600R01005
EPA/#
September 2000
Recommended Methods for
Ambient Air Monitoring of NO, NO2, NOY,
and Individual NOZ Species
edited by
W.A. McClenny
Human Exposure and Atmospheric Sciences Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
-------
NERL-RTP-HEASD-oo-243 TECHNICAL REPORT DATA
1. Report No. EPA/600/R-01/005
2.
3. Recipient's Accession No.
4. Title and Subtitle
Recommended Methods for Ambient Air Monitoring of NO, NO2,
NOY, and Individual NOZ Species
5. Report Date
September 2000
6. Performing Organization Code
7. Author(s)
W.A. McClenny and others
8. Performing Organization
Report No.
9.Performing Organization Name and Address
Atmospheric Methods and Monitoring Branch, Human Exposure
and Atmospheric Sciences Division, National Exposure Research
Laboratory, U.S. EPA, Research Triangle Park, NC, 27711
10. Program Element No.
E-3917
11. Contract/Grant No.
68-D5-0049
12.Sponsoring Agency Name and Address
Atmospheric Methods and Monitoring Branch, Human Exposure
and Atmospheric Sciences Division, National Exposure Research
Laboratory, U.S. EPA, Research Triangle Park, NC, 27711
13. Type of Report and Period
Covered
14.Sponsoring Agency Code
15. Supplementary Notes
16. Abstract The most appropriate monitoring methods for reactive nitrogen oxides are identified subject to the
requirements for diagnostic testing of air quality simulation models. Measurements must be made over one hour or less
and with an uncertainty of equal to or less than 20% (10% for NO2) over a typical ambient concentration range extending
from a lower limit of Ippbv. NO, NO2, HNO3, PAN, and other reactive nitrogen oxides that exist at the Ippbv level and
above, along with the compound sets designated as NOY, NOX, and their difference, NO2 are included in this
measurement requirement. New and/or improved measurement techniques for NO2 monitoring including laser-induced
fluorescence, photolytic conversion/NO.Oj chemiluminescence, differential optical absorption spectroscopy, and
NOj/luminol chemiluminescence are examined with reference to literature citations and to field monitoring as part of the
1999 SOS Summer Field Study in Nashville, TN. Existing approaches to monitoring the other most prevalent reactive
oxides of nitrogen are reviewed. At the lower end of the ambient monitoring range, research-grade instruments are often
needed and operator skill, experience, and close attention are critical to proper instrument operation, calibration, and
maintenance. If the most appropriate methods are used and other species and atmospheric parameters relevant to ozone
production and accumulation are also accurately measured, air quality simulation models can be diagnostically tested and
the bas's for regulatory decisions such as the NO, SIP Call can be evaluated.
17. KEY WORDS AND DOCUMENT ANALYSIS
A. Descriptors - Nitrogen Oxides, Ozone, Nitrogen
Dioxide, Nitrogen Oxide, Nitric Acid
B. Identifiers / Open Ended
Terms
C. COSAT1
18. Distribution Statement
19. Security Class (This
Report)
20. Security Class (This
Page)
21. No. of Pages
22. Price
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development assembled
the information in this report with input from a number of prominent scientists and funded a research effort to
provide field data for this report under Contract 68-D5-0049 to ManTech Environmental Technology, Inc. The
report has been subjected to the Agency's peer and administrative review and has been approved for publication
as an EPA document. Mention of trade names or commercial products does not constitute endorsement or
-------
Abstract
The most appropriate monitoring methods for reactive nitrogen oxides are identified subject to the
requirements for diagnostic testing of air quality simulation models. Measurements must be made over 1 h or
less and with an uncertainty of s20% (10% for NO2) over a typical ambient concentration range extending from
a lower limit of 1 ppbv. NO, NO2, HNO3, PAN, and other reactive nitrogen oxides that exist at the 1-ppbv level
and above, along with the compound sets designated as NOY (all reactive nitrogen oxide compounds), NOX
(NO + NO2), and their difference, NOZ, are included in this measurement requirement. New and/or improved
measurement techniques for NO2 monitoring, including laser-induced fluorescence, photolytic conversion/NO,
03 chemiluminescence, differential optical absorption spectroscopy, andNO2/luminol chemiluminescence, are
examined with reference to literature citations and to field monitoring as part of the 1999 Southern Oxidants
Study (SOS) summer field study in Nashville, TN. Existing approaches to monitoring the other most prevalent
reactive oxides of nitrogen are reviewed. At the lower end of the ambient monitoring range, research-grade
instruments are often needed and operator skill, experience, and close attention are critical to proper instrument
operation, calibration, and maintenance. If the most appropriate methods are used and other species and
atmospheric parameters relevant to ozone production and accumulation are also accurately measured, air quality
simulation models can be diagnostically tested and the basis for regulatory decisions such as the NOX State
Implementation Plan (SIP) Call can be evaluated.
-------
Foreword
The National Exposure Research Laboratory, Research Triangle Park, North Carolina, performs research
and development to characterize, predict, and diagnose human and ecosystem exposure, giving priority to that
research which most significantly reduces the uncertainty in risk assessment and most improves the tools to
assess and manage risk or to characterize compliance with regulations. The Laboratory seeks opportunities for
research collaboration to integrate the work of ORD's scientific partners and provides leadership to address
emerging environmental issues and advance the science and technology essential for understanding human and
ecosystem exposures. One aspect of the Laboratory's mission is to support the iterative process of comparing
model predictions with experimental observations so that ultimately the reconciliation of differences is accom-
plished. In the case of ozone production in the ambient atmosphere through photochemical processes, this
support includes the identification and/or development of monitoring instrumentation of sufficient quality to
evaluate the impact of control strategies.
EPA has established a strategic plan for its scientific program of research and development that includes
production of reports that respond to the perceived needs in support of public interest through compliance with
the Government Performance and Results Act (GPRA). The current report is GPRA Annual Performance
Measure (APM) #442; it contains a summary of published methods for monitoring nitrogen oxides and
identification of the best available methods. Resources for the production of this report are provided as part of
EPA' s N ARSTO program, which includes the study of ozone precursors and the relationship between emissions
of these precursors and the production of ozone in the troposphere. Information provided herein is needed to
understand the priorities for instrumentation improvements and to appreciate the current viable alternatives
available to the atmospheric scientist. The objective of identifying the best methods is ultimately to obtain the
best ambient air data in support of air quality simulation models so as to predict the occurrence of ozone in
terms of the presence of ozone precursor compounds, one category of which is the nitrogen oxides.
This report incorporates information on the need for monitoring methods for nitrogen oxides, on their
required level of performance, on alternatives available in both the commercial and research arenas, on the
operating principles of several monitoring methods, and on the literature that documents information on
methods. A major monitoring need is for diagnostic testing of air quality models (AQMs) so that refinements
can be made to existing models (see Chapter 4). These refinements would then enhance the relevance of model
predictions with respect to recommendations on the control of nitrogen oxide emissions and of ozone
concentrations. Discussion of the various alternatives indicates where improvements in monitoring methods are
most likely to be made so that scarce resources can be allocated accordingly. For all the reactive oxides of
nitrogen considered here, commercial and/or research-grade instruments meet the needs for monitoring with
respect to diagnostic testing of AQMs, although improvements are almost universally indicated to avoid the
often tedious monitoring protocols that are currently required.
Gary J. Foley, Ph.D.
Director
National Exposure Research Laboratory
Research Triangle Park, NC 27711
-------
Contents
Notice ii
Abstract iii
Foreword iv
Figures viii
Tables x
Acronyms and Abbreviations xi
Acknowledgments xii
Dedication xii
Contributors xiii
Chapter 1 Introduction 1
1.1 Rationale for the Need to Measure Nitrogen Oxides 1
1.2 Note about Recommendations of this Report 2
1.3 Definition of Compounds of Interest 2
1.4 Previous Survey of Instrumentation 2
1.5 Point and Path Monitors 2
1.6 Other Important Issues for Tropospheric Monitoring of Nitrogen Oxides 3
1.7 Organization of the Report 3
Chapter 2 Conclusions 4
2.1 Recommendations for Monitoring Total Reactive Oxides of Nitrogen (NOY) 4
2.2 Recommendations for Monitoring Nitric Oxide (NO) 4
2.3 Recommendations for Monitoring Nitrogen Dioxide (NO2) 5
2.4 Recommendations for Monitoring Nitric and Nitrous Acids (HNO3 and HONO) ... 5
2.5 Recommendations for Monitoring Particle Nitrate 5
2.6 Recommendations for Monitoring PAN, PPN, and MPAN 6
2.7 Additional Comments 6
Chapter 3 Recommendations 7
Chapter 4 On the Need for Better Ambient Observations of Important Chemical
Species for Air Quality Model Evaluation 8
4.1 Introduction 8
4.2 Air Quality Process Models in Pollutant Management 8
4.3 Need for Diagnostic Testing in AQM Evaluation 9
4.4 System- and Process-Level Descriptions of the Tropospheric
Photochemistry of 63 10
4.5 Instrumenting AQMs for Diagnostic Testing 15
4.6 A Taxonomy of Diagnostic Tests for Tropospheric Photochemistry 15
-------
4.8 Priority and Utility of Observations Needed for Diagnostic Testing 18
4.9 General Requirements of Observations for Diagnostic Model Evaluation 19
4.10 Summary 21
Chapter 5 Methods for Nitrogen Oxides Monitoring: Discussion and Recommendations 22
5.1 Introduction 22
5.2 Methods for Total Reactive Oxides of Nitrogen (NOY) 22
5.2.1 Method Recommendations for NOY 23
5.3 Methods for Nitric Oxide (NO) 23
5.3.1 NO, O3 Gas-Phase Chemiluminescence 23
5.3.2 Differential Optical Absorption Spectroscopy 24
5.3.3 Tunable Diode Lasers 24
5.3.4 Other Point Monitoring Techniques 24
5.3.5 Method Recommendations for NO 24
5.4 Methods for Nitrogen Dioxide (NO2) 24
5.4.1 NO, O3 Gas-Phase Chemiluminescence 24
5.4.2 Luminol Chemiluminescence 25
5.4.3 Laser-Induced Fluorescence 25
5.4.4 Differential Optical Absorption Spectroscopy for NO2 25
5.4.5 Tunable Diode Laser Absorption Spectroscopy—Middle Infrared 26
5.4.6 Systems Using Visible/Near Infrared Radiation Laser Sources or LEDs ... 26
5.4.7 Method Recommendations for NO2 27
5.5 Methods for Monitoring Nitric and Nitrous Acids (HNO3 and HONO) 27
5.5.1 Thermodenuders 27
5.5.2 Dual-Channel Chemiluminescence Monitors 28
5.5.3 Wet Denuders 28
5.5.4 The Mist Chamber Technique 29
5.5.5 Chemical lonization Mass Spectrometry 29
5.5.6 Tunable Diode Laser Absorption Spectroscopy and DO AS Techniques ... 29
5.5.7 DNPH Derivatization and HPLC Analysis 29
5.5.8 Method Recommendations for HNO3 and HONO 30
5.6 Methods for Monitoring Particle Nitrate 30
5.6.1 Method Recommendations for Particle Nitrate 30
5.7 Methods for Monitoring PAN, PPN, MPAN, and Other Organic Nitrates 30
5.7.1 Gas Chromatography with Specific Detection 31
5.7.2 Gas Chromatography/Negative Ion Chemical lonization
Mass Spectrometry 31
5.7.3 Method Recommendations for PAN, PPN, MPAN, and Other
Organic Nitrates 31
5.8 Methods for Monitoring the Nitrate Radical 31
5.8.1 Method Recommendations for the Nitrate Radical 32
Chapter 6 Use of Commercially Available Systems for NO2 Monitoring in the
Nashville SOS '99 Study 33
Chapter 7 References 37
Appendix A Correction to Point Monitor Readings of O3, NO, and NO2 Due to the Reaction
of NO and O3 during Transport of Ambient Air to the Point of Measurement 44
A. 1 Development of Technical Guidance for NO2 Monitoring 44
A. 2 Conclusion on Gas-Phase Reaction between O3 and NO 45
-------
Appendix B Determination of Atmospheric Concentration of Nitrogen Oxides by
Differential Optical Absorption Spectroscopy 46
B.I Introduction 46
B.2 Theory of DOAS 46
B.2.1 Principle 46
B.2.2 Practical Considerations 47
B.3 Experimental Realization 48
B.3.1 Experimental Setup 48
B.3.2 Analysis of DOAS Spectra 49
B.3.2.1 Determination of the Absorption Cross Section
for the Instrument 49
B.3.2.2 Filtering Procedure 49
B.3.2.3 Separation Algorithm 50
B.3.2.4 Errors 50
B.4 Theoretical Considerations about DOAS 51
B.4.1 Linearity 51
B.4.2 Random Error of the Measurement 51
B.4.3 Possible Systematic Errors 51
B.4.4 Accuracy of DOAS 52
B.5 Current State of the Art 52
B.6 Summary 53
B.7 References 55
Appendix C Photolytic Conversion of Ambient NO2 57
C. 1 Physics and Chemistry of the Measurement 57
C.2 Instrumental Application of NO2 Photolysis 58
C.2.1 General Considerations 58
C.2.2 Practical Considerations 58
C.2.3 Current State of the Art 59
C.2.4 Data Reduction Requirements 59
C.2.5 Summary on Photolytic Converters 61
Appendix D Laser-Induced Fluorescence Detection of NO2 62
D. 1 Sampling and Calibration 64
D.2 Field Trials and Intercomparisons 64
D.3 Future LIF Systems 65
D.4 Acknowledgments 66
D.5 References 66
Appendix E Application of a Luminol-Based Approach to Monitoring NO2, NOX, and NOY 69
-------
Figures
4-1 High-level conceptual diagram of tropospheric photochemistry showing the
major elements of radical initiation, propagation, and termination and the
resultant NO and NOX cycles 11
4-2 Depiction of the nonlinear ozone response surface through plotting daily peak
ozone concentrations as predicted by a photochemical box model for a range of
initial concentrations of VOCs and NOX 12
4-3 Two-dimensional view of Figure 4-2 12
4-4 Cyclic reaction of methane oxidation (as a surrogate for hydrocarbon oxidation)
involving OH and conversion of nitric oxide to nitrogen dioxide, termination to
peroxides and nitric acid, and concomitant formation of ozone 13
5-1 HNO3 and HONO with 10-min resolution as measured using a wet-wall denuder
system for a three-day stretch in the Atlanta SOS study during the summer of 1999 29
6-1 Results of a monitoring comparison at the Cornelia Fort during the period 28 June-
5 July 1999. Three instruments were used: the UV/DOAS system with a 200-m
total optical open pathlength at an average aboveground height of 8 ft and
two point monitors, the LMA-3 and the TEI42, at a height of 30 m 34
6-2 Results of a monitoring comparison at the EPA facility in Research Triangle Park, NC,
during the period 4-11 October 1999 using the same instruments as in Figure 6-1 34
6-3 Results of comparison data for NO2 over a 24-h period, 20-21 June 1999, at the
Cornelia Fort site—UV/DOAS open-path monitor, LIF (Cohen, UC-Berkeley),
and chemiluminescence monitor with photolytic converter (Williams, Aeronomy
Laboratory, NOAA, Boulder, CO) 35
6-4 Results of comparison data for NOY during the period 28 June-5 July 1999
at the Cornelia Fort site—LMA-3, TEI 42, and TEI 42C 36
6-5 Results of comparison data for NOY during 20-21 June 1999 at the Cornelia Fort
site—LMA-3, TEI 42, and TEI 42C 36
A-l Effect of O3 on NO sampling line losses 45
B-1 Schematic overview of DOAS measurements 46
-------
B-2 The principle of DOAS 47
B-3 Known differential absorptions and mercury emission lines in the wavelength
range typically used to monitor NO2 and HONO 50
B-4 Intercomparison of a chemiluminescence detector with photolytic converter and a
DOAS system with a 15-mm-long multireflection cell setup 53
B-5 Intercomparison of a long-path DOAS system (1.3-km light path length)
with an in situ chemiluminescence analyzer with photolytic converter
during the SOS '99 field intensive in Nashville, TN 54
C-l Configurations for photolytic converters 60
D-l Schematic of the UC Berkeley LIF NO2 instrument 63
D-2 NO2 concentration plotted as 30-s averages vs. time for the 2-week period
16-29 August 1998 at UC Blodgett Forest Research Station 65
D-3 NO2 concentration plotted as 1-min averages vs. time at Cornelia Fort
Airpark, Nashville, TN 67
D-4 Southern Oxidants Study, Nashville 1999, Cornelia Fort Airpark NO2
intercomparison plots 68
E-l LMA-3 field setup for SOS '99, Nashville, TN 70
E-2 LMA-3 NOY multipath converter 71
-------
Tables
4-1 Taxonomy of Photochemical Diagnostic Elements 16
4-2 Summary of Diagnostic Tests 21
4-3 Priority Measurements 21
-------
Acronyms and Abbreviations
APM annual performance measure NOX
AQM air quality model NOY
BRD balanced ratiometric detector
CF conversion factor NOZ
CIMS chemical ionization mass NRC
spectrometry Ox
CMAQ Community Multiscale Model for
Air Quality PAMS
CRDS cavity ring-down spectroscopy
DIAL differential absorption lidar PAN
DOAS differential optical absorption PC
spectroscopy PDA
BCD electron capture detector PDE
FTIR Fourier transform infrared pm
GC gas chromatography PM
GC/MS gas chromatography/mass PMT
spectrometry ppbv
GPRA Government Performance and ppbC
Results Act ppbv«m
HC hydrocarbon subset of VOCs
1C ion chromatography ppmv
IR infrared PPN
LIF laser-induced fluorescence pptv
LOD limit of detection PPS
MIESR matrix isolation electron spin RADM
resonance ROC
MPAN peroxymethacrylyl nitrate SDM
MS mass spectrometry SIP
NAAQS National Ambient Air Quality slpm
Standards SNR
NARSTO National Atmospheric Research SOS
Strategy for Tropospheric Ozone TDL
NHX sum of gaseous ammonia and TOLAS
paniculate ammonium
NIST National Institute of Standards and TOP AS
Technology
nm nanometer TTFMS
NOAA National Oceanic and Atmospheric
Administration UAM-IV
non-(NOY)i nitrogen-containing compounds not UV
part of NOY (e.g., NH3, HCN) VOC
sum of NO plus NO2
sum of NOX and other reactive
nitrogen compounds
NOY - NOX
National Research Council
sum of all species that can act as
reservoirs for atomic oxygen
Photochemical Assessment
Monitoring Stations
peroxyacetyl nitrate
photolysis-chemiluminescence
photodiode detector array
partial differential equation
picometer
paniculate matter
photomultiplier tube
parts per billion by volume
parts per billion carbon
parts per billion by volume times
meters
parts per million by volume
peroxypropionyl nitrate
parts per trillion by volume
pseudo-photostationary state
Regional Acid Deposition Model
reactive organic compounds
slotted disk machine
State Implementation Plan
standard liters per minute
signal-to-noise ratio
Southern Oxidants Study
tunable diode laser
tunable diode laser absorption
spectroscopy
Tropospheric Optical Absorption
Spectroscopy
two-tone frequency-modulated
spectroscopy
Urban Airshed Model version IV
ultraviolet
volatile organic compound
-------
Acknowledgments
This report is a joint effort involving members of the monitoring and modeling communities who have
graciously provided information in their areas of expertise or have carefully peer reviewed the manuscript.
Special acknowledgment goes to Dr. I.E. Sickles of the National Exposure Research Laboratory, U.S. EPA,
for his advice on various aspects of this document; to Dr. G.M. Russwurm of ManTech Environmental
Technology, Inc., for his assistance in understanding the subtleties of differential optical absorption spectro-
scopy; to Dr. E.H. Daughtrey, Jr., of ManTech Environmental Technology, Inc., for his contributions to
organizing the ManTech field studies that provided some of the data used in this report; and to Dr. P.K.
Dasgupta of the Texas Tech University who has freely shared his knowledge on microchemistry, an area in
which he pioneers. Thanks also to Dr. F. Fehsenfeld of the NOAA Aeronomy Laboratory, Dr. D. Stedman of
the University of Denver, and Ms. Joann Rice of the OAQPS, U.S. EPA, for peer review prior to publication.
Dedication
This report is dedicated to the memory of Dr. J.A. Hodgeson, an EPA research chemist who pioneered
the use of gas-phase chemiluminescence for the monitoring of ambient trace gases, particularly the nitrogen
oxides. His contributions in the early years after EPA's formation were significant in setting the direction for
subsequent progress in monitoring technology.
-------
Contributors
J.R. Arnold, Atmospheric Modeling Division, National Exposure Research Laboratory, U.S. EPA, Research
Triangle Park, NC
R.C. Cohen, University of California at Berkeley, Berkeley, CA
R.L. Dennis, Atmospheric Modeling Division, National Exposure Research Laboratory, U.S. EPA, Research
Triangle Park, NC
K.G. Kronmiller, ManTech Environmental Technology, Inc., Research Triangle Park, NC
D.J. Luecken, Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory,
U.S. EPA, Research Triangle Park, NC
W.A. McClenny, Human Exposure and Atmospheric Sciences Division, National Exposure Research
Laboratory, U.S. EPA, Research Triangle Park, NC
J. Stutz, Department of Atmospheric Sciences, UCLA, Los Angeles, CA
J. A. Thornton, University of California at Berkeley, Berkeley, CA
G.S. Tonnesen, CE-CERT, University of California at Riverside, Riverside, CA
M. Wheeler, ManTech Environmental Technology, Inc., Research Triangle Park, NC
E.J. Williams, Aeronomy Laboratory, NOAA Environmental Research Laboratories, Boulder, CA
P.J. Wooldridge, University of California at Berkeley, Berkeley, CA
-------
Chapter 1
Introduction
by
W.A. McClenny, U.S. EPA
The objective of this report is to recommend the best
monitoring methods for NO, NO2, NOX, NOY, and speciated NOZ
to support the current national strategy on regional transport of
ozone. Definitions for the compound groupings NOX, NOY, and
NOZ are given in section 1.3. Such instrumentation will be used
to evaluate the effectiveness of NOX emissions reduction as
initially envisioned in the September 1998 final Regional Trans-
port of Ozone Rule (NOX State Implementation Plan [SIP] Call).
As part of this document (Federal Register, 1998), the Federal
government originally proposed that 22 eastern states and the
District of Columbia reduce source emissions of NOX (NO +
NO2) so as to lower the ground-level O3 concentrations at down-
wind sites. An average targeted reduction in NOX emissions of
28% was determined assuming that reasonable, cost-effective
control measures were applied. Subsequently, on 25 May 1999,
the U.S. Court of Appeals for the D.C. Circuit issued an order
partially staying the implementation of the NOX SIP call.
Arguments before the Court were presented on 9 November
1999, and on 3 March 2000 the Court of Appeals generally
upheld the NOx SIP Call provisions. On 22 June 2000, the
deadline for states to submit SIPs was set as 30 October 2000,
and on 30 August 2000 the compliance deadline for the SIP call
provisions was extended to 31 May 2004.
In related actions, on 17 December 1999 EPA granted
petitions filed by four northeastern states seeking to reduce ozone
pollution through reductions in nitrogen oxide emissions from
other states. These petitions were filed under Section 126 of the
Clean Air Act. Understanding the effectiveness of the actions
carried out as a result of the NOX SIP Call and the Section 126
petitions will ultimately require experimental measurements of
ozone and ozone precursors including the nitrogen oxides.
Ideally, these measurements along with those of other trace
species and atmospheric variables will vindicate the current
national strategy of controlling O3 production by selectively
reducing NOX and/or reactive organic compound (ROC) emis-
sions. Measurements before-and after implementation of emis-
sion controls will be required to establish a basis for quantifying
changes.
1.1 Rationale for the Need to
Measure Nitrogen Oxides
A detailed discussion of the rationale for obtaining
measurements of individual nitrogen oxide species including NO,
NO2, HONO, HNO3, organic nitrates, particle nitrate, and NOy
is presented in Chapter 4. This discussion, titled "On the Need
for Better Ambient Observations of Important Chemical Species
for Air Quality Model Evaluation," emphasizes diagnostic testing
of air quality models (AQMs). In diagnostic testing, field mea-
surements of key trace species and of process rates (reaction
rates, photolysis rates, etc.) are compared to the model results at
preliminary output levels in the modeling procedure for
predicting ambient ozone. Given sufficiently accurate and precise
measurements, the results of these comparisons identify the
strong and weak points in the model simulations of the ambient
atmosphere. Measurements of key trace species concentrations
taken with a short temporal resolution compared with typical
significant changes in daytime ozone concentration are most
desirable. Chapter 4 establishes the required frequency of mea-
surement as significantly less than one hour. The discussion gives
specific recommendations for prioritizing variables for a broad
range of parameters, including those nitrogen oxides of specific
interest, along with some estimates of the measurement certainty
required for meaningful comparison tests with the predictions of
AQMs. Measurements of NO2 with an uncertainty of 10% at the
1-ppbv and higher concentration levels and of no greater than
20% for individual NOZ components are given as two require-
ments. Measurements of individual NOZ species such as HNO3
and PAN-like compounds are used in "air mass history resultant
-------
photochemistry tests" where their ratios to other compounds are
formed (see below). Additional requirements on these monitors
may be imposed.
1.2 Note about Recommendations
of this Report
The recommendations of this report are for monitoring
methods that have been shown to meet the rather stringent mon-
itoring requirements for diagnostic testing and atmospheric
research. A method's potential may only have been realized with
customized instrumentation assembled from component parts in
the research laboratory. While this is most often the case, the
existence of the National Ambient Air Quality Standards
(NAAQS) for NO2 and other criteria pollutants has created a
large market for commercial instruments designed for routine
monitoring. In some cases redesign of these instruments has
improved performance levels to meet research requirements.
Therefore commercial instruments will be mentioned as part of
the discussion of methods in cases where they have been used to
experimentally demonstrate a potential for research monitoring
applications. One example is the monitoring of total NOY and
NO by NO, O3 chemiluminescence, which is the reference
method for NAAQS criteria monitoring of NO2. In this case,
commercially available instruments have been appropriately
modified and used to demonstrate the potential of the chemi-
luminescence method. The different monitoring methods are
mentioned in Chapter 2 as part of the conclusions and are
discussed in more detail in Chapter 5 where information on
alternative methods is given. Often several methods have been
successful in achieving the monitoring requirements of diagnostic
testing of AQMs and the scientist must choose among these
based on the perception of advantages and disadvantages.
1.3 Definition of Compounds of Interest
As noted by Kliner et al. (1997), NOY consists of all
oxides of nitrogen in which the oxidation state of the N atom is
+2 or greater, i.e., the sum of all reactive nitrogen oxides
including NOX (NO + NO2) and the remaining set of compounds
that collectively are denoted by NOZ, i.e., NOY - NOX = NOZ.
For the purposes of this document, the most prevalent subset of
NOZ compounds will be treated. These compounds are known to
account for almost all of the reaction products of NOX in ambient
air. This operational set of NOZ compounds consists of nitric and
nitrous acids (HNO3 and HONO); the organic nitrates including
peroxyacetyl nitrate anhydride (PAN), peroxymethacrylyl nitrate
(MPAN), and peroxypropionyl nitrate (PPN); and particulate
nitrates. The nitrogen oxides covered in order of presentation are
NOY, NO, NO2, HNO3, HONO, PAN, PPN, MPAN, RONO2,
and particle nitrate.
1.4 Previous Survey of Instrumentation
Sickles (1992) provides excellent documentation in the
open literature of sampling and analysis techniques for ambient
oxides of nitrogen and related species including references up to
a certain point in 1990. His work is also available in Volume I,
Chapter 6, of EPA publication EPA/600/8-9 l/049aF (August
1993) titled "Air Quality Criteria for Oxides of Nitrogen." For
the present report, two types of references are listed in the
reference section (Chapter 7): those resulting from a literature
search for the period between 1990 and 2000 and certain pre-
1991 references that are required to establish discussion points
in the text. The reader is referred to Sickles (1992) for a com-
prehensive listing of the pre-1991 references.
Parrish and Fehsenfeld (2000) provide a review of
methods for gas-phase measurements that includes the nitrogen
oxides. Their article is part of a special issue of Atmospheric
Environment titled the NARSTO Ozone Assessment—A Critical
Review.
1.5 Point and Path Monitors
Most major contributors to the nitrogen oxides can be
monitored by either point or open-path monitors. Point monitors
sample ambient air through an inlet to the interior of the monitor
where the sample is probed and examined in a controlled envi-
ronment. Sample conditioning to selectively eliminate interfer-
ences and/or to concentrate the sample is possible. Instrument
calibration is typically performed by providing a known con-
centration of the target gas at the inlet. Responses to ambient air
samples are then referenced to a calibration curve to infer the
ambient air concentration of target species.
Open-path monitors typically monitor by probing the
ambient air with a beam of radiation over significant distances
(0.1-5 km). Radiation missing from the transmitted spectra is
attributed to a combination of absorption and scattering by
specific gases and airborne particles distributed along the path
length. No sampling is performed and hence no sampling arti-
facts occur. The sampling environment is uncontrolled so that
sample conditioning to remove interferences or to cause sample
concentration is impossible. The parameter measured by open-
path monitors is the product of gas concentration and distance
usually stated in parts per billion by volume times meters
(ppbv*m). Dividing this product by a known path length, the
average concentration is obtained. As a result of the spatial
averaging, the extent of variations in concentration along the path
is reduced from that monitored by a point monitor, but the path-
averaged value is more representative over the extended dis-
tances typical of AQMs. Calibration is performed by comparing
field spectra to reference spectra taken in the laboratory (ideally
with the same instrument) and/or by placing calibration cells
-------
beam along a portion of the measurement path. Certain types of
open-path measurements can be range resolved by analyzing the
backscattered light from pulsed radiation sources. For example,
short pulses of laser radiation are backscattered from atmo-
spheric particles and received at the laser location as a function
of time. By using multiple wavelengths to establish a difference
in absorption along the path, range resolved measurement of
target gases can be obtained. These systems are referred to as
differential absorption lidar (DIAL) systems and are most
frequently used to map dispersing source plumes. Pulsed laser
systems can also be used for column burden measurements
(ppbvm). In this case the difference in absorption at multiple
wavelengths is obtained using an aircraft-mounted source/
receiver and the earth's surface as a diffuse backscatter of
radiation.
1.6 Other Important Issues for
Tropospheric Monitoring of
Nitrogen Oxides
Calibration of the different trace species can be a limiting
factor in achieving the level of accuracy required to meet the data
quality objectives needed for diagnostic testing of AQMs. Pro-
cedures are firmly established for on-site calibration of NO and
NO2 monitoring instruments due to the NAAQS program, but
have to be extended to measure the low to sub-ppbv levels sought
in atmospheric research measurements. Reliable standards for
other nitrogen oxides are not so routinely available and require
a concentrated effort to establish. Although recommendations for
calibration techniques are beyond the scope of work attempted
in this document, the method references listed in Chapter 7 docu-
ment the procedures for calibration used for specific methods.
Measurements in the vertical dimension are required for
a complete description of the air quality mixture affecting
ground-level ozone. Measurements from aircraft, from tethered
balloons, from radiosondes, and by remote monitoring systems
such as sodar, radar, and lidar are made to provide this infor-
mation. The use of instrumentation for measuring the vertical
distribution of nitrogen oxides is not specifically discussed in this
document, but a number of the ground-based methods have been
adapted to aircraft and/or modified to meet the special require-
ments of rapid response and orientation insensitivity that are
required. Results of comparison studies of instrumentation used
in aircraft, including NO instrumentation (Hoell et al., 1987),
NO2 instrumentation (Gregory et al., 1990a), and PAN instru-
mentation (Gregory et al., 1990b), have been documented by
scientists at NASA Langley.
1.7 Organization of the Report
The report is organized to provide a basis and rationale
for recommending monitoring methods for tropospheric nitrogen
oxides. Following the Introduction, Conclusions, and Recom-
mendations, Chapter 4 presents the rationale for using measure-
ments for diagnostic testing of AQMs. The instruments available
for NOY and individual reactive nitrogen oxides are discussed in
Chapter 5, including the rationale for choosing among alterna-
tives. Chapter 6 reviews some of the data obtained with com-
mercially available NO2 monitoring systems used in the Nash-
ville Southern Oxidants Study (SOS) '99 summer field study.
Supplemental information to highlight recent and/or pertinent
experimental and theoretical work is provided in the Appendices.
Appendix A explains the importance of the NO, O3 gas-phase
reaction in the inlet to sampling systems for NO, NO2, and O3.
Appendix B provides documentation on the design features and
performance characteristics of the open-path monitoring systems
based on differential optical absorption spectroscopy (DOAS).
Appendix C discusses the characteristics of photolytic converters
as an alternative to thermal converters for conversion of NO2 to
NO prior to chemiluminescence detection of NO. Appendix D
provides documentation of recently completed research on the
use of laser-induced fluorescence (LIF) for direct NO2 moni-
toring. Appendix E is a discussion of the applications of luminol-
based monitoring of NO2 and related procedures for monitoring
-------
Chapter 2
Conclusions
by
W.A. McClenny, U.S. EPA
Chapter 4 explains that many factors determine the pro-
duction and accumulation of ozone downwind of fuel-fired
electric utilities and other major point sources of NOX emissions.
AQMs have been developed to incorporate representations for all
these factors and to assign the proper role to each. Based on
AQM predictions, the U.S. EPA has determined that reduction in
NOX emissions from major sources in the Northeast U.S. will
lead to significant reduction of downwind ambient O3 concen-
trations and thereby significantly mitigate the detrimental effects
of high ozone concentrations on human health and the ecology.
Responsibilities with respect to transport of ozone and ozone
precursors across state boundaries in the Northeast and across the
international boundary with Canada are also of concern. To
evaluate implementation of controls on these sources, AQM
predictions will be compared to experimental measurements both
before and after control implementation. The effect of the emis-
sions reduction must be separated from other factors to show the
real consequences. Since separation of factors must occur, the
adequacy of all model constructs must be evaluated using a
holistic or "one atmosphere" approach. Among the critical com-
ponents is the accurate measurement of reactive nitrogen oxides.
The accomplishment of this objective and others of
comparable importance is equivalent to being able to show
relationships among ozone, its precursors, and other variables
that are consistently interpretable and lead to changes that have
known attribution. This process is referred to later in this docu-
ment as diagnostic testing of the models. Only through diagnostic
testing can the accuracy and validity of the model be determined
and a consistent understanding of the interplay of different causal
factors contributing to ozone production and accumulation be
achieved.
