, Toxic air pollution emission factors - a compilation for
selected air toxic compounds and sources. U.S. Environmental Protection Agency, Research
Triangle Park. 1988.
6S.A. Levin, K.D. Kimball, et al. "New perspectives in ecotoxicology," Environmental
Management 8:375-442. (1984).
7R. Stottlemyer, R., "Evaluation of anthropogenic atnospheric inputs on U.S. National Park
ecosystems," Environ. Manage. 11:91-97. (1987).
8D.G. Fox, A.M. Barruska, and others, A screening procedure to evaluate_air pollution effects
on Class I wilderness areas. U.S. Dept. Agricult., Foiest Serv,, Fort Collins. 1989, 36 pp.
9R. Lillie, Air pollutants affecting the performance of domestic animals. A literature review.
U.S. Dept. Agricult., Washington. 1970, 192 pp.
10M. Root, "Biological monitors of pollution," IJioScuince 40:83-86. (1990).
659
-------
UW. Beyer, Trace Substances in Environmental Health-XXII. A Symposium. University of
Missouri, Columbia. 1988, pp. 249-262.
12W. Grodzinski, T.P. York, "Species and ecosystem level bioindicators of airborne pollution:
an analysis of two major studies," Water. Air. Soil Poll. 16:33-53. (1981).
13J.R. Newman, R. K. Schreiber, "Animals as indicators of ecosystem responses to air
emissions," Environ. Manage. 8:309-324. (1984).
MC. D. Wren, "Mammals as biological monitors of environmental metal levels," Environ.
Monitoring and Assess. 6:127-144. (1986).
15K.R. Dixon, B.D. Murphy, Animals as monitors of environmental pollutants. National
Academy of Sciences, Washington. 1979, pp. 15-24.
16J. Jacknow, J.L. Ludke, N.C. Coon, Monitoring fish and wildlife for environmental
contaminants: The National Contaminant Biomonitoring Program. Fish and Wildlife Service,
Washington. 1986. 15 pp.
17D.W. Connell, "Ecotoxicology - A framework for investigations for hazardous chemicals in
the environment," Ambio 16: 47-50. (1987).
660
-------
EFFECTS OF AIR POLLUTANTS ON COLD-DESERT CYANOBACTERIAL-LICHEN
CRUSTS AND ROCK LICHENS: CHLOROPHYLL DEGRADATION, ELECTROLYTE
LEAKAGE AND NITROGENASE ACTIVITY'
Jayne Belnap
Resource Management
Canyonlands National Park
125 West 200 South
Moab, Utah 84532
Exposure of cold-desert cyanobacterial-lichen crusts on three
different substrates (sandstone, limestone a'nd gypsum) to
different pollution sources showed that while urban pollutants
in the Los Angeles basin, especially particulates, significantly
degraded chlorophyll on all three substrates, simulated acid rain
(pH 3.5, 4.5, 5.5 and 6.5; 1:1 sulfuric and nitric acid) had an
opposite, fertilizing effect on sandstone and limestone crusts.
Studies around a coal-fired power plant, comparing sites 9 and
12 km away from the plant with a control site 42 km away, showed
the same fertilizing effect on surrounding sandstone crusts.
However, less pH-buffered rock lichens had significantly
increased electrolyte leakage and chlorophyll degradation at the
nearer sites; nitrogenase activity in a crustal soil lichen was
depressed as well. When exposed to power plant effluents or
simulated acid rain, the degree of contact with the pH-buffering
substrate was important: cyanobacteria, embedded in soils that
buffered acidity, may use nitrates and sulfates as fertilizers.
Rock and soil lichens, with less contact and less buffering,
showed opposite effects. Chlorophyll degradation in crusts by
urban pollutants, especially particulates, suggests that
pollutants other than acid-producing or gaseous ones injure
crusts as well. Combined, these data suggest that the
deleterious effects seen in this study from power plant emissions
and simulated acid rain are caused by different agents than those
injuries due to urban pollutants.
661
-------
Introduction
Many studies have demonstrated the usefulness of non-
vascular plants for biomonitoring of air pollutants in humid
environments. Much less work has been done on species located
in semi-arid or arid regions. This study examined the effects
of different types of air pollutants on cold desert
cyanobacterial-lichen soil crusts and rock lichens from the
Colorado Plateau, using chlorophyll degradation, electrolyte
leakage arid nitrogenase activity as indicators of stress. Study
sites included a coal-fired power plant, dry deposition chambers
in the Los Angeles Basin, California, and wet deposition chambers
in Riverside, California.
Methods
The rock lichens Rhizoplaca melanophthalma. Lecanora
argopholis and Xanthoparmelia taractica, a soil lichen, Collema
tenax, and cyanobacterial crusts dominated by Microcoleus
vaqinatus and Scytonema sp. were used for this study. Samples
around the power plant were collected along a transect running
north-northeast from the Navajo Generating Station, with study
plots located at increasing distances from the plant (6, 12, 21,
and 42 km) . Plots were located within 150 m of the shores of
Lake Powell, and were accessed by boat. All plots were located
on Navajo sandstone, at approximately the same elevation (1200
m) and the same • exposures (north-northeast to north for rock
lichens, flat areas for cyanobacterial crusts and soil lichens).
Ten to 20 samples of each species was collected at each site.
Samples used in the dry and wet fumigation studies were collected
from Canyonlands National Park (sandstone-derived soils), Arches
N.P (gypsiferous soils) and Bryce Canyon N.P. (limestone-derived
soils).
Chlorophyll measurements were made using techniques outlined
in Ronen and Galun1. Absorption spectrums were measured in a
Hewlett-Packard diode array spectrophotometer. Optical densities
used for measurements were determined by scanning between 700 and
400 nm both non-acidified and acidified (using IN HC1) extracts.
Peaks were found at OD435 (chlorophyll a) and OD415 (phaeophytin)
for rock lichens; extracts of the cyanobacterial soil crusts
showed peaks at OD398 (chlorophyll a) and OD362 (phaeophytin).
Results were analyzed using analysis of variance (ANOVA) and
Duncan's multiple range test.
For nitrogenase activity, samples of the soil lichen Collema
tenax were incubated for 4 hours at 26 C in 2.5 cm diameter,
clear, gas-tight tubes with a 10% acetylene atmosphere.
Incubation was in a chamber lighted with ChromoSO (5000 K) and
cool white fluorescent bulbs; samples were analyzed on a Carle
FID gas chromatography equipped with a 8 foot, 8% NaCl on alumina
662
-------
column, using helium as carrier gas (30 ml/min). Injections were
.25 ml. Replicates were generally 10 per site.
Electrolyte leakage (membrane permeability) of the lichen
thalli was determined by methods outlined by Pearson2. Lichen
thalli analyzed were selected for similarity of surface area.
Thalli were humidified for 2 hours, rinsed for 3 seconds, and
then submersed in deionized water for 5 minutes. Conductivity
of the water was measured before and after immersion of the
thalli with a Fisher Scientific Conductivity Meter. Thalli were
then dried and weighed.
Results
Effects of Power Plant Emissions
The Navajo Generating Station ij> a coal-fired power plant
located near Page, Arizona, that is surrounded by a calcium
carbonate-rich sandy substrate. Effects of the plant on both
cyanobacterial crusts and rock lichens in the vicinity were
evaluated. Samples were collected at distances of 6, 12, 21
and 42 km from the plant. Chlorophyll degradation and
€ilectrolyte leakage were measured in rock lichens, while
chlorophyll degradation and nitrogencise activity were measured
for the soil crusts. Results for the rock lichens (Figure 1)
showed greatest chlorophyll degradation in Leconora argopholis
at the 12 km site; values were significantly different from those
at the 6 and 42 km sites. For Xanthoparmelia taractica. greater
degradation was measured at site 12 as well, but the differences
could not be shown to be statistically significant. Rhjzoplaca
melanophthalma showed significantly more chlorophyll degradation
at the 6 and 12 km stations than at the 21 and 42 km sites.
FLhizoplaca also showed significantly increased electrolyte
leakage at the two nearer sites when compared to the two more
distant sites (Figure 2).
The pattern of chlorophyll degradation in the cyanobacterial
crusts (dominated by Microcoleus vaginatus) near the power plant
was opposite that observed for rock lichens (Figure 3). There
was significantly less chlorophyll degradation at 12 km than at
the 6, 21, or 42 km sites on three of four sampling dates. On
the fourth sampling date, there was significantly less
degradation at the 21 km site. Chlorophyll degradation levels
for the soil lichen Collema tenax did not differ significantly
among the sites.
Nitrogenase activity was measured for Collema tenax in 1988,
1989, and 1990 (Figure 4). At all sample dates, nitrogenase
activity was depressed at sites 6, 9, 12 and 21 km from the plant
relative to control sites at 42 and 225 km from the plant. When
the 6, 9, 12 and 21 km sites were combined to make a basin
average for the area surrounding the power plant, that mean was
663
-------
statistically lower than that for the control site on four of the
five sampling dates.
Effects of Urban Air: Dry Deposition
Cyanobacterial-lichen soil crusts on soils derived from
three different substrates (sandstone, limestone and gypsum) and
a rock lichen, Dermatocarponmoulinsii, were exposed to urban
pollutants in the Los Angeles Basin using dry deposition chambers
(Figure 5). In 1988, an eight week exposure resulted in
significantly greater degradation of chlorophyll in crusts from
limestone and sandstone-derived soils, relative to controls in
filtered air chambers. Chlorophyll levels in mosses from all
three soils was unaffected by dry deposition, while the rock
lichen and crusts from gypsum soils showed significantly more
chlorophyll in the unfiltered chambers. Crusts in which
chlorophyll was degraded were dominated by cyanobacteria, while
the gypsiferous crust was dominated by lichens with green algal
phycobionts. The rock lichen tested also had green algal
phycobionts. The data suggest that green algae may be more
resistant to chlorophyll degradation from urban air pollutants
than are cyanobacteria. These observation were supported by the
1989 exposure of these organisms. In those trials, all crusts
were cyanobacteria-dominated, including the gypsiferous crust,
and all showed significant degradation of chlorophyll.
Effects of Wet Deposition
Cyanobacterial soil crusts were exposed to wet deposition
as well (Figure 6). Simulated acid rain (pH 3.5, 4.5, 5.5 and
6.5; 1:1 sulfuric and nitric acid) resulted in a negative
correlation between chlorophyll levels and pH (i.e. chlorophyll
increased as pH decreased) in crusts from limestone and sandstone
derived soils dominated by the cyanobacteria Microcoleus
vaginatus. This difference was statistically significant for the
sandstone soil crusts although statistical difference could not
be shown for the limestone. Sandstone soil crusts with high
levels of another Cyanobacterial species, Scytonema sp., showed
no significant effects. Gypsiferous soil crusts showed increases
in chlorophyll from pH 6.5 to pH 4.5, but crusts exposed to rain
of pH 3.5 showed a decrease in chlorophyll (p < 0.07).
Conclusions
The data suggest that the effects of power plant effluents
and simulated acid rain are strongly modified by the degree to
which organisms are intermingled with their pH-buffering
substrate. Cyanobacteria, embedded in soils that buffer acidity,
may use nitrates and sulfates washed out of the atmosphere as
fertilizers. Rock and soil lichens that have less contact with
the substrate are apparently less buffered and show opposite
effects, including significantly increased electrolyte leakage,
664
-------
increased chlorophyll degradation and decreased nitrogenase
activity. Chlorophyll degradation in soil crusts exposed to
urban pollutants, especially urban particulates, suggests that
pollutants other than acid-producing effluents injure crusts as
well. Combined, these data suggest that the deleterious effects
seen in this study from power plant emissions and simulated acid
rain are caused by different agents than those that cause
injuries in urban air. Finally, cyanobacteria may be more
susceptible to injury from urban sources than green algae.
References
1. R. Ronen, M. Galun, "Pigment extractions from lichens with
dimethyl sulfoxide and estimation of. chlorophyll degradation, "
Envir. Exp. Hot. 24:239-245.
2. Pearson, L. C. "Air pollution damage to cell membranes in
lichens: I. Development of a simple monitoring test," Atmospheric
Environment 19: 209-212.
665
-------
OD 435/416
OD 398/302; 436/416
05
OS
05
RHIZOPLACA LECANORA XANTHOPARMELIA
• 6KM Hih2 KM CD 21 KM • 42 KM
0.7
CRUST 12/88 CRUST 3/89 CRUST S/89 CRUST 11/89 COLLEMA 12/88
6 KM
12 KM Z • 21 KM ^^ 42 KM
Figure 1. Chlorophyll degradation ratios in three rock lichens near the Navajo
Generating Station in Page, AZ (expressed by the spectrophotometric optical
densities of 435/415, equal to chlorophyll a/phaeophytin). Letters a, b differ at
p<0.05.
Figure 2. Chlorophyll degradation ratios in cyanobacterial crusts and the soil
lichen Collema tenax near the Navajo Generating Station in Page, AZ. Letters a,
b differ at p<0.05.
-------
UMHOL/Md/ML
HICUMULtS/HR
12 21 42
DISTANCE FROM PLANT, KM
3/88 12/88 1/89 3/89
Hi PAGE BASIN ^^ 42 KM
1/90
at
OS
CHLOROPHYLL a/PHAEOPHYTIN RATIOS
OD 435/415; OD 398/362
GYPSUM LIMESTONE SANDSTONE 3 SANDSTONE B
• pH 8.5 iH pH 6.6 EZD pH 4.6 iH pH 3.5
1 -
0.8
o.e-
0.4
0.2
0 I
CHU«T COLLI
I CLEAN AIR f££2D\RTf MR
Figure 3-6: Electrolyte leakage in the rock lichen Rhizoplaca melanophthalma at different distances from the Navajo Generating Station. Letters a,
b differ at p<0.01. 4. Nitrogenase activity in the Page basin (combination of sites 6, 9, 12, and 21 km from the plant) that surrounds the power
plant, compared to a control site 42 km away. Letters a, b differ at p < 0.02. 5. Soil crusts on three different substrates exposed to dry deposition
in the San Bernadino Mountains, California. Letters a, b differ at p<0.05. 6. Soil crusts on three different substrates exposed to wet deposition
(1:1 sutfuric and nitric acid) at pH 3.5, 4.5, 5.5, and 6.5. Letters a, b differ at p<0.05.
-------
EVALUATION OF A DIFFERENTIAL OPTICAL ABSORPTION
SPECTROMETER AS AN AIR QUALITY MONITOR
R.K. Stevens, R.J. Drago, W.T. McLeod,
J.B. Bell, and R. Ward
U.S. Environmental Protection Agency, RTF, NC 27711
Y. Mamane
Technion, Haifa, Israel
H. Sauren
Agricultural University, Wageningen, The Netherlands
Abstract
Differential optical absorption spectrometer (DOAS) has been used
by a number of investigators over the past 10 years to measure a wide
range of gaseous air pollutants. Recently OPSIS AB, Lund, Sweden has
developed and made commercially available a DOAS instrument which has a
number of features which make the unit attractive for field monitoring
studies in remote and urban areas. The DOAS is composed of a broadband
light source (emission between 200-1000 nm) and a receiver-spectrometer
assembly. The spectral signals from the spectrometer are processed in
real time using a personal computer to calculate the concentrations of
the pollutants programmed to be monitored by the system. The distance
between the light source and receiver can range from 100 m to 2,000 m
depending on the pollutant to be monitored and species concentrations.
In September and October 1989 an OPSIS AB DOAS was operated in the
Research Triangle Park, NC on the roofs of the two main EPA laboratories.
The distance between the light source and receiver was 557 m and the
pollutants monitored were S02, N02, 03, HCHO and HN02. Comparisons
between the Federal Reference and Equivalent Methods measuring S02, 03
and N02 and simultaneous data derived from the DOAS showed excellent
agreement with correlations typically greater than 0.90.
668
-------
Introduction
Since 1979 several studies1'2 have been published describing the
application of Differential Optical Absorption Spectrometer (DOAS) to
measure over long paths, ambient concentrations of nitrous acid (HN02),
ozone (03), sulfur dioxide (S02), nitrogen dioxide (N02), formaldehyde
(HCHO), and nitrate radical (N03). Recently a commercial version of the
DOAS system has been made available by OPSIS AB, Lund, Sweden.3 This
OPSIS (DOAS) is composed of a 150 watt high pressure xenon lamp assembly,
optical light receiver, fiber optic cable, spectrometer and personal
computer (PC). The spectrometer and software signal processing system
rapidly converts UV-visible light intensity signals at 10-60 second
intervals into ug/m3 concentrations of a large number of gaseous
pollutants.
In most urban areas of the United States, the Environmental
Protection Agency (EPA) requires of a number of air pollutants (eg, S02,
N02 and 03), termed criteria pollutants, to be monitored with EPA
designated methods. These methods are typically based on instrumental
continuous monitoring principles and once designated by EPA are approved
as Federal Reference Methods (FRM's) or Equivalent Methods. These FRM's
are operated at fixed site locations. The EPA has an interest in
comparing FRM ambient pollutant monitoring procedures at fixed sites with
methods that determine the average concentration of the criteria
pollutants over an open path. Comparisons between FRM's and a system
such as a DOAS would assist in determining the influence of atmospheric
pollutant inhomogeneities on FRM measurements and also assess the
possibility of using long path procedures as FRM's.
In September and October 1989 a commercially available DOAS system
was operated in the Research Triangle Park, NC. The system was
programmed to measure S02, N02, 03, HCHC and HN02. During this same
period FRM measurements of S02, 03 and N02 were obtained at the same
location as the DOAS spectrometer receiver for comparing fixed site and
long path pollutant monitoring measurements. While all five of these
pollutants were measured in this study, this report will focus on the
DOAS and FRM measurements of S02, 03 and N02.
Experimental
Federal Reference Methods. In this study sulfur dioxide measurements
were made using a TECO Model 43 pulsed fluorescent S02 system ( EPA
equivalent analyzer EQSA 0276-009). The N02 measurements were made with
a Combustion Engineering N0-N02-N0x Model 8101-B, operating in the 0-
0.5 ppm range (EPA reference analyzer RFNA 0479-038). The ozone was
measured using a TECO Model 49 UV photometric 03 Analyzer operating in
the 0-0.5 ppm range (EPA equivalent analyzer EQOA 0880-047). The FRM
instruments were all calibrated at the beginning and the end of the study
with standard calibration methods using dynamically generated gas
mixtures. The ozone analyzer was calibrated with a TECO UV photometric
03 calibrator model 49PS which was previously certified NBS traceable.
Data from these reference methods were collected by a Campbell Model 21X
data logger and hourly averages stored. Periodically data was retrieved
from the data logger with a PC.
The reference analyzers were operating from the 2nd floor of the EPA
Annex in the RTP, NC as shown in Figure 1. Ambient air samples were
drawn into the monitors through a 4 meter by 4 mm ID Teflon tube inlet.
The inlet was extended through the ceiling above the analyzers to an
669
-------
outside sampling rain shield on the roof of the Annex adjacent to the
DOAS receiver.
Description of Differential Optical Absorption Spectrometer. The DOAS
light transmitter was positioned on the roof of the US EPA Environmental
Research Center (ERG). The DOAS light receiver (Figure 1) was positioned
on the roof of the EPA Annex adjacent to the inlets of the EPA reference
analyzers. The light path was approximately 20 m above ground and passed
directly over a portion of Interstate 40. Visible and ultra violet light
from a xenon lamp, collimated by a parabolic mirror, passed through the
outside air over the 557 m path from the roof of the ERG to the roof of
the Annex. On the roof of the Annex the light from the xenon lamp is
recollimated in a receiver assembly and focused onto the entrance fitting
coupled to a fiber optic cable. The fiber optic cable transmits the
light to a spectrometer, composed of a mirror, grating, chopper and
photomultiplier tube, located 4 meters below in an air conditioned room.
The DOAS system used in this study was calibrated to measure S02, N02,
03, HCHO and HN02. The calibration3 of the DOAS is based on the use of
reference gases introduced into special DOAS spectrophotometric cells of
precise dimensions. The reference intensities and spectra are stored in
the computer to statistically compare with ambient measurements to obtain
concentration data.3
DOAS Measurements were performed by integrating for 1 minute the
wavelength (selected windows from 260 to 460 nm) for each of the five
pollutants in sequence. The computer processed the spectral absorption
signal and displayed the updated concentrations for each pollutant every
five minutes. Twelve one-minute readings were used to calculate the
hourly average concentrations reported by the DOAS.
Results
Figures 2 and 3 are time series and correlation plots of DOAS and
FRM measurements for S02 and 03 obtained in September and October 1989
in the Research Triangle Park, NC.(similar plots were obtained for N02).
These data show that in most instances the FRM's and DOAS measurements
are in good agreement. As shown in Figure 3, during some short periods
the DOAS 03 values were significantly lower than the FRM values. These
variations between the DOAS and FRM's may be explained by atmospheric
inhomogeneities produced by such factors as the diffusion of nitric oxide
(NO) emissions from mobile source traffic on Interstate 40 into portions
of the DOAS absorption path. The nitric oxide rapidly reacts with the
03 thus reducing the 03 concentration in the open path compared to the
FRM 03 measurements. Conversely the small variations in N02 measurements
between the DOAS and FRM may also be explained by the higher
concentrations of N02 over portions of the DOAS absorption path
associated with the roadway emissions.
During the study the S02 concentrations fell below the detection
limit of the TECO pulsed fluorescence monitor. Also, the DOAS instrument
appears to record S02 concentrations 2-3 ppb above the TECO instrument.
The uncertainty in the calibration of the TECO S02 monitor is thought to
be the source of this variation. The standards used to calibrate the S02
instrument are no better than + 27, at the 100 ppb level.
During the study there were several periods when the DOAS signal
dropped to zero. These periods were associated with heavy rain and high
humidity. Addition of heating tape to the receiver optics eliminated the
fogging of mirrors in the receiver, which was thought to be part of the
670
-------
problem. However, during periods of very high humidity, rain and fog,
the light from the xenon lamp was scattered enough to interfere with the
DOAS absorption measurements. This occurred for a few hours during the
one month study.
The pollutants measured by DOAS are species that originate from a
variety of sources in the troposphere. Since the monitoring was
performed on the roofs of EPA facilities rather than at ground level, the
influence of single vehicles or other point sources on the DOAS and FRM
measurements was assumed to be minimal. The data from this study suggest
this assumption was reasonable.
Conclusions and Recommendations
DOAS long path measurements for S02, N02 and 03 in Research Triangle
Park, NC were in excellent agreement with FRM instrumentation. The
correlation coefficients between FRM and DOAS were very high (r >0.9)
and showed no clusters of data. Some variations between the FRM's and
DOAS instruments maybe associated with atmospheric inhomogeneities. The
major advantages of the DOAS instrumentation are its low detection
limits, (0.5 ug/m3 for the 557 m absorption path for S02, N02, and 03),
multiple pollutant monitoring capability, long term calibration stability
and rapid response characteristics.
A disadvantage of the DOAS system is the loss of signal during
periods of fog associated with humid conditions. This maybe minimized
by reducing open paths to 200-300 m. The DOAS system would be an
excellent tool for measuring area concentrations of air pollutants to
determine the sources of pollutants that impact air quality.
The DOAS instrument is capable of monitoring a number of gas phase
air pollutants over different paths simultaneously. This multi-open
path DOAS configuration will be evaluated this summer as part EPA's
VOC/03 study in Atlanta.
Acknovledgemen ts
The authors wish to thank R. Karlsson, L. Olson, W. Karches for
assisting in installation and operation oE the DOAS.
References
1. U. Platt and D. Perner, "Detection of nitrous acid in the atmosphere
by differential optical absorption," Geophvs. Res. Lett. 6: 917
(1979).
2. U. Platt and D. Perner, "Direct measurement of atmospheric
CH20,HN02,03 and S02 by differential optical absorption in the near
UV," J. Geophvs. Res. 85: 7453 (1980).
3. R. Karlsson, "Environmental control using long path measurements,"
Proceedings of the 1990 EPA/AWMA International Symposium, Res. Tri.
Park, NC. (May 1990).
671
-------
HWYM
Figure 1. Location of the FRM's and the path over which the DOAS data
in the Research Triangle Park, NC were obtained.
672
-------
DOAS ,FRM ++ +
-5-
18-Sep-89 22-Sep-89 25-Sep-89 29-S*p-89 02-Oct-89
20-Sep-89 23-Sep-89 27-Sep-89 30-Sep-89 04-Oct-89
Date
25
x.
IX
O
O
Q
-5
-5
15
25
10
20
FRM Monitor, SO2 in ppb
Figure 2. Time series (top) and correlation (bottom) plots of DOAS and
FRM measurements for S02 obtained in Research Triangle Park, NC from
September 18 to October 4, 1989. The square of the correlation
coefficient is 0.96 and the regression equation is:
DOAS(S02) = 1.01 * FRM(S02) +1.78
673
-------
c.
CL.
50-
O 20-
"1 U •mi m 1111111111 mill i UN 11 mini
19-Sep-89 21-Sep-89 24-Sep-89 27-Sep-89 03-Oct-89
20-Sep-89 23-Sep-89 25-Sep-89 01-Oct-89 04-Oct-89
Date
50
a.
d
O 20-
3
o
Q
-10
20
40
60
10
30
50
FRM Monitor, O3 in ppb
Figure 3. Time series (top) and correlation (bottom) plots of DOAS and
FRM measurements for Ozone obtained in Research Triangle Park, NC from
September 18 to October 4, 1989. The square of the correlation
coefficient is 0.89 and the regression equation is:
DOAS(Ozone) = 0.90 * FRM(Ozone) - 2.46
674
-------
Environmental control using
long path measurements
Ronald Karlsson, Ph. D.
OPSIS AB
Ideon Research Park
S-223 70 Lund, Sweden
INTRODUCTION
Identifying, measuring and recording the types and quanti-
ties of molecules in a gaseous mixture has historically been
very difficult. Gases comprise various types of molecules,
and the problem has been to differentiate them in terms of
qualitative and quantitative measurements. The technology
known as differential optical absorption spectroscopy (DOAS)
provides a solution to a large part of the problem. DOAS is
a purely physical measurement method that is based on
simple, well-known laws of physics.
An effective system for analysis of air pollution on the
basis of DOAS technology was developed at the Department of
Atomic Physics at the Lund Institute of Technology in Sweden
by a group of research scientists that included Svante
Wallin and Leif Uneus.
On the basis of their experience, Wallin and Uneus estab-
lished the OPSIS company in 1985. Today, the company has
more than 25 employees and is recruiting new personnel at an
average rate of one per month.
The following presentation describes the measurement prin-
ciples applied in OPSIS technology as well as a number of
typical applications.
675
-------
OPSIS measurement technology
The OPSIS system for analysis of pollutants in gaseous form
measures the average value of pollutants in a beam of light
between a transmitter and a receiver. An OPSIS system
consist of the following units, as shown in Fig. 1:
- Transmitter
- Receiver
- Power pack
- Opto-analyzer
- Modem.
Measurements are based on optical absorption spectroscopy.
Analysis of the parts of the spectra which are missing in
the light received reveals the presence of various substan-
ces. Measuring the guantity of light that is missing enables
the concentrations of these substances to be determined.
Quantitative determination is based on Lambert-Beer's law of
absorption:
C = Log (I'o/I)
(eL)
where
C = concentration
I'o = light intensity before differential absorption
I = light intensity after absorption
e = average differential absorption
L = length of the absorption path.
The light from the transmitter is generated by a high-pres-
sure xenon lamp in the form of a concentrated beam of light
that includes wavelengths from short-wave UV to long-wave
IR.
The receiver is trained on the transmitter, which captures
the light and sends it over a fiber-optic cable to the opto-
analyzer, the central unit in the system.
The opto-analyzer comprises a spectrometer, electronics for
collection and processing measurement values, a hard disk, a
computer for presentation and communication, and a modem.
As noted, light is sent from the receiver over a fiber-optic
cable to the analyzer, in which it is broken up into spectra
by a grating. This grating is mounted on a stepper motor
that enables complete control of the wavelength interval.
The spectrometer also contains a system of mirrors and a 50-
mm photomultiplier. In front of the photomultiplier is a
rotating disc with 20 axially oriented slits about 0.2 wide.
The disc rotates at about 5 revolutions per second.
676
-------
This arrangement enables about 40 nm of wavelength interval
to be scanned in 10 ms over about 1000 channels, with a
resolution of about 0.2 nm and an overlap of 80%.
The computer thus collects about 100 spectra per second of
the wavelength in question. These spectra are converted to
digital signals and stored in a nulti-channel memory. The
computer then compares these spectra with a pre-calibrated
reference spectrum, wavelength by wavelength. Comparison can
cover up to 1000 wavelengths, each of which contributes to
the determination of concentration. A proprietary calcula-
tion program then computes the concentrations of the sub-
stances which the computer is programmed to analyze.
The program reports the margin of e:rror for each measurement
value and also reports on light transmission, i.e. it indi-
cates how much of the light transmitted actually reached the
receiver. These values can be used to check the function of
the system.
The analyzer operates completely automatically. Measurement
results are stored on the hard disk and can be presented on
the VDU and/or a printer and can also be transmitted over
the built-in modem to a host computer. The analyzer can also
be remote-controlled through the modem.
Calibration
An OPSIS measurement instrument is calibrated by measuring a
known quantity of a specific substance in the OPSIS labora-
tory, under precisely controlled conditions of pressure and
temperature. The solvent content is» determined on the basis
of a table of steam-pressure data.
The absorption curve is stored in digital form, after which
the instrument is ready for use. Recalibration is required
only in exceptional cases.
An instrument which is often moved between measurement
points should be zero-calibrated once a month using an OPSIS
CA 075 calibration unit. Zero-calibration takes about 5
minutes per gas. In some cases, it may be necessary to check
system performance using standard gases. This is a simple
procedure and involves specially designed measurement cells
of a precise length, normally 5-10 mm. The cell is connected
with 2 fiber-optic cables between the analyzer and a trans-
mitter fitted with the same type of lamp used in actual
measurements. Gas is allowed to flow through the cell, and
concentrations are recorded as in a normal measurement pro-
cedure. Three to five consecutive measurements are usually
made for each gas, and possible deviations between measured
and theoretical concentrations are adjusted with the change
span and offset programs in the analyzer.
677
-------
ESTIMATION OF ERROR MARGINS
Two types of error can arise during OPSIS measurements:
1. Absolute errors resulting from calibration faults, incor-
rect values for the measurement path, or incorrect values
for temperature or pressure.
2. Relative errors, depending on the measurement instrument.
Absolute errors
Errors relating to calibration of gas normally deviate from
the "correct" value by a maximum of ±5%. The length of the
measurement path can usually be determined to within ±1%.
Temperature and pressure can normally be measured with an
accuracy of ±1%.
Relative errors
The signal-to-noise ratio generates a maximum error of
±0.5%.
It can be seen that the total error during normal measure-
ments amounts in the worst case to about ±8%. In most cases,
measurement error is about ±5%. Under extreme conditions,
measurement errors of 10-15% have occurred.
MEASUREMENT
The flexibility of the OPSIS measurement system enables it
to be used for a wide range of applications. The system is
currently used for the following types of operations:
A. Monitoring of pollutants in ambient air
B. Monitoring of flue-gas emissions
C. Process control
D. Analysis of air pollution in the working environ-
ment.
I shall now briefly describe typical systems for these app-
lications.
678
-------
Monitoring of pollutants in ambient air
For surveillance of the air-pollution situation in a city,
the transmitter and the receiver are normally positioned on
rooftops, at distances of 100-2000 meters. The use of fiber-
optic cable enables the analyzer to be installed in a shel-
tered place.
With a multiplexer connected, the analyzer can support up to
12 pairs of transmitters/receivers, which means that a very
large area can be monitored within the city.
In Gothenburg, the second largest city in Sweden, four mea-
surement stations have been positioned in areas with inten-
sive pollution. Each measurement station includes 3-5 mea-
surement paths of 200-2200 meters in length, for a total of
16 paths, which in effect covers the entire city. See Fig.
2.
All measurement data from these stations are transmitted
together with meteorological data to a central computer, in
which a specially developed program based on a dispersion
model generates precise determinations of pollution levels.
The results are used for traffic planning, among other
things.
Almost 100 OPSIS systems have now been installed in Scandi-
navia and the rest of Europe.
Other applications for monitoring air pollution in ambient
air include:
- Measurement of pollution at street level
- Monitoring of emissions from a factory
- Measurement of background emissions.
Fig. 3 shows a summary of the measurement parameters, measu-
rement areas, detection limits and other factors which are
relevant to the applications described.
Monitoring of flue-gas emissions
In terms of flue-gas monitoring, OPSIS has focused on
systems for power plants, heating plants, co-generation
plants and waste incineration facilities.
Precise, cost-effective control of emissions is a vital
aspect of operations at modern power or heating plants.
OPSIS remote-controlled systems provide unique benefits for
emission control. A single system can be used to measure
SO2, NO, N02, NH3, Hg, H2O and C02 at one or more measure-
ment stations. A typical system configuration is shown in
Fig. 4.
679
-------
Emission control is also a very important activity in con-
nection with incineration of household and/or industrial
waste, particularly with reference to monitoring of HC1
concentrations.
OPSIS systems have proved very attractive to operators of
waste-incineration plants, as the technology provides fast,
accurate information on concentrations of NOX and Hg as well
as other parameters.
Configuration and installation of OPSIS systems for these
applications is similar to those for power and heating
plants.
Process control
The great flexibility of OPSIS technology has led to instal-
lations for monitoring of a number of industrial processes
throughout Europe, such as production of sulphuric acid. A
single OPSIS system can be used to measure S02 concentra-
tions, from strong gases with 10-15% SO2 to residual gases
where concentrations are measured in ppm.
NEW DEVELOPMENT
In terms of analysis, the OPSIS system has been expanded to
cover sections of the infrared spectrum, up to about 1,800
nm. This involves fitting the analyzer with an extra grating
and an IR-sensitive detector. A large number of hydrocarbons
as well as a number of other substances can thus be detected
and measured.
A combined receiver/transmitter system has also been de-
signed, both fixed and computerised. These units are com-
bined with retroreflectors that enable the wavelength of the
light to be doubled, and thus provide a 100% increase in
sensitivity. The retroreflector does not have to be instal-
led on a fixed foundation, which gives greater flexibility
for choice of measurement paths. This type of transmit-
ter/receiver system can be mounted on a container or the
roof of a car.
680
-------
In conclusion, I would like to answer a question that is
often asked, namely "How can we be sure that the values
obtained are reliable?"
As previously mentioned, OPSIS technology is based on simple
physical laws that have been well-known for many years.
However, the technology is new in a commercial sense and our
measurements have been repeatedly compared with conventional
technology both in outdoor air and in process plants.
Figs. 5 and 6 show the results of several comparative tests.
SUMMARY
Feed-back from customers who have installed OPSIS equipment,
from measurement assignments and from comparative tests has
clearly shown that OPSIS measurement technology is highly
reliable and that is has a very promising future. The main
benefits of OPSIS technology include:
- No physical handling or preparation of samples
- High sensitivity and accuracy
- Large dynamic measurement area
- Measurement paths from 0.1 to 2000 meters
- Multipath measurements (12 per station)
- Simultaneous or sequential measurement of different
substances
- On-site measurement and direct presentation of
results
- Minimal calibration requirements
- Easy installation, ready-to-irun
- Minimal maintenance requirements.
REFERENCES
1. U. Platt, D. Perner and H.W. Patz; J.Geophys. Res. (1979)
84 6329-.
2. H. Edner, A. Sunesson, S. Svanberg, L. Uneus and S.
Wallin; Applied Optics (1986) 25(3) 403-.
681
-------
Figure 1
Outdoor air
Transmitter
Power supply
unit
Opto-analysis unit
Figure 2
The four measurement stations are strategic-
ally located. All measurement data are
displayed directly on a computer monitor at
the office of the Gothenburg Municipal
Association.
682
-------
Figure 3
PERFORMANCE DATA
Parameter:
Measurement range:
Minimum detectable quantities:
(fcomtQ"nf p»in 500 m. m«injrefti»m timt 5 i""l
Calibration
2«fO'i»o*m canor»iion
Other calibration
Z«fe-pasm n*o«*'ty (W maittfH-
Linearity
Recommended length of
monitoring path:
Maximum length of
fiber-optic cable:
NO/NH,
0-2000
Mg/m'
2^g/m"
CA075
±4 «^m]
±1%
100-200 m
10m
NO/SOj
0-2000
^g'm1
lM6/m'
CA075
±2«g/m"
±1%
300-flCX)m
50m
Q,
0-200C
nt/m*
3M^mJ
CA07S
tfiufc'IlT'
±1%
300-800 ti
25m
NO,
0-500
pj/m1
0-1 jig/m*
CA075
±0.2 (j|/m*
±1%
300-800 m
50m
MNO^Form-
ildenyde
0-2000
(igfln1
2Mfr'T1J
CA075
±4ngA»<'
±1%
3GO~8GOm
SOm
H,0
o-ioo
gmi'
0-lg/m'
CA075
±0.2gm»
±1%
3QO-SOQ m
SOm
BftfUtnft^
ro*y0^ft?3£y^"c
0-2000
f»g/mj
S^g/m*
CA075
±10 MA'"1'
±1%
30O-800IH
25m
Figure 4
MukipJexer
Typical system layout for a power plant.
683
-------
OZON (03)
12. - 14. April 1989
Figure 5
yg/m—3
0-00 !2.00 0.00 12.00 0.00 tt.OO
• DOAS —«— MESSWAGEN
Figure 6
STICKSTOFFDiOXlD (NO2)
12. - 14. April 1989
yg/m—3
0-00 12.00 0.00 12.00 0.00 12.00
' DOAS —»- MESSWAGEN
From: Karstonales Amt fuer Industrie, Gewerbe und Arbeit.
Pachstelle Luftreinhaltung
Lauper.str. 22, CH-3011 Bern, Schweiz
684
-------
OCMFftKISCN OF HUG ROH FP-IR DRTA TO M»LE ALR
G\NISTER CKm FECK A CCKISDUfD DEWDJD POINT !30QRCE
M..L. Spartz, M.R. Witkowski, J.H. Fateley, R.M. Haimnaker, W.G.Fateley,
Department of Chemistry/ Willard Hall, Kansas State University,
Manhattan, Kansas 66506;
R..E. Carter, M. Thomas, D.D. Lane, Department of Civil Engineering,
G..A. Marotz, Department of Physics and Astronomy, Learned Hall,
University of Kansas, Lawrence, Kansas 66045;
B..J. Fairless, T. Holloway, J.L. Hudson, J. Arello,
U.S. Environmental Protection Agency Region VIII,
25 Funston Road, Kansas City, Kansas 66115;
D.F. Gurka, U.S. Environmental Protection Agency/
Quality Assurance Division, Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada 89119.
Two air monitoring techniques for atmospheric volatile organic
compounds (VOCs) : long path FT-IR and whole-air canister GC/FID have been
tested simultaneously and the data compared. The long path FT-IR
spjectrometer system is being developed at Kansas State University to measure
air toxics over varying distances. The whole-air canister GC/FID method has
been used for several years at the University of Kansas for air toxics
analysis and is an ideal system to test the performance of the FT-IR system.
The experiment, which took place at the University of Kansas, has the
two techniques set-up co-planar with and perpendicular to a controlled
upwind point source. The path length between the FT-IR spectrometer and
source was varied between 50 and 232 meters. Five or six collection
canisters were then placed directly along the IR path, evenly spaced,
perpendicular to the wind-stack plume center line. The plume was generated
from a 5 foot stack and allowed to drift unimpeded downwind to the co-
monitoring systems. For these experiments the collected data were
concurrently acquired over 12 and 27 minute time periods. During this time
period meteorological data was also acquired for the calculation of
concentration in ppb.
Many different parameters and compounds were tested during these co-
monitoring experiments. The monitoring set-up, data analysis, various
parameters of concern, and comparison of the two methods will be addressed.
685
-------
Introduction
A field transportable Fourier transform infrared (FT-IR) spectrometer
system for monitoring volatile organic compounds (VOCs) in the atmosphere is
presently undergoing rigourous testing. A testing and evaluation plan was
devised by Jody Hudson U.S. - EPA Region VII and edited by the Kansas State
University (KSU) and University of Kansas (KU) groups. The three phases of
this plan are the controlled releases at KU, the uncontrolled releases at a
well characterized site and the uncontrolled releases at a complex VOC site
where little or no pre-test characterization data have been acquired.
During phase one testing, many modifications were made to the FT-IR
spectrometer system including instrumentation, data collection and data
analysis. Seme data have been accumulated in all of these phases, however
the majority of work to date has been with controlled releases at KU.
Phase one will be the focus of this paper and presentation. This phase
is broken into six parts which have been completed over the last year.
These six parts were designed to find the practical limitations of the FT-IR
system. The six areas are qualitative and quantitative measurements of
single VOCs, qualitative and quantitative measurements of mixtures with
similar and dissimilar VOCs, VOC detection limits, path length variations,
meterological parameters including humidity and wind velocity, and
scattering by particulate matter.
The collection of data from phase one is now completed, with the
exception of testing at different vertical heights. The data are being
interpreted by both groups at present and the analysis will be completed in
the near future. Comparisons between the KSU and KU data, taken in the
spring and fall of 1989, exhibited differences in quantitative results . In
the most recent measurements we attemped to address these discrepancies.
The recent and past work will be discussed along with modifications to the
experimental design and data collection, that may alleviate these
discrepancies.
Experimental Methods
Equipment
The spectra are collected using a Bomem DA02 FT-IR spectrometer,
equipped with a germanium on potassium bromide (Ge/KBr) beam splitter. This
spectrometer is equipped with a f/4 input collimator and a f/1 fast output
optic to the detector. A broad band MCT (Mercury Cadmium Telluride)
detector, 5000-500 cm-1, is operated at liquid nitrogen temperatures,
approximately 77K. The instrument is purged with dry air from a portable
air dryer. The collection optic is a 10 inch Cassegrainian telescope with
an effective focal length of 150 cm, Figure 1. The primary source optic is
a 20 inch f/4 paraboloid with the appropriate elliptical flat secondary
optic and both are gold coated front surface mirrors. The Dobsonian mount
for the mirrors forms a Newtonian collimating telescope. The infrared
source, which is located at the focal point of the Newtonian collimating
telescope, is a high temperature Nernst glower operated at 2073 K. The
source is powered by a 1600 Watt portable generator. Distances between the
source and instrument are measured accurately using electronic distance
measuring equipment. Mobility is provided by a Journey mobile home designed
to house and transport the necessary measurement instruments and crew.
Field Testing
The phase one testing at KU incorporated a single pass geometry with
the IR source remotely placed from the spectrometer. The infrared source
was placed at the focal point of the remote telescope to produce the
collimated beam which was then passed to the spectrometer, Figure 2.
686
-------
Various path length measurements at KU have been carried out from 50 to 232
nieters. Initial observations at different industrial sites involved various
open path lengths from 40 to 600 meters. In phase one, the KU group placed
a VDC plume generator upwind and perpendicular to the IR path, Figure 3.
Colocated along the IR path were 5 or 6 evacajated stainless steel canisters
that were opened during the IR data collection to collect whole-air samples,
see Figure 2. The stainless steel canisters are then analyzed using gas
cliromatography / flame ionization detection (GC/FID).
!>rocedure
Modifications to the FT-IR spectrometer have been recently made to
alleviate discrepancies between the KSU and JU quantitative data. A silicon
filter was placed just before the detector to mask the higher frequency
region (when not needed) and only pass frequencies below 1600 cm . Ihis
filter was added to cut down on the detector nonlinearity at high quantum
fluxes. By masking out the high frequency region more low frequency energy
iray fall onto the detector and produce a better signal to noise ratio. The
present scan parameters have been changed from our initial work and are
compared in Table I. The spectrometers quantitative accuracy was tested by
acquiring a spectrum of a 50 micron polystyrene film as a pre-test after
every new setup. Polystyrene placed at the aperture before the actual data
collection should afford a reproducible spectrum in frequency and intensity
if the spectrometer is functioning correctly. This test is simple because
polystyrene can be slid in and out of the IR beam between the input
telescope and the aperture and polystyrene has a number of IR bands that can
be readily measured.
The actual data acquistion has changed slightly as well and is as
follows: measure instrumental emission, measure a background (upwind if
passible or when VDCs are not being released) and then measure the sample
spectrum. The background and sample spectrum are corrected for instrumental
emission and are then ratioed to produce the absorbance spectrum.
Originally, instrumental emission was omitted due to the long analysis time
and what were believed to be small energy fluxes. Instrumental emission is
now being acquired and take into account to produce more correct
absorbances.
Analysis
The resultant spectra are analyzed by interrogating the spectra by eye.
The major absorbances from recognizable compounds are analyzed first,
followed by major absorbances at unrecognized frequencies. The unknown
frequencies are typed into a in-house frequency data base and are searched.
These components are then stripped out of the absorbance spectrum using
calibration spectra, if available, to obtain qualitative data, to obtain
quantitative data and to reveal residuals. The residuals are then analyzed
as previously mentioned.
Results
The qualitative results from earlier KU work were very encouraging to
the authors; however, the quantitative results for this FT-IR method seem to
be consistently lower than those for the canister method by 20 to 50
percent. These discrepancies were addressed and the system re-evaluated.
The addition of the silicon filter for a decrease in saturation
nonlinearities and the polystyrene as a standard test will afford us the
ability to have the instrument performing tiie same in the field as in the
Iciboratory calibrations. The very preliminary IR results from the recent KQ
project are seen in Table II. New calibration curves will be produced by
using the new data acquisition method. l,l,l-jrrichloroethane was released
at: KU as a single component sample and the preliminary results from both
687
-------
methods have been calculated. The whole-air canister data produced an
average concentration over a 100 meter path length of 300 ppb. The FT-IR
data analyzed this single component to be 175 ppb. This discrepancy could
be a result of different vertical measuring zones in the atmosphere. The
spectrometer when placed on the present tripod has a viewing height of
approximately six feet. The stainless steel canisters were at a height of
about 3 feet. Future testing at KLJ will address these height differences.
Qualitative and quantitative results were obtained for single
component samples, for mixtures, for different IR path lengths, for varing
meteorological conditions, and for particulate added plumes. None of these
changes seemed to cause any degradation of the spectrometers performance.
At longer IR path lengths however, the absorbances due to water and carbon
dioxide can limit analysis in certain spectral regions. The mixtures
tested at 100 meters had smoke added to the plume at the VOC generator as
seen in Figure 3. The smoke was added to see what problems particulate
matter would cause in the collection of the infrared data. No major
problems were noted, however the IR beam was slightly attenuated by the
smoke.
Conclusions
We believe that once the bias of the two data set techniques is within
5 to 10 percent with respectable precision the system will be site ready.
Further research at the University of Kansas may be necessary to address
height profiles of the VOC plume. Work is presently scheduled at a site in
south central Kansas and other sites outside of Kansas. The Kansas site
will include canister analysis to further help characterize the FT-IR
systems capabilities.
Acknowledgements
This work was supported by funds from the U.S. Environmental Protection
Agency, Quality Assurance Division, Environmental Monitoring Systems
Laboratory, Las Vegas, Nevada, under cooperative agreement numbers, CR-
814059-01 , CR-814059-02, and CR-814059-03.
A Phillips Petroleum research fellowship was awarded to Martin L.
Spartz, which partially supported his fourth year graduate work salary.
Referer
1. J. L. Hudson, "Project Overview for Performing a Comprehensive
Evaluation of a Field Transportable FTIRS Remote Sensing
Instrument," Region VII, Environmental Services Division, U.S.
Environmental Protection Agency, (1989).
2. M. L. Spartz, M. R. Witkowski, J. H. Fateley, J. M. Jarvis, J. S.
White, J. V. Paukstelis, R. M. Hammaker, W. G. Fateley, R. E.
Carter, M. Thomas, D. D. Lane, G. A. Marotz, B. J. Fairless, T.
Holloway, J. L. Hudson, and D. F. Gurka, "Evaluation of a mobile
FT-IR system for rapid VOC determination, Part I: Preliminary
qualitative and quantitative calibration results," American
Environmental Laboratory, Nov.: 15 (1989).
3. G. A. Marotz, D. D. Lane, R. E. Carter Jr., R. Tripp, J. Helvig,
"Preliminary results from a rapid deployment field study of heavy
gas detection and dispersion using a whole-air technique," 80th
Annual Meeting of APCA, New York, New York, 1987.
688
-------
Although the research described in this paper has been funded by the
U.S. Environmental Protection Agency, it has not been subjected to Agency
review and, therefore, does not necessarily reflect the views of the Agency/
and no official endorsement should be inferred. Mention of trade names or
comraercial products does not constitute endorsement or recommendation for
use.
Table I
mi-riai ana Ftesenc ^
Parameter
Resolution:
Scan Speed:
Coadded Scans:
Apodization:
Lew pass filter:
Optical filter:
Data Acquisition Time:
jecLTuneijer bear i ***
Initial
0.1 cm"1
1.5 cm/sec
256 scans
Boxcar
50 KHz
none
27 min
Present
0.5 cm"1
1.5 cm/sec
512 scans
Hamming
10 KHz
Si for <1600
12 min
cm
Data may be taken at other parameters and still compared. This table
is present to illustrate the variability of data acquisition.
Table II
Preliminary FT-IR Data from FD work in April 199O
Ctroound Distance fm)
1,1, 1-Trichloroethane
Methylene Chloride
Methylethyl Ketone
Toluene
t-Butanol
50
50
50
50
50
FT-IR
Mixture 1
56
136
52
35
a
Concentration (tekO
Mixture 2
138
31
151
85
a
Mixtures
38
106
107
233
a
1,1,1-Trichloroethane 100 b
Methylene Chloride 100 b
Methylethyl Ketone 100 b
Toluene 100 b
t-Butanol 100 b
1,1,1-Trichloroethane 200
Methylene Chloride 200
Methylethyl Ketone 200
Toluene 200
t-Butanol 200
35
80
32
18
a
16
42
10
20 C
a
75
31
56
116 C
a
34
13
29
30
a
18
63
80
146
a
7
29
36
42
a
a Calibration curve on t-butanol has not been produced.
b Smoke was added to check the affect of airborne particulates.
c The major bands of toluene are strongly overlapped and measurement is
difficult.
689
-------
Figure 1 Coplanar alignment of the FT-IR spectrometer system with the
evacuated stainless steel canisters. The IR source (not seen) is
positioned just outside the frame on the left hand side.
690
-------
Figure 2 A picture of the inside back of the FT-IR mobile laboratory
displaying the instrumental computer and spectrometer controlling
electronics. The spectrometer can be seen just outside the cargo
door upon the tripod and movement stage.
691
-------
Figure 3 The University of Kansas plume generator, meterological station
and smoke tray. The smoke was added to ascertain whether
particulate matter will affect the PT-IR performance.
VCO. CM-'
Figure 4 An atmospheric mid infrared absorbance spectrum of a five
component VDC mixture generated at the University of Kansas. The
spectrum is of mixture 1 at_100 meters. The compounds are 1,1,1-
trichloroethane (726 cm. , 1090 cm ), toluene (729 cm ) ,
roethylene chloride (749 cm, 1270 cm ) , t-butanol (918 cm ) ,
and methylethyl ketone (1174 cm" ).
692
-------
Analysis of Volatile Organic Compounds
With An Ion Trap Mass Spectrometer
David W. Berberich
Jay M. Wendling
Robert G. Orth
Monsanto Company
Environmental Sciences Center
St. Louis, MO 63167
The analysis of ambient air for trace levels of organic
compounds in the field requires instrumentation which is
sensitive and versatile but is not complicated in operation.
This, of course, is not fulfilled by any one device or method of
analysis. The ion trap mass spectrometer (ITMS) appears to at
least address some of the desirable points mentioned. Therefore,
an evaluation of the ITMS for detection limits using a direct
inlet system was undertaken for 10 organic compounds in air. The
instrument was able to detect the organic compounds in a high
background of phosphoric acid. The detection limits for the
compounds examined all appear to be in the mid ppb to low ppb
regime. The detection limits for the aromatic compounds are
conservative since these figures are based on 250 ppb standard.
The overall detection limits are also conservative since the
interface may still be improved. The ability to detect specific
components in a complex mixture was evaluated using two air
samples. The ITMS was able to determine a variety of components
in these samples. The ITMS was able to confirm the
identification of each of the components by mass
spectrometry/mass spectrometry methods which yield a collision
induced fragmentation spectrum characteristic of each component.
Although this device is not commercially available for ambient
air analysis these results suggest the instrument could fulfill
many of the initial needs.
693
-------
INTRODUCTION
The accurate measurement and determination of volatile
organic compounds in the field requires a device which is
sensitive, versatile, and easily maintained. Additionally, the
device should be able to provide real time monitoring of the air
being analyzed. The use of such a device would allow for an
analysis of ambient air with a fast response time (minutes) . One
instrument which meets many of the attributes is the TAGA 6000
mobile MS/MS laboratory (1).
An instrument that is not commercially available as an
ambient air analysis device but does have many of the above
mentioned attributes is based upon ion trap technology (2-3).
A number of research groups are currently evaluating similar
instrumentation for volatile organic compound analysis in ambient
air (4-7). The ion trap may be operated in a variety of modes
including mass-selective storage (8-9). This mode of operation
can be thought of as compound-selective storage, since it allows
the device to be used to trap ions from a specific compound (i.e.
ions having one m/z ratio) from a complex mixture of compounds.
This connotes that the ion trap may be used to obtain very low
detection limits in complex mixtures. Another advantage of this
device is that it can analyze the selected compound (ion) further
by use of fragmentation (MS/MS) techniques. The fragmentation
techniques involves colliding the compound (ion) of interest with
neutral molecules (collisionally induced dissociation [CID])
resulting in the formation of collision induced ion fragments
which are characteristic of the compound and can be used to
identify this compound in a mixture (10). The second MS/MS
method technique that could be used is selective reactions, which
are analogous to acid-base chemistry in the liquid phase, except
the reactions occur in the gas phase inside the device (9).
R. G. Orth and coworkers presented a paper at the 1939
EPA/APCA symposium with preliminary data obtained using the ion
trap mass spectrometer (ITMS) for ambient air analysis (11).
This evaluation is an extension of the initial work utilizing
different operational modes and interface. The evaluation
included the determination of the detection limits of the ITMS
for several compounds in calibrated mixtures sampled directly.
A second evaluation of air samples was performed with the ITMS.
EXPERIMENTAL
Figure 1 shows a schematic cross section of the ITMS and the
sample introduction system as utilized in this evaluation (3).
The sample introduction system consisted of a fine metering valve
and a heated transfer line. An uncoated megabore 0.83 urn (~3
meter) gas chromatographic column is used as a transfer line.
The samples were analyzed using two ionization modes electron
impact (El) and chemical ionization (CI).
The samples were introduced through a fine metering valve,
VI, to control the flow rate into the device. The static
pressure of air in the device was generally 1x10 tcrr while the
total pressure in the ITMS was in the 10 ~" torr range. The
difference in total pressure was due to the addition of helium
(added co-axially to the sample inlet stream).
€94
-------
The samples analyzed for this analysis include NBS standard:
of a mixture containing benzene, toluene, chlorobenzene, anc
bromobenzene at concentrations of 10 ppm and 250 ppb pej
component in a bath gas of nitrogen. Certified standards froi
Scott Specialty Gas of a mixture containing acetonitrile, 1,3 -
butadiene, chloroform, carbon tetrachloride, and dichloromethane
The Scott standards were at concentrations of approximately :
ppb, 10 ppb, 100 ppb, 1 ppm, and 10 ppm in a bath gas 01
nitrogen. Two air samples were collected in one liter summc
vessels and were analyzed.
RESULTS AND DISCUSSION
The air standard containing carbon tetrachloride was
admitted directly into the ITMS via the sample introductior
scheme in Figure 1. The majority of the sample is N,. Since the
ion trap can be thought of as a box filled with ions from the
sample, the predominate ions contained in the box would be those
arising from N2. This makes it difficult to detect the ppm tc
ppb level constituents of interest, since the static pressure of
the compounds of interest is on the order of 3x10 torr. Tc
accomplish the detection at these trace levels, the device car
remove the ions that arise from N2 and any other ions which are
not of interest (compound-selective storage).
Mass-selective storage provides a means for the minimizatior
of interferences which originate at the trace level such as
hydrocarbons. In the identification of carbon tetrachloride
(CC14) for the standards, interferant ions present in the ITMS
make the identification difficult. Figure 2a shows the result of
combining the use of the ion trap in the mass-selective storage
mode, i.e. selecting only m/z 117 to be stored in the trap and
ejecting all other ions. The identification of m/z 117 as a
characteristic ion from carbon tetrachloride can be accomplished
by the use of MS/MS methods (10) (Figure 2b).
The detection limit was based on the signal (provided by the
analyte) to the noise (present in the background) as established
by continuously monitoring m/z 82 (the fragment ion of m/z 117)
at 1 ppb, 10 ppb, 100 ppb and 10 ppm levels. The detection limit
was determined to be -500 ppb. This is a conservative detection
limit since the interface for sampling was not maximized (i.e.
the fine metering valve was unheated). The detection limits for
both EI/MS and EI/MS/MS mode of operation are summarized in Table
1. The detection limits for the aromatic compounds are very
conservative and are based upon data obtained from the 250 ppb
and 10 ppm standards. The detection limits for the aliphatic
compounds are based upon data obtained from standards ranging in
concentration from 1 ppb to 10 ppm. The detection limits for all
the aromatic compounds examined should be able to be improved to
the low ppb to sub ppb regime. Three compounds in the mixture
were unable to be quantified using the current experimental
arrangement. Future work will include heating the metering valve
and optimization of the data system which should allow for the
quantifiable detection of all the components.
The second evaluation involved the analysis of two air
samples collected using one liter Summa vessels. Sample one
contained chloroform, since this was; also used as a standard for
695
-------
the earlier evaluation the concentration of chloroform was
calculated to be -100 ppb. Sample two was directly analyzed and
the resulting electron ionization mass spectrum is very complex.
The specificity of MS/MS is exemplified below using air sample
two. This was done using collisional induced dissociation (CID)
for each m/z ion which would represent a component in the summa
vessel.
In the identification of benzene (CgHg.. from air sample two,
interferent compounds present make the mass spectrometric
identification very difficult. The ions at m/z 77 can arise from
a number of sources, but the ion at m/z 79 can arise from fewer
sources such as protonated benzene or from hydrocarbons. Figure
3a shows the mass spectrum that results when the m/z ions 77-80
have been selectively retained (compound-selective storage).
This is followed by kinetically exciting the m/z 79 ions and
colliding them with He in the ion trap to yield the
characteristic fragment ions of protonated benzene shown in
Figure 3b. The fragments are typical for protonated benzene
based upon knowledge of mass spectrometric fragmentation. Three
components in sample two were identified by using EI/MS/MS
analysis. This sample also contained ions associated with
hydrocarbons which were not identified.
CONCLUSIONS
The ITMS from this evaluation has distinct promise as an
analytical tool for the determination of trace organics in
ambient air. The simple interface which samples air directly was
shown to have detection limits in a complex background at the ppb
to sub-ppb (carbon tetrachloride) levels for selected compounds.
This limit is once again conservative and probably can be
extended with improvements in the interface and data system
utilized for this evaluation. The ITMS was able to identify and
confirm the components of standard mixtures to the ppb region
using EI/MS/MS. The ITMS was also able to identify components in
two air samples by use of EI/MS/MS analysis. The small size of
devices based upon ion trap technology (ITD and ITS40) indicate
that the device may be miniaturized to a size that would be
easily transportable (approximately the size of a gas
chromatography system) . Although near real time analysis v/as
realized with this study, it should be possible with a properly
designed interface and data system to obtain real time analysis.
REFERENCES
1. B. I. Shushan, G. Debrou. Proceedings of the 1987 EPA/APCA
Symposium On Measurements of Toxic and Related Air
Pollutants, page 218.
2. E. Fischer, Zeitschift Fur Physik, 1959, 156, pp. 1-26.
3. G. C. Stafford, Jr., P. E. Kelley, J. E. P. Syka, W. E.
Reynolds, J. F. J. Todd, Int. Journ. Mass Spectrom. & Ion
Processes, 60, 1984, pp, 85-98.
4. S. A. McLuckey, G. L. Glish, K. G. Asano, Analytica Chemica
Acta. 1989, 225, pp. 25-25.
696
-------
5. M. B. Wise, M. V. Buchanan, R. H. Ilgner, Presented at the
35th Conference on Mass Spectroraetry and Allied Topics, May
21-26, Miami Beach, FL. , 1989, pp. 1435-1436.
6. N. S. Arnold, K. A. Roberts, W. H. McClennen, H. L. C.
Meuzelaar, Presented at the 3E>th Conference on Mass
Spectrometrv and Allied Topics, May 21-26, Miami Beach, FL. ,
1989, pp. 1425-1426.
7. C. P. Leibman, T. M. Cannon, M. A. Wolf, P. H. Hemberger,
Presented at the 35th Conference on Mass Spectrometrv and
Allied Topics. May 21-26, Miami Beach, FL., 1989, pp. 1429-
1430.
8. M. Weber-Grabau, P. E. Kelley, J. E. P. Syka, S. C.
Bradshaw, J. S. Brodbelt, Presented at the 35th Conference
on Mass Spectrometry and Allied Topics, May 24-29, Denver,
CO., 1987, pp. 1114-1115.
9. D. W. Berberich, M. E. Hail, J. V. Johnson, R. A. Yost, Int_._
Journ. Mass Spectrom. & Ion Processes, 94, 1989, pp. 115-
147.
10. F. W. McLafferty, Ed., Tandem Mass Spectrometrv, J. Wiley &
Sons, 1983.
11. R. G. Orth, D. Haile, F. D. Hilerian, M. Weber-Grabau, P. E.
Kelley, Presented at the EPA/APCA International Symposium:
Measurement of Toxic and Related Air Pollutants, 1989, pp.
291-298.
Table 1. EI/MS and EI/MS/MS Limits of Detection for Volatile
Organic Compounds
Compound (m/z) EI/MS LOD's EI/MS/MS LOD's
Acetonitrile (42) NQa NQa
1,3 - Butadiene (54) NQa Nof
Benzene (77) 60 ppb 60 ppb
Benzene (78) 60 ppb 60 ppb
Dichloromethane (83) 5 ppb 5 ppb
Chloroform (84) 5 ppb 5 ppb
Toluene (91) 60 ppb 60 ppb
Chlorobenzene (112) 125 ppb 125 ppb
Carbon tetrachloride (117) 250 ppt 500 ppt
Bromobenzene (157) NQcl N2a
a NQ - Not quantifiable with current experimental arrangement
697
-------
EXPERIMENTAL ARRANGEMENT
1TMS
Heated Transfer Line
Sampla
Introduction ^f—
Helium Buffer Gas
Figure l.
Schematic of experimental
arrangement for air sampling with
the Ion Trap Mass Spectrometer.
50
45
117
65
. , ', I'T'T I I !"
35
m/z
32
105
125
100-
;
50 -j
-
:
-
-
n —
4
b)
!
5 65
1
,
85 105
m/z
17
[
125
100-
50-
50
50
79
60 70 SO
m/z
77
uu -
-
50 :
b)
d
JL ,
79
60 70 30
m/z
Figure 2.
a) Detection of carbon tetra-
chloride in the 10 ppb standard
using selected mass storage.
b) The collisionally induced
dissociation of m/z 117 which
yields fragments characteristic
of carbon tetrachloride.
Figure 3.
a) The use of the ion trap to
isolate m/z 77 to 80 where 79
is believed to be due to the
protonated benzene molecule.
b) The collision induced
dissociation of m/z 79 which
yields fragments that
represent protonated benzene.
698
-------
Evaluation of a Non Cryogenic Automated
Multitube Thermal Desorption System
for the Analysis of Air Toxics
William F. Boehler
Ronald L Huttie
Kenneth M. Hill
Suffolk County Department of Health Services
Center For Forensic Sciences, Hauppauge, NY.
Introduction
Improved methods^ in air monitoring technology which employ multi-layer sorbent tubes,
followed by Thermal Desorption/Gas Chromatography/Mass Spectrophotometry (GC/MS) have been
shown to be a more efficient method for the collection and analysis of volatile organic compounds
(VOC's), than existing preconcentration techniques. Despite its distinct advantages, the Thermal
Desorption/GC/MS analysis of multi-layer sorbent tubes, suffers from the disadvantage that it is not
amenable to total automation.
In an effort to automate the above mentioned "Improved Method," the Suffolk County
Department of Health, working with Envirochem Inc. (Kembleville, PA) over a period of six months
(12/88 - 5/89) evaluated and helped develop the Model #8916 Mulitiple Tube Desorber Unit (MTD).
This paper will discuss and evaluate an entire integrated air analysis system. The analytical
system is composed of Envirochem's Model #8916 MTD, Envirochem's 81OA Concentrator, Hewlett
Packard Model 5890 GC and Model 5970 MSD.
Incorporation of the MTD in the abovementioned system, is shown to provide the following:
(1) Automation Features:
(a) Increased productivity, unattended operation for the thermal desorption of 16 multi-
layer sorbent tubes without the need for cryogenic focusing.
(b) Ability to back load sorbent tubes, therefore permitting extended use of GC/MS
equipment.
(2) Improved accuracy and recovery, particularly for more volatile compounds.
(3) Improvement in overall Quality Assurance due to ability to run more standards and blanks.
(4) Ease of Operation.
Principle of Operation
The MTD (see Figure 1.) employs an electronically controlled, air actuated 16 position multiport
valve, to direct carrier gas flow through a preselected sorbent tube which is about to undergo a thermal
699
-------
desorption cycle.
The start of the thermal desorption cycle is normally controlled by a remote start signal,
received from the GC/MS; however an automode (w/o remote start), and a manual advance button can
also be used to initiate a cycle, and/or advance positions,
A Typical MTD Cycle:
(1) At the start of the thermal desorption cycle the MTD allows a 1 min purge of carrier gas, to
free the sorbent tube and transfer lines of oxygen. This dry purge also acts as diluent and
carrier for moisture in the analytical system.
(2) Thermal desorption of sorbent tubes occurs with the aid of a sleeve heater at a flow of
60mL/min helium at a preselected temperature, for a preselected time.
(3) Through the use of a heated transfer line the VOC's are then transferred to the large bore
trap of the concentrator, (see Figure 1.)
(4) After allowing additional time and flow for proper transfer of organics, the large bore trap is
rapidly heated and the VOC's are transferred onto a micro bore trap.
(5) Similarly, the micro bore trap is ballistically heated and a discrete slug of concentrated
organics is transferred to the capillary direct interface of the GC/MS analytical system.
(6) The MTD is instructed to start a new cycle via contact closures on the GC/MS, once the
analytical run is complete and equilibration temperatures are reached.
Experimental
Materials
Envirochem ST-032 multi-layer sorbent tubes were utilized as the collection media. The
air monitoring tubes were constructed of Pyrex glass (20cm long x 6mm OD and 4mm ID) and
contained a glass frit, sequential layers of glass beads, Tenax, Ambersorb XE-340, and charcoal
adsorbants.
Analytical standards were obtained from Aldrich Chemicals (Milwaukee, Wl), Chem
Services (West Chester, PA), and Scott Specialty Gases (Plumsteadville, PA).
A Supelco (Bellefonte, PA) model 2-3800 high capacity purifier, with an OM-1 indicating
purifier was used to scrub ultra high purity helium of undesirable oxygen, moisture, CO, and
CO2, prior to entry into the analytical system.
Standard Preparation/Storage
Prior to use, air monitoring sorbent tubes are conditioned for 20 min at 310°C, while
being purged at 60ml_/mtn with ultra high purity helium. Upon completion of the heating cycle,
with the aid of continued purge flow, the sorbent tubes are allowed to cool to room
temperature, and are then placed in individually labeled glass/teflon screw capped storage
containers.
After injection with an internal standard (4-bromofluorobenzene) the conditioned tubes
are provided additional protection by storing dated batches of tubes in "Chain Of Custody" heat
sealed 0.006 mil polyethylene bags.
After completion of sampling, the glass/teflon protected tubes are once again batch
heat sealed in the polyethylene bags, and refrigerated @ 4°C to await analysis.
Standards were prepared using the EPA static dilution bottle method^. Additional
check standards were also employed, utilizing a Scott Specialty Gases certified blend of target
compounds, contained in a single compressed gas cylinder.
All prepared gas standards were injected into conditioned sorbent tubes via septum in
a pre-purged tee, which was housed in a 60°C oven. A purge flow of 60ml_/min was continued
for 5 min to assure complete transfer of VOC's onto the sorbent air collection tubes.
Analytical Parameters
M.T.D. - Temperature settings used were; valve compartment - 290°C, fitting
compartment = 120°C, with all transfer lines set to 240°C. A complete 8 min cycle includes a 1
min prepurge (no heat) followed by 1 min heating cycle to 200°C, and finally a holding
temperature of 200°C for the remaining 6 min.
Concentrator - After a 3 min secondary carrier flow time, the large bore trap was heated
to 265°C in 22 seconds. A 3 min trap to trap transfer time was then used to translocate VOC's
700
-------
to the micro bore trap. After the micro bore trap was; heated to 275°C in 22 seconds, the total
trap to column transfer and GC/MS run time was 52 min.
GC - A Restek (Port Matilda, PA) 60 meter, RTx Volatiles (crossbonded phenytmethyl
polysiloxone), 0.25mm ID, custom made 1.5M film thickness, capillary column was used. The
GC oven initial temperature was set to 0°C wKh a 2 min hold, then programmed at a rate of
2°C/min for 20 min, 4°C/min for 5 min, rC/min for 13 min, and 10°C/min for 7.9 min with a final
hold time of 4.1 min (at the resultant 230°C temperature). The total run time was 52 min.
MS - Autotuned every morning with periluorotributylamine. Ran at autotuned settings.
During analysis the MS was operated in full scan mode from 19 amu to 260 amu.
Calibration is performed daily. System (MTD/CONC/GC/MS) analysis of a sorbent
tube calibration standard which contained a 34 compound toxic VOC mix is illustrated in Figure
2. and listed in Table IV.
Results & Discussion
(Evaluation of Autosampler)
Thirty-five volatile organic compounds (34 toxic VOG's and 1 internal standard) were utilized to
evaluate the MTD in conjunction with the entire concentrator coupled/GC/MS analytical system .
Advantages and shortcomings are listed below:
(1) Elimination of Crvofocusing - Moisture based chromatographic problems have been
reported^ in systems which employ a cryogenic based trapping system. Semi-permeable
membrane tubing (i.e., Nation dryers) can be used on such systems to eliminate water,
however such dryers also remove polar compounds. Therefore, one may be sacrificing the
integrity of the original sample. Since the MTD does not employ the cryogenic principle,
analytical difficulties associated with cryofocusing have been eliminated.
(2) Increased Productivity - The labor intensive procedure of attended one-at-a-time desorption
has been eliminated by allowing unattended sequential thermal desorption of 16 multi-layer
sorbent tubes. Typically, before leaving for the evening, one "back loads" the MTD with
additional air monitoring tubes; therefore, depending on analytical cycle time, 24 hour use
of a GC/MS is now possible.
During unattended operation, the unit can ba set into a dual remote start "safety mode",
where, in the event of power failure occurrence, the maximum number of samples that can
be lost is one (1). Once placed into the safety mode the MTD will not advance until the GC
is "Ready" to accept samples.
(3) Improved Precision and Accuracy - Individual "System Precision Checks" are listed for each
of the target analytes in Table III. The mean imprecision for all 35 compounds was 4.9%
R.S.D. (Relative Standard Deviation) which is lower than previously reported4'^ acceptable
values obtained without the use of an MTD.
Since the EPA "Static Dilution Bottle" method used to prepare our standards, lists an
imprecision of ±10%, we find the analytical system to be extremely precise.
Improved accuracy, resulting from increased recovery of volatile organics (with a
boiling point of less than 50°C) has been observed, when utilizing the MTD. Low or erratic
recoveries often occur when employing a stand alone Model 81OA Concentrator. When a
sorbent tube is placed into the thermal desorption chamber of the Concentrator a purge
flow goes over the tube to the vent position. Since the desorption chamber is at a
temperature of 40-45°C this causes losses of highly volatile organics, especially if the
sample is not fired immediately. This condition cannot occur when utilizing the MTD,
because each sorbent tube is dead ended with no purge flow until the desorption cycle
starts.
Results from our most recent EPA audit for Group V compounds are listed in Table V.
With the exception of carbon tetrachloride quanritation accuracy at the 5ppb level was
satisfactory (less than ±30% tolerance). Our poor results for carbon tetrachloride are
currently under investigation.
(4) Improved Quality Assurance - Ability to run more standards and blanks results in an overall
improvement in quality assurance. For example, in a given 24 hour period, we have the
ability to run a 5 point calibration curve, periodic check standards, field blanks, method
blanks, and a post run standard, all appropriately sequenced into a normal sample run.
701
-------
Additional System Checks
(5) Linearity - The analytical system was found to be linear in the 0-50nl_ range for each of the
target analytes. Results obtained for a 5 point calibration curve in a 32 compound mix,
resulted in individual compound correlation coefficients ranging from 0.969 to 1.000, with an
average of 0.993. Figure 2. shows a typical calibration curve.
(6) Storage - Acceptable storage of VOC's ranged from less than 1 week for few of the
extremely volatile compounds, to greater than 5 months for less volatile compounds.
Please refer to Table I for compound specific storage information.
(7) Break Through Studies - Results from a "100 Liter - Dry Purge" break through study
indicated satisfactory recoveries for 31 of the 34 target compounds, inconclusive results
were obtained for Freon 12, vinyl chloride, and bromomethane.
A similar break through study was conducted using a "100 Liters of 100% Humidified
purge" flow. The humidified study indicated a 40% breakthrough for Freon 11, and a 50%
breakthrough for methylene chloride, inconclusive results were obtained for Freon 12,
Freon 114, vinyl chloride and bromomethane. Recoveries for the remaining 28 compounds
were satisfactory. Compound specific information is presented in Table II.
CONCLUSION
Use of the Multiple Tube Desorber (MTD) in the prescribed GC/MS system provides for
automated analysis of air toxics without the need of cryofocusing. Improvements in productivity,
precision, accuracy and overall quality assurance have been demonstrated.
Acknowledgements
We gratefully acknowledge Sigmund M. Menchel, M.D., and Leo A. Dal Cortivo, Ph.D. for their
project support and technical assistance.
We would also like to thank Paul Ames, Barry Passin and Ida Punturieri for the computer
assisted production of the manuscript and visual aids.
References
1. Betz, William R., "Monitoring A Wide Range Of Airborne Organic Contaminants", in
Proceedings EPA/APCA Symposium on Measurement Of Toxic And Related Air Pollutants,
APCA, Pittsburgh, PA, pp. 761-770, (1987).
2. Berkley R. E., Swanson D. H., Bumgarner J. E., "Standard Operating Procedure For The
Preparation And Use Of Standard Organic Mixtures In A Static Dilution Bottle", U.S.