Based on the requirements for adequate diagnostic testing
of AQMs as delineated in Chapter 4, methods for monitoring
ambient reactive nitrogen oxides at stationary, ground-level
monitoring sites must have temporal resolution of significantly
less than 1 h and a measurement uncertainty at 1 ppbv and higher
of 20% for NO, NOY, and the major components of NOZ (PAN,
PPN, MPAN, HNO3, MONO, and nitrate paniculate [as equiva-
lent NO]) and 10% for NO2 (see Chapter 4). Monitoring methods
that can meet these requirements or that appear to have the
potential are discussed in Chapter 5. Those methods that are
recommended for diagnostic testing and atmospheric research
based on documented experimental testing are summarized in
sections 2.1 through 2.6.
2.1 Recommendations for Monitoring
Total Reactive Oxides of Nitrogen (NOY)
Thermal conversion/NO, O3 chemiluminescence.
Use of a thermal converter to convert reactive nitrogen oxides to
NO followed by detection of NO by its chemiluminescence
reaction with an excess of O3 is recommended.
Reference (among others): Williams etal. (1998)
Note: Conversion of all reactive nitrogen oxides to NO,
followed by chemical conversion to NO2, and then detection by
LIF or luminol fluorescence is an option, but involves two con-
versions for all species except NO2.
2.2 Recommendations for Monitoring
Nitric Oxide (NO)
NO, O3 chemiluminescence. Detection of NO by its
chemiluminescence reaction with an excess of O3 is recommended.
Note: All point monitors are subject to sampling losses of
NO in the inlet lines due to reaction with co-collected ambient O3
and the walls of the inlet lines. Thus, rapid transit of the lines is
-------
including both the lines and any sample conditioner (e.g., photo-
lysis cell), ensures that NO is reduced no more than 10% before
reaching the detector (see Appendix A).
References (among others): Fontijn et al. (1970),
Bolllnger (1982), Ridley et al (1988a)
2.3 Recommendations for Monitoring
Nitrogen Dioxide (NO2)
There are at least four successful methods to monitor NO2
over its ambient concentration range, as noted below. The use of
a photolytic converter with NO, O3 chemiluminescence is prob-
ably the most viable at this time, while the use of LIF in its
current incarnation is an excellent research tool. The use of
DO AS open-path measurements and luminol chemiluminescence
is also viable since interferences are generally known and can be
accounted for.
Luminol chemiluminescence. Detection of NO, by
liquid-based luminol chemiluminescence has been used as a
simple, sensitive method subject to interferences (PAN, O3, and
MONO) that can be reduced if not eliminated by adjusting the
composition of the luminol solution and using chemical scrub-
bers (see Appendix E). The method is subject to nonlinearity of
response at the lower end of the ambient NO2 concentration
range. At low NO2 concentrations these issues must be addressed
if accurate monitoring is to be achieved. GC/luminol detector
combinations are used to obtain PAN, MPAN, PPN, and NO2
over short cycle times (see section 2.6.).
References (among others): Kelly et al. (1990), Maeda et
al. (1980), Wendel et al. (1983)
Differential optical absorption spectroscopy.
Detection of NO2 by DOAS (see Appendix B) is direct and does
not require sampling; thus, no sampling losses occur. Uncertainties
in this measurement are subject to the choice of spectral fitting
routine and the ambient environment (mixture of permanent and
trace species, some of which may be unknown interferences). It is
clear that agreement between DOAS techniques and the point
monitoring techniques listed above is possible across the typical
monitoring range of NO2 in ambient air, provided a path of suf-
ficient length is used. The existence of unknown spectral inter-
ferences may require post-run analysis of spectra.
Reference (among others): Plan (1994)
Photolytic conversion/chemiluminescence. Detec-
tion of NO2 by photolytic conversion to NO followed by its
chemiluminescence reaction with an excess of O3 achieves
monitoring objectives and is known to compare well with LIF
and DOAS measurements in field tests (see Appendix D). HONO
is a significant interference since it is photolyzed to an extent
depending on the exact nature of the spectral bandwidth of
radiation being used (37% is photolyzed in one state-of-the-art
system [Ryerson et al., 2000]). See Appendix C for a discussion.
References (among others): Ryerson etal. (2000), Gao et
al. (1994)
Laser-induced fluorescence. Detection of NO2 by LIF
is now a demonstrated, field-proven monitoring method using a
practical, frequency-doubled Nd3+: YAG laser to pump a dye laser
and is a direct method for NO2 monitoring over the entire range of
ambient concentration (see Appendix D). As solid-state laser tech-
nology advances, tunable diode lasers are expected to replace the
dye laser excitation systems, yielding a more compact, lightweight,
and technically simplified instrument.
Reference: Thornton et al. (2000)
2.4 Recommendations for Monitoring Nitric
and Nitrous Acids (HNO3 and HONO)
Note: Nitric acid is highly soluble in water and interaction
with surfaces has been a problem for sampling into point monitors.
Diffusion scrubber/ion chromatography. This tech-
nique involves the capture of HNO3 and HONO (as well as SO2
and other acid gases) in water solution followed by separation by
ion chromatography and detection of anions. The use of some type
of wet-wall gas-phase scrubber to remove HNO3 and HONO from
the sample stream while passing nitrate particles is required.
Reference (among others): Simon andDasgupta (1995a)
Chemical ionization mass spectrometry. This tech-
nique measures nitric acid with more than adequate sensitivity
(15 pptv in 1 s) and is the preferable approach aboard aircraft
due to its rapid response.
References (among others): Huey et al. (1998), Mauldin
etal. (1998), Fehsenfeld et al. (1998)
Chemiluminescence differencing technique. This
technique involves two sample streams, each passing through its
own sampling channel. One channel includes a nylon filter that
retains nitric acid, and both channels have identical thermal
converters. The difference in signal is due to HNO3. Decomposi-
tion of ammonium nitrate particles during sampling is a potential
problem.
Reference (among others): Tanner et al. (1998)
2.5 Recommendations for Monitoring
Particle Nitrate
Collection on filters placed downstream of acid gas
denuders followed either by water extraction and ion chromato-
graphic analysis or by thermal desorption and conversion to NO
-------
References (among others): Simon andDasgupta (1993),
Stolzenburg and tiering (2000)
2.6 Recommendations for Monitoring
PAN, PPN, and MPAN
The GC/ECD and GC/luminol chemiluminescence methods
have the time resolution and sensitivity to make these measure-
ments. One version of the GC/luminol method has been designed
to provide NO2 measurements as well.
References (among others): Roberts etal. (1989), Gaffney
etal. (1998)
2.7 Additional Comments
The discussion in Chapter 5 on methods for monitoring
the reactive nitrogen oxides provides the rationale for these
choices in terms of the interpretation of information obtained
from the scientific peer-reviewed references cited in the text and
the experimental work performed by those who have contributed
to the current effort. The recommended methods have the poten-
tial to meet the monitoring objectives to support diagnostic
testing of AQMs. However, this has not been done on a routine
basis with automated calibration and monitoring such as done for
NAAQS monitoring. To use these monitoring methods success-
fully, quality assurance measures, including calibration, main-
tenance, and instrument performance checks (e.g., converter
efficiency checks, background determinations), are critical and
must be performed with extreme care and diligence, especially at
the lower end of the ambient concentration range.
The goals of diagnostic testing can best be accomplished
by making a complete set of measurements, including speciated
volatile organic compounds (VOCs), OH, and HO2, and process
variables such as the photolytic production rates of OH from
HCHO, etc. Typically, these measurements are performed by
professional scientists in field-intensive studies. These studies
bring together instrumentation based on different principles
and/or instruments that use different protocols for the mea-
surement of each individual species. Agreement among these
methods means additional confidence can often be assigned to
critical measurements. This is particularly important at the lower
end of the concentration monitoring range where interferences
-------
Chapter 3
Recommendations
by
W.A. McClenny, U.S. EPA
The following recommendations address opportunities for
improvements in the quality of measurements for reactive nitro-
gen oxides and are based on the detailed information provided in
Chapters 4, 5, and 6.
• Continue support of comprehensive field intensives for
building consensus on the best methods for monitoring reac-
tive nitrogen oxides, and for diagnostic testing of models
(specifically the modeling of changes in ozone photo-
chemistry as the NOX SIP Call reductions are implemented
and the modeling of regions with significant persistent ozone
exceedances).
• Support research and stimulate commercial interest to im-
prove methods for sampling and analyzing reactive nitrogen
oxides, which includes the following actions:
- Identification and testing of a more stable thermal con-
verter configuration by redesign or by identification of
a different type of converter.
- Further development and commercialization of an aux-
iliary photolytic converter for conversion of NO2 to NO
so that the specificity of NO2 measurements can be
improved over that available with the thermal conver-
ters used in the current U.S. installed instrument base of
NO, O3 chemiluminescence monitors.
— Further development of a cost-effective LIF monitor by
redesign as cost-effective component parts become
available.
- Further development of a DOAS open-path system with"
full access to spectra and spectral fitting routines that
will facilitate postprocessing of data.
- Further development of a near real-time nitric/nitrous
acid monitor based on wet denuders and ion chroma-
tography or on an alternate approach.
- Further evaluation and application of a near real-time
monitor for NO2/PAN measurements based on gas
chromatography (GC) coupled to a NO2-specific
detector.
• Investigate interest by NIST in the production of the
following:
- An ultraviolet (UV)/visible spectral database similar to
the NIST reference library of infrared (IR) spectra.
- A NO2 reference photometer and monitor similar to
those used for O3. Recent research published in the
open literature implies this is now possible.
- Calibration gas standards (e.g., «-propyl nitrate) for
checking thermal converter efficiency in NOy mea-
-------
Chapter 4
On the Need for Better Ambient Observations of Important
Chemical Species for Air Quality Model Evaluation
by
Robin L. Dennis andJ.R. Arnold, Atmospheric Modeling Division, NERL, and
Gail S. Tonnesen, CE-CERT, University of California at Riverside
Chapter 4 describes a suite of measurements required to
evaluate the predictive accuracy of AQMs, an explanation of why
these measurements are needed, and at what concentration levels
and accuracy they are needed. This information is of significant
value in the information exchange between modelers and moni-
toring groups and is given here for the nitrogen oxides sum-
marized in this report and also for the entire array of mea-
surements associated with AQMs. The reader interested only in
a listing of measurement requirements will find them in section
4.9, "General Requirements of Observations for Diagnostic
Model Evaluation," and Tables 4.2 and 4.3.
4.1 Introduction
Air quality is determined by a complex system of coupled
chemical and physical processes. These processes include emis-
sions of pollutants and pollutant precursors, chemical reactions,
transport, and deposition. For ozone production, this system is
nonlinear (Seinfeld and Pandis, 1998). Multiple spatial and tem-
poral scales, multiple gaseous and paniculate pollutants, and
environmental issues such as acid or nutrient deposition to eco-
systems and visibility degradation are involved. The integrated
nature defined by these environmental factors and the feedback
cycles in them means that tropospheric air quality can best be
treated as "one atmosphere" (Dennis, 1998) in which multiple
precursors, intermediates, and products interact and react over
several scales of time and space.
B ut even from the integrated one-atmosphere perspective,
O3 is a pollutant of especially high concern because of its key
position in the processes and cycles affecting other pollutants and
its widespread effects on ecological and human health (National
Research Council [NRC], 1991). For reasons explained in
Chapter 5, the oxidized nitrogen species collectively referred to
as NOY are crucial to the process of O3 and paniculate matter
(PM) formation and fate. Because characterizing the dynamics of
the formation, transport, and fate of tropospheric O3 conveniently
simplifies the more complex interactions among multiple pollu-
tants, tropospheric O3 is being used to illustrate the utility of
high-quality ambient measurements of NOy species and other
important aspects of the urban and regional troposphere for
evaluating numerical photochemical AQMs.
4.2 Air Quality Process Models in
Pollutant Management
AQMs represent the chemical and physical dynamics of
the polluted troposphere using a set of coupled nonlinear partial
differential equations (PDEs) to describe mathematically the
mass conservation equation for each chemical species (Russell
and Dennis, 2000):
(4-1)
fori = l N
where N is the number of chemical species represented in the
AQM, Q is the concentration of species i, U is the wind vector,
KC is the turbulent diffusion coefficient, E, is the source due to
emissions for species i, Rj is the removal term for species i via
various processes (e.g., dry and wet deposition), P, is the
chemical production term for species i, Lj is the loss rate of
species i via gas-phase chemical reactions, and (dC/dt)dollds is the
-------
chemical processes. The set of PDEs is coupled because P| and
Li are functions of the species concentration vector (C). Equation
4-1 cannot be solved analytically, but various numerical methods
can be used to obtain approximate solutions. AQMs generally
use operator splitting to solve each of the terms in Equation 4-1
separately, greatly shortening solution times (McRae et al., 1982)
and improving numerical accuracy, to predict the temporal
evolution of the species concentration fields. Due to computa-
tional constraints, AQMs typically do not save or output the
contributions of each of the processes in Equation 4-1 to the
change in species concentrations; however, AQMs can be
"instrumented" to store this process rate information and make
it available for additional postprocessing and analysis. (Instru-
mentation of the model is described in section 4.5.)
Among the most recent and scientifically advanced AQMs
is U.S. EPA's Models-3/Community Multiscale Model for Air
Quality (CMAQ) (Byun and Ching, 1999), which has been
designed and built specifically to treat problems from a one-
atmosphere perspective. CMAQ and other AQMs were developed
for the purposes of better understanding tropospheric dynamics and
assisting with pollutant management strategies. Both purposes rely
on model predictions of future O3 concentrations ([O3]) that might
result from reductions in emissions of nitrogen oxides (NOX - NO
+ NO2) and/or VOCs (or HCs, the hydrocarbon subset of VOCs),
the two precursors necessary for O3 buildup in the troposphere.
For each purpose—scientific understanding and pollutant
management—the fundamental task of the model is predicting
pollutant concentrations for the nonlinear photochemical system
of the troposphere. However, to advance our understanding of
the photochemical system and to allow us to specify better, more
insightful empirical measures of that system as simulated by the
model, these predictions must be accompanied by causal expla-
nations of how the model produced them. Explanation is in fact
a unique distinction of the numerical process models since there
exists no means for explaining observations using only the
observations themselves. That is to say, there is no currently
available set of measurements that can be used to fully account
for how any one observation at one place came about. However,
analysis of results from a properly instrumented model can be
used to build explanations of the model's predictions and should
be done so as to increase confidence in using AQMs as guidance
for air quality policy making. Without a proper understanding of
the causal mechanisms for the model's predictions available in
this analysis of model processes, there is greater risk for error
when using AQMs in developing and selecting emissions control
strategies. Since it is estimated that the annual cost of compliance
with regulations for managing urban and regional O3 alone in the
U.S. exceeds US$1 billion (U.S. EPA, 1997), the cost of bad
guidance can be very high.
4.3 Need for Diagnostic Testing
In AQM Evaluation
Because of the large economic and social costs of
decisions affecting air pollutant control, we wish to avoid poten-
tial errors where possible by using scientifically advanced AQMs
like CMAQ to provide a realistic simulation of future conditions
and an accurate appraisal of the type and amount of emissions
control necessary to meet mandated air quality goals. This need
to use the models entails adequate testing in an evaluation that
• indicates the validity of the model's scientific formu-
lations,
• assesses the realism of the model's simulations, and
• characterizes the credibility of the model' s realism rela-
tive to its intended applications.
There remains, however, a difficult problem for providing a
meaningful evaluation of the model to guide decision making:
how to tell whether the model is producing seemingly appro-
priate results from incorrect model formulations or bad input
data, i.e., how model performance can appear right for the wrong
reasons.
Evaluation efforts to date have been largely inadequate
for addressing the problems of AQM evaluation for regulatory
use or scientific understanding (Russell and Dennis, 2000). This
inadequacy derives from a general lack of high-quality ambient
measurements in a sufficiently dense spatial and temporal
domain and the concomitant tendency to base the model evalu-
ation on tests using simple statistical measures on outcome
variable residuals (Concentrationpredicted - Concentrationobserved)
only for O3 and, less often, for what is typically reported as NOX.
Because O3 concentrations are a nonunique function of precursor
emissions, and because the system contains several nonlinear
feedback effects that effectively buffer the [O3] (i.e., reduce the
variability of its response to changes in any one part of the
system [see section 4.4]), the realism of model simulations
cannot be usefully evaluated using only comparisons of modeled
and observed ambient [O3] data. Such an analysis cannot test
whether good fits of the model to observations might be due to
compensating errors in the representation of modeled processes;
this greatly reduces our confidence in the usefulness of models
for providing guidance in air quality management decisions.
Inadequate evaluations have significant consequences
for use of the model in that they allow the possibility of finding
a model acceptable for an application when in fact it is not. The
lack of diagnostic tests for a model and its component modules
-------
judged acceptable for several applications when large and
significant errors remained. For example, using the U.S.
EPA-recommended performance evaluation statistics for [O3]
(U.S. EPA, 1991) of
• ±5 to ±15 % bias,
• ±30 to ±35% gross error, and
• ±15 to ±20% unpaired peak prediction accuracy,
urban applications of the Urban Airshed Model version IV (UAM-
IV) (U.S. EPA, 1991) have performed acceptably well even when
errors in the meteorological model produced wind speeds of zero
in all grid layers above the surface (Tesche and McNally, 1995). In
other cases (described in Tesche et al., 1992), data in VOC
inventories used to set up a model for application evaluations were
later shown to be underestimated by a factor of 2 or more, yet, at
the time, model applications were found to be acceptable using
these recommended performance statistics.
Moreover, because model evaluation for air quality appli-
cations is carried out to assist in control strategy development
and selection, a failed model evaluation for applications such as
these leaves the potential for undocumented bias in estimating
the effect of control strategies. Given an undocumented bias in
a model, policy makers could select an incorrect level of future
reductions, or even the wrong type of control—NOX or VOC.
This possibility has been demonstrated in two recent studies
using two different models to simulate the New York domain for
July 1988. A series of sensitivity-uncertainty tests performed
with UAM-IV (Sistla et al., 1996) and with the U.S. EPA
Regional Acid Deposition Model (RADM) (Li et al., 1998) have
shown that emissions and meteorological uncertainties in the
model setup affect final predicted [O3] to the extent that pre-
ferences for control strategies can shift. The high risk of getting
the wrong control strategy has costly economic and social dis-
benefits. The analyses of [O3] time-series plots and residual
statistics that have been the mainstays of model performance
evaluation cannot by themselves reveal such potential biases in
the use of a model. Thus, undiscovered biases present a sub-
stantial negative implication for any evaluation procedure that
uses only residual statistics and other outcome measures.
It is important to note that the examples described above
are failures of evaluation more than of the model: better evaluation
procedures, including enhanced diagnostic testing, might have
detected any errors or problems in the models and the com-
pensating errors in their setup. Hence it is crucial that AQM
predictions be evaluated diagnostically, i.e., in a way that both
assesses the model's predictive performance and reveals why the
model behaves as it does using explanations at the level of the
model's own processes and mechanisms. With diagnostic testing
the operation of these physically based process models is rendered
more transparent, which may increase confidence in their use
(Saltelli and Scott, 1997; Helton and Burmaster, 1996). Diagnostic
testing is in situ testing of a model's processes and can be
performed either internally with one model or across several
models, or, most significantly, in comparisons using specially col-
lected aerometric data that emphasize atmospheric processes (see
for example Parrish et al., 1993; Trainer et al., 1993).
The need for diagnostic testing of AQMs has long been
recognized. Fox's report from an early model evaluation workshop
(Fox, 1984) noted explicitly that improvement in model perfor-
mance is tied to understanding the scientific basis for model
behavior in well-defined physical situations, a point repeated in
Seinfeld (1988). More recently, Tesche et al. (1992) and Reynolds
et al. (1994) have described the need for "stressful" diagnostic
testing in a better, more comprehensive model evaluation method-
ology. And, Arnold et al. (1998) have proposed a revised meth-
odology for both diagnostic and application evaluations that
emphasizes the importance of diagnostic tests and the key role of
a conceptual model of atmospheric processes to guide their
development and interpretation. However, limitations both in the
form of the models (see section 4.5) and in the availability of high-
quality data (see section 4.8) have slowed incorporation of diag-
nostic tests in actual evaluations. Hence, evaluations to date have
focused more on failure analysis of module components than on
true process diagnostics of the full model.
To carry out in situ diagnostic testing of a model's
processes requires first a clear and concise conceptual model of
those processes, one that codifies our understanding of the sys-
tem interactions in basic statements at a very high and general-
ized level. The statements of the conceptual model declare what
we know from theory or experience to be true or must act
instrumentally as if it were true; they provide a generalized
description of what we understand the important chemical and
physical processes and interactions to be. A simplified schematic
for one such high-level conceptual model for the system cycling
of O3 is shown in Figure 4-1. Additional details of the conceptual
model this diagram represents are given just below.
4.4 System- and Process-Level Descriptions
of the Tropospheric Photochemistry of O3
While the conceptual model illustrated in Figure 4-1
provides a general appraisal of the system cycling, it lacks details
necessary to understand the process interactions that constitute
it. For that we require the richer description of processes and
mechanisms at the level of the photochemical processes them-
selves. That description follows here.
Experimental work in environmental smog chambers and
with early numerical models more than 20 years ago showed that
the chemistry of O3 formation and accumulation is highly non-
linear (Dodge, 1977). That is to say, although changing either
NOX or VOC emissions can alter the production of O3 (P(O3)),
changes in P(O3) are not monotonic with changes in the emis-
sions of precursors, especially for changes in NOX. In some
circumstances NOX reductions lead to increases in P(O3) and
ultimately to higher [O3]. To use an AQM for successfully
-------
\l/x
o^
OH Cycle
\HCHO
Initiation
\New
\OH
VOC ^ >
r
missions ~*~
NO ^
missions >
1
Ozo
Produ
k
OH
Propagation
NO2
Ox Production
NOx Cycle
P(Os) J
r
ne
ction
T
e
r
m
i
n
a
t
i
0
n
H202,
Organic
^Peroxides
HNOs,
Organic
Nitrates
Figure 4-1. High-level conceptual diagram of tropospheric photochemistry showing the major elements of
radical initiation, propagation, and termination and the resultant NO and NOX cycles.
r
predicting a photochemical system's O3 response to emissions
changes, we must know both the position of that system on its O3
response surface and the shape of that response surface so that
we can successfully predict the system's sensitivity to changed
inputs. Figure 4-2 shows the response surface of peak [O3] to
various levels of NOX and VOC emissions from a simulation for
Atlanta, GA, made using a trajectory model. (A description of the
model and its setup is found in Tonnesen and Dennis, 2000a.)
The isopleth lines of the response surface are derived by fitting
contours to the peak predicted [O3] in multiple model simulations
using different NOX and VOC emissions. Note that many dif-
ferent combinations of NOX and VOC emissions can produce the
same [O3]. This is equivalently showing that solutions to the
mass conservation equation (Equation 4-1) governing P(O3) are
nonunique and can be reached through many different combi-
nations of the equation's positive and negative terms. This fact
can be seen more easily in the two-dimensional projection shown
in Figure 4-3.
The heavy line drawn across the ridges of the surface
contour lines is the ridgeline of maximum [O3] in the simulation
and divides the response surface into two domains (A and B). In
these two domains, P(O3) is limited in different ways: Below the
ridgeline (domain B), P(O3) is limited by NOX availability, and
reductions in NOX decrease P(O3) while VOC reductions have
little influence; above the ridgeline (domain A) radical avail-
ability limits P(O3), and reductions in NOX can increase P(O3)
while VOC reductions reduce P(O3). The shape of an O3
response surface like the one in Figure 4-2 is in fact determined
by the nonlinearities in P(O3), which change over time as the
NOX and VOC levels change throughout the day and from one
location to another. The O3 system response will also be modi-
fied by perturbations in the model other than in NOX and VOC
emissions levels, such as by changes in physical parameters of
the meteorology driver or in specific details of the chemical
mechanism. Uncertainties such as these in a modeling series may
in fact alter the spacing and shape of the response surface
contour lines, affecting both the change in [O3] throughout the
day and the predicted change in O3 for changing VOCs and/or
NOX. Consequently, the sensitivity of O3 to emissions control of
these two precursors remains a key variable to the vexing prob-
lem of secondary oxidant formation; forecasting that sensitivity
is the central problem for regulators seeking successful control
strategies for O3 and is the motivating factor in probing the
model at the level of its internal systems and mechanistic
processes.
The structure of the O3 response surface in Figure 4-2
depicts explicitly how P(O3) in the troposphere varies with VOC
and NOX concentrations. But, in addition to the concentrations of
-------
r
200
Ozone Response Surface
160
Initial NOx
Concentration
(ppbv)
Initial VOC
Concentration
(ppbC)
Figure 4-2. Depiction of the nonlinear ozone response surface through plotting daily peak ozone
concentrations as predicted by a photochemical box model for a range of initial concentrations of VOCs
and NOX. The heavy line across the top indicates a ridgeline of maximum peak ozone concentrations
achieved with the initial concentrations.
160
Peak [Os] Response Surface (ppbv)
200 400 600 800 1000
Initial VOC (ppbC)
1200
1400
Figure 4-3. Two-dimensional view of Figure 4-2.
-------
these two precursors, P(O3) is governed by the linked initiation,
propagation, and termination of a few important radical species
that preferentially attack either VOCs or NOX as concentrations
of all species in the pollutant mix change through time and space.
The diagram in Figure 4-4 shows a simplified conceptualization
of the relations between the most important radical species and
emitted precursors using the most basic HC, CH4, to illustrate.
Here we describe some additional details of that conceptual
model in brief system- and process-level overviews of the chem-
istry of O3 formation in the troposphere.
P(O3) is initiated by creation of the hydroxyl radical (OH)
through photolysis of O3 or formaldehyde (HCHO) or HONO as
in R1-R5:
O3 + hv -» O('D) + O2
O('D) +H2O -» 2 OH
HCHO + hv -» CO + 2 HO2
H02+NO-»OH + N02
HONO + hv -» OH + NO
Rl
R2
R3
R4
R5
R1-R5 are currently believed to be the major sources of radical
initiation; however, there may be other photolysis and decompo-
sition reactions important as radical sources, including photolysis
reactions of peroxides and the decomposition reaction of PAN or
other unstable intermediate organic species, but there is currently
large uncertainty about these.
Once created, the OH radical can react with carbon
monoxide (CO) or VOCs to produce peroxy radicals (HO2 and
RO2) as in R6. At high [NO], peroxy radicals react with NO to
split an oxygen-oxygen bond, thereby creating an odd oxygen
(Ox) in the form of NO2, as in R7:
OH + VOC
RO2 + NO -
O2 -» RO2
RO + NO2
H2O
R6
R7
The alkoxy radical (RO) produced in R7 can undergo further
reactions that re-create the original OH radical:
RO + O2 -» carbonyl + HO2
HO2 + NO -» OH + NO2
R8
R4
creating a chain of radical propagation reactions in which a
single new OH radical can catalytically mediate the decomposi-
tion of several VOC molecules.
r
HNOs
Figure 4-4. Cyclic reaction of methane oxidation (as a surrogate for hydrocarbon oxidation)
involving OH and conversion of nitric oxide to nitrogen dioxide, termination to peroxides and
nitric acid, and concomitant formation of ozone.
-------
We define Ox as the sum of all species that can act as
reservoirs for atomic oxygen, primarily O3, NO2, and PAN with
minor contributions from some short-lived radical species and
HNO3 and RONO2. In the troposphere, the primary mechanism
for production of Ox (and hence O3) is the oxidation of NO to
NO2 in R4 and R7. A pseudo-photostationary-state (PPS) equili-
brium exists between NOX and O3 such that at high [NO], in
conjunction with radical limitation, the Ox preferentially inter-
converts to NO2 through O3 titration by R9, while at low [NO]
the Ox may be converted to O3 by RIO and Rl 1:
O3 + NO -
NO2 + hv -
(O3P) + O2
NO2 + O2
NO + (O3P)
-» O3
R9
RIO
Rll
The PPS equilibrium relationship is traditionally written as
[03]=jm[N02]/k9[NO]
(4-2)
where j and k represent photolysis and reaction rate constants,
respectively. Equation 4-2 can also be rearranged in terms of
[NO] and written as:
[NO]=j10[NOJ/ki[0J
(4-3)
Note here that RIO re-creates the original NO, and this re-
creation can catalytically mediate multiple conversions of peroxy
radicals to Ox.
P(O3) may proceed by R1-R11 in an autocatalytic cycle
until either OH or NOX is destroyed in termination reactions.
Important termination reactions include
OH + NO2 -» HNO3
RO + NO2 -» RONO2
HO2 + HO2 -» H2O2
HO2 + RO2 -» ROOH
R12
R13
R14
R15
but preference for one or another pathway shifts with changing
concentrations. At conditions with low VOC/NO2 ratios, OH radi-
cals preferentially attack NO2 to produce relatively inert HNO3 as
in R12, and this is the predominant termination reaction for both
OH and NO2 at low VOC/NOX ratios (Sillman et al., 1990). For
systems with high ratios of O3/NO2, the PPS equilibrium causes
[NO] to decrease, thereby favoring the termination reactions R14
and R15 over the propagation reaction R4.
In this process-level view, the photochemistry of O3
formation begins with the initiation of new radicals primarily
from O3, HCHO, and HONO. In rural areas, or NOx-limited
areas, the great majority of new OH comes from photolysis of O3.
The subsequent propagation of radicals through the system as
RO2 and HO2 are formed following OH attack on VOCs, and the
re-creation of the OH radical sets up an OH cycle. The OH
propagation efficiency (PrOH) is defined as the average fraction
of OH re-created for the cycle of OH reactions. In relation, the
OH chain length is defined as the average number of times a new
radical cycles through the system until being removed in a termi-
nation reaction. The OH chain length can be calculated as
OH chain length = 1 + PrOH + PrOH2 + PrOH3 + ...
= l/(l-PrOH)
(4-4)
As Figure 4-1 shows and this process description makes
clear, a companion cycle with a similarly defined chain length is
set up with the conversion of NO to NO2 in which NO is re-
created or cycled until terminated. Note that the number of mole-
cules potentially processed by this cycling over a period of time
can be large when compared to the instantaneous individual
species concentrations, i.e., the rates of production and loss
control species concentrations.
With this basic understanding, the O3 ridgeline on the
response surface in Figure 4-2 can now be explained in terms of
the VOC-NOX-OH cycles. The rate of O3 production is approxi-
mately proportional to the rate of OH attack on VOCs, and this
rate is maximized for conditions that maximize the total rate of
OH production (P(OH)), which is the product of OH initiation
and chain length as in Equations 4-5 and 4-6:
P(O3) = k(OH + VOC)
= OHWllaledxl/(l-PrOH)
(4-5)
(4-6)
Because of the inverse dependence of P(OH) on (1 - PrOH), P(O3)
will be maximized for conditions that maximize PrOH, thereby
forming the [O3] ridgeline. This ridgeline separates the response
surface into two domains and marks a transition area for the
system [O3] response to changes in NOX and VOCs. Figure 4-2
shows that as NOX and VOC levels are changed the photochem-
ical system will change its state and move over this response
surface, often crossing the ridgeline into the other domain. This
change of state as the system moves over the response surface
has important implications for understanding and using an AQM
since the photochemical domains above and below the ridgeline
are substantially different in important ways.
Systems with a low ratio of VOC/NOX emissions (domain
A in Figure 4-2) are found in the domain above the ridgeline where
P(O3) is limited by the availability of radicals. Under conditions of
high [NOX] in this radical-limited domain, NO2 reacts with OH and
terminates to HNO3, as in R12, removing both OH and NO2, which
limits P(O3) by reducing the production of OH. Furthermore,
excess NO in this high NOX region titrates O3 back to NO2, as in
R9, thereby reducing O3 photolysis as a source of new OH for
additional cycling. For photochemical systems in these radical-
limited conditions, the efficiency of P(O3) per NOX terminated is
low, and P(O3) is more responsive to reductions in VOCs than in
NOX. In fact, as Figure 4-2 makes clear, reducing NOX emissions
without concomitant VOC reductions for photochemical systems
in this domain will cause increases in [O3].
-------
Conditions below the ridgeline with high VOC/NOX emis-
sions ratios, however, present a very different case. There, [NO] is
relatively low, allowing the peroxy radicals to self-terminate as in
R14 and R15. This reduces the number of times OH can be
propagated and so lowers the efficiency of P(OX) per radical.
Termination of NO2 by OH as in Rl2 is also reduced because the
higher VOC/NO2 ratio causes NO2 to compete less effectively for
the available OH radicals. Hence in these cases, although the
efficiency of P(O3) per NOX terminated is high, less NOX is
available for reaction, resulting in lower P(O3) and a lower final
[O3]. In contrast to the radical-limited domain above the ridgeline,
in this NOx-limited domain P(O3) is more responsive to reductions
in emissions of NOX than of VOCs. However, although VOC
emission reductions are less effective in reducing [O3] for
photochemical systems in this domain of the response surface, in
no case will reductions in either precursor increase [O3].
4.5 Instrumenting AQMs for
Diagnostic Testing
Although the process details described just above inform
the structure of the numerical model and are the explicit system
pathway components that should be tested diagnostically, they
are generally not available in standard AQM output owing to
computational processing and storage limits. Therefore, it is
necessary to instrument the model code in such a way that these
process details are made available for diagnostic testing and
analysis. Opening up the system processes by instrumenting the
model and then using ambient data to evaluate the accuracy of
the model-simulated component processes is useful because it
helps identify whether the predictions may have been produced
through compensating errors or other unacceptable anomalies in
the model. In this way, the model's explanatory power is greatly
increased, its predictions are more conclusively confirmed, and
confidence in the evaluation is better justified.
The calculated concentrations typically retained at the end
of each process step in an AQM simulation run represent only the
net effect of the coupled processes; no attempt is made to deter-
mine the separate contributions of individual component pro-
cesses to the final predicted result. But, the model can be
instrumented by modifying source code to output the contribution
from each process and making that available with the species
concentrations already saved by the model, thereby making
explicit the model's internal component rates. On a finer mecha-
nistic scale, individual processes such as the nonlinear coupled
photochemistry may also be instrumented separately in order to
learn more precisely the effects of this chemistry on model
predictions. We have applied the techniques of Jeffries and co-
workers (Jeffries and Tonnesen, 1994; Tonnesen and Jeffries,
1994) to several AQMs developed at U.S. EPA including RADM
and CMAQ. Additional details and results of model testing with
the instrumented RADM can be found in Dennis et al. (2000).
Information specific to instrumenting CMAQ can be found in
Gipson (1999); evaluation of the instrumented CMAQ with these
process analysis techniques is currently under way.
4.6 A Taxonomy of Diagnostic Tests for
Tropospheric Photochemistry
Developing diagnostic or process-oriented measures for
understanding model behavior brings together elements and
relations of our conceptual mental model with the explicit repre-
sentation of them in the instrumented numerical model. The pur-
pose here is to devise and assess system-level and process-
oriented probes of both the numerical model and the physical
environment it represents using in situ tests with measures that
can be made in either or, preferably, both systems. We have
developed these measures with process-oriented studies using
theoretical assumptions, model-derived explanations, and results
from instrumented models that can range from 1-D box models
to the full 3-D photochemical modeling system; we note that such
studies must always include the complete 3-D model so that
results are not distorted by the incomplete representations of the
more limited models.
A three-level taxonomy of diagnostic measures can usefully
be constructed using the terms and relations presented in the
schematic diagram of O3 formation and accumulation depicted in
Figure 4-1 and described more fully in Table 4-1.
The first level of the taxonomy holds the individual
components of the photochemical process relations, including
initiation of new radicals from O3 and HCHO and HONO and
termination of radical cycling as either peroxy radicals self-
combine and/or OH and NO2 are converted to oxidized nitrogen
products that no longer react on time scales relevant for the urban
and near-regional settings (NOZ). NOZ is often represented as the
quantity (NOY - NOX). The most important species of NOZ to
define individually are paniculate NO3, HNO3, PAN(s), and
RONO2(s).
The second level of the taxonomy holds the photochemi-
cal process groupings, including the propagation of radicals
through the system as HO2 radicals are formed and convert NO
to NO2, and the related concepts of PrOH and chain length. Recall
from section 4.4 that PrOH is the average fraction of OH re-
created for each OH radical entering the photochemical cycle and
is always less than 1 because termination reactions will remove
some fraction of radicals as well. Recall also from reactions
R1-R11 that the cycle of OH radicals through attack on HCs and
production of HO2 is closely bound to the conversion of NO to
NO2 such that excess availability of one species, NO, ensures OH
propagation, and excess availability of the other, NO2, leads to
OH termination. Linking these two cycles together, PrOH is
approximated by the product of the two halves of these con-
nected cycles that together propagate the OH. This relation can
be represented as
pr ~ f v f (A. 1\
rIOH 'OH+HC A 'HO2+NO V*~')
-------
Owing to the PPS obtained between OH and VOCs and
NOX, the two propagation terms have an opposite dependence on
[NOX]: [NOX] increases the fraction of HO2 being re-created as OH
increases, but the fraction of OH terminating with NO2 also
increases; thus, the fraction of OH attacking HCs to create HO2
must decrease. The net result of the interrelation of these cycles is
that PrOH is maximized for some intermediate level of NOX that
maximizes the product of these two terms. In so doing, it maxi-
mizes the production rate of O3 and creates the [O3] ridgeline.
The third level of taxonomy holds elements for under-
standing the integrated response of the system such as that
represented on the O3 response surface shown in Figure 4-2 and
discussed in section 4.4.
The taxonomy is summarized in Table 4-1. Additional
description for understanding and using these measures as tests
of the model is given in section 4.7.
Table 4-1. Taxonomy of Photochemical Diagnostic Elements
Individual
Component
Aspects
Radical initiation
Radical termination
Competition between termination pathways
Air mass aging
Process
Aspects
OH production
Radical propagation
Radical propagation efficiency, Pr(OH)
OH chain length
NOX chain length
P(O3) efficiency per NOX termination
Response
Surface
Aspects
System state relative to ridgeline
Location of ridgeline in response space
Slope of radical-limited response surface
Slope of NOx-limited response surface
4.7 Diagnostic Model Evaluation Tests
We have proposed (Arnold et al., 1998) and are using
(Tonnesen and Dennis, 2000a and 2000b) two types of diag-
nostic tests for evaluating and understanding the behavior of the
chemical mechanisms in AQMs:
• Process diagnostics that relate to the first two levels of
the diagnostic measures taxonomy
• Response surface diagnostics for interpreting the inte-
grated response of the model at the third level of the
taxonomy
To date we have applied these tests mostly in the model
since process-oriented measurements in the atmosphere are diffi-
cult to make and so are very sparse in space and time. However,
we strongly desire ambient measurements to compare the model
against; hence, we provide here a discussion of the important
measurements we would require to evaluate the model using the
diagnostic tests we have developed and preliminarily explored.
This discussion follows the order and groupings of the model
process taxonomy introduced just above but uses the additional
distinction of local photochemistry and cumulative photochem-
istry over the history of an air mass.
Process diagnostics in the more restricted sense we intend
from this point are specific tests of key pathways and interactions
in the AQM's chemical mechanism. They assess the model's
ability to represent actual atmospheric interactions by examining
pathways and processes in the model, often using special
measurements designed to indicate the activity of such processes
in the atmosphere (see, for example, Parrish et al., 1993; Trainer
et al., 1993).
The first taxonomic level defines a small number of key
photochemical relations that can usefully be summarized as
initiation and termination of radicals. Tests can be identified that,
in various combinations, would allow examination of the process
steps that make up the initiation and termination of radicals in the
system. Examples of such tests and how they illuminate model
processes include the following:
• Radical initiation pathways using comparisons for [O3],
[HCHO], [HONO], [peroxides], j(NO2), j(O3), and
other spectral irradiance measures. Total radical initia-
tion in the system is the sum of individual contributions
from each of the initiating species.
• Radical termination pathways using comparisons of
[HNO3], [PAN], [RONO2], and [ROOH]. Total system
radical termination is the concentration sum of the
species terminated that involve either NOX (i.e., NOZ =
NOY - NOX, or an approximation of NOZ using NO3,
HNO3, PAN(s), and RONO2) or peroxides (i.e., species
comprising ROOH and H2O2).
• The balance between radical initiation and termination.
• Competition among radical termination pathways by
comparing production of HNO3 and RONO2 to that of
H2O2 and ROOH.
• Speciation of NOZ to compare competition among NOZ
termination pathways.
• Speciation and relative fractions of NOX vs. NOY, and
PAN and total NO3 vs. NOY to delineate pathways of
NOX processing in the history of an aging air mass.
-------
Key concepts in the second taxonomic level of process
groupings are PrOH, OH chain length, and NOX chain length.
Combinations of ambient observations can be defined that allow
testing of these process groups; however, describing these tests
requires some additional explanation using details we have
learned using the instrumented model.
Recall from Equation 4-7 that PrOH ~ fOH+HC x fH02+NO'
Using the conceptual model, in situ diagnostic measures can be
derived for these two fractions. This is done by listing all chemi-
cal reactions involving the relevant species and dropping any
minor terms, resulting in two explicit expressions to approximate
the two fractions:
I(NO,t-R02
I(HC,NO,) = -
(4-8)
(4-9)
where I(NO, t-RO2) approximates fHo2+NO and I(HC, NO2)
approximates fon+Hc- ^n Equations 4-8 and 4-9, then, we have a
list of species and rates with which we can approximate PrOH,
and, because of the relation given in Equation 4-5, the OH chain
length as well. While not all species and rates in these two
expressions can currently be measured, with just a few more
high-quality observations of the radicals and NO2, for example,
together with some basic assumptions to cover terms not cur-
rently measured, it should be possible in a small number of
places to begin using this important diagnostic evaluation tool.
The analogous situation for NOX is not as straightforward.
Although the NOX chain length can be approximated by examin-
ing the day-by-day association between O3 and the NOX termina-
tion products NO3, HNO3, PAN(s), and RONO2 for midday
temporal stratifications, this approximation is always attenuated
by unaccounted deposition loss processes. Hence, we still desire
a better test for this important process group.
We note that some process diagnostic tests from these two
taxonomic levels will involve comparisons that reflect both local
photochemistry and photochemistry over the history of an air
mass. However, local photochemistry and the history of an air
mass should be carefully distinguished, and it will be necessary
to develop diagnostic measures to evaluate model treatment of
each. For example, O3 production rates and OH chain length
might be estimated for instantaneous local photochemistry using
measured [RO2], [NO], [NO2], and [VOC], while cumulative
P(O3) and average NOX chain length over an air mass history can
be approximated using [O3] and the ratio [O3]/[NOZ]. Addi-
tionally, the two expressions in Equations 4-8 and 4-9 are among
the process diagnostics and species ratios useful for character-
izing the radical budgets for local photochemistry, for example,
while radical budgets for air mass history might be characterized
with these diagnostic tests:
• [O3]/(2*[ROOH]+2*[H2O2]+[NOZ])
the slope of which over several days' measurements
approximates average OH chain length for an aged air
mass
• 2*[ROOH]+2*[H2O2]+[NOZ]
which describes the major radical termination pathways
in an aged air mass
• 2*[ROOH]+2*[H202] vs. [NOZ]
which compares competing radical termination path-
ways in an aged air mass
Furthermore, ratios of species can be calculated in the
model for comparisons at the surface and aloft to provide addi-
tional diagnostic information on a model's process treatment of
intermediate and product species in an aged air mass. For exam-
ple, chemical dynamics measures calculated using partitioned
NOX and NOY species for aloft and surface concentrations could
provide useful information about air mass aging pathways since
the species aloft would be more susceptible to transport and
dispersion effects and less susceptible to dry deposition effects,
and the chemistry can proceed further to completion there. Also,
RO2-RO2 reactions should be tested in the surface layer of urban
areas, where the reaction has little significance, and again aloft
at low [NOX], where the reaction is significant for [O3] and
[HA].
Other process diagnostic tests for characterizing air mass
aging would use CO data with other ratios of species having
varying lifetimes to derive additional model-estimated chemical
budgets and processing rates. These comparisons would serve as
an aid to interpreting results from specific outcome variable
testing of the chemical mechanism, helping to separate the
influence of chemistry from other physical model processes.
The diagnostic tests from the third taxonomic level, the
response surface diagnostics, test a model's ability to track sys-
temic modulation and generally involve use of indicator species
and ratios of species thought to correlate consistently with VOC-
and NOx-sensitive P(O3) in the model. Measures developed for
the propagation efficiencies from Equations 4-8 and 4-9 can be
adapted to examine the O3 response surface. Hence, with I(NO,
t-RO2) and I(HC, NO2) we now have two explicit gauges of a
photochemical system's position on the O3 response surface and,
consequently, its relative distance from the [O3] ridgeline. These
measures of the sensitivity of O3 to NOX and VOC changes are
characteristics of specific photochemical systems in that each
system may have particular combinations of NOX and VOCs that
determine its position on the O3 response surface. However, we
have observed in the model that individual systems exhibit a
strong diurnal behavior, too, moving from strongly radical lim-
ited in the morning to the transition state near the O3 ridgeline
and on into the NOx-limited domain of the response surface late
in the day. Tracking this air mass history with cumulative diag-
-------
nostics is important for understanding the model's behavior
relative to its chief intended use of prediction for control strategy
development and assessment. This is because, as described in
section 4.4, the position of a system in one or the other domains
of the response surface determines whether VOC or NOX con-
trols are to be preferred.
Several indicators of a system's O3 sensitivity to changing
precursor levels have been proposed and are currently under
development and active testing in several models (Tonnesen and
Dennis, 2000a and 2000b; Sillman, 1995; Chang et al., 1997;
Kleinman, 1994), and many appear promising as diagnostic tests
for model evaluations. The motivation in developing these indi-
cators is the need to have a robust value that tells whether the
model correctly simulates the real-world sensitivity of O3 to
changing precursor emissions. Some examples include [O3]/
[HNO3], [O3]/[NOX], and [O3]/[NOZ], which show strong cor-
relations to O3 sensitivity, correctly predicting conditions either
strongly NOX or strongly radical limited (Tonnesen and Dennis,
2000b). Because these indicator tests bear directly on the chief
policy use of the models, it is important to evaluate their signi-
ficance in the model against values calculated from ambient
observations for a wide area across the modeled domain. We
hasten to add, though, that indicator values calculated from
ambient observations will not replace numerical process models
because observation values cannot reveal the overall effective-
ness of a control strategy nor even whether a particular control
strategy might lead to attainment of an air quality standard as the
model can. But, the indicators are uniquely helpful for diag-
nosing the capability of the models for this important policy goal.
4.8 Priority and Utility of Observations
Needed for Diagnostic Testing
Our work with diagnostic tests in the model suggests that
there can be no single definitive test constructed from any
elements on any level of the taxonomy. Rather, to be useful for
understanding the behavior of the system, diagnostic tests will
need to probe as many different aspects of the physical and
modeled systems as can be done with available analytical tech-
nology. Wherever possible, more than one way of testing a par-
ticular element or process of either system will be preferred. All
possible measures, and hence all species contained in them, will
be needed to develop a distortion-free picture of the true
behavior of the model. Although we have presented species
priorities derived from our taxonomy of diagnostic elements, in
the end we would require that all of the species in the diagnostic
tests described in section 4.7 be measured simultaneously at a
single surface site or aloft in an aircraft and with a response time
significantly less than 1 h, perhaps even approaching the 1- to
5-min range in order to evaluate AQMs diagnostically.
Generating a reduced, ordered list of ambient species
from the large list of species ultimately needed for testing models
is made difficult by the tight interconnections of processes and
feedbacks in the troposphere. This means that it very often
appears that modelers are asking for all species to be measured
everywhere at all times. The taxonomy presented in section 4.6
is our attempt to dispel this appearance and offer a priority
ordering of species ranked according to their utility for process
comparisons and response surface diagnostic testing, both local
and cumulative. In this section, we list the species we would
require for making the diagnostic tests described in section 4.7
and show more specifically how ambient observations of them fit
together in diagnostic tests of the model.
Recall from section 4.2 that the key policy use of the
model is its prediction of O3 sensitivity to NOX or VOC control
for attainment or maintenance of an air quality standard. Like-
wise, recall from sections 4.1 and 4.4 the impossibility of making
only a single change in the tightly linked one-atmosphere concept
that characterizes the troposphere and that any single concen-
tration such as [O3] is a nonunique solution to the mass conserva-
tion equation (Equation 4-1), i.e., that any [O3] can have been
produced through very many different combinations of NOX and
VOCs. Taken together, these suggest that we must understand
both the processes of O3 production and accumulation that lie
behind the final predicted O3 concentration and the potential for
changing production and accumulation with changing NOX and
VOCs if the model's predictions are to be used with confidence.
Hence, it is somewhat disturbing to note that previous
model evaluations (summarized in Russell and Dennis, 2000)
have documented several problems that appear to be common to
many AQMs such as generally underpredicted NOX values and
overly fast NOX aging to NOZ that may be pointing to systematic
errors in model processing pathways or rates, which in turn may
be producing a bias in the model-predicted preference for VOC
or NOX controls. Without better diagnostic tests of the model we
are currently uncertain whether the errors would best be ascribed
• to wrong emissions setups from bad or incorrectly
interpolated data,
• to inaccurate meteorology, both clouds as reactors and
as engines of transport and considerations of incorrect
actinic flux, or
• to the chemistry, possibly maladjusted through inappro-
priate tuning exercises of the chemical mechanism
and/or larger model to fit specific observed [O3].
The diagnostic testing program we present here is one
attempt to narrow this uncertainty by examining process issues
in the chemistry associated with model-predicted preferences for
VOC or NOX control. And the key to these process issues is
understanding the linked OH and NOX reaction cycles as
diagrammed and described above, both for local instantaneous
photochemistry and cumulative chemistry as the air mass ages.
For this reason, the diagnostic tests described here are focused on
-------
understanding O3 production and the production of Ox, which
determines P(O3) as a means for explaining O3 accumulation and
the model's final predicted [O3],
Recall from section 4.4 that O3 is but one component of
Ox and that P(OX) can be defined as the rate of reaction of HO2
and RO2 with NO, minus small losses:
or
P(OX) = k(HO2 + RO2)NO - losses
P(Ox) = P(OH)xfOH_HC
0" initialed X ^OH-HC
(4-10)
(4-11)
U-Pr,
OH '
P(OX) can further be defined as P(O3) plus the rate at which NO
is oxidized to other forms of NOY, or equivalently, P(O3) as the
rate of Ox production remaining after NO has been oxidized.
Thus, our first-priority desired observations are the species
concentrations and reaction rates required to populate the terms
in these expressions and thereby provide a means for diagnostic
testing of P(O3) in the model and the atmosphere. Those priority
observations are
• accurate actinic flux measurements for O3, NO2, and
HCHO;
• accurate concentrations for O3, H2O, HCHO, and HONO;
• accurate concentrations for NO, NO2, PAN(s), HNO3,
RONO2 and total NO3, and NOY (where total NO3 =
nitric acid + paniculate nitrate);
• accurate concentrations for CO and speciated VOCs;
• accurate concentrations for H2O2 and total peroxides;
and
• accurate concentrations for HO2 and RO2 (may be
together with OH).
With this set of measurements we could approximate local
P(OX) using [HO2], [RO2], and [NO] as in Equation 4-10. We
could also approximate local OH chain length using the combi-
nation of concentration measures and rates defined in Equation
4-7, although these may be more difficult to make.
Cumulative diagnostics of the history of an air mass at a
site suitably downwind from the urban core are somewhat more
easily approximated with concentration observations than are
local ones. For example, we can estimate cumulative radical ini-
tiation and termination with 2*[peroxides]+[NOz] and with the
ratio [peroxides]/[NOz]. The cumulative OH chain length at a
downwind site can be approximated with the slope of the ratio
[O3]/(2*[peroxides]+[NOz]) calculated for several days at the
site, while the concentration sum of HNO3, RONO2, total NO3,
and the PAN(s) would approximate a measure of NOX termi-
nation over the air mass history and the sum of [O3], [NO2], and
[NOZ] would serve as a measure of cumulative P(OX). In addi-
tion, using the first-priority observations we could calculate
ambient values for some of the indicators of O3 sensitivity to
emissions reductions as described above, including the ratios
[O3]/[NOX] and [HCHO]/[N02].
We recognize that this first set of desired observations is
quite large, but reiterate that multiple measures are required
because there are significant concerns regarding the usefulness
of any particular diagnostic test. For example, there can be large
uncertainties in the cumulative diagnostics due to the complexity
of the chemistry and other heterogeneous effects. The diagnostic
2*[peroxides]+[NOz] does not provide a reliable measure of
cumulative radical termination because NOZ can be produced by
heterogeneous reactions that do not involve radical termination
(Dentener and Crutzen, 1993) and because both peroxides and
NOZ can be removed from the system by deposition and cloud
processes.
Thus, many diagnostic tests are required to perform a
comprehensive model evaluation, and we would note that most
of the observations listed here are useful in more than one test,
making them very efficient for model evaluation.
In addition, though, a second set of measures, lower in
priority than the first, would provide still more information useful
for testing models and moves down in specificity from the rather
more aggregate measures given in the first-priority set. This
second set includes
actinic flux measurements for HONO and H2O2,
OH,
NO3 radical,
speciated RONO2,
methacrolein and methylvinylketone,
speciated RCHO, and
aerosol mass and size distribution,
which will be useful in directed tests of intermediate chemical
products and in fleshing out our understanding of their processing.
4.9 General Requirements of Observations
for Diagnostic Model Evaluation
One distinguishing characteristic of the diagnostic tests is
that they involve combinations of multiple variables, either
outcome variable concentrations or process variables like reac-
tion rates. Although potential indicators and gauges in new and
different combinations are still being developed, the absolute
number of species involved in all possible measures has now
plateaued, but additional speciation within classes is still desired.
-------
As a consequence, the number of species and process variables
needed to test the models diagnostically is approaching comple-
tion, or, at the least, all general categories of measures we might
desire have been identified. But there remain, in fact, no mea-
surements of process rates in the atmosphere, a key variable type
for many potential tests of the model. We leave aside further
discussion of this topic, but note that, given our understanding of
the chemistry presented in sections 4.4 and 4.7, it appears to be
an area that would repay continued work by modelers and
measurement developers. Even without addressing the problem
of process measures, though, the list of measurements required
for structured and complete diagnostic model testing we gave just
above is a daunting one. Thus, here we consider the questions of
which diagnostic observations can be made at present and what
precision and accuracy we require to use them in tests against the
model.
In recent years new, state-of-the-science research instru-
mentation has been developed and put into production in field
campaigns that is capable of making ambient observations of
many of the key species and variables needed to construct the
diagnostic tests described in section 4.7. This represents a signifi-
cant advance on the part of the analytical community since these
observations are made over very short time scales and often at
very, very small concentrations. Here we have reordered the list
of species we desire for model evaluation according to the
current or near-future viability of making the ambient obser-
vations of them:
• Species or variables easily measured accurately: O3,
NO, CO, HCHO, and PAN
• Species or variables measured accurately, but with
difficulty: NO2, NO3, HNO3, NOY, other PAN(s), H2O2,
HONO, peroxides, some alkyl nitrates, j(O3), and
j(N02)
• Species for which analytical techniques are being devel-
oped and tested and are moving slowly into more
general use, if not production: OH, HO2, total RO2,
speciated RO2, and NO3 radical
• Missing classes:
Alkyl nitrates from biogenic hydrocarbons
Oxygenated hydrocarbons
Process rates
We note that observations of only some of the species in
the first rank are routinely made in current ambient networks, and
it would be our desire to see observations of HCHO and the
PAN(s) more routinely made. We are encouraged, given the
large list of species required for diagnostic testing, that a signi-
ficant number of key species are close to being made more
routinely both at the surface and aloft.
The specificity of these ambient measures is also crucial
for using them successfully as diagnostic tests of the model.
Distinguishing individual NOY species provides a good example,
as can be seen from the number of individual NOY species in the
lists given above. There is concern that observations reported as
NO2 from the installed ambient networks using chemilum-
inescence NOX boxes are contaminated by unknown amounts of
other NOY species due to the nonspecificity of the chemilumi-
nescence reaction and the setup of the monitor. This presents a
problem for using these NO2 values in both process and response
surface diagnostic tests since several of the ratios are designed to
indicate the separate contributions of NOY species and is a
special problem since NO2 is a key species in the transformation
of Ox through the VOC-NOX-OH chains. Thus, only with more
precise NO2 observations will we be able to test the model's
process representation of the OH reaction cycle through the
various forms of NOY, to provide better estimates of air mass
aging (i.e., the extent to which NOX has been converted to NOZ),
and to better estimate the efficiency of P(O3) for each NOZ. For
these tests we also need to directly measure the key constituents
of NOZ. Accurate and precise NO2 measures are also required for
meaningful tests of the proposed indicators of O3 sensitivity to
emissions changes since NO2 appears in the denominators of
several of the ratios including O3/(NO + NO2), HCHO/NO2, and
the OH rate-constant-weighted ratio of VOCs to NO2
For comparisons relating to the efficiency of P(O3), the
currently feasible test is to compare the association between O3
and NOX termination products for midday conditions as defined
by the slope on a scatterplot of O3 versus NOZ. We need to be
able to reliably distinguish differences in slopes of O3 versus
NOZ. If slopes differ by 25% or more we will be concerned.
Hence, we want to be able to distinguish reliably differences on
the order of ±20%. At regional sites for a given value of daytime
O3, a range of 3 ppbv in hourly NOZ with an overall mean NOZ
of 6 ppbv is not uncommon. We can obtain an estimate of how
well we need to know NOZ if we assume O3 is measured very
well, within 1 ppbv, and characterize observations for regional
sites. We also assume that three-quarters of the observed NOZ
range represents irreduceable nonmeasurement error and that the
observations are independent. Then, for a modest number of
observations, to have a chance at distinguishing slope differences
greater than 20%, we need the combined measurement error in
determining NOZ to be between ±15% and ±20%.
The species whose concentrations need to be quantified
in situ to calculate the I(NO,RO2) and I(HC,NO2) indicators are,
by and large, not easy to measure in the field. In fact, some of the
methods are still under development. Thus, it is useful to quantify
to the fullest extent possible from our modeling studies what we
expect will be required of the measurement methods to enable us
to calculate ambient values of these indicators with sufficient
accuracy to resolve ridgeline conditions. For I(NO,RO2), we
require measurement techniques for total RO2+ HO2, NO, and O3
concentrations. Accurate methods are readily available for NO
-------
and O3, and experimental methods are being developed for HO2
and RO2. Conditions that are near the P(OX) ridgeline typically
have [NO] levels of about 1 ppbv and total [RO2] of 10 to 40
pptv, so a lower detectable limit of 10 pptv and accuracy of ±10
pptv for the total [RO2] method would be required to distinguish
the ridgeline conditions.
For I(HC,NO2), we require measurement techniques for
a subset of the total HC mixture and for NO2. In our analysis
using model simulations, we are able to calculate I(HC,NO2)
using concentrations of CO, CH4, alkanes, alkenes, aromatics,
and isoprene. A sufficient number of these hydrocarbon species
are routinely measured at Photochemical Assessment Monitoring
Stations (PAMS) sites (U.S. EPA, 1994) and can be easily
measured in field studies; however, CO and CH4 are not part of
the PAMS suite of measurements. Methods for NO2 need to
provide a true measurement of NO2 without interferences, and
while these methods are not yet routinely available, NO2 has been
measured both directly and indirectly (see Chapter 5). In our
model simulations, NO2 levels for ridgeline conditions typically
range from 1 ppbv in rural areas influenced by anthropogenic
NOX sources to more than 10 ppbv in urban areas. For a rela-
tively low NOX urban cell, an uncertainty of 10% in the [NO2]
measurement would cause the estimate of I(HC,NO2) to range
from 70% to 74%. For a high-NOx urban cell, an uncertainty of
10% in the [NO2] would cause I(HC,NO2) to range from 90% to
92%, so ±10% accuracy [NO2] measurements would be adequate
to calculate values of I(HC,NO2) that could distinguish NOX- and
VOC-sensitive conditions.
4.10 Summary
The process diagnostic tests we propose for the taxo-
nomic levels reflecting both local photochemistry and the history
of an air mass are summarized in Table 4-2. The priority obser-
vations and species concentration measurements are summarized
in Table 4-3.
Table 4-2. Summary of Diagnostic Tests
Local Photochemistry Tests
Table 4-2. Continued
Air Mass History Resultant Photochemistry Tests
Ox production, P(OX)
O3 production, P(O3)
OH chain length
OH production, P(OH)
OH propagation, PrOH
Radical initiation
Radical termination
O3 ridgeline indicators
(NOX vs. VOC preference)
Measure of: fOH+Hc
Measure of: fH02*No
P(OH) x fOH+HC
{P(OH)xfOH+HC}-P(NOz)
1/0 - PI-OH )
OHinllialed x { 1/(1 - PrO
OHnitiated
P(NOZ) + P(peroxides)
'OH+HC
'H02+NO
[03]/[NOX]
, NO2
UNO, t-RO2)
Cumulative Ox production
Cumulative O3 production
Indicative OH chain length
Radical termination—
NOX pathway
Radical termination—
peroxide pathway
Indicative NOX chain length
O3 ridgeline indicators
(NOX vs. VOC preference)
Air mass age indicators
Cumulative daily [OJ + [NO2] + [NOJ
Cumulative daily [O3]
Multiday slope of cumulative
[O3]/{2*[total peroxides] + [NO*]}
or {[HNOJ + partic.[NO3 ] + PAN(s) +
RONO2(s)}
2*{ [ROOH] + [H2O2]) = 2*[total
peroxides]
Multiday slope of daytime O3 vs. NO2
[OJ / [NOX]; and
[HCHO] / [NO2]; and
[NOX] / [NOY]; [HNO3] / [NOY];
[PAN(s)]/[NOY];
[CO] combined with [NOX] or with
[VOC(s)] covering a range of
lifetimes
Table 4-3. Priority Measurements
First Priority
Variables Actinic flux: j(NO2), j(O3), j(HCHO)
Species O3, NO, NO2, NOY, H2O, HCHO, HONO,
PAN(s), HNO3, RONO2 and total NO3, H2O2 and
total peroxides, HO2, RO2, CO, speciated VOCs
Second Priority
Variables j(HONO), j(H2O2)
Species OH, NO3 radical, speciated RONO2, speciated
RCHO, methacrolein, methylvinylketone,
aerosol mass and size distribution
-------
Chapter 5
Methods for Nitrogen Oxides Monitoring:
Discussion and Recommendations
by
W.A. McClenny, U.S. EPA
5.1 Introduction
NOY is defined operationally as the number of NO mole-
cules resulting after passing ambient air through a thermal
converter assuming reduction of all reactive nitrogen oxides to
NO, i.e., the sum of nitrogen atoms in ambient reactive nitrogen
oxides. After accounting for surface deposition in the ambient
air, NOY is the conserved quantity in tracking gaseous NOX
emissions from fossil-fired electric utilities and other NOX
sources. In the absence of NOZ reaction products, nitrogen
dioxide is obtained as NOY - NO = NO2. Many instruments in
the installed instrument base in the U.S. and elsewhere estimate
NO2 by this subtraction and, in so doing, report monitored NO2
as an upper limit on the actual NO2 in ambient air. Since the
NAAQS for NO2 (primary and secondary standards both of
53 ppbv NO2 as an annual average) are not being exceeded in
any locations in the U.S. at this time, this overestimation of
monitored NO2 guarantees that actual NO2 values would show
compliance as well. However, for NOX SIP Call monitoring,
NOY - NO is recognized as consisting of both NO2 and NOZ
where the portion of NOY due to NOZ will depend on the history
of the specific air mass being monitored. Indeed, the quotient
NOX/NOY is often used to suggest the age of an air mass.
5.2 Methods for Total Reactive
Oxides of Nitrogen (NOY)
A recent article by Williams et al. (1998) documents the
results of NOY measurements at a field site in Hendersonville, TN,
near Nashville, during the period 13 June to 22 July 1994.
Scientists from different scientific laboratories (National Oceanic
and Atmospheric Administration [NOAA] Aeronomy Laboratory
[two systems], Environmental Science and Engineering, Georgia
Institute of Technology, Brookhaven National Lab, and the
Tennessee Valley Authority [two systems]) operated instruments
based on thermal reduction of nitrogen oxides to NO followed by
NO detection using NO, O3 chemiluminescence. The thermal con-
verters were of two types. One consisted of a tube of 24-karat Au
used in conj unction with either carbon monoxide (CO) or hydrogen
(H2) as a reductant gas. The other was Mo in mesh form with no
reductant gas required. Concentrations of ambient nitrogen oxides
ranged from 2 to 100 ppbv at this site and were composed
primarily of NOX. Based on these measurements, a number of
conclusions were possible: (1) five of the seven systems agreed
when monitoring ambient air, (2) two systems exhibited problems
that were attributed to either inefficient conversion (particularly of
the NOZ components) or problems with calibrations, (3) some
problems with conversion of NH3 (not a nitrogen oxide) and
variability in conversion of HNO3 were observed when inten-
tionally adding these compounds to the ambient airstream as a
quality control measure, and (4) the average NOY measured by two
Au converters was 5% lower than the average with three Mo
converters. Other observations were that identification of any con-
verter problems (i.e., by spiking ambient air with NOY component
gases or with surrogate nitrogen compounds) should be done
periodically at each installation in order to have a basis for the
post-study corrections to the data. Also, when spiking ambient air
with compounds for converter efficiency checks, the variations in
the ambient signal due to nitrogen oxides introduces uncertainty in
quantitation of the spiked component.
Fahey et al. (1985) and Kliner et al. (1997) have addres-
sed the behavior of thermal converters and offer precautions in
their use. Specifically, after investigating the thermal conversion
of nitrogen oxides for different combinations of reductant gas (H2
and CO) and metal surfaces (Au, Pt, and stainless steel), the latter
concludes that "non-(NOY)i interferences must be individually
-------
assessed for each instrument under appropriate operating and
environmental conditions." Because conversion efficiencies de-
pend on converter surface conditions and other variables, e.g.,
humidity, temperature, ozone concentration, and aging, a "peri-
odic measurement of the conversion efficiencies of the principal
(NOY)| species (e.g., NO2 and HNO3) and, when relevant, non-
(NOY)i species (particularly HCN, CH3CN, and NH3) . . ." is
recommended. The authors caution about the accuracy of NOY
measurements made at rural or remote sites where interferences
from NH3 and other non-(NOY)j components are relatively large
compared to NOX. Harrison et al. (1999) inferred that their
thermal converters may have developed a sensitivity to NHX
species (sum of gaseous ammonia and paniculate ammonium)
that could account for deficits between the sum of individual
NOY species and the larger NOY value.
To be successful in NOY conversion, the thermal converter
must be mounted at the monitor's interface with the ambient air,
typically outside the monitoring shelter. Otherwise HNO3 and other
NOY components are lost to an unknown extent in passing through
the inlet tubing. A thermal converter can effectively destroy any
ozone in the sample stream (Kliner et al., 1997) so that losses due
to NO, O3 reactions in downstream tubing are avoided.
5.2.1 Method Recommendations for NOY
Thermal conversion of nitrogen oxides to NO defines
NOY and the associated analytic method of NO, O3 chemilumi-
nescence is the simplest, most thoroughly evaluated, and most
often used method for measuring the resultant NO. However,
because of the experimentally verified variation of converter
efficiencies for both NOY species and non-(NOY)i species, a
strong program of quality control is recommended under the
direction of a scientist experienced in this area. Considering the
scientific consensus developed in recent field comparison studies
(Williams et al., 1998) and subject to the cautionary guidance
presented in the open literature (Kliner et al., 1997, and Fahey et
al., 1985), this method is recommended for NOY monitoring for
NOX SIP Call applications in urban and suburban locations and
in qualified rural locations. The use of thermal conversion in
rural areas has led to summary statements considering multiple-
day measurements, but can result in values of the sum of
responses from individual NOY species divided by the batch NOY
response, i.e., S(NOY)/NOY, significantly different from 1.0
(Parrish et al., 1993). Obviously, the lower the NOY concentra-
tion, the more important the quality control procedures. In
summary, the limit of detection (LOD) for NOY measurement is
not generally associated with the analytical finish but instead
with the knowledge of the conversion efficiencies for both NOY
and non-(NOY), species over the course of measurements. There-
fore, real-time measurements at the low single-digit ppbv levels
must be substantiated by a strong quality assurance program. As
an indication of the quality of NOY measurements possible, the
hourly averaged data for the best five out of the seven NOY
measurements reported by Williams et al. (1998) in their Table 5
differed by values of 2.1 ppbv at an average of 13.4 ppbv, 1.7
ppbv at 16.2 ppbv, 3.0 ppbv at 21.0 ppbv, and 6.7 ppbv at 26.9
ppbv. These uncertainties are at the upper limit (20%) specified
by modelers (see Chapter 4). Adequate precision is, however,
achievable by any of the methods. Based on information from
this report, the problem with NOY measurement is in the
accuracy, i.e., knowing the conversion efficiencies.