Environmental Protection Agency, EMSL, R.T.P., NC 27711, June, (1983).
3. U.S. EPA, "Method TO-3", Compendium Of Methods For the Determination Of Toxic Organic
Compounds In Air. Sec 6.2 (Interferences), EPA Document No. 600/4-84-041, (1984).
4. Chan, Cecilia C., et al., "Determination Of Organic Contaminants In Residential Indoor Air
Using An Adsorption - Thermal Desorption Technique", Journal Of The Air And Waste
Management Association, Vol. 40. pp. 62, January, (1990).
5. C.C. Chan, et al., (Development Of An Adsorption/Thermal Desorption Technique Coupled
With GC/MS For The Monitoring Of Trace Organic Contaminants In Indoor Air", In
Proceedings EPA/APCA Symposium On Measurements Of Toxic Air Pollutants, Air and
Waste Management Assoc., Pittsburgh, PA, (1986).
702
-------
TABLE I.
STORAGE STUDY
I IT
I STORAGE TIME IN DAYS II
i
COMPOUND
COMPOUND
ACCEPTABLE
(b) I!
INVALID II
STORAGE TIME IN DAYS
ACCEPTABLE
(b)
INVALID
o
w
Freon 12
Freon 114
Vinyl Chloride
1,3-Butadiene
Bromomethane
Freon 11
Freon 113
1,1-Dichloroethylene
Methylene Chloride
1,1-Dichloroethane
Chloroform
1,1, 1-Trichloroethane
Carbon Tetrachloride
1,2-Dichloroethane
Benzene
Trichloroethylene
1,2-Dichloropropane
Toluene
1
4
1
4
1
7
7
7
7
7
7
150
150
150
150
60
150
150
Tetrachloroethylene
1,2-Dibromoethane
Chlorobenzene
1,1,1,2-Tetrachloroethane
Ethylbenzene
M-Xylene
O-Xylene
1,1,2,2-Tetrachloroethane
Int.Std./Bromofluorobenzene
1,2,3-Trichloropropane
1,3,5-Trimethylbenzene
1,2,4-Trimethylbenzene
M-Dichlorobenzene
P-Dichlorobenzene
P-Diethylbenzene
O-Dichlorobenzene
Naphthalene
150
150
150
60
150
150
150
60
60
60
60
60
60
60
60
60
60
_L
Note: (b) Invalid -less than 70% recovery.
Dashes indicate intervals not yet
established.
-------
TABLE II.
BREAK THROUGH STUDY
COMPOUND
100 Liter
Dry Purge
f% Recovery)
100 Liter/100%
Humidity Purge
f% Recovery)
Freon 12
Freon 114
Vinyl Chloride
1, 3 -Butadiene
Bromomethane
Freon 11
Freon 113
1, l-Dichloroethylene
Methylene Chloride
1, 1-Dichloroethane
Chloroform
1,1, 1-Tr ichloroethane
Carbon Tetrachloride
1, 2-Dichloroethane
Benzene
Trichloroethylene
1 , 2-Dichloropropane
Toluene
Tetrachloroethylene
1, 2-Dibromoethane
Chlorobenzene
1,1,1, 2-Tetrachloroethane
Ethylbenzene
M-Xylene
0-Xylene
1,1,2, 2-Tetrachloroethane
Int . Std . /Bromof luorobenzene
1,2, 3-Trichloropropane
1,3, 5-Tr imethylbenzene
1,2, 4-Trimethylbenzene
M-Dichlorobenzene
P-Diethylbenzene
0-Dichlorobenzene
Naphthalene
Front
Tube
I
90%
I
100%
I
100%
100%
100%
96%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
95%
99%
100%
100%
100%
100%
100%
100%
100%
100%
99%
Back-Up |
Tube
I
10%
I
0%
I
0%
0%
0%
4%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
5%
1%
0%
0%
0%
0%
0%
0%
0%
0%
1%
Front
Tube
I
I
I
100%
I
60%
80%
83%
50%
79%
86%
94%
97%
98%
100%
100%
100%
97%
100%
100%
100%
100%
100%
95%
99%
100%
100%
100%
100%
99%
100%
100%
100%
99%
Back-Up
Tube
I
I
I
0%
I
40%
20%
17%
50%
21%
14%
6%
3%
2%
0%
0%
0%
3%
0%
0%
0%
0%
0%
5%
1%
0%
0%
0%
0%
1%
0%
0%
0%
1%
Note:(I) Inconclusive results. Zero (0) or trace amounts
only were found on the back-up tubes; however, less than
expected amounts were found on the corresponding front
end tube. Since sorbent tubes were prepared 7 days prior
to analysis, reported storage problems may have caused or
contributed to the low front tube recoveries. Further
study is indicated.
704
-------
TABLE III.
SYSTEM PRECISION CHECKS
o
tn
COMPOUND
Freon 12
Freon 114
Vinyl Chloride
1, 3 -Butadiene
Bromomethane
Freon 11
Freon 113
1, 1-Dichloroethylene
Methylene Chloride
1, 1-Dichloroethane
Chloroform
1,1, 1-Trichloroethane
Carbon Tetrachloride
1 , 2-Dichloroethane
Benzene
Trichloroethylene
1 , 2-Dichloropropane
Toluene
AVG IN
nL
(n=3)
24.14
24.28
22.72
31.00
22.05
16.41
4.65
6.54
9.11
6.36
6.97
5.48
5.57
7.21
7.76
5.89
5.77
5.34
REL. STD. ||
DEV. ||
(%) II
II
4.0 ||
3.4 ||
9.7 ||
13.1 ||
12.9 ||
11.0 I
3.7 ||
12.8 ||
4.2 |i
4.2 ||
2.6 ||
4.9 ||
5.0 ||
3.9 ||
11.9 ||
5.3 ||
3-5 I
2.8 ||
II
COMPOUND
Tetrachloroethylene
1 , 2-Dibromoethane
Chlorobenzene
1,1,1, 2-Tetrachloroethane
Ethylbenzene
M-Xylene
O-Xylene
1,1,2,2 -Tetrachloroethane
Int . Std . /Bromof luorobenzene
1,2 , 3-Trichloropropane
1,3, 5-Trimethylbenzene
1 , 2 , 4-Trimethylbenzene
M-Dichlorobenzene
P-Dichlorobenzene
P-Diethylbenzene
O-Dichlorobenzene
Naphthalene
AVG IN
nL
(n=3)
5.40
6.62
5.85
5.28
4.62
4.73
4.81
5.98
5.26
5.60
4.33
4.24
4.77
6.36
3.67
5.43
20.15
i
REL. STD.
DEV.
(%)
3.5
2.4
0.9
3.0
3.0
2.1
1.9
3.0
1.9
1.6
1.6
1.7
8.0
3.9
5.2
3. 3
6.0
Mean for 35 compounds
4.9% R.S.D,
-------
Fittinas
Compartment
120°C
S.Tubes &
Sleeve Heaters
Valve Compartment
290 °C
o
OJ
Multitude Desorber
Concentrator
QC / MS
Transfer Lines a 240°C
Selectable 16-Position Valve
Single Flow Path Only
(Remaining 15 traps are Dead-ended)
Capillary Direct
Interface
Figure 1. ANALYTICAL SYSTEM
-------
3.5E+6-
3.0E+6:
2.5E+6:
u
o
5 2.0E+G:
c
_S 1.5E+6-
cr
1.0E+6-
5.0E+5-
0nr~ i n -
. 0L+0
I
1
1
111
T i m«
)
• (i
n i n
30
4 J
4£
3
:JL
50
Figure 2. Chromatogram
(Calibration Sid. S.Tube)
r" •
DETENT ION
: TIME :
s'.w
6.90
8.33
8.63
11.18
13.91
17-25
17.59
20.42
24.57
28.23
: 29.84
30.80
31.12
31.30
33.24
33.65
V •:•
Table
COMPOUND
Frebn 12
Freon 114
Vinyl Chloride
1,3 -Butadiene
Bromo«n«thane
?reon 11
P]:epi> 113
1, 1-DichloToechyLene
Methylen* Chloride
1 , 1-Dlchloroethane
Chloroform
1, 1 , 1-TrichlOroethane
Carbon Tetrachlorlde
1 , 2-Dlchloroethane
Benzene
Trlchloro«thylene
1,2-Dlohloropropan*
IV. COMPOUND IDENTIFICATION ^
CONC.
ni
50.72
50.72
50.72
50 i 72
50.98
32.70
9.70
15 .38
19.08
13.44
14.42
11.62
11.70
15.06
12.92
12.66
12.10
RETEH^'IOH
TIM:
36.75
38.311
39. HI
40.24
: 40.30
40.311
: 40,5V
: :"•':• 41.5J.
42.56
42,8;.
42.8(i
43. k'.t
44.30
45.1(1
45,7;!
46.o:>
49.7ti
COHKKJKD
Toluene
Tetrachloroethylene
1 ,2-DLbronoethan*
Chlorobenzene
1,1,1, 2-T« t rachloroethane
Ethylbenzene
K-Xyl.n«
O-Xylene
1 , 1 , 2 , 2-Tetr»ehloro«thane
Int . Std. /Broa»fluorobengene
1,2, 3-Tr lehl oropropane
1 , 3 , 5-Tr la«thyl!>«ftzene
1,2, *-Trim«t-hyl benzene
M-Dlchlorobensene
P-Dl*thylbenzen.
O-D Ichlorobeneene
Naphthalene
CONC.
nL
11.08
11,32
13.68
11,82
10.98
9,60
9.72
9.78
11.44
10.80
11.44
8.76
8.54
10. OB
7.40
10.76
11.22
J
707
-------
v>
c.
o
3.0E+7-
2.5E+7-1
2.0E+7-
£ 1.5E+71
QL
1 . 0E+7-
5.0E+6-
0.0E+0
111 Trlclethan
Response - 687943(Rmt) + 628563
r^2 - 8.992
0
10
20
Rmt
40
50
Figure 3. 5 Point Calibration Curve
(Linearity Check)
/" TABLE V, ACCURACY CHECK / EPA AUDIT ^v
(Group V Cylinder - Unknown)
Vinyl Chloride
Chloroform
Carbon fetr»chlorld«
Methylen* Chi or Id*
1 , 2-J}iohlQroeth*n«
Trichloro«thyl«ne
Benzene
Iecr»chLoro«thyl«tie
l,3-But*dien«
BrooiocD«th»n«
Freon 11
1,1, l-Tclchi6to*thaii«
\ , 2-DJ.ohlcropropane
1, 2-Dtbroow«th«ne
Toluene
Chlorobenzeni
Ethylbeniene
Ortho-Xylan*
Reported Concentr»tton P.P.B.(V)
Suffolk County
L»bor*tory Rcaults
S.»
5,*
': *-s ..'.....
5,0 :
5.9
5.6
4.8
6.3
6,3
6.7
StPX«ge Ptobleai
5.7 :
5.7
6.2
«,0
6.5
5.*
6.1
B..T.1.
(Reference)
Laboratory Reaulta .
5.04+/-0.3
4.94+/-0.3
4t6«+/-O.J
4.28+/-0.5
5.26+/-0.4
5,97+7-0.3
4.50+/-O.S
5.34+/-0.3
HA .
5.59+/-O.S
5.00+/-0,6
s,ao+/-o.3
5.3J+/-0.3
5.10+/-0.5
S-17+/-O.J
5.33-I-/-0.4
4.S7-f/-0.4
4.70+/-0.4
1 ' ' " . A
708
-------
Stability/Instability of Gas Mixtures Containing
1,3-Butadiene in Treated Aluminum Gas Cylinders
George C. Rhoderick
Gas and Particulate Science Division
Center for Analytical Chemistry
National Institute of Standards and Technology
Gaithersburg, Maryland 20899
The Gas Metrology Research Group of the National Institute of
Standards and Technology (NIST), formerly NBS, has been
involved in research and developnent of gas standards of
volatile toxic organic compounds for many years. Over thirty
toxic organic compounds have been studied in gas mixtures
contained in high pressure aluminum gas cylinders with
specially treated interior surfaces. These mixtures, prepared
using a microgravimetric techniques developed at NIST, have
been studied and tracked for many years to determine long term
compositional stability. Almost all compounds studied to date
have shown very good long term stability at the parts-per-
billion (ppb, nanomole/mole) to parts-per-million (ppm,
micromole/mole) range. One exception, 1,3-butadiene, is a
compound that many scientists and policy makers are very
interested in measuring in the environment. In this paper the
I discuss data obtained over several years for gas mixtures
of 1,3-butadiene in nitrogen, at concentrations of 2 ppm, 100
ppb, and 10 ppb. The data demonstrcite that mixtures of 1,3-
butadiene at the 2 ppm level have remained stable for over
three years. However, gas mixtures of 1,3-butadiene at the
10 ppb level have shown decreases in concentration of more
than 70% over a two year period. Decreases in the
concentration of 1,3-butadiene have been observed immediately
after the preparation of the gas mixture in several cylinders.
This work indicates that gas mixtures of 1,3-butadiene are not
stable at the ppb levels and therefore are not reliable as
accurate calibration standards. This suggests that fresh
standards are needed to calibrate an instrument at the time
of analysis and also that any "grab" samples containing this
compound should be analyzed shortly after being taken.
709
-------
Introduction
Concern has risen in recent years over the occupational
health hazard from exposure to 1,3-butadiene, in particular
its carcinogenic potential<1>. EPA statistics show that there
is an increased risk of cancer around several chemical
facilities in the U.S., and that 1,3-butadiene is one of the
compounds of concern'25. A recent set of EPA data revealed
that cancer risks to persons around some industrial facilities
in the U.S. could be as high as one in ten. Political
campaigners are using this data, as well as other, to press
for stronger emission control provisions in the Clean Air
Act<3>. Studies have been done that compare the permitted
chronic human exposure to the chronic dose rate that induces
tumors in a proportion of laboratory animals; the results
showed 1,3-butadiene to be one of the most hazardous
substances'4'. The current Threshold Limit Value (TLV) for
1,3-butadiene is 10 parts-per-million (ppm, micromole/mole)
in the USA(5) and also in the Netherlands'65. Concerns over
whether to decrease the TLV have been expressed, and as a
result, Norway significantly reduced the permitted levels in
the workplace atmosphere for a number of chemicals, including
1,3-butadiene, to 1 ppm<7).
To properly monitor 1, 3-butadiene levels in the workplace
environment as well as the ambient atmosphere, accurate and
stable calibration standards are needed. The Gas Metrology
Research Group at'the National Institute of Standards and
Technology (NIST) has been studying, with support from the
EPA, the feasibility of producing gas standards of 1,3-
butadiene and other compounds in treated aluminum gas
cylinders. As a result of this research, NIST developed gas
standards for EPA to use in its quality assurance programs for
ambient air monitoring of volatile toxic organics. Two of the
groups of standards for EPA, Groups 4 and 5, contain 1,3-
butadiene as well as other organic compounds.
Gravimetric standards have been prepared containing 1,3-
butadiene ranging in concentration from 2 ppm to 5 parts-per-
billion (ppb, nanomole/mole). These standards have been
analyzed over time, with the introduction of new standards
into the set at useful time intervals to determine the
stability of the 1,3-butadiene gas mixtures. The results of
these stability studies, and the feasibility of producing
accurate, reliable, and stable standards containing 1,3-
butadiene, are discussed.
Experimental
The aluminum gas cylinders used for the mixtures were
treated by Scott Specialty Gases of Plumsteadville,
Pennsylvania, using a proprietary process called "Aculife" to
deactivate the cylinder walls. Gas standards containing 1,3-
butadiene (and other organic compounds not common to every
standard) in nitrogen were prepared in the aluminum gas
cylinders using a micro-gravimetric procedure developed at
NIST(8<9'10). Several standards were prepared at each of three
concentration levels: nominally 2000, 100, and 10 ppb.
710
-------
The standards were intercompared using a gas
chromatograph (GC) equipped with a flame-ionization detector
(FID). Several different packed and capillary chromatographic
columns were used depending on the other compounds present in
the standards. A stainless steel gas sampling valve with a
carbon filled teflon rotor was used to collect and inject the
gas sample onto the column. The gas; sample valve was equipped
with a 1.0 mL stainless steel sample loop for analysis of the
2 ppm set of standards. A 10 mL loop was used for the 100 ppb
standards. For analysis of the 10 ppb standards, an automatic
cryogenic preconcentration system was used. The sample flow
of 50 mL/min into the trap was controlled by a mass flow
controller. The sample was typically collected in the trap
for 5 minutes. The sample trap wa« then electrically heated
and the sample was injected onto the GC column.
Since there is no absolute method to determine
concentrations of organics at these levels, three methods were
used to track the 1,3-butadiene concentration. Method 1
involved the correlation of the set of standards, at each
level, using linear regression (a plot of gravimetric
concentration versus GC area response). A concentration for
each standard was then calculated from the linear regression
fit. If the calculated concentration agreed with the
gravimetric value, or the calculated value from previous
analyses, then the compound was considered to be stable.
Method 2 involved the preparation of new standards at selected
time intervals and intercomparison of the old to the new
standards. A ratio was calculated using the GC area response
of the old to the new standard and a concentration was then
determined based on the gravimetric value of the new standard.
If the concentration determined for the old standard agreed
with previous values, then the compound was considered to be
stable. Any continuous decrease in the concentration outside
of the uncertainty limits of the gravimetric concentration of
the compound would signify instability. Method 3 involved the
presence of an internal stable gas;, such as benzene, along
with the 1,3-butadiene in the standard. The GC area response
for 1,3-butadiene was ratioed to the area response for the
known stable compound. The standard was analyzed periodically
by GC-FID and a ratio was determined. If the ratio changed
in one direction continously, then the suspected compound was
not stable in the gas mixture.
Results and Discussion
Table 1 shows the results for a set of five standards
containing 1,3-butadiene at the 2 ppm level. The gravimetric
concentration, followed by the uncertainty of that value
calculated from the sources of error (weighing, purity of
starting materials) in the preparation, is given for each
standard. The first four standards were prepared within four
months of each other. Method 1 was used to intercompare the
standards in 9/86 and in 5/89. The calculated concentration
was determined from the linear regression of the analytical
data. The results of that plot show excellent agreement
(r2=1.0000) for the set of standards and thus showed no
decrease in the concentration of 1,3-butadiene in the four
standards. However, this did not ensure that decreases were
711
-------
not taking place. It was possible, although unlikely, that
all the standards were changing at the same rate. To
determine whether this was occurring, a new 2 ppm standard was
prepared. Method l was used to intercompare two of the
original four standards (the other two were not available) and
the new standard. The linear regression, although only using
three points, showed excellent agreement (r2=0.9995) between
the new and old standards, as shown under the column heading
9/89 in Table I. The calculated concentrations for these two
original standards were identical to their gravimetric values
determined in 1986. This data confirmed that the 1,3-
butadiene in the gas mixtures at a nominal concentration of
2 ppm was indeed stable, and had been so for a three year
period.
Only a few measurements of 1,3-butadiene stability were
done at the 1OO ppb level. Table II lists those standards
that have been prepared and tracked for stability. Only
method 2 has been used to measure the concentration of the
1,3-butadiene at this level. Standard AAL-19246 was compared
to a 2 ppm standard, AAL-15885, in 4/87. Subsequent
measurements were made by intercomparing the older standards
to a new one. The two oldest standards, AAL-19246 and AAL-
19249, decreased rapidly in concentration over six months and
one month, respectively. Those two standards plus AAL-20701
were analyzed against a new standard, x-138356, shortly after
it was prepared in 1/90. A further decrease in concentration
was observed in each of the three standards, with AAL-20701
having decreased by an absolute amount of 19 ppb in one year.
This represented the largest absolute decrease in 1,3-
butadiene observed for any of the standards studied. The
uncertainty in the gravimetric concentrations listed in Table
II are a factor of ten smaller than the amount of decrease
observed. Therefore, the data for these standards show that
1,3-butadiene is not stable in gas mixtures at the 100 ppb
level, and one might predict similar results at the 10 ppb
level.
Much of the work involving 1,3-butadiene in gas standards
has been at the 10 ppb level. Many standards have been
prepared and studied for stability. The data for a few of
these standards are summarized in Table 3. The first four
listed are the first standards prepared containing 1,3-
butadiene at the 10 ppb level. Method 1, using linear
regression of a set of standards, was used on the data from
the intercomparison of these four standards in 10/86 and 4/87.
The results show that standard AAL-17535 had decreased in 1,3-
butadiene concentration from its preparation in 8/86 to 10/86
when it was first measured. The analysis in 4/87 indicated
further decrease in AAL-17535 and a border-line decrease in
AAL-15940 {the uncertainty bar overlaps the two values). Two
of the standards, AAL-13377 and IL-3437, appeared to be stable
for six months. Since the first four standards in Table 3
were intercompared using linear regression in 10/86 and 4/87,
if one were to assume that all four of the standards changed
a little, then the calculated concentrations may not show the
actual changes in concentration because they are compared on
a relative basis. To determine if indeed the first four
standards in Table III had changed, two new standards were
prepared in 1/88 and intercompared to the first four standards
712
-------
in 2/88. Method 2 was used, ratioing the GC area response for
the first four standards to the new standards. The 1,3-
butadiene concentration was then calculated for each of the
first four standards from the gravimetric concentration of the
new standards. The resulting concentrations determined from
these ratios show that indeed the older standards were
decreasing. These results also show that the rate of decrease
of the 1,3-butadiene is different from cylinder to cylinder,
and therefore a constant rate change cannot be determined and
applied to all cylinders.
Another set of standards are listed in Table IV. These
standards, containing 1,3-butadiene at 10 ppb plus other
organic compounds different from those discussed in Table III,
were prepared and monitored as a function of time. Since IL-
3437, from the original set, appeared to be stable for six
months (see Table III, 4/87 analysis) , it was intercompared
with the five new standards prepared during 5/87. A response
factor (GC area response divided by gravimetric concentration)
was calculated for each of the standards. The response
factors for standards IL-3437, AAL-19256 and AAL-19248 agreed
within 1.6% of each other. All the other standards had
response factors much less than these three. Therefore,
method 1 was used to calculate the concentrations of the other
standards listed in Table IV. The results show that the 1,3-
butadiene had decreased in AAL-19254, AA1-19255 and AAL-19253,
only one month after their preparation. This decrease
continued when some of the standards were analyzed in 6/87.
Two new standards were prepared and intercompared to this set
in 1/88, using method 2. All the old standards had markedly
decreased in concentration over the six months between
analyses. Further decreases were seen again when part of the
set was analyzed against a new standard in 6/88.
The results from these studies indicate that it is very
difficult to determine what the actual concentration of the
1,3-butadiene is in a gas standard or mixture, at any given
time, without preparing new standards. Even then, it is
difficult to ensure the actual concentration since the 1,3-
butadiene concentration might change between the time the
standard is prepared and the time that it is first analyzed.
To emphasize this point, and to follow the decrease of 1,3-
butadiene in a gas mixture from the very beginning, a standard
was prepared and regularly analyzed. This standard also
contained benzene and toluene as internal standards, as well
as some other compounds. Method 3 was used to follow the 1,3-
butadiene concentration with time. The GC area response for
a known stable compound, toluene, was ratioed to that of
benzene, another known stable compound, to make sure the
toluene was behaving as expected. The concentrations of 1,3-
butadiene, toluene, and benzene were all nominal 10 ppb. At
the same time, the GC area response for 1,3-butadiene was
ratioed to that of benzene. Table V shows data for the ratio
of the 1,3-butadiene GC area response to that of benzene and
for toluene to benzene. The ratios for toluene to benzene
did not change throughout the study which covered a period of
over two years. However, the ratios of 1,3-butadiene to
benzene show that there was a steady decrease of the 1,3-
butadiene from the very beginning. The 1, 3-butadiene in this
standard decreased by 72% relative, representing the largest
713
-------
relative decrease observed for any standard or mixture over
such a time period. Other mixtures have been prepared and
analyzed in the same manner with the outcome being the same.
The reason for the loss of 1,3-butadiene in the 100 and
10 ppb standards is not known, and it can only be speculated.
During the study periods, no new peaks were found in the
chromatograms from the analysis of these standards.
Therefore, the reaction and formation of other compounds is
probably not occurring, unless these possible new compounds
are being adsorbed onto the cylinder walls. Another
possibility for the decrease in concentration could be that
the 1,3-butadiene itself is being adsorbed onto the cylinder
walls. Polymerization of the 1,3-butadiene at active sites on
the cylinder walls is also a possibility. The loss of 1,3-
butadiene to heterogeneous polymerization would be minimal at
these concentration levels since it depends on two molecules
of 1,3-butadiene coming together at an active site on the
cylinder wall. It would also be controlled by the number of
active sites on the cylinder walls, which would vary from
cylinder to cylinder. Since the 1,3-butadiene decreases in
the 100 and 10 ppb mixtures, it may also be decreasing in
those at the 2 ppm level. However, this same change in
absolute concentration would not be as detectable in the 2 ppm
standards because the analytical imprecision is on the same
order as the observed decreases.
Conclusions
The research described here has shown that no degradation
in concentration has been detected in gas mixtures containing
1,3-butadiene at the 2 ppm level, with 0.03 ppm precision.
The gas mixtures were contained in treated aluminum gas
cylinders, over a period of three years. Longer term
stability at this concentration level is probable. Reliable
and accurate standards can be prepared at this level and used
to calibrate instruments for measurement and monitoring of
1,3-butadiene in the workplace environment. However, gas
mixtures of 1, 3-butadiene at the 100 and 10 ppb are not stable
and are therefore unreliable even for short periods of time.
The results in Table V show that 1,3-butadiene starts to
decrease in concentration immediately after its preparation,
and other data show that the amount of decrease varies from
cylinder to cylinder. The data also show that it is very
difficult to determine the concentration of 1,3-butadiene in
a gas mixture unless a new standard is prepared and the
analysis performed immediately. Even then, the concentration
determined may not be accurate because the 1,3-butadiene
concentration could decrease before the analysis can be
performed. Anyone using stainless steel cannisters, or
flasks made of other materials, should evaluate these
containers for loss of 1,3-butadiene before using them for
environmental or atmospheric sampling.
Acknowledgment
The author wishes to acknowledge Darryl von Lehmden and
Howard Crist of the U.S. Environmental Protection Agency's
714
-------
Atmospheric Research and Exposure Assessment Laboratory for
their support of this work. This work was supported in part
under Interagency Agreements DW-13931661-01-0, DW-139322187-
01-0, DW-13932911-01-0, and DW-13933643-01-0 with the U.S.
EPA. Mention of commercial products in the text does not
imply endorsement by the National Institute of Standards and
Technology.
References
1. J.Santodonato, Monograph on Human Exposure to Chemicals
in the Workplace: 1,3-Butadiene, Cent. Chem. Hazard
Assess., Syracuse Res. Corp., N..Y., Cited: Chem. Abstr. .
105,26:231597, (1986).
2. D.Rotman, "Doubts Cast on Cancer-Risk Claims", Chem.
Week. 144, 25: 9,12 (June 21,1989).
3. "New Health Data May Affect. Clean Air Rules", C&EN. 68,
4:5 (January 22,1990).
4. L.S.Gold, G.H.Backman, N.K.Hooper, R.Peto, "Ranking the
Potential Carcinogenic Hazards to Workers from Exposures
to Chemicals that are Tumorigenic in Rodents", Environ.
Health Perspect.. 76: 211-219 (Dec. 1987).
5. M.W.Ackley, "Chemical Cartridge Respirator Performances:
1,3-Butadiene", Am. Ind. Hyg. Assoc. J.. 48,5: 447-453
(May 1987) .
6. "Lowering of MAC Value for 1,3-Butadiene", Ned. Chem.
Ind.. 12:7 (June 1989).
7. "More Stringent Standards for Chemical Pollutants in the
Workplace Atmoshpere in Norway", Arbeidervern, 16,4: 12
(August 1988).
8. G.C.Rhoderick, W.F.Cuthrell, W.L. Zielinski, Jr.,
Transactions APCA/ASQC Specialty Conference on Quality
Assurance in Air Pollution Measurements. T.R.Johnson,
S.J.Penkala, Ed.; APCA, Pittsburgh, PA, 239-246 (1985).
9. G.C.Rhoderick, W.L.Zielinski, Jr., Conference on Recent
Developments in Monitoring Methods for Toxics in the
Atmosphere, Boulder, CO (July 1987).
10. G.C.Rhoderick, W.L.Zielinski, Jr., "Preparation of
Accurate Multicomponent Gas Standards of Volatile Toxic
Organic Compounds in the Low-Parts-per-Billion Range",
Anal. Chem.. 7,11: 2454-2460 (Nov. 15,1988).
715
-------
Table I. Stability data for 2 ppn 1,3-butadiene in nitrogen standards in
treated aluminum gas cylinders.
Date
Prepared
5/86
8/86
8/86
8/86
9/89
Sample
Number
Gravimetric
Oancentration3
Calculated Concentration3
9/86 5/89 9/89
AAIr-5921
AAL-15945
AAL-15885
AAL-15941
x-138356
1.436±0.004
2.514+0.008
1,760+0.005
1.650+0. 005
2.014+0.006
1.439+0.015
2.515±0.025
1.7594O.018
1.652+0.016
1.434+0.015
2.514+0.025
1.762+0.018
1.646+0.016
1.767+0.018
1.645+0.016
2.012+0.020
Cfcncentraticns are in ppm (micraiDle/mDle) and the uncertainties are
expressed as two standard deviations.
Table II. Stability data for 100 ppb 1,3-butadiene gas standards.
Date Sample
Prepared Number
4/87
8/87
12/88
1/90
AAL-19246
AAL-19249
AAL-20701
X-138366
Gravimetric
Concentration8
97.4+1.0
138 +1
113 +1
106 +1
Calculated Concentration3
9/87° 10/87 l/90d
98.0
93.2
138
90.9
131
88.0
129
94
aConcentrations are in ppto (nanomole/inQle). The uncertainties for the
calculated concentrations are +1.0 ppto at taro standard deviations.
bCcncentration determined by ratioing to a 2 ppn standard of l,3^^utadiene,
AAL-15885 (see table 1).
cConcentration determined by ratioing to AAL-19249.
-------
Table IV. Stability data for 10 ppb 1,3-butadiene gas standards.
Date Sample Gravimetric Calculated Concentration3
Prepared Number Concentration* 5/87 6/87 1/88 6/88
9/86
4/87
5/87
5/87
5/87
5/87
1/88
1/88
6/88
IL-3437
AAL-19253
AAL-19254
AAL-19255
AAL-19256
AAL-19248
CAL-11243
CAL-11233
AAL-20699
15.4+O.3
9.6+O.2
13.2±0.2
12.340.2
7.4+0.2
4.8+0.1
15.2+0.3
3.9+J0.5
14.0+0.3
4.2
3.6
14.8
3.6
14.0
aOcr)oentraticris are in ppb (naronole/roole). The uncertainties for the
calculated ccnoentraticns are +1.0 ppb at tiro standard deviations.
7.9
11.4
7.1
4.8
5.0
8.1
8.8
4.6
2.2
15.2
3.7
Table V. Ratios of GC area response of 10 ppb 1,3-butadiene to benzene ar
toluene to benzene with time showing instability in gas standard AAL-19256.
GC Area Response Ratios
1.3-Butadiene/Benzene Toluene/Benzene
5/18/87
5/19/87
5/20/87
5/26/87
5/27/87
6/03/87
6/19/87
1/19/88
6/07/88
7/01/88
10/05/88
12/07/88
1/10/89
8/01/89
0.667
0.661
0.659
0.597
0.585
0.563
0.525
0.366
0.328
0.292
0.250
0.243
0.237
0.185
1.23
1.22
1.22
1.19
1.20
1.19
1,
1,
1,
1,
1.
1.
1,
23
20
21
22
23
21
,22
1.20
717
-------
TOXIC ORGANIC GAS STANDARDS IN HIGH PRESSURE CYLINDERS
STEPHEN B. MILLER, Ph.D., MANAGER OF RESEARCH AND DEVELOPMENT
ROBERT B. DENYSZYN, Ph.D., TECHNICAL DIRECTOR
MARK S. SIRINIDES, LAB MANAGER
THOMAS E. SASSAMAN,
ROBERT J. TYSON,
Scott Specialty Gases, Inc.
Route 611
Plumsteadville, PA 18949
The accuracy of low ppb toxic organic gas mixtures in high pressure cylinders has been
evaluated for a number of components in complex gas mixtures.
A review of the process taken to calculate the uncertainty will be described along with the
experimental methods taken to minimize bias and improve precision. Results from such evaluation will
be presented for several multi-component gas mixtures. Results from intercomparison of Scott
standards with NIST standards will also be discussed.
718
-------
1. INTRODUCTION
Several years ago, Scott Specialty Gases began numufacturing low concentration toxic organic
gas mixtures, in response to the demand for standards to support analyses at EPA Superfund sites.
Since that time, there has been a rapid increase in concern over environmental issues in general and
significant improvement in analytical methods for low-level organics. The new Clean Air Act will
mandate that more toxic organic measurements be made. Therefore, there is likely to be
continuing demand for reliable standards covering an expanding list of compounds.
The subject of this talk is the Group V toxic organic mixtures. Scott has manufactured a series
of cylinders containing 5 ppb of each of the following eighteen compounds in a nitrogen balance
gas:
Bromomethane (Methyl Bromide) Trichloroethylene
Dichloromethane (Methylene Chloride) Tetrachloroethylene
Trichloromethane (Chloroform) 1,2-Dichloropropane
Tetrachloromethane (Carbon Tetrachloride) 1,3-Butadiene
Trichlorofluoromethane (Freon-11) Benzene
Chloroethene (Vinyl Chloride) Chlorobenzene
1,2-Dichloroethane Toluene
1,2-Dibromoethane 1,2-Dimethylbenzene (o-Xylene)
1,1,1-Trichloroethane Ethylbenzene
The Group V compounds are all important air toxics and are named on the EPA's target
compound lists of extractable volatile organics.
The objectives of this paper will be to describe the steps in the manufacturing process for the
Group V toxic organic mixtures and to discuss some of the quality characteristics of these
standards, including concentration uncertainties and batch variability. Statistical methods for
uncertainty calculations will be reviewed and the overall uncertainties of Scott mixtures will be
contrasted with those for the NIST SRM 1804. Topics will be presented in the following order:
Pre-Blend Preparations
Blending
Reference Standards
Analyses
Statistical Uncertainty
Scott / NIST Intercomparison
2. EXPERIMENTAL
Pre-blend Preparations
The reagents which are used in these mixtures consist of organic materials and nitrogen balance
gas. The organic reagents are purchased from commercial suppliers such as Aldrich and are
assayed very thoroughly in our own laboratories by gas chromatography. The assay is important for
two reasons. First, we need to detect any potential interferences (i.e., species that are present in
one compound that will affect the concentration of another). Second, we need to know the overall
purity of the material with some degree of certainty.
Nitrogen, used as the balance gas, is Scott's own Research Grade nitrogen (99.9999% purity),
and is also analyzed prior to use. It is by far the largest component of the mixture. Therefore,
interferences, even at the ppm level, are extremely important. A typical assay for Research Grade
nitrogen is shown below:
719
-------
Hydrocarbons (as Methane) < 10 ppb
Halocarbons < 2.0 ppb
Carbon Monoxide < 5.0 ppm
Oxygen < 5.0 ppm
Water < 5.0 ppm
Other Impurities < 1.0 ppm
Analyses show less than 10 ppb non-methane hydrocarbons. The hydrocarbons present are
mostly ethane and ethylene, and therefore, do not interfere with any of the components in the
mixture. The other impurities in the nitrogen do affect the overall certainty of the mixture, but do
not interfere with any of the organic reagents in the mixture.
Cylinders are very carefully prepared prior to use. For mixtures of toxic organics, new
aluminum cylinders of 30 liter capacity are purchased and equipped with a CGA 350 stainless steel
packless valve. The cylinder is subjected to hydrostatic test to 3400 psi. Following the hydrostatic
test, the cylinder is thoroughly dried and the valve/cylinder combination is then checked for gas
leakage.
The cylinder is then subjected to our Aculife IV treatment. The Aculife IV treatment is a
three-step process consisting of 1), a thorough deionized water wash, 2), extended heating and
evacuation, and 3), treatment in the gas phase with a chemical which reacts with chemisorbed and
structural water on the metal surface. The Aculife IV treatment is key to the stability of Group V
organic mixtures.
The Aculife IV treatment has been able to produce stable mixtures of the compounds shown below:
Acetone 1,2-Dibromomethane Freon 12
Acetonitrile m-Dichlorobenzene n-Hexane
Benzene 1,3-Dichloroethane Methylene chloride
Benzyl chloride 1,2-Dichloropropane Methylethyl ketone
Bromomethane 1,4-Dioxane Perchloroethylene
1,3-Butadiene Ethyl benzene Toluene
Carbon tetrachloride Freon 11 1,1,1-Trichloroethane
Chlorobenzene Freon 113 Vinyl chloride
Chloroform Freon 114 Vinylidene chloride
o-Xylene
At Scott, new cylinder treatments are evaluated continually, but no treatment has yet been
identified which can stabilize the organic compounds shown below:
Acrylonitrile Formaldehyde
Aniline Propylene oxide
Bromoform Pyridine
Ethylamine Styrene
Ethylene oxide trans-l,4-Dichloro-2-butene
In general, compounds which are highly polar or which have strong hydrogen-bonding
capability present stability problems.
Cylinder handling and gas transfer operations are accomplished using a specialized blending
manifold. A great deal of care is taken with the components of the blending manifold. Only
bellows or diaphragm valves are used. Stainless steel or nickel tubing is used throughout, and
connections are either orbital welded or made using VCR-type fittings.
Blending
Blending, the process by which the components are combined at the proper concentration,
consists of five steps:
720
-------
1. A master liquid blending mixture of the fourteen liquid components is prepared in a glass
container gravimetrically.
2. A master gas blending mixture is prepared by a three-step process:
• A small quantity of the master liquid blending mixture is introduced gravimetrically to
an evacuated cylinder by the Micrograv™ method. The Micrograv™ method is a
variation of the glass capillary technique reported in the literature.
• The four remaining gaseous components (F'reon 11, vinyl chloride, bromomethane,
and 1,3-butadiene) are introduced to the master gas blending cylinder gravimetrically.
• Nitrogen balance gas is added gravimetrically. The final concentration of each of the
eighteen components in the master gas blending mixture is 500 ppb.
3. The eventual product cylinder is connected to the blending manifold, evacuated and tared.
4. An appropriate quantity of the master gas blending mixture (eg, 50.0 g) is introduced to
the product cylinder gravimetrically.
5. Finally, nitrogen balance gas is added to the product cylinder (eg, 5000 g) to achieve the
final pressure and component concentrations.
Since ah1 operations are performed gravimetrically, very low blending uncertainties can be
achieved. Cylinders can be weighed to ±0.1 g, yielding total uncertainties from 0.1 to 0.5%.
Reference Standards
Even though low blending uncertainties are readily achieved, adsorption/desorption
phenomena can occur within the cylinder which change the blended concentrations of the organic
components. Therefore, each cylinder must be analyzed and reference standards are required for
the analysis.
Two sets of reference standards are required for the Group V toxic organic mixtures. One set
covers the FID-responsive species (16 components) and the other covers the ECD-responsive
species (Freon-11 and carbon tetrachloride). Standards at the appropriate concentration level are
often not readily available for these mixtures, and therefore, it is necessary to generate our own
reference standards.
The steps required to produce the FID reference standard are similar to those used to
produce the original mixtures:
1. A second master liquid mixture of thirteen liquid components is prepared gravimetrically.
This mixture is then diluted by a factor of 10,000 with a suitable solvent in order to give a
weighable quantity of the liquid solution containing low concentrations of the organic
components.
2. The FID standard itself is prepared in a glass flask by a two-step process:
• A small quantity of the master liquid standard mixture is introduced gravimetrically
into the glass flask, which had been pre-treated and evacuated to assure minimum
container effect.
• The three remaining gaseous components (vinyl chloride, bromomethane, and
1,3-butadiene) are introduced individually to the glass flask manometrically.
721
-------
The glass flask mixture is used as a standard for gas chromatography analysis of the
FID-responsive species. The glass flask is used to assign values to a secondary standard, a
previously-prepared cylinder mixture (AAL 18488) of the eighteen components in nitrogen at 270
ppb. Although not essential to the certification process, this procedure provides a relatively large
supply of the secondary standard with known values for future use.
Finally, a working FID reference standard at 10 ppb is prepared in a glass or stainless steel
flask by diluting a small quantity of the 270 ppb standard by a factor of 27 using manometric
techniques. The 10 ppb mixture is used as the standard for the analysis of the batch of Group V
mixture cylinders.
The process for preparing the ECD reference standard is similar. A glass flask, similar to that
used for the FID standard, is prepared and the two gaseous haloearbon materials introduced
manometrically. No intermediate (secondary) standard is prepared.
Analyses
At this point, a batch of cylinders of the 5 ppb Group V mixtures, and two reference standards,
have been prepared. Each cylinder in the batch can now be analyzed. One cylinder is selected at
random from the batch to be the batch master. Since we have only limited quantities of the two
reference standards available, the reference standards are used to assign values to the batch master,
and then the batch master is used to standardize the remaining cylinders. As a QC measure,
analyses are mixed over the course of the analysis process. The batch master is analyzed, then two
batch cylinders, the batch master again, two batch cylinders, and so on. This procedure yields
statistics for the batch master and a measure of long-term drift during the course of the analysis.
The conditions used for the gas chromatographic analyses are as follows:
FID-Responsive Species (All Except F-ll and CC14)
« HP 5890A FID
* 30 ra x 0.53 mm DB-1 (Megabore)
» 9 cc/min Helium
» Temperature Program from Q°C to 165°C
ECD-Responsive Species (F-ll and CC14)
• Shimadzu Mini-2 ECD
• 6 ft x 1/8 in SP-1200 (packed)
* 25 cc/min P-5 (Argon/Methane)
» 80 °C Isothermal
Quality
The overall quality of these mixtures can be described by three key variables:
1. Individual component uncertainty
2, Individual component stability
3. Batch variability
Each of these key factors will be discussed separately.
Virtually every operation in the process, whether it be mechanical or analytical, carries with it
some level of uncertainty. Uncertainties can arise from three main sources:
722
-------
1. Repeatability (the precision of the measurement process)
2. Accuracy (the capability of the measurement device or technique)
3. Purity (assay of the mixture materials)
Uncertainties of all types are additive, ie, uncertainties accumulate as each succeeding
operation is performed. The uncertainty of the final mixture can be determined mathematically by
estimating the accumulated errors using the propagation of errors technique.
For any given operation, the incremental uncertainty can be calculated by taking the
magnitude of the error associated with the operation, dividing it by the magnitude of the operation
of itself, squaring that quantity, and then taking the squEJe root of that result. A 95% confidence
level can be achieved by multiplying the result by 200.
The overall uncertainty in the final mixture is i;he quadrature sum of the incremental
uncertainties for all underlying operations, whether they ;ire of the repeatability, accuracy, or purity
type.
Our experience has shown the level of performance to expect from each step of the operation.
The table below summarizes key process steps and typical relative uncertainties for the step:
SOURCE DOMINANT FACTOR CONTRIBUTION
Raw Materials Purity 0.1%
Blending Scale accuracy (mass dependent) 0.2 - 0.5%
Analysis GC Precision (component dependent) 0.5 - 10%
Standards GC Precision (component dependent) 0.5 -10%
Well-characterized, pure raw materials contribute very little to the uncertainty of the mixture,
0.1% or less. For blending, usually performed gravimetrically, the relative uncertainty contribution
is dependent on the capability of the balance being used and on the absolute mass of material being
weighed. Using a balance capable of ±0.1 g, even relatively small masses (eg, 50 g) can be
measured with an incremental uncertainty of 0.2%. For multi-component mixtures, overall blending
uncertainties range from 0.2 - 0.5%. For the analysis of the mixtures, the dominant performance
factor is the repeatability of peak area in replicate GC analyses. For any given component, the
incremental uncertainty contribution is dependent on the: GC behavior of the component and can
range from less than 1% to more than 10%. In general, compounds which elute late in the
chromatogram or exhibit bad peak shape give poor analytical precision, and therefore contribute
the highest amount of uncertainty in the final product. Total analytical uncertainties range from 2 -
10% for GC analysis, and this process step dominates the total uncertainty for the mixture.
There are a large number of individual operations that occur during the blending and analysis
of an eighteen component toxic organic mixture. The total propagation of error calculation is
extremely laborious to do, so much so that it is impractical to perform for every cylinder produced.
Rather than perform the exhaustive propagation of error.', calculation, Scott relies on a conservative
estimate of 10% overall uncertainty for all components in a 5 ppb mixture. It is very unusual for a
compound to exceed the 10% overall uncertainty criteria.
3. RESULTS
The first quality variable to be discussed is component uncertainty, i.e., the uncertainty for a
given component within one cylinder. If the analytical value determined for the given component,
and its associated uncertainty, includes the gravimetric blending value for that component, and its
uncertainty, then that component is within specifica:ion. For example, a component is
gravimetrically blended at 5 ppb, with a 0.5% relative uncertainty (or ±0.025 ppb), and analytically
determined to be 4.95 ppb, with a 5% relative uncertairty (or ±0.25 ppb. The gravimetric value
would be contained in the analytical value, and this result would meet Scott's criteria for component
uncertainty.
723
-------
The second major quality variable to be discussed is component stability. This variable is
evaluated by performing replicate analyses over an extended period of time to observe
concentration changes. A component is stable if the precision of replicate measurements made
over the time period is within the 10% analytical tolerance. Stabilities for representative
components studied over a nine month period are shown below.
COMPONENT 2/20/8"
Chloroform 9.092
Benzene 8.873
Vinyl Chloride 9.644
Toluene 9.114
Bromomethane 9.815
9.072
8.408
9.905
9.057
9.374
% DELTA
0.22
5.20
2.70
0.63
4.50
The third and final major quality variable is batch variability, or the variability of a given
component within a lot of identical cylinders. For a given component, this parameter is evaluated
by calculating the ratio of the analytical value in the batch master to the analytical value for that
same component in another cylinder from the batch. The calculation is repeated for each cylinder
in the batch and statistics are performed on the set of ratios. A component values are identical
within the batch if the precision of the analytical value ratios is within the 10% analytical tolerance.
As can be seen in the table below, certain of the Group V toxic organics (bromomethane and
1,3-butadiene) exhibit unacceptable batch variation. NIST is unable to certify these components in
the Group V mixtures for this reason.
Freon 11 1.6%
Benzene 2.2%
Chloroform 2.4%
o-Xylene 3.2%
1,3-Butadiene 13.2%
Bromomethane 15.6%
The table below shows a comparison of Scott Group V toxic organic standards to the NIST
Standard Reference Material 1804. The first three columns show the Scott gravimetric value, the
Scott analytical value, and the Scott total uncertainty (based on the 10% analytical tolerance
criteria). The final two columns show the NIST analytical value and the calculated tolerance or
uncertainty. The Scott analytical value, with the 10% tolerance, includes both the NIST analytical
value and the Scott gravimetric value for virtually every component. Further, the NIST analytical
value, with its published tolerance, includes the Scott analytical value for most components. It can
be concluded that the Scott cylinders and the NIST cylinders are essentially identical, except for the
large estimated uncertainty associated with Scott values.
SCOTT SCOTT SCOTT NIST NIST
COMPOUND GRAY. ANAL. TOL. ANAL. TQL.
Benzene 5.0 5.2 ±0.5 5.0 ±0.1
Bromomethane 5.0 5.2 ±0,5 5.4
1,3-Butadiene 5.0 5.4 ±0.5 4.6
Carbon Tetrachloride 5.0 4.6 ±0.5 5.0 ±0.1
Chlorobenzene 5.0 5.5 ±0.5 5.0 ±0.2
Chloroform 5.0 5.0 ±0.5 4.9 ±0.2
1,2-Dibromomethane 5.0 5.6 ±0.5 5.0
1,2-Dichloroethane 5.0 5.2 ±0.5 5.0 ±0.2
Dichloromethane 5.0 5.1 ±0.5 5.1 ±0.3
724
-------
SCOTT SCOTT SCOTT NIST NIST
COMPOUND GRAY. ANAL. TQL. ANAL. TOL.
1,2-Dichloropropane 5.0 5.2 ±0.5 5.0 ±0.2
Ethylbenzene 5.0 5.7 ±0.5 4.7 ±0.2
Tetrachloroethylene 5.0 5.4 ±0.5 5.0 ±0.2
Toluene 5.0 5.2 ±0.5 4.9 ±0.2
1,1,1-Trichloroethane 5.0 5.3 ±0.5 5.0 ±0.1
Trichloroethylene 5.0 5.2 ±0.5 5.0 ±0.2
Trichlorofluromethane 5.0 4.9 ±0.5 5.1 ±0.1
Vinyl Chloride 5.0 5.2 +0.5 5.2 ±0.2
ortho-Xylene 5.0 6.0 ±0.5 5.1 ±0.2
4. CONCLUSION
Scott Specialty Gases, therefore, represents a commercial source for reliable standards for
toxic organic work with near-NIST quality. Many other mixtures of similar quality are also
available.
725
-------
HUMIDIFIED CANISTER STABILITY OF SELECTED VOC'S
Rita M. Harrell, Richard E. Means, and Kenneth J. Caviston
NSI Technology Services Corporation
P.O. Box 12313
Research Triangle Park, NC 27709
William J. Mitchell, PhD
U. S. EPA Environmental Monitoring Systems Laboratory
Research Triangle Park, NC 27711
ABSTRACT
In the environmental monitoring field the need for
reliable standards for auditing and documenting data is well
established. An important factor to take into consideration
when such standards are prepared in canisters is the
stability of the VOC's for the duration of time that they are
to be used. Stability studies involving SUMMA canisters from
several different manufacturers will be described in detail.
INTRODUCTION
NSI Technology Services Corporation serves a contract
laboratory for the preparation of VOC audit materials for the
Environmental Protection Agency. A major portion of these
materials are prepared in SUMMA canisters of the same type as
are used for field sampling. VOC "shelf life" is an
important consideration for such canisters as they are often
used as QA samples periodically during a extended period of
time as well as being used immediately upon receipt. This
report primarily focuses on 6 canisters loaded from the same
VOC cylinder under identical conditions of humidity, etc. ;
but from different manufacturers.
726
-------
EXPERIMENTAL
In order to compare the stability of selected VOC' s in
humidified SUMMA canisters, the same Group V cylinder was
selected to pressurize canisters from several manufacturers
to a undiluted final pressure of 40 psig. All canisters were
humidified by the injection of the appropriate amount of
boiled, ultrapure deionized water to provide 50% relative
humidity.
Following a 24 hour equilibration period, canisters were
analyzed on a periodic basis using an HP 5890 GC/FID equipped
with a cryoconcentrator and an HP-5 (crosslinked) 5% phenyl
methyl silicone capillary column 30ra X 0.55mm with a 0.88 urn
film thickness. Sample sizes were approximately the same for
all analyses.
RESULTS
For simplicity, the 18 compounds present in Group V
cylinders were subdivided into 4 groups according to
classes of compounds and representative compound used for
graphing purposes. Manufacturers are denoted by A, B, C(C1 &
C2 are different types) , D, and E.
Figure 1 represents the class of compounds which
includes benzene, toluene, chlorobenzene, ethylbenzene,
styrene and o-xylene. Stability data is well within
acceptable limits for all canisters throughout an 38 week
period with the exception of A and B. Both of these canisters
show a dramatic drop in concentration at 18 weeks while the
others continue to show good stability.
Figure 2 represents the class of compounds which
includes vinyl chloride, trichloroethylene, and
tetrachloroethylene. For this class of compounds, stability
is good throughout a 38 week period for all canisters except
B which again shows a large decrease in concentration
starting a week 20 or 21.
Figure 3 represents the compounds 1,1,1-trichloroethane,
1,2-dichloroethane, and 1,2-dibromoethane. Stability data for
these compounds show a similar trend as those represented in
Figure 1. A and B begin dropping in concentration at week 21.
Figure 4 represents bromomethane, freon-11, carbon
tetrachloride, and chloroform. Again canisters A and B show
differences from the other canisters with A dropping off at
21 weeks and B at 23 weeks.
CONCLUSION
From the results obtained, it is apparent that for up to
at least 18 weeks Group V VOC's are stable regardless of who
727
-------
manufactured the canisters. However, after this point in time
two of the canisters show a definite concentration drop which
negates their usefulness as audit materials. Thus, careful
consideration of VOC residence times in humidified SUMMA
canisters should be a factor in the selection of canister
vendors.
ACKNOWLEDGMENTS
The authors wish to thank Shirley Henry and Annette King
at NSI Technology Services Corporation for their roles in
canister cleaning and preparation.
REFERENCES
(1). This support was provided under contract number
68-02-4444 with the U.S. EPA.
(2). For details of canister preparation and analysis
procedures see last year's proceedings p.757.
728
-------
. ftROHATIC HYDROCARBON STABILITY
Cnister Rnfacturr CoHpartsans
CD
HALOGENATED ETHYLEHE STABILITY
blister taufjctmr Cmrisons
ffl
Q.
Q.
729
-------
HALOGENATED ETHANE JTABILITY
blister taifartmr Caparisons
0
Q.
a.
HALOGENATED METHANE IT ABILITY
Cnister hnufactinr Cnwrisons
0
1
730
-------
EVALUATION OF A CONTINUOUS SAMPLING AND
ANALYSIS SYSTEM FOR VOLATILE ORGANIC
COMPOUNDS
James M. Hazlett
Robert E. Bailey
Air Toxics Laboratory
Louisiana Department of Environmental Quality
Baton Rouge, Louisiana
ABSTRACT
The Louisiana Department of Environmental Quality has
completed a test project on a gas chromatograph based system
to collect and analyze hourly ambient air samples for Volatile
Organic Compounds. Most ambient sampling strategies for
Volatile Organic Compounds revolve around time integrated
samples collected in canisters or on solid adsorbents. While
this type of sampling strategy can provide much useful
information, it cannot provide detailed data such as diurnal
patterns, daily maximum/minimum concentrations, etc.
The analysis system tested is composed of a XonTech Model
930 Organic Vapor Concentrator coupled to a Hewlett Packard
Model 5890 Gas Chromatograph. The sample concentrator
utilizes a dual Tenax trap system which permits sample
collection on one trap while an analysis is being performed on
the other. The gas chromatograph is fitted with a wide bore
capillary column and FID/HECD Detectors, The system is
programmed to operate 24 hours per day collecting a 56 minute
sample every hour with subsequent analysis. The system
currently provides hourly data on 28 volatile Organic
Compounds.
This paper will include a presentation of ambient
monitoring data, quality assurance data, and an overall
evaluation of the systems performance and reliability.
731
-------
INTRODUCTION
The Louisiana Department of Environmental Quality (LDEQ)
has operated an Air Toxics Monitoring Station since 1984. This
station, located near downtown Baton Rouge, has been one of the
few monitoring sites in the country capable of providing round-
the-clock monitoring data for airborne volatile organic
compounds. The original system consisted of a computer
controlled dual gas chromatograph and data acquisition system.
During the five years this site was in operation the analytical
system was able to produce a vast amount of monitoring data.
However the site was also besieged with a number of problems.
These problems included compound misidentification due to low
chromatographic resolution, high maintenance costs, frequent
breakdowns, and an inadequate data system. During the spring of
1989 a search began for a replacement system.
In order meet the goals of the LDEQ AIR Toxics monitoring
program and address the problems experienced with the existing
monitoring station the new equipment selected had to meet
several requirements.
Hourly Samples of at least 56 minute duration
Concentrator must start GC and Data System
Dual bed sorbent traps to enable capture C2-C10
hydrocarbons
Dual detectors:
Flame lonization Detector
Electrolytic Conductivity Detector (Halogen mode)
Detection levels down to 0.1 ppbv for selected compounds
User selectable sample, zero, and calibration analyses
Data reports at the end of each hourly run
Storage of data in PC compatible files
Ability to recover after a power failure
In addition to the above requirements the replacement system had
to be cost effective by providing a high volume of data with low
costs for operation and maintenance.
SYSTEM DESCRIPTION
In July of 1989 the replacement system was selected and
placed into operation. This system consists of the following
three'sub-systems;
1. XonTech Model 930 Organic Vapor Concentrator equipped
with dual sorbent bed traps.
732
-------
2. Hewlett Packard 5890 gas chromatograph equipped with a
Flame lonization Detector & an O.I. Electrolytic
Conductivity Detector configured in the halogen mode.
3. Dual Hewlett Packard Model 3396 programmable computing
integrators equipped with HP-BASIC.
In addition a GRID lap top MS-DOS computer was purchased in
order to extract the data from the integrators and transfer it
to the agency database.
The XonTech model 930 Organic Vapor Concentrator consists
of two vertically mounted sorption traps which are alternately
activated in order to provide continuous air sampling. The
traps consist of a 1/8" o.d. x 12" long stainless steel tube
packed with a 6" bed of Tenax TA tacked up with a 3" bed of
Carboxen-564. This type of dual bed trap has been shown to
provide excellent recoveries for a wide range of compounds. The
sample flow rate is set at 10 CC/min providing a total sample
volume of 560 CC. Each trap is heated and cooled separately for
independent operation through the following 60 minute cycles:
A. 56.0 minute sampling of ambient air
4.0 minute Helium purge to remove air from sorbent tube
B. 2.5 minute Desorption at 200'C to inject sample
15.0-minute cleaning at 215'C
42.5 minute cooling to ambient temperature
During continuous operation when one trap is performing cycle
A, the other trap is performing cycle B. At the end of the 60
minute cycle the traps switch functions. At the beginning of
each desorption cycle the concentrator sends a remote start
signal to the GC and the integrator data system. The Organic
Vapor Concentrator is connected to the Gas Chromatograph by
means of a heated transfer line.
The gas chromatograph is equipped with a sample splitter
that divides the sample between two separate columns. A SPB-1
capillary column 50 m x 0.53 mm x 2.65 urn is attached to the
Flame lonization Detector while a SPB-1 column 25 m x 0.32 mm x
0.52 urn is attached to the Electrolytic Conductivity Detector.
This combination yields a split ratio of about 5:1 in favor of
the FID. The GC is temperature programmed for an initial
temperature of 30 degrees C. Three minutes after the run start
the temperature is increased at a rate of 2.5 degrees C per
minute up to a final temperature of 130 degrees C. The total GC
run-time is 45 minutes.
733
-------
The monitoring system is calibrated three times a week
using compressed gas cylinders containing certified
concentrations of the compounds of interest. The retention time
of each compound is stored in the integrators along with an
appropriate specific response factor. At the end of each
subsequent analysis the integrators automatically execute a
BASIC program which matches retention times to identify
compounds and quantifies them by using the specific response
factors determined during the calibration analyses. The total
areas of all peaks including unknowns are summed for each
detector. Total area on the Flame lonization Detector is
multiplied by the hexane response factor to yield a value for
total hydrocarbons. Total area on the Electrolytic Conductivity
Detector is multiplied by the ethylene dichloride response
factor to yield a value for total chlorinated VOC. The data is
then written to a daily data file in each integrator and stored
under unique file names. At midnight each day a 24 hour summary
report is printed. Each integrator has enough memory capacity
to hold at least six days of data. Periodically the site
operator will transfer the data files from the integrators to
the lap top computer for editing and subsequent storage in the
agency database.
In addition to the analytical data, the monitoring site has
a meteorological equipment tower which measures wind direction,
wind speed, and ambient temperature. This data is gathered in
the form of hourly averages and stored in the agency database
along with each chromatographic analysis.
Detection levels have been determined to be about 0.1 ppbv
on the Flame lonization Detector for most of the hydrocarbons
and about 0.02 ppbv on the Electrolytic Conductivity Detector
for the chlorinated VOC.
OPERATIONAL RESULTS
Since the monitoring system was placed in operation it has
been able to sample and analyze around 700 ambient air samples
per month. The equipment has been able to recover from most
short term power failures resulting in little loss of data.
There were a few problems initially with background noise on the
Electrolytic Conductivity Detector. Once that problem was
solved both detectors have subsequently performed quite well.
The monitoring system was subjected to several tests in
order to fully evaluate the precision and accuracy. A
performance audit cylinder containing low ppbv levels of several
volatile organic compounds was attached to the sampling manifold
and subsequently analyzed. The results listed in ( Table 1 )
734
-------
show the monitoring system was able to measure most of the
compounds within an accuracy of +/- 20%. In order to evaluate
the precision and repeatability of the system, a gas mixture was
sampled repeatedly over a eight hour period. The results show
a slight difference ( +/- 5% ) in instrument response occurring
on alternating runs. This effect is caused by slight
differences in the trapping and/or desorption efficiency between
the two sorbent traps. Since the concentrator unit has the
capability of injecting an internal standard with each sample,
it is felt that the use of an internal standard calibration
method will eliminate this minor problem.
The monitoring system was co-located with a canister based,
time integrated sampling system which was part of the Urban Air
Toxics Monitoring Program. A comparison of data was made by
taking the mean average of the 24 hourly samples on a given
sample date, and comparing those averages to the results
obtained from the canister analysis. The results contained in
( Table 2 ) show the monitoring system is capable of producing
data of nearly the same quality as the time integrated canister
based sampling system.
.AMBIENT DATA RESULTS
Table 3 shows the mean average concentrations observed for
the period January 1, 1990 through March 31, 1990. During this
period of time the monitoring system collected and analyzed 2003
samples. This represents a data capture of 92.7%.
The most striking fact about; the monitoring data gathered.
is that the levels of volatile organic compounds observed are
highly variable. Often during any g:.ven twenty four-four hour
period the range of measured concentrations can be quite high.
The diurnal pattern of concentrations can be observed by sorting
rhe monitoring data by the hour of the day in which the samples
were collected. As a result it can be seen that most of the
hydrocarbons generally reach their maximum concentrations during
~he 6 - 9 am period and hit their lowest reading during the 12 -
3 pm period ( Figure 2 ) . No noticeable diurnal pattern was
observed for the chlorinated volatile organic compounds.
Another useful type of data analysis involves sorting the
data into groups according to the direction from which the wind
was blowing when the samples were collected. This kind of data
analysis is valuable in pinpointing- the effect local point
sources can have on a monitoring station. The LDEQ monitoring
station is located to the south of several petrochemical plants.
Consequently when the wind is from a northerly direction a
corresponding increase in the observed levels of volatile
735
-------
organic compounds can be seen ( Figure 3 ).
CONCLUSIONS
The continuous monitoring system established by LDEQ
provides a useful alternative to the field collection and
laboratory analysis of ambient air samples. Initial estimates
of the precision and accuracy compare favorably with the
precision and accuracy measured during the early development of
other air toxics monitoring techniques. Improvements in the
precision and accuracy are expected as the knowledge and
understanding of the system operation grows. The overall
performance of the system has been good enough that LDEQ has
begun formulating plans to expand this type of monitoring
program to several sites around the state.
REFERENCES
1. R. J. Sullivan, M. Yoong, and G. Watson (XonTech Inc.),
"Automatic Organic Vapor Concentrator for the Continuous
Measurements of VOCS in Air", Proceedings of the 1989 EPA/A&WMA
International Symposium on the Measurement of Toxic and Related
Air Pollutants.
2. Model 930 Organic Vapor Concentrator - Operations and
Maintenance Manual, XonTech Inc.,Van Nuys, Ca.
3. Robert D. Cox, "Sample Collection and Analytical Techniques
for Volatile Organics in Air", Measurement and Monitoring of
Noncriteria (Toxic) Contaminants in Air Specialty Conference,
Air Pollution Control Association, March 1983.
736
-------
TABLE I
Performance Audit Results
RTI Cylinder AAL-21616
Compound
vinyl chloride
methylene chloride
chloroform
1,2-dtehloroethane
carbon tetraohlorlde
triohloroethylene
perohloroethylene
benzene
toluene
ethyl benzene
o-xylene
RTI Cone. PPB DEO Analysis PPB
6.27
4.41
4.86
6.06
6.06
6.10
6.24
5.59
6.19
4.80
6.29
3.21
3.74
4.86
4.43
4.76
4.66
6.09
6.42
6.04
4.81
6.00
-39.1%
-16.3%
+ 0.3%
-12.6%
-6.2%
-10.8%
- 2.8%
-2.9%
-2.9%
+ 0.3%
-6.6%
TABLE II
Data Comparison
with Canister Based Sampler
Date of Sample Benzene
LDEQ - Canister
Concentration* In PPBV
Toluene
LDEQ - Canister
7/27/89
8/08/89
8/20/89
9/01/89
9/13/89
9/25/89
1.7
2.1
2.2
0.6
1.2
1.3
- 1.80
- 2.12
- 2.13
- 0.88
- 1.12
- 0.79
2.4
3.1
3.7
1.0
2.5
1.1
- 2.61
- 2.30
- 2.55
- 0.96
- 2.14
- 0.88
737
-------
( TABLE 3 )
AVERAGE CONCENTRATIONS MEASURED
FID HYDROCARBONS - BATON ROUGE, LA.
Compound Mean Average Concentration (PPBV)
Propane 8.4
Butane 4.7
2-methylbutane 3.7
Pentane 1. 9
2-methylpentane 1.0
3-methylpentane 0 . 6
Hexane 1.0
Methylcyclopentane 0.5
Benzene 1. 0
2-methylhexane 0.2
2, 2, 4-trimethylpentane 0.3
Heptane 0.2
F-" rhylcyclohexane 0.2
Toluene 1.6
Octane 0.1
Ethylbenzene 0.2
m+p xylene 1.3
o xylene 0 . 6
Cumene 0.2
1,2,4-trimethylbenzene 1.2
Unknowns 6.3
Total Hydrocarbons 35.2
ELCD CHLORINATED VOC - BATON ROUGE, LA.
Compound Mean Average Concentration (PPBV)
Vinyl Chloride 0.4
Methylene Chloride 0.1
Chloroform <0.1
Ethylene Dichloride 0.5
Carbon Tetrachloride 0.1
Trichloroethylene <0.1
Perchloroethylene 0.1
Unknowns 1.5
Total Chlorinated VOC 2.5
738
-------
FIGURE I
Diurnal Variation in Concentration
Baton Rouge, Louisiana
PPBV
'#'
A X
0.6 "
0 1 2346678 910 11 12 13 14 16 161718 192021222324
Hour of Day
—* - B«nz»ra -+- Toluen* -+- Carbon Ttl.
January 1, 1990 - UareH 91, 1990
FIGURE II
Variation of Concentration with
Change in Wind Direction
PPBV
9'r-
-k,
2^
K ' ^ '' " " ' ^'
0
_L.
0 00 MO 270 360
Wind Direction
~-~EDC -+•
January I 1990 - March 31, 1990
739
-------
CANISTER BASED SAMPLING SYSTEMS
A PERFORMANCE EVALUATION
Dave-Paul Dayton
Radian Corporation
P.O. Box 13000
Research Triangle Park, NC 27709
David A. Brymer
Radian Corporation
P.O. Box 201088
Austin, TX 78720
Robert F. Jongleux
Radian Corporation
P.O. Box 13000
Research Triangle Park, NC 27709
Canister based sampling systems have gained wide acceptance for the collection of integrated
whole ambient air samples containing volatile organic compounds. Utilization of this sample collection
method has increased significantly. There are many different canister based sampling systems. The various
sampling system incorporate diverse operating principles but always share one common element, a
SUMMA*-treated, stainless steel canister as the sample containment vessel. The technique and associated
hardware used to time-integrate the sample is the primary difference between the various operating
principles. Integration techniques include the use of several electronic and/or mechanical devices. These
devices include pumps, variable orifices, fixed orifices, and mass flow controllers, either separately or in
combination. Radian Corporation has designed, fabricated, and used canister based sampling systems,
encompassing many of the various operating principles, during several large field sampling programs. A
performance evaluation of several canister based sampling systems, addressing time-integration characteristics
and reliability has been compiled. An estimate of sampler system associated bias is assessed, based on the
suggested certification procedure detailed in Method TO-14, published by the U.S. Environmental Protection
Agency.
740
-------
INTRODUCTION
Canister based sampling systems have gained wide; acceptance for the collection of time-
integrated whole ambient air samples containing volatile organic compounds. There are many canister based
sampling systems which incorporate diverse operating principles but always share one common element, a
stainless steel SUMMA*-treated canister as the sample containment vessel.
The technique and associated hardware used to perform time-integration of a canister sample is
the primary difference among the various operating principles. Time-integration techniques generally involve
the use of electronic and mechanical devices either separately or in combination. Several of these systems
lire commercially available while others are custom built for a specific application.
Radian Corporation has gained vast experience with canister based sampling systems through the
management and operation of the multiyear national Nonmethane Organic Compound (NMOC) and Urban
Air Toxics Monitoring Program (UATMP), under contract to lie U.S. Environmental Protection Agency
(U.S. EPA). Radian Corporation has also managed and operated extensive regional monitoring networks for
private industry. The operation of these government and industry monitoring networks has involved the
design, fabrication, certification, deployment, and operation of approximately 100 canister sampling systems
Encompassing many of the current principles of operation. Mc>dification and repair of the sampling systems
are also an integral part of operating the monitoring networks. During the course of these programs
approximately 14,725 samples were collected at over 218 sites across the country during the past six years.
The following performance evaluation of canister sampling systems is based on experience
gained during the aforementioned programs. The evaluation addresses system associated bias, based on the
certification procedure suggested in Method TO-14.1 In additiDn, time-integration characteristics and
reliability are assessed.
SAMPLING SYSTEM GROUPING
Although the evaluated sampling systems incorporated various operating principles, they can be
categorized into three basic groups. Systems in Group A use negative pressure generated by a pump pulling
against an orifice assembly to deliver an integrated sample into the canister. Group A sampling systems can
t>e used to collect a sample that has either a negative or a positive final pressure. This is usually dependent
upon the needs of the analytical methodology for which the sample is being collected. This paper addresses
only negative final sample pressure collections. Systems in Group B use an electronic mass flow controller
and the vacuum present in the canister to deliver a time-integrated sample into the canister. Systems in
Group C use the vacuum present in the canister pulling agains: a vacuum regulation assembly to delivery an
integrated sample into the canister.
Group B and C systems can be used to collect negative final pressure samples only, unless a
pump is also incorporated into the sampling system. None of ihe models of mass flow controlled or vacuum
regulation sampling systems evaluated incorporated a pump.
SAMPLING SYSTEM CERTIFICATION
Procedure
Sampling systems are subjected to a laboratory certification procedure to quantify any additive
or subtractive biases the sampling systems may contribute to the samples they collect. The certification
procedure, described in Method TO-14,1 involves determination of system specific compound recovery and
contamination. A challenge sample of a blend of organic com[x>unds at known concentrations in humidified
air, is collected through the sampling system. Analysis of the challenge sample is performed and percent
rscoveries of the organic compounds are calculated. The calculated percent recoveries are used as a gauge
of additive and/or subtractive bias on a system specific basis. A humidified zero air blank sample is then
collected through the sampling systems to gauge further additive bias.
741
-------
Results
Twenty-two Group A sampling systems, representing three different models, received sampler
certification. The three models included in Group A were the U.S. EPA NMOC Sampler, the Radian
Corporation RCS Sampler, and the Scientific Instrumentation Specialists AGS-l/B Sampler. Thirty-four
Group B sampling systems, representing three different models, also received sampler certification. The
three models included in Group B were the U.S. EPA UATMP Sampler, the U.S. EPA Modified Type "L"
Sampler, and the Radian Corporation RCS/M Sampler. Four Group C sampling systems, representing one
model, received sampler certification. The model included in Group C was a Veriflow Vacuum Regulator.
A blend of benzene, methylene chloride, toluene, vinyl chloride, and o-xylene, all at approximately 104 ppbv,
was delivered to the Radian Certification System from a high pressure cylinder. Humidified zero air from
the Radian Canister Cleaning System was also delivered to the certification system and mixed with the
organic compounds to produce the desired challenge sample concentrations.
Each of the Group A and Group B sampling systems collected a challenge sample from the
Radian Certification System. The challenge samples were analyzed by multiple detector gas chromatograph
and percent recovery for each compound tested was calculated on a sampler-specific basis. The individual
percent recovery figures were then averaged on a compound-specific basis within the two sampling system
categories. Table 1 presents the sampling systems challenge data. Average recoveries calculated for the
Group A systems ranged from 94.8% for o-xylene to 108.7% for toluene, with an overall mean of
102.8 percent. Standard deviations for the Group A systems ranged from 2.6% for methylene chloride to
7.7% for toluene, with an overall mean standard deviation of 4.5 percent. Average recoveries calculated for
the Group B systems ranged from 80.2% for vinyl chloride to 100.9% for o-xylene, with an overall mean of
91.7 percent. Standard deviations for the Group B systems ranged from 2.1% for o-xylene to 22.2% for vinyl
chloride, with an overall mean standard deviation of 11.2 percent. Average recoveries calculated for the
Group C systems ranged from 91.7% for benzene to 108.0% for o-xylene, with an overall mean of
98.2 percent. Standard deviations for the Group C systems ranged from 1.8% for methylene chloride to
15.9% for o-xylene, with an overall mean of 10.7 percent. Comparing the means of the percent recoveries
for the challenge samples shows there is an apparent difference at the 0.05 level of significance or a 95%
probability between the Group A sampling systems and the Group B systems. There is also an apparent
difference between Group A sampling systems and Group C systems. There is no apparent difference
between Group B sampling systems and Group C systems. Not enough data is currently available to explain
these differences. Although there is a statistical difference between the Group A and Group B and C
recovery data, 92%, 103%, and 98% are generally considered to be reasonable and acceptable percent
recoveries.
When the collection of the challenge samples was completed, all the sampling systems were
purged for 48 hours with zero air that had been humidified to 80% relative humidity in preparation for the
system blank sample collections. The purpose of the system blank samples is to assess sampler cleanliness as
a further gauge of additive bias. It is not a test of sample carryover. The acceptance criterion for sampling
system cleanliness, as described in Method TO-14,1 is less than 0.2 ppbv of any target compound analytes.
After the 48-hour purge, each of Group A, Group B, and Group C sampling systems collected a system
blank sample. The system blank samples were analyzed by multiple detector gas chromatography.
Ninety-one percent of the Group A systems and 75% of the Group C systems were able to meet
the acceptance criterion after 48 hours of purge, while only 32% of the Group B systems were able to meet
the criterion. This is not to say that the Group A and Group C systems are inherently cleaner, but rather
that they can be cleaned faster. It is theorized, that this is due to the ability of the Group A and Group C
systems to operate at a higher flow rate, typically 3 to 5 L/minute. The highest flowrate for the Group B
systems is dictated by the range of the mass flow controller being utilized, typically 20 to 100 mL/minute.
Invariably, all the sampling systems were able to meet the cleanliness criterion with additional purging. The
maximum additional purge time required was 48 hours.
An area that has been lacking in most canister sampling network programs is a method of
performing some form of certification in the field, using humidified standards and humidified zero air. This
in-the-field certification is recommended to be performed on a regular schedule throughout the period of
742
-------
performance of the sampling effort and would, in fact, serve as a performance audit. Regularly administered
audits would give insight into compound recovery variability and sample carryover. The U.S. EPA is
developing a trial in-the-field canister sampling system audit procedure that will be evaluated during the 1990
Urban Air Toxics Monitoring Program.2
SAMPLING SYSTEM TIME-INTEGRATION CHARACTERI7.ATION
Ftocedure
A time-integrated canister sample is defined as a sample that is collected at a constant flowrate
throughout the sample collection period. If the collection flowrate varies during the collection period, the
siimple could be biased or not representative of the ambient air sampled, assuming the concentration of
compounds present in the ambient air changes throughout the sampling period.
To perform time-integration characterizations, a system incorporating a calibrated mass flow
meter and a strip chart recorder was used to measure any flowrate deviations that occurred during varied
sampling conditions. The sampling system being characterized pulled humidified zero air through the
cilibrated mass flow meter and delivered it to the sample canister at a preset collection flowrate. The
electronic output from the mass flow meter was recorded by th; strip chart recorder, and a real-time
assessment of flowrate variation was made. An absolute pressure gauge was plumbed into the outlet transfer
line between the sampling system and the sample canister. This allowed for the real-time assessment of
pressure change within the sample canister in relation to the sample flowrate.
Results
Group A, Group B, and Group C sampling systems were setup to collect 24-hour samples of
humidified zero air, through the time-integration assessment apparatus. Collection flowrates of 3.6 mL/min
for the Group A systems and 33 mL/min for the Group B systems were used. These collection flowrates
were determined to yield time-integrated samples with negative final sample pressures in evacuated six-Liter
SUMMA*-treated canisters. The initial average measured pressure in the 6-L canisters was 29.92 inches Hg
vacuum.
A collection flowrate of 5.6 mL/min was used for the Group C systems. This flowrate
represents the extreme lower end of the range of flow attainable with the Veriflow Vacuum Regulators.
This collection flowrate was determined to yield a time-integrated sample with a negative final sample
pressure in evacuated 15L SUMMA'-treated canisters. The initial average measured pressure in the 15L
(Ministers was 29.78 inches Hg vacuum.
The Group A systems displayed very good time-integration characteristics during the 24-hour,
negative final sample pressure collections. They maintained a constant flowrate as long as the sample
canister pressure remained negative. Group A systems set for EOI average sample flowrate of 3.6 mL/min
maintained this flowrate during the 24-hour test period with no measurable flowrate change. The final
measured sample pressure averaged 3.73 inches Hg vacuum. This represents approximately 87% of the 6-L
canister volume replaced with sample.
The Group B systems evaluated displayed very good time-integration characteristics during the
24-hour, negative final sample pressure collections. Group B systems set for an average collection flowrate of
3.3 mL/min maintained this flowrate during the 24-hour test period with no measurable flowrate change.
The final measured sample pressure averaged 6.1 inches Hg vacuum. This represents 79% of the canister
volume replaced with sample.
A pressure differential across mass flow controllers is required to enable them to maintain a set
flowrate. The pressure difference required varies from model to model and also unit to unit within the same
model type. The average required pressure difference observed during the Group B evaluation was
1.8 inches Hg vacuum. Mass flow controllers intended for use in a vacuum mode should receive a negative
743
-------
pressure calibration instead of the positive pressure calibration that is normally standard from most
manufacturers.
The Group C systems evaluated displayed good time-integration characteristics during the 24-
hour, negative final sample pressure collections. Group C systems set for an average collection flowrate of
5.6 mL/min., maintained this flowrate during the 24-hour test period with no measurable flowrate change.
The final measured sample pressure averaged 12.1 inches Hg vacuum. This represents 54% of the canister
volume replaced with sample.
A pressure differential across vacuum regulators is required to enable them to maintain a set
flowrate. The pressure difference required varies from model to model and also unit to unit within the same
model. The average required pressure difference observed during the Group C evaluation was
12.1 inches Hg vacuum.
Once an individual sampling systems time-integration characteristics are determined, they can be
used as a quality control measure. Any specific sampling system operated at the same collection flowrate
repeatedly should yield the same approximate final sample pressure repeatedly. Any deviation in final
sample pressure greater than 15% is an indication that something abnormal occurred which could impact
sample validity.
SAMPLING SYSTEM RELIABILITY
In general, both Group A and Group B sampling systems operate reliably. Since 1984, Radian
Corporation has collected or overseen the collection of approximately 13,609 samples using 68 Group A
systems at approximately 179 sites across the country. Valid sample collections using the Group A systems
occurred at an average rate of 93.4 percent. Radian Corporation has also collected or overseen the collection
of approximately 1116 samples using 27 Group B systems at approximately 40 sites across the country. Valid
sample collections using the Group B systems occurred at an average rate of 95.0 percent. The reliability of
Group C systems was not assessed because they are normally activated and deactivated manually.
The sampling systems have proven to be mechanically durable. Procedural misoperation of the
sampling systems by the site operators account for the bulk of failed sample collections. Mechanical failures
account for only one-third of the 5.8% of failed collections. The majority of the mechanical problems
involved improper activation or deactivation of the latching solenoid valves, pump failures, or system leaks.
Leaks are considered to be a mechanical problem because they are usually caused by the vibration set up by
the sampling and/or slipstream by-pass pump incorporated into systems. Improper operation of the sampling
systems by site operators accounted for the remaining two-thirds of unsuccessful sample collections. The
primary procedural problems causing unsuccessful sample collection were improper programming of the
sampling system activation/deactivation timing device and failure of the operator to open or close the sample
canister bellows valve either prior to or after sample collection.
CONCLUSIONS
When operated properly, Group A, Group B and Group C canister based sampling systems offer
a reliable method for the collect of representative whole ambient air samples containing volatile organic
compounds. They exhibit good compound recovery characteristics for the compounds tested. They are
acceptably nonbiasing. The area of sample carryover has not been researched in depth. More data to
quantify sample carryover needs to be obtained. The sampling systems will perform time-integrated sample
collections when operated within sampler specific parameters. Sampler specific time-integration
characteristics can be used as a quality control measure to gauge sample validity. Successful sample
collection using the sampling systems described averaged 94.2%, with most of the failures attributed to
improper operation by the system operators. The Group C systems offer a viable sampling approach in
applications where sources of electric power are not available or to accommodate cost considerations.
There was no apparent overwhelming advantages in reliability, certifibility or time-integration
characteristics between Group A and Group B systems. The amount of vacuum remaining in a sample
744
-------
canister at the end of a Group C collection and the inability of Group C systems to use 6L canisters is a
disadvantage. Although not specifically addressed in this paper, there is also no clear advantage in initial cost
or cost to operate between pumped and mass flow controlled sampling systems.
ACKNOWLEDGEMENTS
The authors wish to thank the U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, and the Quality Assurance Division of the Atmospheric Research and Exposure
Assessment Laboratory for their support. The authors also wish to thank E. G. Bowles, W. H. Moore, and
.1. Rice for their analytical expertise and assistance.
REFERENCES
L Winbcrry, W. T., Jr., N. T. Murphy, and R. M. Rsggan, "Compendium of Methods for the
Determination of Toxic Organic Compounds in Ambient Air". U.S. Environmental Protection
Agency, Atmospheric Research and Exposure Assessment Laboratory, Research Triangle Park,
NC, 27711. EPA/600/4-89/017.
2. McElroy, F. F., V. L. Thompson, "Draft-Procedure for Field Audit of the Urban Air Toxics
Monitoring Program (UATMP)", U.S. Environmental Protection Agency, Research Triangle
Park, NC 27711.
745
-------
TABLE I. SAMPLING SYSTEM CHALLENGE RESULTS
Group A Svstems
Compound
Benzene
Methylene chloride
Toluene
Vinyl chloride
o-Xylene
Mean
X
Concentration
(ppbv)
12,7
13.4
12.4
12.9
_LU
13.0
Cases
22
22
22
22
22
X %
Recovery
104.0
101.5
108.7
104.8
94.8
102.8
Standard
Deviation
3.7
2.6
7.7
5.4
.ii
4.5
Cases
34
34
34
34
34
Group B Svstems
X %
Recovery
94.5
91.2
91.5
80.2
100.9
91.66
Standard
Deviation
5.3
18.6
7.7
22.1
2.1
11.2
Group C Systems
Cases
4
4
4
4
4
X %
Recovery
91.7
102.7
90.4
98.1
108.0
98.2
Standard
Deviation
12.9
1.8
14.0
9.1
_L5,9
10.7
-------
MOBILE AMBIENT AIR SAMPLING AND ANALYSIS
EXPERIENCE OF THE TEXAS AIR CONTROL BOARD
The Staff of the Texas Air Control Board
Presentation by Mr. James L. Lindgren
Sampling and Analysis Division
6330 Highway 290 East
Austin, Texas
The content of this presentation will summarize the Texas Air Control
Board's (TACB) past experience in sampling .ind analyzing the ambient air for
organic compounds in the vicinity of chemical and petrochemical complexes.
The basis for and the philosophies of the sampling conducted, area and focused,
will be discussed.
The techniques and methods used for collecting and analyzing ambient air
samples will be discussed. The presentation will include "real-time" methods and
those which use a variety of solid absorbents. The TACB's thermal desorption
experience using carbon molecular sieves will be emphasized.
The process, from conception to final results of conducting a week-long
and intensive mobile sampling effort, will be presented in detail. This
presentation will include duplicate and audit results of the techniques used by
the TACB in ambient air analyses.
747
-------
Introduction
The Texas Air Control Board (TACB) is the state regulatory agency whose
legislative mandate is to protect the ambient air resources of the State of Texas.
The TACB operates under the Texas Clean Air Act. TACB rules and regulations are
aimed at maintaining an ambient air environment that is not a detriment to the
welfare of the general public. Monitoring the ambient air is one of the ways
that the TACB can provide the general public with information concerning the
quality of the ambient air environment to which they are exposed. Other means
of controlling the ambient air are through the permitting process and regular
in-plant inspections conducted by TACB staff. TACB regional investigators
routinely collect ambient air samples in areas of concern to them and in
response to citizen concerns and complaints. These samples are then analyzed
by the central office laboratory. Such sampling and analysis often generates
information which will result in a request for intensified ambient air sampling
in an area of concern. Upon management approval, the Sampling and Analysis
Division of the TACB will plan, coordinate and conduct five days of intensified
sampling and analysis for potential organic ambient air pollutants using its
mobile analytical capabilities. It is the experiences of the mobile capability that
this document will describe.
Initial Efforts (305)
The TACB began mobile sampling in the late 1970's. The early efforts were
conducted using a van outfitted with one or more gas chromatographs (GC)
equipped with flame ionization detectors and selective detectors. In-the-field
electrical power was provided by a gasoline-powered electrical generator. On
occasion, sampling was conducted using instrumentation to detect sulfur dioxide
and hydrogen sulfide. The sampling done with these^ capabilities were point
source oriented. Generally, only one or two compounds were sought and
selective GC detectors and retention times were depended upon for identification.
These early efforts generated the ideas that led to the mobile laboratory
capabilities that the TACB now possesses.
Mobile Laboratory Endeavors
A decision to initiate a mobile laboratory trip is usually based on
information supplied by regional investigators or upon ambient air modeling of
a point source by TACB engineers. The Sampling and Analysis Division, in
consultation with Regional, Enforcement, Permitting and Health Effects
personnel, select the sampling sites and compounds to be targeted during the
sampling trip. Early mobile laboratory trips could be described as those which
involved "area sampling." In these instances, sampling was conducted around
chemical complexes and in the adjacent neighborhoods with a "what's there and
how much" approach. Financial and manpower resources have directed our
mobile laboratory trips to a "focused sampling" concept. In the focused sampling
concept, emission inventories are reviewed, individual compounds are
prioritized based on their potential to cause adverse health effects and the
sampling and analytical effort is designed to detect and quantify a target list of
compounds.
Mobile Laboratory. The TACB mobile laboratory is a 40-foot trailer that
has been outfitted to accept instrumentation necessary to accomplish the
proposed sampling and analytical effort in an area of concern. A gas
chromatograph/ion trap detector (GC/ITD) is permanently mounted in the
trailer. A fume hood and a glove box were initially installed because the type of
sampling media being used required chemical desorption of the adsorbed air
contaminants.
748
-------
Historically, during these endeavors of intensified sampling, the samples
were collected on gram quantities of activatec purified charcoal using modified
high-volume air samplers. Occasionally, other media would be placed after
(before) the charcoal in order to produce a specialized collection technique.
Generally, these samples were collected over a ten to 12 hour period. For short-
time samples of one to three hours, OSHA-type charcoal tubes were employed.
Methanol, methylene chloride, carbon disuliide and/or tetrahydrofuran were
used as solvents to release the air contaminants from the adsorbing media. Thus,
the fume hood and glove box were welcome accessories to the mobile laboratory.
The majority of the mobile sampling endeavors undertaken by the TACB
have been near the coastline of Texas where relative humidities are frequently
high. Experienced samplers and analysts can appreciate the problems that were
encountered using activated charcoal as an adsorbing media. Moisture can
preclude the adsorbance of air contaminants on activated charcoal. Moisture can
also impede the chemical desorption of adsorbed contaminants by not being
soluble in the desorbing chemical solvent. TACB analysts have experienced
conditions where it was necessary to chemically desorb the activated charcoal
with two solvents, analyze both extractions by GC/ITD, and sum the results
without being confident that the adsorbed air contaminants had been efficiently
recovered, due to the effects high humidity can produce.
Current Procedures
For a year and one-half, the TACB has been using carbon molecular sieve
(CMS) tubes to collect ambient air samples. Subsequent analysis involves
thermal desorption of the collected contaminants onto a sub-ambient fused silica
capilliary column.
Collection of the samples is accomplished using an SKC Model 224-PCXR7
pump. This pump has the capability of continuous or intermittent operation, and
can be programmed to sample a set volume of ambient air over a time period of
up to one week. The adsorbing media is Supelco's carbotrap 300 tube, which
consists of carbotrap C/Carbotrap/Carbosieve III. This adsorbent is purported to
work well with U.S. Environmental Protection Agency TO-1, TO-2 and TO-3
methodology. Our experience has shown that retaining 1,2-butadiene is a
problem at parts per billion (ppb) concentrations. Fifty percent of the 1,3-
butadiene will "self desorb" in the first 24 hours and is not detectable after 48
hours of storage when working with ppb concentrations. In addition, attempts to
use the CMS tube to collect sulfur-containing compounds has not been
productive.
Thermal desorption is accomplished using a Dynatherm Model 850 tube
desorber. Chromatography is usually accomplished using a DB-1 or DB-5 column,
30 meters in length, contained in a Varian 3400 GC. Identification and
quantitation of the individual eluting compounds is accomplished with a
Finnigan Model 800 ITD.
Typically, the thermal desorption is done at 325°C. The analytical column
is held at -25°C for four minutes. It is then temperature-programmed at
50°C/minute to 50°C and then at 10°C/minute to 250°C. Helium is used as the
carrier gas, flowing through the CMS sample tube at seven cc/minute. Two cc or
less of this total flow is split to the analytical column for analysis. Internal
diameter and length of the analytical column will determine the upper limit of
flow to the analytical column. Sampling only four liters of ambient air provides
sufficient sample for the analytical scheme described here. With the ITD in full
scan mode, 500 pptv of most of the compounds of interest are routinely identified
and quantified.
749
-------
Those of you who are familiar with the thermal desorption process have
probably already noted several things in this analytical scheme. First, there is
no re-focusing of the sample onto a second adsorber bed and there is no
cryofocusing device. The whole column is held sub-ambient. Admittedly, the
front end chromatography does not develop sharp peaks for the earlier eluting
compounds. Fortunately, to date there have been few compounds of interest to
the TACB eluting in that portion of the chromatogram. Secondly, there is no
flushing of the CMS sample tube to remove potentially collected water. The CMS
tubes are advertised to be hydrophobic. If water is retained by the CMS tubes, the
quantity is such that there has been no observed adverse effects in the operation
of the ITD.
When in the field, the mobile laboratory serves as the base of operations
and support for the other sampling vans. Sampling is conducted 24 hours a day
by two 12-hour shifts. Sampling begins on a Saturday evening and continues
until the following Friday morning. During this time frame, 90-100 CMS tubes
will be used to collect three-hour composite samples. In addition, approximately
1200 real-time samples will be analyzed. The analysis of the CMS tubes is
accomplished using the mobile laboratory ITD and a second ITD temporarily
mounted in a sampling van.
Typical field operations have the mobile laboratory with its ITD and a real-
time GC. Wind speed and direction are monitored at the mobile laboratory. A van
with one or two GC's and the second GC/ITD travel to appropriate sampling sites
for the collection of composite samples and real-time samples in addition to
analyzing CMS tubes by GC/ITD. A second van will be outfitted with one or two
GC's, as appropriate, and travels to sampling sites to collect real-time samples and
composite samples on CMS tubes. A third van is deployed to collect only
composite samples.
Real-time sampling provided an interesting observation during a recent
sampling trip. Winds had been southeasterly for two to three days from a source
of styrene. Due to a cold front, the wind changed 180 degrees. The observed level
of styrene from real-time sampling remained constant for eight-ten hours.
During the next four-six hours, the observed styrene concentrations went to a
non-detectable level. This would infer that under the proper meteorological
conditions that "a cloud of pollutant" could move in one direction, only to return
with a reversal of wind direction.
The focused approach type of sampling attempts to collect the highest
concentration that a point source is emitting to the ambient air. Location of the
samplers in relation to the emission source under a given set of meteorological
conditions is very important, especially when the compound being sought has
little or no odor at sub-ppm levels. The importance of sampler location was
observed during a sampling exercise for 1,2-dichloroethane. Three-hour
composite samples were being collected on CMS tubes. Under the meteorological
conditions existing during that sampling period, one sample resulted in a three-
hour average concentration of 24 ppbv, while the second sample, collected less
than 150 yards away, resulted in a three-hour average concentration of 142
ppbv.
Quality assurance is accomplished by using NIST-traceable standards. To
determine each instrument's response at the beginning of each shift, the GC's
are calibrated with the appropriate gases. The ITD's are tuned using FC-43 as the
reference compound. In addition, a CMS tube is loaded with the appropriate
gaseous standards for analysis by each GC/ITD. One set of duplicate CMS tubes is
collected each shift. This duplicate collection is rotated to a different van each
750
-------
shift. Each shift, an audit CMS tube is prepared for each GC/ITD operator.
Figure 1 depicts the results which can be achieved by this analytical scheme if
attention is paid to detail.
The report subsequent to the sampling and analytical activity tabulates
the results. The real-time and the three-hour CMS composite results are
formatted in a manner such that the sampling location and meteorological data
are also available to the report reviewer. The final results are reviewed by
Health Effects personnel to determine if any of the reported concentrations
could be detrimental to the general public. If substantial exceedances of the
health effects levels are observed, the appropriate compliance and/or
enforcement action is initiated.
Acknowledgements
I want to take this opportunity to acknowledge the cooperation and
support of the following TACB groups:
Logistics Section, Ambient Monitoring Division
Sampling and Analysis Division
Health Effects Division
Technical Services Division
Compliance and Enforcement Divisions
Regional investigators and staff
751
-------
AUDITS OF
COMPOSITE SAMPLES
(ppbv)
Sample
Identification
Audit 2
Audit 3
Audit 4
Audit 5
Audit 6
Audit 7
Audit 8
Benzene
True
10
10
10
10
10
10
10
Found
14
10
14
14
10
11
12
DUPLICATES OF
COMPOSITE SAMPLES
(ppbv)
Sample
Identification
305-1A
305-1AD
650-2A
650-2AD
940-3A
940-3AD
305-4A
305-4AD
650-5B
650-5BD
305-7A
305-7AD
650-8A
650-8AD
Benzene
5
5
1
1
5
6
22
23
3
4
80
130
10
10
Styrene
trace
trace
10
10
90
100
20
20
50
50
380
400
170
160
Figure 1. Results of Audit and Duplicate Samples
752
-------
THE ESTABLISHMENT AND OPERATION OF AN NMOC
AND ALDEHYDE MONITORING PROGRAM --
EXPERIENCE OF A STATE AGENCY
Julian D. Chazin, Mark Allen and John Hillery
Wisconsin Department of Natural Resources (WDNR)
Madison, Wisconsin 53707
ABSTRACT
To meet the requirements of U.S. EPA's Po:;t-1987 ozone SIP strategy for
southeastern Wisconsin, WDNR undertook ar. NMOC and aldehyde monitoring
program. Rather than invest in participation in the U.S. EPA national
NMOC monitoring program, Wisconsin applied the funding to purchase of
the equipment necessary to conduct the sampling and analysis on its own.
In addition to the required NMOC monitoring, WDNR had a need for
speciation as well as to conduct aerial sampling. A description of the
equipment and procedures as well as the problems encountered and the
solutions found over a three year period are presented. The quality
assurance program which was devised is also presented. Special studies
undertaken to deal with issues being addressed nationally are reported,
i.e., pressurized vs. non-pressurized canisters, 7-day vs. weekday
sampling, and interferences in the methods. Results are presented on
hydrocarbon speciation, their application to chemical kinetic mechanism
parameters for photochemical grid models .is well as the use of aldehyde
monitoring data for such mechanism parameters. Data summaries are
presented for NMOCs and compared with the national program, along with
aldehyde and hydrocarbon speciation summaries. Finally, the results of
the quality assurance measures, i.e., data completeness, internal
audits, and interlaboratory comparisons are also presented. Based on
Wisconsin's experience, it should be possible for other state or local
agencies to conduct their own NMOC and aldehyde sampling and analysis
programs.
753
-------
INTRODUCTION
Beginning in 1984, U.S. EPA began a program to measure concentrations of
nonmethane organic compounds (NMOCs) in 39 urban areas in the United
States (1), In September 1986, U.S. EPA announced an expanded effort to
obtain NMOC concentrations in many more urban areas in the United States
"in support of the ozone control program" (2). The U.S. EPA would
supply the necessary sampling equipment, to be collected in passivated
stainless steel canisters, and perform the analysis in its laboratories
at Research Triangle Park, NC, at a cost of $19,000 per site. A minimum
of two sites, one in the core, the other at the fringe of the central
business district (CDB) of each urban area were recommended to measure
ozone and the precursors, nitrogen oxides and NMOC.
Rather than participate in the EPA program, Wisconsin decided to perform
its own NMOC analysis. There were three reasons for Wisconsin's
decision: the cost for EPA's program; the need for additional NMOC data
not provided by U.S. EPA's program; and the need for speciation of the
NMOC samples.
We believed the incoming, upwind precursors (as well as transported
ozone) to the Milwaukee urbanized area to be higher than the EKMA model
assumed in its default values. This is due to upwind large urban
sources to the south and southeast of the Kenosha to Milwaukee corridor.
Therefore we decided we needed to obtain incoming precursor and ozone
concentrations from the south and southeast, over Lake Michigan. This
was accomplished through an aerial study performed in 1987. We will not
discuss that study in this paper but two reports on this study are
available upon request from the authors, one a detailed report on the
equipment, procedures, and data obtained (3) and the other an analysis
of the data (4).
Interest in speciation resulted from our desire to develop canister
techniques for air toxics and to provide the best inputs to the
OZIPP/EKMA modeling effort which was required by the post-1987 ozone SIP
strategy of U.S. EPA. We wanted to confirm that the Milwaukee species
distribution and levels were at least similar to the default levels
suggested for the EKMA modeling.
EXPERIMENTAL
Year 1 - 1987
Equipment. The equipment used was based on U.S. EPA guidance
provided in Method T012 for NMOCs by cryogenic PDFID (5), i.e.,
passivated stainless steel canisters (PSSCs), a sampling system, a
canister cleaning system, and an NMOC lab analysis system. There were
several differences which are discussed below.
Procedures. Collect ambient air samples in canisters at two urban
sites in Milwaukee, 6 to 9 a.m. CDT, seven days per week. The
collection sites were multiple parameter air monitoring stations
equipped with ozone and NOX analyzers. Canisters were returned to a
field lab in Milwaukee daily for PDFID analysis for NMOC. Sixteen
canisters were set aside for the U.S. EPA ASRL lab at Research Triangle
Park, NC for gas chromatographic analysis with an FID detector for over
100 species; ten canisters were sent to the Wisconsin State Lab of
Hygiene (WSLOH) for total NMOC by GC analysis (sum-of-species) . Collect
754
-------
a second canister sample once every 14 days at each site (once per week
at alternating sites) for determination of precision of the sampling
method.
Problems/Solutions. The initial PDriD system included a small GC
with a small oven that could not hold tie cryogenic trapping column.
Instead the trapping column was placed outside the oven and heated with
a hot water bath (6). Inside the GC ovcm a blank unpacked column (a
bypass tube) connected the gas sampling valve to the flame ionization
detector. The 6-port valve for switching gases was also outside the gas
chromatographic oven.
The EPA samplers, used in their national NMOC study, pressurized the
canisters with a Metal Bellows MB-151 pump as an air mover with an
appropriate valve. We were unable to get the Magnelatch™ valves in time
for the study and instead used air operated bellows valves (Nupro™ SS-
4BK-1C). We found that these valves worked well and that a small
cylinder (80 cu ft) of compressed nitrogen was sufficient for our summer
field study. Unlike the Magnelatch, no special circxiit is required to
prevent the valve from heating up. To reduce site visits we built
samplers able to handle 3 canisters at one time (for weekends).
The 1987 canister cleaning system followed the design in the U.S. EPA
Method T012. While the canister system worked, the clean air system was
not consistent in clean air production. The clean air system had no
holding reservoir to store clean air and the direct air flow from the
system was very slow. Canisters were, hum.ldified by direct injection of
HPLC-grade water. After the first year we evaluated the system and made
some significant changes for the 1988 study.
Year 2 - 1988
Equipment and Procedures. The sampling and analysis set-up was
similar to 1987 with the following exceptions:
1) A heated valve compartment was added to the 6-port chromatographic
valve to prevent cold spots; this improved the precision of the method.
2) Cleaning system changes were made, as follows: a) replaced the zero
air source with a commercial tank of zero monitoring air, (a final
liquid argon trap was retained); the tank supplied air of more
consistent quality and allowed faster fLll rates which improved the
cleaning system, doubling throughput from 6 canisters per day to 12 (a
150 cubic foot cylinder of zero monitoring air supplied enough gas to
clean over 80 canisters); b) humidification by direct injection of HPLC
grade water was replaced with humidificatxon of the zero air by passing
it through a high pressure water bubbler (as now recommended in TO-12);
this bubbler was constructed from a two-valve sampling canister; c) we
replaced the critical orifice for ziero ,535 control with a mass flow
controller to better regulate the gas flow and decrease the canister
fill time.
3) Aldehyde sampling was initiated. A sampler consisting of an oilless
pump (Thomas - low volume, low pressure), a timer with an elapsed time
meter, a mass flow controller ('1 ppm), two electrically operated
solenoid valves, with appropriate tubing connected to the glass sampling
manifold used for NMOC analysis. A cartridge, prepared by the WSLOH by
coating purified DNPH on a Waters Associates "SepPAK" silica gel
755
-------
cartridge, was Inserted for each sample. The timer was set to operate
between 6 to 9 a.m. each day, collecting 180 liters of air. The
cartridges were removed and sent to the Occupational Health Lab of the
WSLOH for analysis utilizing U.S. EPA Method T011 (7).
Year 3 - 1989
In 1989 the only change made was that we brought on-line a standard
laboratory GC (Hewlett Packard HP5890) in the PDFID analysis system in
place of the HNU small GC, placing the cryogenic trapping loop inside
the GC oven. This brought improved performance. Analysis time was
reduced from 9 minutes to 5 minutes. The minimum detection limit was
reduced from 50 to 20ppbC. Precision was improved for propane working
standards. In 1989, 79% of all samples were analyzed in a minimum of
two runs with an average deviation of 5.9% for all samples. The
correlation of the new PDFID analysis system with the gas
chromatographic analysis performed by the WSLOH improved from 0.904 to
0.998. However, a 7-10% negative bias was noted for the new system
compared with the old.
Quality Assurance
As with much of the NMOC program at the time, there was little available
information about setting up a quality assurance program for this study.
Drawing on our experience with other monitoring projects, we set up a
program of our own. This included the writing of an EPA-approved
quality assurance project plan as well as standard operating procedures
for sampling and analysis. Specifics of the QA program included:
Field. 1) Co-located canisters for precision (duplicate samples);
2) pre- and post-season sampler audits; 3) co-location of sampling
systems following the sampling season.
Lab. 1) Daily calibration with working propane standards (900
ppbC) certified against a National Institute for Standard SRM; 2) use of
both propane and ambient-air control gases; 3) analysis of selected
samples for total NMOC by gas chromatography conducted at the State
Laboratory of Hygiene (sum-of-species method); and 4) interlaboratory
exchange of samples with U.S. EPA laboratories (EMSL and ASRL).
RESULTS AND DISCUSSION
Pressurized vs. Non-Pressurized Canisters
The Wisconsin program used the MB-151 pump to fill canisters to 30 psia
or greater as described in U.S. EPA Method T012. We analyzed selected
pressurized canisters, vented the canisters to atmospheric pressure and
reanalyzed the canisters. The results showed a positive bias, with the
NMOC concentration consistently higher ( average 12.3%; range <1% to
25%) for the non-pressurized canister. The result of this study does
not show that pressurized canisters are more accurate than non-
pressurized but does show that results from the two methods can not be
compared without some consideration of the bias.
7-Day vs. 5-Day Sampling
Wisconsin decided to collect 6-9 a.m. samples every day of the week
while the U.S. EPA program has remained a 5 weekday program. A
756
-------
comparison of the NMOC results for weekday vs. the weekend samples was
made. The distribution of NMOC data shows that the weekend NMOC data
are not significantly different than the weekday data. The 7-day means
vs. the 5-day means were 0.324/0.338 for the Civic Center Site and
0.454/0.510 for the North site. U.S. EPA's choice of weekday sampling
was based on the theory that the heaviest vehicular traffic occurs in
the early morning on working days in the central business district of
urban areas. While it is true that for many urban areas there is
little vehicular traffic in the central business district on weekends,
that is not necessarily the case for Milwaukee, On weekends leisure
time activities at the lake shore (which is adjacent to the central
business district) attracts very large numbers of people (with their
vehicles). This is confirmed by the results of our weekday vs. weekend
comparison.
Other Studies (of Potential Interferences)
Fluorocarbon dusting spray. In 1988 an intermittent contamination
problem developed with one brand of canisters. The problem was
eventually traced to a fluorocarbon dusting spray used on the canister
valves prior to analysis. The fluorocarbons became trapped in a small
dead space on the valve and would leak out during analysis. We
discontinued use of any dusting spray ani valves are now cleared by a
momentary venting of canister contents.
Aldehyde/Ozone Interactions, Recently Arnts and Tejada (8)
reported ozone interference in the DNPR cartridge method for aldehydes.
He reported that placement of a potassium iodide coated copper tubing
inlet in front of the cartridge appears to eliminate the interference.
We are currently experimenting with that procedures in advance of our
1990 program.
Hydrocarbon Species
As stated in the introduction, we were Interested in determining the
hydrocarbon species present in our NMOC canister samples. Time and cost
consideration made it economically unfeasible to quantitatively analyze
for all compounds in all air samples. During the first year we worked
to develop a list of 37 surrogate compounds to be identified and
quantified in selected air samples based on data from a number of
literature sources. Using data from the 1987 and 1988 field studies,
the list of surrogate compounds was expanded to 50 compounds for 1989.
The. new surrogate list resulted in a reduction of the unidentified
compounds from an average of 32.5% (3.2 to 79.5%) to an average of 19.6%
(6.5 to 33.9%). Gas chromatographic analysis of air samples yielded
concentrations that would be condensed in seven modeling parameters used
for the CB-IV mechanism (9).
A comparison of the computed modeling parameters determined using the 37
surrogate compounds versus an extensive GC analysis provided by U.S. EPA
(for over 100 species) (10) did show a difference in the modeling
parameters calculated. For twelve samples, four of the modelling
parameters - PAR, TOL, XYL, ALD2 - had percent differences greater than
-25% and one modeling parameter - OLE - had a percent difference greater
than -50%.
757
-------
Aldehydes
While the CB-IV mechanism provides for the calculation of a surrogate
aldehyde fraction from the hydrocarbon profile data the accuracy of that
fraction is questionable. When modeling parameters were calculated from
the sampling data the FORM and ALD2 fraction were found to be much less
than the default values in the EKMA model. The U.S. EPA default value
for FORM is 0.021 and for ALD2 it is 0.052. Reduction of our data
yielded FORM fractions of 0.000 in 1987 and 0.002 in 1988 and ALD2
fractions of 0.008 in 1987 and 0.010 in 1988. We attempted to compare
fractions from the aldehyde analysis utilizing Method T011 with that
calculated from the NMOC species. The comparison based on limited
amount of data indicates the actual aldehydes were greater by two to
eight times the computed values, closer to the default values. This
comparison has raised concern and convinced us that the estimation of
the aldehyde modeling parameters from NMOCs is inaccurate and that
aldehyde sampling and analysis is required for accurate CB-IV
computations.
Data Summary
NMOG Summary. A summary of Wisconsin NMOC data results is available
upon request. As with the data reported by U.S. EPA (11) the NMOC
concentrations are lognormally distributed. The geometric mean suggests
that NMOC decreased at both sites from 1987 to 1988. A decrease is seen
at the Civic Center site in 1989 but the North site changed little.
NMOC Comparison. In 1987 the Wisconsin site mean was less than
that reported for the U.S. EPA national study (12). Comparing the
summary data, the UWCC site was 24/34 ranked by mean (21/34 by median)
and the UWN site was 26/34 by mean (29/34 by median). In 1988 the
Wisconsin sites mean (0.365 ppmC) was much less than that reported by
U.S. EPA (0.636 ppmC). The UWCC site would rank 33/40 (34/40 median)
sites and the UWN site would rank 37/40 (36/40 by median) sites by mean.
Aldehyde results. We collected three-hour aldehyde samples on
forecast high ozone days. In 1988 for 37 samples, the mean formaldehyde
was 6.3 ppb and the mean acetaldehyde was 4.2 ppb; in 1989 for 26
samples, the mean formaldehyde was 3.7 ppb and the mean acetaldehyde was
2.2. ppb. A complete data summary is available upon request.
Speciation Results. A complete list of data from the GC analysis
is available upon request. To summarize the 1988 data, the WSLOH
analyzed 37 canisters for 39 surrogate compounds and the U.S. EPA - ASRL
analyzed 11 of the 37 canister samples for over 100 compounds. Of the
50 most common compounds found by ASRL, 29 were paraffins, 7 were
olefins, and 14 were aromatic. The WSLOH found isobutane was the most
common paraffin with a mean concentration of 33 ppbC. Acetylene and
ethylene were the most common olefins and together with ethane had a
mean concentration of 26 ppbC. Toluene was the most common aromatic
with a mean concentration of 25 ppbC. ASRL found similar
concentrations.
Quality Assurance
Data Completeness. The Wisconsin NMOC monitoring plan called for
seven day sampling during the period from Memorial Day until Labor Day.
Data completeness for 1987 was 92%, for 1988 was 96%, and for 1989 was
758
-------
92%. All invalid samples were the results: of sampling problems, such as
incorrectly set timer or canisters improperly attached to the sampler.
This compares with the U.S. EPA reported a completeness of 95.0% in 1987
and 93.4% in 1988 for the national study (11,12).
Internal Audits. Both pre-season and post-season audits were
performed in 1988 and 1989. These audit:; were devised in consultation
with our Quality Assurance Coordinator. Clean canisters were connected
to the samplers at the sampling site. Using a modified CSI 1700 gas
phase titration unit, audit gases were prepared and connected to the
sampler for a short sampling period. The canisters were removed and
returned to the. laboratory for analysis and the results were compared to
the expected concentrations. Four NMOG concentrations, including a
blank, were collected at each sampler. Audit goals of +25% for accuracy
were met for most of the audits samples. There were some failures,
believed to be due to auditing errors more than to system failures.
Problems include incorrectly blending gases and carryover of higher
concentration NMOCs.
Interlaboratory comparison. We exchanged NMOC samples with two
U.S. EPA laboratories located at Research Triangle Park, the EMSL-QA lab
and the ASRL lab. The EMSL lab testing was a direct comparison of PDFID
methods and agreement was quite good: correlation coefficients varied
from 0.936 to 0.998 over three years. The ASRL comparisons, with sum-
of-species, were less favorable. Lonneman (22) reported a "moderately
high positive bias" for the PDFID compared with his GC sum-of-species
method. He attributed this to oxygenated hydrocarbon species, which
showed up on the chromatogram as broad tailing peaks. Those broad peaks
are not integrated by the GC method but are incorporated in the NMOC
peak detected by the PDFID method. By converting from the hot water
bath to the GC oven for the PDFID method, we found better agreement with
the WSLOH sum-of- species data for 1989, believed to be due to better
resolution of the oxygenated species in the PDFID peak integration. We
have not as yet received the 1989 speciation data from the ASRL lab at
U.S. EPA.
CONCLUSION
Based on Wisconsin's experience, it should be possible for other state
or local agencies (or private consultants) to successfully establish,
operate and maintain NMOC and aldehyde sampling and analysis programs.
Techniques of quality assurance should be adapted such as: round-robin
testing; blind audit samples; precision (replicate sampling) and
accuracy (analysis of standards); and external audits. Aldehyde
sampling and analysis is also feasible so long as the ozone interference
as reported by Arnts-Tejada (8) is removed (utilization of a potassium
coated copper tubing ozone cutter). Speciation of hydrocarbon species
is also possible but more difficult and expensive. A major pitfall in
the NMOC analysis is possible interference by oxygenated species. The
selection of appropriate surrogate species is also a consideration.
We recommend that U.S. EPA, through the Quality Assurance Office,
formalize the interlaboratory testing program for NMOC analysis by
PDFID, as well as for aldehydes, as more states develop their own
programs. In addition, goals for precision and accuracy should be
developed by U.S. EPA and audit gases mace available for state use.
759
-------
ACKNOWLEDGEMENTS
V.L. Thompson and H.L, Richter of the QAD, U.S. EPA, assisted with the
PDFID program. W.A. Lonneinan of ASRL, U.S. EPA, assistance with GC
analysis and HC speciation, The Organic Chemistry Unit, Environmental
Sciences Section of the WSLOH for speciation analysis. Finally, the
devotion and skill of Penny Kanable in preparing this paper is
appreciated.
REFERENCES
1. R.L. Seila, W.A. Lonneman, "Determination of Ambient Air
Hydrocarbons in 39 U.S. Cities", APCA, Preprint #88-150.8, 81st
Annual Meeting, Dallas, Texas, June 1988.
2. R.G. Rhoads, U.S. EPA, OAQPS, Research Triangle Park, NC, private
communication, September 30, 1986.
3. J.F. Hillery, The 198JLC)zgne_Aerial Survey. Wisconsin Department of
Natural Resources, Box 7921, Madison, WI 53707, Publ, #AM-030,
August 1988.
4. Wisconsin Department of Natural Resources, 1987 Ozone Study -
Southeastern Wisconsin. Box 7921, Madison, WI 53707, Publ. #AM-028,
July 1988.
5. U.S. EPA, Compendium of Methods for the Determination of Toxic
Organic Compounds in Ambient Air. Publ. #600/4-84-041; Method T01.2,
"NNMOC by Cryogenic PDFID Method", Revision 1, June 1987.
6. R.D. Cox, M.A. McDevitt, et.al., "Determination of low levels of
total norunethane hydrocarbon content in ambient air", Envir. Sci.
Techno1.. 15:57, 1982.
7. U.S. EPA, Compendium of Methods for the Determination of Toxic
Organic Compounds in Ambient Air. Publ. #600/4-84-041; Method T011,
"Method for the determination of formaldehyde in ambient air using
adsorbent cartridge followed by high performance liquid
chromatography", Revision 1.0, June 1987.
8. R.B. Arnts and S.B, Tejada, "2,4-DNPH - coated silica gel cartridge
method for determination of formaldehyde in air: identification of
ozone interference", Env. Sci. Techol.. 23:1428, 1989.
9. K. Baugues, Tech. Support Div., OAQPS, U.S. EPA, Research Triangle
Park, NC, private communication, 1987.
10. W.A. Lonneman, Atmospheric Sciences Research Lab., Research Triangle
Park, NC 27711, private communication, January 1988.
11. U.S. EPA, OAQPS, 1988 Nonmethane OrganicCompound Monitoring
Program. Final Report. Publ. #450/4-89-003, December 1988.
12. H.G. Richter, "Summary report, NMOC-87, NMOC collection and analysis
for state and local agencies", U.S. EPA, QAD, Research Triangle
Park, NC, September 1988.
760
-------
DISCREPANCIES IN AMBIENT NON-METHANE HYDROCARBON
MEASUREMENTS AMONG VARIOUS METHODS
Joel Craig
ABB Environmental Services, Inc.
4765 Calle Quetzal
Camarillo, CA 93010
James Balders, Jan Clover, and Jim McElroy
Monitoring and Technical Services
Ventura County Air Pollution Control District
800 South Victoria Avenue
Ventura, CA 93009
Previous studies have acknowledged a discrepancy in measurement of
ambient Non-Methane Hydrocarbons (NMHC) between continuous analyzers and
manual methods such as the Pre-Concentration Direct Flame lonization
Detection (PDFID) and "sum of species" obtained by gas chromatography.
The Ventura County Air Pollution Control District has been making
comparison measurements between Combustion Engineering 8202A continuous
NMHC analyzers and the PDFID technique since 1987, and additional
comparison with C2-C10 speciation during 1989. Additionally, the EPA's
"Non-Methane Organic Compound" Program made PDFID measurements at twelve
locations in California where Combustion Engineering 8202A analyzers
were making simultaneous measurements. These data as well as the data
from other tests are presented and used to identify some of the factors
that influence this difference in measured NMHC.
These studies have significant iirplications in determining a
representative method for ambient hydrocarbon measurement.
761
-------
Introduction
For some time persons involved vith ambient hydrocarbon measurement
have recognized discrepancies between various non-methane hydrocarbon
monitoring methods. Data from continuous non-methane hydrocarbon (NMHC)
analyzers have not only been shown to compare poorly with manual methods
such as the Pre-Concentration Direct Flame lonization Detection (PDFID)
method. Also, two identical continuous analyzers produce quite
different data when sampling side by side.
Results of comparisons between Combustion Engineering 8202A
continuous analyzers and three hour integrated samples analyzed by the
PDFID method were presented at the 1988 International Symposium on
Measurement of Toxic and Related Pollutants. That study demonstrated
the poor agreement between the two methods, analysis of the data showed
large sample to sample variation between the methods as well as a
distinct difference between the two sites studied.
These studies initiated by the Ventura County Air Pollution
Control District in 1987 was continued during 1988 and 1989, providing
many more comparisons. Additional data was obtained allowing
comparisons between PDFID measurements made by EPA's Non-Methane Organic
Compound Monitoring Program and data generated by the California Air
Resources Board and various California local air pollution districts
using the Combustion Engineering 8202A. An intensive comparative study,
called "the El Rio Shootout", was performed by Ventura County in 1989 to
try to answer some of the questions concerning the observed site to site
differences as well as the sample to sample variations at one site.
This paper reports the findings of these studies.
Experimental Methods
Ventura County A.P.C.D. Comparisons
The PDFID sampling and analytical procedures and equipment as well
as the operation of the Combustion Engineering 8202A's followed the
procedures outlined in Reference 1. In 1989 the PDFID analytical system
was modified by the addition of a Nutec automated cryogenic trap and
other controlling hardware to allow for unattended automated analysis.
Following 'this modification, inter-lab comparisons were performed with
E.P.A.'s Quality Assurance Division as well as California Air Resources
Board and Bay Area Air Quality Management District labs in order to
confirm the continued accuracy of the Ventura County automated
analytical system. Comparative measurements were made at Ventura County
APCD's El Rio and Simi Valley monitoring sites.
EPA PDFID/CARB 8202A Comparisons
The procedures used in EPA's Non-Methane Organic Compound
monitoring program for PDFID measurements are outlined in Reference 2.
The 8202A measurements obtained from the California Air Resources Board
(CARD) data bank were made by both CARB operated and local district
operated 8202A's. The procedures used in operation these 8202A's follow
either the CARB Quality Assurance Manual or the respective local
district Quality Assurance Manual. It was assumed that these procedures
were all based on the CARB operating procedures and similar.
762
-------
The El Rio Shootout
During the winter of 1989 an intensive comparative study vas
performed at Ventura County APCD's El Rio Monitoring site. In this
study, the 8202A used at the Simi Valley site for comparisons since 1988
was brought to the El Rio site and co-located with the existing El Rio
8202A which had been in place at El Rio since 1987. Three hour
integrated samples were collected and analyzed by both PDFID and C2-C12
speciation by capillary Gas Chromatography for comparison to the two
8202A data sets. The speciation analysis followed the procedure
outlined in reference 3. During part of the study collocated integrated
samples were collected and analyzed by PD'fID to allow for estimation of
PDFID sampling precision.
Results
Ventura County APCD Comparisons
The El Rio and Simi Valley comparisons made in 1988 and 1989 are
presented in Figures 1 and 2 respectively. These comparisons show the
same pattern as was seen in the aforementioned 1987 comparisons at these
sites. At El Rio it was observed that at on September 10, 1988, when
the full scale range of the 8202A was chaaged from 0-20 ppm to 0-10 ppm
the sample to sample variability between the two methods decreased
dramatically as can be seen in Figure 3.
EPA PDFID/CARB 8202A Comparisons
Comparisons made in 1987 and 1988 are presented in Figures 4 and 5.
Comparisons made at many sites (each with different 8202A's) reveal a
significant site to site bias between :he two methods and raises a
question as to whether the source of the bias is due to the site
environment or to inherent differences between 8202A's.
El Rib Shootout
Figure 6 presents data from the two 8202A's as compared to the
parallel PDFID measurement. Note that data from the 8202A which was
relocated from the Simi Valley site compares to the parallel PDFID
measurement at El Rio just as it had when this same instrument was at
the Simi Site.
In order to see if the hydrocarbon mix was a influencing factor in
differences between the two methods, the results of the speciated
analyses were used to determine the percent light, medium and heavy
hydrocarbon species in each speciated sample. Light hydrocarbons were
defined as all species lighter than benzene. Medium was defined as all
lighter than toluene but heavier than benzene. Heavy was defined as
all species heavier than toluene. These subdivisions were used because
some people theorize that the 8202A's would have the most difficult
time detecting the heavier species. Therefore samples with higher
percentages of heavy hydrocarbons would be the samples most
underestimated by the 8202A's. Figure 7 and 8 present comparisons of
the percent heavy and light species to tie PDFID/8202A (PDFID measured
value/8202A measured value) ratio of the sample. If there was any
correlation between samples with a high percentage heavy species being
most underestimated by the 8202A, then the samples with high percentage
heavy species would have the highest PDFID/8202A ratio. This analysis
763
-------
showed no such correlation with the heavy species or the light species.
The precision of both methods was estimated by comparison of the
co-located or repeat analysis data. The 8202A precision was estimated
by comparison of the co-located 8202A data during the entire study.
PDFID analytical precision was determined by comparison of repeat PDFID
analysis during the entire study. Total PDFID precision (analytical and
sampling) was estimated by comparison of a limited number of co-located
PDFID samples. A summary of the precision estimates is presented in
Figure 9.
Conclusion
The "site specific" difference in comparisons between the
Combustion Engineering 8202A and the PDFID measurement method appear to
actually be "analyzer specific". Data from this study suggests that
these observed differences are related to some factor(s) inherent to
each individual analyzer. Furthermore, results of this study indicate
that there is no correlation between changes in the ambient hydrocarbon
mix and observed differences between the two monitoring methods.
Using the lowest possible full scale range when operating a
Combustion Engineering 8202A is extremely important. Data from this
study suggests that operating on ranges greater than 0-10 ppm full scale
results in extremely poor precision. Even when the 8202A is operated on
0-10 ppm full scale, the precision of the 8202A is much less than that
of the PDFID method.
References
1. J. Craig, "Field Comparison Study of the Combustion Engineering
8202A and Integrated Grab Sample/Preconcentration Direct Flame
lonization Detection for Ambient Measurements of Non-Methane
Hydrocarbons",1988 International Symposium on Measurement of Toxic
and Related Air Pollutants.
2. Radian Corp., "1987 NMOC and Air Toxics Monitoring Program", Vol.
1, Environmental Protection Agency (EPA-450/4-88-011), 1987.
3. R. L. Seila, E. E. Rickman, "Research Protocol Method for Analysis
of C2-C12 Hydrocarbons in Ambient Air By Gas Chromatography with
Cryogenic Concentration", U.S. Environmental Protection Agency,
1986.
764
-------
a o. l —
•- 0.649
S- -O.OJ7
•- 0.7tl
&• 0.121
r- O.tM
0.1 0.4 O.ft O.I 1 L.2 1.4 1.4
-0.1 -L rorll> 1"^
Figure 1. El Rio 1988& 1888: PDFIDvs. 8202
Figure 2. Slmi Valley 1968 & 1980: PDFIOvt.8202
(1 711/ II *1 Itt 111 121 131 141 151 1(1 171 111 191 tfc
of CoHpariionf
-II
Figures. El Rio 1988& 1980: PDFID/82O2Ratio
Figure 4. 1987 PDFIDvt. 8202: 5 Locations
Figures. 1988 PDFIDvt. 8202: 4 Locations
765
-------
f i mo S202
SMI &202
'00
IQQ -
Figure 6. El Rio Shootout:
8202/8202/POFIO
CL RI
SIU1 &SOS
€L RIO 9302
SIUI &202
Figure 8. % Light v«. PDFID/8202 Slope
0.9
o.a
0.7
0.<3
03
0.1
Figure 7. % Heavy vs. PDFID/8202 Slop*
1.7
IB
1.5
o.r
o.a
0.3
50
PDFID
dOO
*1 fi
5-1
* LIGHT
Overall Precision - 8202 Overall Precision - PDFID Analytical Precision - PDFID
51.667
46.169
19.895
16.298
9.250
6.886
7.017
4,514
7.500
7.583
2.161
2.892
Absolute Difference Mean
Standard Deviation
Percent Difference Sean
Standard Deviation
Figure 9. El Rio Shootout Precision Summery: 8202 and PDFID Methods
766
-------
MEASUREMENT OF ACID DEPOSITION COMPONENTS
IN SOUTHERN COMMERCIAL FORESTS
Robert L. Sutton and Eric R. Allen
Environmental Engineering Sciences Department
University of Florida
Gainesville, FL 32611
The role of acid deposition within rural areas of the southeastern
United States in general, and commercial pine forests in particular, may be
involved in the observed decline of forest productivity in the region. The
University of Florida has established monitoring sites at three rural loca-
tions -- Duke Forest (near Durham, NC), Austin Gary Forest (near Gaines-
ville, FL), and Stephen F. Austin Forest (near Nacogdoches, TX). This
study, which is funded by the U.S. EPA, entails monitoring within a cleared
area in conjunction with controlled artificial exposure studies of pine
seedlings by forestry researchers. Of particular interest are the relative
amounts of wet deposition components at these sites. Collection of wet
deposition samples is performed by an automated wet-dry bucket collector.
Samples are shipped to the Air Pollution program laboratories, Environmen-
tal Engineering Sciences Department, University of Florida, for detailed
analysis by ion chromatography and atomic absorption spectrophotometry.
Comparisons between sites over the period of operation (minimum of 18
months for each site) will be made, as well as observation of seasonal and
annual variations at the sites. In addition, comparisons will be made
between the Florida cleared site and an adjacent site within the forest
canopy.
767
-------
INTRODUCTION
Awareness of the apparent effects of acidic deposition (primarily in
the form of H^O^ and HNO-j) upon the forests of eastern North America has
existed for the past two decades.^'^ However, the focus of attention has
been limited primarily to the northeastern United States and eastern
Canada. No comprehensive study to date has addressed the possible effects
of acidic deposition upon commercial forests in the southeastern United
States. The Southern Commercial Forests Research Cooperative (SCFRC) was
established as part of the Forest Response Program (FRP), National Acid
Precipitation Assessment Program (NAPAP), to thoroughly study commercial
pine forests in regard to the possible effects of acidic deposition on
various species of southern pines.
EXPERIMENTAL
Site Selection
The location of three monitoring sites established by the University
of Florida as part of the Atmospheric Exposure Cooperative (AEC) supporting
the SCFRC program is shown in Figure 1. All sites are situated in cleared
Stephen F, Austin Forest
Austin Cary Forest
a
SOUTHERN COMMERCIAL FOREST
RESEARCH COOPERATIVE
UNIVERSITY OF FLORIDA REMOTE SITES
Figure 1. SCFRC Monitoring Sites
areas within commercial loblolly, slash, and short leaf pine plantations
located at Duke Forest, in north central North Carolina; Austin Cary
Forest, in north central Florida and Stephen F. Austin Forest, in east
central Texas, respectively (hereafter referred to as NC, FL, and TX). The
periods of sampling for the data reported in this paper was from January
768
-------
1988 to December 1989 for the NC site and from July 1988 to December 1989
for the FL and. XX sites.> In addition, sampling has been carried out at a
site in the Integrated Forest Study network (hereafter referred to as IFS).
The latter site is located approximately 2 km northeast of the FL site but
within the forest canopy. The period of sampling at the IFS site was from
May 1988 to May 1989.
Field Sampling
Wet deposition samples are collected at all sites using Aerochem Met-
rics Model 301 Automated Wet-Dry Deposition Collectors (Bushnell, FL).
Samples are analyzed at each site for pH ar.d conductivity using a Markson
Model 4603 Solution Analyzer (San Diego, CA). A 250 mL aliquot of each
event (whenever possible) is shipped to the Air Pollution Analysis Laborat-
ory in a cleaned HDPE bottle under refrigeration. Precipitation amounts
are measured by a Climatronics tipping bucket rain gauge (Bohemia, NY) and
recorded by an Odessa Model DSM-3260 Datalogger (Austin, TX). At the IFS
site, precipitation sample volumes of 500 nL are collected from automated
wet-dry deposition collectors, from stemflcw collectors and from open
containers on the forest floor (throughfall).
Laboratory Analysis
Upon receipt of wet deposition samples, aliquots are taken for
measurement of pH and conductivity. The remainder of the sample is
filtered through a 0.45 micron Nalgene filter assembly and the filtrate
stored in a separate, labelled 125 mL HDPE bottle and refrigerated at 4
degrees C for later analysis. The general methods and protocols for the
detailed analysis of precipitation have been previously reported.
Conductivity and pH measurements are performed with a Markson Model 4603
Solution Analyzer, flame atomic absorption measurements are performed with
a Perkin-Elmer Model 5100 Atomic Absorption Spectrophotometer (Norwalk, CT)
and ion chromatograph measurements are performed with a Dionex. Model 40001
Ion Chromatograph with separate columns for separation and analysis of
inorganic anions and monovalent cations (Sunnyvale, CA).
RESULTS AND DISCUSSION
Annual precipitation amounts observed at the SCFRC sites for the
years 1988 and 1989 are shown in Table 1. At the NC site, there was a 57
percent increase in precipitation from calendar years 1988 to 1989.
Because the FL and TX sites began operation at the start of the third
quarter of 1988, comparison of annual precipitation amounts cannot be made.
However, the amounts of precipitation at the FL and TX sites during the
last half of 1989 are lower than for the last half of 1988 by 29 percent
and 23 percent, respectively.
Table 1. Precipitation Amounts at: SCFRC Monitoring Sites
Site Amount in 1988 (cm) Amount in 1989 (cm)
NC
FL
TX
99.52
88.27*
42.65*
155.98
109.25
128.22
* - Last two quarters of 1988 only
769
-------
pH and Conductivity
Volume weighted averages for hydronium ion concentration and conduc-
tivity at the SCFRC sites (shown as annual averages) are presented in
Figure 2. It is apparent that the NC site has the most acidic precipita-
tion of the three monitoring sites in the network. The FL and TX sites
show nearly equivalent values for hydronium ion concentration and conduc-
tivity on an annual basis. Typical weather patterns show an uptake in
moisture from the Gulf of Mexico at these sites, while the NC site weather
patterns indicate the possibility of long-range transport of acidic
deposition precursors from inland areas.
Comparison of pH values recorded at the sites and later at the Air
Pollution Analysis Laboratory show a general increase in pH values (usually
less than 0.1 pH units) during sample transit. It has been reported that
an apparent increase In pH is likely due to microbes decomposing acetic and
formic acids initially present in the sample. '
Major Inorganic Anions
The project protocols require that only the major inorganic anions
(Cl, NO^, total SO^ and HPO^) are to be analyzed. The annual volume -
weighted averages for these anions at the SCFRC sites are presented in
Figure 2. The levels of SO^ and NC>3 at the NC site are higher during the
period of study, reflecting the increases in hydronium ion concentration
and conductivity at this site. Elevated levels of SO^ and NO-j are also
apparent for the FL and TX sites, although the concentrations on an annual
basis are nearly identical. However, the Cl levels at the FL and TX sites
are noticeably higher during the period of study, whereas the NC site
showed no marked annual variation. Again, the apparent weather patterns
for the uptake of water vapor involving the FL and TX sites might account
for this unique variation.
Major Cations
The annual volume-weighted averages for the major cations at the
SCFRC sites are presented in Figure 2. At the NC site, higher concentra-
tions of Ca and NH^ are indicated for precipitation during 1988. All
levels of cations at the NC site are higher in 1988 than in 1989 due
probably to decreased precipitation amounts in 1988. At the FL and TX
sites, the annual averages for the divalent cations (Ca and Mg) and K are
nearly identical between years and sites. However, the annual average for
Na is noticeably higher in 1988 at the TX site and in 1989 at the FL site.
This may indicate the relative contribution of sea water uptake at these
sites.
SCFRC - IFS INTERSITE COMPARISON
The IFS site utilizes three different types of wet deposition collec-
tors: automated wet/dry deposition collectors in a cleared area and within
the forest canopy (abbreviated as WF), stemflow collectors for measurement
of deposition that runs down the exterior of the tree (abbreviated as SF)
and throughfall collectors placed in specified areas on the forest floor
within the canopy (abbreviated as TF).
The data for the major components measured in wet deposition at the
FL and the IFS sites are presented in Table 2. Comparison of hydronium ion
volume-weighted concentration values shows that collection outside of or
within the forest canopy is not a crucial factor. In addition, values for
770
-------
most analytes are similar for the WF data from the IPS site and the FL
SCFRC site. Samples gathered by SF and TF collectors show a large increase
in the concentration of most analytes, which might be caused by leaching
from the exterior surfaces of the trees or by washing off of particles dry
deposited the tree surfaces. The only exceptions were for NH^ and for NO-j,
which indicate that those components are presumably utilized by the trees
or other organisms during and after the precipitation event.
Table 2. Statistical Summary of FL and IFS Wet Deposition Data
(all concentrations in mlcromoles per liter)
Site and Collector S0^= NCy Cl- HPOii= H+ HH^+ K+ Na+ Ca++ Mg+
IFS
IFS
IFS
FL 1988
FL 1989
UF Collectors
TF Collectors
SF Collectors
13.34
17.35
30.66
44.72
180.70
10.05
15.25
13.95
18.24
11.24
11.69
25.50
17.70
44.96
210.45
0.72
1.19
0.8:5
1.30
2.39
25.50
27.76
22.99
23.12
211.40
5.88
12.16
9.49
4.71
2.38
0.84
1.54
1.10
5.77
16.65
18.34
31.01
21.61
56.26
136.24
11.99
15.79
21.67
48.68
161.82
5.67
7.72
11.70
25.02
80.53
CONCLUSIONS
It has been found that wet and dry deposition samples collected at
the NC site contained higher levels of nearly all acidic deposition
precursors than the FL or TX sites while samples obtained at the FL and TX
sites are remarkably similar. One disturbing trend was found in the
calculation of ion balances at all sites. Typically the anion budget in
wet deposition (in microequivalents per liter) is lower than the cation
budget by a factor of twenty to thirty percent. The likely cause of this
deficit is the failure to preserve and immediately analyze samples for the
weakly ionized organic and inorganic species. Earlier research has found
that organic acid components in wet deposition may contribute between 15 to
35 percent of the free acidity present. >^ Currently field samples are
being preserved by the addition of chloroform on site before shipment in
order to analyze for these weakly acidic components.
ACKNOWLEDGEMENTS
We wish to thank our site operators in the SCFRC network (Rafael
EePaz, David Anthony, David Vermillion, Asgar Mirza, Robert Yeh, P. Brian
Morlock) for the collection of samples and maintenance of the sites. In
addition, the efforts of our undergraduate students in the Air Pollution
Analysis Laboratory (Richard Hutton, Allen Preston, Jose Garcia, Rodney
Phillips, Daniel Green, Ken Kohn, Michael Pri.ng, Mark Roberts, Kim Owen)
for the prompt analysis of wet deposition samples are greatly appreciated.
Finally, the cooperation of research colleagues at the IFS network site
(Dr. Henry Gholz, David Nolletti and Steve Siaitherman of the Department of
Forestry, University of Florida) is gratefully acknowledged.
771
-------
DISCLAIMER
This research was supported by funds provided by the U.S. Environ-
mental Protection Agency within the joint U.S. Environmental Protection
Agency - USDA Forest Service Forest Response Program in cooperation with
the National Council of the Pulp and Paper Industry for Air and Stream
Improvement (NCASI). The Forest Response Program is part of the National
Acid Precipitation Assessment Program. This paper has not been subject to
EPA or Forest Service peer review and should not be construed to represent
the policies of either agency,
REFERENCES
1. G.E. Likens, R.F. Wright, J.N. Galloway and T.J. Butler, "Acid Rain,"
Scientific American. 243: 43-51 (1979).
2. N.R. Glass, D.E. Arnold, et.al., "Effects of Acid Precipitation,"
Environ. Sci. Technol.. 16: 162A-169A (1982).
3. M.E. Peden, S.R. Bachman, et.al., Development of Standard Methods for
the Collection and Analysis of Precipitation. EPA CR810780-01, U.S.
Environmental Protection Agency, Cincinnati, March 1986.
4. W.C. Keene and J.N. Galloway, "Organic Acidity in Precipitation of
North America," Atmos. Environ.. 18 (11): 2491-2497 (1984).
5. J.N. Galloway, G.E. Likens, W.C. Keene and J.M. Miller, "The Composi-
tion of Precipitation in Remote Areas of the World," J. Geophys.
Res.. 87 (11): 8771-8786 (1982).
772
-------
%
(D
Si
O
3
o
rt
O
H-
O
3
ANNUAL VOLUME WEIGHTED AVERAGES
SCFRC MONITORING SITES - 1988 AND 1989
•• H+J MM CONDUCTIVITY
uS/cm
NC 88 FL 88 TX 88 NC 89 FL 89 TX 89
ANNUAL VOLUME WEIGHTED AVERAGES
SCFRC MOMTQFJNG SJTES — 1968 AfO ^969
ANNUAL VOLUME WEIGHTED AVERAGES
SCFRC ^.'1C^.I!TOP.!^!G
S3
(D
rt
(0
n>
i-i
P
cw
(D
in
NC SB FL 88 TX 88 NC 89 FL 89 TX 89
NC88 FL88 TX88 NC89 FL
TX 89
-------
PM-10 HI-VOL COLLECTION AND QUANTI-
TATION OF SEMI-VOLATILE METHOXYLATED
PHENOLS AS TRACERS OF WOOD SMOKE
POLLUTION IN URBAN AIR
Steven B. Hawthorne, David J. Miller,
Mark S. Krieger, and John J. Langenfeld
University of North Dakota
Energy and Environmental Research Center
Grand Forks, North Dakota 58202
Methoxylated phenols, including guaiacol (2-methoxyphenol) and its
derivatives have been shown to have consistent concentrations in the smoke from
28 different wood stoves and fireplaces regardless of the type of wood burned or
operating conditions, suggesting that they would be useful organic tracers of
atmospheric wood smoke pollution. To test this hypothesis, a PM-10 Hi-Vol was
modified to utilize quartz filters backed by polyurethane foam sheets (PUF) to allow
quantitative collection of the methoxylated phenols while maintaining the 40 cfm
air flow required for a 10-um cutoff. Twenty-four winter Hi-Vol samples were
collected in Minneapolis and Salt Lake City to represent air impacted by hardwood
and softwood burning, respectively, and approximately 60 semi-volatile and
particulate-associated organics including phenols, methoxylated phenols, alkanes,
and PAHs have been quantitated in each filter and PUF extract. GC/MS analysis
of the unfractionated extracts showed that phenols and methoxyphenols
accounted for the largest quantity of the semi-volatile organics collected, while n-
alkanes (C14 and larger) and PAHs (acenaphthylene and larger) accounted for an
average total of only ca. one-third of the semivoiatiles. All of the guaiacol
derivatives (except formylguaiacol) had nearly identical relative concentrations in
chimney smoke and in the urban air samples indicating that guaiacols have
sufficiently long lifetimes to be useful as wood smoke tracers. 14C analysis was
performed on each filter to determine "new" (assumed to be from wood smoke)
versus "old" particulate carbon, and preliminary results show good correlation
between guaiacol derivative concentrations and 14C analysis.
774
-------
Introduction
Methoxyphenols, including guaiacol (2-methoxyphenol) and its derivatives,
may be useful tracers of atmospheric wood ;;moke pollution because they are
emitted at consistent concentrations from wood stoves and fireplaces regardless
of the type of wood burned or operating conditions12. As pyrolysis products of
wood lignin, methoxyphenols should be unique to wood smoke in winter urban air.
In an attempt to validate the use of these organics as tracers of wood smoke
pollution in urban air, comparisons between the concentration of the
methoxyphenols and "new" and "old" carbon (based on 14C measurements) have
been conducted on 24 winter Hi-Vol samples collected in Minneapolis (primarily
hardwood burning) and Salt Lake City (primarily softwood burning). The PM-10
sampler was modified to include polyurethane foam (PDF) sorbent sheets so that
semi-volatile organics could be quantitatively collected under flow conditions
needed to maintain the 10-um cut-off, and 60 of the most concentrated organics
(including phenols, methoxylated phenols, PAHs, and n-alkanes) found in the
unfractionated filter and PDF extracts were quantitated in each sample.
Experimental Methods
Three sites were chosen for each city to include residential with low vehicle
traffic, residential with high vehicle traffic, and non-residential (e.g. downtown)
locations. Both day and night air samples were collected at each location using
a PM-10 Hi-Vol sampler that was modified to utilize quartz filters backed by two
7" X 9" X 2" polyurethane foam sheets (PUF), This modification allowed
quantitative collection of the methoxyphenols and semi-volatile hydrocarbons under
air flow conditions needed to maintain the 10-um cut-off. After a 1 2-hour sample
collection (at 68 m3/hr), the filters and PUF sheets were exhaustively extracted
with acetone and the unfractionated extracts were analyzed by GC/MS.
Approximately 60 individual phenols, methoxyphenols, n-alkanes, and polycyclic
aromatic hydrocarbons (PAHs) were quantitated in each extract based on
calibration curves generated from over 50 individual standard species.
Results and Discussion
The modified PM-10 Hi-Vol sampler allowed quantitative collection (based
on the absence of significant species on the back-up PUF sorbent sheet) of phenol
and alkylphenois, all of the methoxyphenols (including guaiacol and syringol
derivatives), acenaphthylene (M = 152) and higher molecular weight PAHs, and
C14 and larger n-alkanes. As shown in Figure 1 by a GC/MS analysis of the
unfractionated filter and PUF extracts from a typical Hi-Vol sample, the phenols
and the majority of the guaiacol derivatives were collected on the PUF sheets,
while the less volatile methoxyphenols were collected primarily on the filter. As
would be expected, the largest percentage of the more volatile n-alkanes (C14 to
C18) and PAHs (up to phenanthrene) also pcissed through the filter and were
collected on the PUF.
775
-------
The quantitative analyses of the unfractionated PUF and filter extracts from
the urban air samples have yielded three major results:
1) Phenols and methoxyphenols are among the most concentrated semivolatile
organics in winter urban air. Of the species that were quantitated in the Hi-Vol
PUF and filter samples, phenols averaged 45% of the total (on a weight basis),
methoxyphenols averaged 24%, n-alkanes averaged 20%, and semi-volatile PAHs
(ranging from acenaphthylene to PAHs with M = 252, not including alkyl-
substituted PAHs} averaged 11%. As would be expected, the proportion of
phenols and methoxylated phenols is higher in the residential samples (because of
the higher influence of wood burning) than in the downtown samples.
2) The relative proportions of the individual guaiacol tracers found in the Hi-Vol
samples is similar to those previously reported for wood smoke samples collected
from chimneys2, with the notable exceptions of the 4-propenylguaiacol and 4-
formylguaiacol species (Figure 2). These results indicate that the majority of
guaiacol derivatives emitted from wood burning have sufficiently long life-times in
winter urban atmospheres to be useful tracers.
3) Although not all.of the 14C results were available at the time of this report and
only preliminary data analysis has been possible, the concentrations of guaiacol
derivatives show good correlations with the percentage of "new" particulate
carbon (assumed to be from wood smoke) determined by 14C analysis, indicating
that the measurement of guaiacol derivatives could provide a reliable method for
determining the relative contribution of residential wood burning to the inhalable
particulate concentrations in urban air (Figure 3).
Future work will focus on determining the best individual guaiacol tracers,
and will investigate the relationship between n-alkane concentrations (assumed to
be primarily from vehicle exhaust in winter air) and "old" (petroleum derived)
particulate carbon.
Acknowledgements
A.J.T. Jull, D.J. Donahue, and L. Toolin of the University of Arizona
Accelerator Mass Spectrometry Facility are thanked for performing the 14C
analyses. The financial support of the U.S. Environmental Protection Agency,
Office of Exploratory Research (grant number R-813257-01-0) is also gratefully
acknowledged.
References
1. S.B. Hawthorne, D.J. Miller, R.M. Barkley, M.S. Krieger, Environ. Sci.
Technol. 22: 1191-1196 (1988).
2. S.B. Hawthorne, M.S. Krieger, D.J. Miller, M.B. Mathiason, Environ. Sci.
Technol. 23: 470-465 (1989).
776
-------
<-m.lhyl I3« 154
"
,— tit -,
4-proprlguilacol 4-tc«l
-------
-3
oo
CO
03
CO
>
1_
CD
Q
"5
o
co
CO
u
CD
"ra
o
in
o
en
o
CN
guaiacol
methylG
ethylG
formylG
propenylG
propylG allylG
acetonylG
acetylG
•
Ambient air (Hi-Vol)
Chimney
Figure 2: Relative average concentrations of individual guaiacol derivatives collected in urban winter air and from
the chimneys of 28 different wood stoves and fireplaces.
-------
CO
11
10
O
E-
•M
O 7
8
I
O
10
40
SO
60
70
% New Carbon
Figure 3: Correlation of % guaiacol (compared to total organics quantitated) collected from the Hi-Vol winter air
samples with the fraction of "new" participate carbon based on 14C analysts.
-------
SOLID SORBENT AIR SAMPLER FOR THE CHARACTERIZATION
OF CONTAMINANTS IN SPACECRAFT ATMOSPHERES
Thomas F. Limero
KRUG International,
1290 Hercules Dr.
Houston, Texas. 77058
Theodore J. Galen
Lockheed Engineering and Sciences
2400 NASA Road 1
Houston, Texas. 77058
John T. James
Biomedical Operations and Research Branch
Mail Code: SD4
NASA, Johnson Space Center
Houston, Texas. 77058
Assessment of air quality aboard the Shuttle during early missions was
trusted to daily 300 ml. grab samples acquired with evacuated stainless
steel cylinders. Additionally, a small sample of the charcoal used to
scrub the cabin air as part of the Environmental Control and Life Support
System (ECLSS) was analyzed by gas chromatography/mass spectrometry
following thermal desorption and cryotrapping.
In later missions the number of sample cylinders manifested was reduced
such that daily grab samples were no longer possible. This, combined with
the greater number of compounds detected in the charcoal samples, led to
the development of the Solid Sorbent Air Sampler (SSAS) by the Biomedical
Operations and Research Branch at the Johnson Space Center.
This paper will present a detailed description of the SSAS which is
capable of acquiring eight 24-hour composite samples with minimal crew
involvement. The entire system, which is battery operated, is contained
in a cylinder of only 4.5" diameter x 8 " length. Furthermore, each of
the eight Tenax sampling tubes can be desorbed without exposing the
adsorbent to the laboratory atmosphere. Results from Shuttle missions
will be presented to demonstrate the effectiveness of the SSAS in
characterizing the internal cabin atmosphere.
780
-------
Introduction
The enclosed environment of a spacecraft, like any partially closed
atmospheric system, is prone to the accumulation of volatile organic
compounds (VOCs). Some of these volatile organic contaminants arise from
the familiar terrestrial sources: human metabolic products, materials
offgassing, leaks and/or spills of onboard chemicals, clothing, foods,
hygiene products, thermodegradation of synthetic materials, and fire
extinguishant. Sources of VOCs uniqut; to spacecraft include: the
Environmental Control and Life Support System (ECLSS) and extravehicular
activities (EVA). The analysis of charcoal taken from the ECLSS
atmospheric scrubbing system during the 7ipollo flights demonstrated the
presence of numerous VOCs in the spacecraft atmosphere.1 Although the
concentrations of most VOCs were below 1 part per million (ppm) , crew
safety was still a concern given the physiological changes that occur in
space and the uncertainties associated with long-term exposure to low
concentrations of VOCs.
Efforts to define a sampling and analytical method capable of
detecting sub ppm levels of volatile organic contaminants drew heavily
from accepted air sampling methods of the time2 and led to the development
of an atmospheric volatile concentrator for Skylab3 and the Air Sample
Assembly (ASA) for Shuttle. Both systems employed Tenax-GC as a medium
upon which VOCs were trapped from the spacecraft's internal atmosphere
during flight. These archival samples wer« subsequently returned to Earth
and analyzed. Cross-contamination of tubes, tube contamination during
ground analysis, and excessive crew time for sampling were a few of the
disadvantages of these devices which produced less than satisfactory
results.4 Complimenting the adsorbent-based sampling technology on
Shuttle was the collection of 300 ml grab samples in evacuated stainless
steel cylinders each flight day. Increasing demands on payload volume and
weight requirements have resulted in reducing the sampling protocols to a
pre-flight sample and a single inflight sariple collected the last day of a
mission. Additionally, the grab sample is a single point in time and
space; therefore, it may not be a true representation of the atmosphere
over the entire mission.