5.3 Methods for Nitric Oxide (NO)
5.3.1 NO, O3 Gas-Phase Chemiluminescence
Nitric oxide in ambient air has been measured by both
point and open-path monitors. One of these methods is the homo-
geneous gas-phase reaction between NO and excess ozone
(Fontijn et al., 1970), one product of which is excited-state NO2
molecules. Subsequent radiative decay of NO2 produces photons
(chemiluminescence) in the visible beginning at 620 nanometers
(nm) and extending into the near infrared. To monitor NO,
ambient air is channeled into a reaction chamber to which excess
O3 is added. The reaction is rapid and contained within a small
volume, typically a few tens to hundreds of cubic centimeters,
and occurs mostly within the reaction cell. The arrangement is
designed as a rate sensor for which, within certain constraints, the
chemiluminescence signal is proportional to the sampling rate
and inversely to the reaction chamber pressure. Since the excess
ozone needed for the reaction is generated using filtered ambient
air, no compressed gases are needed for monitoring (calibration
gases are still needed). Instruments designed on this method are
generally small in size and light in weight. The reaction chamber
design and operating parameters for efficient generation and
collection of photons have been optimized as both a research tool
and a commercial instrument (see, for example, Steffenson and
Stedman, 1974; Delany et al., 1982; Dickerson et al., 1984;
Drummond, 1985), and the instrument is widely used for ambient
monitoring. The method is a benchmark of air monitoring tech-
nology and, along with the similar chemiluminescence method
for ozone, marked EPA's departure from predominantly wet
chemical techniques for monitoring trace gases. Any other tech-
nique for ambient NO monitoring must be competitive.
During ambient monitoring, care must be taken to avoid
NO losses in inlet sampling lines due to the gas-phase reaction
with ambient O3 and to the reaction of NO at the tubing wall.
These considerations are treated in some detail in Appendix A,
where it is shown that a 2-s or less residence time should limit
losses from the gas-phase reaction to 10% or less even at the
highest ozone concentrations typically encountered. Wall losses
can be determined by the use of standard additions of NO to
ambient air entering the sampling manifold. Accumulation of this
data at times when different ozone levels exist in the ambient air
provides a database from which the combined effect of gas-phase
reactions and wall reactions are defined. By calculating losses
expected from the gas-phase reaction and comparing them with
-------
the experimental total first-order loss rate of NO, the effect of
wall reactions can then be inferred (Fehsenfeld et al., 1990).
In most sensitive NO chemiluminescence monitors, a pre-
reaction chamber is used. Its purpose is to allow the NO, O3
reaction to occur just before reaching the reaction chamber and
to use any residual signal from the reaction chamber as back-
ground for the subsequent measurement of NO. An estimate of
the LOD (SNR = 2) for research instrumentation is 10 pptv for
a 1-s integration time and ±10% of the measured concentration
well above the detection limit (Fehsenfeld et al., 1987).
5.3.2 Differential Optical Absorption
Spectroscopy
DOAS research systems for NO have components like
those discussed in Appendix B, Determination of Atmospheric
Concentration of Nitrogen Oxides by Differential Optical
Absorption Spectroscopy. The procedure and system elements
for determination of NO or any other reactive nitrogen oxide are
similar to that for NO2. However, since NO is typically detected
with wavelengths in the short UV wavelength region, a com-
bination of low source light intensity in this region and the
attenuation of radiation by molecular oxygen in this spectral
region limit the path length. Operation of the system involves
acquiring an oxygen spectrum for the path length used during NO
measurement in order to correctly account for the oxygen
attenuation. Literature provided by a commercial manufacturer
OPSIS AB (Furulund, Sweden), indicates a detection limit of 1-2
ppbv over a 200-m path length with an averaging time of 1 min.
Optical components that transmit/reflect efficiently in the lower
wavelength UV region must be used.
5.3.3 Tunable Diode Lasers
Tunable diode laser systems have been used as reference
methods in field study programs because these systems offer an
unambiguous identification of NO through its unique spectral
absorption features. More information on tunable diode laser
systems is given below in section 5.3.5.
5.3.4 Other Point Monitoring Techniques
Other techniques have been used such as LIF, two-tone
frequency-modulated Spectroscopy (TTFMS), and passive sam-
pling tubes for sampling combined with an analytical finish
(usually ion chromatography). These techniques are covered for
the pre-1990 period by Sickles (1992). The simple and inexpen-
sive passive sampling devices have become widely used for time-
integrated sampling of trace gases and are of considerable utility
in establishing pollutant distributions across major urban areas
for the purpose of locating the best monitoring locations (Yarns
et al., 1999) even though these techniques cannot be used for
diagnostic testing. For the most recent work in these areas see
relevant citations in the references (Chapter 7).
5.3.5 Method Recommendations for NO
The simple and sensitive NO, O3 chemiluminescence
technique is well researched and established as the best com-
mercial technology for ambient air monitoring of NO. Ground-
based instruments with sufficient sensitivity for ambient-level
monitoring at any tropospheric location are commercially avail-
able, although research-grade instruments are often assembled
from components for special applications such as for tropo-
spheric aircraft flights. DOAS technology is available for NO but
a lower detection limit at the sub-ppbv level is difficult to obtain.
5.4 Methods for Nitrogen Dioxide (NO2)
NO2 is a criteria pollutant in the U.S. for which the
NAAQS for health and welfare is an annual average of 53 ppbv.
As a result of the NAAQS requirements for monitoring, the
commercial incentive for instrument development has spurred
innovative development of instrumentation, resulting in many
techniques including both point and path monitors. The point
monitors measure NO2 directly by fluorescence, by absorption in
the visible or IR, or by luminol chemiluminescence. NO2 is
monitored indirectly by NO, O3 chemiluminescence after NO2
conversion to NO with thermal or photolytic converters. DOAS
open-path instruments use a relatively interference-free spectral
region in the UV/visible to measure NO2 directly.
As noted earlier in Section 5.2.1, the gas-phase reaction
of NO and O3 in the manifold leading from the ambient air to the
location of a point monitor results in the creation of NO2 and the
loss of NO and O3. As noted there, a 2-s or less residence time
ensures that the loss of NO is less than 10% due to its gas-phase
reaction with ozone. Since NO is typically less than NO2, the
gain in NO2 concentration will also be less than 10%.
5.4.1 NO, O3 Gas-Phase Chemiluminescence
Thermal converters. The installed instrument base for
measuring ambient nitrogen dioxide in the U.S. consists mainly
of chemiluminescence NO monitors equipped with NO2-tO-NO
thermal converters. The NO2 concentration is estimated by sub-
tracting the instrument response when bypassing the converter
from the response when passing the air sample through the con-
verter. These instruments provide data that systematically over-
estimate NO2. There are two reasons: (1) thermal converters
convert other reactive nitrogen species as well as NO2, as noted
in Section 5.2; and (2) the NO, O3 reaction increases the NO2/NO
ratio as the sampled air mass is passed through inlet tubing or
through addition volumes (e.g., photolytic converters) before
reaching the detection chamber. When only one reaction chamber
is time shared between measurements of NO and NOY, the
difference NOY -NO is taken between sequential measurements.
Hence temporal variability in response can result in errors when
the difference NOY - NO is used for NO2.
-------
Even non-NOy compounds such as NH3 and amines can
be converted to NO if converter temperatures are excessive
(Williams et al., 1998). The NO2 is estimated by subtracting the
signal due to NO (bypassing the thermal converter) from the
signal obtained from an airstream passed through the thermal
converter (NO2 + NO + NOZ). The largest percentage errors in
NO2 occur typically in the afternoon of warm, sunny days when
the photochemical conversions of NOX to oxidation products
such as HNO3, HONO, PAN, and other organic nitrates are
optimum. A large percentage of these compounds are converted
to NO in the chemiluminescence monitor's thermal converter.
Although special inlets using nylon filters to remove nitric acid
and nitrate as well as diffusion scrubbers to selectively remove
acid and/or basic nitrogen-containing gases are available to
condition the sample stream prior to entering the thermal con-
verter, the vast majority of instruments do not have this feature.
In fact, the use of an external thermal converter (converters
mounted outside monitoring stations and using extremely short
inlet lines) ensures high-efficiency conversion of reaction pro-
ducts of NO and NO2 and minimizes losses of these species to
wall adsorption so that an accurate measure of NOY can be
obtained.
Photolytic Conversion of Ambient NO2. Photolytic
conversion of NO2 using wavelengths below 400 nm is being
perfected with respect to choice of light sources and other design
features that allow efficient conversion while limiting undesirable
consequences such as NO, O3 reactions and excessive heating in
the photolysis cell. Eric Williams of NOAA addresses these
considerations in Appendix C and reviews the different designs
that are currently in use. The implementation of photolytic
conversion in research instrumentation has already been realized
in several laboratories, and a commercial version has been
marketed by Eco Physics (Diirnten, Switzerland) as part of a
NOX monitor for several years. Current research and routine
monitoring needs for a specific and sensitive monitor of ambient
NO2 is spurring additional development. The prospect of a
photolytic converter that can be added to existing chemi-
luminescence NO monitors appears to be the most practical
current approach to obtaining a near-term ambient air database
for NO2 concentrations.
5.4.2 Luminol Chemiluminescence
As noted by Maeda et al. (1980) and Wendel et al.
(1983), NO2 reacts with luminol in water solution and the
resulting compound emits visible radiation around 465 nm,
resulting in a real-time, direct measurement. In one version of the
method, the reaction takes place in a lightproof chamber, on the
surface of a wick saturated with the water-based luminol solution
and mounted in front of a photomultiplier tube. The solution
continuously moves from one reservoir down the vertically
mounted wick and into a second reservoir. This reaction is the
basis for commercial Luminox instruments available from Scin-
trex, (Toronto, Canada). The instrument provides a direct mea-
surement of NO2 but is subject to interferences from other com-
pounds such as O3 and PAN. Additives to the luminol solution
are effective in selectively suppressing response to interferences
and enhancing response to NO2. Schmidt et al. (1995) describe
a selective ozone scrubber for application in ambient NO2 mea-
surements using the Luminox instrument. The reaction between
NO2 and luminol is nonlinear and must be corrected for measure-
ments of NO2 below a lower concentration level, typically near
2 ppbv. Kelly et al. (1990) evaluated the performance of the
Luminox, Model LMA-3, and noted that to obtain accurate
measurements of NO2 at low concentrations, the readings must
be corrected for zero offset, nonlinearity, and ozone and PAN
interferences, in that order. Fehsenfeld et al. (1990) describe a
comparison among NO2 monitors, one of which was the LMA-3,
at a remote site near Niwot Ridge, CO. Corrections for O3 and
PAN interferences in the LMA-3 were noted to be "sufficiently
consistent that they could be corrected for by using the measured
values of O3 and PAN down to about 0.3 ppbv NO2." Gaffney et
al. (1998) and Gaffney et al. (1999) have demonstrated the use
of a luminol-type detector for NO2 and peroxyacyl nitrates.
5.4.3 Laser-Induced Fluorescence
LIF was used successfully in the 1970s by combining
fixed-frequency lasers and high-discrimination ratio liquid filters
(Gelbwachs et al., 1972). Photons resulting from the de-
excitation of NO2 molecules were viewed at right angles to the
laser light path through the liquid filters, and sensitive photon
counting techniques were used. Flashlamps replaced the fixed-
frequency lasers (argon ion laser or helium-cadmium laser) as a
light source in later research by this group (Fincher et al., 1977).
However, neither approach proved to be as commercially viable
as the thermal conversion/NO, O3 chemiluminescence. LIF has
now been revived using practical, high-power, tunable laser
sources that permit the matching of light output and the maxima
in the NO2 absorption features. Thornton, Wooldridge, and
Cohen discuss the latest advances for this technique in Appendix
D of this report in a text arranged through the efforts of Dr.
Cohen. Examples are given of recent successful comparison
testing against the NO, O3 chemiluminescence monitoring tech-
nique with a photolytic converter as part of the SOS '99 study in
Nashville, TN. Projections in this discussion include the ultimate
achievement of a version for routine use with a detection limit of
10 pptv.
5.4.4 Differential Optical Absorption
Spectroscopy for NO2
The successful use of DOAS for monitoring NO2 in the
troposphere has been demonstrated using both commercial
instruments and instruments assembled from individual com-
ponents (Edner et al., 1993; Plane and Nien, 1992; Febo et al.,
1996). Stevens et al. (1993) compared the measurement of a
-------
commercial DOAS (OPSIS, Furulund, Sweden) to EPA-
approved fixed-point methods. In the U.S., OPSIS has achieved
equivalency status to the reference technique for monitoring NO2.
The achievement of equivalency means that the instrument meets
certain standards of performance and that data obtained with the
instrument can be reported by state agencies in satisfaction of the
reporting requirements for the NAAQS. McElroy et al. (1993)
describe a method for on-site calibration of the OPSIS for NO2
and other gases.
The commercial instruments using DOAS were developed
using certain proprietary operational techniques. For this reason,
the specific algorithms for the estimation of uncertainty for NO2
and other reactive nitrogen species is not generally known in detail
and the transmission spectra are not necessarily available to the
operator for customized postprocessing. However, the estimated
uncertainty is understood to be based on the residuals obtained by
fitting field spectra to a set of stored reference spectra, including
the target gas spectra and known interferences. High estimated
uncertainty can indicate unexpected spectral interferences, a
change in wavelength calibration (although this can be periodically
corrected), or other problems. Nonlinearity in response to NO2
concentrations is also a potential problem at the spectral resolution
used in typical DOAS instruments. This is an inherent problem
with instruments that use spectral transmission data to calculate
trace gas concentrations (Russwurm and Phillips, 1999). Unfortu-
nately, the extent of this problem cannot be determined without
specifics concerning the algorithm used for signal processing. In
commercial units, real-time data processing is emphasized and a
concentration measurement with estimated uncertainty is the output
of the instrument. The detection limit for NO2 as stated in the
commercial literature (OPSIS, Furulund, Sweden) is estimated at
1 ppbv using a 500-m total optical path length. Uncertainty in the
measurement must be less than ±10% to meet the requirements for
diagnostic modeling. Appendix B of this report authored by Dr.
Jochen Stutz addresses the scientific approach to monitoring of
nitrogen oxides by DOAS.
5.4.5 Tunable Diode Laser Absorption
Spectroscopy—Middle Infrared
The diode laser has been used as the radiation source in
the detection of NO2 and other nitrogen oxides by using the tech-
nique of second derivative absorption spectroscopy. Typically,
the ambient sample is drawn into a glass cell with special optics
to direct the tunable diode laser (TDL) radiation in a folded path
of length up to 100 m (Schiff et al., 1983) or used over an open
ambient path (Chancy et al., 1979). The pressure inside the cell
is reduced (to the order of 25 torr) to minimize pressure-
broadening effects of the absorption line. The TDL is a semi-
conductor device whose band-gap energy is temperature depen-
dent and whose output frequency is determined by joule heating
losses. The laser is held at low temperatures by closed-cycle
coolers and tuned by resistive heating of the diode. The output
frequency is tunable over several tens of wave numbers by
varying the current passing through the diode; it can be made to
scan rapidly and repetitively (1 KHz) over a given wave number
range by providing the appropriate current control. Very high
specificity is obtained because the output frequency has an
extremely narrow bandwidth on the order of 10"3 wave numbers
and therefore does not degrade the shape of the gas absorption
line being monitored. This allows a single absorption feature of
the target gas to be selected, generally removing the effects of
other interfering species.
A sensitive, fast response detector is used to generate the
analytical signal. The laser wavelength is maintained at an absorp-
tion maxima by the use of a reference signal generated by passing
a small fraction of the beam through a reference cell containing the
gas of interest. As the beam is tuned across the absorption feature,
it is slightly modulated in wavelength by modulating the diode
current. This generates a so-called first-derivative signal that
switches signs as the laser is tuned past the line center. This zero
crossing generates a signal that is used as a reference to keep the
laser tuned to line center. The strength of the second-derivative
signal after the beam passes through the multipass cell is calibrated
and used to measure the target gas concentration in the sample.
Ried et al. (1980) indicate that this type of system can achieve
sensitivities on the order of 100 pptv.
Since the TDL provides near unambiguous detection of
target gases, it has frequently been used as a reference method in
field test comparisons. Examples of its use for NO2 include a
comparison with the LMA-3 direct NO2 monitor at a field
monitoring site (Russwurm, 1988) and other pre-1991 research
(Sickles et al., 1990) including an intercomparison of NO2
measurement techniques such as the TDL by Fehsenfeld et al.
(1990).
5.4.6 Systems Using Visible/Near Infrared
Radiation Laser Sources or LEDs
Allen et al. (1995) and Sonnenfroh and Allen (1996)
reported the application of single-longitudinal mode, room-
temperature, semiconductor lasers operating in the visible and
near IR to in situ monitoring of NO2. They report a system design
that uses the relatively new balanced ratiometric detector (BRD),
a novel electronic laser noise-canceling technique. They project
a sensitivity for ambient operation of approximately 1 ppbv for
a 10-m path assuming a detectable absorbance of 10"6. Mihalcea
et al. (1996) report the use of two photometers, one using a
commercially available InGaAsP diode laser with output near
670.2 nm and the second using radiation from a prototype
frequency-doubled GaAlAs diode laser source with output near
394.5 nm. These systems were used to measure the temperature
and pressure dependence of NO2 absorption at those wave-
lengths. The experimental results suggest a detection limit of 10
ppbv for a hypothetical system using the lower wavelength and
assuming a detectable absorbance of 10"5. Fetzer et al. (1998)
also used a frequency-doubled GaAlAs diode laser for demon-
strating a capability for NO2 monitoring.
-------
Fetzer et al. (1998) demonstrated the use of an NO2
photometer using a GaN light-emitting diode (LED) as source of
450-nm radiation and achieving a detection limit of 1 ppbv with
an instrument response time of 1 min. No comparable com-
mercial unit is known to be available at the present time.
However, Jung and Kowalski (1986) modified a commercial
ozone photometer to a NO2 photometer by changing the lamp
source, optical filter, mirrors, detector, and the absorption cell
path length. The unit was used to measure NO2 at the South
Coast Air Quality Management District's air monitoring station
at Pomona, CA, and compared with a commercial chemilum-
inescence monitor.
5.4.7 Method Recommendations for NO2
Photolytic conversion of NO2 to NO followed by NO, O3
chemiluminescence for concentrations across the full range of
ambient NO2 concentrations has been demonstrated using com-
mercially available units. However, unless the time spent in the
inlet sampling line, including the photolytic converter, is 2 s or
less, an O3 monitor must also be included to provide information
to correct NO2 readings for conversion of NO to NO2. Most
chemiluminescence systems in the installed instrument base in
the U.S. have thermal converters and provide an upper limit on
NO2 as explained above. Since the NAAQS for NO2 is currently
not being exceeded at any location in the U.S., criteria moni-
toring is not impacted. However, NO2 monitoring for model
evaluation requires a more selective converter.
A successful, commercially available method for NO2
monitoring in urban and suburban areas is the UV DO AS system.
NO2 is measured directly and without sampling. No sampling
artifacts are possible and there are no known interferences.
Although the exact method of treating uncertainty depends on the
manufacturer, the use of residuals in a least squares fitting
procedure across a portion of the absorption spectrum provides
a measure of uncertainty. Additional understanding of the uncer-
tainty of the DOAS measurement is needed.
The best available technology for the full range of NO2
concentration is the LIF technique, a technique that has high
commercial potential, but is not yet being manufactured by a
commercial vendor. LIF is specific for NO2 and measures NO2
directly. Sample transport from the ambient air to the detector is
subject to the NO, O3 reaction in the sample lines so that the
same constraints on residence time as for any point monitor are
relevant. Photometric measurements of NO2 in the visible look
promising if the new balanced radiometric detector and room
temperature tunable diode laser sources are used. The question
of whether the photometric approach is viable for practical
monitoring of ambient NO2 must be answered with additional
research.
5.5 Methods for Monitoring Nitric and
Nitrous Acids (HNO3 and MONO)
HNO3 and HONO are frequently used for diagnostic
testing of AQMs as noted in Chapter 4. HNO3 is produced during
the day as a reaction product of OH and NO2 and peaks during
the daytime. Because of the high deposition rates of HNO3, this
deposition constitutes the primary termination step for NOX
emitted into the atmosphere. HONO has a short lifetime in the
atmosphere during the day because it is readily photolyzed (OH
is formed as one product) and shows a minimum during the
daytime with a maximum at night. There may be some daytime
sources of HONO maintaining a very small concentration at
steady state. It has been argued that hydrolytic production of
HONO from NO2 is catalyzed by soot surfaces (Zellweger et al.,
1999; Ammann et al., 1998).
In the past, time-integrated systems for sampling and
accumulation of nitric and nitrous acids have been widely used.
These systems use either filter packs or a combination of coated
diffusion tubes and filter packs for sampling. However, the
samples must be transported to a laboratory for analysis. Many
of these systems typically collect samples over a time period
exceeding the time resolution required for performing the diag-
nostic tests mentioned in the Introduction. These techniques will
not be considered for diagnostic testing of AQMs but nonetheless
can be important in establishing comparability among different
methods for measurements averaged over a longer time period.
For information on the denuder/filter pack techniques, the reader
is referred to the results of five pre-1991 comparisons of nitric
acid techniques, including the Atmospheric Environment edition
(Vol. 22, No. 8) devoted to the results of the "Nitric Acid
Shootout" that took place in 1985 in Pomona, CA. Other refer-
ences are Anlauf et al. (1985), Spicer et al. (1982), Walega et al.
(1991), and Sickles et al. (1990).
5.5.1 Thermodenuders
Thermal denuders with selective coatings such as tungstic
acid (Braman et al., 1982; McClenny et al., 1982) have been
used for near real-time monitoring with outputs at one-half hour
or less. Using a tungstic-acid-coated tube as a denuder, nitric
acid chemically combines with the acid and accumulates at the
surface of the tube. Subsequent heating of the tube releases NOX,
which is passed to a sensitive NO, O3 chemiluminescence detec-
tor. Roberts et al. (1987) report the lack of significant inter-
ferences by NO2, HCN, PAN, and n-propyl nitrate under most
conditions, although the extent of interference becomes greater
as the atmospheric humidity is reduced. In field tests conducted
at a rural site, comparisons with a nylon filter collection tech-
nique gave significant differences at the sub-ppbv concentration
-------
levels, although the interfering gases were not identified.
Klockow et al. (1989) describe an automatic thermodenuder
system for measurement of HNO3 and ammonium nitrate in air
for a one-half hour sampling period. The system has two denud-
ers in series. Both are coated with MgSO4 and the first is
operated at room temperature, while the second is operated at
150 "C. Gaseous nitric acid is collected in the first denuder and
ammonium nitrate is decomposed and captured as nitric acid in
the second. Subsequent heating to 700 °C liberates HNO3 and
HONO as NOX, which is detected with a standard NOX chemi-
luminescence monitor.
5.5.2 Dual-Channel Chemiluminescence
Monitors
Commercial chemiluminescence monitors for NO have
been modified to design real-time nitric acid detectors using two
inlets, one with only a particle filter and the second with a
particle filter and a nylon filter. Sample air entering the monitor
from both channels passes through a thermal converter before
reaching the chemiluminescence reaction chamber. The differ-
ence signal is attributed to HNO3, including that formed from the
decomposition of ammonium nitrate. In some instances, HNO3
may be removed by substances already present on the particle
filter. Tanner et al. (1998) measured the sum of nitrate and nitric
acid by using a system comprising two inlet channels, each
containing a gold-CO catalyzed converter to reduce all odd
nitrogen species (NOY) to NO. One channel also contained a
nylon filter to capture the sum of nitrate PM and nitric acid. The
NO resulting in each channel is detected by NO, O3 chemi-
luminescence. This type of system has been used in an inter-
comparison of HNO3 methods (Spicer et al., 1982) and aboard
aircraft for studies of plume chemistry during the 1995 SOS
summer field study in Nashville.
5.5.3 Wet Denuders
Nitric and nitrous acids are often sampled by capture in
water followed by ion chromatographic analysis. Capture of
nitric acid by a diffusion scrubber based on a wet anion exchange
membrane and determination by UV detection was reported more
than a decade ago (Dasgupta and Philips, 1987). However, the
method was insensitive and not sufficiently selective. A subse-
quent study (Vecera and Dasgupta, 1990) demonstrated ambient
measurements of nitrous acid with 7- to 15-min time resolution
and a detection limit of 20 pptv. A porous membrane diffusion
scrubber was used for collection and ion chromatography with
UV detection for analysis. However, the diffusion scrubbers of
this type did not collect quantitatively and calibration efforts
were required. Vecera and Dasgupta (1991) introduced the use
of a wet denuder, rudimentary low-pressure ion chromatography,
and postcolumn colorimetric reaction detection for the determi-
nation of HONO and HN03 with an 11-min time resolution and
80- and 230-pptv detection limits, respectively. More efficient
parallel plate wetted denuders were introduced (Simon and
Dasgupta, 1993), which make this approach more sensitive and
competitive with a more expensive ion chromatographic finish.
However, the diazo coupling chemistry uses a copperized cad-
mium reactor for reducing nitrite to nitrate and that, although a
standard technique used in water analysis laboratories, may not
be considered as environmentally friendly. Taira and Kanda
(1993) describe a wet effluent diffusion denuder that incorpo-
rates a 0.16-cm-i.d. straight glass tube as the denuder. Buhr et al.
(1995) disclosed an automated system for nitric acid, particulate
nitrate, and particulate sulfate. The method uses a semicon-
tinuous wet effluent denuder and a wet effluent frit for sampling,
followed by ion chromatography with conductivity detection.
Sample frequency is 15 min and a detection limit of 10 pptv is
achieved for each compound. Nitric acid measurements over an
ambient range of <0.1 to 4 ppbv were compared with a filter
pack method and excellent agreement was obtained. Near real-
time monitors with short cycle times on the order of 15-30 min
have been reported by Simon and Dasgupta (1995a) for HNO3
and HONO using a wet-wall parallel-plate denuder and a
customized ion chromatograph. The most recent version of this
type of monitor is a commercial prototype that was used in the
summer 1999 Atlanta SOS study by Dr. P.K. Dasgupta of the
Texas Tech University. This latest version differs from the unit
described in the open literature in the method for collection of
airborne PM. In the newer version, an in-line particle filter is
placed downstream of the wet-wall denuder; periodic automatic
extraction of this filter occurs followed by ion chromatographic
analysis. Figure 5-1 shows the variation of HNO3 and HONO
with a 10-min resolution as measured using this system during a
3-day stretch in the Atlanta SOS study during the summer of
1999. As expected, the diurnal maxima of HNO3 occurs in the
daytime whereas HONO disappears during the day and becomes
important at night.
Keuken et al. (1988) developed a wet-wall denuder
formed between concentric cylindrical tubes mounted about a
horizontal axis. The walls are wetted by maintaining a supply of
water in the lower section of the cylindrical annulus and con-
tinuously rotating the cylinders. As ambient air is passed through
the water-free portion of the annulus, soluble ions accumulate in
the water reservoir and are periodically analyzed by ion chroma-
tography. Oms et al. (1996) discuss the use of a later version of
this approach in which analysis of HNO3, HONO, HC1, and SO2
are automated by computer control at a frequency of 2/h. Detec-
tion limits of 2 and 12 ng/m3 are stated for HONO and HNO3,
respectively. A commercial version of this unit has been con-
structed by Anderson Inc., Atlanta, GA. Kanda and Taira (1992)
reported the measurement of HNO3 and HONO as NO after
capturing the two gases in a NaOH solution and chemically
converting the two to NO.
-------
Date - Time
Figure 5-1. HNO., and HONO with 10-min resolution as measured using a wet-wall denuder
system for a three-day stretch in the Atlanta SOS study during the summer of 1999 (Dasgupta,
Texas Tech University, Lubbock, TX).
5.5.4 The Mist Chamber Technique
Cofer et al. (1985) describe a mist chamber technique.
Essentially, water soluble components of ambient air are cap-
tured by mixing water droplets with the ambient air. The droplets
are created inside a collecting vessel by using the venturi action
of sample air passing over a water-filled capillary, thereby gene-
rating water particles with high surface area to volume ratio. The
diffusion of water soluble gases to the particles results in the
capture of the gases. The droplets are subsequently collected by
impaction against a filter and analyzed by ion chromatography.
This system is typically operated with cycle times greater than
one-half hour. Talbot et al. (1990) expanded on the use of the
mist chamber method and examined differences between results
using the mist chamber technique and those using the nylon filter
technique as observed in the field experiments.
5.5.5 Chemical lonization Mass
Spectrometry
The most recent addition to real-time measurement of
nitric acid is the chemical ionization mass spectrometry (CIMS)
technique as reported by Mauldin et al. (1998) and by Huey et al.
(1998). These instruments differ in design and in the reagent ion
used. Fehsenfeld et al. (1998) has documented the comparison of
these two systems and a filter pack technique. Apparent decom-
position of ammonium nitrate particles to nitric acid and ammo-
nia on the front filter of the filter pack released nitric acid to be
captured as nitric acid and hence biased the filter pack results
high. Scatterplots of simultaneous measurements by the two
systems indicate a roughly symmetric scatter with about 1:1
agreement and most comparisons fall within ±30% of being iden-
tical across the ambient measurement range of 2-1000 pptv (see
Figure 4 in Fehsenfeld et al., 1998). CIMS detection limits of
less than 15 pptv for a 1-s integration period are cited. The fast
response of the system is particularly well suited to requirements
of airborne monitoring.
5.5.6 Tunable Diode Laser Absorption
Spectroscopy and DOAS Techniques
Optical techniques have often been used for measuring
nitric and nitrous acid. Real-time systems employing tunable
diode laser absorption spectroscopy (TOLAS) (Schiff et al.,
1983) and another using an open-path, multipass Fourier trans-
form infrared (FTIR) system (Tuazon et al., 1981) have been
successfully operated for monitoring nitric acid and are parti-
cularly useful as reference methods in areas where nitric acid
concentrations achieve ppbv concentrations. The TOLAS has a
detection limit given as 4 ppbv for nitric acid, while the best
FTIR systems provide measurements to 10 ppbv. Commercial
DOAS systems are used for monitoring nitrous acid with a
quoted detection limit of 0.5 ppbv under standard conditions for
a monitoring path of 500 m and a measurement time of 1 min
(OPSIS, Furulund, Sweden). Febo et al. (1996) have measured
HONO with a commercial DOAS system at values exceeding 10
ppbv in Milan.
5.5.7DNPH Derivatization and HPLC Analysis
Zhou et al. (1999) have demonstrated a derivatization
technique for HONO based on DNPH derivatization and HPLC
-------
analysis with UV detection at 309 nm. MONO is scrubbed from
ambient air by passing the sample through a coil sampler along
with an aqueous scrubbing solution. Various potential inter-
ferences have been examined and results indicated a 1:10,000
interference equivalent for NO2 and 1:1000 for PAN. Limited
comparisons with a bubbler/IC method indicate at most a 20%
difference for 4-h averages across a concentration range of
80-1200 pptv. The system has been automated with a 5 min
cycle time.
5.5.8 Method Recommendations for
HNO3 and MONO
The denuder/ion chromatographic method for monitoring
both nitrous acid and nitric acid has been commercialized by
Anderson Inc. (Atlanta, GA) by using a rotating annular denuder
for collection, and a system design by P.K. Dasgupta of Texas
Tech University which uses a wet parallel plate denuder has been
tested in several recent field studies (see Figure 5.1). One addi-
tional attraction of the denuder/ion chromatographic approach is
the information about other compounds that is obtained as part
of the ion chromatographic analysis.
DOAS .methods using manual inspection of individual
spectra may be the most accurate for HONO due to the low
detection limits of =100 pptv, the absence of significant inter-
ferences, and the fact that no sampling is required. However, the
use of manual inspection limits automation of the method.
Chemical ionization mass spectroscopy is apparently the
best available method for nitric acid based on the recent work at
NOAA and the National Center for Atmospheric Research
(NCAR). It is not clear that this technique will meet the detection
limit requirement for diagnostic testing of models, although the
recent side-by-side comparisons showed agreement between two
systems of ±30% at sub-ppbv concentrations. Sampling losses
through the MS inlet system are apparently minimized due to the
rapid airflow rate into the system. Additional comparisons with
other viable techniques seem warranted.
5.6 Methods for Monitoring
Particle Nitrate
Particle nitrate has been measured downstream of water-
soluble gas components in the type of instrument described by
Simon and Dasgupta (1995a and 1995b). In the original paper,
Dasgupta and coworkers designed the system to grow particles
to a large size by adding steam to the sample stream and then to
collect the particles for 1C analysis by impaction in a cooled
maze. Recent design changes include the collection of particles
on a filter followed by automated extraction and analysis.
Khylstov et al. (1995) describe a similar method using a steam-
jet aerosol collector with ion chromatographic detection of inor-
ganic ions. Spicer et al. (1985) describe a thermal decomposition/
chemiluminescence method for determining nitrate in which the
sample or an aqueous extract of the sample is heated to 425 °C
and detection of the resultant gaseous nitrogen oxides is by NO,
O3 chemiluminescence. Yamamoto and Kosaka (1994) also dis-
cuss thermal desorption of nitrates to a chemiluminescence
detector using a different arrangement for thermal desorption.
Stolzenburg and Hering (2000) disclose a method for automated
measurement of fine-particle nitrate in the atmosphere with a
cycle time of 10 min. Interfering gases such as nitric acid are
removed from the airstream by an upstream denuder. Particles
are increased in size by humidification and then collected by
impaction. This is followed by flash vaporization and chemilumi-
nescence detection of evolved nitrogen species. This approach
has been used in the field in several studies and is currently being
commercialized by Rupprecht and Patashnik. Inc. (Albany, NY).
5.6.1 Method Recommendations
For Particle Nitrate
The available technologies are relatively new and need to
be compared in side-by-side field testing. However, the systems
based on thermal desorption/chemiluminescence and on filter
collection followed by extraction and 1C analysis are being used
frequently and, in the case of the former a commercial status has
been achieved.
5.7 Methods for Monitoring PAN, PPN,
MPAN, and Other Organic Nitrates
Kleindienst (1994) summarizes the different measurement
techniques for PAN up to a certain point in 1993 in the context
of a comprehensive review of the properties of PAN and dif-
ferent aspects of its part in atmospheric chemistry. Table 2 in this
article includes a summary of methods for detecting PAN in
tropospheric measurements. These include three gas chromato-
graphic methods—electron capture, luminol chemiluminescence,
and NO, O3 chemiluminescence—and two continuous methods—
infrared spectroscopy and PAN, amine chemiluminescence. The
use of capillary chromatography with either electron capture
detectors (ECDs) or luminol chemiluminescence continues to be
prevalent in current monitoring efforts for PAN and other
organic nitrates. Some researchers still use the older packed
chromatographic column technology. A number of additional
references, including more recent references, are given below.