The development of a new sampling device, which eliminated the
problems observed in previous efforts, culminated in the construction of
the Solid Sorbent Air Sampler (SSAS). The SSAS is a compact, lightweight
unit capable of collecting seven 24 h composite air samples. The eight
air samples, each representing a volume o:: 3 liters pulled over Tenax-GC
adsorbent, can be collected using 4 "C" cell alkaline batteries.
Additionally, an entire sampling procedure, including laboratory analysis,
can be completed without exposing t.he Tenax tubes to a contamination
source. Furthermore, crew time required to obtain the samples is lesa
than one minute per day.
This paper will describe the design and operation of the Solid
Sorbent Air Sampler, and present results from data collected during
Shuttle missions. The SSAS, being a compact, easy to operate, quiet
sampler, can also be applied to studies of indoor air quality.
Experimental Methods
The success of the Solid Sorbent Air Sampler can be traced to its
simplicity (Figure 1). The SSAS is housed in a metal cylinder (anodized
aluminum) 4.5" in diameter and 9.5" in length with a total weight of 4.25
Ibs. It is composed of two subassemblies: an electronic assembly and a
gas flow assembly. The electronic subassembly contains a battery pack, a
diaphragm pump (model 11-0002, AeroVironment, Inc., Pasadena, CA) , and a
timing circuit that pulses the pump, thus permitting a sampling rate of
781
-------
approximately 3 liters/24 h. An ON/OFF switch is available to activate
and deactivate the pump. The heart of the gas flow assembly is an 18-port
(1/32" fittings) microvalve (Valco, Houston, Tx.) that connects the inlet
and outlet of all eight glass-lined Tenax tubes, the sample inlet, and the
pump. The sample inlet and the inlet/outlet ports of the Tenax tubes all
have 1 mm screen filters to eliminate particulate clogging of the inlet
and prevent loss of Tenax. Each 1/4" O.D. [.158" (4mm) I.D.] x 5.91" (15
cm) glass-lined tube (SGE, Inc., Austin, Tx.) contains 0.5 grams of Tenax-
GC (Alltech, Deerfield, IL.). The valve rotates easily through eight
positions indicated by the inscribed numbers (1-8) on the top of the unit.
If the valve is pointing to #1, then the sample inlet, the inlet and
outlet of tube #1, and the pump are in line; and, a sample of air is being
pulled across the Tenax in tube #1 by the pump. When the valve is facing
#2, then tube #2 would likewise be in line and an air sample would be
collected in that tube.
The sampling procedure is brief and can easily be written on a small
label attached to the unit and still leave space for recording sampling
times. Since an "in line" Tenax tube is exposed to the environment even
when the pump is deactivated, the valve is always positioned on Tube #8
when the SSAS is not operating. Consequently, samples collected on Tube
#8 will be of dubious value because of volatile organics diffusing in and
out of the tube from the surrounding environment. The three steps
required to initiate a seven day sampling cycle with the SSAS are: switch
the unit "ON" to start the pump, turn the valve to the first sample tube
(usually #1), and record the start time. At the end of a specified
sampling period, the user manually switches the valve to the next Tenax
tube and records the time. At the conclusion of the sampling operations,
the valve is again switched to position #8, the time is recorded, and the
SSAS is deactivated by switching the unit to "OFF". The SSAS unit can
now be sent to a laboratory for analysis without further preparation since
initial testing on the unit demonstrated no cross-contamination between
tubes .
Once the Solid Sorbent Air Sampler is received in the laboratory,
the electronic subassembly is removed and a helium purge line is connected
to the pump line. The sample inlet, with the filter removed, is attached
to a valve on a gas chromatograph leading to a cryotrap, and the SSAS
valve is positioned so that the tube to be desorbed is in line. The tube
o
is then heated to 200 C for 20 minutes with a 20 cc/min helium flow over
the Tenax, while the cryotrap is held at liquid nitrogen temperature. A
flow diagram of the gas chromatograph inlet system is shown in Figure 2.
At the conclusion of the desorption process the cryotrap is heated and
VOCs are transferred onto a GC column where subsequent GC/MS analyses are
performed to obtain qualitative and quantitative data. Each tube is
desorbed in a similar manner with a total run time of two hours per tube.
Once the desorption process is completed, the SSAS unit is attached to
another source of helium flowing at 10 cc/min and each tube is "cleaned"
o
for 24 hours at 250 C. Following this cleanup procedure, a blank is run
on each tube, and, with satisfactory results the unit is again ready for
use. It is noteworthy that the entire protocol is performed without
removing the Tenax tubes from the SSAS unit; thereby, eliminating a
potential contamination step.
Results
The Solid Sorbent Air Sampler has been restricted to Shuttle
missions that either contain the Spacelab or involve the first flights of
new or extensively refurbished vehicles. These restrictions have limited
the inflight operation of the SSAS to seven missions over the past five
years. Unfortunately, the archival samples collected have been analyzed
on a variety of GC/MS instruments. However, some trends in the data do
surface and with the SSAS manifested on several missions scheduled for the
782
-------
near future, a more comprehensive characterization of the Shuttle internal
atmosphere should be forthcoming.
The results presented in Table I5'6 indicate that the total mass of
volatile organic compounds collected is generally in the range of 1.5 to
2.5 mg/m^. As noted, these data do not include the contributions from
ethanol and 2-propanol because these compounds are introduced into the
cabin atmosphere at inordinately high levels by hygiene and medical wipes
respectively. The concentration of ethano.. rises dramatically, especially
near the conclusion of mission, when housekeeping activities usually
increase in anticipation of the return to Earth. The total mass of
volatile organics collected on Tube #7 of STS 32 was 5.5 mg/rrP; however,
ethanol and 2-propanol contributed 3.6 mg/rn^. Subtracting the
contributions of the ethanol and 2-propanol leads to the observation that
the concentration of VOCs appears to stabilize and rises only slightly
with time during a mission.
The data presented in Table II7 illustrate several important points.
First, the distribution of volatile organic compounds in the spacecraft
atmosphere is substantially different from that typically found in an
indoor environment on Earth. Terrestrial samples of indoor air usually
contain as high as 80%-90% hydrocarbons (alkanes, alkenes, etc.).8 As
noted in Table II, the spacecraft atmosphere contains more oxygenated
hydrocarbons and halogenated compounds than alkanes, alkenes, and
aromatics. This composition may reflect careful materials selection and
rigorous offgas testing applied to all nonmetallic substances under
consideration for use in flight hardware. Although offgassing of
materials in flight still occurs, the major contributors to the VOCs in
the atmosphere seem to be from metabolic processes and utility chemicals.
A second point to be ascertained from TabI.e II is that this air sampling
technique can indicate a problem has a occurred or the success of a
correction effort. For example, the levels of halogenated hydrocarbons
which were fairly high in the pre-Challenger era (3 flights with the
SSAS), appear to have been reduced. Furthermore, in the post-Challenger
era (4 flights with the SSAS) the level of silicone compounds would, appear
to be remarkably high; however, this can bet traced to a anomaly on board a
single mission (STS 28). Finally, in spite of different sampling devices,
personnel, and analytical systems, the trends of VOC composition in
spacecraft atmosphere appear to be consistent throughout the entire 17
years since Skylab.
Conclusion
The Solid Sorbent Air Sampler has proven to be a valuable means of
collecting volatile organic compounds from the Shuttle atmosphere for
subsequent ground analysis. The ability to collect, analyze, and clean
each tube without exposing the Tenax to the atmosphere is an important
advantage of this device. The SSAS has provided information on the
character of the volatile organics in the Shuttle atmosphere in both the
ppm and sub ppm ranges. The analyses of the SSAS samples and appropriate
trend analysis are tools that can be employed to identify anomalies or
modifications in the Shuttle atmosphere. Facing the challenges presented
by long duration missions, such as Space Station, necessitates exploring
the potential of multi-sorbent tubes and automated valve switching to meet
the demands of extended spaceflight. Additionally, studies comparing the
SSAS with Passive Sampler Devices (PSDs) is a high priority in the search
for sampling strategies for Space Station. Moreover, the design features
of the SSAS for spaceflight (lightweight, compact, quiet, simple
operation, and inexpensive) are the very same attributes required for a
successful indoor air sampler intended for long-term, broad base studies
on Earth.
783
-------
References
1. T. Godish, Air Quality. Lewis Publishers, Inc., Chelsea, MI. 1985,
pp.186-190.
2. W. J. Rippstein Jr., Biomedical Results of Apollor R.S. Johnston,
L.F. Dietlein, and C.A. Berry, eds., NASA SP-368, Washington
D.C. 1975, pp.153-159.
3. A. V. Shannon, Biomedical Results From Skylabf R.S. Johnston
and L.F. Dietlein, eds., NASA SP-377, Washington, D.C. 1977,
pp 478-481.
4. W. J. Rippstein, "Atmospheric Analysis Reports", STS 1 and
STS 2, Biomedical Sciences Division, NASA, Lyndon B. Johnson
Space Center, 1981.
5. M. Coleman, "Atmospheric Analysis Reports", STS 51B, 51J, 61A,
Biomedical Sciences Division, NASA, Lyndon B. Johnson Space
Center, 1985.
6. M. Coleman, "Atmospheric Analysis Reports", STS 26, 27, 28, 32,
Biomedical Sciences Division, NASA, Lyndon B. Johnson Space
Center, 1988-1990.
7. H, Kaplan, "Definition and Description of Contaminants in the
Spacecraft Environment", J. Waligora, ed., NASA 1045,
Washingtion D.C. 1979, p 25.
8. W. Bertsch, A. Zlatkis, H. Liebich, and H. Schneider,
"Concentrations and Analysis of Organic Volatiles in Skylab 4",
J. Chromatog-., 99: 673. (1974).
784
-------
TABLE I. TOTAL VOLATILE ORGANICS
COLLECTED (mg/m3)*
00
CJ1
MISSION
STS-51B
STS-51J
STS-61A
STS-26
STS-27
STS-28
STS-32
SOLID SORBENT AIR SAMPLER TUBE NUMBER
1
4.1
1,4
2 9
0.6
2.3
2.3
1.0
2
3.1
1.4
1 9
0.4
2.3
1.6
0,8
3
11.6
1.3
2.0
,81
1.8
1.6
1.5
4
1,7
2 4
2.3
1.8
1.5
5
1.2
22
1.2
1.5
6
1.7
1.5
7
1,0
1.9
8
2.0
* mg/m3 OF ETHANOL AND 2-PROPANOL ARE NOT INCLUDED
IN THESE VALUES
-------
Table II. COMPARISON OF AVERAGES FOR VOLATILE
ORGANIC COMPOUND CLASSES IN
SPACECRAFT ATMOSPHERE BY ERA
-4
CO
OS
MISSIONS
SKYLAB*
PRE -
CHALLENGER
POST-
CHALLENGER
COMPOUND CLASS
ALKANES
ALKENES
AROMATICS
12.5**
8,4
7,0
* AN ESTIMATED AVERAGE
KETONES
ALDEHYDES
ALCOHOLS
ESTERS
42.5
41,3
50.1
SILICONS
COMPOUNDS
3,0
2,9
22,7
HALOGENATED
HYDROCARBONS
32,5
47.4
20.2
** ALL VALUES ARE mg/m3
-------
TENAX GC TUBE (1 OF 8)
00
BATTERY
PACK
TIMER
CIRCUIT
PUMP
SWITCH
1
VALVE
n
GAS FLOW
In
ELECTRONIC Out
Figure 1. Schematic of the solid sorbent air sampler electronic and gas
flow subassemblies. Arrows indicate the flow of an air
sample through the sampler.
-------
VACUUM
VACUUM
VALVE #2
d
PRESSURE
SENSOR
TO MSD
10 PORT
SAMPLE VALVI
HELIUM
INLET VALVE #3
HELIUM PURGE SUPPLY
HELIUM
INLET VALVE #4
FLOW
CONTROLLER
THREEWAY VALVE
SAMPLER VACUUM LINE
Vj] VACUUM
^ VALVE #1
n
PRESSURE
SENSOR
INLET SELECTOR
VALVE
0
STANDARD INLET
SAMPLER INLET
SOLID SORBENT
SAMPLER
CYLINDER INLET SYSTEM
STANDARD INLET SYSTEM
SOLID SORBENT INLET SYSTEM
Figure 2. GC/MSD system valving and inlet paths showing
the sample valve in position for cryotrapping.
788
-------
SULFUR ON SURFACES OF ATMOSPHERIC MINERALS
AND SPORES
Yaacov Mamane
Environmental Engineering, Technion, Haifa 32000 ISRAEL
Thomas G. Dzubay and Rachel Ward
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711, USA
Atmospheric particles were analyzed by electron microscopy to obtain
direct evidence for S enrichment on minerals and spores in the 0.5 to 10 /jm
diameter range. The particles were collected at Research Triangle Park for
two weeks during summer, 1988, using dichotomous samplers that were modified
for use with electron microscopy. One of the samplers was equipped with an
annular denuder that removed acidic gases upstream of the filters to avoid
artifact reactions on the filter between gases and particles. A comparison
of X-ray spectra for atmospheric particles and locally collected soil and
mushroom spores revealed significant S-enriclunent of atmospheric silicates and
spores. It was concluded that the S enrichments were caused by atmospheric
reactions and not by sampling artifact.
789
-------
Introduction
The goal of this study is to obtain direct evidence for enrichment of
sulfur on minerals and spores in the atmosphere as a function of diameter in
the 0.5- to 10-/jm range. To meet that goal, we compared ambient particles
with locally collected soil particles and mushroom spores. To avoid confusion
about whether some of the observed sulfur enrichment is caused by SO_ reacting
with particles on the filter during sampling, we analyzed samples from a
second simultaneously operated dichotomous sampler that was equipped with an
annular denuder that removed SO- upstream of the filters. Instead of using
standard dichotomous samplers, which collect fine particles as multiple layers
that are unsuitable for microscopy analysis, we used modified dichotomous
samplers that collect fine and coarse particle samples that are suitable for
both microscopy and bulk sample analysis . We used X-ray fluorescence (XRF)
to obtain the elemental content of the samples and scanning electron
microscopy (SEM) with an energy dispersive X-ray (EDX) spectrometer to analyze
individual fine (<2.5 /jm aerodynamic diameter) and coarse (2.5 to 10 fjm)
particles.
Methodology
Ambient particles were collected simultaneously in two modified
dichotomous samplers , placed on the roof of EPA laboratory 10 m above ground
in Research Triangle Park, NC, for 10 consecutive 24-h periods between June
28 and July 9, 1988. One of the samplers was equipped with an annular denuder
to remove SO_, HNO, and HNO,.
All of the Teflon filters (coarse and fine) were analyzed by gravimetry
for mass and by XRF. Several coarse Teflon and fine Nuclepore filters were
analyzed manually in a SEM equipped with an EDX spectrometer. Size, X-ray
spectrum and morphology were recorded for several hundred particles.
Soil and road dust, and fresh mushroom spores from the area were sampled
on Teflon and Nuclepore filters. The Teflon filters were analyzed by
gravimetry and XRF, the Nuclepore filters by SEM-EDX.
Results and Discussion
Source sample analysis
Spores. The three major elements detected by XRF are P, S and K, and
the S/P and S/K ratios are about 0.4 and 0.5, respectively. Such ratios for
spores are similar to results reported for pollen .
Figure lisa SEM micrograph and an X-ray spectrum of mushroom spores.
The spores are quite uniform in size and shape. The spores contain P, K, and
S with ratios of 0.3- 0.4 and 0.7-0.8 for S/P and S/K, respectively.
Soil and Road Dust. SEM analysis of soil and road dust sample showed
that particles were irregular in shape and composed mostly of silicate
minerals. Sulfur was not associated with these minerals. Only few particles
analyzed in the microscope had detectable amounts of S.
XRF results for ambient samples
The mass and the complete set of XRF data for fine and coarse particles
led to the following findings:
(a) The total particle mass concentration ranged from 20 to 50 pg/m , and the
ratio of fine to coarse mass concentrations ranged from 1.3 to 5.3.
790
-------
(b) Sulfur, as ammonium aulfate, accounts for about 50% of the fine mass
concentration. No other element (>5%) was measured by XRF. Sulfur was
detected by XRF in the coarse fraction although the concentration was low.
(c) Alumino-silicate minerals account for about 50% of the coarse mass.
Volatile nitrates, spores, other organic material, and water can contribute
to both size fractions but were not measured by XRF.
(d) Day to day variations in mass and elemental concentrations are as large
as a factor of 2 for consecutive samples. The relative changes in the coarse
particle composition are not identical to those of the fine fraction.
(e) Particle losses in the annular denuder were very small.
Three cases were chosen for SEM-EDX analysis based on S in fine fraction
and Si (minerals) in the coarse fraction:
Case I (June 28): low fine S and nearly median Si,
Case II (July 3): high fine S and lowest Si and
Case III (July 5): lowest fine S and high Si.
Case II represents S concentrations that are typical of summertime conditions
in the East and Midwest, and cases I and III represent conditions of unusually
low S-concentrations in the eastern United States.
SEM results for ambient samples
Coarse particles, analyzed by SEM-EDX, were classified into categories
according to morphology and X-ray spectrum ' . Table 1 shows particle counts
in each category. For the above three cases particles consisted of 60 - 80%
minerals, 10 - 30% spores and pollen, and 1-2% fly ash.
Almost all fine particles smaller than 0.5 /jm were sulfates. The non-
sulfate fine particles, 0.5 to 2.5 jum, were 50-60% organics, 30-45% minerals,
5-10% spores, and 0-3% fly ash.
S enrichment in ambient samples
Figure 2 show SEM-EDX X-ray spectra of atmospheric silicates and spores
that appear to be enriched in S relative to the spore and mineral source
samples. To systematically characterize particles according to S enrichment,
coarse fraction silicates and spores were classified as follows:
Ns < 150 - no S enrichment
150 < Ns < 350 - some S enrichment
350 < Ns - high S enrichment
where Ns is the number of S X-ray counts acquired in a 30-s analysis. Fine
fraction silicates and spores were classified similarly except that the count
limits were 30% lower than those shown above. On the basis of these
definitions of s-enrichment, all the fresh spores and soil minerals (see
Figure 1), are in the category of "no enrichment".
What is the cause of the S-enrichments in atmospheric particles? First,
we eliminate the possibility that SO_ reacted with particles on the filter
during sampling because the denuder in one of the samplers removed SO2
upstream of the filter. The small differences between denuder and non-denuder
data in Table 1 are due to counting statistics) and cannot be attributed to the
denuder. Second, our use of modified dichotomous samplers caused particles
to be spaced far enough apart on the filter so that each X-ray spectrum are
representative of only the particle being examined . Third, the silicates and
spores in our source samples were not enriched in S.
Thus, we conclude that the observed S-enrichments are a consequence of
reactions that occur in the atmosphere while particles are airborne. A
791
-------
possible mechanism is that SO- and oxidants diffuse into and react in a water
layer that forms on particles in clouds and at high humidity.
Data in Table 1 indicate that S-enrichment occur in all three cases
studied for coarse and fine particles. The amount of enrichment was greatest
for case II, which represents typical summertime high fine S concentrations,
and was least for cases I and III.
Summary and Conclusions
The main results are summarized in the following:
(1) No significant difference in the S-content of atmospheric spores and
silicates was obtained for samples collected with and without a denuder.
Thus, there is no evidence for gas-particle reactions occurring on the filter
during sampling.
(2) The sources analyzed (soil and spores) were not enriched in S.
Sulfur was not detected by XRF in the soil minerals. Sulfur was found in
spores as part of their intrinsic composition but was always less than the
amount of P and K in each spore studied.
(3) All ambient samples that we analyzed by SEM-EDX contained silicates
and spores that were enriched in S. Moreover, on July 3, a day when S
concentrations were typical of Eastern summertime values, most of the
silicates and spores were either enriched or highly enriched in S. The
S-enrichments were caused by atmospheric reactions. Although this study
demonstrated that both fine and coarse fraction minerals and spores become
enriched in S while airborne, it does not determine the mechanism by which
this happens.
Ac knowledgement s
We thank R. Stevens, T. Lemmons, R. Kellogg and M. Wilkins for their
assistance during this project. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use. This research was
supported in part by the National Research Council, Washington, D.C.
References
1. Mamane, Y. and T.G. Dzubay (in press) " Dichotomous samplers modified for
use with electron microscopy". Aerosol Sci. and Technol.
2. Vossler T.L., Stevens R.K., Paur R.J., Baumgardner R.E. and Bell J.P.
(1988) " Evaluation of improved inlets and annular denuder systems to measure
inorganic air pollutants". Atmos. Environ. 22, 1729-1736.
3. Stanley R.G. and Linskens H.F. (1974) Pollen-Biology, Biochemistry.
Management. pp 119-127, Springer-Verlag, Berlin.
4. Dzubay T.G. and Mamane Y. (1989) " Use of electron microscopy data in
receptor models for PM-10". Atmos. Environ. 23, 467-476.
792
-------
Table 1. Particle count by category for coarse particles in the 1.5- to 10
diameter range *
Case I: June 28 Case III: July 5 Case II July 3
no no no
denuder denuder denuder denuder denuder denuder
Silicates
No S enrichment 30 37 67 45 11 7
Some S enrichment 29 11 12 21 24 24
High S enrichment 46 57 18 34
Spores and pollen
No S enrichment 15 13 44 34
Some S enrichment 11 12 4 2 47
High S enrichment 14 13 55 16 11
Others # 17 25 16 23 24 30
Total 120 117 113 107 100 117
* The total area scanned in four fields wasj 38000 pm per sample.
# Include other minerals, fly ash, orqanicss, and metal rich particles.
793
-------
Fig. 1. (A) Micrograph of mushroom spores at X5000 magnification. (B) x-ray
spectrum of typical mushroom spore. Note that the S peak (abundance) is lower
than that of P or K.
794
-------
Fig. 2. X-ray spectra of coarse particles, enriched in S (cursor line is
on the S peak) in July 3 samples: (A) mineral (major peaks are Al, Si)
and (B) spore (high background spectrum characterizes biological material)
from sample with no denuder.
795
-------
DEVELOPMENT OF ATMOSPHERE GENERATION, MONITORING, AND CLEAN-UP SYSTEMS FOR
ACID AEROSOL ANIMAL EXPOSURES
M.A. Higuchi, D.L. Ross, and G.F. Hudson
Inhalation Exposure Group, NSI Technology Services Corporation
Research Triangle Park, North Carolina 27709
D.W. Davies
Inhalation Exposure Section, Pulmonary Toxicology Branch
Environmental Toxicology Division, Health Effects Research Laboratory, U.S. EPA
Research Triangle Park, North Carolina 27709
i
The subchronic effects of ambient levels of acid aerosols has yet to be determined, due
mainly to the highly reactive nature of these compounds. This paper explains the development of
an acid aerosol exposure system for animals. The animal chambers are continuously monitored
and maintained for the proper aerosol concentration, temperature, relative humidity, and air
changes. The quality of aerosol delivered is determined by the aerosol generation system. The
generation system was composed of a constant output atomizer (T5I, Inc.), which was linked to a
diffusion dryer, aerosol neutralizes and heat exchange tee. To maintain a uniform aerosol
distribution, the Hazleton 2000 Inhalation Chamber was used to house the animals. Continuous
monitoring was accomplished by use of a sulfur analyzer and a tapered element oscillating
microbalance (TEOM)3 linked to each chamber. Aerosol species identification and
characterization was determined by annular denuder samples analyzed by ion chromatography
(1C). The aerosol particle size was determined by use of cascade impactors and particle-sizing
instruments. Each chamber effluent was conditioned by a scrubber (Mystaire) for removal of acid
aerosols and gases before being exhausted to the atmosphere. The design of these systems
provides the operator with a high degree of freedom to accomplish routine analysis of chamber
samples while continuously maintaining four inhalation exposures.
Disclaimer
This is an abstract of a proposed presentation and does not necessarily reflect EPA policy.
796
-------
Introduction
The term acid aerosols includes many acid species, including two sulfated species-
and NHUHSOa, which are both droplets in ambient air. Although HMOs is considered an "acid
gas" because of its ability to form a gas molecule from aqueous solutions, in dilute aqueous
solution {0.1 M), HNC>3 is strong, being about 93% dissociated1. This indicates that an aerosol
generator will deliver HNOa asan aerosol and gas mixture.
The subchronic effects of ambient levels of acid aerosols (K^SO/j, NH4HSO4, and HNOs) have
not been examined thoroughly using state-of-the-art measurement and exposure techniques.
This paper describes the development of an acid aerosol inhalation exposure system for guinea
pigs. The exposure chambers provide the animals with the proper temperature, relative
humidity, and adequate ventilation. The exposure system was developed to deliver
predetermined concentrations of acid aerosol within a specific particle size range (0.4 to 0.8 ^im
mass median aerodynamic diameter [MMAD]}; to minimize acid neutralization by NHs; measure
the aerosol concentration continuously and quantitate the different chemical species present,
and exhaust chamber effluents that are essentially acid free.
Methods
Generation
The aerosol generation systems for each test material (h^SO^ NH4HSO4, and HNOa) are
described below. These systems provide stable, accurate test agent flow rates to the exposure
chambers to provide target aerosol concentrations and aerosol particle size.
The generation equipment is mounted on a 3 ft wide by 4 ft long by 2 in. thick platform
mounted to the wall and ceiling 6 ft above the floor, and includes metering pumps, constant
output atomizers, diffusion driers, and particle neutralizes. These items are available
commercially. A solution of either H2SO4, NH4HSO4, or HNOa is metered into a nebulizer and
atomized as a polydispersed aerosol, with the excess solution being delivered to a waste reservoir.
The aerosol becomes more uniformly sized by the removal of excessive moisture in the diffusion
dryer. The mechanical action of generators can create highly charged aerosols2. These electrical
charges are neutralized to a normal Boltzmann equilibrium by passing the aerosol through a
particle neutralizer (2.0 mCi of 85Kr). Finally, the aerosol is delivered to the exposure chamber
through a diluter tee to prevent aerosol loss and ensure proper mixing with the chamber supply
air. The initial aerosol particle size is determined primarily by the acid solution concentration
delivered to the nebulizer. Final aerosol particle size is influenced by chamber humidity and
temperature. The exposure chamber concentration is. controlled by the metering pump and
chamber air flow rates.
Monitoring and Control
During operation of the acid aerosol exposure facility, all inhalation chambers are
monitored continuously and have automatic control and alarm for temperature (73 ±3 °F),
relative humidity (60 ± 10%), and chamber air flow (minimum 15 air changes/h). In addition, the
exposure chambers are monitored continuously for contaminate flow rate (within 1% of actual
flow rate), chamber static pressures (between negative 0.1 and 0.5 in. h^O), and nominal
concentrations (± 10%). These three parameters are controlled manually.
The chamber NHs levels are monitored on a routine basis before the start of every daily
exposure. This procedure includes sampling every chamber by drawing a known volume of
chamber atmosphere through an impinger and analyzing the sample with an NHa analyzer. Once
the NH3 levels in each chamber are acceptable (less than 20% neutralization of target acid
concentration), all exposures begin simultaneously. An NHj probe is used to verify the NH3
analyzer operation on a routine basis.
797
-------
Acid aerosols generated are monitored in the exposure chambers for particle size
distributions and total particle concentrations. To ensure complete aerosol size determinations,
two instruments with overlapping particle size ranges are used to determine the aerosol particle
distribution. An electrical aerosol analyzer ([EAA] TSI, Inc.) is dedicated to each exposure chamber
and one optical particle analyzer (Climet, Inc.) is used for confirmation and quality assurance (QA)
of each EAA, Each exposure chamber will be monitored continuously for particle size
distributions to ensure that uniform aerosols are generated during the entire exposure period.
Aerosol mass concentrations will be analyzed by a tapered element oscillating microbalance
(TEOMp ambient particulate monitor (Rupprecht & Patashnick Co, Inc.). This analyzer enables
precise and accurate measurement of all aerosols in the chamber as on-line, real-time data. These
data provide the operator information helpful for precise control of the generation system, as
well as continuous aerosol mass concentration analysis. One analyzer will be dedicated to each
exposure chamber to provide on-line, real-time aerosol concentrations.
The speciation determination during acid aerosol exposures is essential because of the
highly reactive nature of these aerosols and trace NHg levels created continuously by the animals.
EPA method IA09, using annular denuder system (ADS) followed by 1C analysis, will be used to
speciate the aerosol in the exposure chambers (EPA Method IA09, draft). This system was
developed to measure reactive acidic and basic gases and particulate matter which are contained
in indoor ambient air. The following chemical species can be measured by the ADS: gaseous S02,
HNO2, HN03, and NH3 and particulate SCV", NO3", NH4*, and H*. Other similar chemical species
can be collected sucessfully with just a few simple modifications (i.e., changing the denuder
coating solutions). Once collected, the species concentrations are quantified by 1C analysis and/or
NHs analyzer. The ADS sampling is performed once a day on every exposure chamber to monitor
for changes in chamber atmosphere species. All NH3 analyses were performed with an NH3 probe
(Fisher Scientific) in 20 ml of deionized water. The samples were collected for 30 to 60 min with a
midget impinger.
The total sulfur in the exposure chambers also is determined using an on-line real-time
total sulfur analyzer. This analyzer is multiplexed between all acid aerosol exposure chambers for
quality control over sulfur species (SO42', SO2).
Chamber Systems
Hazleton 20004'5 whole-body inhalation exposure chambers are used to conduct the
animal exposures (Figures 1 and 2). The 2.0-m3 chambers are constructed of stainless steel and
glass and are designed for animals to live in the chambers5. Each chamber contains six cage
modules with integral food troughs and automatic watering. Catch pans are mounted below
each cage module to collect the animal waste products. The animal caging capacity is 60
guinea pigs per chamber. All chamber air is filtered first by a prefilter, then by a bed of
activated carbon. The incoming air is then humidified to ~60% RH and is passed through a bed of
Type CA charcoal, two beds of Purafil®, and finally a high-efficiency particulate air (HEPA) filter.
This has proven to be an effective method of removing ambient NH3. The supply then is
dehumidified and rehumidified to ~50% RH. This conditioned supply air is delivered to a plenum
maintained at 1.0 in. positive pressure which supplies all inhalation chambers. The exhaust air
system has redundant exhaust blowers. When a mechanical failure occurs, the system
automatically switches over to the backup blower without loss of exhaust air flow.
The chambers are operated to minimize the neutralization of the H2SO4 by NH3 forming
(NH4)2SO4. This situation is addressed by maintaining strict cleanliness in the exposure
chambers. To maintain the lowest NH3 levels possible, animals are exposed in one set of
chambers for 4 h, then they are transferred to a second set of chambers (or filtered clean racks)
for holding overnight. Exposure chambers are cleaned so as to remove the animal waste products
splattered on the cages and chamber wails which are not removed by simply replacing the drop
pans and cage boards after exposures. Preliminary data collected from the proposed exposure
facility indicate that the above procedures are a necessary first defense in the effort to
minimize the NH3 problem. The average NH3 concentrations for 30 rats in the Hazleton 2000
chambers (15 air changes/h) with no control measures ranges from 0.2 to 1.0 ppm, depending on
798
-------
the number of hours the animals are housed in the same chamber. The same analysis has been
run for 30 guinea pigs, with the average NHs ranging from 0.5 to 1.5 ppm.
Aerosol Distribution
Aerosol distribution of H2$O/i in the Hazleton 2000 has been studied4. Additional
distribution testing will be conducted in the exposure chambers prior to initiating the proposed
animal exposures for toxicological evaluation. A TEOM v/ill be used to monitor the concentration
in the exposure chamber after the aerosol generator has stabilized. A reference position and
several points will be selected that are located in the breathing zone of the test animals. The
sample probe of the aerosol mass monitor will be rotated through the selected sampling
positions allowing sufficient time for the sample to stabilize at each position. The data will be
analyzed by a statistician for temporal and spatial variability. The aerosol distribution will be
evaluated thoroughly at all dose levels, then spot checks will be performed on the remaining
exposure chambers to verify the performance of each cha mber.
Waste Disposal
The removal of acid aerosol-laden atmospheres from the inhalation exposure chambers is
accomplished with a wet scrubber (Mystaire, Heat Systems, Inc.). This device physically removes
the aerosol particles from the air with water from a high-pressure nozzle sprayed on a nylon
mesh. The water is recirculated through a 10-gal holding tank mounted beneath the pump.
Acidified water is removed, neutralized, and replaced on a regular schedule. (Figure 2).
Results
Current research on the neutralization of NH3, which is created by the test animals (rats
and guinea pigs) and is present in ambient air, indicates a signficant problem with whole-body
exposures. The reactive nature of NHj can only be controlled by complete neutralization of it as it
is formed.
The most significant sources of NHs is the animals themselves, from action of urease
bacteria from animal feces on the urea found in the urin*-'- and from the expired air.7 The results
of the NHs levels of rats vs. guinea pigs are shown in Table I. The current control methods include
using chambers coated with 0.1N HjSC^ solution applied with an aerosol generator for 2 h and
allowed to dry overnight before loading the animals. The use of neomycin-treated DAC paper in
catch pans located under the guinea pigs and rats has greatly reduced NHs levels.
Future research plans include the use of an NH3-neutralizing substance incorporated into
disposable microencapsuiated absorbancy pads, in place of the neomycin DAC paper. It is hoped
that this procedure will give the control necessary for complete neutralization of NH3 generated
by the animals.
Ambient air is an additional source of NH3.8 This NHs rnust be removed from ambient air
which supplies the chambers. The method for removal is discussed previously under chambers.
This system is being modified to provide NHs-free air for the studies, and the preliminary results
of chamber room air and chamber supply air monitoring are presented in Table II.
The room air was sampled in the vicinity of the chamber, whereas Ports 3 and 4 located on
the middle tier in the Hazleton 2000 inhalation chamber (Port 3, right side, and Port 4, left side)
sampled the chamber air (see Figure 1). The decrease in NHs levels was due to increases in the
supply air relative humidity.
Future plans include using acid-soaked prefilters in the supply lines for bettter
removal. If needed, a scrubber may be employed (such as the one in Figure 2) to remove NH3 to
prevent the neutralization of generated acid aerosols.
799
-------
References
1. F.A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry: A Comprehensive Text.
Interscience Publishers. 1962.
2. K. Willeke, Generation of Aerosols and Facilities for Exposure Experiments, Science
Publishers, Inc., Ann Arbor. 1980.
3. H. Patashnick and E.G. Rupprecht, "Continuous PM-10 measurements using the tapered
element oscillating microbalance." Draft.
4. R.L. Seethe, R.K. Wolff, L.C. Griffis, CH. Hobbs, and R.O. McClellan, "Evaluation of a recently
designed multi-tiered exposure chamber," Inhalation Toxicology Research Institute,
Lovelace Biomedical & Environmental Research Institute. Report prepared under contract
number EY-76-C-04-1013. {Nov. 1979).
5. M.G. Brown and O.R. Moss, "An inhalation exposure chamber designed for animal
handling," Lab. Animal Sci. pp. 717-720. (1981).
6. P.P. Phalen. Inhalation Studies: Foundations and Techniques, CRC Press, Inc. 1984.
7. C.S. Barrow and W.H. Steinhagen, "NH3 concentrations in the expired air of the rat:
Importance to inhalation toxicology," Toxicol. Appl. Pharmacol. 53:116-121. (1980).
8. E. Buijsman, H.F.M. Maas and W.A.H. Asman, "Anthropogenic NHs emissions in Europe. "
Atmos. Environ. 21, pp. 1009-1021 (1987).
TABLE I. AMMONIA LEVELS FOR RATS VS. GUINEA PIGS
Rats
(ppb)*
227.4
210.2
295.5
236.6
278.6
249.1
415.6
259.7
Mean = 271. 6
Rats
(ppb)**
81.2
77.3
48.9
66.7
40.9
22.7
34.0
Mean = 53.1
Guinea Pigs
{ppb)**
54.6
30.0
62.4
37.6
58.4
56.0
79.8
56.0
Mean = 54.4
Chamber air flow rate was 500 Urn in (15 air changes/h) and the readings were taken over 4 h (30 rats).
Same chamber air flow with animals using neomycin-treated DAC paper and the chamber was coated with 0,1 N
H2SO« solution generated as an aerosol (30 rats or 30 guinea pigs).
TABLED. AMBIENT LEVELS OF AMMONIA
Room
(ppb)
15.0
15.4
4.6
6.9
4.5
4.2
4.8
Chamber Port 4
(ppb)
11.2
3.4
Chamber Port 3
(ppb)
35.4
16.6
19.1
6.4
7.2
800
-------
Figure 1. Acid Aerosol Exposure System-Front View. Acid aerosol will be generated from a
© pressurized nebulizer fed with acid solution by a liquid metering pump. Electrical
charges on the aerosol will be neutralized by a ® 85Kr source, ® diluted, and delivered
to the ® exposure chamber. The acid aerosol is removed from the chamber exhaust by
a ® water scrubber. The scrubbing liquid is re-circulated through a ® reservoir.
Figure 2. Acid Aerosol Exposure System - Side View.
801
-------
The Atmospheric Transformation of Automobile Emissions and Measurement of the Formation of
Mutagenic Products
T.E. Kleindienst, D.F. Smith, E.E. Hudgens, R.F. Snow
NSI Technology Services, Inc. Environmental Sciences
Research Triangle Park, NC 27709
E. Perry
Environmental Health Research and Testing
Research Triangle Park, NC 27709
L.T. Cupitt, L.D. Claxton
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Automobile emissions were irradiated in a 22.7-m3 environmental chamber. The hydrocarbons
for the irradiated mixture was taken from the exhaust (50%), and from a seven-component surrogate
mixture (50%), representative of evaporative emissions from automobiles. NOj came entirely from the
exhaust and was brought to an initial concentration of 516 ppbv in the environmental chamber before the
irradiation. The irradiation was conducted at an initial HC/NO, ratio of 15. The mutagenic activity of
the mixtures were measured using a variant of the Ames test. In this procedure, the bacteria, Salmonella
typhimurium, strain TA100, was exposed in individual exposure chambers to the irradiated and non-
irradiated effluent. The results show that the gas-phase mutagenic activity of the products is significantly
greater than that of the reactants with a doubling of the irradiated mixture (over background activity) after
four hours of exposure. Collection of the paniculate matter from the reactant or irradiated mixtures did
not yield sufficient mass to obtain either chemical or mutagenicity data above background levels.
Experiments at higher concentrations were thus conducted to obtain the mutagenic activity of the
paniculate phase components. The measured activities of the gas phase (43 revertants/h) and the
paniculate phase (0.6 revenants//ig) were convened to activities per unit volume, that is, the mutagenic
density which is the number of revertant/m3 of effluent. (The mutagenic density includes the potency of
chemical mutagens in the mixture and the amount of these compounds present.) For the gas phase
products the measured mutagenic density was 2730 revertants/m3; the mutagenic density for the paniculate
phase products was approximately 7 revertants/m3. The difference in mutagenic density for the two phases
mostly represents the much higher mass of products present in the gas phase.
802
-------
Introduction
Automobiles emit a significant fraction of the urban levels of hydrocarbons and NOr Automotive
hydrocarbon emissions result from exhaust emissions, running losses, and other evaporative losses. In
urban environments many emission sources, in addition to automobiles, contribute to the total
hydrocarbon loading. On the other hand, automobiles generate a large fraction of the nitric oxide found
in urban areas and on average accounts for around 60% of th? urban loading for oxides of nitrogen [1].
This is due in part to the fact that most control strategies for reducing urban ozone has focused more on
automotive hydrocarbon control than on NOj control.
Several studies have been conducted to examine the effect automotive emissions have on the
generation of photochemical pollutants, particularly the production of ozone. These studies have generally
involved experiments performed in photochemical smog chambers, with quantitative measurements for the
removal of the reactants and the generation of products. \VTule some of the information from these
studies could be used in risk assessment models, the studies are generally not designed, and are certainly
not optimized, for this purpose.
In addition to ozone control, the biological effects of automobile exhaust are also of concern.
Most current assessments of risk attributed to automobile txhaust are based on the composition of
cliemicals found in direct tailpipe emissions rather than the photochemical transformation products.
However, evidence has accumulated to suggest that products generated from the photochemical trans-
formation of the components of combustive emissioas may ^ntribute more significantly to potential
health risk than do the direct emissions themselves [2],
Over the past few years research studies from several laboratories have examined mutagenic
activities attributed to the photooxidation products of various atmospherically important hydrocarbons,
particularly olefins and aromatics. The mutagenic activity of the gas-phase oxidation products of individual
hydrocarbons (notably, propylene [3], and toluene [4,5]) have been measured and it has been
determined that the gas-phase photooxidation products from these reactions can be significantly more
mutagenic than are the initial reactants (i.e., notably the hydrocarbon, NO, and NO2). In some of these
studies extensive efforts have been made to identify the origin of the observed activity in terms of
mutagenic contributions from individual oxidation products. While the approach has met with limited
success, the results have generally indicated that a large fraction of the activity in the toluene and
propylene oxidation systems is probably attributable to products (perhaps nitrogenated products) formed
in low yields in the system. Experimentally measured compounds contributing to the observed activity
included peroxyacetyl nitrate (PAN), the dicarbonyls- glyoxal and methyl glyoxal, and several bifunctional
organic nitrates [3].
This study has been conducted to evaluate potential mutagenicity from the components of
automobile exhaust and the associated photooxidation producis. The results of the study are expected to
indicate the extent to which there are components in automobile exhaust which lead to the formation of
mutagenic products by photooxidation processes. In addition, with current interest in alternate fuels and
reformulated gasolines, the study will give baseline data from which to compare the impact these fuels and
their oxidation products might have with respect to the generation of mutagenic compounds.
Experimental Methods
The main components of the apparatus include: (1) a portable dynamometer and associated
dilution tunnel, (2) a photochemical reaction chamber, and (3) biological exposure chambers. Automobile
ethaust was generated on-site by operating a light-duty passenger vehicle under load on a portable
dynamometer. The dynamometer and accessory equipment necessary for its operation were mounted on
a single flatbed trailer for turn-key operation at a remote site. A single automobile, a 1977 Ford Mustang
with 40,000 miles, was employed for this study. The entire exhaust from the automobile passed through
a heated, stainless-steel transfer line of 7.6-cm internal diameter which was connected to the dilution
tunnel. The exhaust was diluted with ambient air that was driwn through the tunnel using a 20 m3/min
turbine. Effluent from the tunnel then passed through a 7-m length of 3.2-cm Teflon tubing to reach the
point where the dilute exhaust could be accurately meiered into the reaction chamber.
The reaction vessel is a 22,700 L cylindrical chamber constructed of 5 mil Teflon film. The
803
-------
reaction chamber is sealed to aluminum end plates that are coated with fluorocarbon paint. The radiation
is generated by a combination of fluorescent sunlamps and blacklights that surround the chamber
longitudinally. NO2 photolysis rate, kj, as previously determined experimentally, was 0.22-0.24 min"1.
Reactants introduced into the reaction chamber first passed through a 150 L inlet manifold where they
were mixed with ultra-zero air. Dilution air added to the inlet manifold was supplied using a high volume
clean air generator.
Four 190 L, Teflon-coated biological exposure chambers were used for the exposure of the gas-
phase test mixtures to the biological assay. The activity of irradiated gas-phase effluent was measured in
an exposure chamber situated in close proximity to the reaction chamber to minimize transfer losses for
reaction products of low volatility. The gas-phase mixture entering this exposure chamber was filtered
with a 135 mm Teflon-impregnated glass fiber filter. The exhaust from this chamber was then fed into
a second exposure chamber in series. This chamber, the carry-over chamber, was employed to determine
the extent to which mutagenic compounds pass through the initial exposure chamber. Also included were
chambers to measure the activity of the reactants and the background activity of clean air. All exposure
chambers were operated at a flow of 14 L/min.
The bacteria Salmonella typhimurium, strains TA100 and TA98 were employed as the biological
assay for this study. 45 ml of base agar (Difco) was poured into glass petri dishes. The plates were seeded
with bacteria using 0.1 ml of the S, typhimurium culture containing 1 x 108 bacteria mixed in a 3 ml of an
agar overlay. The overlay was poured at a temperature of 45 °C. In formulating these plates minimal
histidine at the same final concentration as prescribed was placed in the base agar rather than in the top
agar. Since only direct activity was determined in these experiments, metabolic activating substances such
as S9 were not added to the active mixture. Colony counting of the expressed bacteria was performed
using a Biotran III automatic colony counter. All other procedures using the biological assay followed
those previously described [6].
The reacting hydrocarbons for this study derived from two sources, the automobile exhaust and
a hydrocarbon surrogate mixture. The gasoline used to power the automobile in this study was a
commercially available winter-grade unleaded fuel. The surrogate component was derived from a seven-
component chemical mixture representative of vehicular evaporative losses.
The experimental procedures generally followed those previously described for gas phase exposures
of both simple hydrocarbons, as well as those for other complex mixtures. The procedures for these
experiments were refined to address the constraints of this combustion source with respect to viable
operation of the reaction chamber. Automobile exhaust and the surrogate hydrocarbon mixture were
admitted into the chamber using the following procedure. The automobile was operated using a 23
minute driving cycle followed by a ten minute idle. The automobile exhaust from the dilution tunnel, was
mixed with zero air in the inlet manifold and injected into the reaction chamber at the requisite flow. The
integrated HC/NO, ratio of the automobile exhaust over the complete 33 minute cycle was between 7 and
8. The exhaust was metered into the inlet manifold at a rate necessary to yield an integrated NO, chamber
concentration of approximately 500 ppbv. The total hydrocarbon contribution from the exhaust amounted
to 3.75 ppmC, approximately 40% of which was generated during the driving cycle and 60% during the
automobile's idling period. The remaining 3.75 ppmC was provided from the evaporative surrogate
mixture by bubbling nitrogen at a flow of about 0.5 ml/min through the liquid surrogate mixture and
feeding the headspace vapor simultaneously with the exhaust into the inlet manifold.
The irradiation was conducted by bringing the reaction chamber up to the desired reactant
concentrations with automobile exhaust and the surrogate mixture. The irradiation began by operating
the chamber in a static mode to reach the desired extent of reaction. The initial mixture was irradiated
until the ozone maximum was attained which required six to eight hours. When the ozone maximum
concentration was reached, the operation mode for reaction chamber was changed from static to dynamic
mode. This change required a short transition period (ca. 2 h.) for the effluent to reequilibrate to the
exact conditions of the experiment. Following the transition period the biological exposure chambers were
filled with approximately 25 plates each of TA100 and TA98. Also included were survivor plates for each
exposure condition. These plates served as an initial screen to evaluate the degree to which the biological
assay might experience toxicity from the test mixtures. The exposure chambers were flushed with the
appropriate effluent until the concentrations in the exposure chambers reached a steady state value. The
petri dish covers were then removed beginning the exposure. Predesignated groups of test and survivor
plates were covered during the course of the exposure to generate a dose-response relationship. For
804
-------
statistical purposes, 7-8 plates were employed for a single exposure. Before and during the exposure,
chemical analyses of the components in the reaction chamber, exposure chambers, and mixing manifold
were obtained at regular intervals.
Results and Discussion
Initial conditions for the photooxidation experiment are shown in Table I. An initial
concentration of 516 ppbv was used. This concentration lies at the upper end of ambient
concentrations which have been measured [7]. NO, generated in automobile combustion is largely
composed of NO and in this experiment NO comprised 88% of the total NOr initially present. The
values for NO and NO, were determined from a weighted contribution of the reactant material present
at the start of the static irradiation and that added during dynamic mode operation. Although the
concentrations of added exhaust components were not constant with time, the car was operated on a
driving cycle giving a cyclic and reproducible input of reactants.
Table I also gives product information for the expeiiment before and following irradiation. The
residence time of the mixture in the reaction chamber was 6.3 hours. The dominant products detected
other than ozone were aldehydes and to a lesser extent peroxyacyl nitrates for which aldehydes are often
the dominant precursors. Formaldehyde and acetaldehyde increased by at least an order of magnitude by
the time steady-state was reached. Most of the other carbonyl and dicarbonyl compounds increased from
below the detection limit to concentrations between 5 and 50 ppbv. PAN was present at a concentration
of 88 ppbv. Peroxypropionyl nitrate (PPN) and peroxyburyryl nitrate (PBN) were also detected on the
same chromatogram as was PAN. Although standards were not available for the calibration of PPN and
PBN, their concentrations based on the instrumental sensitivity for PAN could be estimated. PPN
amounted to approximately 9% of the PAN concentration and PBN to approximately 6%.
The activity measured in TA100 for the reactant and product mixture is shown in Figure 1. The
mutagenic activity of the reactants (dotted line) showed an activity of 8 revertants/h, although this activity
was not reproducible in other measurements of the reactant mixture. The exposure of the photooxidation
products to TA100 gave a substantial increase in the mutagenic activity above that of the reactants. As
given in Figure 1, the slope of the dose-response curve (solid line) produced an activity of 43 revertants/h.
The dashed line in Figure 1 represents the activity of the carry-over component of the effluent and gave
a value of 21 revertants/h.
The mutagenic activity of the gas phase can also be placed on an air volume basis, that is, the
number of revertants generated from gas-phase mutagenic species per m3 of effluent. The calculation is
made using the following relationship [8],
Mp; (rev/m3) = m^ (rev/plate-h) x N (plates) / V^ (m3/h) (I)
where M^ is the mutagenic density of the gas phase, m^ is the mutagenic activity of the gas phase as
given above, N is the total number of plates in the exposure chamber, and V is the volume of effluent
passing through the chamber per hour of exposure. For the reactants with an activity of 8 rev/plate-h the
calculated mutagenic density is 510 revertants/m3. For the effluent with an activity of 43 rev/plate-h, an
activity of 2730 revertants/m3 is determined.
The number and volume distributions for the paniculate matter were measured before the
irradiation and at the maximum in the ozone concentration. Before irradiation, the integrated number
distribution (between 0.01 and 1 jim) was 1.3 x 103 particles/cc. This value rose following irradiation to
1.1 x 104 particles/cc and suggested the occurrence of secondary particle formation [4]. The total volume
of the particles between 0.01 and 1 fj.rn increased by a factor of five following irradiation from 0.7 nl/m3
to 3.6 nl/m3.
Paniculate matter from the products collected on filters during the exposures was insufficient to
obtain reliable identifications of individual chemical components by GC/MS. (The paniculate
concentrations of the reactant mixtures were already extremely low.) This was indicated by filter weights
and the total volume measurements of the EAA which is related to the total mass between 0.01 and 1.0
fan. An analogous problem was encountered for the biological analysis. Given the low absolute amount
805
-------
of sample collected and the statistical uncertainties using the Ames assay, the measured revertant levels
were not statistically different from the background revertant levels. This occurred for both the reactants
and effluent. However, an experiment was conducted using reactant concentrations an order of magnitude
higher than that of the previous experiments. In this experiment the initial total hydrocarbon value was
80 ppmC and the initial NOX value was 5 ppmv. Photooxidation of this mixture yielded values for the
mutagenic activity of the reactants (1.1 rev//xg) and products (0.6 rev//ig). Given the approximate
paniculate concentrations in the effluent of 2 /xg/m3 for the reactants and 11 /ig/m3 for the products, this
lead to mutagenic densities for the paniculate phase of 2.2 rev/m3 for the reactants and 6.6 rev/m3 for the
product mixture.
Thus, for the reactants and products in this experiment, the mutagenic density of the gas-phase
components outweigh the contributions of the paniculate-phase components. However, it must be
recognized that this report represents a single experiment employing a single vehicle, using a single fuel
and operated under a single driving cycle; any generalizations from the results of this study must take this
into account and may not be applicable to different conditions.
Disclaimer
Although the research described in this article has been funded wholly or in part by the U.S.
Environmental Protection Agency through Contract 68-02-4443, it has not been subject to the Agency's
peer and policy review. Therefore, it does not necessarily reflect the views of the Agency, and no official
endorsement should be inferred.
References
1. Atkinson, R. 1988. Atmospheric Transformation of Automotive Emissions. In Eds. A.Y. Watson,
R.R. Bates, and D. Kennedy, "Air Pollution, the Automobile, and Public Health", National
Academy Press, Washington, D.C.
2. Shepson P.B., Kleindienst T.E., and Edney E.O. 1987. The production of mutagenic compounds
as a result of urban photochemistry. EPA-600/3-87-020. U.S. Environmental Protection Agency,
Research Triangle Park, NC.
3. Kleindienst T.E., Shepson P.B., Edney E.O., Cupitt L.T., and Claxton L.D. (1985). The mutagenic
activity of the products of propylene photooxidation. Environ. Sci. TechnoL, 19, 620-627.
4. Shepson P.B., Kleindienst T.E., Edney E.O., Namie G.R., Pittman, J.H., Cupitt L.T., and Claxton
L.D. (1985) The mutagenic activity of irradiated toluene/NOx/H2O/air mixtures. Environ. Scl
TechnoL, 19,249-255.
5. Shiraishi F., Hashimoto S., and Bandow H. 1986. Induction of sister chromatid exchanges in
Chinese hamster V79 cells by exposure to the photochemical reaction products of toluene plus
NO2 in the gas phase. Mutat. Res. 173, 135-139.
6. Claxton L.D., Toney S., Perry E., and King L. 1984. Assessing the effect of colony counting
methods and genetic drift on Ames bioassay results. Environ. Mutagen. 66, 331-342.
7. Finlayson-Pitts, B.J. and Pitts, J.N. (1986) "Atmospheric Chemistry," Wiley-Interscience,
New York, NY, pp. 368-369.
8. Kleindienst T.E., Shepson P.B., Edney E.O., Cupitt L.T., and Claxton L.D. 1986. Wood smoke:
Measurement of the mutagenic activities of its gas-and particulate-phase photooxidation products.
Environ. Sci, TechnoL 20, 493-501.
806
-------
Table I. Concentrations of measured reactant and product species from the photooxidation of
automobile exhaust and evaporative surrogate mixture.
Species
(initial)
NO
NO,
THC.ppmC
HCHO
CH3CHO
HC/NO^
Cone
(ppbv)
454
516
7.9
16
3.6
15.3
Species
(final)
NO
NOX
THC.ppmC
HCHO
CH3CHO
03
PAN
'Cone
(ppbv)
0
375
5.7
176
89
516
88
Species
(final)
CH3C(O)CH3
MEK
C3H7CHO
C4H9CHO
C6H5CHO
CHOCHO
CH3CHO
j
Cone
(ppbv)
54
25
11
8
2
11
7
Figure 1. TA100 activity from automobile emissions and their irradiated products
750
1
03
r
tu
CD
500
250
Effluent
Carry Over
Reactants
4 6
Exposure Time (h)
8
10
807
-------
VALIDATION OF METHODOLOGY FOR DIRECT BIOASSAY
OF ATMOSPHERIC VOLATILE ORGANIC COMPOUNDS
C. M. Sparacino*, T. J. Hughes, N. P. Castillo, J. Warner, C. Rawn
Research Triangle Institute
Research Triangle Park, NC
C. Pietarinen
New Jersey Department of Environmental Protection
Division of Science and Research
Trenton, NJ
The Ames/Salmonella assay has been used to detect mutagenicity of
organic extracts of ambient air particulate and this mutagenicity has been
correlated with industrial activity. Little is known about the potential
mutagenicity of volatile organic compounds (VOCs) because adequate bioassay
techniques have not been available. A need therefore exists to directly
assess the mutagenic activity of ambient air VOCs. Studies were conducted
to validate methodology developed for the bioassay of ambient air VOCs as
collected via Tenax cartridges.
Tenax-based collection and chemical analysis of VOCs are well-estab-
lished methods, and were used, with modifications, for this work. A large
glass cartridge containing 12 g of Tenax was used in order to sample suffi-
cient quantities of air for bioassay detection of ambient levels of most
VOCs. A technique for measuring the mutagenic activity of volatile com-
pounds was recently developed, and was validated with known quantities of
specific target compounds. The two procedures, i.e. Tenax collection and
VOC bioassay, were combined as a single technique that will permit direct
assessment of the mutagenic potential of atmospheric samples. Four repre-
sentative VOCs (halogenated and epoxy organics) were individually spiked
into large cartridges and desorbed into the Tedlar bag bioassay system.
Response curves were obtained for duplicate experiments for each compound.
The results showed good correlation between spiked quantities and mutagenic
response; acceptable precision was demonstrated for each VOC. Dose ranges
for each VOC were at levels of several hundred micrograms, which can be
accommodated by the cartridges used for this study.
808
-------
Introduction
Background
The Ames Salmonella assay has been shown to detect animal genotoxic
substances as mutagens in a cost- and time-efficient mannerl.2. The assay
has been utilized to detect mutagenicity of organics from ambient air par-
ticulate and the mutagenicity has been correlated with industrial acti-
vity3. The literature contains many references on the use of the Ames
assay for similar studies, but little is known about the potential mutage-
nicity of volatile organic compounds (VOCs) because adequate bioassay tech-
niques have not been available. The volatile component of ambient air
pollutants represents a potentially greater human health risk than conden-
sed phase materials since their atmospheric load is more than an order of
magnitude greater^. Thus a means of collecting and testing VOCs using the
same general Ames system that is used for other pollutant mixtures would
provide a useful tool for assessing total mutagenicity/carcinogenicity of
collected air samples.
The interface of volatile (boiling point < ~ 130*C) and, in many cases,
lipophilic compounds with an aqueous based assay conducted at elevated
temperatures presents many problems. These problems were largely circum-
vented following research efforts at the RTI and EPA laboratories some
years ago^. An assay method was developed that utilized a closed system
(Figure 1) wherein vapor phase substances were equilibrated and incubated
with standard pour plates. The method uses a small Tedlar bag (~ 500 mL)
fitted with a septum-containing inlet, and containing one (or more)
standard pour plate.
To serve as a practical means of assessing mutagenicity of ambient air
levels of VOCs, it is necessary to effect an interface with a VOC collec-
tion method. One of the most effective of the many existing methods is the
use of Tenax polymeric sorbent for entrapment of VOCs from air. This
methodology was developed at RTI, and is well validated, having been used
for the collection and analysis of several tens of thousands of samples.
An obvious choice then for the overall methodology of collecting and deter-
mining the mutagenic potential of ambient air VOCs involved a combination
of Tenax based sampling and Tedlar bag bioassay. The validation of the
combined techniques is the subject of this manuscript.
Experimental Approach
The general scheme for method validation involved loading Tenax
cartridges with known amounts of representative VOCs, and thermally desorb-
ing the VOCs into a Tedlar bag bioassay system. Mutagenic response was
assessed by recording the number of revertants at several dose levels, and
comparing these data with experiments involving direct injection of known
quantities of VOCs into the Tedlar bags. The study was limited to VOCs
that were known to yield linear responses in doses ranging from several
hundred to over 1000 /ig. These ranges were required by the sensitivity of
the bioassay test system, which then also dictated the amount of Tenax
needed for the study. To trap hundreds of .nicrograms of airborne VOCs
requires the use of a large cartridge with a capacity of more than 10 g of
Tenax. Optimization studies were conducted to establish effective trap-
ping/desorbing/analysis conditions for the large cartridge system.
Replicate experiments were conducted with each of four VOCs. Compounds
studied to date include epichlorohydrin (EP3), ethylene dibromide (EDB),
809
-------
propylene oxide (PO) and butylene oxide (BO). Recovery studies were
carried out for two of these analytes. Precision and recovery data, and
results from the comparison of the Tenax-introduced versus the direct
injection methods, provided the requisite procedure validation.
Experimental Methods
Tenax Methodology
The general methodology for Tenax sampling and analysis of ambient air
has been described elsewhere, most recently by Pellizzari et al.6 All work
detailed in this report for Tenax preparation, loading and desorption
followed the published methods. Minor modifications were incorporated into
the method because of the use of a larger cartridge, and to allow direct
injection of desorbed VOC into a Tedlar bag. A schematic of the desorption
system is shown in Figure 1. Glass cartridges were fabricated to contain
~12 g of Tenax (35-60 mesh). Cartridges were 3 cm wide, 21 cm long, and
contained tapered ends (to 1.5 cm width) to accommodate existing connectors
that are used routinely for analytical Tenax cartridge work. A thermal
desorption unit was also fabricated to contain the larger cartridge; this
unit is identical, with the exception of scale, to analytical cartridge
units. A small bore access port was fashioned into the desorption chamber
cap to allow direct injection (syringe) of test compounds for recovery
studies. For introduction of desorbed/cryotrapped compounds into the
Tedlar bag system, a short section of heated SS tubing (~ 150 x 1 mm OD)
was led from the valve (Figure 1) and coupled with a 5 cm section of Teflon
tubing. A SS needle (21 gauge) was inserted in the end of the Teflon
tubing for injection of the VOC into the Tedlar bag. For desorption
studies, ethanolic solutions of test compounds were spiked onto Tenax
cartridges via a flash evaporation system. Routinely, 1-3 ^L of solution
were used for preparing all dose levels.
Bioassay Method
The methodology utilized for all Tedlar bag bioassay work has been
describedS, in order to maintain the integrity of the closed test system
(Figure 2), each bag was sealed with tape, then leak-checked with a leak
detector. Each bag, prior to the introduction of Tenax-desorbed compound,
was physically compressed to remove as much internal air as possible. This
allowed for the ready accommodation of 250-350 mL of vapors from the
desorption system. For this study, only a single tester strain of bacteria
(TA100) was used, without S9 activation. Positive controls consisted of
propylene oxide (500 /Jg) , 2-anthramine (2.5 /ig) and sodium azide (2.5 yjg) .
Blanks consisted of bacteria with and without ethanol. Bags were incubated
for 48-56 hours in a 37*C incubator. Colonies were counted with an
automatic colony counter.
Results
Optimization/Recovery Studies
The general conditions for trapping and desorbing the smaller, analy-
tical-scale Tenax cartridges were used as a starting point for establishing
similar conditions for the large cartridges. Although a trap constructed
from capillary tubing is used for analytical work, the larger quantities of
desorbed components for this study led to the use of a large bore (2.1 mm
ID) stainless steel trap. Several experiments were conducted to determine
the optimum trap configuration. All such work utilized liquid nitrogen as
810
-------
a coolant for the trap. Based on previous wcrk conducted at RTI and by
ethers, glass beads were also used in the optimization studies. The trap
geometry finally adopted for all validation work included a "U" shaped
section of tubing, 30 cm in length, containing a 10 cm bed of glass beads
(200 mesh). Although initial work utilized a boiling water bath for
desorption of trap contents, a higher temperature was shown to be more
effective. Final overall trapping/desorption conditions are shown in Table
I; also included in the table are the GC conditions used in the recovery
studies.
Recovery data was generated for the first two compounds studied,
epichlorohydrin (EPC) and ethylene dibromide (EDB). Desorbed compounds
were diverted to a capillary GC system (Table I) for flame ionization
detection and data acquisition. All recovery experiments were conducted by
comparison of area counts for peaks associated with desorbed material from
spiked cartridges, with area counts from similar quantities of material
injected directly into the desorption chamber. For EPC, an initial experi-
ment was conducted at a single dose level (500 /Jg); for EDB, recoveries at
three dose levels were determined. Experiments were conducted in tripli-
cate or quadruplicate for the test compounds. Given the excellent preci-
sion associated with direct injection for the EPC work, only single injec-
tions were made for the EDB study. As shown, the recoveries were high
(average ~ 83 Z) and were virtually independent of compound type or dose
level.
Bioassay Studies
Mutagenecity results for each of the four compounds studied are shown
in summary form (Table III) and as plots (Figures 3 and 4) of response
versus dose. Each plot contains data for results from directly injected
and Tenax-introduced test compound. Dosing levels, response ranges, and
pertinent statistical data are shown in Table III for all four compounds.
Each figure shows two plots corresponding to each of two separate
experiments. In general, linearity of response was quite good, although in
some instances, e.g. EDB and BO, deviation from linearity at higher dose
levels was evident. The mean correlation coefficient for all sixteen
experiments was approximately 0.93, while overall precision (mean of 2RSD
for all data points for all sixteen runs) was less than 12 ZRSD. In all
cases, mutagenicity for Tenax-introduced assays were less than for direct
injection assays. This is presumably due to simple differences in recovery
far the two modes of introduction of test compound. The data reflect very
adequate responses and show acceptable levels of linearity and precision.
Canclusions
The procedure for introduction of VOCs into a previously validated
closed mutagenicity bioassay system using Tenax desorption techniques has
been shown to be effective, precise and reliable for recovery and detection
of mutagenic VOCs. The successful results obtained for four mutagenic and
volatile test compounds serves to validate the desorption/bag interface,
and thus represents a possible method for the direct assessment of mutage-
nic potential of ambient or other types of atmospheric samples. Field
validation experiments using the combined techniques are in progress.
Acknowledgement
This research was supported by the New Jersey Department of Environmen-
tal Protection, Division of Science and Research, under contract number
811
-------
P31010. For helpful advice and discussions on the design and conduct of
the research described here, the contributions of J. Louis and S. Qwan of
that Division are gratefully acknowledged.
References
1. D. M. Maron, B. N. Ames, "Revised methods for the Salmonella muta-
genicity test," Mutation Res. 113: 173 (1983).
2. R. W. Tennant, B. Margolin, M. Shelby, E. Zeiger, J. Haseman, J.
Spalding, ¥. Calspary, M. Resnick, S. Stasiewicz, B. Anderson, R.
Minor, "Prediction of chemical carcinogenicity in rodents from in
vitro genetic toxicity assays," Science 236: 933 (1987).
3. T. J. Hughes, E. Pellizzari, L. Little, C. Sparacino, A. Kolber,
"Ambient air pollutants: collection, chemical characterization and
mutagenicity testing," Mutation Res. 116: 51 (1980).
4. R. A. Duce, "Speculations on the budget of particulate and vapor-
phase nonmethane organic carbon in the global troposphere," Pur_e_
Appl Geophys.. 116: 244 (1978).
5. T. J. Hughes, D. M. Simmons, L. G. Monteith, L. D. Claxton,
"Vaporization technique to measure mutagenic activity of volatile
organic chemicals in the Ames/Salmonella assay," Environ. Muta-
genesis, 9: 421 (1987).
6. E, D. Pellizzari, K. ¥. Thomas, Total Exposure Assessment Methodo-
logy (TEAM); 1987 Study in New Jersey, Parj: II; Protocols, EPA
Contract No. 68-02-4544, Work Assignment No. 8, October, 1987.
812
-------
Table I. Final system operating conditions.
GC Conditions
Desorptlon Parameters
Purge Gas Flow Rate: 50 mLVmin
Desorption Chamber Temperature: 270 °C
Desorption Time: 15 min
Trap Temperature (collection): -196*0
Trap Temperature (purge): 200 CC
Purge Time: 5 min
Trap Design: ss "U" tube, 300 x 2.1 mm ID,
with 10 cm bed of glass beads, 200 mesh
Column Type: Fused Silica, SE-54
(30 m x 0,32 mm ID)
Column Temp (background check):
45 °C (5 min), 45 -160 °C @ 40/min,
160°C(5min)
Column Temp (compound analysis): 45 °C
Carrier Gas Flow Rate: 0.96 mL/min
Injector Temp: 160°C
Detector: FID
Detector Temp: 300 °C
Split Ratio: 47:1
Table II. Recovery data.
Peak Area Counts (xlO4) Recovery
Compound (Amount) Cartridge
Direct Injection
Epichlorohydrin
(500 ng)
Ethylene Dibromide
(600 jig)
Ethylene Dibromide
(840 jig)
Ethylene Dibromide
(1080 jig)
604.6
522.8
612.0
579.8 (8.5% RSD)
201.0
208.9
183.7
190.2
196.0 (5.7% RSD)
285.6
276.7
251.1
28Q.Q
273.4 (5.6% RSD)
368.5
342.8
336.0
530.4
344.4 (4.9% RSD)
730.6
741.1
734.4
735.4 (0.7% RSD)
78.8
227.4
326.7
86.2
83.7
417.6
82.5
813
-------
Table III. Bioassay test data.
CO
Dose Response
Compound Range dig) Range (rev.) Linearity^ Precisionb
EPC 100-600 -500-1400 0.933 -11% RSD (Dir. Inj.)
-13% RSD (Tenax)
EDB 500-1080 -300-700 0.877 -7% RSD (Dir. Inj.)
~12%RSD(Tenax)
PO 400 - 960 -500 - 1100 0.944 -11 % RSD (Dir. Inj.)
-9% RSD (Tenax)
BO 600-1200 -400-800 0.963 -5% RSD (Dir. fnj.)
-15% RSD (Tenax)
aAverage correlation coefficient from least squares regression for all 4
runs.
bAverage of all response values for the two Dir. Inj. experiments and the
two Tenax experiments.
-------
Figure 1. Closed bioassay system for organic volatiles.
Purge
Gas
Thermal
Desorption
Chamber
6-Port Vatve
(heated)
Vent
m
Cryotrap
(w/heater)
'Carrier
Gas
Capillary
GC
Detector
(MS, FID, etc.)
Figure 2. Schematic of Tenax desorption system.
815
-------
Epichlorohydrin (1)
0>
DC
1400
1200 -
1000 -
800 -
600 -
400 •
200 -
0
Epichlorohydrin (2)
Dir
100 200 300 400 500 600
100 200 300 400 500 600
CO
CO
c
to
C
o
0)
DC
Ethylene Dibromide (1)
Ethylene Dibromide (2)
1400
200 400
600 800
Dose (jig)
1000 120
200 400
600 800
Dose (ng)
1000 1200
Figure 3. Bioassay results - duplicate experiments.
-------
Propylene Oxide (1)
tn
**
to
K
o»
DC
Butylene Oxide (1)
ROO
<' 600 -
200 400 600 800
Dose (|ig)
1000 1200
Butylene Oxide (2)
200
400 600 800
Dose (|jg)
1000 1200
-------
ISOLATION OF MUTAGENS IN AMBIENT AIR PARTICULATE EXTRACTS USING BIOASSAY-
DIRECTED FRACTIONATION
Daniel J. Thompson
NSI Technology Services Corp.
Research Triangle Park, NC
Ron Williams, Lance Brooks
EHRT
Durham, NC
Douglas A. Bell
National Research Council, U.S. EPA
Research Triangle Park, NC
Marcia G. Nishioka
Battelle Columbus Division
Columbus, OH
The bioassay-directed fractionation of mutagens in Boise, ID ambient air
particulate extracts was initiated using two levels of fractionation.
First, two composite samples, a wood smoke composite (WSC) and a mobile
source composite (MSC), were fractionated using nonaqueous anion solid
phase extraction (anion-SPE). In five runs, recoveries of mass and
mutagenicity from both samples were consistently good with ca. 85% to 95%
of the mass and ca. 85% to 100% of the mutagenicity being recovered. A
standard mixture containing basic, neutral and acidic compounds was also
fractionated to verify acid/neutral-base separation. The anion-SPE strong
acid fraction was chosen for further component separation using normal
phase cyanopropyl HPLC (CN-HPLC). In the second level fractionation,
quantitative recoveries of both mass and mutagenicities were recorded. The
preliminary data describing the distribution of mass and rautagenicity
between fractions In both levels of fractionation are described.
Introduction
Isolation and identification of potential cancer causing compounds from
the innumerable components of ambient air complex mixtures is, at best, a
formidable task. However, by using the process of bioassay-directed
fractionation, mutagens may be isolated from the non-mutagens by developing
methods that target those sample fractions that exhibit the greatest
mutagenicity.1 The general approach is to initiate a multi-level
fractionation based upon the broad chemical properties of the sample
components. In each successive level, the most mutagenically potent
fractions are chosen for additional, more selective fractionation. The
principal goals for all developed methods are quantitative mass and
mutagenicity recoveries, Isolation of mutagen enriched fractions, and the
subsequent identification of the mutagenic compounds.
We initiated our fractionation scheme in two distinct fractionation
levels that separate sample components based on ionic character and
818
-------
polarity (Figure 1). For the first level we developed a method similar to
that reported by Bell et al.2 for isolating acidic sample components from
neutral and basic components. For the second level, we developed a
fractionation procedure that separated the sample into nonpolar, moderately
polar and polar fractions. Reported here are the preliminary data on the
fractionation of two ambient air particulate extract composite samples and
a standard mixture.
Experimental
Reagent*;
All solvents, methanol (MeOH), dichloromethane (DCM) and hexane, were
obtained from Burdick and Jackson (Muskegon, Mi). The anion exchange
resin, AG MP-1, was obtained from Bio-Rad (Richmond, CA). The
trifluoroacetic acid (TFA) was obtained from Aldrich (Milwaukee, WI) and
the C02 was obtained from National Specialty Gases (Raleigh,NC).
Samples
Standard mixture. The standard mixture (STD), in DCM, was composed of
pyrene (1.81 mg/mL), 1-naphthol (1.47 mg/mL), benzoic acid (0.57 mg/mL),
7,8-benzoquinoline (1.46 mg/mL), 2-ni trob£:nzoic acid (1.57 mg/mL), and 4-
nitrophenol (1.37 mg/mL), all obtained from Aldrich (Milwaukee, WI). The
mixture also contained 1-nitronaphthalene (1.61 mg/mL) from Chem Service
(West Chester, PA).
Composite Samples.5 The composite samples were prepared from the
extractable organic matter (EOM) of 100 Boise, ID, Elm Grove Park (EGP) and
Fire Station samples (50 each divided between daytime and nighttime). A
single element tracer multi-linear regression (MLR) model was applied to
the EOM mass of each sample using soil corrected fine particle potassium
and fine particle bromine as tracers for wood combustion and mobile
sources, respectively. Samples showing the greatest wood smoke impact were
combined until a total of ca. 2.5 grams had been gathered to form the wood
smoke composite (WSC) sample. The mobile source composite (MSC) sample was
prepared in the same manner using samples more heavily impacted with mobile
source combustion. Based upon trace modeling, the percent wood smoke
contribution to the final two composite samples were ca. 89% for the WSC
and ca^ 64% for the MSC.
Level 1: Nonaqueous Anion Solid Phase Extraction (Anion-SPE)
Resin Preparation. The AG MP-1 resin (100 - 200 mesh) was prepared for
use by washing 200-300 mL resin with 300 irL 1% TFA/MeOH for 10 min, then
500 mL MeOH for 10 min, followed by 500 ml, 0.10 M KOH/MeOH solution for 10
min, then another 10 min MeOH wash and finally rinsed with copious amounts
of deionized water. The resin was then Soxhlet extracted sequentially in
500 mL MeOH for 16 hours, 850 mL DCM for 24 hours and finally 850 mL MeOH
for 24 hours. A portion of the clean resin (15 mL) was then placed into a
30 cc syringe plugged with silane treated glass wool. The syringes were
then placed onto a Baker solid-phase extraction vacuum manifold (SPE 10
system).
Fractionation Procedure. Samples, ca. 20 to 30 mg in DCM, were
concentrated and placed directly onto the resin in 250 uL aliquots per
column. The five collected fractions were each eluted with 24 mL aliquots
of the respective solvent. The elution order was DCM, MeOH, C02/MeOH
(prepared by bubbling C02 in methanol for 30 min), 2% TFA/MeOH and for runs
819
-------
3 through 5 an additional 10% TFA/MeOH fraction was eluted. The DCM, MeOH
and C02/MeOH fractions were rota-evaporated to ca. 2 mL. To ensure adequate
removal of the TFA, we took advantage of the TFA/MeOH azeotrope; the 2% and
10% TFA/MeOH fractions were diluted to ca^. 50 mL with MeOH, rota-evaporated
to ca. 5 mL, rediluted to ca. 50 mL and concentrated to a final volume of
ca. 2 mL, Finally, all samples were quantitatively transferred to a 10 mL
volumetric flask and diluted to volume with DCM.
Reversed Phase HPLC. The HPLC experiments were performed with a
Hewlett Packard Model 1090 instrument in conjunction with a Valco
Instruments six-port HPLC valve and a Waters Model 440 UV detector (254
nm). A Du Pont (Wilmington, DE) Zorbax ODS reversed-phase HPLC column
(0.46 x 25 cm; 10 /im particles) was used. The distribution of the standard
mixture components between the fractions was determined using a linear
gradient of 50% to 100% MeOH/H20 in 10 min at 2.0 mL/min.
Microbial Mutagenicity Methods. The collected fractions were assayed
for mutagenicity toward Salmonella typhimurium strain TA98 with and without
exogenous metabolic activation (+ or - S9, respectively) using the Ames
plate incorporation assay.4 The fractions were tested with duplicate plates
at four to five dose levels (5 ug - 200 ug), depending on the availability
of sample mass. Mutagenic potency was defined as the slope of the dose-
response data, calculated by simple linear regression on the first four
dose levels for each sample.
Level 2: Cyanopropyl HPLC (CN-HPLC)
CN-HPLC. The HPLC experiments were performed with a Varian Model 5000
instrument in conjunction with a Valco Instruments six-port HPLC valve and
a ISCO Model 1840 UV detector. A Keystone Scientific (State College, PA)
Deltabond cyanopropyl (CN) HPLC column (0.46 x 25 cm; 5 jUm particles) was
used. The CN-HPLC fractionations of the strong acids from the anion-SPE
experiments were performed in three 25 min steps of 100% hexane, 100% DCM
and 100% MeOH, respectively at 2.0 mL/min. Three minute gradients were
used to progress from one step to the next. The collected fractions were
then concentrated to a final volume of 1.0 mL.
Results
Level 1: Nonaqueous Anion Solid Phase Extraction (Anion-SPE)
To characterize the nonaqueous anion-SPE method, five fractionations of
each sample (WSC, MSC, and STD) were performed with corresponding blank
runs. The mutagenicity and mass recoveries were calculated for each run
and averaged for each sample (Table I). The anion-SPE fractions of the
standard mixture were analyzed by reversed phase HPLC and the percent
distribution of the individual standards between fractions determined
(Table II). As shown in Figure 2, both the DCM and TFA fractions
demonstrated the greatest mutagenic potencies. Although both samples would
eventually be used in the second level fractionation, we chose the TFA/MeOH
fractions for the continuation of this study.
Level 2: Normal Phase Cyanopropyl HPLC
Two TFA/MeOH samples were used in the second level normal phase HPLC.
Sample #1 was prepared from the TFA/MeOH fractions of two anion-SPE runs
that were performed on old AG MP-1 which caused slightly anomalous results;
most of the mass in the TFA/MeOH fractions eluted in the 2% TFA/MeOH
fraction rather than an even distribution between the 10% and 2% TFA/MeOH
820
-------
fractions. Although the exact nature of s.ample #1 was somewhat ambiguous,
the sample proved useful in the initial stages of level 2 methods
development. Sample #2 was composed of the residual mass in the TFA/MeOH
fractions of three anion-SPE runs discussed above. To characterize the CN-
HPLC method, three runs were made using sample #1 (ca. 0.70 mg, 1.23 mg,
and 2.23 rag injections) and two runs made using sample #2 (ca. 870 pg
injections). The mass and mutagenicity recoveries are reported in Table
III. A comparison of the percent mass distribution between fractions of
sample #1 and #2 (all runs) are shown in Figure 3 along with the mutagenic
potencies of each fraction.
Conclusion
Although the data presented here is preliminary, the results are very
promising. We have achieved consistently high recoveries of both mass and
mutagenicity. In addition, the level 2 HPLC experiments have yielded a
nonpolar acid fraction which contains ca. 45% of the sample mutagenicity
while containing only ca. 20% of the total sample mass. This novel,
mutagenically potent fraction is in the process of being analyzed using
GC/MS in hopes of identifying some of the components. Because of the
limited mass available for GC/MS analysis, we have only been able to
tentatively identify what appears to be a hydroxylated nitro compound. We
are also in the process of characterizing a new anion-SPE protocol which
differs from the original procedure in three principal ways. We are now
using only 1.5 mg of resin for the same sample mass and collecting the
strong acids in only one 15 mL fraction of 10% TFA. In addition, we are
testing the use of the AG MP-1 resin without the extensive soxhlet
extraction pretreatment.
Acknowle dgement s
We thank Gwendolyn Belk, Virginia Houk and Maria Taylor for their
invaluable assistance in this research. We also thank Roy B. Zweidinger
and Joellen Lewtas of the U.S. EPA and William Ellenson of NSI-ES for their
critical comments and helpful suggestions throughout this project. We also
thank the U.S. EPA Integrated Air Cancer Project for funding this research
through contracts #68-02-4443 to NSI and #68-02-4456 to EHRT.
Although the research described in this article has been funded wholly
or in part by the U.S. EPA, it has not been subjected to Agency review and
therefore does not necessarily reflect the views of the Agency. Mention of
trade names or commercial products does no~ constitute endorsement or
recommendation for use.
References
1. D. Schuetzle, J. Lewtas, "Bioassay-directed chemical analysis in
environmental research" Analytical Chemistry 58: 1060 (1986).
2. D.A. Bell, H. Karam, R.M. Kamens, "A nonaqueous ion exchange separation
technique for use in bioassay-directed ::ractionation of complex
mixtures: application to wood smoke particle extracts." Envir. Sci.
Techno1. in press (1990).
3. R. Zweidinger, Personal communication.
4. D. Maron and B. N. Ames, "Revised methods for the Salmonella
mutagenicity test" Mutat. Res. 113: 173 (1983)
821
-------
Table I. Level 1: anion-SPE
average mass and mutagenicity
recovery.
Table III. Level 2: CN-HPLC
recovery of mass and mutagenicity.
Sanple
MSC
(RSD)
WSC
USD)
STD
(ESD)
Mass
PUCEBT
Recovery
91*
(.05)
83*
(.06)
99*
(.22)
Hutagenici ty
Recovery
(+/• ) S9
951/102%
(.14/.10)
851/88*
(,1/,08)
Reconstituted
(*/• ) S9
151*/131I
(.12/.17)
im/112t
[.IB/. 14)
SAMPLE
RON #
1
2
3
SAMPLE
RUN (t
1
2
i
MASS
Recovery
1 5 4 %
1 1 9 %
9 7 %
2
MASS
Recovery
1 0 5 %
1 0 7 %
Mutagenicity
Recovery
1 1 0 %
1 0 8 %
9 0 %
Mbtagenicity
Recovery
1 3 9 %
1 2 7 %
Table II. Level 1: anion-SPE distribution of Individual
standards.
STANDARD
7 , 8-BENZOQUINOLINE
PYRENE
1 -NITRONAPHTHALEKE
1-HAPHTHOL
4-NITROPHENOL
BENZOIC ACID
2-NITROBENZOIC ACID
DCH
100%
100%
100*
-
-
KeOH
-
-
25%
-
C02/KeOE
-
-
-
65%
26%
-
-
2t
TFA/HeOH
-
•
10%
34%
52%
10%
104
TFA/MeOH
-
-
-
-
40%
48%
90%
BlOlSSir DIRECTED P R A C I I 0 H » T I 0 F
CO 2/HeON TFA/KeOK
Normal - Phase
Cyanopiopyl HPLC
Figure 1. Schematic diagram of
Bioassay Directed Fractlonation.
LEVEL 1: AHION-SPE
AVERAGE HDTAGENIC POTBMCY
HoOK C01 2» TFAlOt TFA
Elution Solvent
MSC (.35) ^»SC <-S5]^gv3C (*Si>
Figure 2. Average mutagenic
potency from 5 anion-SPE runs.
822
-------
n
o
•H
4J
JJ
c
0)
u
H
-------
APPLICABILITY OF GC/MS INSTRUMENTATION FOR THE ANALYSIS
OF UNDRIED AIR TOXIC SAMPLES
L.D. Ogle, R.B. White. D.A. Brymer, and M.C. Shepherd
Radian Corporation
P.O. Box 9948
Austin, Texas 78766
Increased demand for Volatile Organic Compound (VOC) analyses of
ambient air samples by Gas Chromatography/Mass Spectrometry (GC/MS) has led
to the use of a variety of analytical instruments. The desire to analyze
ambient samples containing polar organics such as alcohols, ketones,
aldehydes, nitriles and esters can lead to the injection of a relatively
large amount of water onto the analytical system. Water in the samples can
cause some instruments to shut down due to pressure increases or degrade
the performance of the system such that the necessary sensitivities cannot
be obtained. Pleil, et.al. (1), reported the use of a Megabore column and
a mass selective detector for the analysis of polar compounds, but also
observed a loss of sensitivity due to peak spreading. The poor peak shape
was a result of a slow temperature program to avoid an overpressure fault
in the MS caused by water.
Sample moisture can be removed prior to cryogenic concentration
by the use of a Nafion8 dryer. Removal of moisture using a Nafion membrane
does not affect the recovery of nonpolar compounds, but will restrict the
analytes which can be observed (2,3). EPA Compendium Method TO-14 reports
the use of a Nafion drier with a Hewlett Packard Model HP-5970 mass
selective detector (4). However, this method reports that water and other
light, polar compounds permeate the walls of the dryer into a dry air purge
stream flowing through the annular space between the Nafion and the outer
tubing. Cox and Earp (5) reported low recoveries of ketones and zero
recovery of low molecular weight alcohols through Nafion membranes. They
determined that other drying agents, such as potassium carbonate, also
remove organic compounds.
Therefore, the analysis of polar compounds generally requires an
analytical system which can tolerate a concentrated volume of water. The
volume of water injected onto the system is dependent upon sample humidity
824
-------
and load volume. A typical VOC analysis of a 250 mL volume of undiluted
sample with 70% relative humidity results ;ln 2 to 3 milligrams of water
being introduced into the system. This amount of water is approximately
1000 times the mass of the target analytes in a typical ambient sample.
Use of turbomolecular pumped GC/MS systems with limited pumping
capacity, such as the Hewlett Packard 5970 MSD, for the analysis of polar
compounds in humid samples is not generally considered a viable option
without the use of a jet separator and even then performance may be
questionable. These systems are incapable of handling the pressure
increases caused by the injection of even email amounts of water. Such
systems turn themselves off when such a rise in pressure is detected.
Therefore, only differentially pumped systems were considered for analysis
of polar compounds. This paper discusses the use of two such systems, a
Finnigan Model 4500 and a Hewlett Packard Model 5988. for the analysis of
undried samples. The effect of moisture on detector sensitivity and
reproducibility will be discussed. Other variables, such as the type of
cryogenic focusing hardware, instrument configuration, and desorption time
will also be discussed.
EXPERIMENTAL METHODS
Two approaches can be taken for the analysis of polar VOCs by
GC/MS. A wide (0.32 mmid) bore column can be used with the effluent
directed into the source of the mass spectrometer or a Megabore (0.53 mm or
larger i.d.) column and jet separator can be utilized. The wide bore
column approach can suffer from the column plugging with ice causing
shifting retention times, but provides better resolution. The Megabore
column approach requires the use of a jet separator and has poorer
resolution than the narrow bore, but does not have the ice blockage problem
associated with the wide bore column and provides higher loading
capacities.
Radian uses the Megabore column with a jet separator when
analyzing samples for the Urban Air Toxics Program for the USEPA. A
Finnigan 4500 GC/MS is used to analyze the humid samples for 38 non-polar
compounds. A 0.125 inch o.d. stainless steel trap packed with silanized
glass beads is used to cryotrap organic compounds from the ambient pressure
samples. The sensitivity and resolution wss not adversely affected by the
70% relative humidity in this standard. Reproducible retention times and
correlation coefficients of greater than 0.995 are routinely obtained for
these relatively nonpolar compounds.
Analysis of polar compounds has also been accomplished on this
system without the use of a drier. These compounds include acetone,
several alcohols, diethyl ether, 1,4-dioxane and 2-methyl-l,3-dioxolane.
These compounds exhibit excellent peak shape, sensitivity and
reproducibility. Using SIM techniques, estimated detection limits for
1,4-dioxane and 2-methyl-l,3-dioxolane approaching 0.5 ppbv could be
obtained for both compounds. Estimated detection limits obtained in the
full scan mode were much higher, on the order of 10 to 15 ppbv.
A similar GC/MS system was configured with a wide bore column
(0.32 mm i.d., 1.0 urn DB-1) to provide confirmational analysis for air
samples analyzed on Radian's gas chromatography systems with multidetectors
(GC/MD). The interface used on this system loaded from pressurized
canisters using mass flow controllers and utilized a 0.0625 inch o.d.
nickel cryotrap packed with silanized glass beads. This cryotrap is
similar to the one described in Method TO-14. The column was placed
directly into the source and a jet separator was not used. Excellent
825
-------
reproducibility was obtained from low humidity samples. A relative
standard deviation of less than 5% was obtained for one internal standard,
toluene-d8, over a series of 15 analyses. Peak shape and resolution were
very good for most compounds, with the exception of acetone which tailed on
the DB-1 methyl silicone column (Figure 1).
Subsequent analyses of higher humidity samples on this analytical
system resulted in ice plugs restricting flow through the small diameter
trap. Coinciding with the ice blockage was an increased variability in
internal standard response. Loading a smaller volume of sample solved the
problem and allowed analysis to continue, though sacrificing sensitivity.
High humidity standards formed ice plugs in the cryotrap, but did not
affect peak shape or resolution (Figure 2). However, a drop in the nominal
area counts for toluene-dg from 25000 (30% RH) to 19000 occurred.
Restricted sample flow through the trap corresponded with increased
analytical variablility.
For comparison, the interface with the small cryotrap was moved
to a Hewlett Packard Model 5988 GC/MS system. Analysis of a dry standard
gave excellent peak shapes and reproducibility. However, calibration with
the Urban Air Toxics list of compounds at 30% relative humidity resulted in
very large relative standard deviations in the response for individual
compounds, as shown in Table 1. The reason for the variability in the
analytes and internal standards was not understood as dry standards worked
well.
Water was suspected as the cause of the problem, so mass 18 was
monitored during analysis of humid samples. Figure 3 presents an elution
profile of water from a 75% relative humidity sample. At 30% relative
humidity, the response of toluene-d8 is reduced by 10 to 20 percent. At
75% relative humidity, the internal standard does not respond at all. The
elution time of toluene-d8 coincides with the maximum of the water peak.
To prove water diminishes the detector response and the cryogenic
interface is not involved in this phenomenon, a toluene-d8 standard was
prepared in water and in methylene chloride so direct liquid injections
could be made into the instrument. The relative standard deviation of
d8-toluene in methylene chloride was 6.1% for four replicates and 64% for
three replicates in water. Lowering the instrument source temperature from
250° C to 150° C did not significantly effect the variability caused by
water.
Numerous attempts were made to limit the amount of water going
into the mass spectrometer so the effects of water could be minimized. The
interface desorption time was reduced in an attempt to limit the amount of
water transferred to the GC/MS. This attempt reduced the volume of water
going into the mass spectrometer, but was not adequate to obtain
reproducibility at higher relative humidities. An inlet splitter was
installed to divert a portion of the water before it reached the column.
Limited success was obtained using this technology. A plug in the HP
system used to increase the pressure during CI mode operation was removed
from the source to try opening up the source so more of the water could be
removed. This attempt was not successful. In addition, the filament
emission current was increased from 300 uA to 400 uA in an attempt to
increase ionization efficiency. This attempt also was not successful.
A jet separator was then added to the Hewlett Packard instrument
so a majority of the water could be removed before being transfered into
the mass spectrometer source. Monitoring for water during the analysis of
a wet standard resulted in very little water being observed, as shown in
826
-------
Figure 4. The relative standard deviation of repetitive analyses was 6.1%
for toluene-d8 and 2.6% for bromofluorbenzene for a 65% relative humidity
standard on a VOCOL 0.75 mm i.d. large bore column. However, use of the
jet separator has resulted in a net loss of 20 to 50% in sensitivity
(compound dependent).
The internal diameter of the cryotrap was also increased from
0.254 mm i.d. to 1.35 mm i.d. Increasing the internal diameter of the trap
and leaving the first inch of the trap unpacked increased maximum load
volume for water saturated standards from less than 100 mL to more than
250 mL of undiluted sample before ice in the trap started to restrict
sample flow. Larger sample volumes can be concentrated using an interface
similar to the one utilized on the UATMP project which loads using a vacuum
source to pull the sample through the cryotrap. Load rate drops as the
sample volume is increased, but does not effect the measurement of sample
volume as it is determined by differences in pressure and not by a constant
flow rate.
CONCLUSIONS
Analysis of undried ambient samples is best performed on GC/MS
instrumentation utilizing a jet separator for the removal of water to
maximize analytical reproducibility. Undried samples can be analyzed using
a column placed directly into the source of the Finnigan 4500, but sample
volumes must be limited if reproducibility is to be achieved. Lower sample
volumes result in higher detection limits and somewhat more variable
results due to fact that response may be closer to the noise level of the
instrument. Large amounts of water cannot be tolerated by either
instrument and a separator must be used to obtain reproducible results.
Trap design also plays an important role in the analysis of
undried samples. Cryogenic concentration of samples containing 70% or more
relative humidity requires a trap with an internal diameter of 1.35 mm or
greater to concentrate 250 mL of sample when using an interface design
based upon a consistent load rate. After this point, restricted flow
results in lower load rates and irreproducible sample volumes. Use of an
interface designed to load on pressure differentials allows even larger
sample volumes to be concentrated.
ACKNOWLEDGEMENTS
Radian would like to thank the U.S. Environmental Protection
Agency for their support of the Urban Air Toxics Monitoring Program. We
would also like to acknowledge Finnigan MAT and Hewlett Packard Instruments
for their support and advice in the development of these instrumental
procedures.
References:
1. Pleil, J.D., W.A. McClenny, and K.D. Oliver, "Dealing with water
in GC/MS analyses of whole air samples", presented at the 1989
Pittsburgh Conference and Exposition on Analytical Chemistry and
Applied Spectroscopy, Atlanta, GA, March, 1989.
2. McClenny, W.A., J.D. Pleil, M.W. Holdren, and R.N. Smith,
"Automated cryogenic preconcentration and gas chromatographic
determination of volatile organic compounds in air". Anal.
Chem.. 56: 2947-2951. 1984.
827
-------
3. Pleil, J.D., K.D. Oliver, and W.A. McClenny, "Enhanced
performance of nafion dryers in removing water from air samples
prior to gas chromatographic analysis", JAPCA. 37:244-248, 1987.
4. EPA Compendium of Methods for the Determination of Toxic Organic
Compounds in Ambient Air. Method TO-14; EPA-600/4-84-041, U.S.
Environmental Protection Agency, AREAL, Research Triangle Park,
NC, 1984.
5. Cox, R.D., R.F. Earp, and K.W. Lee, "Removal of Water Vapor from
Gases Prior t^ Analysis for Low Level Organics", presented at
1982 Pittsburgh Conference on Analytical Chemistry and Applied
Spectroscopy, Atlantic City, NJ, 1982.
Table I. Relative Standard Deviations of Response Factors for Selected
Compounds from Wet Standards on the HP 5988
Compound RSD of RF (Percent)
Vinyl Chloride 19
1,3-Butadiene 5.1
Methylene Chloride 64
Bromochloromethane 72
Benzene 77
Cyclohexane 48
Cis-l,3-Dichloropropene 14
Chlorobenzene 102
o-Xylene 2.6
1,2-Dichlorobenzene 30
828
-------
Low Humidity Ambient Air Sample On A 60m x 0.32 mm W
DB-1 Column - Finnlgan 4500 GC/MS
Figure 1
T rtcW oroethyton*
TohwfW -<«
Acetone
70% Relative Humidity Standard On A 60m x 0.32mm id
DB-1 Column - Finnigan 4500 GC/MS
Figure 2
M^—^
Elutlon Profile Of Water From A 75H Relative Humidity
Standard On A 60m x 0.32mm id Column-HP 5988 GC/MS
Figure 3
829
-------
LABORATORY EVALUATION OF MICROSENSOR
TECHNOLOGY GAS ANALYZER
Joseph S. C. Chang, Sydney M. Gordon
IIT Research Institute
Chicago, IL 60616-3799
Richard E. Berkley
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
A Mlcrosensor Technology M200 dual-column gas analyzer was evaluated
for specific identification and quantitative analysis of simple and complex
mixtures of organic vapors at low concentrations to assess its suitability
for analysis of trace pollutants in air. Independent and simultaneous
analyses were performed on nonpolar and moderately polar narrow-bore
columns. Detection limits below 1 ppmv for trichloromethane, tetrachloro-
methane, tetrachloroethylene, benzene, and toluene were found. Relative
standard deviations of retention times were less than 0.1%, but attempts to
correlate data from the two columns were hampered by column bleed from one
of them and by software constraints on run time.
Introduction
The measurement of volatile organic compounds (VOCs) in air is
usually carried out by collecting samples at a field site and returning
them to a laboratory for analysis. These samples normally have to be
preconcentrated before analysis, because the typical levels of VOCs in air
are too low for direct detection. This approach often results in a loss of
components or inadvertent contamination of the sample.
Portable gas chromatographs (GCs) are capable of producing immediate
results, with or without preconcentration, and can therefore avoid many of
the difficulties associated with batch analytical approaches in a central
laboratory.1-2 Two commercial!y-avaiTable GCs in this category are the
Photovac 108-series and the Microsensor Technology M200. The Photovacs use
wide-bore wall-coated open tubular columns and extremely sensitive
photoionization detectors that make sample preconcentration unnecessary.
830
-------
These instruments have been evaluated by Berkley,1-2 who showed that the
manufacturer's claim of a detection limit below about 0.1 part per billion
(ppbv) for benzene and similar compounds is; reasonable.
The Microsensor Technology M200 microchip GC contains two narrow-bore
(4 m x 0.1 mm ID) fused silica columns, each of which is coated with a
different stationary phase for improved selectivity.3'4 The instrument also
includes a miniature sampling pump and injector, as well as two independent
miniature high-performance thermal conductivity detectors. The two high
resolution capillary columns can simultaneously analyze air samples with
total run times of about 1 minute.4 Because the M200 gas analyzer was
originally Intended for industrial hygiene work, we undertook a study to
evaluate its potential usefulness for ambient air monitoring, including Its
ability to identify and quantify toxic organic vapors in complex mixtures.
This paper summarizes the results of this •nvestigation.
Experimental Methods
The portable M200 gas analyzer contains two completely independent GC
modules, each with its own miniature sample injector (injection volume 25-
250 nL), column, column heater, and therma"! conductivity detector (dead
volume 2 nL).3 The modules share the M200's electronics and front panel
controls. Method parameters, such as column temperature, injection time,
and detector sensitivity can be controlled quite simply from the front
panel of the instrument. The system has an automatic sampling mode which
can also be activated from the front panel. For our evaluation, the M200
was configured with two 4 m x 0.1 mm ID fused silica columns: nonpolar
OV-73 and moderately polar OV-1701 stationary phase coating. The different
phases, and the fact that samples can be analyzed either simultaneously or
independently of each other, provide a system with considerable analytical
flexibility.
Besides offering control from the front panel with detector output to
an integrating strip chart recorder, the instrument also offers operation
via a personal computer using the EZChrom 2:00 software package supplied by
the manufacturer. The program, which was developed to control, acquire,
and process data from the M200, runs in the Microsoft Windows™ environment.
To evaluate the M200 GC, standard gas mixtures containing benzene,
toluene, trichloromethane, tetrachloromethane, and tetrachloroethylene were
prepared in static glass dilution flasks at five nominal concentrations
over a range from 1 to 50 parts per million (ppmv) for each compound.
Besides these standards, stainless steel canisters containing 41 different
VOCs at known concentrations were obtained from the Atmospheric Research &
Exposure Assessment Laboratory, EPA (Research Triangle Park, NC). Three of
the canisters contained all of the compounds present in the fourth canister
at concentrations of about 10 ppmv. The levels of the 41 compounds in the
fourth canister were between 2.3 and 4.1 ppmv.
Samples were injected into the M200 unit using either the manual or
autosampler injection modes. For autosampler injection, fixed volumes of
the standards were injected repetitively by connecting one end of a Teflon
tube to the sample injector port and the other end to a Swagelok connector
that screwed into the neck of the static dilution flask or canister.
831
-------
Re s u11s And JP1 sc u. ssj op
The effect of column temperature on retention time and resolution was
studied by Injecting fixed amounts of the five-component standard onto the
columns and varying the temperature from 30°C to 120°C. The plots in
Figure 1 for the nonpolar column show a marked decrease in the retention
times of the components as the column temperature is increased. The plots
also illustrate the high speed of analysis achievable with the M200,
especially at the higher temperatures used. Benzene and tetrachloromethane
in the standard co-elute at all temperatures between 30°C and 120°C on this
column but are fully resolved from one another on the moderately polar
column at temperatures below 50*C. These results clearly demonstrate the
usefulness of having two columns of differing polarity available 1n the
M200 for making positive identifications of unknown VOCs.
The retention time and peak area stability of the instrument were
assessed by performing replicate Injections of the five-component standard
mixture on both columns in the manual injection mode and the autosampling
mode under computer control. As an example, Table 1 lists the retention
times and peak areas obtained for the nonpolar column at 70°C. The
relative standard deviations for the retention times are less than 0.1% at
all analyte concentrations. The peak area precision is not as good but is
generally satisfactory at higher concentrations, especially for the more
volatile compounds such as trichloromethane, tetrachloromethane, and
benzene. Despite higher background levels, the moderately polar column
gives comparable results for retention time and peak area precision.
Detection limits were estimated by measuring peak heights of the
standards as a function of concentration at different temperatures, and
performing linear regression analysis. The detection limits, defined as
the signal corresponding to three times the standard deviation of the
baseline noise level divided by the slope of the best-fit curve, were
determined for both columns and are summarized in Table 2. The results
indicate that the detection limits decrease as the column temperature
increases. In addition, they are about an order of magnitude lower on the
nonpolar column than on the moderately polar column, due to the higher
column bleed from the latter.
To assess the suitability of the instrument for the analysis of
organic pollutants in air, we prepared static dilution flasks of three
calibration standards of the five-component mixture (at 1, 6, and 10 ppmv)
as well as two mixtures (at 4 and 8 ppmv) to serve as "unknowns".
Calibration curves were prepared and used to generate peak tables,
calibration tables, and timed events for the two columns at 70°C, using the
EZChrom 200 data system. Then, the two "unknown" mixtures were analyzed
separately on each column under the same operating conditions.
The results obtained are summarized in Tables 3 and 4. It is clear
that, on the basis of retention times, positive identifications of the five
"unknown" compounds can be performed with a high degree of reliability,
using the calibration libraries defined by the EZChrom 200 software. There
is also close agreement between measured and actual concentrations,
especially at 8 ppmv. At the 4 ppmv level, the average measured values for
the nonpolar column lie between 3.3 and 3.7 ppmv with relative standard
deviations (RSDs) of 8-14*. For the moderately polar column, the average
measured concentrations are between 3.7 and 4.3 ppmv with RSDs of 11-44%.
The relatively large scatter noted here is due to the much higher
background noise levels and baseline drift observed with this column than
with the nonpolar column.
832
-------
Attempts to extend this approach to the analysis of a complex
multicomponent mixture were less successful. Three canisters containing 41
different VOCs at concentrations of about 10 ppmv each were used to
generate calibration libraries for the two columns at 30°, 70°, 90', and
110°C. This was followed by the analysis of the contents of the fourth
canister, which contained all of the compounds present in the other
canisters. In running these samples, we were limited by constraints on the
run time imposed by the EZChrom 200 software. The maximum run time is
limited by the software to 100 seconds, whereas it can be extended to 255
seconds in the manual mode. The short run time prevents the detection of
later eluting peaks, especially at the lower column temperatures used.
Raising the temperature to 90°C resulted in the elution of all components
within the 100-second time limit, but caused a significant loss of
resolution. The sample was re-run in the manual mode at 70°C and with a
maximum run time of 255 seconds. This permitted the identification of
several more peaks but, even under these conditions, there was considerable
peak overlap, making identification and quantification of most constituents
of the mixture practically impossible.
Conclusions
The Mlcrosensor Technology M200 dual-column gas analyzer is a truly
portable and useful rapid screening device that can produce valuable
information about organic vapors in air. The Instrument can provide
independent or simultaneous analyses using narrow bore columns of different
polarity and yields detection limits of less than 1 ppmv. It can be
operated manually or under the control of a sophisticated, user-friendly
software program. The latter, however, limits the maximum analytical run
time to 100 seconds, which hampers efforts to analyze complex air mixtures
of environmental interest. Work currently 1n progress in our laboratory is
directed toward evaluating an alternative approach to improving overall
sensitivity and compound specificity, based on coupling the M200 GC to an
ion trap mass spectrometer.
Acknowledgement
Although the research described was funded by the U.S. Environmental
Protection Agency (contract number 68-D8-0002), it has not been subjected
to the required peer and administrative review and does not necessarily
reflect the views of this agency and no official endorsement should be
inferred.
References
1. R. E. Berkley, "Analysis of Toxic Organic Vapors in Air Using a
Portable Photoionization Gas Chromatograph," Proceedings of the 1988
EPA/APCA International Symposium on Measurement of Toxic and Related
Air Pollutants, Research Triangle Park, North Carolina, May 1988, pp.
352-357.
2. R. E. Berkley, "Field Evaluation of Photovac 10S70 Portable Gas
Chromatograph," Proceedings of the 1989 EPA/APCA International
Symposium on Measurement of Toxic and Related Air Pollutants, Raleigh,
North Carolina, May 1989, pp. 19-26.
3. G. Lee, C. Ray, R. Siemers and R. Moore, "Recent Developments in High
Speed Chromatography," Amer. Lab. 111 (Feb. 1989).
833
-------
E. B. Overton, L. H. Grande, R. W. Sherman, E. S. Col lard, and C. F.
Steele, "Rapid and Reliable Analysis of Volatile Organic Compounds
with a Field Deployable Gas Chromatograph," Proceedings of the 1989
EPA/APCA International Symposium on Measurement of Toxic and Related
Air Pollutants, Raleigh, North Carolina, May 1989, pp. 13-18.
Table 1. Reproducibility of retention times and peak areas for
nonpolar column1 using EZChrom 200 data system
Compound
T r 1 ch 1 oromethane
Concn.
ppmv
1
6
10
Ave. Ret. Time (s)
20.43 ± 0.01 (0.04)2
20.46 ± 0.01 (0.05)
20.46 ± 0.01 (0.05)
Ave. Peak Area
1041 ± 24 (2.3)
2183 ± 77 (3.5)
3071 ± 72 (2.3)
Tetrachloromethane + 1
Benzene 6
10
Toluene
Tetrachloroethylene
1
6
10
1
6
10
24.99 ± 0.01 (0.04)
25.04 ± 0.01 (0.04)
25.03 ± 0.01 (0.05)
43.32 ± 0.02 (0.04)
43.40 ± 0.03 (0.08)
43.37 ± 0.02 (0.05)
56.03 ± 0.03 (0.05)
56.14 ± 0.04 (0.08)
56.12 ± 0.02 (0.03)
1652 ± 106 (6.4)
3430 ± 36 (1.0)
4844 ± 103 (2.1)
478 ± 92 (19)
1231 + 49 (4.0)
1705 ± 23 (1.3)
655 ± 65 (9.9)
1506 ± 116 (7.7)
2134 ± 107 (5.0)
1 Column temperature 70°C; 2 Average of 5 measurements ± SO (XRSD)
Table 2. Limits of detection (ppmv) for five-component standard mixture
Compound
Trichloromethane
NONPOLAR COLUMN
30°C 50°C
0.2 0.2
MODERATELY POLAR COLUMN
30'C 40°C 50°C
2.5 2.2
Trichloromethane +
Tetrachloromethane
Tetrachloromethane
Tetrachloromethane +
Benzene
1.1
2.4
1.9
0.1
0.1
Benzene
Toluene
Tetrachloroethylene
0.8
0.8
0.5
0.4
3.3
9.4
nd1
2.6
6.8
5.4
2.2
4.8
4.0
nd=not detected; retention time exceeded maximum run time
834
-------
Table 3. Measurement of target compounds in "unknown" gas mixtures for
nonpolar column1 using EZChrom 200 data system.
Compound
Trichloromethane
Tetrachloromethane +
Benzene
Toluene
Tetrachloroethylene
Ret. Time (s)
Callb. Measured2
20.45 20.43 ± 0.01
20.46 ± 0.01
25.02 25.00 ± 0.01
25.04 ± 0.01
43,36 43.33 ± 0.02
43.38 ± 0.02
56.10 56.05 ± 0.02
56.14 ± 0.03
Concentration (ppmv)
Actual
4
8
4
8
4
8
4
8
Measured2
3.4 ± 0.3
8.2 ± 0.4
3.3 ± 0.4
8.4 ± 0.6
3.4 ± 0.3
7.9 ± 0.8
3.7 ± 0.5
8.4 ± 0.2
1 Column temperature 70°C; 2 Average of 5 measurements ± SO
Table 4. Measurement of target compounds 1n "unknown" gas mixtures for
moderately polar column1 using EZChrom 200 data system.
Compound
Trichloromethane +
Tetrachloromethane
Benzene
Toluene
Tetrachloroethylene
Ret. Time (s.)
Calib. Measured2
24.26 24.24 ± 0.01
24.27 ± 0.00
27.15 27.14 ± 0.02
27.18 ± 0.01
49.13 49.10 ± 0.02
49.17 + 0.02
56.24 56.21 ± 0.02
56.29 ± 0.04
Concentration (ppmv)
Actual
4
8
4
8
4
8
4
8
Measured2
3.9 ± 0.5
8.3 ± 0.3
4.3 ± 1.9
8.1 ± 0.7
3.7 ± 0.9
7.9 i 0.5
3.8 ± 0.4
8.2 ± 0.5
1 Column temperature 70°C; 2 Average of 5 measurements ± SD
E
CD
CC
50 70 90 110
Temperature ("C)
Figure 1. Effect of temperature on retention time for nonpolar column.
835
-------
A REAL-TIME MONITOR FOR CHLORINATED ORGANICS IN WATER
Joseph R. Stetter
Transducer Research, Inc.
1228 Olympus Drive
Naperville, IL 60540
Zhuang Cao
Illinois Institute of Technology
Department of Chemistry
Chicago, IL 60616
A low-power and portable solid-state gas sensor with a
selective response to chlorinated hydrocarbons has been
combined with a simple tubular silicone rubber permeator. The
permeator and sensor combination have been tested in a way that
simulates the on-line analysis of chlorinated hydrocarbons in
water. The detection system can provide information on whether
or not the sampled stream contains chlorinated hydrocarbons as
well as the quantitation of the chlorinated hydrocarbon in the
sample. The selective gas sensor and permeation apparatus
offers a new and convenient method to analyze the contents of
an aqueous sample. It is important to note that the approach
allows the use of a gas sensor to analyze a liquid stream in
near real-time. Since many more types of gas sensors are
available than liquid sensors, this approach may be more
generally useful if other gas sensors are interfaced to the
liquid sampling system by means of semi-permeable membrane
technology.
This analytical method combines a selective permeator and
a portable gas sensor. The new approach may find use in field
screening of mud, water, or other complex sample matrices for
volatile halogenated organic compounds.
836
-------
INTRODUCTION
Chlorinated hydrocarbons are potentially toxic vapors
that commonly occur in the workplace*, environment, and process
stream. Recently, the Environmental Protection Agency has added
25 organic chemicals to the list of compounds regulated as
toxic in wastes under the Resource Conservation & Recovery
Act1. Sixteen of these organic chemicals are chlorinated
hydrocarbons and are illustrated in Table I. It is important
and useful to develop a real-time, inexpensive analytical
method for the determination of chlorinated hydrocarbons in
waste water.
Chemical sensor technology is playing an increasing role
in analytical science. It not only provides real-time, on-site
information about the presence and concentration of chemicals
in a given environment, but also is a less expensive alterna-
tive to large analytical instruments. Unwin and Walsh2 have
described a sensitive chlorinated hydrocarbon sensor which is
based on the electrical conductivity change effected by the
adsorption of chlorine, produced by decomposition of the
chlorinated hydrocarbon over a heated platinum coil, on films
of lead phthalocyanine. This paper describes a rare-earth
sensor that is selective for chlorinated hydrocarbons.
Membrane technology is developing significant uses in the
analytical sampling systems. It not only can be used to extract
or separate organic vapors from an aqueous sample matrix, but
also offers a technique for continuous sampling. Hellgeth and
Taylor3 have described a membrane based method for on-line
high-performance liquid chromatography/Fourier transform
infrared spectrometry. This method involved the detection of
the organics in an aqueous/organic segmented stream through a
flow cell which is constructed by using multiple layers of
Teflon membrane. Blanchard and Hardy4 introduced a separation
method based on the permeation of volatile organic compounds
through a silicone polycarbonate membrane from an aqueous
sample matrix into an inert gas stream. A portion of this
stream was then injected into a capillary gas chromatograph.
Melcher5 developed a silicone membrane flow injection system
for the determination of trace organic compounds in aqueous
samples. Recently, Lauritsen6 reported using a silicone
membrane for on-line monitoring of dissolved, volatile organic
compounds with mass spectrometry.
Combining membranes and,chemical sensors will allow the
development of new in situ detection and quantification methods
for pollutants and trace chemicals in waste water. Our previous
paper7 presented the possibility of the use of a rugged
permeation membrane in combination with a low cost, small,
selective gas sensor for on-line and near real-time analysis
of chlorinated hydrocarbons in an aqueous phase. This paper
837
-------
will focus on the feasibility of this real-time monitor for
the determination of chlorinated hydrocarbons such as chlorob-
enzene, 1,1,1-trichloroethane, chloroform in water.
EXPERIMENTAL METHODS
Sensor
A schematic diagram of the sensor is shown in Fig. 1. A
dc potential of about 4 V across the electrodes is maintained
and the conductance is measured using Ohm's law and .the voltage
drop across a known resistor. In absence of a chlorinated
hydrocarbon vapor, a high resistance is observed and the
conduction between terminals 1 and 6 is very small. But in the
presence of a chlorinated hydrocarbon vapor, the resistance
decreases significantly and this increases the current flow
between the terminals. The magnitude of the sensor background
current at constant dc bias voltage is very sensitive to the
presence of chlorinated hydrocarbon vapors in the atmosphere
surrounding the sensor since these gases alter the conductance
of the semiconductor surface. This conductance change in the
presence of chlorinated hydrocarbon vapors is the analytical
"signal" from this sensor, i.e., the conductance is a function
of the concentration of the chlorinated vapor present. The
sensor is insensitive to many hydrocarbon contaminants such as
benzene, hexane "and can be operated in air.
Permeator
Since the chlorinated hydrocarbon sensor is a gas sensor,
it is a necessary to convert the aqueous sample containing the
analyte into a vapor sample suitable for analyses by the gas
sensor. The permeation cell has been described7 and consists of
thin wall permeable tubing (Silastic Medical-Grade Tubing,
0.012 in. i.d. x 0.025 in. o.d. manufacture by Dow Corning)
that is submerged in the aqueous sample. Silicone materials
preferentially allow organic compounds to permeate while
rejecting water and other highly polar molecules.
Instrumentation
The experimental system schematic is given in Fig. 2. The
microprocessor was used to record the experimental data and to
control the sensor and the sample exposure time. The collected
data are analyzed by the PC computer to provide a graphic
display of the results. In order to measure the net response
for organic compounds we used two permeation cells. One is sub-
merged in clear water for the sensor baseline. The other is in
water containing the organic compound at a low concentration.
The two permeation cells were purged with the same flow of air
carrier gas and connected to a three-way solenoid valve that
838
-------
was placed in front of the gas sensor inlet. The solenoid valve
is controlled by a microprocessor. By switching the valve it
was possible to make a very fast change of air flow over the
sensor from pure water to sample. The flow rate was maintained
constant at 170 cc/min during the experiments.
RESULTS
At constant temperature the conductivity of sensor bead
changes with the partial pressure of the analyte7. This
pressure dependence of conductivity of takes the form
a/a0 = k P" (1)
for chemisoption on a transition-metal-oxide system, where a
is the conductivity, P is the partial pressure (or gas con-
centration) of the reacting gas, k and m are constants. Since
the ratio of the conductivities, a/a,, is equal to the ratio of
the conductances, C/C0, equation (1) can be written as
In C/C0 = In k + m In P (2)
In this paper, we use the ratio of the conductance of air
containing sample to that of clean air, C/C0, as the sensor
signal. The sensitivity of the chlorinated hydrocarbon sensor
is highly dependent upon the sensor temperature and gas flow
rate7.
A series of different low concentrations (0.5-10 ppm) of
chloroform in water were prepared, and the signals are given
in Fig. 3. Three different chlorinated hydrocarbons, chlorofo-
rm, chlorobenzene, and 1,1,1-trichloroethane, in water were
(individually) analyzed by using this monitor (see Fig. 4) . The
responses at different low aqueous concentrations (0.5-10 ppm)
were recorded by using an amplifier with a gain of 100 to boost
the signal. There is a good linear relationship between the
response for all three chlorinated compounds and the concentra-
tion and the results of the regressive analysis is given in
Table II. This linear response was obtained using the simple
silicone tube permeator and the entire benchtop instrument
system.
Fig. 5 presents the selectivity of the chlorinated
hydrocarbon sensor. Three chemicals, chlorobenzene, benzene,
and hexane with the same concentration (100 ppm in a gas phase
air mixture) were detected by the same sensor. There is a great
difference in sensor response between chlorinated and un-
chlorinated hydrocarbons. We estimate that the response to
chlorinated hydrocarbons is at least 10J and perhaps 105 times
greater than for hydrocarbons. The origin of this unusual
selectivity is not completely understood at this time.
839
-------
The lifetime of this sensor is suitable for practical
application (see Fig. 6). The life of this small inexpensive
sensor is over 20 days with continuous exposure to high
concentrations of chlorobenzene vapor. This is more than 2,000
sample analyses at high concentration. The stability of sensor
response can be found in Fig. 7 and sensor is quite stable over
a day with variations of about 10% and stable over its entire
lifetime with variation in sensitivity of less than 50%. The
lifetime appears to be a function of the concentration exposure
history of the sensor and, perhaps, would be greatly improved
if used at lower levels of sample.
CONCLUSIONS
The chlorinated hydrocarbon sensor and permeator system
has high sensitivity and selectivity and can used for the
determination of chlorinated hydrocarbons in aqueous samples.
We have examined concentrations between 500 ppb and 10 ppm.
Lower levels may be possible with a higher gain amplifier.
Using a semipermeable membrane allows the application of
relatively inexpensive gas sensors to the analysis of liquid
samples. The relatively low cost, small, and rugged equipment
needed for this analysis has the potential for continuous on-
line analysis of aqueous streams as well as the field screening
of water samples. The sensor is quite stable over its entire
lifetime. The simple silicone permeation tube is rugged,
resistant to many chemicals, and easily inserted into a
concentrated (acid, base, etc) water stream and can provide the
analyte to the gas sensor in a form suitable for quantitative
analysis.
Further work will be focused upon understanding the
sensor mechanism. In addition, optimum geometry for the
permeation system will be investigated to improve the analyti-
cal performance of the sampling system. The characterization
of the analytical method will be extended by using field
samples as well as laboratory standards.
ACKNOWLEDGEMENTS
We wish to acknowledge the funding for Mr. Cao from Dow
Chemical Company and the inciteful comments of Dr. Robert
Bredeweg on this work, especially for suggesting the silicone
membrane technology for the permeator.
840
-------
REFERENCES
1. David Hanson, C & EN. 4. (March 12, 1990).
2. J. Unwin and P. T. Walsh, Sensors and Actuators. 18: 45.
(1989) .
3. J. W. Hellegeth and L. T. Taylor, Anal. Chem. 59: 295.
(1987) .
4. R. D. Blanchard and J. K. Hardy, Anal. Chem. 58; 1529.
(1986).
5. R. G. Melcher, Anal. Chim. Acta 214; 299. (1988).
6. F. R. Lauritsen, Int. J. Mass Spectro. and Ion Proc. 95:
259. (1990).
7. J. R. Stetter and Z. Cao, Anal. Chem. 62: 182. (1990).
Table I. These organics in wastes will be regulated as
hazardous by EPA.
Samples
Regulatory level
(mg per L)
Benzene
Carbon tetrachloride
Chlordane
Chlorobenzene
Chloroform
o-Cresol
m-Cresol
p-Cresol
1 , 4-Dichlorobenzene
1 , 2-Dichloroethane
1, 1-Dichloroethylene
2 , 4-Dinitrotoluene
Heptachlor
Hexachlorobenzene
Hexachloro-1 , 3-butadiene
Hexachloroethane
Methyl ethyl ketone
Nitrobenzene
Pentachlorophenol
Pyridine
Tetrachloroethylene
Trichloroethylene
2,4, 5-Trichlorophenol
2,4, 6-Trichlorophenol
Vinyl chloride
0.5
0.5
0.03
100.0
6.0
200.0
200.0
200.0
7.5
0.5
0.7
0.13
0.008
0.13
0.5
3.0
200.0
2.0
100.0
5.0
0.7
0.5
400.0
2.0
0.2
841
-------
Table II. Least-Square Regression Results from Fig. 4.
samples slope y int R2
chloroform
chlorobenzene
1,1, 1-trichloroethane
0.5379
0.5814
0.7226
0.5043
0.6280
0.6890
0.997
0.988
0.993
Rare earth
semiconductor
Gas in
LNXX \XV\I
Gas outJ
I K ohm
Recorder
(Bias voltage)
Figure l. Schematic diagram of sensor and conductance measure-
ment circuit.
842
-------
Signal generation
-sensor(s)-
Sampling subsystei
-permeator-
Sample
handling
Electronics/power
&
Data acquisition
Microprocessor control
&
display
PC control
&
data analysis
Figure 2. Schematic functional diagram of the experimental
apparatus.
-------
400
00
0
20
40
60 80
Time (min)
100
140
Figure 3. Examples of 3-minute exposures of the sensor to the
permeator containing chloroform analyte at different
concentrations separated by 17-minute to the
permeator containing clean water.
-------
CO
4k
tn
o
O
2.5
2-
1.5
1-
0.5-
• 1,1,I-TCE
t3 C6H5C1
A CHC1,
0
-0.5
0
0,5 1
In(ppm)
1.5
2.5
Figure 4. Example of the relative (linear) sensitivity of the
sensor to aqueous concentrations of three analytes.
-------
o
O
O
0
C6H5CI
20
C6H6
40
60
TIME (MIN)
C6H14
80
100
120
Figure 5. Example of the selectivity of the sensor for
chlorinated organics verses aliphatic or aromatic
hydrocarbons.
-------
oo
O
O
O
100
200 300 400 500 600 700 800
X1000 PPM-HOUR/GRAM
Figure 6. Sensor lifetime as a function of sensor exposure to
analyte and sensor size.
-------
00
*»
00
80
70
60
50
_j
CD 40-
30-
20-
10
245
250
255 260 265
TIME (HOUR)
270
275
Figure 7. Short term (l-day) stability of the sensor response.
Repeated injections (3-minute exposures) of 100 ppm
chlorobenzene in air.
-------
PERFORMANCE OPTMIZAT10N OF PHQTOVAC
10S70 PORTABLE GAS CHROMATOGRAPH
R. E. Berkley, Environmental Protection Agency, Atmospheric Research and
Exposure Assessment Laboratory, Research Triangle Park, NC 27711.
Keith Kronmiller and Karen Oliver, MSI Technology Services Corporation,
2 Triangle Drive, Research Triangle Park, NC 27709
Photovac portable gas chromatographs (PGC) are designed to automatic-
ally sample ambient air for industrial hygiene analyses in which typical
analyte concentrations are a few parts-per-mi11ion levels, so many
sampling errors which might affect preconcentration-based methods can be
avoided. Data are obtained in near real-time, which facilitates rapid
screening of emission sources. However, these PGC's were originally
intended for industrial hygiene analyses in which analyte concentrations
typically exceeded one part pei million. Using them at one part per
billion or less requires modified operating procedures which are discussed
herein.
849
-------
Experimental
Two Photovac Model 10S70 PGC's equipped with constant-temperature
column accessories were used. A 0.53 millimeter inside diameter X 10 meter
long methy1si1icone-coated open-tubular column was used to analyze com-
pounds of intermediate-to-1ow volatility. A KC!/Alumina porous-layer open-
tubular column of the same size was used to analyze more volatile com-
pounds. Ultrazero air (less than 0.1 part per million carbon) was used as
carrier gas. and high-purity helium was used to clean columns. A sample
inlet probe consisting of 2 millimeter inside diameter X 9 meter long
stainless steel tubing was extended from the "SAMPLE IN" port of the PGC
through a vehicle window to a point at least 2 meters above ground. Calib-
ration standards prepared by flow-dilution of commercial 10 part per mil-
lion mixtures of organic compounds in nitrogen (Scott Specialty Gases) were
pumped into 6-liter passivated canisters to a positive pressure of about
100 kiloPaschals (25 pounds per square inch gauge) using a rnetaI-be 11ows
pump. Data were stored on disk by iBM-compatib1e laptop computers using
Photovac DANDI (Data Acquisition aN_d Disk interface) software.
Results and Discussion
Power Requirements
Photovac PGC's were originally powered by an internal 12 volt battery
which recharged automatically when the unit was connected to 110 volt
mains. An auxiliary 12 volt battery could be connected to extend operation
time. Twelve ampere-hour gel-cell battery packs (Globe Battery Division,
Johnson Controls) lasted about 6 hours, provided they were in good con-
dition and fully charged at start-up. The constant-temperature column
accessory increased power consumption significantly, so that 12 ampere-hour
gel-cells discharged much sooner. Although gel-cells may be recharged many
times, a single deep discharge can ruin them, and such damage is not immed-
iately apparent. A panel volt-meter attached to the power lead is recom-
mended for monitoring battery performance, since external battery voltage
is not otherwise displayed, and much time could be lost by trying to oper-
ate with a dead external battery. Larger gel-cells (80 ampere-hours) last
more than 24 hours and consequently are more reliable. Marine batteries
(100 ampere-hours), which are tolerant of deep discharge, may also be used.
Gel-cells can be shipped by air, but marine batteries cannot.
Column and Valve Contamination
Operating at high gain (typically 200 or higher) to detect ambient
background levels of pollutants results in a split-level baseline because
trace contaminants bleed from solenoid valves when they are energized.
When the valves de-energize to resume backflush the baseline falls to its
original level. This disruption lasts about a rainute, during which eluting
peaks cannot be quantitated, so the time when backflush resumes must not be
too near retention times of target compounds. Also, if backflush resumes
too early, late-eluting compounds may be partially lost and not detected.
Because of this, the backflush portion of the sampling cycle must generally
be too short to clear less-volatile compounds from the precolumn or to pre-
vent their breaking through into the analytical column. As a result, col-
umn bleed gradually increases until visible baseline drift appears. Even-
tually the baseline buckles as backflush resumes, and finally it drops
below zero at that point and disappears from the strip chart. Usually it
will have returned to a positive value by the end of the chromatogram, so
it has a large positive slope between resumption of backflush and the end
of the chromatogram. Incorrect placement of integration baselines by the
microprocessor is more likely to occur with a sloping baseline, especially
if the slope is positive. The only remedy for such "bad integrations" is
850
-------
to retrieve chroraatograras stored on disk and correct them manually, using
the "BASELINES" feature of DANDI, a tedious procedure which requires saving
the corrected chromatogram to a separate file. Large volumes of data can
be damaged in this way, and for a long time the cause of the problem was
believed to be failure of solenoid valves. We have found it effective to
bake columns in a drying oven at 90°c for 4 hours while backflushing with
helium, but drying ovens are not field-portable. One of us (Kronmi11er),
after conferring with Photovac, designed and built a portable column heater
which uses a strip of heating tape around the outside edge of the column to
heat it. Four temperature sensors, built into the column assembly, are
used to control thermal output. The unit is powered by 110 volt mains,
and it is about half the size and weight of a briefcase.
Detection Limits
An analytical detection limit is usually taken to be the quantity of a
compound which produces a signal just largi? enough to be distinguished from
baseline noise. Criteria for this distinction are not widely agreed upon,
but the microprocessors in Photovac PGC's don't process peaks smaller than
5 millivolt-seconds. This clearly defines a theoretical detection limit
for each compound at each gain level. It Ls the quantity sufficient to
produce a peak with 5 millivolt-second are.a. However, the real detection
limit could be higher because the microprocessor does not reliably notice
small peaks when the baseline is not level.
Da ta Hand 1 ing
Mobile monitoring in a vehicle is an important use for PGC's, but they
generally produce more data of higher quality by stationary autosampling.
Sampling at 15-minute intervals round the clock produces large volumes of
data which must be validated and tabulated. Sample files can be copied
from floppy disk to a virtual disk in computer memory to facilitate rapid
screening for improper integrations and other problems using the "LIST AUTO
VIEW" feature of DANDI. Quantitation Iist:5 can be printed for manual data
tabulation, or data can be copied directly from the computer screen or
plotter tape to a table, but it is safer not to risk scrambling data by
passing them through a human brain. The DANDI "EXPORT TO TEXT" feature can
be used to output from each sample file an ASCII text file which can later
be imported to a Lotus 123 spreadsheet. Individual sample files must be
manually retrieved and assigned filenames ':o accomplish this. Once impor-
ted into Lotus, they can be edited and save»d as worksheet files. One of us
(Berkley) has written a Lotus macro which automatically imports ASCII files
one-by-one from a filename list, removes unneeded portions, transposes from
linear to columnar f'ormat, and saves them as Lotus worksheet files. Macros
for combining a list of these worksheet files into tables of concentrations
or retention times were also written. These macros require Lotus 123, Ver-
sion 2.01 or higher. They can be modified to suit the needs of individual
users. They are not protected by copyright, and interested parties may re-
quest copies of them.
Calibration Libraries
New operating conditions were devised to expand calibration libraries
to maximum capability of the PGC. The methyl si 1 icons column at 40°C. 6
milliliters per minute, and gain 200, was used for compounds of intermedi-
ate volatility. The same column at 50CC, 17.7 milliliters per minute, and
gain 500, was used to analyze 1 ess-vo1 ati1 £ compounds, including isomers of
trimethy1 benzene and dichIorobenzene. The KCl/Alumina column at 40°C. 12
milliliters per minute, and gain 1000 can te used to analyze highly vola-
tile compounds such as vinyl chloride and vinylidene chloride.
851
-------
Low Temperature Operation
The extent to which winter temperatures might tax battery capacity or
impair column temperature stability was investigated. No difficulty was
found during all-day operation above 0°C when equipment was taken outdoors
and energized while still warm. However, responses to airborne compounds
of intermediate-to-1ow volatility began to diminish near 10°C, apparently
due to sample condensation on the inlet tube. Concurrent canister samples,
taken for comparison, did not pass through an inlet tube. Data in TABLE 1
show results of concurrent PGC and canister sampling at different tempera-
tures. At 7° C the PGC found less of each compound than did the canisters.
At higher temperatures the two methods were in better agreement. It should
be remembered that canister and PGC samples are not equivalent, because the
PGC analyzes one of seventy milliliters of air which enter the sampling
probe, while a representative sample is analyzed from the entire six liters
which enter the canister during the same period. Usually canister and PGC
samples resemble each other closely, but concentrations in air downwind of
a point source (such as a dry cleaning plant) may be nonhomogeneous and can
lead to significant discrepancies, as seen in TABLE 1. Concurrent PGC and
canister samples for less-volatile compounds are shown in TABLE 2. The PGC
apparently responded to an unknown compound which coeluted with m,p-dichlo-
robenzene and 1,2,4-trimethy1 benzene.
Retention Time Precision
We have previously shown, for a single series of runs, that the cons-
tant-temperature column accessory can reduce relative standard deviations
of retention times at constant carrier flow by an order of magnitude to
about 1% (4). However, rotameters were inadequate for duplicating carrier
flow rates from one series of runs to another. A digital bubble meter,
which can measure flows to within 0.i milliliter per minute, worked much
better. It also revealed that a rotameter attached to the chromatograph
can cause fluctuations in carrier flow. The rotameter ball remained per-
fectly steady while carrier flow varied continuously over a range of half a
mi lli liter per minute. Removal of the rotameter reduced fluctuation to
less than 0.1 milliliter per minute. Setting a preselected flow accurately
(within 0.1 mi 11i1iter/minute) is still difficult because the 3 needle
valves which control it perform erratically, but patience can be rewarded
by day-to-day retention times of good precision. TABLE 3 shows 5 series of
sampling runs during a month in late winter. The data came from two PGC's
equipped with similar columns. Such precision makes it possible to inter-
change libraries. A library saved to disk from one PGC can be used in an-
other, a useful option in case of emergency, but not a recommended proce-
dure for routine operation.
Cone 1 us ions
Photovac portable photoionization gas chromatographs produce valid
results which compare reasonably well with other methods when adequate
operating procedures are followed. To get good results, battery power must
be sufficient, columns must be clean, flow rates must be properly set, the
sample intake line must be clean and reasonably warm, and accurate calib-
ration standards must be used.
852
-------
Ref erences
1. R. E. Berkley, EPA/600/4-86/041, PB67-132858.
2. R. E. Berkley. Proceedings of the 1987 EFA/APCA Symposium on Measurement
of Toxic and Related Air Pollutants. 232, 1987.
3. R. E. Berkley, Proceedings of the 1988 EF'A/APCA International Symposium:
Measurement of Toxic and Related Air Pollutants, 352. 1988.
4. R. E. Berkley, J. L. Varns, U. A. McClenny, and J. Fulcher, Proceedings
of the 1989 EPA/APCA International Symposium: Measurement of Toxic and
Related Air Pollutants. 19, 1989.
Di sc1 aimer
The information in this document has been funded by the United
States Environmental Protection Agency. It has been subjected to
agency review and approved for publication. Mention of trade names or
commercial products does not constitute endorsement or recommendation for
use.
TABLE 1. CONCURRENT PGC AND CANISTER SAMPLES
AT SEVERAL AMBIENT TEMPERATURES
Tri Tetrs-
chloro^ chioro- Ohloro- Ethyl- m,p- o-
Benzene ethene Toluene ethene benzene benzene Xylene Xylene
Monday February
PGC
Cani s ter
Monday Februar
PGC
Canis ter
1.
2.
y
0.
Friday March 26,
PGC
Can is ter
PGC
Canister
0.
0.
26. 1990
29
10
26. 1990
ND
50
1990.
ND
68
ND
76
ND
ND
f
ND
ND
28.
ND
ND
ND
ND
7° C .
9
18°C.
0.
3°C.
0.
0.
Mai 1
ND
90
Dry
ND
50
Dry
ND
56
ND
62
be ! ow
0
3
c
29
9
cl
1
2
4
3
.56
.90
lean
.05
.40
ea ni
.41
.38
.09
. 15
parking deck.
ND
0.50
ing plant.
ND
ND
ng p 1 ant .
ND
ND <•:
ND
< 0.20 <
ND
0.60
ND
ND
ND
0 . 20
ND
0.20
0.
1.
0.
0.
0.
78
20
ND
30
ND
35
ND
29
ND
0. 70
ND
ND
ND
0.20
ND
', 0.20
853
-------
TABLE 2. CONCURRENT PGC AND CANISTER SAMPLES
OF LESS-VOLATILE COMPOUNDS
Monday March 12, 1990. 20<>c. Mall below parking deck.
m,p-Dich1oro-
benzene
or
1,3,5-Tri- 1,2,4-Trimethyl-
methy1 benzene benzene
or or o-Dichloro-
o-Xylene Bromobenzene 4-Ethyltoluene Benzyichloride benzene
PGC
Cani ster
PGC
Cani ster
PGC
Cani ster
0.
0.
0.
0.
0.
0.
10
35
03
25
22
45
0. 01
ND
0.03
ND
ND
ND
0.
0.
0.
<0.
04
ND
37
ND
72
20
13.
<0.
11.
0.
10.
0.
17
20
07
29
51
68
ND
ND
ND
ND
ND
ND
TABLE 3. RETENTION TIME STABILITY DURING FIVE DAYS OF SAMPLING
BETWEEN FEBRUARY 22 AND MARCH 23, 1990.
Ambient Temperature Range 7 - 30°C.
Trichloro- Chloro- Ethyl-
Benzene Ethene benzene benzene m,p-Xylene Styrene o-Xylene
Runs 32 32 32 32 32 32 32
Peaks 13 4 6 12 22 3 13
Mean 128.7 165,1 489,6 567.2 606.1 688.5 735.0
SDEV 2.3 3.5 16.4 19.1 18.5 5.5 IS.8
KSDEV 1.8 2.1 3.4 3.4 3.1 0.8 2.6
854
-------
COMPUTER SOFTWARE FOR GAS CHROMATOGRAPHY IN THE FIELD
C. F. Steele, J. J. Stout, K. R. Carney and E. B. Overton
Institute for Environmental Studies
•12 Atkinson Hall
Louisiana State University
Baton Rouge, La 70803
The availability of small, portable gas chromatographs makes their
use at or near the site of sample collection an attractive prospect. Such
systems can be used for real-time diagnostic purposes, for sample screening
or a combination of both.
The computer hardware requirements of data acquisition and processing
have long been in the range satisfied by off-the-shelf desk-top and lap-top
computers. The state of the art in chromatography software, however, is
not so advanced. Commercially available software packages are designed to
operate in a laboratory and require considerable support and supervision
from the laboratory staff. Under field conditions, the time of the
qualified staff available is too valuable to be allocated except where
necessary and may be unavailable altogether. The authors have developed
and are currently testing a software package that minimizes its need for
the supervision of qualified analytical staff.
The three areas of most concern in chromatography software are the
peak-picker/iritegrator, the analyzer and the calibrator. In the case of
the integrator, most of the need for supervision has been eliminated by
using improved algorithms. The two most troublesome aspects of user
supervision in chromatographic integration, integrator events and slope
sensitivity settings are entirely dispensed with. Identification is
accomplished by game-playing to make optimal, feasible identifications
based upon elution time on two columns, relative peak size and known
elution order. Where possible, coelution is accounted for. The calibrator
reduces the calibration process to a relatively intuitive "point-and-click"
process and supports both single-point and multi-point calibration in a
logically consistent manner.
855
-------
Introduction .
The cost of laboratory analysis, in terms of both dollar cost and
turn-around time make field analysis, where possible, an attractive
alternative. It may be that the general character of a site can be
assessed more quickly and efficiently using field methods and, once that is
known, a more efficient sampling strategy may be established to collect the
samples that need be sent to a contract laboratory. Even in cases where
preliminary analysis is not used, it may still be efficacious to screen
collected samples in order to determine which ones should be sent to a
contract laboratory. Where volatile organics are of concern, gas
chromatography is often the method of choice. Accomplishing
chromatography in the field means that a chromatograph and its associated
data system are operated outside the laboratory environment for which which
most chromatographs and data systems were designed. The absence of
laboratory support necessitates changes to both the gas chromatograph and
its data system hardware and software.
The necessary hardware changes have already been accomplished by
chromatograph and computer manufacturers. There are several portable or
transportable gas chromatographs such as those manufactured by PhotoVac and
by Microsensor Technologies. Portable and transportable computers have
been in use since the late 1970's and simple MS/DOS-based machines are now
under 5 pounds. Advanced systems that are being used as gas
chromatographic and mass spectral data systems, such as the Toshiba 3200
and Macintosh Portable are under 18 pounds.
Gas chromatographic data system software, however, has not changed
significantly since the late 1970's. It has not had to change, being well
suited to its target environment, an analytical laboratory. The main
aspect of this environment from a programmer's point of view is that there
are staff available whose job it is to see that their equipment, both
hardware and software, work properly, including providing the software with
whatever data that it needs to function. In the case of gas
chromatography, this includes partial descriptions of the chromatograms,
such as integrator events and peak detection thresholds, as well as data
pertaining to internal and external standards for the purposes of peak
identification and quantitation. The availability of these data from
external sources makes a programmer's job much easier. He (or She) need
only provide some rudimentary form-entry mechanism to acquire the needed
data.
In the field, things may be quite different. It is appropriate to
minimize the reliance of the data system on the the available personnel in
order to reduce the cost of operating it. The more the data system can
"figure out for itself", the more it matches the ideal case of a black box
that unobtrusively provides answers rather than creating additional demands
upon the available staff. Of course, the answers must be sufficiently
reliable to be useful. In an attempt to address the reliance/reliability
issue, a gas chromatographic software package,"M2001", has been developed
at the Institute for Environmental Studies, Louisiana State University
under contract with the National Oceanographic and Atmospheric
Administration in cooperation with the U.S. Environmental Protection
Agency. Reliance upon external data sources has been minimized while
preserving the reliability of the data provided to the field staff.
Software Design.
The software described here (Figure 1) runs on Apple Macintosh
computers and is designed to acquire and analyze dual-column chromatograms
from a Microsensor Technologies M200 or P200 chromatograph. The main
design criteria were ease of use, achieving improved peak identification by
using elution patterns from both columns and minimizing operator input
while maintaining credible output. This has resulted in significant
improvements to the integrator, calibration system and analyzer as compared
to standard chromatographic software.
856
-------
The Integrator.
The integrator requires almost no operator input. It does not use
integrator events, nor peak detection thresholds. The main uses of
integrator events, establishing baselines and tangent skimming are handled
automatically. The need for peak detection thresholds is avoided by
avoiding first and second derivative tests for peak start and end
detection. Although when necessary, the operator can graphically edit the
integration by changing peak starts and ends, locating missed peaks and
removing false peaks, the goal of the design of the integrator is to remove
completely the need of such remedies.
Peak starts are located by the "Increasing Tangent" method of Woerlee
and Mol1. In this method, the start of a peak is deemed to have occurred
when the tangents of the angles between a point and each of the four
succeeding points increase. This method requires no threshold and
therefore no operator input.
The ends of peaks are detected quite differently. A peak is deemed
to have ended when a line drawn from the start of the peak merges with the
chromatogram. This also requires no threshold setting and, hence, no
operator input.
The scan for starts of peaks begins at the start of the chromatogram
and proceeds until one has been found. After the apex of the peak has been
found, both peak starts and peak ends are acceptable, allowing the start of
one peak to occur on the tail of another (i.e. before the proper end of the
first peak has been found).
When all the peaks have been located, it is generally the case that
groups of them will share common domains (often, a common end point) . Such
clusters of peaks are initially treated as if they each were tangent .
skimmed from the immediately previous peak in the cluster. The clusters
are then post-processed to locate cases of incorrect tangent skimming and
to convert such cases to merged peaks,
Libraries.
The calibration and peak identification process require at
least the names of the compounds that may be identified. Each
compound may be thought of as having a distribution of peak pairs
(one from each of the two columns) existing in a two-dimensional
detection space. When both the names and the statistical
descriptions of the detection space distributions of the detectable
compounds are known, the detection space is said to be characterized
and identification is largely a matter of matching observed peak
pairs with the known distributions. The data pertaining to the
distributions of the possible eluents and their names are stored in
separate files called libraries.
The Calibrator.
Before further analysis can be accomplished, it is necessary to
provide data on the state of the chrornatogcaph being used. This is loosely
termed "calibration" and pertains to both quantitation and qualitation.
Quantitative calibration consists of providing the data necessary for the
program to build a function that maps peak areas into concentration or
amounts. Qualitative calibration provides the program with the data
necessary to calculate the expected retention time of a compound of
interest under the current instrumental conditions. Although qualitative
and quantitative calibration are mathematically separate operations, each
requiring its own independent data and procedures, they come together at
two fundamentally important places; when the operator provides the
calibration, both qualitative and quantitative data will be available and
when the final output is calculated, both qualitative and quantitative
results will be expected. In our software, both qualitative and
quantitative calibration are carried out a': the same time.
857
-------
Quantitative calibration is multi-point with single-point calibration
supported as a special case via a calibration editor (Figure 2). From one
or more previously acquired and integrated chromatograms, the operator
chooses peaks with known identities and concentrations by first selecting
to the name of the compound from the current library, then entering the
known concentration and then pointing to the peaks that are due to that
compound. This process may be repeated for as many different chromatograms
or compounds as appropriate. When all the relevant peaks have been pointed
out, the calibrator builds a quantitative calibration for each compound
involved and a qualitative calibration for each column. It is assumed that
all the calibration chromatograms represent the same instrument at the same
conditions.
For quantitative calibration a regression is calculated from all the
observed area/concentration pairs of each compound, the calibration
function for any specific compound is simply
[c] = Ac + Bc«areac (1)
where
[c] is the concentration of compound c
AC is the intercept of the line of regression
Bc is the slope of the line of regression
areac is the area of a peak supposedly due to compound c
Qualitative calibration consists of the results of two regression
analyses, one performed upon peaks from each column. Guardino's method2 is
used to estimate the mathematical dead time and the slope and intercept of
the line of regression so that
I = A + B-log10 a
prefect match, 0 => not a match), the number of peaks upon which the
identification is based, the natural log of the peak size ratio and the
amount represented by the smallest peak in the identification. The lower
window is one of the two peak list windows, there being one for each
column. It contains the retention time, index, area, amount, and
identifications of each peak detected on a column. The last three columns
pertain to the identification. The column labeled "CF" contains the
certainty factor of each identified component. It is possible to have a
single peak result in two different identifications if there is evidence of
858
-------
coelution as in the case of the peak at 11.28 seconds. The column
immediately to the left the certainty facto:: contains the percent of the
peak accounted for by the specific identification. It is the size of the
peak implied by the amount given in the inventory window divided by the
total size of the peak. If the percentages for a given peak add up to less
than 100%, there may be an unidentifiable coeluent. It is, of course,
possible for a peak to be completely unidentifiable as in the cases of the
peaks at 39.52 and 43.60 seconds.
The analyzer uses a newly developed procedure to convert the
identification process for peaks from multiple chromatographs to a game
and, thence to a zero-one integer (mathematical) program3, the details of
which are beyond the scope of this paper. Briefly, each possible
combination of peaks chosen from each column together with a possible
identity chosen from the library is considered an hypothesis that must be
tested. Each hypothesis (Figure 4) is characterized (from left to right)
by the numbers of a peak from column A, the number of a peak from column B,
a certainty factor, the natural log of the peak size ratio and a possible
identity. An hypothesis may be interpreted as a statement that a specific
peak from column A and a specific peak from column B were both caused by
•;he elution of a specific substance. The certainty factor and log size
ratio are used to assess the validity of tha.t statement. The analyzer
accepts those hypotheses that are mutually consistent and have the largest
•sum of certainty factors while preferring those that involve matching
similarly sized peaks (low log size ratio).
After having performed this task, it may, at the operator's
discretion, analyze the accepted peak pair identifications to see if any
involve matching peaks that do not have similar sizes. If this is the
case, the larger peak is mathematically split into a peak that perfectly
matches the smaller and a left-over peak. The identification is then
recomputed taking account of the left-over peaks which allow the program to
identify some cases of coeluted peaks.
Conclusion.
The authors have developed a relatively easy-to-use aoftware package
(M2001) for doing gas chrornatography in the field. Unlike standard
laboratory packages, M2001 requires little or no user input for integration
and is capable of identifying chemical substances even when standards are
hard to come by.
Acknowledgement.
The authors gratefully acknowledge the: extreme patience of Lawrence
Kaelin, (Roy F. Weston, Inc.) and Michael Solecki (NOAA liaison to EPA-ERT.)
during the beta-testing phase of this project.
The development of the software described here was funded wholly by
the National Ocean Service of the National Oceanic & Atmospheric
Administration, U.S. Department of Commerce under Contract number
50-ABNC-7-00100.
References.
:. Woerlee, E.F.G. and J.C. Mol, "A real-time gas chromatographic data
system for laboratory applications".J. Chromatoyr. Sci . 18. :258(1980)
2 Guardino, X., J. Albaiges, G. Fripo, R. Rodriguez-Vinals and M.
Gassoit,"Accuracy in the determination of kovats retention index.
Mathematical dead Mme.. ". J. Chromatocrr . 118 :13 (1976)
3 Steele, C.F. and E.B. Overton,"Zero-one integer programming as an
identification mechanism in parallel multidimensional
chromatography", in preparation, (1990)
859
-------
Figure 1
M2001 Application Structure
File Edit M200 Sample Trace Display Windows
Calibrate: Cal Gas
Pook
List
fl
Peak
List
LIBfiflRV
Ho. RTfl.
11.28
13.20
19.36
34.88
14296.373
20449.700
223.472
265.560
279.023
ftupynt
14297
21446
223
265
279
4.
4.
ZO1 .
.25
.85
47
.56
.02
74
.03
60
S511
100
100
100
1
1
100
"
.96
.00
.00
.00
.81
.33
00
ppm
ppm
ppm
ppm
ppm
ppm
ppm
25»
II
IOC*
1OOX
100*
loan
CF
816
095
825
997
998
979
007
Substance
air-
voter
butane
pen tone
hexana
heptane
octane
D
H«
a
0
i
a
t
3
3
3
3
4
4
3
3
6
v
24 Hypotheses (cutoff = 0)
8» CF
0 907
2 44
1 772
1 44
2 772
10 351
9 331
5 12«
4 129
4 973
5 973
8 330
7 330
9 241
!•» -Xl*>
R Sub* tone*
0 air
6 Mter
0
6
0
0 Hthul *thul k.torv.
0
1 htxana
0
1
0
0 ftthgl acetate
0
0 cMorofa™
t
o
IP'
ill
;!|i
:=vl
j^l
i||;
Ipi
lip
-iji'i
ijjji
rn
v
Figure 4
Some Hypothetical Identifications
Figure 3
Peak Identification Windows
-------
In EJ.uti.Qn Time Data from Microchip Gag Chromatograifiha i
Ramifications for Sample Component Identification
K.R. Carney, E.B. Overton, C.F. Steele and R.L. Wong
Louisiana State University, Baton Rouge, LA
The introduction of the microchip gas chromatograph in 1975 wae an
early application of photolithography to the construction of analytical
instrumentation1. The result was the "gc on a chip", a complete gas
chromatograph on a silicon wafer the size of a microscope slide.