PAN thermally decomposes in the atmosphere to form
NO2 and the acetylperoxy radical, and at 298 K (or 77 T)
calculation of the thermal decomposition rate results in an
atmospheric lifetime of nearly 1 h, although an increase in
ambient temperature to 305 K (or 90 °F) reduces this to roughly
15 min (see Figure 1 of Kleindienst, 1994). Sample integrity of
PAN and PAN-like compounds during the measurement process
in moving through a point monitor, contacting surfaces, and, in
some cases, moving through capillary or packed columns, is a
major concern. Thermal decomposition of PAN-like compounds
on a precolumn converter is sometimes used to check for the
presence of interferences when gas chromatographic separations
-------
with short retention times (e.g., during aircraft measurements) are
being used. There is no primary standard for PAN-like com-
pounds, and they are usually synthesized in the laboratory and
stored for subsequent dilution at the monitoring site. Conversion
of PAN to NO in the well-characterized thermal converters used
for NOY measurements (see above) often provide a surrogate
standard.
Methods for sampling and analysis of alkyl nitrates are
reviewed by Parrish and Fehsenfeld (2000) in their article on
methods for gas-phase measurements of ozone and aerosol pre-
cursors. Direct sample injection orpreconcentration followed by
GC/ECD analysis is typical.
5.7.7 Gas Chromatography with
Specific Detection
The most common current methods for analysis of PAN
and PAN-like compounds involve separation of organic nitrates
by GC and detection using either an ECD or a luminol chemilum-
inescence detector for NO2 (after thermal conversion of nitrates
to produce NO2). As pointed out by Roberts et al. (1989),
capillary columns offer better sensitivity of detection compared
to packed columns and can separate PAN, PPN, and a number of
C1-C5 alkyl nitrates from many light halogenated compounds
that have similar retention times. Blanchard et al. (1990) com-
pared two packed-column GC/ECD systems for monitoring
ambient air (0.15 to 2 ppbv) and obtained slope and intercept
values of 1.14 ± 0.01 and -0.03 ± 0.05, respectively, with a
correlation coefficient of 0.995. A comparison between one of
these systems and a system with a packed column and a GC/
luminol detector for the same monitoring sequence gave slope
and intercept values of 1.08 ± 0.03 and 0.08 ± 0.24, respectively,
with a correlation coefficient of 0.864. Blanchard et al.(1993)
demonstrated the potential of a packed-column GC method for
PAN using a postcolumn chemical amplifier and a NO2 luminol
detector. For PAN concentrations < 1 ppbv, a NO2 amplification
factor of 180 ± 20 was observed when 6 ppmv of NO and 8% of
CO were added to a postcolumn reactor; this approach was
proposed for extending the lower range of detection for PAN. De
Santis et al (1996) used an annular denuder coated with a sodium
carbonate solution to capture PAN. The PAN is retained as
nitrate and the nitrate is extracted for measurement by ion
chromatography. Danalatos and Glavas (1997) used a capillary
column at subambient temperatures along with an ECD to
demonstrate improved response compared to ambient tempera-
ture runs and to separate PAN from interfering gases (mainly
halogenated compounds). Nikitas et al. (1997) used thermal
conversion of PAN to NO2 to generate a signal of PAN + NO2 on
a luminol detector and subtracted the signal obtained from NO2
alone by bypassing the thermal converter; PAN was obtained by
subtraction. Gaffney et al. (1998) have suggested a method com-
bining chromatographic separation of NO2, PAN, MPAN, and
PEN (peroxybutyryl nitrate) on a 3-m-long, 0.53-mm-diameter
capillary column and a modified commercial luminol-based
nitrogen dioxide detector for field measurements. The column is
coated with 3 Jim of DB -1 stationary phase and operated at room
temperature. Helium is used as the carrier gas for 1- to 5-mL
samples to obtain a chromatogram in 1 min and to achieve a
sensitivity in the tens of pptv with synthetic samples. Field
measurements of PAN, PPN, and MPAN are now being made
with capillary column/ECD systems at total uncertainties of ± 5
pptv + 15%) to infer the contributions of biogenic and anthro-
pogenic sources to ozone formation (Roberts et al., 1998). Detec-
tion limits as low as 5 pptv have been achieved. Williams et al.
(2000) extended the GC/ECD approach to airborne measure-
ments of PAN, MPAN, and PPN.
While ECDs have been used for many years as
chromatographic detectors for PANs and other organic nitrates,
the use of radioactive material, typically 63Ni, is coming under
increasing scrutiny and restriction. Zedda et al. (1998) have
recently applied a new pulsed discharge electron capture detector
to the measurement of PAN, PPN, and other atmospheric
nitrates. The new detector employs no radioactive components
and therefore requires no special consideration for transport and
use at field sites.
5.7.2Gas Chromatography/Negative Ion
Chemical lonization Mass Spectrometry
Tanimoto et al. (1999) report the measurement of PAN,
PPN, and MPAN at pptv levels by gas chromatography/negative
ion chemical ionization mass spectrometry. For PAN the detec-
tion limit is given as 15 pptv with good linearity at the pptv
levels. All three compounds are measured within-10 min.
5.7.3 Method Recommendations for PAN,
PPN, MPAN, and other Organic Nitrates
PAN and other organic nitrates are monitored using
capillary columns for separation and sensitive detectors such as
the luminol-based chemiluminescence system or the ECD for
detection. These units can be assembled from commercially
available products. The sample integrity of PAN during a mea-
surement is subject to its thermal decomposition in the mea-
surement instrument and to potential interferences even in the
presence of gas chromatographic separation. Careful experi-
mental and calibration techniques are required to accurately
measure PAN and PAN-like compounds, so any routine or
network application should only be attempted with a high level
of dedication and preparation.
5.8 Methods for Monitoring the
Nitrate Radical
The free radical NO3 is highly reactive and its contri-
bution to NOy is generally negligible. However, it is important
in nighttime atmospheric chemistry because it oxidizes many
primary organic pollutants to form nitric acid, peroxy radicals,
-------
and other products. Platt et al. (1980) recorded the diurnal
variations of NO3 using the DOAS technique. A recent paper
(Geyer, 1999) lists the various techniques and shows a favorable
comparison between the two most frequently used, i.e., DOAS
and matrix isolation electron spin resonance (MIESR). King et
al. (2000) describe the first use of cavity ring-down spectroscopy
(CRDS) to detect NO3 in the laboratory with a system achieving
a noise equivalent mixing ratio of 2 pptv for a 30-s averaging
period.
5.8.1 Method Recommendations
For the Nitrate Radical
Research-grade DOAS systems have been used recently
in summer intensive studies by the Southern Oxidants Study
group and appears to be the best available method.
-------
Chapter 6
Use of Commercially Available Systems for
NO2 Monitoring in the Nashville SOS '99 Study
by
K. Kronmiller and M. Wheeler
ManTech Environmental Technology, Inc.
Interest in gathering information on NO2 monitoring
technology led EPA to participate in the SOS '99 summer field
study in Nashville, TN, at the Cornelia Fort site northeast of
midtown Nashville. Three other groups were using research-
grade instrumentation for the reactive nitrogen oxides at this site.
The EPA effort was organized by W.A. McClenny of the Atmo-
spheric Methods and Monitoring Branch, Human Exposure and
Atmospheric Sciences Division, National Exposure Research
Laboratory, Office of Research and Development, U.S. EPA, and
the site coordinator, E.J. Williams of the NOAA Agronomy
Laboratory. EPA contractor ManTech Environmental Tech-
nology, Inc., prepared and operated commercial instrumentation,
in some cases with modifications, for NOY, NOX, NO2, and NO
monitoring. This activity resulted in valuable field monitoring
experience and in a database for certain commercial and modi-
fied commercial instruments for monitoring nitrogen oxides.
Additional monitoring was carried out after the field study at the
EPA facility in Research Triangle Park, NC, in order to repeat
the types of monitoring executed in Nashville.
Three monitoring systems were operated inside one of the
office trailers located at the base of the sampling tower at the
Cornelia Fort site. One system consisted of a Model LMA-3
luminol-based NO2 monitor (Scintrex, Toronto, Canada) with a
sampling train designed with alternate routes to condition the
sample air before entering the instrument's reaction chamber
(Spicer, et al., 1995). This monitoring arrangement is explained
more fully in Appendix E. The gases NO2, NOX, and NOY could
be monitored in sequence by using the different sampling routes.
Each monitoring mode was maintained for 5 min and repeated
every 15 min. The conditioned sample air was also monitored
using a TEI Model 42 chemiluminescence monitor with internal
thermal converter (Thermo Environmental Instruments, Franklin,
MA). The TEI Model 42 provided NO, NO2, and NOY during the
first 5 min (no conditioning) of each 15-min cycle. A third
monitoring system, a TEI Model 42C was operated inside the
office trailer with a Mo thermal converter mounted on the
sampling tower. This unit was used to obtain some of the NOY
measurements shown below.
A fourth system, a commercially available UV/DOAS
open-path monitoring system (OPSIS, Furulund, Sweden), of a
similar type to the research system described in Appendix B was
operated adjacent to the sampling tower at Cornelia Fort. A
source/receiver was placed at one end of the monitoring path and
a corner-cube retroreflector at the opposite end so as to establish
a 200-m total optical open path at an average aboveground height
of 3 m. This system provided essentially real-time measurements
of ambient NO2 by measuring the differential absorption of
UV/visible radiation along the measurement path.
Figure 6-1 shows the results of a monitoring comparison
at the Cornelia Fort during the period 28 June through 5 July
1999 using three of the instruments described above. Figure 6-2
shows the results of a monitoring comparison at the EPA facility
in Research Triangle Park, NC, after the Nashville summer study
during the period 4-11 October 1999. The same three instru-
ments were used, although the height of the optical path length
for the path monitor averaged the same as the inlet height of the
point monitors, i.e., 3 m. In both figures, NO2 is monitored with
sufficient accuracy to establish compliance with the NAAQS for
NO2, i.e., an annual average concentration of less than 53 ppbv.
However, it is obvious that the monitoring requirements cited in
Chapter 4 for modeling, i.e., ±10% at 1 ppbv and above, is not
uniformly achieved. Signal offsets occurred at the lower end of
the monitoring range and instrument problems caused the loss of
a number of monitoring sequences. Since the LMA-3 was
-------
r
SOS99 Cornelia Fort Site June 28 - July 5 1999
EPA N02 Measurement Method Comparison
55
45
35
25
15
179.0 179.5 180.0 180.5 181.0 181.5 182.0 182.5 183.0 183.5 184.0 184.5 185.0 185.5 186.0 186.5 187.0
Julian Day
Figure 6-1. Results of a monitoring comparison at the Cornelia Fort during the period 28
June-5 July 1999. Three instruments were used: the UV/DOAS system (OPSIS, Furulund,
Sweden) with a 200-m total optical open pathlength at an average aboveground height of 8
ft and two point monitors, the LMA-3 (Scintrex, Toronto, Canada) and the TEI 42 (Thermo
Engineering Instruments, Waltham, MA), at a height of 30 m.
EPA Annex RTF NO, Study
October 4 - 11 1999
N02 Comparison
70
65
60
55
50
45
40
35
30
25
20
15
10
5
277.0
278.0
279.0
280.0
281.0 282.0
Julian Day
283.0
284.0
285.0
286.0
Figure 6-2. Results of a monitoring comparison at the EPA facility in Research Triangle Park,
NC, during the period 4-11 October 1999 using the same instruments as in Figure 6-1.
r
-------
calibrated at high concentrations and corrections for the response
nonlinearity (lower slope of response vs. concentration at low
concentrations), the lower readings for the LMA-3 are under-
standable. If not for this effect, the OPSIS and LMA-3 readings
should be comparable and somewhat lower than the TEI Model
42 since the Model 42 is responding to NOY - NO and therefore
establishing an upper bound to NO2.
Figure 6-3 shows a data comparison for a 24-h period
(20-21 June 1999) during the Nashville SOS '99 study between
the UV/DOAS system and two research-grade instruments, one
based on LIF and the other on NO, O3 chemiluminescence with
photolytic conversion of NO2 (see Appendices). All these sys-
tems are essentially real time and, aside from the difference in
height (3 m for the open path and 30 m for the inlet to the point
monitors), the readings should agree. As noted on the figure, the
average value for those times at which all three monitors were
operating is extremely close. Differences between the point and
path monitors are also evident, although these differences may be
due to the distribution of NO2 with height. The correlation of
responses in time show both positive and negative differences,
which would not be the case for a true bias.
Figures 6-4 and 6-5 show comparisons of two chemilumi-
nescence instruments and the LMA-3 for NOY. Again the LMA-3
appears to read somewhat low at low concentrations of NO2 for
reasons noted above. However, as shown in these figures, the
average value of data from the three units is within 12 ppbv for
each of the monitoring sequences. Differences at the low-ppbv
levels observed during the daytime and at certain other times are
greater than the modeling requirements of ±20% for concen-
trations k 1 ppbv.
These experimental measurements indicated that com-
mercial instruments as operated in the studies agreed with
research-grade instruments to the extent that the conclusions for
NAAQS monitoring and trends monitoring of daily averages or
maximum values would be valid using either. However, biases
due to uncorrected nonlinear response and the possibility of
uncompensated interferences from PAN and O3 in the LMA-3
and from NOZ compounds in the TEI Model 42 prevent the
agreement of the three monitors during periods when NO2 con-
centrations are low and photochemical activity produces inter-
fering NOZ compounds. Note from the Figures 6-1 and 6-2 that
agreement among the three monitors was significantly better in
October 1999 when photochemical activity was lower. The use
of single-channel point monitors did not allow the high tem-
poral resolution most desirable for diagnostic monitoring. Also,
the time sharing of one channel to determine a difference mea-
surement for NO2 can lead to erroneous values during conditions
of high NOY and NO concentration variability. Open-path
measurements of NO2 (3-m average above ground-level height)
were made at a different height than the point monitors (30-m
inlet) and were often at the stated detection level of the
commercial instrument (1 ppbv over a 500-m path length or
about 2.5 ppbv over the 200-m path length used) during daylight
monitoring.
3
70
60
50
40
30
20
10
-10
171.4
SOS99 Cornelia Fort Site June 20 • 21 1999
N02 Comparison Data
24hr Avg(ppb): Opsis 22.69, LIF 21.77, NOAA 22.68
Opsis NOZ
LIF N02
NOAA PC N02
171.6
171.8
172.0
Julian Day
172.2
172.4
172.6
Figure 6-3. Results of comparison data for NO2 over a 24-h period, 20-21 June 1999, at the
Cornelia Fort site—UV/DOAS open-path monitor, LIF (Cohen, UC-Berkeley), and chemi-
luminescence monitor with photolytic converter (Williams, Aeronomy Laboratory, NOAA,
Boulder, CO).
-------
c
Nashville SOS99 EPA NOy Data
June 28 - July 5, 1999 Cornelia Fort Site
Means: LMA3 17.1, TEI42 17.5, TEI42C 16.6 (ppb)
120
100
80
60
40
20
179
180
181
182 183
Julian Date
184
185
186
187
Figure 6-4. Results of comparison data for NOY during the period 28 June-5 July 1999 at
the Cornelia Fort site—LMA-3, TEI 42, and TEI 42C.
Nashville SOS99 EPA NOy Data
June 20-21, 1999 Cornelia Fort Site
Means: LMA3 44.9, TEI42 43.7, TEI42C 45.9 (ppb)
160
140
120
100
I 80
Z 60
40
20
171.6 171.7 171.8 171.9 172.0 172.1
Julian Date
172.2
172.3 172.4
172.5
Figure 6-5. Results of comparison data for NOY during 20-21 June 1999 at the Cornelia Fort
site—LMA-3, TEI 42, and TEI 42C.
r
-------
Chapter 7
References
Allen, M.G., Carleton, K.L., Davis, S.J., Kessler, W.J., Otis,
C.E., Palombo, D.A., and Sonnenfroh, D.M. 1995. Ultrasensitive
dual-beam absorption and gain spectroscopy: Applications for
near-infrared and visible diode laser sensors. Appl. Opt.
34(18):3240-3249.
Ammann, M., Kalberer, M., lost, D.T., Tobler, L., Rossler, E.,
Piguet, D., Gaggeler, H.W., and Baltersperger, U. 1988. Hetero-
geneous production of nitrous acid on soot in polluted air masses.
Nature 395:157.
Anlauf, K.G., Fellin, P., Wieve, H.A., Schiff, H.I., Mackay, G.I.,
Braman, R.S., and Gilbert, R. 1985. A comparison of three
methods for measurement of atmospheric nitric acid and aerosol
nitrate and ammonium. Atmos. Environ. 19:325-333.
Arnold, J.R., Dennis, Robin L., and Tonnesen, Gail S. 1998.
Advanced techniques for evaluating Eulerian air quality models:
Background and methodology. In Preprints from the 10th Joint
Conference on the Applications of Air Pollution Meteorology
with the AWMA, pp. 1-5, Phoenix, AZ. Boston: American
Meteorological Society.
Atkinson, R., Baulch, D.L., Cox, R.A., Hampson, R.F., Kerr,
J.A., Rossi, M.J., and Troe, J. 1997. Evaluated kinetic photo-
chemical and heterogeneous data for atmospheric chemistry:
supplement V, IUPAC subcommittee on gas kinetic data eval-
uation for atmospheric chemistry. J. Phys. Chem. Ref. Data
26:521-1011.
Blanchard, P., Shepson, P.B., Schiff, H.I., Bottenheim, J.W.,
Gallant, A.J., Drummond, J.W., and Wong, P. 1990. A com-
parison of calibration and measurement techniques for gas
chromatographic determination of atmospheric peroxyacetyl ni-
trate (PAN). Atmos. Environ. 24A:2839-2846.
Blanchard, P., Shepson, P.B., Schiff, H.I., Drummond, J.W.
1993. Development of a gas chromatograph for trace level
measurement of peroxyacetyl nitrate using chemical amplifica-
tion. Anal. Chem. 65(18):2472-2477.
Bellinger, M.J. 1982. Chemiluminescent measurements of the
oxides of nitrogen in the clean troposphere and atmospheric
chemistry implications. Ph.D. thesis, Chemistry Department,
University of Colorado, Boulder.
Braman, R.S., Shelley, T.J., and McClenny, W.A. 1982. Tungstic
acid for preconcentration and determination of gaseous and
particulate ammonia and nitric acid in ambient air. Anal. Chem.
54:358-364.
Buhr, S.M., Huhr, M.P., Fehsenfeld, F.C., Holloway, J.S., Karst,
U., Norton, R.C., Parrish, D.D., and Sievers, R.E. 1995.
Development of a semi-continuous method for the measurement
of nitric acid vapor and paniculate nitrate and sulfate. Atmos.
Environ. 29(19): 2609-2624.
Burkhardt, M.R., Maniga, N.I., Stedman, D.H., and Paur, R.J.
1988. Gas chromatographic method for measuring nitrogen
dioxide and peroxyacetyl nitrate in air without compressed gas
cylinders. Anal. Chem. 60:816-819.
Butcher, S.S., and Ruff, R.E. 1971. Effect of inlet residence time
on analysis of atmospheric nitrogen oxides and ozone. Anal.
Chem. 43:1890-1892.
Byun, D.W., and Ching, J.K.S. 1999. Science Algorithms of the
EPA Models-3 Community Multiscale Air Quality (CMAQ)
Modeling System. U.S. EPA report no. EPA/600/R-99/30. Wash-
ington, DC: U.S. EPA, Office of Research and Development.
Chancy, L.W., Rickel, D.G., Russwurm, G.M., and McClenny,
W.A. 1979. Long-path laser monitor for carbon monoxide:
System improvements. Appl. Opt. 18:3004-3009.
-------
Chang, T.Y., Chock, D.P., Nance, B.I., and Winkler, S.L. 1997.
A photochemical extent parameter to aid ozone air quality
management. Atmos. Environ. 31:2687-2794.
Gofer, W.R., III, Collins, V.G., Talbot, R.W. 1985. Improved
aqueous scrubber for collection of soluble atmospheric trace
gases. Environ. Sci. Technol. 19:557-560.
Copeland, H. TV A, Mussel Shoals, AL, personal communi-
cation, September 1999.
Danalatos, D., and Glavas, S. 1997. Improvement in the use of
capillary columns for ambient air peroxyacetyl nitrate moni-
toring. J. Chromatogr. A 786:361-365.
Dasgupta, P.K. and Philips, D.A. 1987. A diffusion scrubber for
the collection of gaseous nitric acid. Sep. Sci. Technol.
22:1255-1267.
De Santis, P., Allegrini, I., Di Filippo, P., Pasella, D. 1996.
Simultaneous determination of nitrogen dioxide and peroxyacetyl
nitrate in ambient atmosphere by carbon-coated annular diffusion
denuder. Atmos. Environ. 30(14):2637-2645.
Delany, A.C., Dickerson, R.R., Melchoir, F.L., Jr., and
Wartburg, A.F. 1982. Modification of a commercial NOX
detector for high sensitivity. Rev. Sci. Instrum. 53:1899-1902.
Dennis, R.L. 1998. The Environmental Protection Agency's third
generation air quality modeling system: An overall perspective.
In Preprints from the 10th Joint Conference on the Applications
of Air Pollution Meteorology with the AWMA, pp. 225-258,
Phoenix, AZ: American Meteorological Society.
Dennis, R.L., Arnold, J.R., and Tonnesen, G.S. 2000. Probing
the shores of ignorance. In Forecasting Environmental Change,
ed., M.B. Beck. New York: Elsevier Press (in press).
Dentener, F.J., and Crutzen, PJ. 1993. Reaction of N2O5 on
tropospheric aerosols: Impact on global distributions of NOX, O3,
and OH. J. Geophys. Res. 98:7149-7163.
Dickerson, R.R., Delany, A.C., and Wartburg, A.F. 1984. Further
modification of a commercial NOX detector for high sensitivity.
Rev. Sci. Instrum. 55:1995-1998.
Dodge, M.C. 1977. Combined use of modeling techniques and
smog chamber data to derive ozone precursor relationships. In
Proceedings of the International Conference on Photochemical
Oxidant Pollution and Its Control, ed., B. Dimitriades, U.S. EPA
report no. EPA-600/3-77-001b. Research Triangle Park, NC:
U.S. EPA, Office of Research and Development.
Drummond, J.W., Volz, A., and Ehhalt, D.H. 1985. An
optimized chemiluminescence detector for tropospheric NO
measurements. J. Atmos. Chem. 2:287-306.
Edner, Hans, Ragnarson, Par, Spannare, Stefan, and Svanbery,
Sune. 1993. Differential optical absorption spectroscopy (DOAS)
system for urban atmospheric pollution monitoring. Appl. Opt.
32(3):327-333.
Fahey, D.W., Eubank, C.S., Hubler, G., and Fehsenfeld, F.C.
1985. Evaluation of a catalytic reduction technique for the
measurement of total reactive odd-nitrogen NOY in the atmo-
sphere. 7. Atmos. Chem. 3:435-468.
Febo, A., Perrino, C, and Allegrini, II. 1996. Measurement of
nitrous acid in Milan, Italy, by DOAS and diffusion denuders.
Atmos. Environ. 30(21):3599-3609.
Federal Register. October 27, 1998. 63(207):57355-57404.
Fehsenfeld, F.C., Dickerson, R.R., Hubler, G., Luke, W.T.,
Nunnermacker, L.J., Williams, E.J., Roberts, J.M., Calvert, J.G.,
Curran, C.M., Delany, A.C., Eubank, C.S., Fahey, D.W., Fried,
A., Gandrud, B.W., Langford, A.O., Murphy, P.C., Norton, R.B.,
Pickering, K.E., and Ridley, B.A. 1987. A ground-based inter-
comparison of NO, NOX, and NOY measurement techniques. 7.
Geophys. Res. 92:14710-14722.
Fehsenfeld, F.C., Drummond, J.W., Roychowdhury, U.K.,
Galvin, P.J., Williams, E.J., Buhr, M.P., Parrish, D.D., Hubler,
G., Langford, A.O., Calvert, J.G., Ridley, B.A., Grahek, F.,
Heikes, E.G., Kok, G.L., Shelter, J.D., Walega, J.G., Elsworth,
C.M., Norton, R.B., Fahey, D.W., Murphy, P.C., Hovermale, C.,
Mohnen, V.A., Demerjian, K.L., Mackay, G.I., and Schiff, H.I.
1990. Intercomparison of NO2 measurement techniques. 7.
Geophys. Res. 95(D4):3579-3597.
Fehsenfeld, F.C., Huey, L.G., Sueper, D.T., Norton,
R.B.,Williams, E.J., Eisele, F.L., Mauldin, R.L., III, and Tanner,
D.J. 1998. Ground-based intercomparison of nitric acid measure-
ment techniques. 7. Geophys. Res. 103(D3):3343-3353.
Fetzer, Gregory J., Miao, Linshi, Chilla, Juan L.A., Pikal, Jon
M., and Menoni, Carmen S. 1998. NO2 photometer based on
solid-state light sources. Appl. Opt. 37(24):5590-5595.
Fincher, C.L., Tucker, A.W., Birnbaum, M., Paur, R.J., and
McClenny, W.A. 1977. Fluorescence ambient NO2 monitor with
flashlamp pumping. Appl. Opt. 16(May):1359-1365.
Fontijn, A., Sabadell, A.J., and Ronco, R.J. 1970. Homogeneous
chemiluminescent measurement of nitric oxide with ozone:
-------
Implications for continuous selective monitoring of gaseous air
pollutants. Anal. Chem. 42:575-579.
Fox, D.G. 1984. Uncertainty in air quality modeling: A summary
of the AMS workshop (September 1982, Woods Hole, MA) on
quantifying and communicating model uncertainty. Bull. Am.
Met. Soc. 65:27-36.
Gaffney, J.S., Bornick, R.M., Chen, Y.-H., and Marley, N.A.
1998. Capillary gas chromatographic analysis of nitrogen dioxide
and PANs with luminol chemiluminescent detection. Atmos.
Environ. 32(8): 1445-1454.
Gaffney, J.S., Marley, N.A., Steele, H.D., Drayton, P.J., and
Hubbe, J.M. 1999. Aircraft measurements of nitrogen dioxide
and peroxyacyl nitrates using luminol chemiluminescence with
fast capillary gas chromatography. Environ. Sci. Technol.
33(19):3285-3289.
Gao, R.S., Keim, E.R., Woodbridge, E.L., Ciciora, S.J., Proffitt,
M.H., Thompson, T.L., Mclaughlin, R.J., Fahey, D.W. 1994.
New photolysis system for NO2 measurements in the lower
stratosphere. J. Geophys. Res. 99(D10):20673-20681.
Gelbwachs, J.A., Birnbaum, M., Tucker, A.W., and Fincher, C.L.
1972. Fluorescence determination of atmospheric NO2.
Optoelectron. 4:155-157.
Geyer, A., Alicke, B., Mihelcic, D., Stutz, J., and Platt, U. 1999.
Comparison of tropospheric NO3 radical measurements by
differential optical absorption spectroscopy and matrix isolation
electron spin resonance. J. Geophys. Res. 104(D21):26097-
26105.
Gipson, G. 1999. Chapter 16: Process analysis. In Science
Algorithms of the EPA Models-3 Community Multiscale Air
Quality (CMAQ) Modeling System, eds., D.W. Byun and J.K.S.
Ching, U.S. EPA report no. EPA/600/R-99/30. Washington, DC:
U.S. EPA, Office of Research and Development.
Gregory, G.L., Hoell, J.M., Jr., Carroll, M.A., Ridley, B.A.,
Davis, D.D., Bradshaw, J., Rodgers, M.O., Sandholm, S.T.,
Schiff, H.I., Hastie, D.R., Karecki, D.R., Mackay, G.I., Harris,
G.W., Torres, A.L., and Fried, A. 1990a. An intercomparison of
airborne nitrogen dioxide instruments. J. Geophys. Res.
95(D7): 10103-10127.
Gregory, G.L., Hoell, J.M., Ridley, B.A., Singh, H.B., Gandrud,
B., Salas, L.J., and Shelter, J. 1990b. An intercomparison of
airborne PAN measurements. 7. Geophys. Res. 95:10077-10087.
Harrison, R.M., Grenfell, J.L., Yamulki, S., Clemitshaw, K.C.,
Pendett, S.A., Cape, J.N., and McFadyen, G.G. 1999. Budget of
NOY species measured at a coastal site. Atmos. Environ.
33:4255-4272.
Helton, J., and Burmaster, D.E. 1996. On the treatment of
aleatory and epistemic uncertainty in performance assessment for
complex systems. Reliability Engineering and Systems Safety,
Special Issue 54:91-94.
Hering, S.V., Lawson, D.R., Allegrini, I., Febo, A., Perrino, C,
Possanzini, M., Sickles, J.E., II, Anlauf, K.G., Wiebe, A., Appel,
B.R., John, W., Ondo, J., Wall, S., Braman, R.S., Sutton, R.,
Cass, G.R., Solomon, P.A., Eatough, D.J., Eatough, N.L., Ellis,
E.C., Grosjean, D., Hicks, B.B., Womack, J.D., Horrocks, J.,
Knapp, K.T., Ellestad, T.G., Paur, R.J., Mitchell, W.J., Pleasant,
M., Peake, E., Maclean, A., Pierson, W.R., Brachaczek, W.,
Schff, H.I., Mackay, G.I., Spicer, C.W., Stedman, D.H., Winer,
A.M., Biermann, H.W., and Tuazon, E.G. 1988. The nitric acid
Shootout: Field comparison of measurement methods. Atmos.
Environ. 22:1519-1539.
Hoell, J.M., Jr., Gregory, G.L., McDougal, D.S., Torres, A.L.,
Davis, D.D., Bradshaw, J., Rodgers, M.O., Ridley, B.A., and
Carroll, M.A. 1987. Airborne intercomparison of nitric oxide
measurement techniques. 7. Geophys. Res. 92:1995-2008.
Huey, L.G., Dunlea, E.J., Lovejoy, E.R., Hanson, D.R., Norton,
R.B., Fehsenfeld, F.C., and Howard, C.J. 1998. Fast time re-
sponse measurements of HNO3 in air with a chemical ionization
mass spectrometer. 7. Geophys. Res. 103(D3):3355-3360.
Jeffries, H.E., and Tonnesen, G.S. 1994. A comparison of two
photochemical reaction mechanisms using mass balance and
process analysis. Atmos. Environ. 28:2991-3003.
Jung, J., and Kowalski, J. 1986. Direct ambient nitrogen dioxide
measurement by visible light absorption. In Proceedings of the
EPA/APCA Symposium on Measurement of Toxic Air Pollutants,
Raleigh, NC, APCA publication VIP-7 (EPA Report No. 600/9-
86-013).
Kanda, U., and Taira, M. 1992. Simultaneous determination of
atmospheric nitric acid and nitrous acid by reduction with
hydrazine and ascorbic acid with chemiluminescence detection.
Analyst 117:883-887..
Kelly, T.J., Spicer, C.W., and Ward, G.F. 1990. An assessment
of the luminol chemiluminescence technique for measurement of
NO2 in ambient air. Atmos. Environ. 24A(9):2397-2403.
Keuken, M.P., Schoonebeek, C.A.M., Van Wensveen-Louter, A.,
and Slanina, J. 1988. Simultaneous sampling of NH3, HNO3,
HC1, SO2, and H2O2 in ambient air by wet annular denuder
system. Atmos. Environ. 22:2541-2548.
-------
Khlystov, A., Wyers, G.P., and Slanina, J. 1995. The steam-jet
aerosol collector. Atmos. Environ. 29(17):2229-2234.
King, M.D., Dick, E.M., and Simpson, W.R. 2000. A new
method for the atmospheric detection of the nitrate radical (NO3).
Atmos. Environ. 34:685-688.
Kleindienst, T.E. 1994. Recent developments in the chemistry
and biology of peroxyacetyl nitrate. Res. Chem. Intermed.
20:335-384.
Kleinman, L.I. 1994. Low- and high-NOx tropospheric photo-
chemistry. J. Geophys. Res. 99:16831-16838.
Kliner, D.A., Daube, B.C., Hurley, J.D., and Wofsy, S.C. 1997.
Laboratory investigation of the catalytic reduction technique for
measurement of atmospheric NOY, J. Geopyhs. Res. 102(D9):
10759-10776.
Klockow, D., Niessner, R., Malejczyk, M., Kiendl, H., Berg, B.
vom, Keuken, M.P., Wayers-Ypelaan, and Slanina, J. 1989.
Determination of nitric acid and ammonium nitrate by means of
a computer-controlled thermodenuder system. Atmos. Environ.
23(5):1131-1138.
Li, Y., Dennis, R.L., Tonnesen, G.S., and Pleim, J.E. 1998.
Regional ozone concentrations and production efficiency as
affected by meteorological parameters in the Regional Acid
Deposition Modeling system. In Preprints from the 10th Joint
Conference on the Applications of Air Pollution Meteorology
with theAWMA. Phoenix, AZ: American Meteorological Society.
Maeda, U.K., Aoki, K., and Munemori, M. 1980. Chemilumi-
nescence method for the determination of nitrogen dioxide. Anal.
Chem. 52:307-311.
Mauldin, R.L., Tanner, D.J., and Eisele, F.L. 1998. A new
chemical ionization mass spectrometer technique for the fast
measurement of gas-phase nitric acid in the atmosphere. J.
Geophys. Res. 103(D3):3361-3367.
McClenny, W.A., Gailey, P.C., Braman, R.S., and Shelley, T.J.
1982. Tungstic acid technique for monitoring nitric acid and
ammonia in ambient air. Anal. Chem. 54:365-369.
McElroy, F.F., Hodgeson, J., Lumpkin, T.A., Rehme, K.A.,
Stevens, R.K., Conner, C.P., and Hallstadius, H. 1993. Simu-
ltaneous calibration of open-path and conventional point mon-
itors for measuring ambient air concentrations of sulfur dioxide,
ozone, and nitrogen dioxide. In Proceedings of the International
Specialty Conference on Optical Sensing for Environmental
Monitoring, Atlanta, GA, October 11-14, 1993 AWMA pub-
lication SP-89. Pittsburgh, PA: Air &Waste Management
Association.
McRae, G.J., Goodin, W.R., and Seinfeld, J.H. 1982.
Development of a second-generation mathematical model for
urban air pollution—I. Model formulation. Atmos. Environ.
16:679-696.
Mihalcea, R.M., Baer, D.S., and Hanson, R.K. 1996. Tunable
diode-laser absorption measurements of NO2 near 670 and 395
nm.Appl. Opt. 35(21):4059-4064.