Technical difficulties at the time, namely etching a good, symmetrical
column into the silicon wafer, precluded large scale development. The
technology was, however, successfully used to fabricate small injection
systems and small thermal conductivity detectors. This was an excellent
and timely complement to the evolving technology for production of good
microbore capillary columns. The Michromonitor 500 and the newer M200
portable gas chromatographs are the descendants of these two
technologies. The M200 uses micromachined injector assemblies and
detectors with two 4-meter, 0.100 mm columns and thus provides the
capability for chromatographic analyses of volatile organic compounds in
under 2 minutes. In addition to short analysis times this combination
provides for excellent short term and long term stability, evident in
retention times and in detector response. The use of two columns with
differing stationary phases also greatly enhances the analytical power
of the system over a single column system. Clearly, the M200 has unique
capabilities with respect to both qualitative and quantitative analysis.
The "M2001" software, written at Louisiana State University, takes
advantage of the multidimensional nature of the information in the
output of the M200. The two chromatograms are searched for peak pairs
corresponding to compounds that may be in the sample. The possible
compounds are distributed in a multidimensional space as points
described by their retention times on each of the two columns. From the
list of possible constituents the software uses an optimization
procedure to select the most likely composition of the sample. The
details of this identification procedure have been described elsewhere^.
Theory
Of primary importance in the use of the M2001 identification
procedure is the availability of a reliable database from which a list
of possible constituents.can be drawn. This database must contain
information about the retention of each compound on each column and also
information about the expected variations due to random error. This
paper will focus on the generation of such a database or "library" and
will discuss important sources of error and ways of avoiding them.
The quality of this database can be addressed in terms of its
accuracy and precision. The accuracy of the database or "library of
compounds" can be assessed by how well th« values in the library predict
experimentally determined retention times, The variable nature of
retention times is well known and, although the M200 generally yields
very reproducible retention times, the likelihood of obtaining the same
retention times on different instruments or even on a single instrument
over an extended length of time is small. The use of a retention index
system for removing the effect of small changes in operating conditions
has been used since the Kovats3 system. Thus the library of Kovats
retention indices was created for a number of volatile organic
compounds.. Because retention indices were known to vary, to a small
degree, with time4 a unique second normalization step was added to the
861
-------
procedure. Prior to analysis a "qualitative calibration" is performed
in which the retention index library is essentially recalculated as
retention times based on the retention times for a mixture of known
compounds. The use of the resulting "pseudo-retention indices"for
component identification is made possible by the high degree of
consistency in retention times obtained with the M200.
Because the identification algorithm uses a statistically based
optimization procedure, accurate estimates for the expected variability
in observed retention times are important. The standard deviations
given in the library for each compound will affect the statistical
weight df the various mixtures proposed by the software. Incorrect
standard deviations can thus lead to misidentifications by, generally,
discriminating against including compounds for which the standard
deviation is underestimated and in favor of including those for which
the standard deviation.is overestimated It is important to note that
this requirement is for accurate standard deviation estimates rather
than small standard deviations per se. The latter is required for what
might be called "qualitative resolution", the ability of the software to
meaningfully distinguish between different but similar compounds. It is
an instrumental requirement well met by the M200 but not what is meant
here.. Accurate estimates of the standard deviation, whether large or
small, are required in order to prevent systematic bias in favor of
choosing particular compounds in the library.
The retention index library was based on the Kovats system, using
the homologous series of n-alkanes as reference compounds. The
retention index (I) for each compound was calculated on the basis of the
log-linear relationship between adjusted retention time and number of
carbon atoms in n-alkane molecules
In(t-tQ) = al + b (1)
where I=number of carbon atoms times 100, t=absolute retention time,
to=column dead time, and a and b are the slope and y-intercept
respectively of the line. The column dead time was estimated by
iterative linear regression of In(t-tQ) on I as per Guardino efc aJ.^
Incorrect estimates for to result in curvature in the In(t-tQ) vs. I
relationship. The curve is concave upward if tQ is underestimated and
concave downward if it is overestimated. The relationship is linear
only if the proper value is chosen for tQ. Thus tg was adjusted to
maximize the correlation coefficient for the linear regression and the
values so obtained for a and ta were used to calculate I for the compound
of interest.
Experimental
Headspace vapors of 35 compounds were individually mixed with
nitrogen containing 100 pprri each of butane, pentane, hexane, heptane and
octane. Concentrations of the 35 compounds were kept between 100 and
1000 ppm to prevent column overloading. Compounds that eluted near one
of the n-alkanes were diluted v-:ith ambient air rather than the alkane
mixture to avoid interference from unresolved peaks. Samples were
initially run at column temperatures of- 60°C using mean carrier gas
velocities of 45 to 50 cm/sec (0.21 to 0.23 ml/minute). Zero grade
helium was used as the carrier gas throughout. Each of the 35 compounds
was run in duplicate. Approximately 3 days were required to complete
the 35 samples. After an approximately 7 day interim, the 35 compounds
were again analyzed in duplicate over a three day period. The analysis
sequence was scrambled to reduce the effect of possible time dependent
862
-------
changes. The entire analysis was repeated a third time again after a 7
day interval. Thus three sets of data were available for calculating
retention indices at 60°C.
After obtaining the 60° retention data, the process was repeated
at temperatures of 80°, 50°, 40°, and 70°. The column head pressure was
not changed so the analyses were actually performed with slightly
different carrier gas flows due to the temperature dependence of the
carrier gas viscosity. This amounted to a variation of about 5% between
the flow at 40° and that at 80°. Each compound was again run in
duplicate but the entire sequence of compounds was only performed once
at each'of these other temperatures.
Results
The retention times of the n-alkanet; proved to be highly
reproducible. The sequence chart in Figure 1 shows both the short term
variability in the retention times as well, as their general stability
over a period of several weeks. .The relative standard deviations of the
alkane retention times were well under O.i>% over the course of each
three day period. Furthermore, the retention times showed no
significant difference between three day periods if the instrument was
left running. The 7 day periods separating each three day series are
represented by solid vertical rules in Figure 1. The instrument was
shut down and the gas flow was turned off for several days between the
first and second series but the column teriperature gas flow was
maintained during the week between the second and third series. The
retention times show a considerable change between the first and second
series while those in the second and third series are essentially the
same. The important feature here is that if the conditions are
unchanged then the retention times are remarkably stable.
.Because of the highly stable nature of the alkane retention times,
the linear regression estimate of equation (1) could be obtained using
the average retention time of each alkane over the approximately 70 runs
in the relevant three day period. This method had two advantages.
First, the confidence intervals for the average alkane retention times
were much smaller; they provided an order of magnitude improvement in
precision over a single observed value. Secondly, by using the average
values for the entire series as reference retention times retention
indices for all compounds are calculated on the same basis rather than
using internal references for some compounds and external references for
those that coelute with or.e of the n-alkanes. A small portion of the
library calculated on the basis of the retention times in Figure 1 is
shown in Table I. They correspond tc the first, second and third series
in Figure 1. The important feature to not.e here is that the retention
times are reproduced very •.-.-ell even though there is a definite shift in
absolute retention times between the first and second series.
Regardless of its stability with time, the library is of limited
use if it is not transportable to clifferent instruments. As an initial
test of library portability, some library compounds were run on new M200
columns. The library was created on an instrument that has been in
continuous service for approximately 18 months. The carrier flow on the
new columns was adjusted to that at which the library was originally
created and a sequence of n-alkanes was analyzed at 4 temperatures. The
results are summarized in Tables II and ITI by the deviations from a
reference chromatogram obtained at the same time on the M200 used to
create the library. Also included are results from a chromatogram
obtained 8 weeks previously which had actually been used in creating the
original library. It appe-irs from the average deviations, shown in
Table II, that one must Irwer the temperature to 59° in order to
863
-------
reproduce the results from the reference M200. Looking at the
deviations in terms of per cent deviation (Table III), a different
conclusion may be drawn. While the 58° result still provides the lowest
total deviation, the 60° result provides an essentially constant
relative deviation from the reference chromatogram. Upon the
logarithmic transformation used in both the the retention index
calculations, relative errors are transformed into absolute errors.
Thus a constant relative error of 5% becomes a constant absolute error
of 0.05 in the logarithmic data. This is exactly the type of error that
is welj. treated by retention index comparison. In other words, even
though the 58° result appears to more closely match the reference data
in terms of absolute deviations, the deviations in the 60° data will be
essentially eliminated as a constant offset value in the logarithmically
based library. This is demonstrated by the summary in Table IV. The
nominal results are those in which the new M200 columns were set at the
same flow rate and temperature as the reference M200 (i.e. 60°C). The
tuned results were obtained by adjusting the column conditions until the
n-aikane retention times most closely approximated those obtained with
the reference instrument (i.e. 58°C). This is somewhat like physical
creating a retention index system. These results parallel those for the
n-alkanes above; while the retention times were forced into a much
better agreement by adjusting the column conditions, deviations were
adequately handled by the mathematical retention index system.
Ideally one should be able to use this retention index library
without regard to temperature. Unfortunately, the temperature
dependence of Kovats retention indices is well known and may, in fact,
contain some useful qualitative information. The results from the
several temperatures at which the compounds were run clearly show a
correlation between temperature calculated retention index. For most
compounds in this library, the variation was less than 8 retention index
units'for a 40° swing in temperature. This can be a significant problem
for early eluters for which a change of 8 retention index units can
correspond to quite a large difference in the retention predicted by the
software. For later eluters the problem may be less significant.
Furthermore, the nature of the correlation between I and temperature is
clearly structure related; all compounds which showed a negative
correlation were oxygen containing compounds such as ketones, esters and
alcohols. All other compounds, including those with some polarity (e.g.
chloroform and ethyl bromide) as well as nonpolar compounds and
aromatics exhibited negative correlations. While the extent of these
variations has not been well characterized at present it seems likely
that one would be able, under some circumstances, to use the library at
temperatures other that that at which it was created, albeit with a loss
of reliability in the resulting identifications. If, however, the
general class of compounds one expects to find in the sample are known
(e.g. halocarbons), then using compounds of that class as standards in
the qualitative calibration would allow use of the library over a range
of temperatures without such a loss in reliability for those compounds.
Reliability for compounds not in that class would, of course be severely
compromised.
Summary
A retention index library containing a number of volatile organic
compounds has been created for use with a statistically based
identification algorithm. The high reproducibility of retention times
obtained from the M200 allows the use of external retention index
standards. These external standards can be used to "calibrate" the
library thus allowing the library to be used with slightly different
864
-------
column conditions or with different M200 instruments. This calibration
creates "pseudo-retention indices" based on observed retention times of
the standards and entries contained in the main retention index library
presented here. The standards need not be, and occasionally should not
be, the normal alkanes required under the true Kovats system. An
observed temperature dependence limits the use of this particular
library to temperatures quite near 60°, Given the fact that many
compound show similar types of correlation between I and temperature,
the usable temperature range may be increased for a similar class of
compounds by using members of that class as qualtitative standards.
References
1. Angell, J.B.; Terry, S.C.; Earth, P.W.; Scientific American,
April, 1983, p.44.
2. Steele, C.F.; Stout, J.J.; Overton, E.B.; Carney, K.R.;
This symposium.
3. Kovats, E,; Helv. Chim Acta, 41,1915 (1958)
4, Guardino, X.; Albaiges, J.; Firpo, G; Rodrigues-Vinals, R.;
Gassiot, M.; J. Chromatogr., 118, 13-22 (1976)
865
-------
UV-73 column at 60°C
u
-------
TABLE II
Absolute deviation (seconds)from reference M200 retention
times
Reference M200
8 weeks
previous
n-pentane
n-hexane
n-heptane
n-octane
0
0
0
0
.15
.15
.13
.04
-0
-0
-1
-2
60°
.52
.88
.45
.24
5
-0
-0
-0
-0
go
.44
.60
.65
.08
58°
-0
-0
0
?.
.32
.24
.31
,56
55°
-0
0
3
9
.04
. 68
.03
.96
AVERAGE
0.116 -1.272
-0.442
0.578
3.408
TABLE III
Per cent relative deviation from reference M200 retention
times
Reference M200
8 weeks
previous
n-pentane
n-hexane
n-heptane
n-octane
1
0
0
0
.08
.75
.37
.06
-4
-4
-4
-3
60°
.06
.62
.32
.33
59
-3.
-3.
-1 .
-0.
o
42
11
89
12
58°
-2 .
-1 .
0.
3.
46
22
88
55
55°
-0.
3.
7.
12 .
30
30
96
52
AVERAGE
0.56 -4.08
-2.13
0.19
5.87
TABLE IV
Relative standard deviations between instruments
Substance
Acetone
Methylethylketone
Dichlorobromomethane
Dibromochloroinethane
Rentention Time
Nominal
6.70
7.17
8.85
8.17
Tuned
0.9C
0.64
0.01
0.14
Retention Index
Nominal
0.17
0.14
0.04
0.05
Tuned
0.13
0.02
0.07
0.05
867
-------
A FRACTIONAL (FRACTAL) BROWNIAN MOTION MODEL
OF ATMOSPHERIC DIFFUSION
F. A. Gifford
109 Gorgas Lane
Oak Ridge, TN 37830
Present atmospheric di-f-fusion model• assume, directly or implicitly,
that atmospheric turbulence motions are o-f the classical Brownian type. A
nineteen-month series of 12-hour averaged values of Kr85 concentrations
measured at Murray Hill, NJ, 1050 km downwind from the source near Aiken SC
has been analyzed using a method o-f fractal geometry, the R/S (renormal-
ized-range) statistic. The result achieves a high degree of organization of
the initially extremely variable time-series of observed air concentration
values. The slope of the R/S curve indicates that large-scale atmospheric
diffusion is controlled by large-scale, random air motions that are quali-
tatively different from the usual assumption of Brownian motion. Real atmo-
spheric motions at large scales are richer in high-frequency components than
their Brownian counterpart, and exhibit correlation at all scales. The frac-
tal nature of this scale of turbulent atmospheric motions, and a computer
algorithm for generating the appropriate fractal Brownian motion for use in
diffusion models, are briefly discussed.
868
-------
INTRODUCTION
In very general terms, turbulent tropospheric motions appear to be a mix-
ture of two rather distinct kinds: large, quasi-horizontal, essentially two-
dimensional motions o-f scales greater than several hundred kilometers; and
smaller, three-dimensional eddy motions. The -former cascade eddy enstrophy
(mean-squared random vorticity) from the very large scales of eddy kinetic-
energy generation (several thousand km) to scales of a few hundred km. The
latter dissipate this random vorticity by rapidly attenuating, distorting and
concentrating it and cascading the eddy kinetic energy to the very small «1
cm) scale of viscous dissipation (Gifford^). This view of the nature of
atmospheric turbulence is supported by the extensive measurements of eddy-
energy spectra reported by Nastrom and Gage19 and by the atmospheric dif-
fusion data summarized by Gifford7»8 and Barr and Gifford2.
A cloud or plume of contaminant material diffusing in such a field of
random eddies is in effect sampling the turbulence structure through an ever-
increasing volume of the atmosphere. At first, in the energy-cascade region,
spreading is rapid because, as was pointed out by Batchelor', the range
of eddy sizes present always includes those that are just the size of the
plume. Later, after many hours of downwind travel, the cloud expands into
the enstrophy-cascade range of eddy sizes. This range (Gifford et al.12t
'Bifford11) is characterized by rapid cloud distortion due to the large
eddies, accompanied by diffusion at a slower rate. Large clouds quickly
develop very irregular outlines, as shown by the long volcano plumes studied
by Gifford1* and the evolution of the Chernobyl plume discussed by
Smith20. Although it has so far always been ignored in (short-range)
•studies of atmospheric diffusion of clouds and plumes, the large outer
time-scale of turbulence, about 10^ sec, implied by this two-range
view of atmospheric turbulence structure (Gifford9*H»*2) has been shown
also to influence diffusion at smaller scales (Gifford10),
FRACTAL DIMENSION OF POLLUTANT CLOUDS
The observed shapes of energy spectra in the earth's troposphere, es-
pecially the extensive GASP spectra19, strongly suggest a large-scale,
enstrophy-cascade range and a smaller-scale, energy-cascade range of turbu-
lent atmospheric motions. The eddy dynamics of each range are governed, re-
spectively, by random enstrophy and energy transfer and can be assumed to be
quantitatively described solely in terms of the corresponding similarity
parameters (Gifford11). It follows that passive scalar functions of the
turbulence, such as pollutant concentration, will exhibit self similarity.
That is, changes in concentration must depend on time in such a way that
i.n a statistical sense, for any two times tj and t2»
C «
-------
the -Fractal dimension, D, o-f the perimeters of clouds and rainbands and by
inference o-f atmospheric turbulent eddies, equals 1.35 over a broad range
of eddy scales. Giffordl3 showed that this applies up to linear scales
o-f about 300-400 km, or time scales of a -few times 10^ seconds, but that
at larger scales, in the enstrophy-cascade region, DEI. 8, indicating turbu
lence o-f a markedly different type.
Squaring equation (1) and averaging gives the variance
a $t2H ; (3)
SC is the concentration increment over the time lag, £t=t2-ti. Also
from Eq.(l), the so-called "renormalized-range" statistic (Mandelbrot*8)
can be written as
T) - Cmin(T)]/ffc(T) • R(T)/S(T) = br , (4)
where R is the range and S <=l/2 VSii jt, or of R/S vs. T should contain broad, linear ranges
if atmospheric turbulent motions are self -similar. Atmospheric values of the
similarity exponent H, and of the fractal dimension D o-f atmospheric turbu-
lence, follow directly from these slopes.
FRACTIONAL BROWNIAN MOTION IN ATMOSPHERIC DIFFUSION MODELING
All atmospheric -diffusion models assume that turbulence can be simulated
by Brownian motion. This is done either directly, as in models of the
Lagrangian particle motion (e.g. Hanna14; Gifford6) or indirectly,
through the assumption of Gaussian turbulence statistics in plume models or
by the use of K-theory. The turbulent velocity v(t), and consequently the
position of a pollutant particle, can be expressed as
v(t+*t) = v(t) + r(t) (5)
(Hanna*4), where r(t) is a random velocity with Gaussian statistics; r(t)
is the integral of a random acceleration, a(t), which has Brownian motion
(white noise) statistics,
r(t) • a(x)dx . (6)
The quantity r is called the trace of the Brownian motion process, a. Such a
Gaussian random process has the well-known properties, =0, and
(= represents ensemble averaging.
Mandelbrot18 proposed the following generalization of the classical
Brownian motion model;
a t2H , (B)
where 0
-------
narily white. By fitting various kinds of long, natural time-series (river
discharges, tree ring data, varves, etc.) to Eq.(4>, it has been found
(Feder5) that often H ~ Q.7-Q.B. The assumption that atmospheric turbu-
lence is of the Brownian motion type, implying that H = 1/2, is known to re-
isult in fairly good estimates of atmospheric diffusion (cf. Bifford^, for
example) which is, of course, why it is a staple of diffusion modeling. Vet
•systematic departures are known to occur and should be of concern to model-
Mrs. Theoretical arguments, for instance by Hentschel and Procaccia^, in
terms of fractal geometry for the case of fu). ly-developed homogeneous turbu-
lence indicate that H should be about 1/3.
Fractal Brownian motion (fBm) curves for various values of H are quali-
tatively quite different in shape. For H-values of 0.7-0.8, corresponding
to the river-discharge, varve, and tree-ring data, high-frequency spikes
are small and the fBm curve is relatively smooth. For H=l/2, the white-noise
of ordinary Brownian motion, all frequencies are equally represented and
the curve is uniformly rough. For small values of H, closest to the case of
atmospheric turbulence, the fBm curve is decidedly rougher at all scales.
The physical meaning of H is perhaps most readily conveyed by relating
fBm to the more familiar power-spectrum representation. In a similarity
region of the spectrum, the slope is equal to l/f&, f being frequency.
The fractal dimension of the spectrum equals (Mandelbrot^)
D«E+1-H«E + <3-G>/2 (9)
where E is the spectrum's Euclidean dimension. For time spectra £»! and
D=2-H. Since 0l/2 the fBm exhibits persistence; de-
partures from the mean tend to be followed by even larger departures. In the
limit of large time, the correlation remains positive. For H
-------
Values of the statistic R/S, Eq.(4), Here calculated far time intervals
ranging from 30 to 1158 half-days from the start of measurement. The result-
ing curve, Figure 2 (open circles), is quite interesting) it contains three,
roughly linear segments, each having about the same slope but separated from
its neighbor by a sharp jump in R/S. Inspection of the concentration ob-
servations shows this to be due to the presence of two sharp spikes in the
Murray Hill data; these may have been caused by Kr85 from the Oyster Creek
(NJ) reactor, the only nearby source capable of producing such features. The
two spikes were removed by a linear interpolation between adjacent concen-
tration points, a minor adjustment of 4 data values. The resulting modifi-
cation to the R/S values is shown as the dotted curve (2) in Fig.2. The ex-
traordinarily high degree of organization produced by this simple transfor-
mation o-f the initially random data of Fig.l has no parallel in the writer's
experience of atmpospheric turbulence data and argues strongly for the frac-
tal nature of atmospheric turbulence at these large (>300-400km) scales.
A straight line fitted to the log-log plot of the adjusted R/S-curve
points produces the value H=0.36, corresponding to the fractal dimension D"
1.64. It is much too early to draw final conclusions about the general sig-
nificance of this value, except that it certainly agrees with both the theo-
retical results by Hentschel and Procaccia*6 for the fractal dimension of
large-scale relative diffusion, and with the observed value for aqueous
clouds spreading in the enstrophy-cascade range found by Gifford**. it
appears, from the result, that atmospheric turbulence is not in general
characterized by H=l/2, as usually assumed. At least at larger scales it has
an H-value on the order of 1/3. Fortunately, the methodology for introducing
fBfli with proper H-values already exists, having been developed in connection
with the seemingly unrelated problem of the video-simulation of natural
landscapes and clouds (Mandelbrot19; see also Feder5 and Barnsley ejt
al.1). Computer algorithms for generating the fBm curves discussed in
the last two references are based on the following generalization of Eq.<6),
r(t) » C1/P(H
t
(t - t')^ dr(t') , (10)
which is seen to reduce to (6) when H»l/2.
CONCLUSION
The possibility of introducing realistic fBm into atmospheric diffusion
models is one of the most attractive applications of the fractal geometry
methodology and distinguishes it sharply from the usual similarity theory of
turbulence spectra, to which it is related by Eq.(9). Verifying that H*l/3
for large-scale motions, and finding its value at other atmospheric scales,
are urgent research goals. This approach has the potential to free diffusion
modelers from the need to make an assumption about the nature of atmospheric
turbulence that now appears to be unrealistic.
REFERENCES
1, Barnsley, M. F., R, L. Devaney, B. B. Mandelbrot, H.-O. Peitgen, D. Saupe,
and R. F. Voss, The science of fractal images. Springer-Verlag,NY,19BB.
2. Barr, S. and F. A. Gifford, "The random force theory applied to region-
al scale atmospheric diffusion," Atmos. Environ.. LJS21, 1737-1741.
(1987).
872
-------
3. Batchelor, G. K.p "Diffusion in a field of homogeneous turbulence, II.
The relative motion of particles," Proc. Camb. Phil. Soc,. U\48,
345. (1952).
•4. Farmer, J. D. and J. J. Sidorowich, "Exploiting chaos to predict the
future and reduce noise," in Evolution. Learning, and Cognition. Ed.
by Y. Lee, World Scientific Pub. Co.,
Singapore, p277. (1988).
15. Feder, J., Fractals. 2B3pp, Plenum Press, NY, 1988.
6. Gifford, F. A., "Horizontal diffusion in the atmosphere! a Lagrangian,
dynamical theory.Atmos. Environ.. 16, 505-512.(1982).
7. Gifford, F. A., "Atmospheric diffusion in the mesoscale range: the evi-
dence of recent plume width observations, ' Preprint Vol., 6th Symp. on
Turbulence and Diffusion, March 22-25, 1983, Amer. Met. Soc. (1983).
3. Gifford, F. A., "The random-force theory: application to meso- and large-
scale atmospheric diffusion," Boundary Layer Meteor..38.159-75.(1984).
c?. Sifford, F. A,, "Atmospheric diffusion in the range 20 to 2000 km," in
Air Poll. Modeling and its Applications. Plenum Press, 247-265. (1985).
10.Gifford, F. A., "The time-scale of atmospheric diffusion considered in
relation to the universal diffusion function fj," Atmos, Environ..
21, 1315-1320, (1987).
1.1.Gifford, F. A., "A similarity theory of the tropospheric turbulence
energy spectrum. "J. Atmos. Sci.. 4Ji, 1370-3,379. (1988).
12.Gifford, F. A., Sumner Barr, R. C. Halone, and E. J. Mroz, "Tropospheric
relative diffusion to hemispheric scales," fltmos. Environ.. 22,
1871-1879.(1988).
13.Gifford, F. A., "The shape of large tropospheric clouds, or 'Very like a
whale'," Bull. Am. Meteor. Soc.. 7_P_, 468-475. (1989).
14.Hanna, S. R., "Some statistics of Lagrangian and Eulerian wind fluctu-
ations," J. APDl. Meteor., it, 518-525. (1979).
IS.Heffter, J. L., J. F. Schubert, and G. A. Mead, "Atlantic Coast Unique
Regional Atmospheric Tracer Experiment (ACURATE)," NOAA Tech. Memo. ERL
ERL ARL-130, 60pp. (19B4).
16.Hentschel, H. G. E, and I. Procaccia, "Fractal nature of turbulence as
manifested in turbulent diffusion," Phvs. Rev. Letters. A, £7,
1266-1269. (1983).
17.Lovejoy, S., "The area-perimeter relationship for rain and
cloud area," Science. 216. 185-187. (1982).
18.Mandelbrot, B. B., 1983s The fractal geometry of nature. 486 pp,
W. H. Freeman Co., NY, 1983.
19.Nastrom, G, D., and K. S. Gage, "A climatology of atmospheric wavenumber
spectra of wind and temperature by commercial aircraft," J. Atmos. Sci..
42, 950-960. (1986).
873
-------
20.Smith, F. B., "The deposition of Chernobyl ce*sium-137 in heavy rain and
its persistent uptake by grazing sheep," ftqric. and Forest Meteorology.
17, 163-177. U9B9).
o
E
o
CL
O
O
o
• en-
o :
00~
4 4
^0
4 41
200
4
i i ! i r | i i
400
4
4
4
1 1 1 1 1 1
600
4
T 1 t
4
800
4
I I t
1
I I | I I t 1
1000
1 TTT r j
1200
TIME, HALF-DAYS
Figure 1 Successive 12-hour average values of Kr85 air concentrations
measured at Murray Hill, NJ, -for 19 months beginning March, 1982. Down
arrows indicate zeros and up-arrows indicate points above 25
10-
"1—I I I I I
i—r i i
10
100
1000
TIME, HALF-DAYS
Figure 2 Renormalized range, R/S, vs. time interval, T, for Murray Hill
Kr^S data, based on Eq.(4). Open circles, curve (1), are based on the
1158 unmodified data points of Fig.l. Small, solid dots, curve (2), show
the effect of smoothing the two highest data peaks (4 data points).
874
-------
Numerical Simulations of the Mountain Iron Tracer Data
Tetsuji Yam ad a and Susan S. Bunker
Los Alamos National Laboratory
Los Alamos, New Mexico
Extensive field experiments were conducted during 1965 and 1966 near Van-
denberg Air Force Base. The experiments included wind speed and wind direction
measurements at several towers, upper air soundings by radiosondes, and fluorescent
particle releases to characterize the diffusion processes.
The data provide a unique opportunity to test numerical models under re-
alistic boundary conditions: land-sea contrast and complex topography. We used
HOTMAC (High Order Turbulence Model for Aii Circulations 1, a three-dimensional
mesoscale model based on simplified turbulence-closure equations to simulate tem-
poral and spatial variations of wind, temperature, mixing ratio of water vapor, and
ttirbulence distributions.
Surface concentrations were computed by using a three-dimensional transport
and diffusion model RAPTAD (Random Puff Transport and Diffusion). RAPTAD
is a Lagrangian puff code based on the Monte Carlo statistical diffusion process.
The center location and standard deviation of concentration distribution for each
puff are computed by using wind and turbulence modeled by HOTMAC. Then, the
concentration at any location is computed by summing concentrations contributed
by all the puffs.
875
-------
Introduction
The purpose of this study is to simulate the transport and dispersion of at-
mospheric poDutants in the complex terrain surroundings at Vandenberg Air Force
Base (VAFB) by using the Los Alamos National Laboratory (LANL) atmospheric
models HOTMAC and RAPTAD.1
HOTMAC is a prognostic model and solves a set of time-dependent physi-
cal equations such as conservation equations of momentum, internal energy, and
mixing ratio of water vapor. Prognostic models can forecast three-dimensional dis-
tributions of wind speed, wind direction, temperature, mixing ratio of water vapor,
and turbulence variables.
HOTMAC provides RAPTAD both mean and turbulence variables to simu-
late transport and diffusion processes of airborne materials. Only a few mesoscale
atmospheric models can forecast three-dimensional variations of atmospheric turbu-
lence. Therefore, HOTMAC and RAPTAD offer a considerable improvement over
the current emergency response management models at VAFB that are extremely
simple.
The Mountain Iron Diffusion Experiments
The Mountain Iron (MI) diffusion experiments2 were conducted at VAFB
during 1965 and 1966 to establish quantitative diffusion predictions for use as range
safety tools in the "South Vandenberg" (SV) ballistic and space vehicle operations.
The experimental site, SV, is located along the California coast approximately
160 km west-northwest of Los Angeles. The coastline is oriented in approximately
a north-south direction along the western side of SV, but changes abruptly at Point
Arguello to an east-west direction. The coastline gradually changes to a north-
south direction down to Point Conception and then changes again to an east-west
direction. The Santa Ynez Mountains form an east-west barrier along the coastline
far south of SV.
Fluorescent pigment zinc sulfide particles with a geometric mean of 2.5 microns
in diameter were released to understand transport and diffusion processes and derive
an empirical formula for the pollutant concentration distribution in the SV area.
The effective release height was 2 to 6 meters above ground. The primary sampler
used was a membrane filter inserted in a disposable polyethylene holder. The bulk
samples from the field were assayed by use of the Rankin counter, which uses an
alpha emitter to activate the fluorescent pigment deposited on the membrane filter.3
876
-------
7J5
x (km)
725
Figure 1. Modeled horizontal wind
vectors at 6m above the ground at
1300 1st, June 13, 1966. Terrain
is contoured by solid lines with an
increment of 200 m. Dashed lines
indicate contours halfway between
the solid contours.
_ J m/s
730
x (km)
725
730
Figure 2. Same as in Figure 1 except
observed wind vectors in the surface
layer are shown.
0 2 4 6 8 10
wind speed (m/s)
0 90 180 270 360 278 284 290 296302 306 3H
wind direction pot temp (k)
Figure 3. Vertical profiles of the modeled wind speed, direction, and potential
temperature at 1300 1st, June 13, 1966 at B22. Solid circles indicate observations.
877
-------
Results and Discussions
A total of 113 tracer releases was made from which we selected the MIST, MI90
and MI91 cases to evaluate the performance of HOTMAC and RAPTAD. Only the
results from the MIST simulations are reported here.
The initial potential temperature profile was determined by averaging the
four upper air soundings at 1300 local standard time (1st) in SV. The potential
temperature lapse rate was approximately 0.044" C/m from the sea surface to 460 m
mean sea level (msl), 0.0142° C/m between 460 m and 960 m, and 0.0045° C/m
above 960 m msl. Wind speed and direction were determined by examining five
upper air soundings (four locations mentioned above plus VIP-1) at 1300 1st. Intial
upper air wind speed and wind direction were estimated to be 2 m/s and 225 degrees,
respectively.
The computational domain is 40 x 48 km2 with a horizontal grid spacing of
1 km. To resolve the details of topography in the vicinity of the release site, we
decided to nest a fine resolution grid 15 x 16 km2 with horizontal grid spacing of
0.5 km.
Integration started at 0500 1st, June 13, 1966 and continued for over 12 hours.
The plume was released at 1310 1st for 30 minutes as was done in the experiment.
The plume was followed for 4 hours in the model computation. By that time the
plume was transported far away from the sampling areas.
Figure 1 shows the modeled horizontal wind vectors in the inner computational
grid at 6 m above the ground at 1300 1st, June 13 (Julian day 164). Although the
upper air wind direction is 225 degrees (southwest), upslope flows develop in the
surface layer due to heating at the sloped surfaces.
The modeled wind distribution (Figure 1) is in good agreement with the ob-
servation (Figure 2). The observed winds show much more variations in space than
the modeled winds. Observations adjacent to each other show considerable varia-
tions in direction and magnitude. On the other hand, the modeled wind field varies
more slowly in space than the observed since the model neglects subgrid scale vari-
ations of the surface (the grid resolution is 500 m). Nevertheless the simulation
successfully reproduced many features observed.
Less satisfactory results are obtained in comparison of the vertical profiles of
the modeled wind speed and wind direction with observations (Figure 3). Wind
speed and wind direction become highly variable in space and time when the pre-
vailing wind speed is small. It is noted that the observations were instantaneous
values whereas the modeled results are ensemble averages. On the other hand, po-
tential temperature is relatively stationary unless synoptic scale disturbance such
as fronts pass through the measurement area.
Significant changes in the modeled wind direction occurred at around 600 m
above the ground. This is caused by the mass conservation constraint to compensate
the divergence and convergence of the wind distributions in the boundary layer
878
-------
(Figure 3). Observed wind direction profiles appear to support such variations
but the changes appear to occur at heights much closer to the ground than those
modeled.
Figure 4 shows the modeled ground level concentration contours and Figure 5
shows the corresponding observation. Figure 5 also shows the observed wind speeds
and wind directions at the ground stations. The observed wind direction close to
the release site is west-southwesterly, but changes to westerly at the station slightly
north of the release site. Our simulation (Figure 1) indicates that the wind direction
is close to westerly at the release site. The observed plume apparently transported
to the east-northeast direction despite the fact that wind directions measured at
the ground stations suggest the plume should be transported to the east-southeast
which is the case for the modeled plume.
-. 3mf$
3838. -
3833.
4J
3
3828.
3823
•MS.
720. 725.
utmx (km)
730.
715 720 725 730
x (km)
Figure 4. Modeled ground level con- Figure 5. Observed ground level con-
centration (accumulated). centration (accumulated).
Although the modeled plume direction did not match the observed one, the
modeled ground level concentration along the plume axis are in good agreement
with observations as shown in Table I. It is not known why the observation at
3310 m from the source shows the largest value among the observations.
879
-------
Table I: Normalized Concentrations
Modeled
Distance from
the source
(m)
716
1253
3312
4403
Concentration
8.04 x 10~6
2.55 x 10~6
9.06 x 10-7
7.08 x 10"7
Observed
Distances from
the Source
(m)
720
1260
3310
4400
Concentration
3.97 x 10~6
1.08 x 10~6
5.39 x 1Q-6
3.02 x 10~7
Acknowledgements
The authors are grateful to Dr. W. Clements for reviewing and K. Coen for
typing the manuscript. The work was supported by the U. S. Air Force Engineering
Service Center, Tyndall Air Force Base, and was performed under the auspices of
the U. S. Department of Energy at Los Alamos National Labortory.
References
1. T. Yamada, and S. Bunker, "A Numerical Model Study of Nocturnal Drainage
Flows With Strong Wind and Temperature Gradients," J. Appl Meteor., 28,
545-554 (1989).
2. W. T. Hinds, and P. W. Nickola, "The Mountain Iron Diffusion Program:
Phase I South Vandenberg: Volume I," AEC Research and Development
Report, Pacific Northwest Laboratory, AFWTR-TR-67-1, BNWL-572 Vol 1
(1967).
3. W. T. Hinds, and P. W. Nickola, "The Mountain Iron Diffusion Program:
Phase I South Vandenberg: Volume II," AEC Research and Development
Report, Pacific Northwest Laboratory, AFWTR-TR-67-1, BNWL-572 Vol II
(1968).
880
-------
THE INCLUSION OF POLLUTANT REMOVAL PROCESSES
IN URBAN AIR QUALITY MODELS
K. Shankar Rao
Atmospheric Turbulence and Diffusion Division
Air Resources Laboratory, NOAA
Oak Ridge, Tennessee
and
James M. Godowitch*
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina
Gaseous and particulate pollutants emitted into the atmosphere are removed by
several natural processes. Important among them are the dry deposition of pollutants
at the earth's surface, and chemical transformation in the atmosphere. These removal
mechanisms affect the pollutant concentrations and residence times in the atmosphere
and, therefore, it is necessary to account for them in air quality models. This is
particularly important for urban air quality models, which are often used to assess
the risk associated with chronic exposure of population to toxic air contaminants.
This paper describes a methodology for including dry deposition and a first-
order chemical transformation in urban air pollution models based on the Gaussian-
diffusion framework. The concentration algorithms for point sources are derived from
analytical solutions of a gradient-transfer model. In the limit, when deposition and
settling velocities and the chemical transformation rate are zero, these expressions
for various stability and mixing conditions reduce to the familiar Gaussian plume
diffusion algorithms without the removal processes. The point-source algorithms
are integrated to obtain the concentrations due to emissions from distributed urban
area sources. A new mathematical approach, based on mass budget considerations,
is outlined to derive simple expressions for ground-level concentrations. The
concentration and deposition flux formulations described in this paper are currently
used in several of EPA's air quality models.
On assignment from the National Oceanic and Atmospheric Administration (NOAA),
881
-------
Introduction
Pollutant gases and particles released into the atmosphere are transported by the
wind, diffused and diluted by turbulence, and removed by several natural processes.
Among the important removal mechanisms are chemical transformation in the
atmosphere, and dry deposition of pollutants at the earth's surface by gravitational
settling, turbulent transfer, chemical adsorption, and other effects. Depletion of
airborne pollutant material by these physical processes affects its concentrations and
residence times in the atmosphere. Surface deposition of pollutants may adversely
impact on human health, local ecology, structures, and monuments. It is necessary,
therefore, to consider the removal processes in air quality models in order to obtain
reliable estimates of the concentrations and surface deposition fluxes.
This paper describes a methodology for incorporating the dry deposition and
first-order chemical transformation of gaseous or particulate pollutants in urban air
pollution models based on Gaussian plume-diffusion assumptions. The concentration
algorithms are derived from analytical solutions of a gradient-transfer model1'2.
Model Formulations
We consider the steady state form of the three-dimensional atmospheric
advection-diffusion equation (see Rao2) for the concentration C of the pollutant:
U dC/dx = Ky d2C/dy2 + Kz d2C/dz2 + W dC/dz - C/rc (la)
Here, x,y, and z are the horizontal downwind, crosswind, and vertical coordinates,
respectively; U is the constant average wind speed, Ky and Kz are the eddy
diffusivity coefficients, W is the gravitational settling velocity of the pollutant, and
TC = l/&t is the time scale associated with the chemical transformation which
proceeds at a known rate kt. For a continuous point source, which emits pollutant at
a rate Q from (x = 0, y = 0, z — H), the initial and boundary conditions are given by
,*) = Q/U.6(y).8(z-H) (U)
C(z,±oo,z) = 0 (lc)
[Kt dC/dz + W C] = [Vd C] at z = 0 (Id)
C(z,t,,oo) = 0 (le)
where 6 is the Dirac delta function. Equation Id states that, at ground-level, the sum
of the turbulent flux of pollutant and its downward settling flux due to the particles'
weight is equal to the net flux of pollutant to the ground, resulting from an exchange
between the atmosphere and the surface. The deposition velocity Vj, characterizes this
exchange. When deposition occurs, the turbulent flux of the pollutant at the surface
(z = 0) is given, from Equation Id, by - we = Kz dC/dz = (Vd-W) C > 0
which implies that Vd > W > 0. The deposition boundary condition (Equation Id),
suggested by Monin2 and Calder3, was discussed by Rao1.
The exact analytical solution of Equation 1 can be written (see Rao1'2) as
C(x,y,z) = Q/U . gi(x,y}/Ly . g'2(x,z}/Lz (2)
882
-------
where g\ and g'2 are nondirnensional functions, and Ly and Lz are length scales
characterizing the plume diffusion. In order to facilitate the practical application of
this solution, we express Ky and K2 as
Ky = 0.5 U da2y/dx , Kz - 0.5 U da2Jdx (3)
where ay and az are the widely used Gaussian plume dispersion parameters. This
will permit utilization of the empirical data on these parameters for a variety of
meteorological and terrain conditions.
Parameterization of Concentration
In order to parameterize and simplify the expressions for concentration, we
define, following Rao1'2, the following nondiraensional "capped" quantities:
Vd = Vd/U , W = W/U , Vj = (Vd - W/2)/U , V2 = (Vd - W)/U (4a)
x = x/\/2ffz , z = z/V2
-------
where g'4(x) = $^[g'<2/ LZ\H=Q dz. For gases or small particles (Vd ^ W), this
integration yields
g't(x) = ezp(-/32-z/rc)[(Vi/V2)e«r/c(£) - (W/2V2) efterfc(P)] (66)
where £ = 2V\x and /3 = Wx. The expression for large particles (Vd = W) is
(6c)
where £ = V^i = Wx. When deposition and chemical loss are negligible, Equation
6 reduces to the familiar Gaussian plume model with g\ = g± — 1. The plume
is generally considered to be well-mixed for x > 2xm , where xm is the downwind
distance x at which 2.15 crz(x) = L.
In the region xm < x < 2xm, where the plume is considered to be trapped
between the ground and the stable layer aloft, the mixing depth L should be explicitly
included in the concentration algorithms. This can be done by writing the equation
for g'3(x, z] following Rao1, incorporating multiple reflections of the plume from both
the ground and the stable layer. Alternately, the ground-level concentrations in this
region can be estimated by linearly interpolating between the concentration values at
xm and 2xm on a log-log plot of concentration versus downwind distance.
Once the ground-level concentration is determined, the surface deposition flux of
the pollutant can be calculated directly from D(i,y) = Vd C(x,y,0), where D gives
the amount of pollutant deposited per unit time per unit surface area.
Area Sources
The urban area-source emission inventory is developed by dividing the city into
equal-sized grid cells, each typically a square of 2-5 km side, and representing the
total of all low-level emissions of the pollutant in each grid cell by an equivalent area-
source emission. The concentration from an area source is generally calculated by
integrating the point-source algorithms over the area. We consider two equal grid
squares (see Figure la), one of them containing the area-source emissions Q, assumed
to be located at the center of the square, and the other containing a ground-level
receptor R at its center. The wind U blows along the line from Q to R as shown.
Then the surface concentration C A at R due to the area source Q is given by Rao5
as
CA = (Q/U) fX\g'2(x,Q)/Lz}dx (7)
Jxi
where x\ and £2 are the distances from the receptor R to the downwind and upwind
edges, respectively, of the emission grid square. Since the two grid squares are equal
in size, these distances can also be measured from Q to the upwind and downwind
edges, respectively, of the receptor grid sqaure, as shown in Figure la.
If deposition and chemical transformation are neglected, and if the urban area-
source emissions are assumed to occur at ground-level (H = 0), then Equation 7
reduces to
CA = V^ (Q/U) I"' dx/
-------
atmospheric stability, to derive simple algorithms for the concentrations from area
sources.
Equation 7 can be easily adapted to account for distributed urban area sources
and multiple receptors2'5. Assuming that az — axb and H — 0 in Equation 7,
Rao2 outlined an elegant new mathematics! approach, based on mass budget
considerations, to derive a simple expression for ground-level concentration from area-
source emissions. This is schematically illustrated in Figure Ib. For each receptor
grid square box formed by the ground surface and two imaginary vertical planes at
x — xi and x = x2, the pollutant mass budget can be written as
Incoming flux — outgoing flux ;t flux gain/loss due to chemical
transformation = surface deposition flux.
Substitution for the various terms above leads to the final expression2 for CA'-
CA = Q/[2(l-b)Vd}{g[(xi)- g((x2)-l/(Urc) f" g't(x) dx} (9)
Jx\
where g'4(x) is the point-source algorithm (Equation 6) in the well-mixed region.
Equation 9 is computationally efficient2'7 because it permits one to use the same
subroutine for both point and area sources.
All of the equations given above for C'4 ignore horizontal diffusion. This is
justified on the basis of Gifford's narrow plume hypothesis6, which postulates that the
concentration at a receptor is influenced only by the distributed area sources located
in a fairly narrow, plume-shaped upwind sector. The concentration downwind of the
center of the emission grid is then the same as that if the area source were infinitely
wide in the crosswind direction.
Conclusions
In this paper, we have briefly described a realistic methodology for including
deposition, gravitational settling, and first-order chemical transformation in applied
urban air pollution models. The concentration expressions given here can be thought
of as analytical extensions of the familiar Gaussian plume dispersion algorithms
to include these removal processes. Empirical values of deposition velocities for a
wide variety of pollutants and surface and atmospheric conditions can be found in
the literature. Some guidance is provided in Rao''2 for specifying the deposition
and settling velocities in the concentration algorithms. These algorithms are now
optionally available in EPA's air quality models, MPTER-DS, PAL-DS, INPUFF-2,
and PAL-2.
Frequently, the product of a chemical reaction may be the pollutant of primary
concern, rather than the reactant itself. A well-known example is the atmospheric
transport and transformation of SOa to sulfate. In general, the secondary (product)
pollutant will have deposition and settling velocities which are different from those
of the reactant species, and it may also be directly emitted from the sources. The
concentration algorithms for such coupled pollutant species, which are also derived
by Rao2 from the gradient-transfer model, are considerably more complex than
those given here. These algorithms are used in the EPA's multi-source urban air
quality model PEM-2 (see Rao5), which is designed to predict short term (1 to 24
hr) ground-level concentrations and deposition fluxes of two gaseous or particulate
pollutants, with or without the chemical coupling, at multiple receptors.
885
-------
Acknowledgements
This work was accomplished under interagency agreements among the U.S.
Department of Energy, the National Oceanic and Atmospheric Administration, and
U.S. Environmental Protection Agency under IAG-DW13930021-01. The authors
are grateful to Jack Shreffler, Ray Hosker, and Bruce Hicks for their many helpful
comments and encouragement during the course of this work. It has been subjected
to EPA review and approved for publication.
References
1. K. S, Rao, "Analytical solutions of a gradient-transfer model for plume
deposition and sedimentation," EPA-600/3-82-079 (1982). PB 82-215 153, NTIS,
Springfield, VA.
2. K. S. Rao, "Plume concentration algorithms with deposition, sedimentation,
and chemical transformation," EPA-600/3-84-042 (1984). PB 84-138 742, NTIS,
Springfield, VA.
3. A. S. Monin, "On the boundary condition on the earth surface for diffusing
pollution," Adv. Geophys. 6: 435 (1959).
4. K. L. Calder, "Atmospheric diffusion of particulate material considered as a
boundary value problem," J. Meteorol. 18: 413 (1961).
5. K. S. Rao, "User's guide for PEM-2: Pollution Episodic Model (version-2)," EPA-
600/8-86-040 (1986). PB 87-132 098, NTIS, Springfield, VA.
6. F. A. Gifford, S. R. Hanna, "Modeling urban air pollution," Atmos. Environ.
7: 131 (1971).
7. K. S. Rao, M. M. Stevens, "Pollution Episodic Model user's guide," EPA-600/8-
84-008 (1984). PB 84-164 128, NTIS, Springfield, VA.
GRID SQUARE
WITH EMISSIONS
GRID SQUARE
WITH RECEPTOR,
(a)
-Ax-
(b)
Figure 1. Schematic diagram for area-source algorithms showing (a) an emission
grid square and a receptor grid square, and the distances; (b) a cross-section of the
receptor grid square, and the incoming and outgoing normalized fluxes of pollutant.
886
-------
THE INFLUENCE OF MEASUREMENT UNCERTAINTIES Oil THE
PERFORMANCE OF AIR POLLUTION DISPERSION MODELS
Steven R. Hanna, David G. Strimaitis, and Joseph C. Chang
Eigma Research Corporation
234 Littleton Road, Suite 2E
Westford, MA 01886
ABSTRACT
Uncertainties of several classes of air pollution dispersion models are
estimated using comparisons with field data for continuous sources in the
boundary layer. Source scenarios include ground level point sources, tall
power plant stack plumes, dense gas releases, and releases from offshore oil
platforms. Model predictions of maximum concentration (independent of
position) show typical mean biases of ± 10 to 4O% for the best-performing
models. Typical root-mean-square errors are about 60% of the mean value, with
a range from 30% to 100% for the best models.
The model uncertainties are partly due to measurement uncertainties in
input data and in observed concentrations. For example, the wind speed is
seldom known within ± 10%. Monte Carlo methods of estimating the influence
of these measurement uncertainties on the model predictions are reviewed.
INTRODUCTION
The past decade has seen a growth in interest concerning the uncertainty
in air pollution dispersion models [1,2,3,4]. The old rule of thumb was that
dispersion models carried a "factor of two" uncertainty. Recent model
development programs have had the objective of significantly reducing this
uncertainty, but researchers have discovered that there is a large amount of
irreducible uncertainty due to stochastic processes in the atmosphere and due
to the limitations of measurements. The purpose of this paper is to provide
several examples of the magnitude of the uncertainty associated with air
887
-------
pollution dispersion models, using results from a variety of sites, source
emission scenarios, and models. Methods of quantifying measurement
uncertainties and accounting for their effects on model predictions shall be
outlined.
EXAMPLES OF UNCERTAINTIES ASSOCIATED WITH AIR QUALITY MODEL PERFORMANCE
We have recently been involved in a broad range of air quality model
development and evaluation exercises, involving the use of field data from
eight independent full-scale experiments [5,6,7]. The sources in all
experiments are continuous point sources within the lowest 100 m of the
boundary layer. An overview of the results of each of these studies is given
below.
Prairie Grass Experiment: The 1956 Prairie Grass experiment resulted in a
comprehensive dataset containing 44 separate runs, where SO was released
from a near-surface continuous point source over flat terrain, and detailed
supporting meteorological data were taken. This dataset has been analyzed by
dozens of researchers and used as the basis for the development and
evaluation of numerous dispersion models over the past 30 years. It
represents the optimum research-grade field experiment for which data
uncertainties are minimized. We used multiple linear regression procedures
to fit a simple model to these data, and the resulting model predictions are
plotted versus observations in Figure 1 (where concentration, C, is normalized
by source emission rate, Q). Maximum concentrations on five downwind
monitoring arcs (50 m to 800 m) are included. The model explains about 93% of
the variance in the observations, and the rmse or scatter of the predicted C/Q
about the best-fit line at any observed C/Q averages about 20 to 30%.
Dense Gas Jet Experiments: Pressurized ammonia and hydrogen fluoride were
released continuously from a pipe near the ground, resulting in a dense
aerosol jet. Seven separate runs were made (4 for NH_ and 3 for HF) in which
concentrations were observed at downwind distances ranging from 100 m to
3000 m. The performance of 14 hazardous gas models was evaluated with this
dataset [5], and the_results are summarized in Figure 2. The relative mean
bias, (C - C )/0.5(C + C ), is plotted on the abscissa, and the relative
o p __o p^
2~ — —
mean square error, (C - C ) /C C , is plotted on the ordinate. It is seen
M o p o p K
that some models perform very poorly, with a relative bias of ± 1.0 (i.e.,
100%) and a relative mse of 2 or 3 {i.e., 200 or 300%). However, there is a
cluster of 8 models that perform relatively well with relative biases of about
0.0 to 0.4 (i.e., 0 to 40%), and relative mse of about 0.25 Ap, 0.50 (i.e.,
25 to 50%). The relative rmse, which equals (relative mse) , is therefore
about 50 to 70% for these eight models.
Overwater Tracer Experiments: The Offshore and Coastal Dispersion (OCD)
model was developed to estimate the on-shore impact of pollutants released
from offshore oil platforms. The OCD model was evaluated using 101 individual
runs from field experiments at four separate locations [6]. Tracer gas was
released at elevations of 10 to 20 m above the water, at distances 1 to 15 km
from the shoreline, and concentrations were observed by lines of monitors at
the shoreline. Table 1 contains the results of the model evaluation exercise
at the four sites, where two or three independent_se_ts of experiments were
conducted at each site. The ratio of the means, C /C , ranges from 0.65
(i.e., a 35% underprediction) to 2.13 (i.e., a 113% overprediction) over all
the experiments, with a median of 1.07. The relative rmse ranges from 0.42 to
888
-------
Table 1. Comparisons of OCD Model Predicted Concentrations with Observed
Concentrations during Overwater Tracer Experiments at Four Locations.
Number of DataRatio of Relative RMSE
r
Experiment
Cameron - Winter
Cameron - Summer
Carpinteria - SF
Carpinteria - Fumigation
Carpinteria - CF Br
Pismo - Winter
Pismo - Summer
Ventura - Winter
Ventura - Fall
U J.11 L3
n
17
9
18
9
10
15
16
8
9
c A:
P o
O.B5
O.BO
2. :.3
0,65
1.07
1.49
1.94
1.23
0.(>8
((C - C )2/C C )1/2
p o op
0.60
0.72
1. 17
0.98
0.65
0.94
0.90
0.42
0.60
1.17 (i.e., 42 to 117'/.) from site to site. It is interesting how the model
will underpredict at one site and overpredict at another, suggesting that mode]
biases obtained at a single site should not he extrapolated to other sites, anc
that one should not jump to conclusions if data from only one or two sites are
analyzed.
Experiments with Stack Plumes in Urban Areas: For the final example in this
section, we present results for the urban Hybrid Plume Dispersion Model
(urban-HPDM), which was developed and evaluated using about 80 hours of
tracer data from the buoyant plume from an SCI m power plant stack in
Indianapolis. Maximum predicted and observed concentrations on downwind arcs
ranging from 0.25 to 12 km were considered for a wide range of stability
conditions [7]. It was found that the ratio of the means, C /C , equaled
1.08 (i.e., a 7% overprediction), and the relative rmse was calculated to be
C.35 (i.e., 35%). There is little variation of the ratio C /C with wind speed
stability, mixing height, and hour, implying that the modelpgenerally satisfies
the requirement that model errors should, be randomly scattered and should not
be functions of any input variables. We emphasize that the same data were used
to both develop and evaluate the model. In the next few months, the model will
be evaluated with about 80 hours of "independent" data from the same
experiment, which have been reserved especially for the final evaluation
exercise.
UNCERTAINTIES IN METEOROLOGICAL MEASUREMENTS
Part of the uncertainty in air pollution dispersion model predictions is due to
uncertainties in input parameters, such as wind speed and stability [8]. We
have surveyed a large number of reports and journal articles in order to
estimate the meteorological data uncertainties listed in Table 2. It is
assumed that typical averaging times of 10 tc 60 minutes are used.
Furthermore, it is recognized that there are two classes of instruments in
operation: research-grade and routine, where recommended QA/QC procedures are
followed in all cases. In general, remote sounders are found to provide
adequate estimates of mean wind speed and direction, but do not yield
satisfactory observations of the turbulence components, a- and cr [9]. Routine
wind sensors are subject to large errors when the wind speed drops below the
threshold (0.5 to 1.0 m/s), and mixing depth observations become inadequate at
values less than about 200 m.
889
-------
Table 2. Typical Uncertainty in Meteorological Measurements (10-60 min avg. )
Parameter
Wind Speed
Wind Direction
Lateral Turbulence
(0~ )
Vertical Turbulence
(cr )
AT
Mixing Depth
Instrument
Sonic
Cups, Props,
Remote Sounders
Research-Grade
Routine
Remote Sounders
Research Grade
Remote Sounder
Research Grade
Remote Sounder
Research Grade
Routine
Radiosonde
Remote Sounder
Uncertainty
0.1 m/s (threshold < 0.1 m/s)
0.3 m/s (threshold 0.5 - 1.0 m/s)
1.0 m/s
O
2
o
o
10
0. 1 m/s
0. 5 m/s
0. 1 m/s
0.5 m/s
O.IK
0.4 K
100 m
100 m (minimum 50 m)
ESTIMATING THE EFFECTS OF MEASUREMENT UNCERTAINTIES
There are two methods for estimating the effects of uncertainties in
meteorological observations on dispersion model predictions: (1) an analytical
method based on differentiation of the model equations and (2) a Monte Carlo
method where the model is run many times for randomized input data [10]. The
first method is practical only if the data uncertainties are small (< 10%) and
if the model equations are relatively simple and can be easily differentiated.
Because it is possible to run most models on a computer in minimal time, the
Monte Carlo method (2) is the choice in most uncertainty analyses. In order to
apply this method, it is necessary to estimate the means and variances for each
input meteorological parameter. To be strictly correct, known correlations
among meteorological variables (e.g., very strong stable conditions cannot
occur with high winds) should be accounted for when the random variables are
selected. However, in most applications of the Monte Carlo method, these
correlations are ignored.
A simpler alternative is to run the models for only two extreme values
of each input parameter, rather than hundreds, in order to bracket the
solution, or determine the sensitivity of the solution to variations in the
parameter. To demonstrate this procedure we have applied the SLAB dense gas
model to an area source emission of chlorine gas, where Q = 5 kg/s, source
radius = 5 m, source duration = 5 min, and concentration averaging time =
5 min. The base calculation is made for a wind speed u of 5 m/s, a surface
roughness z of 0.10 m, and a Monin-Obukhov length L of oo m (i.e., neutral
stabilities". The model was then run for u = 4 m/s and 6 m/s, z =0.05 and
0.15 m, and L = -100 m and 100 m, where only one input parameter at a time is
varied. These variations in L represent uncertainties in stability of about
one Pasquill-Gifford class. Resulting centerline concentration predictions at
downwind distances of 100, 800, and 2800 m are listed in Table 3. It is seen a
+ 20% variation in wind speed causes a + 20% variation in predicted
concentration. The variation in predicted concentrations is about ± 30 to 50%
for the stated variation in stability class, and is about ± 15 to 25% for the
stated variation in roughness length. If these input parameters are not known
with any better accuracy, then it can be concluded that errors in model
predictions of at least 50% are possible.
890
-------
Acknowledgements: Portions of this research were sponsored by the U.S. Air
Force, the Minerals Management Service, the U.S. Army, the American Petroleum
institute, and the Electric Power Research Institute.
Table
u
Cm/s)
4
5
5
5
5
5
6
3. Sensitivity Runs of SLAB Model for Chlorine Pool.
(S)
0. 10
0. 10
0.05
0. 10
0. 15
0. 10
0. 10
L
(m)
00
100
00
00
CO
-100
CO
Comment
Low Wind
Stable
Less Rough
Base Case
More Rough
Unstable
High Wind
Concentration (ppm) at Downwind Distances
100 m 800 m 2800 m
2010
2180
2130
1660
1420
1190
1390
46.6
53.6
46.3
37. 1
31, 9
24.0
30. 7
3.84
4.66
3.84
3. 14
2.75
2.00
2.69
References:
1. D.
Fox,
"Uncertainty in air
quality modeling,"
Bull. Am.
Meteorol. Soc.
65: 27-36 (1984).
2. D. J. Carson, "A report on the symposium in uncertainty in modelling
atmospheric dispersion," Atmos. Environ. 20: 1047-1049 (1986).
3. M.M. Benarie, "The limits of air pollution modeling," Atmos. Environ.
2J.: 1-5 (1987).
4:. A. Venkatram, "Inherent uncertainty in air quality modeling,"
Atmos. Environ. 22: 1221-1227 (1988).
5. S.R. Hanna, J.C. Chang, "Revision of the Hybrid Plume Dispersion Model
(HPDM) for application to urban areas," Proceedings, 18th ITM on Air
Pollution Modeling and Its Application, Vancouver, B.C., NATO/CCMS,
(1990).
6. S.R. Hanna, D.G. Strimaitis, J.C. Chang, "Evaluation of 14 hazardous gas
models with ammonia and hydrogen fluoride' field data," Submitted to
J. Hazardous Materials. (1990).
7. D.C. DiCristofaro, S.R. Hanna, M.T. Baer, "The Offshore and Coastal
Dispersion (OCD) model," Proceedings, 19th ITM on Air Pollution Modeling
and Its Application, Vancouver, B.C., NATO/CCMS, (1990).
8. W.S. Lewellen, R.I. Sykes, "Meteorological data needs for modeling air
quality uncertainties," J. Atmos. and Ocean Tech. 6: 759-768, (1989).
9. P. Chintawongvanich, R. Olsen, C.A. Biltcft, "Intercomparison of wind
measurements from the acoustic Doppler scdars, a laser Doppler lidar, and
in situ sensors," J. Atmos. and Ocean Tech. 6: 785-797, (1989).
10. D.L. Freeman, T.R. Egami, H.F. Robinson, J.G. Watson, "A method for
propagating measurement uncertainty through dispersion models,"
J. Air Poll. Control Assoc, 36: 246-253, (1986).
891
-------
10
-1
10
-2
10
-3
10
-4
e
I
10
-5
-6
10
10
10-5 10-
-------
FLOW AND DISPERSION OF POLLUTANTS WITHIN
TWO-DIMENSIONAL VALLEYS
William H. Snyder*
Atmospheric Sciences Modeling Division
Atmospheric Research and Exposure Assessment laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Leon H. Khurshudyan
Main Geophysical Observatory
Leningrad, U.S.S.R.
Igor V. Nekrasov
Institute of Mechanics
State University of Moscow
Moscow, U.S.S.R.
Robert E. Lawson, Jr.* and
Roger S. Thompson
Atmospheric Sciences Modeling Division
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Wind-tunnel experiments and a theoretical model concerning the flow structure and
pollutant diffusion over two-dimensional valleys of varying aspect ratio are described
and compared. Three model valleys were used, having small, medium, and steep slopes.
Measurements of mean and turbulent velocity fields were made upstream, within and
downwind of each of these valleys. Concentration distributions were measured downwind
of tracer sources placed at an array of locations within each of the valleys. The data
are displayed as maps of terrain amplification factors, defined as the ratios of maximum
ground-level concentrations in the presence of the valleys to the maxima observed from
sources of the same height located in flat terrain. Maps are also provided showing the
distance to locations of the maximum ground-level concentrations. The concentration
patterns are interpreted in terms of the detailed flow structure measured in the
valleys. These data were also compared with results of a mathematical model for
treating flow and dispersion over two-dimensional complex terrain. This model used the
wind-tunnel measurements to generate mean flow fields and eddy diffusivities, and these
were applied in the numerical solution of tne diffusion equation. Measured
concentration fields were predicted reasonably well by this model for the valley of
small slope and somewhat less well for the valley of medium slope. Because flow
separation was observed within the steepest valley, the model was not applied in this
case.
*
On assignment from the National Oceanic and Atmospheric Administration, U.S.
Department of Commerce.
Disclaimer: The information in this document has been funded by the United States
Environmental Protection Agency. It has been subjected to Agency review and approved
lor publication.
893
-------
Introduction
This report presents results of the Joint Soviet-American Work Program for studying
air flows and dispersion of pollutants within valleys . It is a natural complement to
earlier work wherein similar measurements were made over two-dimensional hills. An
extensive data set was collected in a wind tunnel on the flow structure and
concentration fields resulting from sources placed within valleys with different
width-to-depth ratios. In addition to furthering basic understanding, one of the main
purposes of the experiments was to test the performance of a diffusion model for
calculating ground-level concentrations (glcs) resulting from point sources placed
within the valleys.
The primary results are presented in terms of terrain amplification factors (TAFs),
defined as the ratios of maximum ground-level concentrations in the presence of the
valleys to the maxima from sources of the same height in flat terrain. This definition
does not involve the locations of the maximum glcs, which will generally occur at very
different downwind distances in the two situations. We present the primary results as
contour plots of constant TAP. This allows the further introduction of "windows" of
excess concentration. If a source is far enough upstream of the valley and the
pollutant is released at low level, the maximum gic will occur upstream of the valley,
so that the effect of the valley will be negligible. If the source is tall enough, the
maximum glc will occur downstream of the valley, so that, again, the effect of the
valley will be negligible. If pollutants are released within the valley, the TAP will
generally exceed unity, and its value will depend upon the source location within the
valley. Hence, a "window" will exist such that pollutants emitted within that window
will result in "excessive" glcs.
3
Lawson ef al. have presented "windows" of excess concentration (for typical shapes
of two- and three-dimensional hills) that extended as far as 10 to 15 hill heights
upstream and downstream of the hills. The purpose of the current study was to determine
to what extent valleys might influence maximum glcs, that is, to establish windows of
excess concentration for typical valley shapes. The shapes chosen were two-dimensional,
with three different aspect ratios, n=a/h = 3, 5, and 8, where a is the valley half-width,
and h is the valley depth. These shapes were chosen to represent a fairly typical range
of realistic valley shapes. The maximum slopes were 10° (valley 8), 16° (valley 5), and
26° (valley 3). The flow structure in these valleys differed dramatically from one
another. In valley 8, the flow did not separate and, to some extent, resembled
potential flow. In valley 3, the flow clearly separated on the upstream slope, and a
mean recirculation region was formed inside the valley. In valley 5, the separation
might be described as incipient; the mean flow was downstream everywhere, but
instantaneous flow reversals were commonly observed. Thus, pollutants emitted within
this flow were frequently wafted back and forth before being transported downstream.
4
Concurrently with the measurement program, a theoretical model was used to
calculate terrain amplification factors for sources located within the valleys. This
model used wind-tunnel data on the flow structure as input for numerical solution of the
turbulent diffusion equation. The model provided quite reasonable predictions of TAFs
for the valley of intermediate slope and somewhat better predictions for the
gently-sloped valley. Because the model cannot handle separated flows, it was not
applied to the steep-sloped valley.
Apparatus, Instrumentation, and Measurement Techniques
The model valleys were placed within the EPA Meteorological Wind Tunnel, which has
a test section 3.7 m wide, 2.1 m high and 18.3 m long. The approach flow was a
simulated atmospheric boundary layer. Extensive measurements of both the flat-terrain
boundary layer and the flow structure within the valleys were made using hot-wire and
pulsed-wire anemometry. Ethane gas, used as a tracer, was released from numerous
894
-------
positions within each valley through a perforated sphere to simulate a neutrally buoyant
point source. Concentration measurements were made downstream using fiame-ionization
detectors. More extensive descriptions of the experimental apparatus and measurement
techniques may be found in Snyder et at5 or Khurshudyan ef a/.1.
Presentation and Discussion of Experimental Results
The pulsed-wire anemometer proved to be quite useful within the very highly
turbulent, separated flows within the valleys, bemuse it can sense flow reversals.
Probability density distributions of longitudinal velocity fluctuations were constructed
from the pulsed-wire measurements. They showed that at the lowest levels within valley
5, mean velocities were quite small, but instantaneous flow reversals were very common
(up to 40% of the time). In valley 3, mean velocities were negative below h/4, very
close to zero at h/2, and some reversals occurred sven at the valley top h. in spite of
the flow reversals and very large turbulence intensities (up to 170%), for many
practical purposes, the distributions were closely Gaussian in character
Mean streamlines were calculated from the mean velocity measurements within the
valleys (Figure 1). At first glance, the streamline pattern over valley 8 is
reminiscent of potential flow; but closer examinalion reveals it is asymmetrical, with
the lower streamlines being slightly closer to the surface on the downwind slope than on
the upwind one. The streamline pattern over valley 5 is clearly asymmetrical, and
because the streamlines diverge strongly away from the surface, it is clear that the
velocity is reduced markedly at the valley center; indeed, it appears that a stagnation
region exists in the valley bottom. In valley 3, the streamline pattern clearly shows a
recirculation region, with separation occurring a short distance down the upwind slope
and reattachment occurring about halfway up the dov/nwind slope. The three valley shapes
thus result in three fundamentally different flow patterns. These basic flow structures
are fairly typical and cover the range of patterns to be observed at full scale, albeit
in neutral stratification.
Figure 2 illustrates some typical comparisons between surface concentration
profiles measured from sources placed above the valley centers and that from a source of
the same height in flat terrain. In all cases, the stack height Hs was equal to
one-half the valley depth h. Xg denotes the distance from the source. The increased
concentrations caused by the valleys are dramatic and the TAFs range from about 2.5 in
valley 8 to about 12 in valley 3. As the concentration increases, the distance to the
maximum decreases. The location of the maximum for valley 3 was actually slightly
upwind of the source. For valley 5, the location of the maximum glc was downstream, but
very close -- about 2 stack heights away -- and the TAP is about 6.
Measurements such as these were made at an array of source locations and heights in
the vicinity of each of the valleys, and the TAF:> were determined for each location.
Maps of these TAFs are shown in Figure 3, where isopleths of constant TAP have been
drawn. The first impression is that the patterns are symmetrical about the vertical
centerline, but closer examination reveals some asymmetry. Nevertheless, the
near-symmetry and the overall similarity in shape amongst the three valleys is quite
surprising in view of the very different flow patterns observed. In contrast, the
magnitudes of the maximum TAFs differ widely, from 2.5 in valley 8 to 17 in valley 3.
These differences, of course, reflect the effects of the different flow structures.
Contours with TAP values of 1.4, 2, 4, etc., have been drawn where appropriate.
Note that these contours form "windows" within which the maximum glc exceeds the glc
that occurs in flat terrain by 40%, 100%, 300%, efc. The longitudinal extent of the
window of 40% excess concentration extends over approximately 60% of the width of valley
8, 80% of the width of valley 5, and more than 90% of the width of valley 3. The
vertical extent of the 40% window is 1.5, 2.0, and 2.5 valley heights above the valley
fop for valleys 8, 5, and 3, respectively.
895
-------
Application of the data in Figure 3 is straightforward. Let us consider a source
which is located in the center of a rather broad valley, say one similar in shape to
valley 8; and the height of the source is half the valley height. Figure 3c suggests
that the maximum glc would be about 2.5 times that expected from a source of the same
height but located in flat terrain. On the other hand, rf the valley were considerably
narrower, say close to valley 5, Figure 3b suggests the maximum glc would be about 7
times as large. Although precise interpolation of these results for valleys
intermediate in shape to those examined here may be difficult, the results allow us to
place some useful limits on the effects of valleys of intermediate shape.
Figure 4 shows the loci of source positions leading to the same locations of
maximum glc. These loci have been identified by marking them with the position of the
maximum glc (in valley heights from the centers of the valleys). Note that the
"undisturbed" or flat-terrain loci (dotted lines) are simply parallel, nearly straight,
diagonal lines. Within the valley, these loci are distorted, as shown by the solid
lines. The diagrams may be used as follows: for any given source position, we may plot
that position on the diagram, then follow the locus to the ground; the intersection of
that locus with the ground is, of course, the location of the maximum glc. Conversely,
from a knowledge of the location of the maximum glc, we may use these diagrams to
determine the line along which the source was positioned. These loci become highly
distorted near the valley centers, and the steeper the valley, the higher the
distortion. As the distance (both longitudinal and vertical) from the valley center
increases, these loci gradually relax to their undisturbed or flat-terrain values.
Numerical Model and Comparisons with Experimental Results
One of the main purposes of the experimental study described above was to test the
applicability of a diffusion model for the evaluation of maximum glcs resulting from
elevated continuous point sources placed in a curvilinear neutral atmospheric boundary
layer. The model used data from the wind-tunnel measurements of wind velocities and
turbulence characteristics as input parameters to calculate two-dimensional,
mass-consistent flow fields. These flow fields were then applied in the numerical
solution of the diffusion equation. Such a model was developed primarily for
evaluation of pollutant dispersion in complex terrain. The present version of the model
does not incorporate a longitudinal diffusion term, and therefore does not calculate the
spread of pollutants in separated flows. Thus, no attempt was made to apply the model
to valley 3, and calculations were made only for valleys 5 and 8.
Contour maps of constant TAF as predicted by the model are shown in Figure 5.
These are to be compared with the measurements shown in Figure 3. The maps for valley 8
show generally similar overall patterns, but they differ in several details. The
vertical extent of the 40%-excess window (the TAF = 1.4 contour) extends to about 1.5 h
based on the measurements, but to only 1.25 h based on the model predictions. The
horizontal extent of the measured window is larger than that of the model-predicted
window. The model-predicted window is shifted slightly upstream and, whereas the
resolution of the grid used for the experimental measurements was rather coarse, a hint
of an upstream shift is also observed there. The model generally predicts larger TAFs
to occur at lower elevations, whereas the measurements show elevated maxima. Both
predicted and observed TAFs were less than unity when the source was at the upstream or
downstream edge of the valley. Maximum TAF values are quite close to one another.
Similar statements may be made when comparing the calculated and observed TAF maps for
valley 5, but the differences are somewhat larger.
Conclusions
The model valleys cover the range of a majority of valleys to be found at full
scale, at least in terms of the basic classes of flow structure that may be observed.
Valley 8 was rather gentle in slope, and the flow over it may be characterized as
relatively smooth and well-behaved. Valley 5, being steeper in slope, caused the flow
to separate intermittently, but not in the mean. In valley 3, the steepest, the flow
896
-------
clearly separated a short distance from the upstream edge, and a recirculating flow was
formed within the valley. Pollutants released at the same relative locations within
each of these valleys behave very differently from one another, and the resulting
surface concentration patterns are dramatically different.
The overall effects of the valleys on surface concentrations are characterized in
terms of terrain amplification factors (TAFs). Maps of these TAFs are provided for each
valley. Also provided are maps detailing the distances to locations where the maximum
ground-level concentrations occur. These maps allow a practitioner to quickly and
easily assess the likely impact of a source located in a valley and to identify the
location where that maximum impact will occur.
A two-dimensional theoretical model that uses a variational analysis technique was
applied to the wind-tunnel measurements to produce mass-consistent mean wind fields.
Measurements of the turbulent fluctuating velocities were also used to calculate
vertical and crosswind eddy diffusivities. The diffusion equation was then solved
numerically to obtain maximum ground-level concentrations from elevated point sources of
various heights near valleys 8 and 5, as well as over flat terrain. Comparison of
calculated and measured TAFs for valley 8 showed satisfactory agreement. Valley 5
exhibited more severe streamline distortion and a stagnation region with large
fluctuating velocities near the bottom of the valley, and therefore the differences
between calculated and measured TAFs were significant in some cases.
References
1. LH. Khurshudyan, W.H. Snyder, I.V. Nekrasov, R.E. Lawson, Jr., R.S, Thompson, F.A.
Schiermeier, "Flow and dispersion of pollutants within two-dimensional valleys:
summary report on joint Soviet-American study", EPA Report, U.S. Envir. Prot. Agcy.,
Res. Tri. Pk., NC, 85p. (1990).
2. L.H. Khurshudyan, W.H. Snyder, I.V. Nekrasov, "Flow and dispersion of pollutants
over two-dimensional hills: summary report on joint Soviet-American study",
EPA-600/4-81-067, U.S. Envir. Prot. Agcy., Res. Tri. Pk., NC, 143p. (1981).
3. R.E. Lawson, Jr., W.H. Snyder, R.S. Thompson, "Estimation of maximum surface
concentrations from sources near complex terrain in neutral flow",
Atmos. Envir, 23: 321-31. (1989).