National Research Council .1991. Rethinking the Ozone Problem
in Urban and Regional Air Pollution. Washington, DC: National
Academy Press.
Nikitas, C., Clemitshaw, K.C., Oram, D.E., and Penkett, S.A.
1997. Measurements of PAN in the polluted boundary layer and
free troposphere using a luminol-NO2 detector combined with a
thermal converter. J. Atmos. Chem. 28:339-359.
Oms, M.T., Jongejan, P.A.C., Veltkamp, A.C., Wyers, G.P., and
Slanina, J. 1996. Continuous monitoring of atmospheric HC1,
HNO2, HNO3, and SO2 by wet-annular denuder air sampling with
on-line chromatographic analysis. Int. J. Environ. Anal. Chem.
62:207-218.
Parrish, D.D., Buhr, M.P., Trainer, M., Norton, R.B., Shimshock,
J.P., Fehsenfeld, F.C., Anlauf, A.G., Bottenheim, J.W., Tang,
Y.Z., Wiebe, H.A., Roberts, J.M., Tanner, R.L., Newman, L.,
Bowersox, V.C., Olszyna, K.J., Bailey, E.M., Rodgers, M.O.,
Wang, T., Berresheim, H., Roychowdhury, U.K., and Demerjian,
K. 1993. The total reactive oxidized nitrogen levels and the
partitioning between the individual species at six rural sites in
Eastern North America. J. Geophys. Res. 98:2927-2939.
Parrish, D.D., and Fehsenfeld, F.C. 2000. Methods for gas-phase
measurements of ozone, ozone, precursors, and aerosol pre-
cursors. Atmos. Environ. 34(12-14):1853-2332.
Plane, John M.C., and Nien, Chia-Fu. 1992. Differential optical
absorption spectrometer for measuring atmospheric trace gases.
Rev. Sci. Instrum. 63(3): 1867-1876.
Platt, U. 1994. Differential optical absorption spectroscopy
(DOAS). In Monitoring by Spectroscopic Techniques, ed. M.W.
Sigrist. New York: John Wiley.
Platt, U., Perner, D., Harris, G.W., Winer, A.M., and Pitts, J.M.
1980. Detection of NO3 in the polluted troposphere by dif-
ferential optical absorption. Geophys. Res. Lett. 7:89-92.
-------
Reynolds, S.D., Roth, P.M., and Tesche, T.W. 1994. A Process
for the Stressful Evaluation of Photochemical Model Per-
formance. Glendale, CA: Western States Petroleum Association.
Ridley, B.A., Carroll, M.A., Gregory, G.L., and Sasche, G.W.
1988a. NO and NO2 in the troposphere: Technique and measure-
ments in regions of a folded tropopause. J. Geophys. Res.
93:15813-15830.
Ridley, B.A., Carroll, M.A., Torres, A.L., Condon, E.P., Sachse,
G.W., Hill, G.F., and Gregory, G.L. 1988b. An intercomparison
of results from ferrous sulphate and photolytic converter
techniques for measurements of NOX made during the NASA
CTE/CITE 1 aircraft program. J. Geophys. Res. 93(D12):
15803-15811.
Ried, J., El-Sherbiny, M., Garside, B.K., and Ballik, E.A., 1980.
Sensitivity limits of a tunable diode laser spectrometer with
application to the detection of NO2 at the 100 ppt level. Appl.
Opt. 19:3349-3354.
Roberts, J.M., Norton, R.B., Goldan, P.O., and Fehsenfeld, F.C.
1987. Evaluation of the tungsten oxide denuder tube technique
as a method for the measurement of low concentrations of nitric
acid in the troposphere. J. Atmos. Chem. 5:217-238.
Roberts, J.M., Fajer, R.W., and Springston, S.R. 1989. Capillary
gas chromatographic separation of alkyl nitrates and peroxy-
carboxylic nitric anhydrides. Anal. Chem. 61:771-772.
Roberts, J.M., and Bertman, S.B. 1992. The thermal decom-
position of peroxyacetic nitric anhydride (PAN) and peroxy-
methacrylic nitric anhydride (MPAN). Int. J. Chem. Kinet.
24:297-307.
Roberts, J.M., Williams, J., Baumann, K., Buhr, M.P., Goldan,
P.D., Holloway, J., Hubler, G., Kuster, W.C., McKeen, S.A.,
Ryerson, T.B., Trainer, M., Williams, E.J., Fehsenfeld, F.C.,
Bertman, S.B., Nouaime, G., Seaver, C., Grodzinsky, G.,
Rodgers, M., and Young, V.L. 1998. Measurements of PAN,
PPN, and MPAN made during the 1994 and 1995 Nashville
Intensives of the Southern Oxidant Study: Implications for
regional ozone production from biogenic hydrocarbons. J.
Geophys. Res. (Atmos.) 103(D17):22473-22490.
Russell, A.G., and Dennis, R.L. 2000. NARSTO critical review
of photochemical models and modeling. Atmos. Environ, (in
press).
Russwurm, G.M. 1988. Development of tunable diode laser: A
reference method for nitrogen species. Research Triangle Park,
NC: Northrop Services, Inc.-Environmental Sciences.
Russwurm, G.M., and Phillips, B. 1999. Effects of a nonlinear
response of the Fourier-transform infrared open-path instrument
on the measurements of some atmospheric gases. Appl. Opt.
38(30):6398-6407.
Ryerson, T.B., Williams, E.J., and Fehsenfeld, F.C. 2000. An
efficient photolysis system for fast-response NO2 measurements.
J. Geophys. Res. (Atmos.) (in press).
Saltelli, A., and Scott, M. 1997. The role of sensitivity analysis
in the corroboration of models and its links to model structural
and parametric uncertainty. Reliability Engineering and Systems
Safety 52:1-4.
Schiff, H.I., Hastie, D.R., Mackay, G.L, Iguchi, T., and Ridley,
B.A. 1983. Tunable diode laser systems for measuring trace
gases in tropospheric air. Environ. Sci. Technol. 17:353A-364A.
Schmidt, R.W.H., Kames, J., Kanter, H.J., Schurath, U., and
Slemr, F. 1995. A selective ozone scrubber for application in
ambient nitrogen dioxide measurements using the commercial
Luminox (LMA-3, Scintrex/Unisearch Inc.). Atmos. Environ.
29(8):947-950.
Seinfeld, J.H. 1988. Ozone air quality models: A critical review.
J.AirPollut. Control Assoc. 38:616-645.
Seinfeld, J.H., and Pandis, S. 1998. Atmospheric Chemistry and
Physics: From Air Pollution to Climate Change. New York:
Wiley.
Sickles, I.E., II. 1992. Chapter 2: Sampling and analysis for
ambient oxides of nitrogen and related species. In Gaseous
Pollutants: Characterization and Cycling, ed., Jerome O.
Nriagu. New York: John Wiley & Sons, Inc.
Sickles, J.E., II, Hodson, L.L., McClenny, W.A., Paur, R.J.,
Ellestad, T.G., Mulik, J.D., Anlauf, K.G., Wiebe, H.A., Mackay,
G.L, Schiff, H.I., and Bubacz, O.K. 1990. Field comparison of
methods for the measurement of gaseous and paniculate
contributors to acidic dry deposition. Atmos. Environ. 24A(1):
155-165.
Sillman, S., Logan, J.A., and Wofsy, S.C. 1990. The sensitivity
of ozone to nitrogen oxides and hydrocarbons in regional ozone
episodes. J. Geophys. Res. 95(D2):1837-1851.
Sillman, S. 1995. The use of NOY, H2O2, and HNO3 as indicators
for O3-NOX-HC sensitivity in urban locations. J. Geophys. Res.
100(D7):14175-14188.
-------
Simon, P.K., and Dasgupta, P. 1993. Wet effluent denuder
coupled liquid/ion chromatography systems: Annular and parallel
plate denuders. Anal. Chem. 65:1134-1139.
Simon, P.K., and Dasgupta, P. 1995a. Continuous automated
measurement of gaseous nitric and nitrous acids and particulate
nitrite and nitrate. Environ. Sci. Technol. 29:1534-1541.
Simon, P.K., and Dasgupta, P. 1995b. Continuous automated
measurement of the soluble fraction of atmospheric particulate
matter. Anal. Chem. 67:71-78.
Sistla, G., Zhou, N., Hao, W., Ku, J., Rao, S.T., Bornstein, R.,
Freedman, F., and Thusis, P. 1996. Effects of uncertainties in
meteorological inputs on Urban Airshed Model predictions and
ozone control strategies. Atmos. Environ. 30:2011-2055.
Sonnenfroh, D.M., and Allen, Mark G. 1996. Ultrasensitive,
visible tunable diode laser detection of NO2. Appl. Opt.
35(21):4053^1058.
Spicer, C.W. Howes, J.E., Jr., Bishop, T.A., Arnold, L.H., and
Stevens, R.K. 1982. Nitric acid measurement methods: An
intercomparison. Atmos. Environ. 16:1487-1500.
Spicer, C.W., Joseph, D.W., and Schumacher, P.M. 1985.
Determination of nitrate in atmospheric particulate matter by
thermal decomposition and chemiluminescence. Anal. Chem.
57:2338-2341.
Spicer, C.W., Kelly, T.J., and Ward, G.F. 1995. An Evaluation
of 'Two Approaches for Improved Nitrogen Oxides Monitoring in
Urban Atmospheres, U.S. EPA report no. EPA-600/R-95/031.
Columbus, OH: Battelle, EPA Contract 68-DO-0007.
Steffenson, D.M., and Stedman, D.H. 1974. Optimization of the
operating parameters of chemiluminescent nitric oxide detectors.
Anal. Chem. 46:1704-1709.
Stevens, R.K., Drago, R.J., and Mamane, Y. 1993. A long path
differential optical absorption spectrometer and EPA-approved
fixed-point methods intercomparison. Atmos. Environ. 27B:
231-236.
Stolzenburg, M.R., and Hering, S.V. 2000. Method for the
automated measurement of fine particle nitrate in the atmosphere.
Environ. Sci. Technol. (in press).
Taira, M., and Kanda, Y. 1993. Wet effluent diffusion denuder
for sampling of atmospheric gaseous nitric acid. Anal. Chem.
65:3171-3173.
Talbot, R.W., Vijgen, A.S., and Harriss, R.C. 1990. Measuring
tropospheric HNO3: Problems and prospects for nylon filter and
mist chamber techniques. J. Geophys. Res. 95(D6):7553-7561.
Tanimoto, H., Hirokawa, J., Kajii, U., and Akimoto, H. 1999. A
new measurement technique of peroxyacetyl nitrate at part per
trillion by volume levels: Gas chromatography/negative ion
chemical ionization mass spectrometry. J. Geophys. Res. 104
(D17):21343-21354.
Tanner, R.L., Valente, R.J., and Meagher, J.F. 1998. Measuring
inorganic nitrate species with short time resolution from an
aircraft platform by dual-channel ozone chemiluminescence. J.
Geophys. Res. 103(D17):22387-22395.
Tesche, T.W., Roth, P.W., Reynolds, S.D., and Lurmann, F.W.
1992. Scientific Assessment of the Urban Airshed Model (UAM-
IV). Crested Butte, CO: Alpine Geophysics.
Tesche, T.W., and McNally, D.E. 1995. Assessment ofUAM-IV
Model Performance for Three St. Louis Ozone Episodes.
Covington, KY: Alpine Geophysics, LLC.
Thornton, J.A., Wooldridge, P.J., and Cohen, R.C. 2000.
Atmospheric NO2: In situ laser-induced fluorescence detection
at parts per trillion mixing ratios. Anal. Chem. 72:528-539.
Tonnesen, G.S. and Dennis, R.L. 2000a. Analysis of radical
propagation efficiency to assess ozone sensitivity to hydro-
carbons and NOX. Part 1: Local indicators of instantaneous odd
oxygen production sensitivity. J. Geophys. Res. (in press).
Tonnesen, G.S., and Dennis, R.L. 2000b. Analysis of radical
propagation efficiency to assess ozone sensitivity to
hydrocarbons and NOX. Part 2: Long-lived species as indicators
of ozone concentration sensitivity. J. Geophys. Res. (in press).
Tonnesen, G.S., and Jeffries, H.E. 1994. Inhibition of odd
oxygen production in the Carbon Bond 4 and generic reaction set
mechanisms. Atmos. Environ. 28:1339-1349.
Trainer, M., Parrish, D.D., Buhr, M.P., Norton, R.B., Fehsenfeld,
F.C., Anlauf, K.G., Bottenheim, J.W., Tang, Y.Z., Wiebe, H.A.,
Roberts, J.M., Tanner, R.L., Newman, L., Bowersox, V.C.,
Meagher, J.F., Olszyna, K.J., Rodgers, M.O., Wang, T.,
Berresheim, H., Demerjian, K.L., and Roychowdhury, U.K.
1993. Correlation of ozone with NOY in photochemically aged
air. J. Geophys. Res. 98:2917-2925.
Tuazon, E.G., Winer, A.M., and Pitts, J.N., Jr. 1981. Trace
pollutant concentrations in a multiday smog episode in the Cali-
fornia south coast air basin by long path length Fourier transform
infrared spectroscopy. Environ. Sci. Technol. 15:1232-1237.
-------
U.S. Environmental Protection Agency (U.S. EPA). 1991.
Guideline for Regulatory Application of the Urban Airshed
Model, U.S. EPA report no. EPA-450/4-91-013. Research
Triangle Park, NC: Office of Air Quality Planning and
Standards, Technical Support Division.
U.S. Environmental Protection Agency (U.S. EPA). 1993. Air
Quality Criteria for Oxides of Nitrogen, U.S. EPA report no.
EPA/600/8-91 /049aF. Washington, D.C.: Office of Research and
Development.
U.S. Environmental Protection Agency (U.S. EPA). March 1994.
Photochemical assessment monitoring stations implementation
manual, U.S. EPA report no. EPA-454/B-93-051. Research
Triangle Park, NC: Office of Air Quality Planning and
Standards.
U.S. Environmental Protection Agency (U.S. EPA). 1997.
Regulatory Impact Analyses for the Paniculate Matter and
Ozone National Ambient Air Quality Standards and Proposed
Regional Haze Rule. Research Triangle Park, NC: Office of Air
Quality Planning and Standards (OAQPS).
Yarns, J.L., Mulik, J.D., Sather, M.E., Lister, M.K., Glenn, G.,
Smith, L., Williams, D.D., Berz, E.A., and Hines, A.P. 1999. The
Passive Ozone Network in Dallas (POND Concept)—A
Modeling Opportunity with Community Involvement, poster
presentation at the 31st Annual Air Pollution Workshop, April
26-29, Corvallis, OR.
Vecera, Z., and Dasgupta, P.K. 1991. Measurement of ambient
nitrous acid and a reliable calibration source for gaseous nitrous
acid. Environ. Sci. Technol. 25:255-260.
Williams, E.J., Baumann, K., Roberts, J.M., Bertman, S.B.,
Norton, R.B., Fehsenfeld, F.C., Springstone, S.R., Nunner-
macker, L.J., Newman, L., Olszyna, K., Meagher, J., Bartsell, B.,
Edgerton, E., Person, J.R., and Rogers, M.O. 1998. Inter-
comparison of ground-based NOY measurement techniques. J.
Geophys. Res. 103:22261-22280.
Williams, J., Roberts, J.M., Bertman, S.B., Stroud, C.A.,
Fehsenfeld, F.C., Baumann, K., Buhr, M.P., Knapp, K., Murphy,
P.C., Nowick, M., and Williams, E.J. 2000. A method for the
airborne measurement of PAN, PPN, and MPAN, in preparation
for submission to J. Geophys. Res.
Winer, A.M., Peters, J.W., Smith, J.P., and Pitts, J.N., Jr. 1974.
Response of commercial chemiluminescence NO-NOX analyzers
to other nitrogen-containing compounds. Environ. Sci. Technol.
8:1118-1121.
Yamamoto, M., and Kosaka, H. 1994. Determination of nitrate
in deposited aerosol particles by thermal decomposition and
chemiluminescence. Anal. Chem. 66:362-367.
Zedda, D., Keigley, G.W., Joseph, D.W., and Spicer, C.W. 1998.
Development of a new high sensitivity monitor for peroxyacetyl
nitrate and results from the west-central Mediterranean region. In
Air Pollution VI—Sixth International Conference on Air Pollu-
tion, 79-88. eds., C.A. Brebbia, C.F. Ratio, and H. Power,
Southampton: WIT Press.
Zellweger, C., Ammann, M., Hofer, P., and Baltensperger, U.
1999. NOY speciation with a combined wet effluent diffusion
denuder aerosol collector coupled to ion chromatography. Atmos.
Environ. 33:1131-1140.
Vecera, Z., and Dasgupta, P.K. 1991. Measurement of atmo-
spheric nitric and nitrous acids with a wet effluent diffusion
denuder and low-pressure ion chromatograph-postcolumn reac-
tion detection. Anal. Chem. 63:2210-2216.
Zhou, X., Qiao, H., Deng, G., and Civerolo, K. 1999. A method
for the measurement of atmospheric HONO based on DNPH
derivatization and HPLC analysis. Environ. Sci. Technol.
33:3672-3679.
Walega, J.G., Dye, J.E., Grahek, F.E., and Ridley, B.A. 1991.
Compact measurement system for the simultaneous deter-
mination of NO, NO2, NOY, and O3 using a small aircraft. In
Proceedings of SPIE—The International Society of Optical
Engineering, 1433:232-241. Bellingham, WA: International
Society for Optical Engineering.
Wendel, G.J., Stedman, D.H., and Cantrell, C.A. 1983. Luminol-
based nitrogen dioxide detector. Anal. Chem. 55:937-940.
Williams, E. NOAA Agronomy Laboratory, personal communi-
cation, January 2000.
-------
Appendix A
Correction to Point Monitor Readings of O3, NO, and NO2 Due to the Reaction
of NO and O3 during Transport of Ambient Air to the Point of Measurement
by
William A. McClenny and Deborah J. Luecken
Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, U.S. EPA
The measurement of NO, NO2, and O3 by point monitors
requires the transport of sample air from the ambient air through
a length of tubing and eventually into the measurement region of
the monitor. During this transport the gas-phase reaction of NO
and O3 occurs, leading to their mutual decrease and a cor-
responding equal increase in the NO2 concentration. This is a
widely recognized sampling artifact (Ridley et al., 1988a;
Sickles, 1992) that should be minimized and/or accounted for to
allow accurate monitoring of NO and NO2 in ambient air.
Butcher and Ruff (1971) showed there could be significant
changes (changes greater than the precision of measurement) in
these gas concentrations under certain situations, e.g., in rural air
where ozone is high compared to NO and N02, an ambient
temperature of 25 °C or higher is assumed, and a transport time
on the order of 10 s occurs.
In revisiting this issue in order to correct monitoring data
taken in Nashville during the 1999 Southern Oxidants Study
(SOS '99) summer study, the reaction rate for NO and O3 was
taken from the latest compilation by Atkinson et al. (1997). This
rate constant was used to determine losses of NO in the sampling
manifold assuming a sample gas temperature of 30 °C (86 T)
corresponding to a hot day in the southeast USA. Significant
percentage decreases in NO concentration were noted to occur in
a sample containing high O3 concentrations. Depending on
ambient NO2 concentrations, significant percentage increases in
NO2 could also occur in the sampling manifold. Since there is no
official requirement referred to in the Federal Register other than
that the residence time should be less than 20 s (40CFR58,
Appendix E, Section 9), additional information has been
developed to provide guidance in NO2 monitoring using point
monitors and is included in this report to clarify an important
aspect of NO and NO2 monitoring.
A.1 Development of Technical Guidance
for NO2 Monitoring
The ambient NO concentration is exponentially attenuated
during transport through the inlet tubing by its reaction with
ozone, such that, assuming a constant ozone concentration well
above the NO and NO2 concentrations, the final NO concen-
tration is given by
[NO]f=[NO]iexp(-k[03]lt)
(A-l)
where [NO]f = final (entrance to detection chamber) NO
concentration in ppbv
[NO], = initial (ambient) NO concentration in ppbv
k = 0.044267*exp(-1370/T) in 1/ppbv-s
T = temperature in "Kelvin; pressure assumed at 1 atm
[O3]j = initial O3 concentration in ppbv which is assumed
constant
t = residence time in seconds
As examples of NO concentration losses, values of the ratio
[NO]/[NO]j for 50 and 100 ppbv O3 concentrations are shown in
Figure A-1.
For an ozone concentration of 100 ppbv in air samples at
temperatures of 30 °C, the residence time for ambient NO in a
sampling line must be less than 2.20 s to limit the NO losses
to 10% (and 1.06 s to limit the NO losses to 5%). In reality, the
ozone concentration decreases by the same amount as the NO
concentration, and NO losses will be slightly less than predicted
under the assumption of constant O3 concentrations.
-------
Effect of O3 on NO Sampling Line Losses
- 100 ppbV 03 (lower curve)
- 50 ppbv 03
10 15
Time, Seconds
Figure A-1. Effect of O3 on NO sampling line losses.
Short residence time values are not always achieved in
practice. In such cases, a correction to the data should be made
by rearranging Equation A-1, i.e., by multiplying the recorded
NO concentrations, [NO]f, by the factor exp(k [O3]i t):
[NO]i=[NO]f[exp(k[03]it)]
(A-2)
Also, the decrease in NO concentration [NO] f[exp(k [O3]i
t) - 1] must be subtracted from the recorded NO2 concentrations
to correct it, i.e.,
[N02]s = [NOJ, - [NO], [exp( k [OJ, t) - 1]
(A-3)
For the most general treatment to account for any set of
initial O3, NO2, and NO concentrations, numerical integration
software was used to determine O3 and NO losses (and NO2
gains) during transport. Information on this software is available
from Human Exposure and Atmospheric Sciences Division,
National Exposure Research Laboratory, U.S. EPA (Deborah
Luecken, [919-541-0244]). However, the corrections obtained
based on Equation A-1 are nearly identical to those obtained by
numerical integration in most ambient monitoring situations in
which corrections are important.
Experimental monitoring data has shown that the ratio of
final to initial NO concentrations is less than calculated and that
to account for this, the effective reaction rate, k, in the above
equation must be higher. The reaction of O3 and NO at the
surface of the transport tubing is believed to be the cause of this
additional loss, although the exact mechanism has not been
determined. Fehsenfeld et al. (1990) describe experimental pro-
cedures that can be used to determine the effective first-order
reaction rate constant accounting for both NO, O3 reactions and
wall losses. This method involves the standard addition of a
known amount of NO at the entrance of the sampling manifold
at times when different O3 concentrations are present in ambient
air. The experimental value of this reaction rate is expected to be
greater than the gas-phase reaction rate alone.
A.2 Conclusion on Gas-Phase Reaction
between 03 and NO
Sampling manifolds used for establishing compliance
with the NAAQS and for research studies of ambient air should
be examined to ensure that the transit time of air samples through
the sampling manifold and into the monitoring instrument is
short. The longer the residence time the greater the chance for
reaction of NO and O3 to form NO2 and change the concen-
trations of these gases from their values in the ambient air. If the
transit time is not short, the measured NO and NO2 concen-
trations should be corrected as shown in Equations A-2 and A-3.
Wall reactions between O3 and NO to form NO2 have
been neglected here since creation of NO2 at tubing walls is
subject to the type and condition of the wall. In one case, loss has
been observed to be 25% higher than that due to the gas-phase
reaction (Fehsenfeld et al., 1990). However, actual wall losses
must be measured at the specific monitoring location and should
be tested periodically (every 3 months) or as may be indicated by
local conditions, and the manifold should be cleaned to prevent
buildup of particles. Other elements of the inlet manifold system,
notably gas-line filters, may contribute substantially to conver-
sion of NO to NO2 and should be tested accordingly.
If a 10% loss of NO is acceptable at 100 ppbv of O3 and
30 °C, then a manifold residence time of 2.2 s (round to 2.0 s) is
appropriate. Assuming that NOj <. NO2f (NO is generally less
than NO2 away from NO sources) where NO2f <. 20 ppbv, that
[O3] z 100 ppbv, and that T z 30 °C, the overestimation in [NO2]
due to the NO, O3 dark reaction during a 2-s residence time in the
sampling manifold is also calculated to be less than 10%. Lower
ozone concentrations and/or lower temperatures are expected
during almost all sampling periods in any given year at most
locations. Hence corrections to the annual average for NO2
concentrations are expected to be small. However, corrections to
monitoring data will be useful in providing better data for input
to air quality simulation and observationally based models during
periods when ozone concentrations are high. Corrections to the
NO2 at any location and time will obviously depend on manifold
design (residence time), the prevalence of ozone and NO, and the
temperature during the ozone season, as indicated in Equation A-
3 above.
If the chemiluminescence monitoring system includes a
vacuum pump then pressure reduction at the sampling inlet effec-
tively eliminates the homogeneous chemistry inlet losses.
-------
Appendix B
Determination of Atmospheric Concentration of Nitrogen Oxides by
Differential Optical Absorption Spectroscopy
by
Jochen Stutz
Department of Atmospheric Sciences, UCLA
B.1 Introduction
Spectroscopic methods are among the oldest analytical
techniques used to identify and quantify atmospheric trace con-
stituents. In general, their advantages are high selectivity and low
detection limits. And, in the case of open-path measurements, no
calibration is necessary. One of the most successful spectroscopic
techniques in the atmosphere is differential optical absorption
Spectroscopy (DOAS) (see Platt and Perner, 1983; Plane and
Nien, 1992; Platt, 1994; and Plane and Smith, 1995, for reviews).
In the 1970s, DOAS was used to measure stratospheric (Noxon,
1975) and tropospheric (Platt, 1978) nitrogen dioxide, NO2, and
other trace gases (Perner and Platt, 1979; Platt et al., 1979).
Today it is used to monitor trace gases such as NO2, NO, HONO,
NO3, SO2, O3, HCHO, and others (Platt, 1994). The following
text gives a short introduction to the technique and discusses its
advantages, disadvantages, and problems for monitoring nitrogen
oxides.
B.2 Theory of DOAS
DOAS is a method that determines the concentration of
atmospheric trace gases by measuring their narrow-band absorp-
tion. Unlike typical absorption spectroscopic techniques, the light
path of DOAS is placed in the open atmosphere with path lengths
of a few hundred meters to several kilometers. Because the light
path is placed in the open atmosphere, there are several problems
that DOAS has to overcome. The following section describes
how DOAS solves the problem introduced by the location of the
light path.
B.2.1 Principle
A schematic setup of a DOAS instrument is shown in
Figure B-1. Light, with an intensity I0(A.) emitted by a suitable
spectral broadband source, passes through the open atmosphere,
is collected at the end of the light path, and is spectroscopically
analyzed.
-. light
/~\ source
435 445 455
wavelength [nmj
435 445 455
wavelength |nm]
spectrograph
Figure B-1. Schematic overview of DOAS measurements.
-------
The absorption of trace gases can be described by Lambert-
Beer's law:
T/Ti \ _ T (\ \ v 0-O(X)xCxL .D ..
l^/lj — IO^A,,) X e (B-l)
where I0(A) and I(A) are the light intensities at the beginning and
end of the light path with a length L. The absorption of the trace
gas is described by its absorption cross section o(A.) and its
concentration C. As the light travels through the atmosphere, it
also undergoes extinction due to absorption processes by dif-
ferent trace gases and to scattering by air molecules and aerosol
particles. The transmitivity of the instrument (mirrors, grating,
retroreflectors, etc.) will also decrease the light intensity; we will
assume here that this decrease is already included in I0(A).
Therefore, the determination of I0M would require removal of
the respective gas from the air along the light path. The principle
of DOAS is based on the fact that extinction processes and many
trace gas absorptions show very broad or even smooth spectral
characteristics.
The basic concept behind DOAS is the separation of the
absorption spectrum and cross section a = aB + o' into two parts:
OB, which represents broad spectral features, and the differential
absorption cross section a', which represents narrow spectral
structures (see Figure B-2).
Figure B-2. The principle of DOAS.
Equation B-l then transforms to
(B-2)
(B-3)
contains all the broadband parts of the spectrum. Therefore,
considering only o' eliminates interferences with extinction and
other broadband absorptions. It will also eliminate any depend-
ence of the light intensity on the instrument transmitivity, which
typically also shows a broad spectral characteristic. This sepa-
ration can be performed by a suitable filtering procedure (see
below).
By defining a differential optical density, the depth of a
narrow absorption band at a wavelength A', the concentration C
can be calculated:
'!„(*•')'
c =
a'xL
(B-4)
where
In contrast to a simple absorption measurement at one
wavelength, DOAS requires the measurement of a long enough
wavelength interval to perform the described filtering and to
separate overlapping absorptions of different trace gases showing
differential absorption in the same wavelength interval. This
separation is possible because the o are physical constants that
. identify an individual trace gas like a "fingerprint." A calibration
is, in principle, not necessary (we will see below that for real
instruments a calculation to adapt the cross section to the spectral
resolution of the individual instrument is indeed necessary).
Therefore, DOAS is, if applied correctly, an absolute analytical
method for atmospheric nitrogen dioxide and other trace gases.
It should be noted that, for simplicity, the integration of
the product o(A) x C x L over the light path and the sum over a
number of trace gases was omitted here. DOAS measures the
average concentration in the air mass crossed by the light path.
B.2.2 Practical Considerations
A more detailed examination of the physical and mathe-
matical background that is involved in. the measurement of the
light after it has traveled through the atmosphere is necessary to
understand some of the problems associated with DOAS. Since
this is beyond the scope of this paper, only a qualitative descrip-
tion will be given here. For a more thorough mathematical de-
scription of DOAS refer to Stutz and Platt (1996).
Figures B-l A, B-1B, and B-1C illustrate how the spectral
shape of an absorption spectrum—for example, here the NO2
absorption (Harder et al., 1997)—changes during the measure-
ment. After the path through the atmosphere, the shape is
dominated by the natural spectral width of the trace gas absorp-
tion. In some trace gases, like NO2, this leads to structures with
spectral widths in the 100-picometer (pm) range overlaid by
narrower structures in the 1- to 10-pm range (see Figure B-l a).
Other trace gases such as HONO and NO3 do not show this very
narrow absorption, and the absorption bands have widths in the
range of 0.1 to 3 nanometers (nm) (Platt, 1994). Due to the
-------
limited spectral resolution of the spectrographs (in the range
0.1-1 nm full width, half maximum) used in DO AS instruments,
the shape of this spectrum changes during the measurement (See
Figure B-lb). The mathematical description of this process is the
convolution of the spectrum shown in Figure B-la, with the
instrument function H of the spectrometer. Since the shape of the
spectrum changes during the measurement, the depth of the
absorption bands also changes. Therefore, the differential optical
absorption D' will change depending on the instrument. It is thus
essential for DOAS to determine the differential absorption cross
section that corresponds to the respective absorption spectrum
and instrument. As will be described in more detail below, this
can be done either by determining H and simulating the absorp-
tion spectrum or by measuring it with the same instrument. The
last step of the measurement process is detection of the spectrum
at the exit of the spectrograph by a detector, digitizing it in
intensity and wavelength (see Figure B-1C).
B.3 Experimental Realization
DOAS measurements can be subdivided into several
steps. The first step is the actual measurement of the atmospheric.
trace gas absorption, which is accompanied by several auxiliary
measurements. In the next step, the absorption cross sections
have to be adapted to the instrumental function of the instrument,
or alternatively, a cell with the known trace gases has to be
measured with the same instrument. If the DOAS instrument is
stable, this step has to be repeated only from time to time,
particularly after any changes of the instrument. The last step is
the filtering and separation of the absorptions in a so-called
fitting procedure. This last step is also used to estimate the error.
of the actual measurement. The following section will discuss the
different parts of DOAS measurements, pointing out the crucial
aspects of the individual steps.
B.3.1 Experimental Setup
In general, DOAS instruments consist of a light source,
transfer optics to send the light through the atmosphere, and a
spectrograph detector system that records the spectral structure
of the light.
Three different light-path setups are currently used in
DOAS instruments. The classical arrangement developed by Platt
et al. (1979) has a xenon arc lamp combined with a collimating
mirror on one end of the light path and a receiving telescope on
the other. Although this setup is still used in some commercial
DOAS instruments, it has been replaced by an arrangement that
folds the light path once by aiming a combined sending and
receiving telescope at one or more quartz cube corner retro-
reflectors (Axelsson et al., 1990). The main advantages of this
approach are its simpler setup and the ability to aim the telescope
at different retroreflector setups and thus vary length, orientation,
and height of the light path. In the third setup the light path is
folded into a small length (on the order of 1-20 m) by multiple
reflection on mirrors (White, 1942; White, 1976;Ritzetal.,1992).
This setup is experimentally more demanding, but has the advan-
tage of better comparability with in situ techniques (see below).
Several types of spectrographs have been used in the past.
The type is not as important as certain properties of their
construction. The most important of these properties are the
stability of instrument function and the spectral position. Since
changes can often be attributed to temperature drift, the temper-
ature of the spectrometer is usually kept constant to within ±0.5
K by means of regulated heating and insulation (Stutz, 1996).
Another problem, especially in the UV, is stray light from wave-
lengths other than the desired wavelength in the spectrograph
(Pierson and Goldstein, 1989; Stutz, 1996). Careful spectrograph
design and the use of bandpass filters can reduce stray light. In
any case, an accurate characterization of the spectrograph stray
light should be performed.
The spectral resolution chosen has to be high enough to
resolve the differential structures of the different gases, but low
enough to allow a large enough wavelength interval to be ob-
served (Platt, 1994). To avoid aliasing effects, a certain amount
of oversampling is necessary. For pixel array detectors a typical
minimum width of the instrument function of 4 pixels is
necessary.
Two types of detectors are used in DOAS instruments.
The slotted disk machines (SDMs) developed by Platt (1979)
move a slit rapidly over the focal plain of the spectrograph,
scanning the spectrum by converting wavelength-intensity infor-
mation into time-intensity information with a photomultiplier.
More modern instruments use photodiode array detectors (PDAs).
Due to the multiplex advantage of measuring all wavelengths in
an interval simultaneously, these solid-state detectors offer about
60 times higher light throughput than the SDM. Although an
improvement of the detection limit has been achieved, the physi-
cal detection limit imposed by this higher sensitivity has not been
reached due to spectral structures that are most likely induced by
the detector itself (Stutz and Platt, 1992; Mount et al., 1992;
Stutz and Platt, 1997).
The determination of trace gas concentrations by DOAS
requires several auxiliary measurements to ensure optimal per-
formance of the instrument (Stutz, 1996):
• Atmospheric background light intensity caused by aero-
sol scattering of sunlight into the telescope has to be
measured by blocking the lamp during a measurement
cycle. This background light has to be subtracted from
the absorption spectrum.
• The shapes of the emission peaks of xenon lamps, used in
most DOAS instruments, change with time and alignment
of the instrument and should be monitored regularly.
• As described above, the instrument function H and the
spectral position are the most important parameters to
analyze DOAS spectra (see below). Their measurement
-------
is therefore essential. Since atomic emission lines
are two orders of magnitude narrower than the
typical resolutions of DOAS spectrographs,
measurement of these lines offer a sufficiently
accurate description of H and the spectral position.
• One way to overcome the need for regular determina-
tion of H and the spectral position is to calibrate the
instrument with absorption cells containing known con-
centrations of a trace gas. While this seems to be an
easy way to avoid the determination of H, one of the
most important advantages of DOAS, the absolute
measurement of trace gas concentration based on the
absorption cross section (see below) is given up. It is
often also difficult to accurately determine the concen-
tration of a trace gas in a cell. For example, the con-
centration of NO2 in a cell depends on its temperature
due to the equilibrium with its dimer N2O4 and the
photolysis in the light beam (Hofmann et al., 1995).
Special care has to be taken in preparing the cell to
avoid water on the cell walls, which will form nitrous
acid, HONO. Since MONO absorbs in the same wave-
length range as NO2, its presence will introduce prob-
lems in the analysis of atmospheric spectra. Experience
has shown that water-free cells filled with a mixture of
a small amount of NO2 and ultrapure O2 at atmospheric
pressure show reasonable performance if the NO2-N2O4
equilibrium is accounted for (Harwood and Jones,
1994; Hofmann et al., 1995). This approach is experi-
mentally very challenging and expensive for unstable
gases like HONO and NO3.
Inexpensive computers with large storage capabilities and
storage media control today's DOAS instruments. These fast
computers make it possible to perform an on-line analysis, which
delivers real-time NO2 concentrations. Nevertheless, the pos-
sibility of instrumental problems like shifts in spectral position
and aging of the lamp makes it highly desirable, if not necessary,
to store and archive all atmospheric absorption spectra as well as
all auxiliary spectra. Storing the spectra also provides an oppor-
tunity to later reanalyze spectra to prove unequivocally the
presence of a certain trace gas concentration.
B.3.2 Analysis of DOAS Spectra
After the measurements described in the preceding section
are made, the actual concentrations have to be determined by
analyzing the atmospheric absorption spectra (Platt,1994; Stutz
and Platt, 1996). To obtain reproducible DOAS results, it is
necessary that the different steps described below are well docu-
mented for each instrument.
B.3.2.1 Determination of the Absorption Cross Section
for the Instrument
As described above, measurement with an instrument with
a limited spectral resolution changes the shape and size of the
absorptions and also the corresponding absorption cross section.
As long as these changes are identical for the atmospheric mea-
surement and the absorption cross section, Lambert-Beer's law
can be used to determine the concentration.
Two basic approaches can be used to determine instrument-
specific absorption cross sections. The first approach is the cali-
bration of the instrument by measuring the absorption of the gas
in a reference cell with known content. As described above,
several experimental problems are connected with this approach
for NO2, and its application to unstable gases such as HONO or
NO3 is even more difficult. The main disadvantage is the loss of
DOAS as an absolute technique.
The second approach is to use spectrally high-resolution
absorption cross sections (about 20-100 times better resolved
than the actual instrument), which have been published in the
literature, for example, for N02 (Harwood and Jones, 1994;
Merienne et al., 1995; Harder et al., 1997; and Vandaele et al.,
1998). By using the measured instrumental function H and the
spectral position, these cross sections can be adapted to the
individual instrument with high accuracy (Stutz, 1996). The
adaptation algorithm is based on the convolution of a simulated
high-resolution spectrum (calculated with Lambert-Beer's law)
with H, and the integration of this spectrum according to the
spectral position of individual pixels. Since the accuracy of the
published absorption cross sections is approximately 3-10% (see
Harder et al. [1997] for a review of published NO2 absorption
cross sections) and the error of the adaptation procedure is
approximately 1 %, the total error of this method is approximately
3-10%. It should be added that to obtain more accurate measure-
ments of the absorption cross-section concentrations in the
future, past measurements could be corrected. Besides its easier
experimental realization (only the measurement of atomic
emission lines are required), the accuracy is certainly higher than
for the first approach, since it is improbable that a calibration in
the field is more accurate than intensive laboratory studies. The
origin of the absorption cross sections should always be reported
with DOAS results. In all research DOAS instruments, this last
approach is used because it can be considered an absolute
method to determine ambient trace gas concentrations.
B.3.2.2 Filtering Procedure
The separation of the absorptions into differential and
broadband parts is an essential part of DOAS. No unique or
favored method exists to perform this separation. In general, all
high-pass filtering procedures can be used, as long as they do not
change the differential absorption too much. Independent of what
-------
filtering method is used, it is essential that the absorption cross
section, or a simulated reference absorption spectrum used in the
analysis, is treated in the same way and that the differential
absorption cross section fits the filtered absorption spectra. The
filtering procedure should be well documented since it can
strongly influence the size of the absorption and the differential
cross section.
It is also possible to omit a filtering of the spectra and
include a higher order polynomial in the separation procedure to
describe the broadband structures (see below).
B.3.2.3 Separation Algorithm
The analysis of the absorption spectra, is one of the main
problems in the application of DOAS. Figure B-3 shows the
different absorbers and other spectral structures that have to be
separated in the analysis of NO2 and HONO in the wavelength
range from 300 to 400 nrn.
The evaluation procedure is typically based on a model
that describes the physical behavior of the logarithm of DOAS
spectra, ln(I'0(A.)/I(A.)). The model is most often a function that
consists of a linear combination of reference spectra (which are
obtained by simulation or measurement for the specific instru-
ment) and a polynomial to describe broad structures (for a mathe-
matical description, see Stutz and Platt, 1996). A linear least
squares fitting procedure is then used to optimize the scaling
factors for the reference spectra, aj5 and polynomial parameters,
with the goal of minimizing the difference between measured and
modeled spectra. The scaling factors as are the result of the fit
and can be used to calculate the concentration Cj of the respec-
tive trace gases: Cj = a/(0j x L). In the case of filtering
procedures applied to the spectra before the actual separation
process, the degree of the polynomial can be set to 1. If addi-
tional filtering is needed, higher degrees of polynomials can be
chosen.
One of the problems in DOAS is the shift in spectral
position of the atmospheric spectrum. If this shift is large (in the
range of tens of pixels), systematic errors are introduced due to
the imperfect alignment. More advanced analysis procedures
allow the automatic alignment of the references to the measured
spectrum. This can lead to considerable improvement of the
analysis, but also introduces additional problems that supercede
the scope of this report (see Stutz and Platt, 1996, for additional
information).
Numerical simulations have shown that no interferences
between absorbers larger than the error of the concentration
calculated by the fit occur, as long as the reference spectra are
accurately aligned and there are no unidentified structures in the
spectra (Stutz, 1996). Problems can occur if the quality of the
reference spectra is poor or if they contain structures of more
than one absorber (as can be the case of HONO and NO2).
300 310 320 330 340 350 360 370 380
310 320 330
340 350 360 370
wavelength [nm]
Figure B-3. Known differential absorptions and mercury emis-
sion lines in the wavelength range typically used to monitor NO2
and HONO. The region around 370 nm is often preferred for NO2
measurements, because around 440 nm, where the strongest
differential NO2 absorptions are found, Xe lamps have strong
emission peaks.
B.3.2.4 Errors
Besides deriving the trace gas concentration, the task of
the evaluation procedure is also to estimate the random error Aaj
of the scaling parameters 3j and, therefore, of the measured trace
gas concentrations (Stutz and Platt, 1996; Hausmann et al.,
1999). B oth tasks can be solved with linear least squares methods
if no instrumental problems are encountered. Aaj is proportional
to a factor determined by the fitted reference spectra and the
polynomial, and the noise in the spectrum or, if present, unex-
plained structures other than noise (Stutz and Platt, 1996). It is
important to note that, due to changing conditions in the atmo-
sphere, the noise, and therefore the error, of individual measure-
ments can vary and always has to be reported.
While the above-described determination of the error
considers only random errors, several sources of systematic
-------
errors have been found. An uncorrected spectral alignment or
change in dispersion of the reference spectra due to a drift of the
spectrograph, as discussed above, can lead to systematically
wrong concentrations. Also, additional spectral structures caused
by the instrument or by unknown absorbers are often found to
introduce problems.
B.4 Theoretical Considerations about DOAS
As with all experimental methods, several requirements
must be met to make it useful for the intended measurement. The
following aspects have to be considered for DOAS:
B.4.1 Linearity
The linearity of DOAS has so far not been discussed in
the literature. To determine the linear behavior of DOAS,
numerical simulations have been performed as follows: The high-
resolution absorption cross section by Harder et al. (1997) has
been used to simulate absorption spectra in the column density
(column density = C x L) range from 5 x 1015 cm"2 to 2 x 1019
cm'2 (the lower end was determined by truncation errors in the
simulation). These spectra were then analyzed by the DOAS
procedure using the spectrum for 5 x 1016 cm"2 arbitrarily as the
reference spectrum. The concentrations derived by this analysis
were compared to those used in the simulation. If the column
density was smaller than 2 x 1018 cm"2, no deviation from the
linear behavior could be found above 0.1 %. At column densities
above 2 x 1018 cm"2, saturation of the NO2 absorption occurred.
In summary, as long as the high-resolution absorption is not in
saturation, DOAS is linear with respect to the relation of dif-
ferential optical density to concentration. If the limit of saturation
is reached, the analysis of smaller absorption bands of the same
trace gas or shortening the light path might be considered as
possible solutions.
B.4.2 Random Error of the Measurement
As discussed above, the random error of DOAS measure-
ments, in a case where no instrumental problems or unknown
spectral structures are encountered, is proportional to the noise
in the spectrum. A short qualitative description of the behavior
of the random error for only one absorber in different situations
will be given. A more quantitative description can be found in
Stutz and Platt (1996).
If we consider the errors of o(A.) and L as nonrandom, the
error of the concentration C is proportional to the error of the
differential absorption cross section:
AD' =
AC =
1
xAD'
(B-5)
The error of D' depends only on the noise in the spectra and,
therefore, on the error of I(A') and I'0(V):
(B-6)
If we assume, for simplicity, that I'0(A,') = (1 + a) x I(A/), where
the factor a gives a nonlinear measure of the strength of the
absorption, we can gain some qualitative information.
In the case of very small absorption, a will approach zero
and we can write the error of C as:
AC = — X A/2 X .° (B-7)
The error will therefore approach a constant value, which is
dependent only on the light intensity and is independent of the
concentration. In most detectors the intensity is proportional to
the number of photons counted by the detector, and, therefore,
Poisson statistics applies for the error of the intensity. In this case
the error of the concentration will be proportional to (r0(A/))"a5.
The situation is more difficult when we consider large
absorptions. In this case the relative error of I(A/), and also
possibly the error in I'0(A.'), increases because the absorption
(absolute and differential) will decrease the intensity. The error
of the concentration will then depend on the concentration and on
the light intensity. An exact description of this phenomenon can
only be achieved by considering the shape of the absorption cross
section and is beyond the scope of this report.
The material described here only considers the error of
the intensity at two wavelengths. Since DOAS typically uses a
wavelength range, the error is much smaller. Similar to an
averaging procedure, one expects approximately a 10 times
smaller error if 200 pixels of a detector are used for an analysis.
The exact improvement of the analysis depends on the shape of
the spectrum and can be derived by considering a least squares
procedure. Stutz and Platt (1996) have mentioned that one can
derive precise concentrations even if the noise of the spectrum is
larger than the actual absorption structures.
8.4.3 Possible Systematic Errors
The most common systematic errors in DOAS are spectral
misalignment and unknown spectral structures. While spectral
misalignment can often be corrected in measured spectra, this
procedure introduces additional errors and also makes the
analysis very difficult. Also, in contrast to the linear least squares
fit, the nonlinear methods used to correct misalignment are non-
analytical and special care must therefore be taken when
applying them (Stutz and Platt, 1996).
So far no method has been found to correct for unknown
spectral structures, and the only possible way to attack this
problem is to identify the sources of the structures. In the case of
unknown absorbers, the reference spectra of these absorbers have
to be included in the analysis. If the structures are caused by
instrumental problems, they have to be removed by changes in
the hardware.
-------
BAA Accuracy of DO AS
The overall accuracy of DOAS for NO2 and other trace
gases is usually dominated by the uncertainties in the absorption
cross sections. Only in situations of low or very high concentra-
tions do the random errors of the intensity dominate. It is there-
fore desirable to increase the accuracy of the absorption cross-
section data.
B.5 Current State of the Art
To assess the current state of the art of DOAS instruments
one has to distinguish between commercial and research instru-
ments. Not much is known about the available commercial
DOAS instruments because their construction and especially
their analysis software is proprietary. The current state of the art
of research instruments is better documented in the literature
(Platt, 1994; Plane and Smith, 1995; Stutz and Platt, 1997; and
references therein). These instruments are all similar in their
construction, the main differences being in the length of the light
path and in the analysis software. Typical detection limits for
nitrogen oxides are listed in Table B-l (Platt, 1994; Plane and
Smith, 1995; Stutz and Platt, 1997).
Table B-1. Detection Limits of Research DOAS Instruments for
Various Nitrogen Oxides
Trace Gas
NO
NO2
MONO
NO3
Typical Absorption
Path Length (km)
0.2
5
5
10
Detection Limit
(pptv)
200-300
50-100
50-100
1-2
The determination of the accuracy of DOAS instruments
is problematic for several reasons. Since DOAS instruments are
typically not calibrated, the only possibility to compare instru-
ments is by simultaneous measurements in ambient air. The
largest uncertainty of strongly absorbing species like NO2 is
typically the error of the absorption cross section. Experiments
in the past have shown that the agreement between DOAS
systems is improved if all instruments use the same absorption
cross section to derive the concentrations (Camy-Peyret et al.,
1996). Another problem with intercomparing DOAS instruments
and comparing them to other techniques is the spatial averaging
over the absorption path (Camy-Peyret et al., 1996). Comparing
measurements of DOAS systems of several kilometers path
length with in situ measurements close to sources, i.e., in a city,
is particularly problematic. In the following paragraphs several
of these comparisons are discussed.
Two intercomparisons of research DOAS instruments
were performed in the framework of the Tropospheric Optical
Absorption Spectroscopy (TOPAS) project of the EUROTRAC
program (Bosenberg et al., 1997). The first campaign in
Brussels, Belgium (Camy-Peyret et al., 1996), revealed many of
the problems discussed above. Eight DOAS instruments were
compared in a polluted urban environment. Despite the dif-
ferences of optical setups of the instruments, in an environment
with nearby NOX sources the agreement was very good, with
correlation coefficients above 0.97 for the measured concen-
trations of O3, SO2, and NO2. The authors concluded that the
general agreement among the instruments was better than 2 ppbv
NO2 in the relatively polluted environment.
To reduce the problems of spatial averaging, a second
campaign took place in fall 1994 at the Weybourne Atmospheric
Observatory in north Norfolk, U.K. This site typically receives
well-mixed air, and differences due to spatial averaging were not
expected. Seven DOAS instruments were compared to each other
and to in situ analyzers for O3, SO2 (both EPA-approved sys-
tems), and NO2 (chemiluminescence + photolytic converter). The
intercomparison of the DOAS instruments and the comparison
with other techniques showed very good agreement. Unfortu-
nately problems with the chemiluminescence NOX monitor pre-
vented an accurate comparison of the NO2 concentrations. All
DOAS instruments agreed with the in situ O3 and SO2 analyzers
within 5-10% (Bosenberg et al., 1997).
A more recent comparison of state-of-the-art DOAS
instruments was published by Geyer et al. (1999). During a cam-
paign near Berlin in 1998 measurements of NO3 by DOAS and
by MIESR were compared, showing agreement within the error
limits of the instruments. During the same campaign a DOAS
system based on an open 15-m base path-length multireflection
cell was compared to an in situ NO2 chemiluminescence analyzer
with photolytic converter (Patz et al., 2000). In this setup the
DOAS instrument only observed the small air mass inside the
open cell. The cell was mounted within a few meters of the sam-
ple inlet of the in situ analyzer; the values should therefore be
free of any spatial averaging effects. Figure B-4 shows a
correlation plot of the DOAS and chemiluminescence measure-
ments. The two instruments agree within 1% over a 2-week
period (=1000 data points).
The most recent intercomparison was made during the
SOS field campaign in Nashville, TN, in summer 1999. Three
different techniques were compared with each other: a system
based on the photolytic conversion of NO2 and detection of NO
by chemiluminescence, LIF, and the newest generation of DOAS
research instruments with a 1.35-km folded light path (total
absorption length 2.7 km). Figure B-5a shows the measurements
of the DOAS and the chemiluminescence instrument during a 3-
day period in June. To compare the DOAS values, which were
recorded with a lower time resolution, we averaged the chemi-
luminescence measurements over the exposure time of the DOAS
system. The correlation plot (see Figure B-5b) shows that on
average the instruments agree within 2%. The larger scatter in the
data compared to Figure B-4 can be explained by the spatial
-------
averaging in this case. The in situ monitors sampled at a 10-m
height, while the DOAS light beam was located about 100 m
away covering a 2- to 36-m height interval and a horizontal
distance of 1.3 km. The DOAS system had an average detection
limit of 200-400 pptv.
B.6 Summary
DOAS is a powerful technique to monitor air pollutants
like NO2 or NO. Its main advantage is the ability to measure
absolute trace gas concentrations over an extended light path. For
air quality monitoring this averaging is an advantage because
influences of strong local sources are diminished. The compari-
son of Figure B-4 with Figure B-5b show how large these
uncertainties can become.
Intercomparison experiments have shown that the DOAS
instruments agree well with each other and in situ techniques if
the spatial averaging of the DOAS is eliminated. Larger scatter
is observed in cases where the long-path DOAS data are com-
pared with in situ measurements. The detection limits for NO2
reported in the literature vary for different light path lengths, but
they are in general approximately 50-100 pptv for a 5-km
absorption path length (Platt, 1994; Plane and Smith, 1995; Stutz
and Platt, 1997). State-of-the-art instruments give a HONO
detection limit of approximately 50-100 pptv (5-km light path).
DOAS is currently the only technique able to monitor NO3
radicals over an extended period. The detection limit for NO3 is
approximately 1-2 pptv.
The main challenge today for DOAS is integrating the
complexity of the method into an instrument that can be operated
by nonspecialists while avoiding all the possible problems
outlined in this report. Several commercial instruments attempt
this, but it remains to be seen if they will succeed.
22
oo
00
3
Q
ID
i
20-
18-
16-
14-
12-
10-
8-
6-
4-
2 -
0
Y=
R =
(0.036 +/- 0.019) + (1.006 +/- 0.005) *X
0.99
0 2 4 6 8 10 12 14 16 18 20 22
NO2 (ppbv) Chemiluminescence
Figure B-4. Intercomparison of a Chemiluminescence detector with photolytic converter (B.
Alicke, A. Geyer, H.W. Patz, and A. Volz-Thomas unpublished data) and a DOAS system
with a 15-mm-long multireflection cell setup. Both instruments were located within a few
meters from each other. Due to the optical setup of the DOAS system, the problem of spatial
averaging was avoided. The agreement between the two techniques is excellent. The
equation in the graph is the result of a weighted linear least squares fit. Errors are two times
standard deviation.
-------
60
50
£ 40
&,
£ 20
10
0
Chemiluminescence
DOAS
6/21/99 00:00 6/22/99 00:00 6/23/99 00:00
Y = (1.5 +/- 0.1) + (0.983 +/- 0.003) * X
R = 0.94
10 20 30 40 50 60
NC>2 (ppbv) Chemiluminescence
Figure B-5. Intercomparison of a long-path DOAS system (1.3-km light path length) with an
in situ Chemiluminescence analyzer with photolytic converter (see Appendix C in this report)
during the SOS '99 field intensive in Nashville, TN: (a) shows the mixing ratios determined
by both instruments. The differences can mostly be explained by the different air masses
observed by both instruments, (b) shows a correlation plot, where the data of the Chemi-
luminescence analyzer were integrated over the longer integration time of the DOAS mea-
surements. The slope of 0.983 ± 0.03 shows the excellent agreement between the instru-
ments. The scatter is again caused by the observation of the different air masses. This also
explains the non-negligible intercept.
-------
B.7 References
Axelsson, H., Galle, B., Gustavsson, K., Regnarsson, P., and
Rudin, M. 1990. A transmitting/receiving telescope for DOAS-
measurements using retroreflector technique. Techn. Dig. Series
4:641-644.
Bosenberg, D., Brassington, D., and Simon, P.C., eds. 1997.
Instrument Development for Atmospheric Research and Moni-
toring. Berlin: Springer-Verlag.
Camy-Peyret, C, Bergqvist, B., Galle, B., Carleer, M., Clerbaux,
C, Colin, R., Fayt, C., Goutail, F., Nunes-Pinharanda, M., Pom-
mereau, J.P., Hausmann, M., Platt, U., Pundt, I., Rudolph, T.,
Hermans, C., Simon P.C., Vandaele, A.C., Plane, J.M.C., and
Smith, N. 1996. Intercomparison of instruments for tropospheric
measurements using differential optical absorption spectroscopy.
J. Atmos. Chem. 23:51-80.
Geyer, A., Alicke, B., Mihelcic, D., Stutz, J., and Platt, U. 1999.
Comparison of tropospheric NO3 radical measurements by
differential optical absorption spectroscopy and matrix isolation
electron spin resonance. J. Geophys. Res. 104:26097-26105.
Harder, J.W., Brault, J.W., Johnston, P.V., and Mount, G.H.
1997. Temperature dependent NO2 cross sections at high spectral
resolution. J. Geophys. Res. 102:3861-3879.
Harwood, M.H., and Jones, R.L. 1994. Temperature dependent
ultraviolet-visible absorption cross sections of NO2 and N2O4—
Low-temperature measurements of the equilibrium constant for
2NO2 - N2O4. J. Geophys. Res. 99:22955-22964.
Hausmann, M., Brandenburger, U., Brauers, T., and Dorn, H.P.
1999. Simple Monte Carlo methods to estimate the spectra evalu-
ation error in differential-optical-absorption spectroscopy. Appl.
Opt. 38:462-475.
Hofmann, D., Bonasoni, P., Demaziere, M., Evangelisti, F.,
Giovanelli, G., Goldman, A., Goutail, F., Harder, J., Jakoubek,
R., Johnston, P., Kerr, J., Matthews, W.A., McElroy, T.,
McKenzie, R., Mount, G., Platt, U., Stutz, J., Thomas, A.,
Vanroozendael, M., and Wu, E. 1995. Intercomparison of UV-
visible spectrometers for measurements of stratospheric NO2 for
the network for the detection of stratospheric change. J.
Geophys. Res. 100:16765-16791.
Merienne, M.F., Jenouvrier, A., and Coquart, B. 1995. The NO2
absorption spectrum 1. absorption cross-sections at ambient
temperature in the 300-500 nm region. J. Atmos. Chem. 20:
281-297.
Mount, G.H., Sanders, R.W., and Brault, J.W. 1992. Interference
effects in Reticon photodiode array detectors. Appl. Opt.
31:851-858.
Noxon J.F. 1975. Nitrogen dioxide in the stratosphere and tropo-
sphere measured by ground-based absorption spectroscopy.
Science 189:547-549.
Pa'tz, H.-W., Corsmeier, U., Glaser, K., Kalthoff, N., Kolahgar,
B., Klemp, D., Lerner, A., Neininger, B., Schmitz, T., Schultz,
M., Slemr, J., Vogt, U., and Volz-Thomas, A. 2000. Mea-
surements of trace gases and photolysis frequencies during
SLOPE96 and a coarse estimate of the local OH concentration
fromHNOj formation. /. Geophys. Res. 105:1563-1583.
Perner, D., and Platt, U. 1979. Detection of nitrous acid in the
atmosphere by differential optical absorption. Geophys. Res.
Lett. 6:917-920.
Pierson, A., and Goldstein, J. 1989. Stray light in spectrometers:
causes and cures. Lasers and Optronics, Sept:67-74.
Plane, J.M.C., and Nien, C.F. 1992. Differential optical absorp-
tion spectrometer for measuring atmospheric trace gases. Rev.
Sci. Instr. 63:1867-1876.
Plane, J.M.C., and Smith, N. 1995. Atmospheric monitoring by
differential optical absorption spectroscopy. In Spectroscopy in
Environmental Sciences, eds., R.E. Hester and R.J.H. Clark, pp.
223-262, London: Wiley.
Platt, U. 1978. Dry deposition of SO2. Atmos. Environ. 12:
363-367.
Platt, U. 1994. Differential optical absorption spectroscopy
(DOAS). InMonitoring by Spectroscopic Techniques, ed., M.W.
Sigrist, New York: John Wiley.
Platt, U., Perner, D., and Pa'tz, H. 1979. Simultaneous measure-
ment of atmospheric CH2O, O3, and NO2 by differential optical
absorption. J. Geophys. Res. 84:6329-6335.
Platt, U., and Perner, D. 1983. Measurements of atmospheric
trace gases by long path differential UV/visible absorption spec-
troscopy. In Optical and Laser Remote Sensing, eds., D.K.
Killinger and A. Mooradien, pp. 95-105, Berlin: Springer.
Ritz, D., Hausmann, M., and Platt, U. 1992. An improved open-
path multireflection cell for the measurement of NO2 and NO3.
Proc. Europto series. Optical methods in atmospheric chemistry,
eds., H.I. Schiff and U. Platt, 1715:200-211.
-------
Stutz, J. 1996. Messung der Konzentration tropospharischer
Spurenstoffe mittels Differentieller-Opischer-Absorptionsspek-
troskopie: Eine neue Generation von Geraten und Algorithmen.
Ph.D. thesis, Heidelberg: Univ. Heidelberg.
Stutz, J., and Platt, U. 1992. Problems in using diode arrays for
open path DOAS measurements of atmospheric species. Proc.
Europto series. Optical methods in atmospheric chemistry, eds.,
H.I. Schiff and U. Platt, 1715:329-340.
Stutz, J., and Platt, U. 1996. Numerical analysis and estimation
of the statistical error of differential optical absorption spectro-
scopy measurements with least-squares methods. Appl. Opt.
35:6041-6053.
Stutz, J. and Platt, U. 1997. Improving long-path differential
optical absorption spectroscopy (DOAS) with a quartz-fiber
mode-mixer. Appl. Opt. 36:1105-1115.
Vandaele, A.C., Hermans, C., Simon, P.C., Carleer, M., Colin,
R., Fally, S., Merienne, M.F., Jenouvrier, A., and Coquart, B.
1998. Measurements of the NO2 absorption cross-section from
42 000 cm'1 to 10 000 cm'1 (238-1000 nm) at 220 K and 294 K.
J. Quant. Spect. Rad. Trans. 59:171-184.
White, J.U. 1942. Long optical paths of large aperture. J. Opt.
Soc. Am. 32:285-288.
White, J.U. 1976. Very long optical paths in air. J. Opt. Soc. Am.
66:411-416.
-------
Appendix C
Photolytic Conversion of Ambient NO2
by
Eric Williams
NOAA Aeronomy Laboratory, Boulder, CO
Photolytic conversion of NO2 to NO is an attractive
means of measuring NO2 for a number of reasons. Photolysis, as
opposed to thermal or catalytic conversion, is a process that is
understandable from a fundamental physical chemistry stand-
point. Absorption cross-section and quantum yield data for this
molecule have been accurately determined by a number of
investigators. Moreover, photolysis is in principle more specific
than thermal conversion. As such it is not subject to potential
interferences from thermally labile species such as PAN and
peroxynitric acid (HO2 NO2). Since absorption cross sections for
all of the principal reactive nitrogen oxides have been measured,
the wavelength range of radiation used for NO2 photolysis can be
adjusted to eliminate interferences from photolysis of other
compounds such as nitric acid (HNO3), PAN, and N2O5. There
are, however, two reactive nitrogen species that will be inter-
ferences due to photolysis that occurs in the wavelength region
used for NO2; these are nitrous acid (HONO) and nitrate radical
(NO3). In most cases, though, the levels of these species will be
small compared to NO2, and in the case of NO3 photolysis the
interference can be eliminated by the use of optical filters (see
below).
C.1 Physics and Chemistry of the
Measurement
Atmospheric levels of NO2 can be determined readily by
photolysis with UV radiation followed by measurement of the
increase in NO by chemiluminescence. This is effected by flow-
ing the sample gas through a photolytic converter, essentially a
quartz or Pyrex cell, that is irradiated by a source of intense UV
such as a high-pressure Xe lamp. The chemical reaction of
interest is
where hv is in the range 350-400 nm. Since not all of the NO2
entering the cell is converted, a conversion factor is defined that
relates the increase in the amount of NO that exits the cell to the
entering NO2 level. The conversion factor must also account for
a competing reaction that occurs in the cell:
NO + O3 ->• NO2 + O2
R2
where the ozone (O3) comes from ambient O3 or is produced by
the very fast reaction
O + O2
• O3
R3
that occurs in the cell during photolysis. Reactions R1-R3 are the
same reactions that occur in the atmosphere to interchange NO
and NO2 while conserving O3. These reactions are commonly
referred to as the photostationary state. Viewed from the per-
spective of these reactions, the photolytic converter does nothing
more than change the ratio of NO2 to NO during gas transit
through the cell.
The photolytic rate coefficient, J in units of s"1, that
describes the rate of reaction Rl is the wavelength-integrated
product of the absorption cross section, the quantum yield of NO
production, and the UV photon flux in the cell. The second-order
rate coefficient that describes the rate of reaction R2 is k, in units
NO, + hv -> NO + O
Rl
of cm3 molec"' s"1. If conditions in the photolytic converter can be
arranged such that reaction Rl proceeds much faster than reac-
tion R2, then the conversion factor (CF) relating the measured
NO increase to the NO2 concentration is
CF = {J*t/(J*t+k*[O3]*t)}* {l-exp(-J*t-k*[O3]*t) (C-l)
where t is the residence time of the gas sample in the cell and the
assumption is made that [O3] is large compared to [NO]. The
-------
residence time is typically determined by the ratio of the cell
volume to the gas flow rate (usually where both are determined
at STP). The use of this factor relies on knowledge of J, which
usually is difficult to determine because the photon flux through
the cell cannot be readily estimated. In practice, CF is determined
by direct calibration of the instrument with a known concen-
tration of NO2. Moreover, under carefully controlled conditions
such as using zero air as the sample matrix, the CF determined by
NO2 calibration can be used to calculate J using Equation C-l
above.
C.2 Instrumental Application of NO2
Photolysis
C.2.1 General Considerations
The use of photolysis in the determination of NO2 in the
atmosphere requires an instrument capable of sensitively mea-
suring NO, such as a chemiluminescence detector. However,
since NO is generally present in the atmosphere as well, a
photolytic/chemiluminescence NO2 instrument will actually mea-
sure some fraction of NOX (NOX = NO + NO2), that is, all of
ambient NO plus the fraction of NO2 photolyzed in the system.
Thus, in order to determine ambient NO2, a means of separately
determining ambient NO from NO produced during photolysis of
NO2 is necessary. If only one instrument is available, then NO
and NO2 may be measured sequentially and the NO present
during the NO2 measurement can be interpolated. In practice,
rather than turning the lamp on and off, which induces lamp
instability, this has been accomplished by means of a shutter that
is capable of blocking all UV radiation from the photolysis cell.
The interpolated NO is then subtracted from the measured NO2
(actually NOX) and the result divided by CF to yield NO2. A
more desirable arrangement is the use of two instruments, where
one is dedicated to the continuous measurement of NO and the
other dedicated to continuous measurement of NO2. Here, every
measurement of NO2 has a corresponding NO data point; thus, no
interpolation of NO is required, which results in a more complete
and more accurate data set. In this case, calculation of NO2
involves taking a difference between data from two separate
instruments, so care must be taken to ensure that no systematic
errors are present between the two instruments. This is best
accomplished by acquiring sufficient NO data from the NO2
(NOX) instrument to compare to data from the NO instrument.
Calibrations for NO and NO2 must be performed at
regular intervals regardless of the instrumental arrangement.
Instrument calibration can be conveniently accomplished by
addition to the sample airflow of a small known gas flow (<1%
of sample flow) containing a known mixing ratio of the species
of interest. Compressed gas cylinders of calibrated NO (generally
at low ppmv levels in N2) are readily available with uncertainty
to within ±2%. It is prudent, though, to continually evaluate the
standard against others to ensure stability. Compressed gas
cylinders containing calibrated mixtures of NO2 are also avail-
able, but there is some question as to long-term stability. Permea-
tion devices containing N02 are also commonly employed; these
devices must be closely temperature regulated and must be
regularly calibrated against other N02 standards to ensure accu-
racy. A third technique is the use of a small O3 generator
(typically a Hg penray lamp) coupled to the primary NO standard
in a flow system that can produce precisely known levels of NO2
by monitoring the loss of NO from O3 titration. The assumption
here is that mass balance applies with respect to NO and NO2
(i.e., no additional titration of N02 to NO3 and N2O5 with
subsequent loss to surfaces). Regardless of the method, the
concentration of added calibration gas should be similar to the
expected atmospheric concentrations.
An artifact signal is usually present when a photolytic
converter is used. This signal is observed when it is known that
no NO2 is present in the sample gas (e.g., zero air) and is
believed to arise from evaporation of NO or NO2 from material
deposited on the cell walls when irradiated. The artifact appears
to be diminished by use of a dichroic beamsplitter (see below)
and can be substantially reduced by the use of optical filters to
restrict the photolytic wavelength region to 350-400 nm. This
artifact is monitored by periodic evaluation of the system with
zero air, converted to an equivalent NO2 concentration, and
subtracted from the measured NO2 levels. This NO2 artifact level
can vary from very low to negligible (0-0.02 ppbv) to high
values (>0.10 ppbv) depending on the sampling environment. It
can be eliminated by washing the cell with cleaning solution
(dilute KF or KOH) followed by rinsing with deionized water.
However, it is always prudent to perform regular checks of the
system with zero air.
C.2.2 Practical Considerations
Implementation of the photolysis technique is technically
straightforward. A glass cell, which in practice is always a
cylinder, is installed in the sample gas flow stream. The cell is
coated on the outside with a UV reflecting material (common Al
foil works well) and irradiated with an intense UV light source.