4. M.E. Berlyand, E.L Genikhovich, L.H. Khurshucyan, "Use of the results of modeling
of an air stream in wind tunnels for the calculation of air pollution", Atmos. Diff.
Air Poll.. Trudy GGQ Hydromet Press, Leningrad, USSR, 352: 3-15. (1975).
5. W.H. Snyder, LH. Khurshudyan, I.V. Nekrasov, R.E. Lawson, Jr., R.S Thompson, "Row
and dispersion of pollutants within two-dimensional valleys", Atmos. Envir. (1990).
-1:25 -1
.75
1.25
•1.25 -1
1 1.25
Figure 1. Streamline patterns derived from
experimental measurements over the valleys.
Note that the vertical scales are exaggerated.
1.25
897
-------
a Valley 3
a Valley 5
c Valley 8
•» Flat terrain
3-2-101234567B
-S -4 -3 -2 -1 0 1 2 3
2 ••
(b) Valley 5
20
-7.S -5 -2.5
0 2.5 5
K/h
Figure 2. Comparison of surface concentration z
profiles from sources placed above valley
centers with one from source of same height
in flat terrain. Hs = h/2.
-12
0 4
x/h
8 12
(a) Valley 3
Figure 4. Distance in valley heights from valley
center to location of maximum ground-level
concentration. Flat terrain values are indicated
as dotted lines.
-1.5 -1.0 -O.5 0.0 0.5 1.0 1.5
1.0 -0,5 0.0 0.5 1.0
1.5 -1.0 -O.5 0.0 0.5 1.0 1.5
-1.5 -1.0 -0.5 0.0 0.5 1.0
1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Figure 3. Contours of constant terrain
amplification factor derived from experimental
measurements.
Figure 5. Contours of constant terrain
amplification factor derived from model
calculations.
898
-------
WIND-TUNNEL MODELING OF THE DISPERSION
OF ODORANTS AND TOXIC FUMES ABOUT HOSPITALS
AND HEALTH CENTERS
Robert N. Meroney and Thomas Z. Tan
Fluid Mechanics and Wind Engineering Progran
Colorado State University
Fort Collins, Colorado
Abstract. This paper presents results from wind-tunnel dispersion
studies performed around a hospital health center. Large research and
teaching hospitals or health centers frequently include animal laboratories,
contagious disease wards, incinerators, diagnostic laboratories, and
radiological treatment facilities. The resulting building complex thus
includes chemical fume hoods, exhaust ducts and short stacks distributed
almost randomly over the building roofs and walls. Such activities are
sources for odorants, air-borne bacteria or viruses, exotic and often toxic
chemicals, and radioactive gases. Wind-tunnel simulations of the resulting
transport of toxic or odorous scalar products are often needed to optimize
the placement of air handling units or mitigate existing re-entrainment
conditions.
Introduction
The concentration field produced by a source located on or near a
hospital complex can be significantly modified from that predicted by
conventional diffusion formulae. Such formulae contain the implicit
assumptions that the flow field has straight parallel streamlines, modest
velocity gradients, and distribution of turbulence energy and length scales
which result from surface features that remain unchanged over long distances.
Near large hospital buildings the flow field becomes highly complex. Curved
streamlines, sharp velocity discontinuities, and non-homogeneous turbulence
disperse effluents in a complicated manner uniquely related to source
configuration and building geometry. Research and teaching hospitals or
health centers in particular often contain laboratories, disease wards,
incinerators, and radiological treatment facilities which can release
particularly noxious odorants and toxic gase:;. These facilities tend to grow
899
-------
and add extensions, wings or entire buildings which complicate the placement
of air handling units intended to mitigate re-entrainment of effluents.
Some general guidance exists about the flow field around building
complexes in the ASHRAE Handbook and Product Directory, and monograph
chapters on turbulent diffusion near buildings.1'2'3 If concentrations
predicted by such methods are marginal or the building configuration is
unique then field or laboratory (wind tunnel) sampling at critical locations
are proposed. This paper presents the results of case studies of ventilation
problems associated with such a health center.
Cases studies to be considered include a) dispersion from fume hoods
and exhausts of the University of Colorado Health Sciences Center (UCHSC),
Denver, and b) transport of traffic exhaust into air handlers for the
proposed Biological Research Center (BRC), University of Colorado, Denver.4'5
Fluid Modeling Criteria
Successful fluid modeling requires simulation of the characteristic
turbulent scales of the atmospheric boundary layer and the replication of
scaled flow around hospital buildings. Similarity criteria and wind-tunnel
metrology are reviewed by Snyder (1981), Plate (1982), and Meroney (I986)-6'7'8
Critical to accurate estimation of isolated plume dispersion will be the
reproduction of approach wind and turbulence profiles as characterized by
friction velocity, u., and roughness length, z0, or velocity profile power-
law exponent, a.
Often atmospheric turbulence may cause only weak effects compared to
the turbulence generated by a hospital complex and local terrain. Yet the
magnitude of the building induced perturbations depends upon the incident
flow turbulence scale and intensity, details of the hospital shape and
surface roughness, and size of the hospital compared to the boundary layer
depth. Geometrical scaling implies that the ratio of the hospital building
height to length scale must be matched and, of course, that all other
building length scales be reduced to this same ratio.
Golden (1961) measured the concentration patterns above the roof of
model cubes in a wind tunnel.9 Frequently, modelers quote Golden's
experiments as justification for presuming dispersion invariance when
obstacle Reynolds numbers exceed 11,000. Halitsky (1968) observed that for
dispersion in the wake region, no change in isoconcentration isopleths from
passive gas releases was found to occur for values of Reynolds number as low
as 3300.ro
In addition to modeling the turbulent structure of the atmosphere in
the vicinity of a test site it is necessary to properly scale the plume
source conditions. When one considers the dynamics of gaseous plume behavior
the following nondimensional parameters of importance are identified:
Momentum flux ratio, psWs//>ref Ure,; Densimetric Froude number, Fr = p,U ref/[(/9s -
pref)L]; and plume Reynolds number, ReD = WsD/i/. Exhaust gases released from
fume hoods and small ventilators are typically at ambient temperatures and
densities; thus, model plumes are also set to ambient density. Plume exit
Reynolds number need only be large enough to ensure turbulent conditions,
i.e., ReD > 2500.
Data Acquisition and Analysis Techniques
The experiments were performed in the Environmental Wind Tunnel (EWT)
of the Fluid Dynamics and Diffusion Laboratory at Colorado State University.
This wind tunnel, especially designed to study atmospheric flow phenomena,
900
-------
incorporates special features such as an adjustable ceiling, a rotating
turntable and a long test section to permit adequate reproduction of
micrometeorological behavior. Mean wind speeds of 0.1 to 15 m/sec in the EWT
can be obtained. Boundary-layer thickness up to 1.2 m can be developed over
the downstream 6 m of the EWT test section by using vortex generators at the
test section entrance and surface roughness; on the floor. The EWT roof is
adjustable in height to reduce blockage and permit the longitudinal pressure
gradient to be set at zero.
Flow Visualization Techniques. A visible plume was produced by
passing the metered simulant gas through a smoke generator (Fog/Smoke Machine
manufactured by Roscolab, Ltd.) and then out of the modeled stack. The
visible plumes for each test were recorded on either VMS or S-VHS video
cassettes with a Panasonic Professional/Industrial camera/recorder system
(AG-450). Flow visualization provides insight into how architectural
features lead to shortened dispersion paths between source and receptor
locations. Often visualization suggests changes in exhaust location, stack
height, or other architectural modifications which mitigate re-inhalation
situations. Observations of visualization tests were examined to note the
presence or absence of phenomena such as building downwash, plume descent,
and cavity vortices.
Concentration Measurements. The experimental measurements of
concentration were performed using a Hewlett Packard gas-chromatograph and
sampling systems designed by Fluid Dynamics and Diffusion Laboratory staff.
The lower limit of measurement is imposed by the instrument sensitivity and
the background concentration of tracer within the air in the wind tunnel.
Background concentrations were measured and subtracted from all data quoted
herein. Concentrations were presented as normalized concentrations, K =
XUH/Q, for all tests.
Test Program and Data
Exhaust fumes from seed storage and drying operations, animal pens, and
chemical fume hoods are released at many points over the roofs of the UCHSC.
Occasionally these fumes enter the air-handling units resulting in odors and
contaminated air throughout the School of Medicine. Another problem is to
locate inlet ventilators for the new Biomedical Research Center (BRC) in
order to avoid vehicle exhaust entrainment.
A physical modeling study of the UCHSC vent buildings was performed to
assist in predicting environmental impacts for several proposed stack-
building configurations. This involved:
1) The 1:150 reduced scale construction of all buildings within 900
feet of the School of Medicine site,
2) The placement of this model into a wind tunnel facility with the
appropriate upwind roughness for this site,
3) Acquisition of velocity and turbulence profiles approaching and
at the modeled UCHSC site,
4) Video taping of six different model plumes for 16 different wind
directions, and
5) Concentration measurements at either 48 (for the School of
Medicine) or 34 (for the EMomedical Research Center) different
sampling locations for two wind speeds and eight wind directions,
Model Construction. Based on atmospheric data over the UCHSC area,
the size of the concentration grid, and modeling constraints discussed above,
a model scale of 1:150 was selected. Since the Environmental Wind Tunnel has
a 3 m turntable this allowed the reduced scale construction of all
901
-------
significant buildings within a 900 foot radius of the UCHSC site. The
terrain upwind of the turntable area was modeled with a generic 2.54 cm
roughness.
Ventilator buildings incorporated large ventilator plenums and accurate
placement of inlet and exhaust openings. The primary ventilator buildings
are the Biomedical Research Center (BRC) and the School of Medicine (SON).
Sampling and source points surveyed during concentration measurements are
indicated for the UCHSC in Figure 1 and for the BRC in Figure 2.
Test Plan School of Medicine. The emissions from the Research Bridge
and Hospital roofs, a highly toxic gas from a stack on the hospital (ETO),
and fumes from a stack on the SOM roof (EF-91) were all simulated.
Additionally the intakes on the Hospital roof, on the roof of the SOM, and on
the SE annex of the SOM were modeled.
Test Plan Biomedical Research Center. The emission from the Research
Bridge roof tops, the emission from the BRC roof top, and the traffic from
the Colorado Boulevard were evaluated. The observations were divided into
building downwash, plume descent and vortices situations.
Discussion of Typical Results
Selection of the final intake and exhaust stack configuration for the
UCHSC and BRC sites will be based upon the consideration of its visual
appearance, zoning regulations, and mitigation of environmental impact. The
environmental effects of exhaust from the ventilator stacks will depend upon
traffic volume, ventilator flow rates, state and federal ambient air-quality
regulations, building and plume aerodynamics, and local meteorology. This
study evaluates through fluid modeling the influence of building and plume
aerodynamics on plume dilution.
Conclusion from smoke visualization tests. Major conclusions drawn
from observations of the visualization tests are as follows:
1. Emissions from the stack on the SOM roof top do not appear to have much
impact on the SOM itself. However, with a easterly wind there is some
downwash into the larger of the two roof airhandler courtyards on the
SOM. Some building downwash is also evident with a NE wind into the
courtyard on the southern side of the SOM.
2. Emissions from the Research Bridge roof top tend to completely engulf
any region downwind. Consequently there could be a considerable
collection of pollutants from this source which may accumulate in
regions where the air stagnates. The Plaza to the SE of the BRC being
one such example.
3. There is some downwash of the exhaust vented from the BRC into the
adjoining courtyard, especially for N,NW and SW wind directions.
However, for the most part the fumes do not appear to have a strong
effect upon the proposed BRC itself.
4. Vehicle emissions from Colorado Boulevard did have a considerable
effect on the BRC. With wind coming from the N, the eddy in the wake
of the BRC tends to draw the pollutants back into the BRC's SE
courtyard. For winds coming from the NE, E, and SE directions, the
auto emissions tend to impinge on the BRC, concentrating along its west
and north sides. This results in high concentration on the intakes
proposed for the west side of the building. But with winds coming from
the SW and W directions, the vortices caused by the obstacle of the BRC
902
-------
building tends to sweep the traffic exhaust westward or away from the
building, that would cause low concentration at the proposed intake
locations.
Conclusions from concentration measurements. By maintaining flow
similarity between model and field conditions, relative concentrations (x/Q)
for a given source configuration, building configuration and wind direction
will be invariant. The wind tunnel relative concentration measurements for
the UCHSC building complex will be the same as those that could be obtained
during full-scale measurements under the same ambient conditions.
Variation of wind orientation produces a wide variance in sample
concentrations. For the SOM Figure 3 shows concentrations measured at all
the sampling locations from the three exhaust sources for the NE wind
direction. Because the ETO stack is close to the intakes on the Hospital
roof top and the exit velocity for the stack is large, the Maximum K
concentration reaches a value as high as 6,000. Concentrations at the
intakes on the Hospital roof top are always higher than the other locations
sampled.
For the BRC Figure 4 indicates concentrations measured at the sampling
points for a east wind direction for each exhaust source. Sampling point #19
detects the highest K concentration value measured during the entire test
(13,000 from the traffic exhaust). This indicates that the traffic exhaust
strongly effects the proposed intake locations. The wind coming from the E
produces 20 times higher K coefficient thar the wind coming from the west.
This phenomenon was also shown during the visualization program.
Based on the concentration data acquired during this study, there were
two recommendations as follows:
1. The intakes on the Hospital roof top should be closed and removed
in order to avoid the highly hazardous ETO stack.
2. The best location for the potential intakes at the Biomedical
Research Center should be on the roof top near the sampling point
# 4 of that building to avoid the traffic exhaust.
References
1. ASHRAE, Handbook and Product Directory. 1989 Fundamentals. Chapter 14, (1989).
2. Hosker, R. P., Jr., "Flow and Diffusion Near Obstacles, "Atmospheric Science and Power Production,
DOE/T1C-27601, pp. 241-326, (1984).
3. Meroney, R. N., "Turbulent Diffusion Near Buildings,11 Engineering Meteorology. Elsevier Publishing
Co., Hew York, 1982, Chapter 11, pp. 481-521.
4. Tan, T.Z. and Keroney, R.M., "Fluid Modeling of Enhuast Gas Dispersion for the University of
Colorado Health Sciences Center," Final Report for UCHSC, Denver, CER88-89TZT-RMM-17, 54
pp., (1989).
5. Tan, T.Z. and Meroney, R.N., "Wind-tunnel Studies to Mitigate Snowdrift into Rooftop Air-handling
Courts on University of Colorado Health Sciences Center," Final Report for UCHSC, Denver,
CER88-89TZT-RNM-15, 53 pp., (1989).
6. Snyder, w. H., "Guidelines for Fluid Modeling of Atmospheric Diffusion," EPA Report EPA-600/8-81-
009, 185 pp., (1981).
7. Plate, E.J., "Wind-tunnel Modeling of Wind Effects in Engineering," Engineering Meteorology.
Elsevier Publishing Company, New Yor«, 1982, Chapter 13, pp. 573-639.
8. Meroney, R.N., "Guideline for Fluid Modeling of Liquefied Natural Gas Cloud Dispersion: Volume
II: Technical Support Document," Gas Research Institute Report GRI 86/0102.2, xxx pp.,
(1986).
9. Golden, J., "Scale Model Techniques," M.S. Thesis, Dept. of Met. and Ocean., New York University,
42 pp., (1961).
10. Halitsky, J., "Gas Diffusion Near Buildings," Meteorology and Atomic Energy, 1968, editor D. H.
Slade, Atomic Energy Commission, Ch. 5-5, pp. 221-256, (1965).
903
-------
Figure 2 Sampling Point Diagram
for the Biomedical Research Center
Figure 1 Sampling Point Diagram
for the School of Medicine
Wind direction: NE
Figure 3 Concentration Level at NE Wind
Direction for Low Speed Conditions
Wind direction: E
• B • 7 • • 1
Figure 4 Concentration Level for East Wind
Direction and Low Speed Conditions
904
-------
ESTIMATING EXPOSURES DOWNWIND
OF ISOLATED BUILDINGS
John S. Irwin, Alan H. Huber and William B. Petersen1
Atmospheric Sciences Modeling Division
Atmospheric Research And Exposure Assessment Laboratory
Research Triangle Park, NC 27711
Introduction
Using results from video-image smoke dispersion experiments1'2
recently conducted In the Environmental Protection Agency (EPA) wind tun-
nel, an evaluation was conducted of the application of a finite line source
model for simulating the dispersion of a continuous nonreactive,
nondepositing point source release, centered at ground-level on the leeward
side of a building. The model estimates the average concentrations at
surface receptors within the cavity and wake region of the building. With
the vertical dispersion effects simulated using an empirical estimate of
the nondimensional crosswind integrated concentration as a function of
scaled distance downwind from the building, the model construct allows
incorporation of results from various experiments. The dispersion induced
by the building is seen to result from a superposition of two dispersive
processes, that induced by the building, and that which would have occurred
if the building were not present. For buildings not oriented perpendicular
to the predominate wind flow, an additional dependence of building orienta-
tion would occur. Although some work has been accomplished on wind orien-
tation effects3'4, it was decided to limit this discussion to the simplest
case of flow perpendicular to the building, and await the processing of
data regarding wind orientation effects just recently collected in the EPA
wind tunnel during the summer of 1989.
Modeling Equations
The concentration x downwind from an ideal continuous point source
release of Q mass per unit time is often approximated as having a Gaussian
lateral dispersion as,
v--*^f (M
c p I L )
j2nU
where
(2)
1 All authors are on assignment from National Oceanic Atmospheric Adminis-
tration, U.S. Department of Commerce.
905
-------
where U is the dilution wind speed, xy is the crosswind integrated concen-
tration, and the term Ep involving the lateral dispersion parameter ay is
the lateral dispersion model. There are various models available for Xy>
many of which do not rely on Gaussian dispersion in the vertical5.
Integrating Equation 1 over a finite line source of length L oriented
perpendicular to the wind flow, provides the concentration downwind from a
continuous line source emission of q = Q/L mass per unit length per unit
time,
(3)
In Equation 3, Et is the finite line source lateral dispersion model.
As L approaches zero, Equation 4 approaches Equation 2. To avoid inherent
limitations in numerical approximations to the erf function, Equation 2 is
evaluated wherever the solution to (4) is less than 10~6, and the larger of
the two solutions is used.
Lateral Dispersion
For the lateral dispersion parameter, we have chosen a characteriza-
tion by Briggs6 (for rural conditions), that has performed well for point
source tracer experiments within the EPA wind tunnel,
cjy = oax/(l +0.0001x)"2 (5)
where x is downwind distance in meters and oa is 0.08 radians for neutral
stability conditions. (The wind tunnel results for building flow simu-
lations are typically scaled 1:250, corresponding to a roughness length of
about 4 cm.) Tests were conducted at various downwind distances comparing
the lateral dispersion model for a point source release (Equation 2) with
the lateral dispersion model for a finite line source release (Equation 4).
It was found that Equations 2 and 4 yield numerically identical results at
downwind distances of order 100L. But for all practical purposes, Equa-
tions 2 and 4 are quite similar for distances of order 10L and beyond.
Finite Line Source Length
Previous characterizations7 of wind tunnel experiments of dispersion
from surface releases centered on the leeward side of a block shaped build-
ings have noted that beyond lOff, where H is building height, the dispersion
is that of a point source, having some initial vertical and lateral
dispersion. These results when coupled with the similarity of Equations 2
and 4 beyond 10L, and the patterns of vertically integrated smoke behind
various block shaped buildings, suggest that L is of order W, at least for
squat buildings with W/H less than 4. The choice of setting L equal to W
is consistent with the source configurations used in previous puff simu-
lations1, where the puff release positions on the leeward side of the
building (W/H = 2) were spaced along path 1.6H in length. Logically, there
should be a point when further increase in the width to height ratio would
have diminishing affect on the lateral dispersion pattern on an isolated
point source at the surface centered on the leeward side of a building.
Wind tunnel experiments4 suggest that this occurs in the vicinity of W/H of
4. Therefore, we model the length of the finite line source as,
/ V /or V/H<4\
L \4V for Wtf>4/
906
-------
Crosswind Integrated Concentration
Using results from wind tunnel experiments2''1'7, values for the nondi-
raensional crosswind integrated concentration, Cy were derived for surface
receptors for point source surface releases centered on the leeward side of
buildings, where Cy=x?UH/Q, where U is at building height. These data
are shown in Figure 1 and were summarized as,
Cy = a(x///)" a = 5(W//y07 ln(0.45)-ln(a)
ln(100) ' ( }
The fit of Equation 7 to the data at x/H = 10 are shown in Figure 2.
It is evident in Figure 2 that beyond W/H - 12 Equation 7 tends to
underestimate Cy. The observed tendency fo:: Cy to increase for W/H > 12 is
considered real. As the width to height aspect ratio (W/H) increases, the
turbulent eddies induced at the ends of the building have less impact on
the initial dispersion behind the building. Characterizing such details in
the functional relationship of Cy will have more importance when data
become available to substantiate the characterization shown in Figure 1.
Dashed lines have been used in Figure 1 to identify regions where there are
no data currently available for comparison.
Initial Lateral Dispersion
Previous characterizations of dispersion within the wake of squat
buildings4'7 suggest that the initial lateral dispersion ayo is on the
order of Q.7H. The initial dispersion can be found by solving Equation 5
interatively for the distance xa such that the lateral dispersion equals
Q.7H. The lateral dispersion for any distance downwind is then found using
a virtual source model for the lateral dispersion, as 2, to identify any
901
-------
dependency of ayo on building geometry. For the comparisons presented aya/H
- 0.7 was used for both the rectangular building (W/H - 2) and the cubical
building.
The comparisons presented are from wind tunnel experiments which rep-
resent neutral stability simulations. In the model presented, atmoshperic
stability and surface roughness effects have direct influence in the
specification of aa in Equation 5. We anticipate that within the building
cavity and wake region, stability and surface roughness effects will have
little influence on Cy (Equation 7). Wind orientation to the building,
source location and receptor height will be the most important factors
influencing Cy.
It appears, that even if future results were to identify inadequacies
in the characterization of the crosswind integrated concentrations provided
by Equation 7, we can anticipate that the overall model construct would
remain valid. All that would be needed would be an extension of Equation 7
to accommodate the new results. In a similar manner, the characterization
of Equation 7 could be extended to receptors at other heights, as data
become available.
We speculate that extension of the model for wind orientations other
than perpendicular to the building, might be accomplished by viewing the
line source as three contiguous line sources, each of length L/3. The line
source emission rates could be adjusted as a function of wind orientation.
Previous results4 suggest the largest surface concentrations occur for
orientations 45° to the average wind flow. These results also suggest that
Cy is a function of source location. Further analysis of data are needed
to ascertain the functional dependence of Cy for source location and wind
orientation.
Acknowledgements
The authors acknowledge and thank Robert E. Lawson, Jr. for making
available tracer dispersion data from experiments conducted within the EPA
wind tunnel. This work was sponsored in part by the Air and Interdisci-
plinary Research Programs of the EPA.
Disclaimer
Although the research described in this article has been supported by
the United States Environmental Protection Agency, it has not been formally
released by the U.S. Environmental Protection Agency and should not at this
stage be construed to represent Agency policy. It is currently undergoing
internal review and clearance for technical merit and policy implications.
References
1. W.B. Petersen, A.H. Huber, "Concentration fluctuations of a toxic
material downwind of a building," To be presented at the 1990
EPA/A&WMA International Symposium on Measurements of Toxic and Related
Air Pollutants, (1990)
2. J.T. Lee, D.L. Call, R.E. Lawson Jr., W.E. Clements, D.E. Hoard, "A
video image analysis system for concentration measurements and flow
visualization in building wakes," Presented at the Fourth Interna-
tional Workshop on Wind and Water Tunnel Modeling of Atmospheric Flow
and dispersion, Oct. 3-6, 1988, Karlsruhe, West Germany, 32 pp.
(1988).
3. R.S. Thompson, D.J. Lombardi, "Dispersion of roof-top emissions from
isolated buildings, A wind Tunnel Study," EPA-600/4/77-006, U.S.
Environmental Protection Agency, Research Triangle Park, NC, 44 pp.
(1977).
4. A.H. Huber, "The influence of building width and orientation on plume
dispersion in the wake of a building," Atmos. Environ. 23: 2109-2116
(1989).
908
-------
5.
6.
7.
S.E. Gryning, A.A.M. Holtslag, J.S. Irvin, B. Siversten, "Applied
dispersion modelling based on meteorological scaling parameters,"
Atmos. Environ. 21: 79-89 (1987).
F.A. Gifford, "Turbulent diffusion-typing schemes:
Safety 17: 68-86 (1976).
a. review," Nuclear
A.H. Huber, W.H. Snyder, "Wind tunnel investigation of the effects of
a rectangular-shaped building on dispersion of effluents from short
adjacent stacks," Atmos. Environ. 16: 2837-2848 (1982).
Table I. Comparison of observed and estimated nondimensional concentra-
tions for receptors within ± 0.5H of the centerline for the cubical build-
ing (Figure 5), and within ± H of the centerline for the rectangular
building (Figure 4)
x/H
Cubical
Block
1.5
2.0
5.0
Rectangular
Block
3.0
10.0
Number of
values
18
24
42
23
30
Avg. Concentration
Observed
2.10
1.61
0.69
0.71
0.29
Estimated
1.78
1.46
0.74
0.61
0.28
Est./Obs.
0.85
0.92
1.08
0.86
0.98
NoridimensionaI Crosswind
Integrated Concentrations
Noncfimensiono Crosswind Integrated
Concentrations at X/H = 10
>,
o
1,00 -
0,30 -
o
0,60 -
0.40 -
O.?0 -f-r-i
o.co
Figure 1. Nondimensional crosswind
integrated concentrations for several
building width to height ratios (W/H)
as a function of scaled downwind dis-
tance (x/H). Lines are modeled val-
ues given by Equation 7.
5.00 10.00 15.00 20.00
W/H
Figure 2. Nondimensional crosswind
integrated concentrations for several
building width to height ratios (W/H)
at scaled downwind distance x/H — 10.
Solid line is model values given by
Equation 7.
909
-------
3.00
Rectangular Block (W/H = 2)
Nondimensional Centerline Concentrations
' 0.00
0.00
5.00
1000 1500
X/H
20.00
Figure 3. Nondlraensional centerline
concentrations versus scaled down-
wind distance for a building width
to height ratio W/H = 2. Circles
and squares are from wind tunnel
experiments. Solid line is result
using the finite line source model.
Dashed line are INPUFF results.
0.9
Rectangular Block (W/H = 2)
10 and 20 cm Cube
Figure 4. Nondimensional concentra-
tions for a building with a width to
height ratio, W/H = 2, Wind tunnel
results are shown for x/H = 3
(circles), and for x/H = 10
(triangles). Solid lines are finite
line source results, and dashed lines
are INPUFF results.
Figure 5. Nondimensional concentra-
tions for a cube. Wind tunnel
results are shown for x/H =1.5
(circles), for x/H — 2 (triangles),
and for x/H = 5 (squares). Solid
lines are finite line source results
910
-------
CONCENTRATION FLUCTUATIONS OF A TOXIC
MATERIAL DOWNWIND OF A BUILDING
William B. Petersen and Alan H. Huber1
Atmospheric Research And Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Introduction
Investigations of the distribution of pollution in the near-wake
region of buildings are needed to bett€>r understand the complicated disper-
sion processes caused by the disturbed air notions near buildings. Concen-
trations of pollution as emitted into the anbient air are often potentially
hazardous. The dilutions of these emissions with the surrounding air
naturally reduces the pollution concentrations. The aerodynamic wake flow
and thus pollution concentrations are highly dependent on the building
shape and length scales, and the approach boundary layer flow. The lateral
and vertical mixing initially spreads the pollution over the width and the
height of the building. The pollution concentration is also influenced by
the downstream extent and the residence time of the recirculating flow
adjacent to the building1-2.
Wind-tunnel modeling is a practical method for evaluating building-
wake influences. However, the measurement of fluctuating concentrations in
wind tunnel studies requires the use of analyzers having a faster response
time than is required for full-scale field measurement because of the
reduced scaling for wind-tunnel models. There have been some wind-tunnel
studies of fluctuating concentrations and new instruments are now available
for future studies. Li and Meroney3 evaluated the fluctuation in concen-
tration downwind of a cubical model building. Within one building height
downstream of sources on the building, the ;;tandard deviations of plume
centerline concentrations were found to range from 1 to 3 times the mean
concentration. The standard deviation of plume centerline concentrations
beyond three building heights downstream was found to be approximately 0.35
times the mean concentration.
A wind-tunnel study'1'5 of video images cf smoke dispersion in the wake
of a model building was conducted to provide a measure of the vertically
integrated concentration through the plume. These values are well-suited
for identifying vortex shedding and horizontal plume meander. A very
recent study of video images of smoke flow in the wake of a building has
used laser sheet-lighting to provide measures of point concentrations over
a horizontal plane5. In the following discussion, we used the understand-
ing gained from these wind-tunnel experiments for the development of a puff
model for calculating pollutant concentration fluctuations downwind of
buildings. A puff model formulation can provide estimates of the average
and standard deviation of the concentrations in the building wake. Fluc-
tuations in the spatial-temporal scales of the shed vortices and meander in
1 Authors are on assignment from National Oceanic Atmospheric Administra-
tion, U.S. Department of Commerce.
911
-------
the plume can be related to the scales of the mean velocity and the
building. Here, we compare the cross-stream distribution for a series of
video images with the concentration distributions computed with a Gaussian
puff model. The model simulations are intended to characterize concentra-
tion fluctuations in the wake of a building. Also, averaged concentrations
from the puff model were compared with averaged tracer concentrations,
Description of the Wind Tunnel Experiments
The experimental data used in this paper were collected as part of
studies conducted in the U.S. Environmental Protection Agency's Meteorolog-
ical Wind Tunnel, The model building height (H) was either 0.1 m or 0.2 m.
The along-stream building length was equal to the building height. The
primary case being studied had a building width, W equal to 2H which was
oriented perpendicular to the direction of the free-stream flow. Data were
collected for cases with variation in the building orientation angle and
the building width. The mean velocity profile was typical of a neutrally
stratified boundary-layer flow. The mean velocity upstream of the building
was 2 m/s at an elevation of 0.5H. An oil-fog smoke was generated and
emitted into the wind tunnel through a pipe. The stream of smoke simulated
point-source emissions at floor level, midway along the leeward side of the
building. Black-and-white video images of smoke were recorded at a rate of
30 pictures per second with a camera connected to a standard VHS recorder.
The camera was mounted outside the wind tunnel looking down through a ceil-
ing window. In the first studies, the video luminance analog signal was
digitized by a video sequence processor at the North Carolina State
University's Image Processing Laboratory. The most recent studies used a
collection and digitizing system that was set-up in the EPA Fluid Modeling
Facility7. The value above the background level is related to the inte-
grated smoke particle density through the field of view. Thus, the origi-
nal values were corrected by subtracting the estimated background level and
then normalized to give a relative measure of concentration. For the
experiments under general lighting the normalized scale range should corre-
spond to a unitless scale providing an indirect measure of the smoke con-
centration integrated vertically through the field of view. For the laser
light experiments the values are nearly a point measure (a volume defined
by the thickness of the light sheet). Although the relationship between
video image intensity and concentration has not been firmly established, we
believe the data are useful for the development of a building wake model.
Puff Model Description
The puff model used in this study was INPUFF8. INPUFF is a Gaussian
INtegrated PUFF model that was particularly adaptable to simulating concen-
trations in the building wake where rapid puff release rates were neces-
sary. Several changes were incorporated into INPUFF to simulate
concentrations in the wake of a building: (1) Along-wind dispersion is
incorporated into the model, (2) Source position and rate of puff release
are a function of building geometry/orientation and wind speed, and (3) The
initial sigmas (ox, oy, and o^) are a function of building geometry. Param-
eters to characterize the puffs were chosen based on the video image analy-
ses. Analysis of the video image data suggested that vortex shedding in
the lee of a building can be simulated by puffs that are released at three
positions, in sequence along the leeward edge of the building. The source
location oscillates between the building edges with a superimposed random
component. The first puff is released on the centerline, the second at
-0.8H, the third on the center, and the fourth at 0.8H position. This same
pattern repeats for every following cycle except for differences among the
random components. The random component is included here for added realism
to this model simulation of the measurements. The period of the four puff
cycle is computed as follows;
14H where, P is the puff release period,
P ~ fj H is the building height, and
U is the wind speed.
912
-------
For a wind normal to the building the puff release positions are at
the center of the leeward edge of the building and Q.8H on either side.
As the wind direction rotates to a 45 degree orientation the puff release
positions are moved along the leeward edge towards the downwind edge.
Maximum displacement occurs at a wind direction of 45 degrees. No change
is made for wind directions greater than 45 degrees. The video image data
was used to develop this empirical relationship but, building orientation
effects have not been extensively tested and will be reported in later
work. Puffs are always released at the surface. The source strength (Q=l
g/s) is distributed so that each of the two puffs released along the
building sides contained Q/3 and each of the two centerline puffs contained
Q/6. A random component of 0.2Q was also incorporated for added realism.
Virtual source distances are calculated and incorporated into the
model so that at the leeward edge of the. building the dispersion parameters
(ox, Qy, and O were set equal to;
10
n \ ri j ri ri ri
0 I/ { \J "\ I/ 0 \J (it/ \ \>J
^-0.35--0.062S(--2J /or-<10 ^ - 0.3- - 0.3S[ -- 1 0 J/or 1 0 < - < 30
In the puff simulations of dispersion. P-G stability class "D" was
used to characterize the rate of growth of the puffs. ox and ay were
assumed to have the same growth rate.
Results
We have been investigating the performance of the modifications made
to the puff model. The above section pressnts the parameters as they were
used in the analyses reported here. There is a very delicate 'play'
between the choice for the period, the size of ox relative to az, and. the
degree of randomness that is included in the model. The degree of fluctua-
tion is very sensitive to the selected combination of these parameters.
The selected parameters used here are consistent with the general
observations made of the smoke prior to these model simulations.
Figure 1 compares the cross-stream profile of mean values across the
plume at x/H=3. The video image data and the model estimates have been
normalized to match the tracer value at y/H-0. Once normalized, the spa-
tial and temporal distributions of the video data should be a valid mea-
surement. The model and tracer concentrations are non-dimensionalized by
UHZ/Q; where U is the approach velocity at z=H, H is the building height,
and Q is the emission rate. Since the source is located near the building
where the mean wind is reduced in comparison to the free stream approach
flow, some compensation for this effect must be made in the model concen-
tration estimates. The model input includes a puff advective velocity
which applies fully to the effective plume dilution at some distance
downwind of the building (e.g. x/H=15). Thus, to compare with wind tunnel
tracer concentrations which actually feel this reduced wind, a modified
model velocity is used. The adjusted velocity used in these comparisons
was fixed to equal the reduced velocity in the wind tunnel.
u,olfn f xY1-77
-—^ — = 0.95+6,42 - /or3
-------
Figure 2 compares the model and tracer cross-stream profiles at
x/H-10; there are no video data at this downstream distance. Figure 3
compares the model and tracer concentrations along the centerline. The
comparison at distances beyond x/H-10 could be improved by incorporating
dispersion parameters specific to conditions found in the wind tunnel in
place of the P-G dispersion parameters. These figures show that the mean
profiles for the puff model compare well to measurements.
We are interested in seeing how well the puff model can characterize
fluctuating concentrations in the building wake. The cross-stream distrib-
utions of several statistics for the time series of video image and model
data were compared. The video image values and the model values were
normalized by the mean centerline value. Figure 4 compares the profiles
for the value of the standard deviation divided by the mean centerline
value for the series. The standard deviations for the model lie between
the laser and general light values from the video images. Figure 5 pres-
ents example time series values at y/H=0 at x/H=3. Similar comparisons are
observed at other cross-stream positions. Time has been normalized by the
observed period of oscillation (0.64 s for video, 18 s for INPUFF). The
temporal resolution of the video values is 19 per cycle while for INPUFF it
is 18 per cycle. The model values (5a) compares best to the video values
for the general light (5c) . The spatially integrated concentration for the
general light video values is in effect similar to a temporal average over
a shed vortex in the building wake flow and thus should be comparable with
that modeled by an advected puff. The video values for the laser sheet-
lighting should represent short-term (on the order of 30 second) averaged
field values.
Conclusions and Recommendations
The puff model simulates many features of the larger scale variation
observed in the building wake flow but does not simulate the short term
peaks. This could be overcome by adding an empirically determined factor
for the relative short term peaks. However, this feature would not be
necessary for most applications where average concentrations over several
minutes are needed. These comparisons indicate that a 4 puff per cycle
model can be used to simulate the overall characteristics of building-wake
dispersion. The puff model also has the special feature of being able to
simulate the effects of meander and rapidly changing conditions for emis-
sions and meteorology. The building wake formulations developed in the
puff model are based on analysis of the video image data from wind tunnel
studies and are considered preliminary.
References
1. Fackrell J.E. (1984) Parameters characterizing dispersion in the near
wake of buildings, J. Wind Engr. Ind. Aerodyn., 16, 97-118
2. Vincent, J.H. (1978) Scalar transport in the near aerodynamic wakes
of surface-mounted cubes, Atmos. Environ., 12, 1319-1322.
3. Li W. And R.N. Meroney (1983) Gas dispersion near a cubical model
building, Part II, Concentration fluctuation measurements, J. Wind
Engr. Ind. Aerodyn., 12, 35-47.
4. Huber, A.H. (1988) Video images of smoke dispersion in the near wake
of a model building. Part I. Temporal and spatial scales of vortex
shedding, J. Wind Engr. Ind. Aerodyn., 31, 189-223.
5. Huber, A.H. And S.P.S. Arya (1989) Video images of smoke dispersion
in the near wake of a model building. Part II. Cross-stream distribu-
tion, J. Wind Engr. Ind. Aerodyn., 32, 263-284.
6. Huber, A. H., D. S. Trotter, R. E. Lawson Jr., S. A. Rajala, and R. E.
Van Dyck (1990) Evaluation of Laser Light Video Flow Data. (In prep-
aration) .
914
-------
Lee, J. T. D. L. Call, R, E. Lawson Jr., W. E. Clements, and D. E.
Hoard (1989) A Video Image Analysis System for Concentration Measure-
ments and Flow Visualization in Building Wakes. Los Alamos National
Laboratory Report No. LA-UR-88-4014.
Petersen, W.B. And L.G. Lavdas (1986) INPUFF 2.0 A Multiple Source
Gaussian Puff Dispersion Algorithm. EPA/600/8-86/024, U.S. Environ-
mental Protection Agency, Research Triangle Park, NC. 105pp.
Rectongular Block (W/H = 2)
Nondirriensioral Concentrations (X/H = 3)
OOOOO Tracer data
Pijff model
AAAAA Loser dota
-101 23
Y/H (NONDIMENSIOMAL)
>/H (NONDIMENSIONAL!
Figure 1. Nondimensional lateral
concentration at 3 building heights
downwind.
Figure 2, Nondimensional lateral
concentration at 10 building heights
downwind.
Rectangulor Block (W/H = 2)
Nondirl^ensional Centerhne Concentrations
g 0.8
ryi
0.6 -
: 0.4 -
O
u
5 10 15
X/H (NONDIMENSIONAL)
20
Rectonguksr Block (W/H = 2'?
Nondimensioral Concentrations (X/H = 3)
Figure 3. Nondimensional center-
line concentration for tracer data
(circles) and model results (dashed
line).
Figure 4. Standard deviation of
noridimensional lateral concentration
at 3 building heights downwind.
915
-------
5a
4
3-
o
O
Puff Mode!
iy\
~i | r
5
i 1 1 1 1 1 1 i r
1O 15
0
TIME (NON-D)
4
5b
8^'
0
0
Laser light
~i 1 1 1 ! 1 F
-------
HAZARDOUS WASTE INCINERATOR
EVALUATION PROGRAM
BY
SHEKAR VISWANATHAN
ABB ENVIRONMENTAL SERVICES, INC.
(FORMERLY C-E ENVIRONMENTAL, INC.)
39255 Country Club Drive, B-25
Farmington Hills, Michigan 48331
INTRODUCTION
The disposal of hazardous wastes in a safe manner has been a concern since the 1970's.
Indiscriminate dumping of chemical wastes of all typos had been going on since the birth of
chemical industries in the U.S. Methods of disposal, such as incineration, that would have
eliminated wastes permanently, were initially disregarded as being too costly. However, many
of the major companies installed incinerator facilities on their own plant sites; not because it
was the most economical method but because they realized that other methods of disposal
would not be open to them. It is estimated that a tola! of 100 million tons of organic chemicals
wastes are generated in the U.S. every year and only 5% of these wastes are treated through
incineration. Under the amendments to the Resource Conservation and Recovery Act (RCRA)
that were passed in 1985, the Environmental Protection Agency (EPA) is required to ban the
land disposal of many hazardous wastes unless it can be proved that such wastes can be
safely disposed of to the land. As a result, incine-ation as a means of hazardous waste
disposal, especially for organic materials, is becoming popular.
The ultimate goal of hazardous waste incineration is to destroy the waste material with as high
a destruction efficiency as possible. Under the RCRA, every incinerator should show that it can
adequately destroy those hazardous waste constituents which are most difficult to incinerate.
In general, the governing regulatory agency will select compounds within the inlet mixture which
are of sufficient toxicity, concentration, and thermal stability so as to be designated as principal
organic hazardous constituents (POHCS). During the incineration of hazardous wastes,
compounds not identified in the waste feed may be formed. These compounds are known as
products of incomplete combustion (PICS). In order to obtain an operating permit for a
hazardous waste incinerator, it must be shown by trial burn that the incinerator meets or
exceeds the following performance standards:
99.99% Destruction Efficiency
° 0.08 gr/dscf of paniculate emissions corrected to 7% oxygen
917
-------
4 Ib/hr of hydrogen chloride emissions or a 99% removal efficiency
This paper describes the experimental program undertaken to evaluate the performance of two
large hazardous waste in-plant incinerators used in converting liquid and gaseous chlorinated
hydrocarbons into hydrochloric acid product. The sample collection methods, analytical
techniques employed, quality assurance and quality control program instituted along with results
are described in this paper, In particular, this paper discusses the details of dioxin/furan testing
results.
PROCESS DESCRIPTION
The waste gases and liquids from the process area of a chemical plant are fed to two oxidation
units that converts chlorinated hydrocarbons into hydrochloric acid, by passing through an
absorption system. Then the exhaust gases are vented to the atmosphere. The major
components that are fed to the oxidation units are provided in Table 1. It was expected that
these incinerators do not produce any hazardous toxic compounds.
SAMPLING METHODOLOGIES
The sampling program, recovery, and analytical methodologies were designed during this
program to ensure:
Collection of a representative sample;
Minimization of both sample loss in the train and contamination by sampling train
components;
« Quantitative recovery of the sample from all parts of the train; and
Appropriate sample pretreatment to achieve quantitative extraction, concentration, and
negligible interferences during analysis.
Before designing the protocol, it is important to identify the key organic components that are
exhausted from the process, by considering key POHCs. Table 2 provides a typical list of
POHCs and corresponding PICs.
The designated POHCs during this program were chosen by considering:
Concentration of organics in the feed
The rank of incinerability.
In addition, a whole range of PICs were also considered.
The following documents were used to develop the sampling and analytical methodologies for
this program;
ASME/EPA Draft Protocols for the Determination of Chlorinated Organic Compounds in
Stack Emissions [1, 2, 3]
Battelle Laboratory Procedure for Sampling and Analysis of Polynuclear Aromatic
Hydrocarbons [4]
918
-------
Protocol for the Collection and Analysis of Volatile POHC's {Principle Organic Hazardous
Constituents) using VOST (Volatile Organic Sampling Train) [5]
Texas State Department of Health - Compliance Manual for the Measurement of Chlorine
and Hydrogen Chloride [6]
TABLE 1: MAJOR COMPOUNDS THAT ARE FED TO INCINERATORS
Chloroethene/Vinyi Chloride
1,2- dichloroethylene
Dichloroisopropyl Ether
Dichloropropanol
Epichlorohydrin
Ethylene Dichloride
Hexachlorobenzene
Hexachlorobutadiene
Hexachloroethane
Hexachlorobutylene
Hexachloropropane
Octachlorostyrene
Pentachlorobutylenes
Propylene Chlorohydrin Bhane
Propylene Dichloride
Propylene Oxide/Propion aldehyde Acetone
1, 1, 2, 2 - tetrachloroethane
Tetrachloromethane
1, 2, 3 - trichlorobutylene
1, 1, 2 - trichloroethane
TABLE 2: TYPICAL POHCs AND CORRESPONDING PICs (8)
Designated POHC Products of Incomplete Combustion
Carbon Tetrachloride Tetrachloroethylene
Hexachloroethane
Chloroform Tetrachloroethylene
Carbon Tetrachloride
Chlorobenzene Benzene
Toluene Benzene
Trichlorobenzene Cnlorobenzene
Dichlorobenzenes
Ketone Hexachlorobenzene
Pentachloroethane Totrachloroethylene
PCBs Chlorinated dibenzofurans
919
-------
EPA Stationary Source Sampling Methods for the Determination of Total Hydrocarbons
and Vinyl Chloride [7]
Table 3 provides a list of sampling methods used during this program. All procedures
employed here were in accordance with USEPA recommended methods.
TABLE 3: SAMPLING METHODS EMPLOYED DURING THIS PROGRAM
General Compounds
Type of Interest Sampling Method
Most volatile B.P. < 30°C Vinyl Chloride EPA Method 18
Volatile B.P. (30° - 100°C) Methylene chloride VOST
Semi-volatile Chiorobenzenes Modified Method 5
B.P. (100° - 200°)
Non-volatile B.P. > 200°C Dioxins/Furans Modified Method 5
Acids Hydrochloric acid Texas Method
Gases Chlorine Texas Method
PREPARATION OF SAMPLING TRAIN
All recommended procedures including proving (1) were employed for the preparation of
sampling trains. The quality of the pretest cleanup for organics sampling is crucial, due to the
expected low concentration of heavy molecular weight organics. Hence, cleaning and proving
protocols for trains were strictly followed to ensure that contaminants from sampling equipment
would not interfere with the analysis. The protocols include:
Rinsing all glassware and train components with solvents such as methanol, acetone and
methylene chloride
Final washing with pentane
• Sealing precleaned glasswares with proven aluminum foil
Soxhlet extraction and proving tenax and XAD-2 sorbent cartridges using ASME
protocol (1)
The proving of the trains was performed by concentrating and analyzing the final pentane
extract using the analytical methodologies illustrated in the later section of the paper. The total
allowable background concentration of chlorinated organics was limited to 100ng per train
(includes all train components). If the analysis of proving extract indicated a higher level, all
train components were once again subjected to the same cleaning process, before using for
sampling.
920
-------
ON-SITE SAMPLING AND SAMPLE RECOVERY
All sampling was conducted by personnel experienced with test procedures and cognizant of
the analytical techniques for trace organics, particularly contamination problems. A total of 8
tests were conducted to identify variability on emission rates.
Extreme care was exercised during the sampling of heavy molecular weight organics (semi and
non-volatiles). A new set of train components was used for each test. The various components
of the sampling train were leak checked separately prior to assembly. A total vacuum of 15
inches Hg or higher was drawn and maintained for 2 minutes. If a leakage rate was found to
be 0.02 CFM or greater, the specific assembly was dismantled and reassembled until the leak
was adequately reduced or eliminated completely. This step wise procedure was followed
consistently to identify the location of leaks in separate components. Finally the entire train,
including the probe, was checked accordingly, prior to the start of sampling. Leak checks were
performed before and after each traverse of a test.
Actual sampling time of three hours was necessary in all tests to collect a minimum of 80 to
120 cubic feet of stack gas for each test. Isokinetic sampling was achieved through the use
of appropriate pre-test data to develop the isokinetic sampling rate equation needed to predict
the pressure drop across the sampling train orifice for each velocity pressure measured.
At the end of each test, the pump was turned off and post-test leak checks were completed
before transporting the probe-impinger assembly to the on-site laboratory. A blank train was
set up in the same manner as actual test trains for the entire duration of each test with no
sample collection. This provided quantitative residual contamination that could occur under site
conditions.
The VOST and hydrocarbon sampling were conducted in accordance with the respective
protocols. All required, trip, field and laboratory blanks were used in accordance with the
protocol.
Quantitative recovery of heavy molecular weight samples from all parts of the train required a
rigorous and meticulous procedure. In order to avoid any contamination, the stack sampling
train assembly was transported to the on-site laboratory for partial sample recovery. The
samples were recovered in a fashion to provide catches for the analyses of chlorinated and
polynuclear aromatic hydrocarbons (heavy molecular weight organics). The following samples
were collected from each train:
• nozzle, probe and front half of the filter
filter
condenser and back end of the filter
sorbenttube
impinger 1
impingers 2 and 3
All train components were washed with respective solvents as specified in the protocol followed
by pentane wash. This final wash was separately analyzed in each test to provide a
quantitative recovery efficiency.
ANALYSIS AND QUALITY CONTROL
The samples collected from the field during the dioxin/furan testing program were analyzed as
individual catches for heavy molecular weight organics by employing standard methods. This
method of analysis provided details of the various organics that were collected in various parts
921
-------
of the train. Figure 1 illustrates the schematics of the procedures adopted prior to analysis.
Table 4 lists appropriate surrogates and internal standards used during this program. These
were added in the following manner before the extraction procedure was begun:
20 ml of SNA surrogate
200 ml of PCB surrogate
200 ml of dioxin internal standard/surrogate
The extracts to be analyzed for dioxin/furans were cleaned up using the combination base/acid
silica gel as described in the ASME protocol [2]. These were also cleaned using the basic
alumina column described in the ASME protocol [2]. The eluent from the column was then
concentrated to 50 ul before subjecting it to analysis.
The analysis of these extracts were performed for:
dioxins/furans (including members of dirty dozen family as described in Table 4) and other
tetra to octa congener groups
PCBs (all congener groups)
Polynuclear Aromatic Hydrocarbons
Chlorobenzenes
o Chloroethanes
Octachlorostyrenes
using gas chromatography/mass spectrometry.
The quality control/quality assurance program for the dioxin/furan analysis involved:
analysis of samples spiked with stable isotopically labelled internal standards listed in
Table 5;
processing and analysis of one blank sample with each extraction batch;
matrix spike duplicate (MSD) pair of analysis with each extraction batch;
analysis of glassware rinses collected prior to usage of the trains;
analysis of glassware rinses collected after usage of the trains;
analysis of distilled, deionized water used in the preparation of the trains;
analysis of catches collected from the various components of the field blank trains; and
splitting of selected samples for analysis by Regulatory Agency Laboratory.
922
-------
RESULTS AND DISCUSSION
Tables 5, 6, 7 and 8 provide the testing results of this program. Because of the complex
nature of the program, only the important results aro provided in this paper. The numbers
indicated in parentheses were positively identified, whereas the other numbers represent the
limit of detection of individual compounds based on sample volume, extraction procedure, and
analytical method used. The testing program revealed that:
o only total penta-furan and octa-dioxin congeners were found in the exhaust stream;
no members of the dirty dozen family were positively identified;
hexachlorobutadiene and hexachlorobenzene were identified positively;
many of the volatile organics fed to the incinerators were identified positively;
no vinyl chloride monomers were detected positively;
excellent recoveries of surrogates from matrix spike and matrix spike duplicate samples
were obtained (Table 9);
no significant contamination of the blank train was observed (Table 10);
presence of naphthalene was observed in the sorbent trap in all tests;
compounds with higher chlorination produced lower destruction efficiencies; and
an analysis of variance of spiked compound recoveries showed the mean percent
recoveries were equal at a level of significance less than 0.005.
REFERENCES
1. "ASME/EPA Draft-Protocol Sampling for the Determination of Chlorinated Organic
Compounds in Stack Emissions"; Environmental Protection Agency; October (1986).
2. "Analytical Procedures to Assay Stack Effluent Samples and Residual Combustion
Products for Polychlorinated Dibenzo-p-dioxins (PCDD) And Polychlorinated Dibenzofurans
(PCDF)"; Environmental Protection Agency; September (1984).
3. Tiernan, T.O.B., "Analytical Protocol for Determination of Polychlorinated Biphenyls (PCBs)
in Drinking Water, Environmental Water and Waste Water"; Write State University, Dayton,
Ohio; October (1986).
4. "Battelle Laboratory Procedure for Sampling and Analysis of Poly Nuclear Aromatic
Hydrocarbons"; Battelle Laboratory; Columbus, Ohio; (1985).
5. "Protocol for the Collection and Analysis of Volatile POHCS (Principal Organic Hazardous
Constituents) Using VOST (Volatile Organic Sampling Train)"; EPA-600/8-84-007, U.S.
Environmental Protection Agency; March (1980).
6. "Compliance Manual for the Measurement of Chlorine and Hydrogen Chloride"; Texas
State Department of Health; March (1973).
7. Environmental Protection Agency Stationary Source Sampling Methods for the
Determination of Total Hydrocarbons; EPA Methods August (1985).
8. W.S. Smith and T. Wong, "Developing a Trial Burn Plan, "Incinerating Hazardous Wastes,
Technology Publishing Co. (1988).
923
-------
FIGURE 1: SCHEMATICS OF LABORATORY PROCEDURES ADOPTED
FOR HEAVY MOLECULAR WEIGHT ORGANIC ANALYSIS
1/Bth SAMPLE
RETAINED FOR
FUTURE ANALYSIS
3/8th SAMPLE
USED FOR
DIOX1MTURAN
ANALYSIS
1/4th SAMPLE
BASE NEUTRAL
ACID ANALYSIS
IMth SAMPLE
PCB
ANALYSIS
SPLIT
IMtil SAMPLE
TO
REGULATORY AGENCY
V*th SAMPLE
TO
INTERNAL LABORATORY
TABLE 4: INTERNAL STANDARDS AND SURROGATES
ADDED TO FIELD SAMPLES
SURROGATE COMPOUNDS
COMPOUND
37CI-2,3,7,8-TCDD
D5-nitrobenzene
2-fluorobiphenyl
D14-terphenyl
2-fluorophenol
D5-phenoI
2,4,6-tribromophenol
ANALYSIS
Polychlorinated Biphenyls
Dioxin/Furan
Base Neutral Acid
Base Neutral Acid
Base Neutral Acid
Base Neutral Acid
Base Neutral Acid
Base Neutral Acid
INTERNAL STANDARD COMPOUNDS
COMPOUND
13C-2,3,7,8-TCDD
13C-1,2,3,7,8-PCDD
l3C-1,2,3,6t7,8-HCDD
13C-1,2,3,4,6,7,8-H7CDD
13C-1,2,3,4,6,7,8,9-OCDD
ANALYSIS
Tetrachlorodioxins and furans
Pentachlorodioxins and furans
Hexachlorodioxins and furans
Heptachlorodioxins and furans
Octachlorodioxins and furans
924
-------
TABLE 5: RESULTS OF DIOXIN/FURAN ANALYSIS
Compound
Total Tetra-Dioxin
Total Tetra-Furan
Total Penta-Dioxin
Total Penta-Furan
Total Hexa-Dioxtn
Total Hexa-Furan
Total Hepta-Dioxin
Total Hepta-Furan
Total Octa-Dioxin
Total Octa-Furan
Concentration,
Low
0.5
0.7
2.8
1.3
1.3
1.7
2.3
1.0
1.7
1.7
ng/m3
High
3.0
2.2
4.3
[5.3]
12.0
6.5
4.9
2.2
[9.4]
4.5
Blank Train
ng
0.4
0.4
1.7
1.0
0.3
0.3
0.6
0.3
1.3
0.5
[ ] Positive Identification
TABLE 6: RESULTS OF BASE NEUTRAL/ACID COMPOUND ANALYSIS
Compound
Trichlorobenzene
Tetrachlorobenzene
Trichlorophenol
Pentachlorophenol
Hexachforoethane
Hexachlorobutadiene
Hexachlorobenzene
Octachlorostyrene
Pentachlorobenzene
Pentachloroethane
Concentration,
Low
5.0
23
15
23
15
[6.2]
[6.2]
64
25
26
ug/m3
High
7.3
32
22
33
27
[10]
[7.8]
112
45
36
Blank Train
ug
20
90
60
90
60
80
90
250
100
100
[ ] Positive Identification
925
-------
TABLE 7: RESULTS OF VOLATILE ORGANICS
Compound Concentration, ppm
Chloroform 0.009 ± 0.003
Trichloethane 0.028 ± 0.015
Trichloethylene 0.002 ± 0.001
Tetrachloethylene 0.066 ± 0.012
1,2-dichloroethane 0.014 ± 0.009
Vinyl chloride < 0.5
Propylene dichloride 0.0008 ± 0.0002
TABLE 8: RESULTS OF DIRTY DOZEN DIOXINS/FURANS
Concentration, ng/m3
Isomer Low High
2,3,7,8 - TCDD
2,3,7,8 - TCDF
1,2,3,7,8 - PCDD
1,2,3,7,8 - PCDF
1,2,3,4,7,8 - HXCDD
1, 2,3,4,7,8 -HXCDF
1,2,3,7,8,9 - HXCDD
1,2,3,7,8,9 - HXCDF
1,2,3,6,7,8 - HXCDD
1,2,3,6,7,8 - HXCDF
1,2,3,4,6,7,8 - H CDD
1,2,3,4,6,7,8 - H CDD
0.1
0.1
1.4
0.7
0.4
0.4
0.3
0.1
0.2
0.2
0.3
0.3
3.0
2.0
2.2
2.0
3.2
3.0
4.1
3.1
3.2
3.1
2.1
0.9
Note: All values represent instrumental limit of detection and none of the compounds were
positively identified.
926
-------
TABLE 9: TYPICAL MATRIX SPIKE RESULTS - IMPINGER SAMPLES
Compound
TCDD
PCDD
HCDD
H7DD
OCDD
OCDF
Paramenter
Amt. Spiked
Amt. Recvd
% Recovery
Amt. Spiked
Amt. Recvd
% Recovery
Amt. Spiked
Amt. Recvd
% Recovery
Amt. Spiked
Amt. Recvd
% Recovery
Amt. Spiked
Amt. Recvd
% Recovery
Amt. Spiked
Amt. Recvd
% Recovery
Matrix Spike
7.5
7.80
104.00
15
14.875
99.17
15
16.482
109.88
7.5
7.771
103.61
7.5
9.381
125.08
15
16.948
225.97
Matrix Spike
Duplicate
7.5
7.79
103.91
15
15.519
103.46
15
15.602
104.01
7.5
7.599
101.32
7.5
9.296
123.95
7.5
17.978
239.71
Note: All amounts spiked were in nanograms.
TABLE 10: RESULTS OF DIOXIN ANALYSIS FROM A TYPICAL BLANK TRAIN
Compound
Total
Tetra-Dioxin
Total
Penta-Dioxin
Total
Hexa-Dioxin
Total
Hepta-Dioxin
Total
Octa-Dioxin
Probe &
Nozzle
(ng)
<0.1
<0.7
<0.4
<0.8
<1.2
Filter
(ng)
<0.3
<7.9
<10
<2.5
<7.6
Condenser
(ng)
<0.2
<1.1
<0.5
<1.1
<2.0
Sorbent
Tube
(ng)
<0.7
<7.4
<1.3
<5.4
<18
Implnger
#1
(ng)
<0.4
<1.9
<0.6
<0.4
<1.3
Total
ng
<1.8
<1.9
<25
<10
<30
927
-------
Remote Optical Sensing of VOCs:
Application to Superfund Activities
Timothy R. Minnich and Robert L. Scotto
Blasland, Bouck & Lee
Edison, New Jersey
Thomas H. Pritchett
U.S. Environmental Protection Agency
Environmental Response Team
Edison, New Jersey
INTRODUCTION
Ambient air monitoring is conducted for a variety of purposes under the
Comprehensive Environmental Response, Compensation and Liability Act of
1980 (CERCLA or "Superfund"), as amended by the Superfund Amendments and
Reauthorization Act of 1986 (SARA). Remote optical sensing can be an ideal
method for assessing the air migration pathway for many activities spanning
the Superfund process. However, proper application of this emerging
technology requires a clear understanding of its uses and limitations, as well
as definition of appropriate data quality objectives (DQOs) for each activity.
DQOs are defined as the qualitative and quantitative statements that specify
the quality of data required to support a decision-maker's needs.1'2 In this
context, DQOs include ambient air monitoring program objectives and the data
quality needs necessary to ensure their achievement.
This paper is directed toward environmental managers responsible for the
design and implementation of Superfund-type ambient air quality monitoring
investigations. DQOs are identified for a variety of air-related Superfund
activities for which application of this technology is appropriate. Suggested
monitoring approaches and methodologies for achieving each DQO are
presented. Finally, limitations are discussed for the above applications.
OVERVIEW OF MONITORING TECHNIQUE
Remote sensing is generally defined as the ability to detect an object or
phenomenon without having the detector in direct contact with the object
being detected. This technique can be either passive where some natural
source of radiation is received, as in the case of a satellite that maps
infrared (!R) emissions from the earth's surface, or active where the sensing
system probes the object of measurement in some way and infers the
parameter being measured from the object's response, as in the case of
radar. Remote sensing, as applied to air toxics measurements, is an active
928
-------
optical technique that involves the use of long-path spectroscopy, Long-path
spectroscopy has been used in a wide va'iety of environmental applications
during the past 15 years.5'9 The basic principle involves the generation and
propagation of light in a known wavelength, or range of wavelengths, and
the subsequent detection and spectral analysis after some pathlength has
been traversed. By creating a ratio between the detected signal and the
emitted signal and comparing this ratio to a library of known absorption
spectra, individual species can be identified and quantified. Certain
compounds are observed principally in the IR, other compounds are observed
principally in the ultraviolet (UV), and still others are observed equally well
in either.
A fundamental difference between remote optical sensing and traditional point-
sampling techniques involves the nature of the data output. This difference
must be clearly understood by those who interpret and apply remote sensing
data. As will be shown later, the key to the tremendous advantages offered
by remote optical sensing for Superfund application lies in the nature of its
data output. For point-sampling techniques, gaseous contaminant
concentrations are generally reported either in mass of contaminant per unit
volume of gas, such as micrograms per cubic meter (ug/m3), or in volume
of contaminant per volume of gas, such as parts per million (ppm) or parts
per billion (ppb). Open-path remote sensing, however, is a volume-averaging
technique where total contaminant burden is measured within a cylinder
defined by the finite cross-section of the light beam at each end and the
length of the beam itself. This volume average is reported as a path-
integrated concentration typically normalized to a pathlength of 1 meter. If,
lor example, an integrated concentration of 30 ppm-m is reported, no
information concerning the contaminant distribution can be directly inferred,
and the instrument response would be identical whether there was a uniform
concentration of 30 ppm over a distance of 1 meter, 3 ppm over a distance
of 10 meters, 300 ppb over a distance of 100 meters, or 3 ppb over a
distance of 1 kilometer,
DATA QUALITY NEEDS
Ambient air quality monitoring programs conducted under Superfund are
frequently designed and implemented without adequate understanding of the
decisions to be made based on the data. Data collected during programs
employing traditional point-sampling techniques at complex hazardous waste
sites rarely have been of a quality and quantity sufficient to achieve the
'DQOs required in meeting a decision-maker's needs.
The principal deficiencies associated with data collected via traditional point-
£.ampling techniques involve data representativeness and comparability. The
complex nature of air dispersion, combined with the spatial variability of air
emissions typical of Superfund sites, has made it difficult, and often
impossible, to use point-sampling techniques to identify and characterize
contaminant plumes for subsequent assessment of long-term downwind air
innpacts. Compounding this problem of data representativeness is an
increasing need to generate real-time data for a variety of health-related
assessments. On-site and off-site health and safety concerns during site
remediation and emergency response activities frequently dictate the need for
continuous real-time assessment of contaminant emissions and ambient air
concentrations.
929
-------
Data comparability has historically been a problem for point-sampling
techniques because of an inability to collect data with identical meteorological
conditions throughout the sampling period. This, combined with a lack of
adequate documentation of atmospheric transport conditions, has made it
virtually impossible to produce air monitoring data that are comparable when
collected over sequential stages of the Superfund process.
A carefully designed air monitoring program involving remote optical sensing,
together with complementary traditional point sampling and on-site
meteorological monitoring, can surmount these problems and produce
statistically valid, volume-averaged measurements over long distances
(pathlengths) in near real time (less than 15 minutes). The statistical
representativeness intrinsic to path-integrated air measurement data ensures
its comparability throughout the Superfund process.
SUPERFUND APPLICATIONS AND MONITORING METHODOLOGIES
Table 1 summarizes Superfund activities for which application of remote
sensing is appropriate, and presents corresponding DQOs. Three main types
of Superfund activities are identified: site assessment, remedial, and
emergency response. Within each of these types, specific activities are
identified along with their corresponding DQOs. It must be kept in mind that
remote sensing may not be suitable to meet all air monitoring needs, as
migration of contaminated particulate matter cannot be assessed via this
technique. Following is a discussion of each Superfund application and
corresponding DQO.
Site Assessment Applications
HRS Model Scoring
Hazard Ranking System Model (HRS) scoring is performed to assess the
relative degree of risk posed by known or suspected uncontrolled hazardous
waste sites, thereby providing a uniform basis for site placement on the
National Priorities List (NPL) for cleanup. Separate scores are calculated for
each contaminant migratory pathway. A score for the air route is calculated
based either on a demonstrated "observed release" or on the demonstrated
potential for an observed release. Demonstration of an observed release is
preferable, and generally requires air measurement data showing detectable
concentrations off site in the downwind direction together with appropriate
documentation supporting the contention that the site is the source. The
HRS Model does not attempt to relate a non-detect to some maximum level
of risk.
The DQO for HRS Model scoring is to conclusively demonstrate the presence
or absence of off-site migration of air contaminants. Remote optical sensing
is ideally suited to achieve this DQO if the plume is non-buoyant and
originates from a ground-level source (i.e., the plume hugs the ground as it
travels downwind from the source). This is a good assumption for most
hazardous waste sites, as gaseous contaminant releases are generally the
result of volatilization from impoundments or hazardous waste storage areas.
Emissions from such sources can be detected if the long-path beam is shone
along the site perimeter, downwind of the suspected source area and across
the plume. If contaminants are detected, upwind monitoring would then be
930
-------
conducted along the upwind site perimeter to demonstrate that the
contaminants detected downwind are attributable to the site and are not beinc
advected across the site from some upwind source. Because air monitorinc
for HRS Model scoring is typically a one-time event, it is important that i
be conducted under meteorological conditions representative of reasonable
worst-case emissions and transport.
Remedial Applications
Rl Air Monitoring
Remedial investigations (RIs) are conductsd in conjunction with feasibility
studies (FSs) to assess the nature and extent of contamination at NPL sites
This information is used to support the detailed engineering evaluation of all
technically feasible remedial alternatives aid to provide a baseline againsl
which the effectiveness of each alternative, including the no-action alternative,
can be assessed. Because air emissions are a "secondary" effect, indicative
of the presence of waste products or contaminated soils or water, air
monitoring data are not appropriate to support remedial engineering decisions.
Instead, air monitoring data generated during the Rl are typically used as a
baseline against which the effectiveness of each remedial alternative is
assessed, and to support the appropriateness of the no-action alternative.
The DQO for Rl air monitoring is to provide air contaminant data suitable for
generation of accurate emission rate estimates under undisturbed conditions.
These emission rates can then be compared to those predicted to occur
following implementation of various remedial actions, and can also be used
as input into air quality dispersion models to evaluate downwind community
health impacts in support ol evaluation of the no-action alternative.
Remote optical sensing is one of only a few techniques available to measure
emission rates for Rl baseline emission estimates. For large sites with
complex sources of air emissions, remote sensing may be the only technique
that is not cost-prohibitive. Path-integrated concentrations measured at known
downwind distances, together with on site meteorology, are used to satisfy
basic relationships derived from classical Giaussian dispersion theory10 which
lead directly to quantification of source emission rates.11 The measurement
technique requires minimal understanding of site characteristics, and the
resultant emission rates can be easily calcu ated in the field. It is important
that both upwind and downwind measurements be made so that issues of
source apportionment are resolved,
FS Air Monitoring
Air is the only media in which the exposure potential drastically increases
during site cleanups. Because many Superfund sites are very close to
downwind populations that could be affected by these emissions, it is
becoming more common for pilot studies to be conducted during either the
feasibility study or the remedial design itself. Pilot studies simulate, on a
s,mall scale, the ultimate cleanup in order that emission rates for the
contaminants can be determined for each of the cleanup steps. These
emission rates can then be modeled in order to estimate the downwind health
impacts associated with that cleanup option and to determine the necessary
degree of emission controls.
931
-------
The DQO for FS air monitoring is to provide air contaminant data suitable
for generation of accurate emission rate estimates during pilot-scale
assessments of cleanup alternatives. Remote optical sensing can be used to
meet this DQO in a manner identical to that described for obtaining baseline
emissions during the Rl. Once downwind exposures are quantified, the
suitability of various emission controls can be easily evaluated.
Remediation Air Monitoring
Remediation refers to the implementation of waste removal, treatment, and/or
containment actions based on RI/FS results. Air monitoring is conducted
during site remediations to ensure that cleanup activities do not pose adverse
health impacts to on-site workers or downwind residents. Because the
principal health concern usually involves short-term exposure, obtaining results
in real time (on the order of seconds to minutes) is an important data
requirement. As a rule, monitoring methods that require sample collection
and subsequent off-site laboratory analysis cannot meet this requirement.
The DQO for air monitoring during site remediations is to provide real-time
air contaminant data suitable for assessing on-site and off-site health impacts.
Remote optical sensing can generally be used in a cost-effective manner to
achieve this DQO, In a manner similar to the Rl application, path-integrated
concentrations measured at a known distance downwind of a remediation,
together with on-site meteorology, are used to satisfy basic relationships
derived from classical Gaussian dispersion theory.10 However, for this
application, these data lead directly to a slightly conservative estimate of
maximum plume centerline concentration at any downwind distance of concern.
These maximum concentrations (in units of ppb or ug/m3) can then be directly
compared to health-based standards or action levels.
Post-Remedial Applications
Post-Remedial Air Monitoring
Periodic monitoring of a remediated source is sometimes required to ensure
the long-term integrity of the treatment or containment action. Quarterly or
annual ground water monitoring is most often required, but post-remedial air
monitoring is sometimes a provision of the record of decision (ROD) for a
site.
The DQO for post-remedial air monitoring is to ensure that contaminants do
not create adverse off-site community health impacts. This DQO can be met
cost-effectively using remote optical sensing in a manner identical to that
described for obtaining baseline emission estimates during the Rl. However,
the post-remedial air monitoring program is better focused because more
information is known about the site. Because the frequency with which post-
remedial air monitoring is required is not likely to exceed quarterly, it is
again important that it be conducted under meteorological conditions
representative of reasonable worst-case emissions and transport.
The monitoring method involves the measurement of path-integrated
concentrations upwind and immediately downwind of the site so that issues
of source apportionment are addressed. It is important that the detection
limits for the contaminants of concern are low enough so that non-detects
932
-------
measured under given meteorological conditions translate to an acceptable
annual risk to downwind receptors based on local climatology. Depending or
local climatology, on the particular contaminant(s) of concern, on what is
considered an acceptable risk, and on current instrument detection limits, this
may not always be possible.
Emergency Response Applications
Emergency Response Air Monitoring
Emergency response actions are carried out for any situation that poses an
imminent and substantial danger to public health and welfare. Frequently,
this imminent danger is in the form of adverse health impacts arising from
exposure to air contaminants by downwind populations. Air monitoring
frequently represents a major need during emergency response activities.
The DQO for emergency response air monitoring is to provide real-time air
contaminant data suitable for assessing off-site community health impacts.
Remote optical sensing can be used to achieve this DQO in two ways. First,
t can be used at a fixed downwind distance to estimate maximum plume
oenterline concentrations at any downwind distance of concern, in a manner
identical to that described for remediation air monitoring. Second, remote
isensing can provide an accurate source term estimation for input into real-
time emergency response models (e.g., CAMEO) to facilitate emergency
response decisions. Such models can facilitate response decisions by
providing data more refined than those generated by the simple Gaussian
dispersion relationships mentioned above.
LIMITATIONS
Several significant limitations exist concerning application of remote optical
sensing to Superfund activities. Although the technique is very powerful and
can achieve many of the necessary DQOs, its employment in the field without
a comprehensive understanding of each of these limitations can jeopardize the
results of the investigation. These limitations are:
1. Contaminant Identification. Positive identification of contaminants
measured via broad-band long-path spectroscopy is accomplished by
comparing the spectra of the measured contaminant to its library
spectra. The "goodness of fit" between these spectra is typically
established by a statistical method such as a classical least-squares
analysis. As the number of interfering adsorption bands in the
measured spectra increases, the certainty with which the contaminant
of interest can be identified decreases. These interfering adsorption
bands can arise either from other contaminants or from naturally
occurring gases such as water vapor. Because of this, it is
generally advisable to conduct some type of confirmatory sampling
and analysis using traditional point-sampling techniques, especially
when there are more than several contaminants of concern.
2. Detection Limits. Detection limits are derived in the field on a
compound-specific basis. Several factors govern detection limits,
including refraction of the beam due to thermal air currents,
variations of naturally occurring interferers, and high levels at
933
-------
cross-interfering target compounds. The detection limits associated
with most remote optical sensing techniques are not sufficiently low
to facilitate direct comparison to ambient air action levels at
downwind receptor locations; therefore, this is not a recommended
application. Detection limits on the order of 5 to 50 ppm-m are
generally achievable for most compounds of interest.
3. Meteorology. Although not a limitation per se, meteorology is an
important factor governing the interpretation of long-path measurement
data. The most significant meteorological variable is atmospheric
stability, as this controls the extent to which the plume disperses.
Because containment of the plume (in the horizontal) within the
beam is a requisite for Superfund application as discussed, an
understanding of horizontal dispersion as a function of stability is
necessary.
Understanding how stability affects plume dispersion in the vertical
is especially important in the interpretation of the data. At an
arbitrary distance of 50 meters downwind of a source that emits at
a constant rate, the measured path-integrated concentration can vary
by nearly two orders of magnitude, depending on the stability class
and the associated vertical dispersion. This illustrates the need to
have accurate on-site meteorological data.
4. Area Sources. The procedure described to derive source emission
rates assumes a virtual point-source release in the absence of site-
specific emissions information. The modeled downwind concentrations
from large area sources will tend to be slightly higher than actual
concentrations. However, the conservative nature of these
calculations is generally acceptable, and often desirable, for most
Superfund applications.
5. Logistical. Logistical concerns associated with the use of remote
sensing, regardless of its application, can be significant. A clear
line of sight must exist in the crosswind direction both upwind and
downwind of the site, For some applications, this may involve
clearing a path through existing vegetation or simply waiting for the
wind to blow from some pre-determined direction.
934
-------
REFERENCES
1. U.S. Environmental Protection Agency, "Guidance on Applying the Data
Quality Objectives Process for Ambien: Air Monitoring Around Superfund
Sites (Stages I and II)," EPA-450/4-89-015, ' August 1989.
2. U.S. Environmental Protection Agency, "Guidance on Applying the Data
Quality Objectives Process for Ambienl Air Monitoring Around Superfund
Sites (Stage III)," EPA-450/4-90-005, March 1990.
3. M.L. Spartz, M.R. Witkowski, J.H. Fatoley, J.M. Jarvis, J.S. White, J.V.
Paukstelis, P.M. Hammaker, W.G. Fateley, R.E. Carter, M. Thomas, D.D.
Lane, G,A. Marotz, B.J. Fairiess, T. Holloway, J.L. Hudson, and D.F.
Gurka, "Evaluation of a Mobile FTIR System for Rapid VOC
Determination, Part I: Preliminary Qualitative and Quantitative Calibration
Results," Amer. Env. Laboratory, November 1989.
4. R.J. Bath, T.R. Minnich, R.M. Naman, R.D. Spear, O, Simpson, J. Faust,
D.H. Stedman, S.E. McLaren, W.F, Herget, and W.M, Vaughan, "Remote
Sensing of Air Toxics Using State-of-the-Art Techniques," presented at
the 1989 EPA/AWMA International Symposium on Measurement of Toxic
and Related Air Pollutants, Raleigh, NO, May 1989,
5. T.R. Minnich, R.J. Bath, and R.M Naman, "Remote Sensing of Air Toxics
for Pre-Remedial Hazardous Waste Site Investigations," presented at the
82nd Air and Waste Management Association Annual Meeting, Anaheim,
CA, June 1989.
6. S.E. McLaren, D.H. Stedman, G.A. Bishop, M.R. Burkhardt, and C.P,
DiGuardia, "Remote Sensing of Aromatic Hydrocarbons at Hazardous
Waste Sites Using Long Path Ultraviolet Spectroscopy," presented at the
82nd Air and Waste Management Association Annual Meeting, Anaheim,
CA, June 1989.
7. W.F. Herget, "Analysis of Gaseous Air Pollutants Using a Mobile FTIR
System," Amer. Labs, 72: (1982).
!B. V. Platt, D. Perver, H.W. Patz, "Determination of Trace Atmospheric
Species by Long Path UV Spectrometiy," J. Geophys. Res. 84: 6329
(1979).
9. P.L. Hanst, "Air Pollution Measurement by Fourier Transform
Spectroscopy," Appl. Opt. 17: 1360 (1978).
10. U.S. Environmental Protection Agency "Work Book of Atmospheric
Dispersion Estimates," Office of Air Programs, Research Triangle Park,
NC, GAP Publication No. AP-26, D.B, Turner, Revised 1970.
11. R.L. Scotto, T.R. Minnich, and M.R. Leo, "Emissions Estimation and
Dispersion Analysis Using Path-Integrated Air Measurement Data from
Hazardous Waste Sites," to be presented at the 83rd Air and Waste
Management Association Annual Meeting, Pittsburgh, PA, June 1990.
935
-------
TABLE 1
SUPERFUND APPLICATIONS AND ASSOCIATED DQOs
Superfund Application
Site Assessment
MRS Model scoring
for the air route
DQQ
To conclusively demonstrate the
presence or absence of off-site
migration of air contaminants.
Remedial
Rt air monitoring
FS air monitoring
Remediation air
monitoring
Post-Remedial
Post-remedial air
monitoring
Emergency Response
Emergency response air
monitoring
To provide air contaminant data
suitable for generation of accurate
emission rate estimates under
undisturbed conditions.
To provide air contaminant data
suitable for generation of accurate
emission rate estimates during
pilot-scale assessments of cleanup
alternatives.
To provide real-time air contaminant
data suitable for assessing on-site
and off-site health impacts.
To ensure that air contaminants
do not create adverse off-site
community health impacts.
To provide real-time air contaminant
data suitable for assessing off-site
community health impacts,
936
-------
Measurements of Emissions at a Chemical Waste Water Treatment Site with an
Open Path Remote Fourier Transform Interferometer
Orman A. Simpson and Robert H. Kagann
MDA Scientific, Inc.
Norcross, Georgia
Remote measurements of gaseous emissions from a chemical waste water treatment
site were performed using quantitative Fourier transform infrared (FTIR) spectroscopic
techniques. The remote open path FTIR unit was bistatic: the transmitter and the receiver
were placed on opposite sides of the volume of atmosphere to be measured. The beam passed
through the atmosphere in which quantitative measurements of chemical species were to be
made.
Path integrated concentrations of ammonia, methanol, methylene chloride, and sulfur
hexafluoride were measured with the FTIR unit along the downwind side of the waste
treatment site along a horizontal beam path at various heights above ground. The
concentrations as a function of height in most cases followed a monotonic curve with the
greater concentrations nearer the ground as would be expected from simple dispersion
models.
937
-------
We used an FTIR remote sensor to measure the emissions from an equalization pond
ata waste water treatment facility as part of a chemical company program to test emission
models. The setup used consisted of a ir beam tranmitter which was a four inch Cassegrain
telescope with a glow bar ir source at its focus. The ir beam was transmitted through an
atmospheric path of 60 m along the downwind shore of an equalization pond to the receiver.
The receiver is a 14 inch F/6 Cassegrain telescope. The beam was recollimated with a
diameter of 1.5 inches and then passed into the Michelson interferometer of the Fourier
transform spectrometer. We used the simple bistatic configuration to facilitate changing the
height of the beam. Measurements were made at five different heights, 1.5, 3, 5, 12, 18 ft.
The different heights were obtained using two towers with five mirrors on each, as shown in
Figure 1. The setup time for changing the heights was 30 to 40 minutes.
Receiver
5.5 ft
Figure 1. The Bistatic FTIR Remote Measurement Configuration Used to Vary the
Elevation of the FTIR Concentration Measurements. The beam is reflected by
a total of three mirrors in the field, M1, M2 and M3. For each elevation, the
tilt angle of the transmitter and the mirror M3 must be adjusted.
Initially, the infrared field spectra were visually compared to clean air spectra in
order to make qualitative determinations of the chemical emissions. Upon identifying the
presence of methylene chloride, ammonia, methanol and sulfur hexafluoride (which was used
as a tracer), we performed a classical least squares fit1 to reference spectra of precisely
known quantities of these species. These spectra are part of a library of reference spectra of
140 species, which we brought to the field stored on the computer's hard disk. Table 1 shows
the results of the measurements. The concentration are averages over the 60 meter path,
which is slightly larger than the width of the equalization pond (-50 m). We were able to
obtain reliable measurements of sulfur hexafluoride at very low concentrations because of
the extremely strong absorption feature {Q branch) which appears at 944 cm"1. Two of the
measurements ( runs 614.07 and 615.01) were made on the side of the pond parallel to the
wind direction to verify that the source of the chemicals was indeed the pond. These
determinations were much lower than the downwind determinations, but since this
configuration was not truly upwind, small quantities of the chemical emitted from the pond
still crossed the ir beam. True upwind measurements were not attempted because they would
have involved a long setup time.
-------
We are presently investigating the sources of error in the FTIR remote sensor, in
order to determine with confidence the accuracy of concentration determinations. For the
present, we estimate the accuracy to be -15 percent when the concentrations are not too
close to the minimum detectable limit.
Run
Number
613.01
613.02
614.01
614.02
614.03
614.04
614.05
614.06
614,07
615.01
615.02
615.03
615.04
615.05
615.06
615.07
615.08
615.09
615.10
615.11
615.12
616.01
Time
8:07
8:54
15:55
17:20
17:55
18:31
19:07
19:46
20:39
10:43
13:32
14:02
14:41
16:35
17:12
17:37
18:13
18:38
19:23
19:49
20:12
9:16
Beam
Height
(ft)
5
5
5
3
1.5
1.5
3
5
6.25
6.25
1.5
3
5
12
18
18
18
12
5
3
1.5
6.25
Effective
Pathlength
(m)
50
50
60
60
60
60
60
60
125
62
62
62
62
62
62
62
62
62
62
62
62
62
SF6
Tracer
Release
small
small
no
no
no
small
small
small
no
small
small
small
small
small
small
no
large*
large*
large*
large*
large*
no
SF6
(ppb)
2.3(1.5)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
11.3(2.6)
18.2(3.0)
18(13)
25.1(1.9)
25.7(2.0)
0
Methylene
Chloride
(ppb)
607(94)
1568(138)
443(115)
1445(156)
1839(99)
728(126)
0
0
1886(241)
1123(265)
1529(297)
424(205)
0
612(431)
605(428)
416(156)
909(435)
1265(105)
1692(115)
1300(180)
Methanol
(ppb)
1670(10)
680(50)
900(40)
1220(80)
1070(60)
1270(70)
1320(60)
1110(90)
277(20)
160(50)
1770(130)
1450(150)
1200(110)
740(100)
473(30)
548(37)
540(40)
610(50)
1520(240)
1570(80)
1370(90)
0
Ammonia
(ppb)
500(60)
58(32)
850(40)
910(70)
910(70)
970(80)
950(50)
690(70)
176(21)
0
550(110)
210(150)
320(100)
200(150)
200(40)
250(40)
260(60)
270(70)
190(350)
730(70)
840(80)
0
Table 1 Concentration Values for all Field measurements at the Equalization Pond. The first
threee numbers in the file names are coded for the dates June 13, 14, 15 and 16, 1989,
and the suffix is the measurement sequence number. The "effective" pathlengths aree the
portions of the total pathlengths where the chemicals where present. The "small* SF6
tracer release are a factor of 10 smaller than the "large" releases. The small releases
resulted in concentrations at the beam position which were lower than the minimum
detectable limit. The numbers in the parenthese are the estimated standard deviation of
the least squares fit propagated to the concentrations determinations. The italicized
rows (runs 614.07 and 615.01) were measured in an "upwind" configuration.
Reference
1. D. M. Haaland and R. G. Easterling, "Application of New Least-squares Methods for
the quantitative infrared analysis of multicomponent samples," Applied
Spectroscopy, 36, 665-673, (1982).
939
-------
APPLICATION OF A HENE LASER TO HYDROCARBON LEAK
DETECTION OVER AN OIL FIELD
Joel D. Cline
Mobil Research and Development, Dallas Texas
Gilbert R. Jersey
Mobil Research and Development, PaulSboro, New Jersey
Larry W. Goodwin and Michael N. Crunk
Mobil Exploration and Producing, Midland, Texas
Orman A. Simpson
MDA, Scientific Atlanta, Georgia
As a part of an oil field modernization study, a laser-based hydrocarbon leak detection
system was evaluated on the H. O. Mahoney lease in Yoakum County, Texas. The purpose of
the test was to determine if small amounts of methane gas associated with oil leaks could be
reliably detected in the atmosphere by long-path laser measurements.
The test system consisted of a 6 mW HeNe laser tuned to emit radiation at 3.392 (im, a
mechanical chopper, a lead sulfide detector, a lock-in amplifier, and a 15 cm diameter
simultaneous transmitter/receiver telescope. The signal was returned to the detector with a 6
cm cube retroreflector of 30 arc second accuracy. The response of the laser return signal to
both ambient concentrations of methane and to controlled releases of methane gas was
measured over a range of wind speeds and air temperatures.
The cross-plume concentration of methane was determined with the HeNe laser and this
value was compared to predictions from a 2-D plume dispersion model. In general, agreement
between observations and model predictions was excellent, even though the predictability of
the model is uncertain at downwind distances less than 100 m. Poor agreement between the
observed methane column density and the model result usually occurred during light and
variable winds. These observations also showed that the signal-to-noise ratio of the single-
beam system is strongly dependent on temporal fluctuations in the atmospheric concentration of
methane and on local meteorological conditions (e.g., strong solar insolation, dust, etc.)
940
-------
INTRODUCTION
In 1987, Mobil Oil Inc. modernized an oil field production facility in the Wasson field
of west Texas. A part of that program involved the testing of a remote detection system that
would provide early warning of leaking petroleum from storage tanks, flow lines, process
equipment, and pumps on the lease. The chief concern was the environmental detection of
leaking oil, but because oil is not readily detected by remote methods, methane gas, which is a
major component of the produced fluids, was selected as the tracer of interest.
After examining several optical systems, we; selected a continuous wave HeNe laser
tuned to 3.392(j,m. This laser is an ideal choice for the detection of atmospheric methane
because its emission wavelength is nearly coincident with one of the rotational-vibrational
absorption bands of methane (Kucerovsky et al., 1973; Murray, 1978; Patel, 1978; Grant and
Menzies, 1983). Previous studies by Kucerovsky et al., (1973); Grant, (1986) have clearly
demonstrated the utility of this technique to long path-length monitoring of atmospheric Qij.
In this report we describe the response characteristics of a double-ended, single beam,
HeNe laser to point-source releases of methane gas on the H. O. Mahoney lease. The purpose of
the test was to determine: (i) fluctuations in the ambient concentration of methane on the lease and
(ii) response characteristics of the HeNe laser to smell, controlled, upwind releases of methane
gas.
EXPERIMENTAL
Instrumentation
The test breadboard unit consisted of a 6 mW HeNe laser (Spectra-physics Model 127)
tuned to 3.392 )im, a mechanical chopper for beam modulation, a cooled, lead sulfide detector, a
lock-in amplifier for signal processing, a 15 cm simultaneous transmitter/receiver telescope, and
a strip-chart recorder (Figure 1). Both transmitting and receiving signals were propagated along
the same optical axis by the use of a double-sided secondary mirror in the Newtonian
transmitter/receiver telescope. Operational parameters for the laser system are shown in Table 1.
Physjcal getting
The H. O. Mahoney lease is located in the Wasson field of Yoakum County, Texas (Fig
2). The lease contains 16 producing wells (40 acre spacing) and nine CO2 injector wells, with a
centrally located tank battery that consists of storage tanks and process equipment (Fig. 2). Flow
lines on or near the surface of the ground transport oil and gas from the wells to the tank battery,
thus likely sources of leaking oil and methane gas will be at or near ground level. The H. O.
Mahoney lease is particularly well-suited for the long-path length monitoring of methane gas
because the topography is relatively flat and only partially covered with low-growing plants to
approximately 3 ft. The only significant obstacles to the laser beam are the storage tanks near the
center of the lease; they are approximately 20 ft high.
The ambient concentration of methane (includes other absorbing species at 3.392 |j.m)
was determined along three transects located near the center of the lease (Fig. 2). These transects
were chosen to provide ambient atmospheric conditions under varying wind trajectories.
941
-------
Point releases of methane gas were also made along the same three transects, depending
on the local wind conditions and specific objectives. Methane gas (98% pure) from a standard
220 ft3 cylinder was released at rates of 330 mg Cltys and 660 mg Cltys (1500 ft3/d and 3000
ft3/d, respectively) approximately 1 m above the ground. Flow rates of methane gas were
determined with a flow meter attached to the tank and a two-stage pressure regulator. Wind
velocities were determined with a hand-held, digital anemometer, average and range of velocities
were determined over a one minute time period at the beginning and near the end of the release
period. Experimental conditions are displayed in Table 2.
RESULTS AND DISCUSSION
Ambient Conditions
Figure 3 shows a 15 min record of the laser response to the ambient atmosphere
upwind of the tank battery. In order to compress the analog record, two second intervals of the
power ratio (Pj/Pt) were averaged for 12 seconds and the resulting mean value plotted. This
ratio is the received power (Pr) relative to the transmitted power (Pt) as determined on a strip-
chart recorder. Peak-to-peak fluctuations for low wind conditions upwind of the tank battery
(Fig. 3a) were only 4%, a value typically recorded during the early evening and early morning
hours. Not shown in this particular record are the relatively infrequent, but large amplitude,
low frequency fluctuations that occurred during light and variable winds.
During periods of high solar insolation, peak-to-peak fluctuations in the power ratio
generally increased (Fig. 3b). In this particular record the peak-to-peak fluctuation was 5%,
but variations up to 10% were noted during some afternoon tests. Fluctuations of this size are
assumed to be related to density fluctuations in the lower atmosphere brought about by strong
ground heating, but thermally induced electrical effects in the laser and associated signal
processing equipment may also have been important. The afternoon shade temperature reached
106°F on May 23 and May 24.
Figure 4 shows the relative power response of the HeNe laser to small releases of
methane gas over a 90 min period. In this case the analog signal was averaged for 1 min.
Observed reductions in the received signal varied from 10-20% for the small releases (330 mg
CH4/s) to 20-40% for the larger releases (660 mg CH4/s). Rapid fluctuations in the signal
were evident as the methane plume wandered in and out of the laser beam. In some cases the
laser response did not return to baseline (approx. 90%) after the gas had been turned off. It is
likely that insufficient time was allowed for the methane plume to reach stationary conditions.
Plume Dispersion Model
In order to estimate the response of the laser return signal to a leak of methane gas
anywhere on the lease (maximum range approximately 1 km), methane releases at relatively
short distances were modeled with a stationary, 2-D, plume dispersion model (e.g., Csanady,
1973). Because the HeNe laser measures the column density of methane in the atmosphere, an
integrated form of the plume dispersion model was adopted. Mean wind speeds were
estimated from our measurements, while estimates of the dispersion coefficient in the z-
direction (e.g., az) were linearly extrapolated from Gifford's (1961) values. Assuming that the
plume trajectory was orthogonal to the laser beam and, at the same elevation as the laser beam,
the solution to the plume dispersion model is:
Q[l+exp(-2h2/q2)]
Y2~7cua2
942
-------
where Q is the release rate (ing CH4/s), Cm is the column density of methane (mg CH4/m2), u
is the mean wind speed (m/s), and h is the elevation of the laser beam (in this case, 1 m).
Similarly, the laser response to a methane plume can be estimated from the background-
corrected LIDAR equation (Grant, 1986; eqn. 1) The result is:
~ 2a
where Cj is the predicted column density of methane, Pj is power response of the laser in the
methane cloud, P^ is the power response to the ambient methane concentration, and a is the
methane absorption coefficient. The methane absorption coeficient is 770 m'1 atnr1 (Grant,
1986).
Methane concentration, as determined by the HeNe laser (Q) is compared to the
calculated value (Cm) in Table 3. Atmospheric stabilities (i.e., A-D) were estimated from the
wind velocity, solar insolation, and cloud cover at the time of the observations. There is
excellent agreement between the laser response and the prediction of the plume dispersion
model for most of the experimental runs. The average difference between the observed and
predicted values is approximately 30%, which is remarkably good agreement when one
considers the uncertainties in the model predictions at short downwind distances so near the
ground.
Significant departures between the observed and expected results were noted in
experiments 6, 10, and 11. In experimental runs 6 and 10, the methane column density, as
determined by the laser, was lower than the model prediction. There are several possible
reasons for this including, misalignment of the methane plume and the laser beam, mean wind
speed estimate was too low, or the model value for the vertical dispersion coefficient was too
low. In run 6 winds averaged less than 1 m/s, thus there is good reason to believe that the
laser beam only occasionally intersected the methane plume
The difference between observed and model methane concentrations in Run 1 1 was
unique in that the laser response was larger. In this case our estimate of the atmospheric
stability may have been in error. By assuming a more stable atmosphere (C or D), the model
cross-plume methane concentration would have been larger and consequently in better
agreement with the observed value.
CONCLUSIONS
The single HeNe laser has been shown to be a reliable remote sensing instrument for
the detection of methane gas in the atmosphere. Its usefulness as a methane leak detector,
however, is limited by ambient fluctuations in the background atmospheric concentration of
methane and local meteorological conditions. It is anticipated that the signal-to-noise ratio of
the HeNe laser system could have been improved significantly by employing two laser
wavelength differential measurement to minimize fluctuations in the ratioed laser return signals
due to water vapor, atmospheric aerosols, and atmospheric turbulence.
Good agreement between the cross-plume opticiil density of methane and the 2-D plume
dispersion model in this study suggests that the vertical dispersion coefficient of Gifford can be
linearly extrapolated to downwind distance less than 100 m. Alternatively, the HeNe laser may
be used to measure vertical dispersion scales in the atmosphere using methane as a tracer.
943
-------
REFERENCES
Csanady, G. T. (1973). Turbulent Diffusion in the Environment. D. Reidel
Publishing Co., Boston, pp. 66-68.
Gifford Jr., F. A. (1961). Use of routine meteorological observations for estimating
atmospheric dispersion, Nucl. Safety, 2: 47.
Grant, W. B., and R. T. Menzies, (1983). A survey of laser and selected optical systems
for remote measurement of pollutant gas concentrations, J. Air Poll. Control
Assoc., 33: 187.
Grant, W. B., (1986). He-Ne and cw CO2 laser long-path systems for gas detection,
Appl. Opt., 25; 709.
Kucerovsky, Z., E. Brannen, K. C. Paulekat, and D. G. Rumbold, (1973).
Characteristics of a laser system for atmospheric absorption and air pollution
experiments, J. Appl. Meteorol., 12: 1387.
Murray, E. (1978). Remote measurements of gases using differential-absorption lidar,
Opt. Engineer., 17: 30.
Patel, C. K. N., (1978) Laser detection of Pollution, Science, 202: 157.
Table 1. Component parameters for the HeNe laser system.
COMPONENT
HENE Laser
Power
Divergence
Beam Width
Chopper
Frequency
Receiver Telescope
Diameter
Focal Length
Detector
Diameter
D*
Retroreflector
Coating
PARAMETER
Spectra-Physics 127
6mW @ 3.392 urn
0.66 mrad
1.25 mm
110 Hz
Newtonian
22.86 cm
76cm
PbS
2 mm
1.3 x lO1^ mm Hz1/2 W"1
6.35 crn (dia.)
Au with MgF2 overcoat
944
-------
Table 2. Summary of controlled release experiments performed on the H. 0, Mahoney lease,
Wasson field, Texas. Q is the methane release rate, X is the perpendicular distance
between the release point and the laser beam, and II is the distance to the retroreflector.
The location of the transects is given in Figure 1.
Exp.
No.
1
2
3
4
5
6
7
8
9
10
11
12
15
Transect
A
A
A
A
A
B
B
C
c
C
c
c
c
Date/Time
May 23, 1989/1540 hrs
11
May 24, 1989/1600 hrs
11
H
May 23, 1989/2000 hrs
"
May 23, 1989/1300 hrs
M
IT
May 9, 1989/1700 hrs
May 9, 1989/1600 hrs
May 9, 1989/1630 hrs
Q
mg CH4/s
330
660
330
330
660
330
330
330
330
660
330
330
330
X
m
67
67
41
73
73
55
55
36
36
36
61
20
36
R
m
100
100
200
200
200
100
100
100
100
100
100
100
100
Table 3. A comparison of observed (Cj) and predicted (Cm) methane column densities for 11
experimental runs. Atmospheric stabilities according to Gifford (1961) are shown in
parentheses. Stability class A is the most unstable; D the most stable.
RUN
No.
1
2
3
4
5
6
8
10
11
12
15
Q
mg CH4/s
330
660
330
330
660
330
330
660
330
330
330
X
m
67
67
41
73
73
55
36
36
61
18
36
U
m/s
2.5
2.5
3.0
3.0
3.0
0.5
2.0
2.0
4.0
6.0
4.0
q
mg CH4/m2
25
49
48
23
70
92
34
40
32
55
39
Cm
mg CH4/m2
26 (B)
52 (B)
35 (B)
20 (B)
61 (C)
156 (B)
36 (A)
72 (A)
18 (B)
48 (D)
51 (C)
945
-------
REFLECTING TELESCOPE
MIRROR
PbS
{^DETECTOR;
RETRO-
RELECTOR
HENE LASER
A, = 3.392jim
CHOPPER
Figure 1. Schematic diagram of HENE breadboard unit. Operational parameters for
the laser, the Newtonian telescope, and the detector are listed in Table 1.
H.O, Mohoney
Lease
Yoakum Co,
Texas
Yoakum Co.
N:
O Lubbock
Pump
Station
c
Tank
Battery
800
Feet
Miles
A
Figure 2. Location of the H. O. Mahoney lease in west Texas. Measurements were conducted along
transects A, B, and C (see Table 2).
946
-------
CL
^
CL
UPWIND OF TANK BATTE:.RY (B)
MAY 23, 1987, 2000 MRS
RANGE: 100m
WIND: WEST, 0-0.5m/s
T= 1.0s
CD
cc
LU
-z.
LU
LU
LU
DC
90 -
80 -
UPWIND OF TANK BATTERY (A)
(b)
MAY 23,1987, 1400 MRS
RANGE: 100m
WIND: SOUTH, 1-2m/s
i= 1.0s
10 11 12 13 14 15
TIME (MIM)
Figure 3, Relative response of the HENE laser to ambient conditions along transects A
and B. Peak-to-peak fluctuations in the signal were 4% and 5% respectively, for
12 s averages.
UPWIND OF TANK BATTERY (B)
RANGE: 100m (330 ft)
DISTANCE (X): 55m (180 ft)
WIND: VARIABLE, <1m/s (<2mph)
DATE/TIME: 5/23/89, 2000 Mrs
CL
i_
CL
CC
LU
Z
LU
LU
LU
CC
100 -i
80 -
1500 FT3/d
1500 FT3/d
3000 FT3/d
WIND
SHIFT
BACKGROUND TRANSMISSION
10
30 40 50
TIME (MIN)
Figure 4, Relative response of the HENE laser to point releases of methane gas along transect B. The
amounts of gas released were 330 nig CHl4/s (1500 ft-Vd) and 660 mg ClL^s (3000 ft-Vd).
947
-------
EVALUATION OF COMMUNITY EXPOSURE TO AIRBORNE SARA
TITLE III SECTION 313 CHEMICALS EMITTED FROM
PETROLEUM REFINERIES
C. Herndon Williams. Walter L. Crow
Radian Corporation
P.O. Box 201088
Austin, TX 78720
Paul S. Lewandowski
American Petroleum Institute
1220 L Street, NW
Washington D.C. 20005
Radian Corporation conducted an ambient air monitoring study for the
American Petroleum Institute (API) to estimate the impact of petroleum
refinery emissions on air quality at the fenceline. Three refineries
were monitored over three consecutive days in the spring and early summer
of 1988. The target list of chemicals included twenty five (25) SARA
Title III Section 313 air toxics in several classes: hydrocarbons,
chlorinated hydrocarbons, acid vapors, ammonia and metals.
Eight sampling sites were located around each refinery to monitor
the air both upwind and downwind of the emission sources. Meteorological
measurements were made during the three days of sampling. Sampling
methods were based on a combination of filters, sorbents and evacuated
metal canisters. Analytical detection limits in the range 0.2 - 20
micrograms per cubic meter (/ig/m3) were achieved for the target
chemicals.
The elements of the study design will be presented. The methods for
sampling, analysis and data treatment will be reviewed and an overview of
the air monitoring results will be presented. An evaluation of the
monitoring methods will be given, as well as conclusions regarding the
potential for community exposure.
Experimental Methods
The target list of airborne chemicals to be measured consisted
primarily of twenty five (25) species reported by one or more of the
refineries under SARA Title HI Section 313. The target species would be
present in ambient air in the form of a gas, vapor or particulate matter.
No single sampling and analytical method was feasible so the target
chemicals were grouped according to their monitoring methodology. Table
I lists the target chemicals in four groups of primary target chemicals
(groups 1, 4, 6 and 7) and one group (group 9) of secondary target
chemicals.
948
-------
The four primary target chemical groups contain 25 SARA Title III
Section 313 compounds. Group 9 contains .A SARA chemicals and 11 non-
SARA indicator compounds and was monitored at only one third the
frequency of groups 1,4, 6 and 7.
The sampling and analytical approach for each target chemical group
is shown in Table II. The target limits of detection (LOD) apply to 24-
hour integrated air samples. Air sample volumes over 24-hours were
nominally 144 liter for all groups, except: 1440 liter for the group 7
metals and 15 liters for the group 9 canister samples.
Eight sampling sites were arranged around each refinery according to
the strategy shown in Figure 1. This strategy was designed to measure
ambient air concentrations at fixed distances in the range 1000-5000 feet
from the refinery center. Multiple sampling sites would help assure
upwind and downwind coverage even if the wind direction were variable
over 24-hours. Radian operated a portable meteorological station
throughout the three days of air sampling.
Field quality control consisted of a field duplicate set (sites 3
and 4) and a field blank (site 10) for each 24-hour sampling period. In
addition, filter and sorbent media were spiked with selected target
chemicals from groups 1, 3, 6, and 7. These spiked media were air
sampled and analyzed for recovery to help validate the sampling approach.
Air Monitoring Results
The ten target chemicals detected with the highest frequency around
each refinery are listed in Table III along with their measured
concentration range in ^g/m3. Aliphatic hydrocarbons detected in these
ambient air samples were summed into carbon number groups: C4 to C5, C6
to C8, and C8+.
Other than these aliphatic hydrocarbons, benzene, toluene and xylene
were the major target hydrocarbons found in this study, and sulfuric acid
was the only non-hydrocarbon found at all three refineries. Figure 2
summarizes the air concentration data for benzene, toluene, m,p-xylene
and ethylbenzene for the three days of monitoring at each of the three
refineries. The daily mean and maximum air concentrations over all 8
sampling sites is shown.
Conclusions
The sampling and analytical approach was demonstrated to be a cost
effective way to measure ambient air concentrations for several classes
of air toxic chemicals in a short term, ir.tensive monitoring campaign.
Benzene, toluene, and xylene were the target hydrocarbons seen with
the highest frequency around the three refineries studied, but they were
generally present at low concentrations, eg <1 to 23 jUg/™3 for benzene.
949
-------
TABLE I. THE API REFINERY STUDY: LIST OF PRIMARY AND
SECONDARY TARGET CHEMICALS
PRIMARY TARGET CHEMICALS
Group 1
Anthracene Ethyl Benzene Toluene
Benzene Methyl t-Butyl Ether 1,2,4-Trimethylbenzene
1,3-Butadiene Methyl Ethyl Ketone m-Xylene
Cumene Naphthalene o-Xylene
Cyclohexane p-Xylene
Group A
Carbon Bisulfide Ethylene Dibromide Ethylene Bichloride
Group 6
Ammonia Hydrogen Fluoride Sulfuric Acid
Group 7
Cadmium Chromium Nickel Selenium Vanadium
SECONDARY TARGET CHEMICALS
Group 9
Benzene Additional non-SARA
1,3-Butadiene Indicator Compounds:
Cumene t-Butyl Benzene
Cyclohexane n-Hezane
Ethylene Dibromide n-Heptane
Ethylene Dichloride Isohexane
Ethyl Benzene Isooctane
Ethylene Methylcyclohexane
Methyl t-Butyl Ether 3-Methylhexane
Methyl Ethyl Ketone 3-Methylpentane
Toluene n-Pentane
1.2,4-Trimethylbenzene Trichloroethylene
m,p-Xylene Tetrachloroethylene
o-Xylene
950
-------
TABLE II.
SAMPLING AND ANALYTICAL TECHNICAL APPROACH
co
Group
#
1
4
6
7
9
Chemicals
Hydrocarbons
in Group
and C-H-O
Electrophilic Compounds
Acids
Ammonia
Metals
Hydrocarbons
C-H-CI
. C-H-O,
Method
Source
NIOSH
NIOSH
NIOSH
NIOSH
NIOSH
Radian
Sampling
Medium
Fitter +
Charcoal tube
Charcoal tube
Silica ge! tube
Silica get tube
Filter
Canister
Analytical
Method
GC-FID
GC-ECD
!C
Color
ICAP
HRGC-MD3
Target
LOD
(ppbv)
2-20
1-5
5-20
2-10
0.5-5
0.5-1
a
MD = multiple detection = FID + PID + HSD.
-------
TABLE III. THE TEN CHEMICALS DETECTED WITH THE HIGHEST FREQUENCY IN THE
AMBIENT AIR AROUND EACH REFINERY
co
CJi
to
Refinery Chemical Name
1 C4 - C5 Hydrocarbons
Ammonia
C8+ Hydrocarbons
Sulfuric acid
C6 - C8 Hydrocarbons
Ethylene dichloride
Toluene
m.p-Xylene
Benzene
Chromium
2 Toluene
C4 - C5 Hydrocarbons
C6 - C8 Hydrocarbons
Benzene
m.p-Xylene
Sulfuric acid
o-Xylene
C8+ Hydrocarbons
1.2,4-Trimethylbenzene
Ethylbenzene
3 Toluene
C4 - C5 Hydrocarbons
C6 - C8 Hydrocarbons
Benzene
C8+ Hydrocarbons
Sulfuric acid
1.2. 4-Tr imethyl benzene
Ethylbenzene
m.p-Xylene
o-Xylene
Percent Frequency
Detected
89
67
56
44
44
33
22
11
11
7
96
96
93
89
81
41
30
26
22
19
100
100
100
70
63
63
52
48
24
14
Concent rat ion
Range (ug/m )
1
<4
1
<30
2
<1
<1
<1
<2
<0.3
<1
7.7
1
<1
<1
<20
<1
2
<1
<1
2
5
8.6
<1
1
<20
<1
. <1
<1
<1
- 120
- 9.9
- 23
- 56
- 39
- 44
- 3
- 4
- 3
- 0.5
- 42
- 130
- 350
- 19
- 33
- 130
- 11
- 84
- 6.4
- 9.0
- 23
- 190
- 470
- 23
- 40
- 640
- 4
_ C
- 19
- 6.9
-------
Upwind
< 1 ppb isooieth
Downwind
Prevailing Wind
Direction Range
10 • Field BianK
Diagram of the Proposed Placement of
Ten Sampling Sites Around the Refinery
FIGURE 1.
953
-------
FIGURE 2.
CD
Hydrocarbons Detected At High Frequency Daily Maximum
Concentration And Mean For Three Days
O>
c
o
i
I
o
0
-------
DEVELOPMENT OF A TECHNICAL APPROACH FOR DETERMINING THE
VOLATILIZATION RATE OF HYDROCARBONS FROM A LANDFILL USING BOTH
MODELING AND DIRECT EMISSION MEASUREMENT TECHNIQUES
Dr. C.E. Schmidt
Environmental Consultant
1479 Salmon Falls Road
Folsom, California
Mr. David Suder
ENVIRON Corporation
5820 Shellmound, Suite 700
Emeryville, CA 94608
A technical approach was developed to determine the
fugitive total non-methane hydrocarbon emissions during
landfilling operations. The landfill accepts soils contaminated
with hydrocarbon petroleum products.
The landfilling operation was simulated by a series of
waste handling operations that approximated the operation. This
was necessary in order to develop a technical approach that
could generate representative emission rates from the operation.
The operation included a 24-to-30 hour cycle during which soils
were delivered, unloaded, stock-piled, spread over the working
face of the landfill, and compacted. The challenge of the
program was to develop a testing and modeling strategy that
could generate emission data to represent this event.
The technical approach developed to collect these
emission data included a modeling exercise using the CHEMDAT6
models and field testing using thti EPA Surface Isolation
Emission Flux Chamber Technology. The testing strategy included
direct emissions measurement of eight different loads of
contaminated soil conducted over the active operation cycle of
the landfill. These time-weighted emissions data were then used
to estimate the total non-methane hydrocarbon emissions from the
landfill.
955
-------
Introduction
ENVIRON Corporation was contracted by a California
landfill site owner to prepare an Authority to Construct
application. The landfill is a Class I hazardous waste landfill
located at an existing treatment, storage, and disposal facility
(TSDF). In order to prepare the required non-methane
hydrocarbon (NMHC) emission estimates, ENVIRON used the RTI land
treatment model included in the EPA's CHEMDAT6 series of models.
In addition to the estimate of NMHC or reactive organic
compounds (ROC) emissions obtained by using descriptive
information from the landfill operation, total petroleum
hydrocarbon concentration data, and modeling, ENVIRON contracted
Radian Corporation to measure ROC emission rates. A protocol
was developed and employed to obtain ROC emission rates using
the EPA recommended emission isolation flux chamber.
Technical Approach
The basic approach used to study the emission's event
was to model emission rates using available emission models and
waste characterization data and to measure emission rates from
representative landfill waste. Both approaches for obtaining
emission rates have advantages and limitations. The advantage
of the modeling is that an emission rate estimate can be
obtained by imputing representative data into a specific model
using limited measurement data and resources. The disadvantage
of modeling is that the input data may not represent the
volatile potential of the waste material. In addition, models
provide estimates of dominant emission process and are not
expected to accurately represent all facets of the emissions
event, site conditions, and variables that have an affect on the
emissions event. As such, emission estimates derived by
modeling may be limited by discrepancies between the idealized
system considered in the model and actual site characteristics.
Direct emissions measurements provide the most accurate
and representative emission rate estimates with measurable
precision. However, as with all measurement approaches, data
must be collected that are representative of the emissions
event. This may require testing in many locations following a
labor intensive sampling schedule. Most testing is specific to
a waste material, site conditions, and often times environmental
conditions. Thus, measurements may be representative of a
particular waste and specific site conditions, but data from
this particular test may not be useful in understanding the
emission event or the time-dependent emission rate.
For this testing and modeling effort, the objective was
to use model input data and to conduct testing in order to
represent maximum daily ROC emissions. Model inputs were
selected to represent the typical material disposed of at the
landfill. In addition, other model input parameters were used,
including soil moisture and porosity that approximated average
site conditions. The field test included screening waste loads
and selecting materials for testing. Waste selected for
956
-------
measurement included: gasoline contaminated soils (5 tests);
diesel contaminated soils (1 test.) ; refinery waste soils (1
test); and crude oil contaminated soils (1 test). The incoming
loads of waste were screened by reviewing the waste manifest.
The mix of waste types selected for testing conservatively
represented the types of materials disposed of at the landfill.
The measurement event simulated the landfilling
operation by placing the waste in the landfill and spreading it
as is done under normal operation. Testing was conducted over
a 24-hour (or longer) period. Waste samples collected from the
selected loads were analyzed for total petroleum hydrocarbon
(TPH), porosity, and moisture content.
Direct Measurement
The sampling procedure employed was the surface
isolation emission flux chamber technique1. Real-time VOC
emission screening was performed during each isolation flux
chamber measurement using a Foxboro Corporation Organic Vapor
Analyzer (OVA) Model 108 calibrated with methane-in-air gas
standards. Following placement of "he flux chamber, the OVA was
attached to the sampling line of the chamber. After steady-
state conditions were realized in the chamber, samples were
collected by syringe and analyzed by the on-site gas
chromatograph. In addition, samples were collected in stainless
steel canisters and analyzed off-site by gas chromatography.
The results were reported as total non-methane hydrocarbon
[parts per million by volume of carbon (ppmv-C)].
Results
The collection of physical and chemical waste
characteristic data coinciding with the flux chamber
measurements provides a small but informative data base which
can be used to compare emission rates estimated by the RTI land
treatment model contained in CHEMDAT6 to flux chamber
measurements. Of the eight loads tested, five provided
meaningful comparisons. Some or all of the input data required
for the model were unavailable for the remaining loads.
Using the CHEMDAT6 model with load-specific chemical and
physical data, the time sequence of NMHC flux was estimated for
each of the loads for which appropriate input data were
available. Figure 1 provides sample CHEMDAT6 model input and
output. Figure 2 illustrates the measured emission rates for a
load of soil contaminated with gasoline at various times
following initial exposure of the waste, as well as emission
rates estimated by the model where appropriate. From the eight
tests, it was seen that agreement between the model and
measurements was good in some cases, but very poor in others.
It may be that, in cases of poor agreement, most of the volatile
constituents had volatilized before the measurements began. If
that occurred, the characterisation of the remaining
hydrocarbons as gasoline would have produced significant
overestimation of the emission rates, relative to measured
values.
957
-------
Figure 3 is a scatter plot of measured vs. estimated
NMHC emissions for the ten hours following placement of waste in
the landfill. Data for five of the eight loads are plotted.
The diagonal line has a slope of one, representing agreement
between the methods. Load 2 (diesel-contaminated soil) was the
only load for which the measured emissions over the first 10
hours exceeded emission estimates using the model. For the
remaining loads, measured emissions appeared to be well
correlated to estimated emissions, but significantly lower.
The average measured total emissions from the testing of
materials over the first ten hours was 2. IxlCf3 g/cm2. This
value is considerably lower than the emissions estimated a
priori for the period of daily operations of the proposed
landfill using the model. The average emission rate for the
modeled test cases was 6.lxlO~3 g/cm2.
Conclusions
This study is unique in that two very different
emissions estimation approaches were used to estimate emission
rate data representing ROC emissions from the same source, an
active landfill. Both approaches were used properly and
generated emission estimates that, in some cases, compared well
with each other. Either approach used properly and independent
of the other would have provided an acceptable estimate of the
ROC emission rate.
The following conclusions are offered:
1) The CHEMDAT6, RTI land treatment model has
application for estimating air emissions from
hazardous waste landfills, but may over estimate
emissions substantially.
2) Emission rate data can be obtained using modeling
equations, provided that the model represents the
dominant emissions event and input data are
realistic.
3) The EPA recommended emissions isolation flux chamber
has application for measuring emissions from
landfills.
4) Direct emission measurement approaches provide
acceptable data so long as the techniques are
properly executed and the testing strategy
represents the emissions event.
5) Average emissions data from measurements over
appropriate time periods and for representative
waste can be used to describe emissions over larger
cycles (yearly emissions) and for a variety of waste
materials.
6) Indicator parameters, like TPH in solids, may not
provide an accurate representation of volatile
characteristics of waste materials.
958
-------
References
1.
"Measurement of Gaseous Emission Rates From Land
Surfaces Using An Emission Isolation Flux Chamber -
Users Guide". EPA Contract No. 68-02-3889, Work
Assignment 18, EMSL, Las Vegas, December, 1985.
LAND TREATMENT MODEL DATA
(open landfill, waste pile)
L,Loading (g oil/cc soil) o.oo:i72
Enter ci x 10A6 VO ppmw 210000
1,Depth (cm) 67
Total porosity 0.4
Air Porosity (defaulted) 0.:>65
MW oil :.90
VO diss. in water, enter 1 0
Time of calc. (days) 0.25
Biodegradation, enter 1 0
Temperature (Deg. C) 17
Wind Speed (m/s) :i.2
Area (m2) 2:.00
COMPOUND NAME
LANDTREATMENT EMISSION RATES (g/c»2-s)
TIME (hours)
0.25 1 4 12 48
5.52E-07 I1.76E-07 1.38E-07 7.98E-08 3.99E-08
GASOLINE
COMPOUND NAME
GASOLINE
LANDTREATMENT
FRACTION LOST
AIR BIOL.
INTERMEDIATE TIME
0.25 days
AIR BIOL.
1.000
0.000
0.093
0.000
Figure 1. Sample CHEMDAT6 Input and Output Data
959
-------
35
30
25
20
15
10
5
0
GASOLINE IN SOIL MATRIX
Estimated vs. Measured NMHC Emissions
NMHC Emissions (g/cm**2/s) (1E-8)
500 1000 1500 2000
Minutes After Placement
2500
3000
Measured Values
CHEMDAT6 Values
Figure 2. plot Showing Estimated vs. Measured NMHC Emissions
TEN-HOUR EMISSIONS
Estimated vs. Measured NMHC Emissions
Measured 10-Hour Emis. (g/cm**2) (1E-5)
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
Predicted 10-Hour Emis. (g/cm**2) (1E-5)
Figure 3. Scatter Plot Showing Correlation Between 10-hour
Estimated vs. Measured NMHC Emissions
960
-------
TEMPERATURE-CORRECTED DISPERSION MODELING OF
VOLATILE EMISSIONS FROM HAZARDOUS WASTE SITES
Richard J. Hardy and Damon D. Judd
Morrison Knudsen Corporation
Environmental Services Group
Boise, Idaho
Richard L. Graw
Morrison Knudsen Corporation
Environmental Services Group
Denver, Colorado
A modification to the Industrial Source Complex-Short Term (ISCST)
model is proposed to provide the option of correcting volatilization source
terms, on an hourly basis, in proportion to the variation in vapor pressure
that results from the diurnal temperature cycle. Measurements are
presented to demonstrate that air temperature, a standard meteorological
input parameter, is a suitable surrogate for surface soil temperature, a
more direct determinant of volatilization from the surface. The derivation
of the proposed correction term is described, and results from test runs of
the temperature-corrected and the standard model are compared. Diurnal
variations in relative flux, plume dilution, and predicted ambient
concentrations are presented. The effect of this correction on annual
concentration estimates is shown for five compounds.
Annual concentration estimates are 200-300% lower when the hourly
correction option is selected than when a constant flux, determined for a
reference temperature of 25°C is us;ed. In addition, maximum hourly
concentrations, predicted to occur (contrary to popular belief) during
nighttime stable conditions, were 50% Lower with the temperature-corrected
model than with the standard version.
961
-------
Introduction
Air pathway assessment of volatile organic emissions from hazardous
waste sites involves two distinct steps: 1) determining the volatilization
flux and 2) dispersion modeling to determine exposures to people living off-
site. Emission fluxes are commonly either estimated by one of several
exposed-waste volatilization models or measured. In either case, the source
term is often determined at daily maximum temperatures and then assumed to
be constant for the entire 24-hour day (and often, for the entire year).
However, by examination of the models recommended for estimating
volatilization from waste surfaces1, it can be seen that volatilization flux
varies in proportion to the saturated vapor pressure, which in turn varies
with temperature. Thus, for example, for a typical summertime, diurnal
temperature range of 15-33°C, (Boise, Idaho in July2) , the vapor pressure
of toluene, a moderately volatile compound, ranges from 16 mm Hg at night
to 39 mm Hg during midday, a 240% increase, and the volatilization flux is
expected to increase proportionately.
Atmospheric dilution also varies throughout the day, however. The
gaussian plume equation for a ground-level point source3,
C - Q (1)
T U Oy
-------
relationship between air and surface soil (or waste) temperatures,
temperature sensors were installed 0,1 cm below the surface of a sandy/loam
soil and in a shaded shelter one meter above ground level. The time
history of hourly-averaged air and soil temperatures is presented in Figure
1. Surface soil temperatures are highly correlated with air temperatures
during nighttime hours but become significantly warmer (5-20°C) during all
but: one midday solar heating period. The surface temperature cooled to
below the air temperature on the sixth day after the soil was wetted.
Therefore, at least during the critical nighttime periods of reduced mixing,
this limited data set suggests that the t.mbient air temperature is a good
surrogate for the surface soil or volatilization temperature of the organic
waste under dry conditions and is conservatively cooler when the soil or
waste is wet or rapidly volatilizing.
Temperature Correction Algorithm
The temperature dependence of saturated vapor pressure is shown, by
the well known Glausius-Clapeyron equation, to be a function of the heat of
vaporization. A wide variety of vapor pressure estimation methods have
been derived by integration of the Clausius-Clapeyron equation. One of the
simplest methods, with coefficients widely available, is a two-parameter
version reported by Schlessinger4;
log,0 P - (-0.2185 A/T) + B (2)
where P is the vapor pressure in Torrs, T is the temperature in degrees
Kelvin, A (the "A coefficient") is the molar heat of vaporization in
calories per gram mole, and B is a constant. The correction factor for
scaling the source term in proportion to the relative vapor pressure can
be written by solving Equation 2 for the vapor pressures at the ambient
(Pamb) and reference (Pfef) temperatures (Tamb and Tre), respectively) and
substituting both into the ratio Pamb/Pref. The final temperature correction
factor, TCORR is thus obtained;
TCORR - *""» = 10l"'— " V^ref -/-«nb/J (3)
where the correction term requires only the reference and ambient
temperatures and the molar heat of vaporization for the compound of
interest.
The ISCST model was modified, by addition of two self-contained
subroutines, to allow optional interactive selection of the compound or
vapor pressure of choice from a menu listing compounds most frequently
found at Superfund sites. The reference temperature, at which the base
emission rate was measured or calculated, is entered by the user, and
ambient air temperatures are read from the sequential hourly meteorological
file as usual.
Dispersion Modeling
The modified program was exercised for five compounds over a range of
vapor pressures to show the influence of this modification on model
estimates. One year of sequential hourly meteorological data from Denver
Stapleton Airport was used. In all cases, a unit emission flux (1.0
ug/m2-s) was assumed to emanate from a 10 n. by 10 m ground-level source area
963
-------
at a reference temperature of 25°C. Concentrations were computed at 100
ra from the source edge for every 2-degree wind direction interval from 0 to
359 degrees.
Results
The model was initially exercised without the temperature correction.
The maximum 24-hour concentration occurred on a fall day (October 25) with
11 hours of F stability and an average temperature of 7°C. When the
temperature correction option was selected, the maximum 24-hour
concentration occurred, more realistically, on a summer day (June 6th), with
an average daily temperature of 18°C and 6 hours of F stability. The
diurnal temperature variation and relative plume dilution variation for
June 6 are shown in Figures 2a and 2b, respectively.
Three alternative source-term, temperature-correction assumptions for
toluene are depicted in Figure 2c. Curve A represents the relatively common
practice of estimating (or measuring) emissions at the daily maximum
temperature and assuming that emissions are constant at that rate throughout
the day. A somewhat more realistic option is to compute emission flux at
the daily average temperature and keep it constant throughout the day, as
indicated by curve B. Finally, by selecting the proposed hourly temperature
correction option, hourly air temperatures are used to most realistically
compute emissions for each hour (curve C).
As occurs on most days, the atmospheric stability on June 6 ranged
from F at night to A just after noon, resulting in model-predicted relative
dilution (at 100 m downwind) that is 39 times greater at midday than at
night, as shown in Figure 2b. When the emission scenarios for the three
described temperature correction alternatives are modeled with ISCST, the
resulting ambient toluene concentrations, shown in Figure 2d, are obtained.
Remarkably similar curves are reported by Harper, et. al.5 for pesticide
volatilization. Although these three source treatments result in comparable
concentrations during midday, fairly large differences occur at night, with
the hourly-corrected concentration reaching only 50% of the "daily maximum
temperature" case and 70% of the "daily average temperature" case.
It should be remembered that the measured surface soil temperatures
ranged from 10°C cooler, with wet soils, to nearly 20°C warmer than the
air temperature during midday. If the hourly corrected air temperature is
increased by 20°C to simulate hot, dry surface soil temperatures, the
ambient concentration increases insignificantly, as indicated by the single
data point shown at 1300 on Figure 2d.
Perhaps the most striking conclusion to be drawn from Figure 2d is
that the maximum hourly concentration occurs during the night and is 16
times greater than the maximum midday concentration. This result
demonstrates that, contrary to popular belief, "high event" or worst-case
conditions for air impacts from volatilization sources are during clear-
sky evenings or nights when the atmosphere is most stable and temperatures
are moderate, rather than during sunny, warm, midday conditions. This
conclusion is supported by the fairly common observation that odor
complaints at landfills, waste sites, and wastewater treatment facilities
greatly increase during the evening and early morning hours.
964
-------
Annual concentrations for the standard and temperature-corrected
versions for five compounds are shown in Tatle I. Concentrations predicted
by the corrected model are 200-300% lower than concentrations estimated
using the standard (un-corrected) ISCST model. The greatest effect (3,00%
lower) is seen for the lowest-volatility compound. In addition, the
greatest effect is expected for continental climates with large diurnal
temperature ranges, with the utility of this approach diminishing for
coastal areas with small diurnal temperature ranges,
Conclusions
On the basis of limited air/surface: soil temperature measurements and
a theoretically based correction to the IS3ST dispersion model, it may be
concluded that 1) air temperatures serve £.s a good surrogate for surface
soil temperatures when critical, low-dilution conditions occur; 2) annual
concentrations are 200-300X lower, for Denver airport meteorology, when the
hourly correction option is selected; and 3) the temperature-corrected model
appears to provide more realistic diurnal and seasonal variation in ambient
concentrations resulting from surface volatilization sources.
Finally, we would recommend that reference temperatures always be
reported along with calculated or measured emission estimates. In addition,
this analysis demonstrates that nighttime conditions should not be neglected
when conducting "high-event" sampling or modeling for Air Pathway Assessment
of surface-contaminated waste sites.
References
1. "Air/Superfund National Technical Guidance Series, Volume II
Estimation of Baseline Air Emissions at Superfund Sites," EPA-450/1-
89-002, U. S. Environmental Protection Agency, RTF, NC. 1989.
2. "Climatic Atlas of the United States," NOAA, U.S. Dept. of Commerce,
Asheville, NC, 1979.
3. D. B. Turner, "Workbook of Atmo5;pher:.c Dispersion Estimates," U. S.
Environmental Protection Agency, RTF, NC, Revised 1970.
4. "Handbook of Chemistry and Physics,' 51st ed, The Chemical Rubber
Company, Cleveland, OH, 1970, pp. D-146 -D-165.
5. L. A. Harper, A. W. White, R. R. Bruce, A. W. Thomas, R. A. Leonard,
"Diurnal trifluralin volatilization flux," J. Environ. Qual. 5: 236
(1976).
965
-------
Table I Annual Modeling Results
Denver Airport Meteorology
(base flux - 1.0 Mg/m2-s @ 25°C for all cases)
Compound
No correction
1,3-Butadiene
Chloroform
Toluene
Dichlorobenzene
2-Hexanol
Heat of
Vaporization
(cal/gm-mole')
N/A
5,688.2
7,500.5
8,580.5
10,446.8
. 12,386.5
Vapor Pressure
at 25°C
(mm Hg)
N/A
1,848.7
172.0
26.8
2.3
2.5
Maximum
Annual
Concentration
0.0393
0.0221
0.0189
0.0173
0.0153
0.0133
Time History: Air and Soil Temperatures
SOIL TEMPERATURES AT 0.1 CM
M 6A N 6P M 6A N 6P M 6A N 6P M 6A N 6P M 6A N 6P M 6A N 6P M 6A N
TIME OF DAY, (MST)
Figure 1. Time history of air and surface soil temperatures, April 10 - 16.
966
-------
co
o>
O
O
Air Temperature
DAY 157
(2a)
Air
Temperature
HOUR OF THE DAY (MST)
Alternative Source Term Treatments
DAY 157
Curve A (at Max T)
3 ' Curve B
x
Z) 0 '
_>
u_
UJ 0 5
(2c)
Curve C
<" ' (w/hourly
o i Correction)
Vapor Flux
Relative to
ug/m2-s @25°C
HOUR OF THE DAY (MST)
O
z
2
O
Relative Plume Dilution
DAY 157
»\
/ 1
Relative
Dilution
<2b)
HOUR OF THE DAY (MST)
Maximum Hourly Predicted Concentration
DAY 157
o "
hj 0
I 0
o „
At Max T
(2d)
Model-predicted
Ambient
Concentration
- \ \^, At Avg T
^ \ V '-,
Hourly ^
HOUR OF THE DAY (MST)
Figure 2. Diurnal variation In (2a) air temperature, °C; (2b) plume dilution, relative to lowest dilution
observed (- 1) at 0100 hrs; (2c) source volatilization flux, based on 1.0 jig/m2-s toluene flux at 25'C
and assuming three alternative temperature correction approaches, |*g/ni2-s; and (2d) ambient air
concentration (/ig/m3) , resulting from three temperature correction alternatives shown in (2c), June 6,
1964, Denver Colorado Airport Meteorology. The data point (•) at 1300 hours (0.13/jg/m3) in figure 2(d)
represents the concentration predicted at the midday air temperature (25°C) plus 20°C.
-------
THE EFFECT OF WOOD FINISHING PRODUCTS
ON INDOOR AIR QUALITY
Bruce A. Tichenor, Leslie E. Sparks, and Merrill D. Jackson
U. S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
Zhishi Guo and Susan A, Rasor
Acurex Corporation
P.O. Box 13109
Research Triangle Park, NC 27709
The use of stains, polyurethane finishes, and waxes on interior surfaces
can cause elevated levels of vapor-phase organics indoors. Studies were
conducted in an IAQ (Indoor Air Quality) test house to determine the magnitude
and temporal history of indoor concentrations of organic compounds due to the
use of these products. An oak floor was constructed and stained. The
concentrations of total organics and certain individual compounds were
measured at two locations in the house for a period of about 2 weeks. After
10 days, including 2 days of house ventilation, the same procedure was
followed for a polyurethane finish. Finally, a liquid wood-floor wax was used
and similar measurements conducted. During each test period, house air
exchange rates were measured using a CO tracer. The measured concentrations
of organics were compared to results obtained from an IAQ model (INDOOR): good
agreement was obtained. The model used emission rate data developed from
small chamber tests of the same products and the measured air exchange rates.
The model also included an evaluation of adsorption to and re-emission from
interior "sink" surfaces. The studies showed that re-emission from the sinks
was the dominant factor affecting the indoor concentrations from 2 days after
application to the end of the test period.
968
-------
Introduction
A major objective of the EPA's IAQ program at the Air and Energy
Engineering Research Laboratory is to determine the impact of sources on IAQ.
An integrated research program involving chamber investigations, modeling, and
test house studies has been developed to mee" this objective. The program
involves the development of emission factors, research to develop models to
analyze the impact of the emissions on IAQ, and studies in an IAQ test house
to verify the model predictions. This paper describes how this three-phase
approach was used to evaluate the impact of three wood finishing products
(wood stain, polyurethane, liquid wood-floor wax) on IAQ.
Experimental Methods
Small Chamber Studies
Methods for determining emission characteristics of a wide variety of
indoor sources using small test chambers have been developed. A known
quantity of clean air is passed through a chamber containing the indoor source
to be evaluated. The chamber outlet concentration of vapor-phase organics is
measured over time using appropriate sampling' techniques followed by gas
chromatographic analysis. Emissions of both total measured organics and
individual compounds are reported. The resulting concentration vs. time curve
is then analyzed to determine the emission re.te for the source. The emission
rate is generally reported in the form:
ER = ER e"h (1)
0
where ER is the emission rate at time t, ER is the initial emission rate, and
k is the decay constant. The three products discussed herein were tested in
this manner.
IAQ Test House Studies
EPA operates an IAQ test house that is used for experiments to provide
measurements of the effect on IAQ of sources in full-scale indoor
environments. These measurements are used to validate the small chamber
emission rates and to verify an IAQ model (see below). The floor plan for the
test house is shown in Figure 1.
This paper reports on a set of experiments designed to determine the
effects pf three wood finishing products on IAQ. An oak floor (2.44m x 2.44m
= 5.95 m') was constructed and placed in the test house dining area (see
Figure 1). In the first experiment, a wood stain was applied to the floor,
a:id the concentrations of total organics and certain individual compounds were
measured at two locations in the house (living room and corner bedroom) for a
period of about 2 weeks. Samples were collected on Tena*/charcoal sorbent
tubes. The tubes were thermally desorbed and analyzed by a gas chromatograph
equipped with a flame ionization detector. Each test period was followed by a
waiting period of about 10 days, including 2 lays for house ventilation. The
same procedure was followed for a polyurethane finish. Finally, a liquid
wood-floor wax was used and similar measurements conducted.
During each test period, house air exchange rates in air changes per
hour (ACH) were measured using carbon monoxide (CO) as a tracer. The CO was
released in the hall, near the return air ven:, and monitored in the hall and dt
969
-------
IAQ Model
An IAQ computer model, INDOOR, has been developed to calculate indoor
pollutant concentrations based on source emission rates, room-to-ropm air
movement, air exchange with the outdoors, and indoor sink behavior.
Source Emission Rates. The emission rates of solvent based sources
like those used in the test house experiments include an application phase and
a decay phase. INDOOR uses the following expression to calculate the total
emissions:
ERtot = ERapp + EV"kt (2)
where ER, . is the total emission rate and ER is the emission rate during the
application phase, t. .
spp
Sink Adsorption and Re-emission. Research has shown that sinks (i.e.,
surfaces that adsorb and re-emit indoor pollutants) play a major role in
determining indoor pollutant concentrations. INDOOR uses the following
model to describe the behavior of indoor sinks:
dM/dt = kaCtA - kd'(Mt)nA (3)
where dM/dt is the rate of change of mass in the sink, k is the adsorption
rate constant, Ct is the indoor pollutant concentration at time t, A is the
area of the sink, k, is the desorption (re-emission) rate constant, M, is the
mass in the sink at time t, and n is an empirical constant.
Results
Figures 2, 3, and 4 present the results of the three test house
experiments. Each figure shows the concentration of total organics measured
in the living room of the test house over the course of the experiment. The
figures also provide two model predictions based on the input data shown in
Table I. The emission rate data in Table I are based on small chamber studies
of the three products.
Table I. Input Values for IAQ Model Predictions
Variable Wood Stain Polyurethane Floor Wax
E,.. (mg/mj-hr)
ipp
Sp (hr}
ER0 (mg/m2-hr)
k (hr"1)
ACH (hr'1)
12,000
0.4
17,000
0.4
0.3
12,500
1.5
5,000
0.25
0.36
58,000
0.1
10,000
1.0
0.4
970
-------
The "No Sink" predictions were made by iissuming that the interior
surfaces of the test house did not adsorb or re-emit organic vapors. The IAQ
model predictions assumed that the interior surfaces behaved as described by
equation (3), with k, = 0.2, k, = 0.003, and :n = 1.5.
a Q
The figures show that concentrations of total vapor-phase organic
compounds much greater than 100 rag/m were obtained when the three high
solvent content wood finishing products were used indoors. The figures also
clearly indicate the effect of indoor sinks on the time history of the
concentration of vapor-phase organics. The "No Sink" predictions show that,
in the absence of sinks, the concentrations would drop to background levels
within 2 days of using any of the wood finishing products that were evaluated.
The test house data and the IAQ model predictions provide evidence that
adsorption to and re-emissions from the sinks in the test house impact the
indoor concentrations for up to 2 weeks after product usage.
In general, the IAQ model predictions fit the data for all runs quite
well. Some of the lack of fit between the model and the data is due to
varying air exchange rates over the course of the experiment caused by
changing weather conditions, while the model uses a constant air exchange
rate. The model is being modified to allow the air exchange rate to be
varied. The fit could also be improved by using different sink constants for
the different sources and sink materials. However, until knowledge of and
data on the behavior of sinks are improved, this modification is not possible.
Conclusions
A three-phase approach involving small chamber studies, IAQ test house
experiments, and IAQ modeling was used to investigate the effect of wood
finishing products on indoor concentrations of vapor-phase organic compounds.
The experiments showed that indoor concentrations in excess of 100 mg/m of
total vapor-phase organics can occur after the use of such products. The
importance of indoor sinks on the time history of the indoor levels of
organics was demonstrated.
References
1. B. A. Tichenor, Indoor Air Sources: Using Small Environmental Test
Chambers _to _Charac_ter_ijze__Qrganj._c_Emis sions^ from Indoor Materials_ and
Products, EPA Report EPA-600/8-89-074 (NTIS PB90-110131}, Research
Triangle Park, NC, August 1989.
2. B. A. Tichenor and Z. Guo, "The effect of ventilation on emission rates of
wood finishing materials," Healthy Buildings '88, Vol. 3, pp. 423-432,
Swedish Council for Building Research, Stockholm, 1988.
3. L. E. Sparks, Indoor Air Quality Model Version 1.0. EPA Report EPA-600/8-
88-097a (NTIS PB89-133607}, Research Triangle Park, NC, September 1988.
4. B. Berglund, I. Johansson, and T. Lindvall, "Adsorption and desorption of
organic compounds in indoor materials," Healthy Buildings '88, Vol. 3, pp.
299-309, Swedish Council for Building Research, Stockholm, 1988.
971
-------
i
^v.
S
Master
bedroom
^* Bi
Clos
Cla
•^
Clos
5
""an
Corner
bedroom
too.
/ ^^^
i
•)
Master
bath
a
•
mm
mm
i
i
Bath
/
Return p
n
oir
X
X
H
\ ^•'J Clos
^*
Clos
"S
**
A
Utility
Middle
bedroom
fl
M
•
mm
A
a
Den
Kitchen
-
am
, ,. t ,,.._,
i
M
"
L.JL,
Wood floor /
,\
Livi
Clos
ng room
a
MM
an*
Instruments
Garage
n = Organic Sampling Point A = CO Sampling Point
Figure 1 — IAQ Test House
1000
100
IAQ Model Prediction
"No Sink" Prediction
en
CO
y
'c
D
O
o
I-
100
200
300
400
500
Time (hrs)
Figure 2 — IAQ Test House Experiment: Wood Stain
972
-------
1000
100
IAQ Model Prediction
"No Sink" Prediction
en
Jl
eo
o
'c
en
6
~o
-M
O
0
100
200 300
Time (hns)
400
500
Figure 3 — IAQ Test House Experiment: Polyurethane
1000
E 100
IAQ Model Prediction
"No Sink" Prediction
en
E
o
'c
0
o
"D
"o
50
100
150 200
Time (hrs)
250
300
350
Figure 4 — IAQ Test House Experiment: Wood Floor Wax
973
-------
INDOOR AIR TESTING FOR LOW-LEVEL VOLATILE ORGANICS:
A SITE-SPECIFIC TECHNICAL APPROACH
Amy S. Johnson, Jeff B. Hicks, and Eric D. Winegar
Radian Corporation
Sacramento, California
Migration of subsurface contamination from industrial operations into
neighboring residential structures raises concerns about possible human
exposures and creates a demand for new technical approaches to assess
exposure potential. This paper compares and contrasts the results from
indoor air studies at two different sites to emphasize the importance of
developing a site-specific technical approach to provide scientifically
defensible and cost-effective assessments of exposure potential for the
site, taking into account such factors as subsurface conditions, types and
levels of contamination, other sources of contamination, and
characteristics of the residential structures.
A technical approach is described that was used successfully to
evaluate human exposures to potential subsurface contamination consisting
of low levels of volatile organic compounds found in common household
products and urban ambient air. The approach involved indoor air and
outdoor ambient sampling in the study neighborhood, along with indoor and
outdoor sampling in a control neighborhood where no subsurface
contamination was present. State-of-the-art sampling and analytical
techniques were used to measure part-per-billion levels of the target
compounds. A comprehensive quality assurance/quality control program,
including detection limit studies, was conducted to establish limits of
certainty critical to data interpretation. The results of the study
underscored the importance of this site-specific approach: Nearly all
measured concentrations were less than 100 parts per billion, and the
concentration differences were very low - often negligible - between indoor
and outdoor air, and between the study and control neighborhoods.
Results from a second site are considered to demonstrate a different
technical approach. In this case, the subsurface contamination consisted
of indicator hydrocarbons at relatively high levels, with much higher
concentrations measured indoors than in the outdoor ambient air or the
control area. Comparison of the results from the two sites suggests that
unique study design criteria are often dictated by relatively subtle
differences in site factors.
974
-------
Introduction
Migration of subsurface chemicals from industrial operations into
neighboring residential structures can present the potential for human
exposures to indoor airborne chemicals. This paper presents the results of
exposure assessments conducted in two residential neighborhoods, each
located adjacent to an industrial facility where subsurface chemicals had
been measured and could potentially impact indoor air quality in the
residential structures. Although the sites and study objectives were quite
similar, different site-specific technical approaches were required to
provide technically defensible, conclusive data in a cost-effective manner.
The following discussion demonstrates how these technical approaches
evolved in response to preliminary findings, regulatory agency
requirements, community concerns, and the industrial facilities'
philosophies.
Background
Site 1 was an ocean-front residential neighborhood bordered on one
side by a large refinery. Subsurface contamination consisted of liquid
petroleum hydrocarbon product that had leaked from several large storage
tanks, creating a floating lens on the ground-water table. Volatile
components had been measured in the shallow soil vapor in the neighborhood
at levels up to 10,000 ppmv total hydrocarbons (THC) in one area. Types
and relative concentrations of hydrocarbons were characteristic of an
intermediate petroleum product; predominant species included C5 and C6
aliphatics, benzene, toluene, xylene, and various olefins. The outdoor
ambient air was expected to reflect a significant ocean influence, since
on-shore winds prevailed. Odors had been observed at likely points of
subsurface vapor infiltration in some structures, and these were reported
to the health agencies. Real-time screening with organic vapor analyzers
equipped with flame ionization detectors (OVA-FIDs) in selected structures
near the higher soil gas contours indicated high point sources (5000 to
10,000 ppmv THC) at likely points of infiltration, such as floor-wall
joints and cracks or holes in floors or walls. Elevated THC levels (up to
100 ppmv) were also measured in the centers of rooms at approximately five-
foot heights in some structures (referred to as "room ambient" locations).
The neighborhood was located in a very desirable area of expensive
homes. Limited community organization or reaction was exhibited during the
investigations. The regulatory agencies adopted a cooperative, "hands-off"
approach, and much of the oversight was handled by local authorities during
the investigation phases. Priorities shared by the refinery management and
the regulatory agencies were expeditious completion of the investigative
phases and prompt mitigation if needed.
Site 2 was a residential neighborhood located adjacent to a very
large chemical manufacturing facility. The facility manufactures and uses
a wide variety of common solvents, many of which had been measured in
ground water on site. These compounds had not been measured in shallow
ground water at the facility's fenceline; the highest concentrations were
measured in deeper ground water. Soil vapor testing conducted in early
phases of the investigation did not indicate a plume of chemicals in the
soil gas off site. Outdoor air was influenced by typical urban sources and
by certain high-use solvents from the facility. Dispersion modelling
indicated that measurable concentrations (i.e., >0.010 ppmv) of some
solvents could be present in outdoor air.
The manufacturing facility was the primary employer in the area, and
consistent with their longstanding "good neighbor" image, maintained a high
975
-------
level of community involvement during the investigations. The community
reacted to the subsurface findings with significant concern, and several
existing and new groups represented the public interest. These groups were
highly involved throughout the studies; representatives and hired
consultants reviewed workplans and reports and attended regular meetings
with the public agencies. The regulatory agencies also were highly
involved, maintaining day-to-day oversight during all field activities and
providing detailed input on the technical approaches and experimental
protocols. The agencies required a strict comparison approach rather than
design based on chemical toxicity and placed a strong emphasis on the
completeness and thoroughness of the investigations.
Site 1 Approach and Results
The purpose of the indoor testing program for Site 1 was to determine
whether subsurface compounds had entered structures via the surrounding
soils. The indoor testing program for Site 1 was initiated with expansion
of the real-time instrument surveys to include many structures in the
neighborhood. In total, 313 structures were surveyed over a period of
roughly six months. The real-time survey results, especially any elevated
room ambient or point sources at likely infiltration points, were used to
prioritize structures for testing using additional methods. With input
from the regulatory agencies, additional testing was conducted in
structures near the subsurface plumes or where elevated real-time reading
were obtained to help identify any structures requiring mitigation
measures. Charcoal tube sampling over eight-hour periods was conducted to
test for C4 through C12 hydrocarbons at point source and room ambient
locations. Evacuated stainless steel canister grab samples also were
collected at point source and room ambient locations and were analyzed
using gas chromatography with multiple detectors (flame ionization,
photoionization, and Hall electrolytic conductivity) to speciate a wide
variety of compounds and to help identify types and ratios of compounds in
structures for comparison to soil vapor and ground-water sampling data.
Limited sampling was conducted using similar techniques in a control
structure located in an area removed from the subsurface contamination.
Results from the indoor testing ranged from very low concentrations
similar to baseline levels (determined from trip blank samples) and control
structure levels to significantly higher levels of compounds similar in
type, level, and ratios to those measured in the subsurface sampling.
Likely presence of subsurface compounds in some structures was indicated
by: elevated real-time THC readings at room ambient or suspect point
source locations; elevated THC or total non-methane hydrocarbon (TNMHC)
levels in charcoal tubes, with dominant C5 and C6 species at relative
concentrations characteristic of the subsurface mixture; and canister
sample species common to the subsurface at levels above baseline and in
similar ratios to those in subsurface samples. Results for certain
structures appeared to be very similar, suggesting various categories of
data. Data from structures in the "Level 1" category typically included
real-time readings above 30 ppmv THC at room ambient readings, elevated
hydrocarbon point sources, subsurface-type hydrocarbons at levels well
above baseline concentrations, and tracer species at ratios common to the
subsurface plume. Data from structures in the "Level 2" category included
real-time readings between 10 and 30 ppmv at room ambient locations and
elevated point sources or subsurface-type hydrocarbons. Structures not in
the Level 1 or Level 2 categories typically did not exhibit elevated room
ambient or point source measurements, and the types and levels of any
hydrocarbons measured above baseline levels were not similar to those in
the subsurface plume. Of the 313 structures, 4 fell into the Level 1
category, and 5 fell into the Level 2 category. Average concentrations of
976
-------
some key species are presented In Table I, where the relationship between
the types and levels of subsurface compounds and those measured in the
indoor samples is demonstrated.
Level 1 structures were prioritized for mitigation. Short-term
mitigation measures were developed in coordination with the regulatory
agencies prior to startup of a vapor recovery system, which served as the
long-term mitigation measure. Short-term mitigation activities included
sealing and caulking floors, cracks, and holes that could serve as vapor
infiltration points. Follow-up testing was conducted in these structures
following the completion of the short-term mitigation measures, and
additional follow-up testing was conducted after startup of the vapor
recovery system.
Site 2 Approach and Results
Design of the indoor testing program for Site 2 was based primarily
on the very low levels and common nature of compounds expected to be
measured indoors, and on specific requirements of the regulatory agencies.
As for Site 1, the purpose of the indoor testing was to determine whether
subsurface compounds had entered structurijs through the surrounding soils.
The target compound list was provided by the regulatory agencies and
included eight relatively common solvents. Results of subsurface
investigations indicated little or no likely influence on indoor levels.
Some or all of the compounds were likely ~o be present in household
chemical products and in the outdoor ambient air. In addition, three of
the target compounds (benzene, toluene, and cyclohexane) had been measured
in natural gas. Since most of the structures were supplied with natural
gas for heating and hot water, this was another potential indoor source of
the target compounds.
At the request of the regulatory agencies, the offer of testing was
extended to every structure along the fenoeline in the study area; 49
structures were tested over approximately six months. To provide data
representative of typical residential levels of target compounds, control
area testing was conducted. The control area was selected to be as similar
to the study area as possible without the presence of any subsurface
contamination. In selecting the control area, features such as age, size,
value, construction, heating types, and garage types of the homes were
considered, along with the predicted ambient air quality based on
dispersion modelling. Twenty control area structures were tested to
provide a representative data set that could be compared to the study area
data set. Simultaneous outdoor testing was conducted at each study area
and control area location to indicate outdoor levels of target compounds
during the indoor testing.
Gas chromatography/mass spectroscopv was used to analyze eight-hour
evacuated stainless steel canister air samples for the target compounds.
In addition to the target compounds, the ten compounds producing the
largest GC response were reported to further characterize the air samples.
Real-time instruments were used to screen for point sources of natural gas
or other point sources, and a complete inventory of household chemicals was
performed while the samples were being collected.
Detection limit studies and a high level of quality assurance/quality
control (QA/QC) testing were performed to evaluate the limits of certainty
associated with individual low-level results. Results of these assessments
indicated good performance of the sampling and analytical systems, and
provided method and total system detection limits, along with measures of
the inherent variability and false positive/negative potential that were
977
------- |