These light sources consist of the lamp, a housing, a stable power
supply, and power cables. Because commonly available UV
sources also output light in other wavelengths, some means of
isolating the 350- to 400-nm region is required. The use of a
Pyrex cell will provide short-wavelength cutoff at about 320 nm,
and the use of a dichroic beamsplitter can separate UV from
longer wavelength light. This latter mirror prevents visible and
IR radiation from reaching the photolytic cell, which reduces the
temperature increase of the gas in the cell. This, in turn, reduces
interferences from thermally labile species such as PAN.
Additional optical filters must be used to limit the light entering
the cell to 350-400 nm. The benefits of these filters are (1)
increased measurement specificity to NO2 and (2) reduction (or
elimination) of the NO2 artifact. The disadvantage is a loss of
conversion efficiency from attenuation of the lamp output beam.
-------
The amount of the attenuation depends on a number of factors
such as the number of filters used, filter placement, and align-
ment. A shutter device is also installed between the light source
and the cell since there is always the need to measure NO levels
in the system, regardless of whether one instrument or two is
used for the measurement of NO2.
Equation C-l indicates that only J and t influence the
conversion of NO2 to NO in the cell. With modern flow control
devices it is usually an easy matter to control gas flows (and
residence time) very precisely. It is therefore imperative that light
sources be used that not only have the required intense output in
the UV but that also are highly stable in output. Other require-
ments also exist depending on the sampling platform for the
instrument (e.g., aircraft) or the measurement program (e.g.,
research vs. monitoring). A commonly used light source is a 300-
W high-pressure short-arc Xe lamp with an integral reflector.
This type of lamp can be obtained with a built-in 320-nm short-
wavelength cutoff filter, which is required since otherwise O3
would be produced in the sample gas stream (and in the ambient
environment around the lamp). Power supplies to drive these
lamps can provide stable output regulated to better than ±5%.
This type of lamp is very convenient to employ because the light
beam is collimated and thus easily introduced into photolytic
converter cells. Using these lamps and cells with gas residence
times of 3-5 s, CF values of 0.4-0.5 are readily obtainable.
Under these conditions J is approximately 0.1-0.2 s"1. This
provides more than sufficient sensitivity to NO2 for most chemi-
luminescence NO detectors. However, gas residence times of 3-5
s also provide sufficient time for reaction R2 above to become
significant, and this situation requires that the measured concen-
trations be corrected during the data reduction process. Further,
longer gas residence times also may allow thermolytic conver-
sion of compounds like PAN, which can be a significant
interference with respect to NO2. A better arrangement is to
reduce t to less than 1 s in Equation C-l above, but this requires
that J be increased to 0.6 s"1 or greater. This has been
accomplished.
C.2.3 Current State of the Art
To date three approaches have been employed to obtain
adequate conversion efficiency at short gas residence times (see
Figure C-l). One technique has been to employ two separate Xe
UV sources to irradiate one cell; a second method uses a newly
developed quasi-focused Xe UV lamp; and a third method uses
a long-arc Hg-Xe lamp system.
A. One approach has been to use a cylindrical (3 cm i.d.
x 20 cm) photolytic converter with a conventional 3QO-W Xe
lamp shining into one end and a mirror at the other end. This
arrangement, with the use of optics for focusing the beams,
substantially increases the photon flux in the cell. Dichroic
beamsplitters are used to remove visible and IR radiation, but
because the energy input to the cell is substantial, the cell is
cooled with a water bath and the gas pressure inside the cell is
maintained at subambient (=250 torr) to eliminate condensation
of water from ambient air onto the inner walls of the cell.
Conversion efficiency for NO2 in this system has been measured
at 0.5 at a gas residence time of 0.5-1 s. Artifact levels of
equivalent NO2 with this system have been negligible, even when
ambient NO2 levels have been close to the detection limit.
B. Another approach has been to use a newly developed
Xe lamp that provides a quasi-focused output. By focusing, the
lamp output can be introduced into a smaller cell, increasing the
photon flux density. This system uses a 500-W Xe lamp (with a
coating to eliminate O3 production) and an aluminum-foil-coated
cylindrical quartz cell (1 cm i.d. x 30 cm) with a UV reflecting
mirror on the end opposite the lamp. The beamsplitter is not used
in this system in order to improve the conversion efficiency. The
cell is water jacketed and cooled by a recirculating water bath
with a water reservoir that is maintained at ambient temperature.
Gas pressure in the cell is slightly subambient due to the gas flow
(2000 seem) through it. Under these operating conditions no
condensation has been observed in the cell even with sample gas
dew point temperatures to 28 °C. Depending on the specific lamp
used, the NO2 conversion efficiency for this system has been
0.4-0.8 at a cell residence time of 0.6-0.7 s. Artifact levels for
this system have been observed at less than =0.04 ppbv in a rural
environment to slightly greater than 0.10 ppbv in an urban
environment.
C. A third approach uses a short-arc high-pressure 100-W
Hg lamp housed in an ellipsoidal reflector that focuses the lamp
output into a cylindrical cell. The Hg lamp provides a greater
fraction of radiation in the 350- to 400-nm region than do Xe
lamps; thus, a lower overall power is required. Long-wavelength
and short-wavelength cutoff filters are used to minimize artifact
levels to <10 pptv equivalent NO2. A small fan provides forced-
air cooling to the photolysis cell. Further, because the power
supply for the Hg lamp is quite compact and lightweight, this
entire system (cell, lamp, power supply, cooling components,
etc.) is comparable in size and weight to the power supplies of
the systems described in 1 and 2 above. During operation a
constant cell pressure (=250 torr) is maintained by a pressure-
flow feedback system, and under these conditions the conversion
efficiency (again, depending on the specific lamp used) was =0.5
at 0.8 s residence time.
C.2.4 Data Reduction Requirements
There are a number of data reduction issues that must be
kept in mind when the photolytic NO2 converter is used. The
most obvious requirement is the need for simultaneous
measurements of NO. If a single instrument is used to collect NO
and NO2 data, interpolation of NO between NO2 determinations
becomes necessary. If the NO concentrations are not varying
widely in the atmosphere, linear interpolation can provide a
reasonable approximation of the actual NO levels while NO2
(actually NOX) is being measured. On the other hand, if NO is
-------
(A)
UVMmr
CjHrtz Gel w Ughl
(3 X 20cm)
/
SampteOut
Stutter
(B) QjartzCMI
I
UVMrnr
Lamp Housng
\ \
r
^npteOut
&mpteh
j
. Cooing Fai
SOOWXeLanp
(focused output)
Stutter
Mnoied
<>Jartz
-------
changing significantly, then the calculation of NO2 becomes
highly uncertain. This uncertainty can be reduced somewhat by
only calculating NO2 levels at the time when the instrument
switches between NO and NOX measuring modes. However, this
technique reduces the total amount of data available. If another
instrument can provide simultaneous NO data, then the only
issues are that the data be truly simultaneous, that is, any time
lags between the measurements be compensated for or elimi-
nated, and that the NO concentrations from each of the
instruments be identical (and correct). Under these conditions
NO2 levels can be calculated at high time resolution (1 Hz or
faster). A potential problem under these circumstances is the
attenuation of concentration fluctuations in the NOX system due
to the presence of the photolytic cell, which acts as a mixing
chamber. This problem is manifested in the data reduction by the
presence of sharp negative calculated peak NO2 levels in the
middle of a broader positive peak. This problem is due to lower
time resolution in the NOX channel than in the NO channel,
which may not have an in-line photolysis cell. The solution to
this problem is to match exactly the plumbing configuration and
flow rates in the two instruments. This can be accomplished by
including a nonilluminated photolysis cell in the NO instrument
plumbing and matching the lengths of tubing between the two
instruments. This also eliminates any time lags between the mea-
sured NO and NOX and greatly simplifies data reduction.
Unfortunately, this solution can also generate another
problem. Since the photolytic converter can be viewed as a
mixing volume with a gas residence time determined by the flow
rate, there is the possibility that some of the NO in the ambient
gas sample can be converted by O3, according to reaction R2
above. For example, considering only the homogeneous NO/O3
reaction with ambient O3 at 100 ppbv, a gas residence time of 4 s
in the cell will convert almost 20% of ambient NO to NO2.
Moreover, peroxy radicals (HO2, RO2) in the atmosphere can
have an effect comparable to O3 and there also appears to be a
heterogeneous process (though smaller in magnitude) involving
conversion of NO to NO2 on the walls of the cell. Clearly, these
processes affect the accuracy of both the NO and NO2 data, since
these same reactions may reconvert some of the NO derived by
NO2 photolysis back to NO2. On the other hand, for a residence
time of 1 s in the cell only 5% of ambient NO is converted with
ambient O3 at 100 ppbv; similar effects are expected for con-
version by peroxy radicals and heterogeneous processes. It is
therefore apparent that short cell residence times are highly
desirable, but must be balanced by the need for adequate NO2
conversion efficiency (i.e., sensitivity). In any case, ambient O3
levels should be measured along with NO and NO2 in order to
correct the data for these effects. The methods for performing
these corrections have been published, but they are beyond the
scope of this document.
Assuming that instrument parameters are selected so that
adequate sensitivity is available with low levels of interferences
and artifacts, the calculation of NO2 from the measured NOX can
itself be a significant issue. Under many sampling conditions in
the atmosphere, the ratio of NO2 to NO is greater than 3-4, and
at night in many cases NO virtually vanishes due to titration with
ambient O3. However, near strong sources of NO, the ratio of NO
to NO2 can be close to, or even exceed, unity. Since the calcu-
lation of NO2 involves taking a difference between measured
NOX and NO, under the latter conditions the calculation involves
a small difference between relatively large numbers. This situa-
tion, of course, may generate large uncertainties in the derived
NO2 concentrations and points out a fundamental limitation of
this technique for NO2: The background levels against which
NO2 (again, actually NOX) is measured are usually varying,
sometimes rapidly, and are sometimes large enough to make the
NO2 determination highly suspect. Finally, the optimal imple-
mentation of this technique actually requires two independent
measurements to be conducted simultaneously, though it may be
argued that NO2 data without corresponding NO data are not
very useful.
C.2.5 Summary on Photolytic Converters
Photolysis has been used in the determination of NO2
since the late 1970s and has been shown to provide high-quality
data, based on results from field-based instrument intercom-
parisons. The technique is suited for ground as well as aircraft
operation, where fast time response is required. The method is
reasonably specific for NO2, with direct interferences known to
occur only with HONO and NO3. When implemented correctly,
interferences from NO3 and thermally labile species, most
notably PAN, are insignificant. For these reasons this method has
been the one of choice for researchers in atmospheric chemistry.
While generally reliable and straightforward to implement, even
via retrofit to existing instruments, the technique nevertheless
requires more expensive equipment, both initially and over the
long term, and more care and attention to detail during operation
than other methods of conversion such as with thermal con-
verters. Also, with the use of high-voltage power supplies and
very hot, high-pressure lamps that produce intense UV radiation,
safety concerns become paramount.
-------
Appendix D
Laser-Induced Fluorescence Detection of NO2
by
J.A. Thornton', P.J. Wooldridge' and R. C. Cohen1'2
'Department of Chemistry, University of California at Berkeley
^Department of Geology and Geophysics and Environmental Energy Technologies Division,
Lawrence Berkeley National Laboratory
Laser-induced fluorescence (LIF) detection of NO2 offers
the promise of a combination of simplicity, sensitivity, specificity,
accuracy, and long-term reliability that is unparalleled in existing
commercial or research-grade techniques for NO2 detection. Most
existing techniques rely on liquid-phase luminescent reactions of
NO2 or on chemical or photolytic conversion of NO2 to NO
followed by detection of the NO. These approaches are subject to
an array of interferences from other trace atmospheric species (see
Chapter 5). LIF detection, by contrast, can be configured so that it
has remarkable specificity for NO2. As we show below, the signal
can be generated using a high-resolution laser tuned to a specific
rotational feature in the electronic spectrum of NO2. The high
spectral resolution eliminates the possibility of a substantial false
signal from another molecule since the probability of another
molecule having identical rotational structure in its absorption
spectrum and comparable fluorescent behavior is near zero.
A typical LIF apparatus for NO2 detection uses a laser
oriented at right angles to both the sample flow and to a photo-
multiplier tube (PMT). Figure D-l is a schematic of the laser and
detection assemblies developed at the University of California,
Berkeley. Fluorescent photons are imaged onto the PMT and the
average count rate is proportional to the NO2 mixing ratio. The
detection limit is set by noise in the stray light that reaches the
PMT, and the sensitivity is determined by a product of the laser
power and the absorption and fluorescence efficiencies of NO2.
Early attempts at LIF detection of NO2 were not competitive with
other approaches because laser and detector technology were not
sufficiently advanced to allow high sensitivity or high specificity
in a compact, portable instrument.
Also, the demand for more specific techniques was low because
the limitations of other techniques were not as well understood
as they are today.
Recently, we have incorporated new laser and detector
technologies into an LIF sensor that is accurate (±5%, lo),
sensitive (15 pptv/10 s, S/N = 2; 1-pptv detection limit), pre-
cisely calibrated (drifts of less than 1%/month), free from inter-
ferences, and relatively low maintenance (Thornton et al., 1999).
This research-grade system was designed for aircraft use and in
the remote troposphere where NO2 concentrations below 100
pptv are typical. We have also made substantial progress toward
an LIF system based on a commercially available tunable diode
laser that demonstrates a sensitivity of 10 ppbv/10 s. These
results suggest that within a year or two LIF instrumentation with
all the capabilities necessary for atmospheric monitoring in urban
and rural areas will be available at a price competitive with
existing commercial instrumentation.
a) Early designs for LIF NO2 sensors. Before 1997,
designs for LIF detection of NO2 did not emphasize specificity.
The lasers used in these instruments were not tunable, precluding
discrimination against interferences by tuning on and off of a
rotational resonance unique to NO2. For example, Tucker et al.
(1975) used a He-Cd laser with nonresonant detection and
achieved a detection limit of 2040 pptv/10 s. George and
O'Brien (1991) used a frequency-doubled Nd3+:YAG laser (532
nm) with detection of fluorescence from 580-900 nm to achieve
a sensitivity to NO2 of 2500 pptv with 10-s averaging in a more
portable configuration. They took advantage of the long fluor-
escence lifetime of NO2 to discriminate against some of the
resonant background scatter and used an NO2 scrubbing step
-------
YAG -.
Figure D-1. Schematic of the UC Berkeley LIF NO2 instrument.
The core of the instrument is mounted on a breadboard, one side
holding the laser subsystem and the other side the detection
axis. A frequency-doubled Nd3+:YAG laser (YAG) at 532 nm
pumps a custom-built dye laser (DL). The output (585 nm) is
sampled by fused-silica beam splitters to monitor power, fre-
quency (by measuring transmittance through an N02 reference
cell shown as a cube), and line width (measured with an external
etalon [E]). Six photodiode detectors (PD) are used to measure
laser power at various points along the beam path. A set of
dispersion prisms (DP) is used to separate the 585-nm light from
the 532-nm light, which is then dumped. The 585-nm light is then
sent through a hole in the breadboard to the detection side to the
multipass white cell (WC). The pressure in the WC is measured
with a manometer (100-torr Baratron [B]). NO2 fluorescence is
collected and sent through a series of optical filters housed in the
filter changer (FC) to a photomultiplier tube in its TE-cooled
housing (PMT).
(reducing NO2 to NO using FeSO4) to demonstrate specificity of
the technique to NO2. In the first attempt to use spectroscopic
features of NO2 to ensure specificity in an atmospheric sensor,
Fong and Brune (1997) developed an LIF instrument employing
a tunable dye laser as the excitation source. They excited NO2 at
564 nm with detection using a red-sensitive PMT (8% quantum
efficiency) filtered to a bandwidth of approximately 200 nm. By
tuning the laser on and off of a rotational resonance in the
excitation spectrum, this instrument demonstrated its specificity
for NO2. However, the sensitivity of the Fong and Brune proto-
type was limited by the combination of a modest signal
rate—0.012 count/s/pptv—and a high background count rate—
approximately 1000 counts/s. The reported instrument sensi-
tivity, 1085 pptv/10 s, was only about a factor of 2.5 better than
that of George and O'Brien.
b) LIF systems with parts-per-trillion sensitivity. A UC
Berkeley-Harvard University collaboration produced a tremen-
dous advance in LIF sensitivity to NO2 through attention to
eliminating sources of laser scatter and at the same time design-
ing a detection cell with efficient collection of fluorescent
photons. The instrument was designed to detect NO2 aboard the
ER-2 aircraft in the lower stratosphere (Perkins et al., in prepara-
tion, 1999). This aircraft-borne instrument uses a Nd3+:YAG
pumped dye laser that is tuned on and off an NO2 resonance near
585 nm. Detection of the total fluorescence integrated from 750
nm to the long-wavelength cutoff of a GaAs PMT (850 nm)
occurs simultaneously with the pulsed laser. The ER-2 instrument
participated in NASA's 1997 Photochemistry of Ozone Loss in
the Arctic in Summer (POLARIS) experiment (Newman et al.,
1999). During this deployment, Perkins et al. (in preparation,
1999) demonstrated a detection limit of 10-50 pptv, limited by
systematic noise in the background, and a sensitivity of 40-80
pptv/10 s at the ambient pressure (< 200 torr). An informal
intercomparison with a photolysis-chemiluminescence (PC)
instrument developed by Del Negro et al. (1999) showed strong
correspondence between the two approaches. A linear regression
of 13075 simultaneous measurements by the two instruments
obtained on 27 separate flights over a 6-month span gives
[NO2]UF=1.07*[NO2]rc and an R2= 0.95. At pressures above 200
torr, the ER-2 instrument's sensitivity degrades due to pressure
broadening and an increase in background noise associated with
2nd Stokes Nitrogen Raman scattering. Nonetheless, the sensi-
tivity of this instrument is more than a factor of 10 improvement
over the LIF instrument designed by Fong and Brune and within
a factor of 2 of the best PC instruments.
At UC Berkeley, we improved on this approach, reducing
the detection limit to 1 pptv, enabling operation at any pressure
within the atmosphere without loss of sensitivity, and improving
the sensitivity by more than a factor of 3, to 15 pptv/10 s. The
instrument is described in detail in Thornton et al. (in press,
November 1999). To achieve these results, the instrument is
operated at low pressure in combination with a time-gated photon
counting technique. We also redesigned the detection chamber
moving the multipass optics further away from the fluorescence
collection region to reduce the probability of scattered light
reaching the detector. The combination of reduced pressure and
the time-gated photon counting technique substantially reduces
background noise (by more than 95%) associated with the laser
light scattering off of the mirrors, walls, and gas sample and
reaching the photocathode. Photon counting begins only after the
laser pulse has left the detection chamber (a delay of about 50
ns), at which point the laser-induced background noise has
decayed to zero while =85% of the NO2 fluorescence signal
remains due to the long lifetime (t = 115 ns) at 4 torr. Although
the signal rate of this experiment is low, approximately 0.1
count/s/pptv, the noise is also extremely low, about 4 counts/s.
Thus, a signal-to-noise ratio of 2 for 15 pptv NO2 in 10 s is
achieved using the Berkeley LIF instrument. Another important
improvement over the ER-2 instrument was the redesign of the
dye pump system, which extended the mean time between fail-
ures from tens of hours to several weeks.
-------
D.1 Sampling and Calibration
The Berkeley LIF instrument samples the atmosphere by
continuously pumping air through the instrument at a rate of 1.5
standard liters per minute (slpm). This results in a residence time
in the tubing and the detection chamber of approximately 1 s. A
stainless steel pinhole held in a stainless steel fitting is used as
the inlet, and the tubing used to move air from the sampling
region to the detection chamber is PFA Teflon. On the down-
stream side of the pinhole, the pressure is maintained below 100
torr by a 10-cfm rotary vane pump. The low pressure essentially
eliminates concerns that the NO+O3 reaction can cause an
increase in the NO2 concentration during the transit time from the
inlet to the detection point since the rate of this reaction drops by
almost a factor of 10 from its value at atmospheric pressure. The
calibration procedure includes a series of tests of instrument
performance, stability, and response to a known source of NO2.
1. Calibrations are performed in dry air every 3-5 h using
a cylinder of zero air and a cylinder of 5.08 ± 0.10
ppmv NO2 in N2 (Scott Specialty Gases with
ACULIFEC1 coating) for dynamic dilution over the
range 2.5-150 ppbv. Two mass-flow controllers are
used to obtain the desired concentration range. An addi-
tional solenoid valve directly following the NO2 flow
controller provides a hard zero for the calibrations and
prevents contamination of the sample flow.
2. Standard additions are performed to an atmospheric
sample every 15 h. During a standard addition, NO2 is
added to the atmospheric sample stream just prior to the
detection cell. This provides a diagnostic for the effect
environmental factors such as water vapor have on the
fluorescence signal. We observe a 3.5% effect in the
fluorescence signal due to variation of 0.01 in the mole
fraction of atmospheric water vapor.
3. The background signal in zero-grade dry air is mea-
sured by adding dry air every 3-5 h to maintain an
accurate and precise measure of the instrumental zero.
4. The sensitivity to N2-Raman scatter is measured every
20 min. In the N2-Raman detection mode, zero air is
flowed into the detection chamber at various rates to
observe the number density dependence of the Raman
signal. We then use the measured Raman sensitivity to
normalize for changes in laser alignment and optical
throughput from the signal. Extensive observations in
the lab and in the field show that the ratio of the in-
strument response to N2 Raman and to NO2 is constant
to better than ±2%.
D.2 Field Trials and Intercomparisons
Performance of the UC Berkeley LIF instrument has been
tested during three field deployments, two in the foothills of the
Sierra Nevada (25 July-31 October 1998 and 1 September-15
November 1999) and one in Nashville, TN, as part of SOS '99
(15 June-15 July 1999). The Sierra experiments were at a rural
research site located on a ponderosa pine plantation (owned and
operated by Sierra Pacific Industries) 50 miles east of Sacra-
mento, CA, at an elevation of 4000 ft. The plantation is adjacent
to the University of California Blodgett Research Forest in
Georgetown, CA. The site is characterized by a strong diurnal
cycle in atmospheric composition with abundances of NOX,
VOCs, and O3 typical of the clean continental boundary layer
observed regularly in the morning and increases in the concen-
tration of all of these species as air is transported from the
Central Valley into the foothills every afternoon. In the evening
the flow reverses and downslope flow brings in clean air again.
Figure D-2 shows observations of NO2 (30-s averages) from a
14-day period, 16-29 August 1998, that indicate the regularity of
the meteorology and of the sensitivity of the LIF technique.
Substantial improvements were made to our approach
prior to deployment from 15 June-15 July 1999 as part of SOS
'99 in Nashville, TN. These improvements are the basis for our
current estimates of sensitivity (15 pptv/10 s, S/N = 2; 1-pptv
detection limit). The Berkeley LIF NO2 measurements were part
of an extensive suite of chemical measurements made at the
Cornelia Fort Airpark 8 km northeast of downtown Nashville
http://www.al.noaa.gov/WWWHD/pubdocs/SOS/SOS99.html.
Part of our objective was an informal intercomparison of the
relatively new LIF approach with existing techniques for obser-
vation of NO2. The suite of NO2 measurements at Cornelia Fort
included NOAA Aeronomy Laboratory PC (Williams), EPA PC
(Kronmiller), EPA luminol (Wheeler), EPA DO AS (Kronmiller),
and the University of Heidelberg/UCLA DO AS (Stutz).
As of this writing (November 1999), only the PC obser-
vations by Williams are considered final and available for com-
parison to our LIF measurements. The Cornelia Fort experiment
was designed with this intercomparison in mind. Air was brought
down from the sampling tower through a 5-cm-i.d. PFA tube
using a 200-L/min blower common to both instruments, which
were located in adjacent trailers. The Berkeley instrument drew
off the main flow approximately 5 cm below the NOAA instru-
ment at a flow rate of 1.5 1/min using the pinhole sampling
system described above. After the pinhole, the air was pumped
through approximately 5 m of 6-mm-i.d. PFA tubing to the
detection cell of the LIF instrument by a 10-cfm rotary vane
pump. Both instruments had similar residence times in the
plumbing from the sample point to the instrument. Although the
pressure drop in the LIF instrument reduces the impact of gas-
phase processes, effects of surfaces in the sampling lines, if any,
should be common to both instruments.
-------
Blodgett Forest, CA, Aug. 16-29,1998
228 230 232 234 236 238 240
Figure D-2. NO2 concentration plotted as 30-s averages vs. time for the 2-week period 16-29
August 1998 at DC Blodgett Forest Research Station. Points that are far from the mean are
due to air from the site's diesel generator.
NO2 concentrations measured at Cornelia Fort Airpark by
the Berkeley LIF instrument and by the NOAA PC instrument
were compared on a point-by-point basis for the period from
17 June 1999 to 14July 1999. The time recorded by the Berkeley
instrument was synchronized to the NOAA time base by com-
paring structures in the 10-s average Berkeley and 1-min NOAA
measurements. The Berkeley measurement time base was then
shifted to match the NOAA time base and the Berkeley mea-
surements were averaged to 1 min. Figure D-3 shows obser-
vations for the two instruments from 12 p.m. on 20 June to 12
p.m. on 21 June. Figure D-4 shows an intercomparison of
observations using data collected by the Berkeley LIF instrument
and by the PC instrument operated by Williams of NOAA. The
results show remarkable agreement. For the first two-thirds of the
campaign, the PC measurements were on average 4% higher than
the LIF measurements. The zero offset between the instruments,
if any, is below 20 pptv. An R2 = 0.984 for the linear fit is an
indication of the near perfect agreement between the two obser-
vations. Nonetheless, the effect of variations in instrument
calibration on individual days is evident as stripes in the data
with different slopes at high concentration. We have not yet
identified which instrument has a varying calibration, although
efforts to show the calibration of the LIF experiment are precise
to a few percent give us confidence that all of the variations are
not due to the LIF instrument (Thornton et al., in press,
November 1999). After 6 July 1999, a substantial change in the
calibration of one of the two instruments occurred. From this
point forward to 15 July, when the experiment ended, the PC
instrument reported measurements 9% lower than the LIF
observations with a measurable offset of 97 pptv. The R2 = 0.988
again indicates the strong correspondence between observations
from the two instruments. We do yet not understand the source
of the shift observed. In this context, it is important to note that
it is only the extreme precision of the two instruments that allows
these relatively subtle variations to be observed and evaluated.
Few NO2 measurements are of the quality necessary to investi-
gate differences of less than 10%.
D.3 Future LIF Systems
Rapid progress in laser and detector technology make it
likely that LIF approaches to NO2 detection will soon be routine.
In our group at UC Berkeley we have developed a prototype LIF
instrument based on a commercial tunable external cavity diode
laser. The diode laser operates in the 638-nm range at 4 mW
average power, and red-shifted fluorescence is collected and
-------
imaged onto a cooled GaAs PMT. Currently, the prototype has
a signal-to-noise ratio of approximately 3 at 1 ppbv NO2 after
10-s averaging. Improvements in laser technology, the use of
large area single photon counting avalanche photodiode detec-
tors, higher quality optics, and incremental improvements to the
design of the optical system for collecting fluorescence are ex-
pected to increase the sensitivity of this prototype by a factor of
1000 over the next year. The instrument will weigh less than 80
Ib, will be smaller than 0.5 m3 including all components, and will
be robust enough to be operated continuously for months in
routine monitoring situations with only minor maintenance. We
estimate the cost of parts for such an LIF instrument will be on
the order of $25,000.
Thornton, J.A., Wooldridge, P.J., and Cohen, R.C. 2000.
Atmospheric NO2: In situ laser-induced fluorescence detection
at parts per trillion mixing ratios. Anal. Chem. 72(3):528-539.
Tucker, A.W., Birnbaum, M., and Fincher, C.L. 1975. Atmo-
spheric NO2 determination by 442 nm laser induced fluor-
escence. Appl. Opt. 14(6):1418-1421.
D.4 Acknowledgments
This work was supported by the Office of Science,
U.S. Department of Energy, under Contract No. DE-AC03-
76SF00098, through the UC Energy Institute, and by NASA,
through its Instrument Incubator Program under Contract No.
NAS1 -9905 3. Observations in Nashville, TN, were supported by
NOAA's Office of Global Programs under Grant No. NA96-
GP0482.
D.5 References
Del Negro, L.A., Fahey, D.W., Gao, R.S., Donnelly, S.G., Keim,
E.R., Neuman, J.A., Cohen, R.C., Perkins, K.K., Koch, L.C.,
Salawitch, R.J., Lloyd, S.A., Proffitt, M.H., Margitan, J.,
Stimpfle, R.M., Bonne, G.P., Voss, P.B., Wennberg, P.O.,
McElroy, C.T., Swartz, W.H., Kusterer, T.L., Anderson, D.E.,
Lait, L.R., and Bui, T.P. 1999. Comparison of modeled and
observed values of NO2 and JNO2 during the POLARIS mission.
J. Geophys. Res. 104(D21):26687-26703.
Fong, C., and Brune, W.H. 1997. A laser induced fluorescence
instrument for measuring tropospheric NO2. Rev. Sci. Instrum.
68(ll):4253-4262.
George, L.A., and O'Brien, R.J. 1991. Prototype PAGE
Determination of NO2. J. Atmos. Chem. 12:195-209.
Newman, P.A., Fahey, David W., Brune, William H., Kurylo,
Michael J., Kawa, S. Randolph. 1999. Preface: Photochemistry
of ozone loss in the arctic region in summer (POLARIS). Special
section, J. Geophys. Res. 104(D21):26,481-26,495.
Perkins, K.K., 2000. An airborne laser-induced fluorescence
instrument for in situ detection of stratospheric NO2, Ph.D.
thesis, Harvard University.
-------
a)
SOS 99: Cornelia Fort Airpark NO? Intercomparison
12:00 15:00 18:00 21:00 00:00 03:00 06:00 09:00 12:00
June 20-June 21,1999 (CST)
b)
Q.
C?
SOS 99: Cornelia Fort Airpark NO? Intercomparison
!:00 13:00 14:00 15:00 16:00 17:00 18:00
June 20,1999 (CST)
Figure D-3. NO2 concentration plotted as 1-min averages vs. time at Cornelia Fort Airpark,
Nashville, TN. Data measured by the Berkeley LIF instrument are shown as open circles, and
those measured by the NOAA PC instrument are shown as crosses, a) Data for the entire
1-day period from 12 p.m. (CST) on 20 June to 12 p.m. on 21 June 1999. b) Data for the
period from 12 p.m. to 6 p.m. (CST) on 20 June when the concentration of NO2 was less than
5 ppbv.
-------
a)
SOS 99 Cornelia Fort Airpark: Intercomparisen before 7/6/99
70
60
40
O 30
a.
20
10
PC-1.04*LIF
R2-0.984
N-18,119
10 20 30 40 50 60 70
Berkeley LIFNOjtppbv)
b)
SOS 99 Cornelia Fort Airpark: Intel-comparison after 7/6/99
50
40
30
O 20
C
10
PC-0.914-LIF +0.097
R2 = 0.988
N-9214
10 20 30 40
Berkeley LIF N02 (ppbv)
Figure D-4. Southern Oxidants Study, Nashville 1999, Cornelia Fort Airpark NO2 inter-
comparison plots. NO2 data measured with NOAA's PC instrument is plotted vs. that
measured with DC Berkeley's LIF instrument. A least squares best fit line is drawn through
the data, a) Data from the period 17 June to 7 July 1999 are plotted and a best fit line forced
through (0,0) has a slope of 1.04 and an R2 = 0.988. b) Data from the period 7 July to 14 July
1999 are plotted and a best fit line has a slope of 0.914 with an R2 = 0.988. The figures show
that while over the entire campaign the two instruments agreed very well on average, there
were at least two distinct populations in which the two instruments were calibrated differently
to more than 10%.
-------
Appendix E
Application of a Luminol-Based Approach to Monitoring NO2, NOX, and NOY
by
William A. McClenny, U.S. EPA, and
K. Kronmiller and M. Wheeler, ManTech Environmental Technology, Inc.
The U.S. EPA has funded the evaluation of two ap-
proaches for improved nitrogen oxides monitoring in urban
atmospheres at the Battelle laboratories in Columbus, OH. The
detector used in these two approaches is a luminol-based detector
obtained commercially as the Model LMA-3 available from
Scintrex (Toronto, Canada). Based on the results as documented
in EPA Report 600/R-95/031 (Work Assignment 43 of EPA
Contract 68-DO-0007—see Spicer et.al., 1995), a similar system
was assembled for monitoring at the Cornelia Fort monitoring
site near downtown Nashville, TN, in the 1999 SOS summer
field study in that city. The system was operated successfully at
this site and led to some of the monitoring results that are
presented in Chapter 6 of this report.
The system is shown schematically in Figures E-l and
E-2. Figure E-l shows the principal components of the system.
The thermal converter for NOY to NO conversion was placed at
the top of the monitoring tower at the Cornelia Fort monitoring
site so that compounds such as HNO3 would be immediately
converted to NO and thereby minimize losses to tubing walls.
Residence times in the sampling lines were minimized by the use
of auxiliary air pumps to pull ambient air through the relatively
long sampling lines that stretched from the top of the monitoring
tower to the trailer that housed the monitoring equipment. As
shown in Figure E-l, a second pump pulled air through the
sample conditioner and exhausted it into a sampling manifold for
the luminol-based monitor and for a supplemental measurement
by a Model 42 chemiluminescence monitor (TEI, Franklin, MA).
Routing of the sample streams inside the multipath converter is
shown in Figure E-2. By means of valves V1-V3, the sample air
was routed in three ways, each path being available for 5 min and
repeating every 15 min. With a path through VI and V3, NO2 in
the air sample was directly monitored. With a path through VI,
V2, and V3, NO in the sample stream was converted to NO2 by
the chromium trioxide oxidizer so that NOX was monitored by
the LMA-3. With V2 and V3 open, the sample passed through
the Mo thermal converter and then through the chromium
trioxide converter to provide a number of NO2 molecules equal
to the NOY.
The approach illustrated in Figures E-l and E-2 provides
NO2, NOX, and NOY, and NOZ can be obtained by difference.
Thus, four priority components (see Chapter 4) requested by the
scientists working with AQMs are obtained with the luminol-
based detector.
-------
TOWERTOP
INLET
NOTE: PATH A BY-PASSES ALL CONVERTERS AND IS THE NO2
CHANNEL.
PATH B SAMPLE FLOWS THROUGH THE CHROMIUM
TRIOXIDE OXIDIZER AND IS THE NOX CHANNEL. __,__,,__.__
ROTOMETER
PATHC SAMPLE PASSES THROUGH THE MOLY
CONVERTER AND CHROMIUM TRIOXIDE OXIDIZER AND
IS THE NOY CHANNEL.
SAMPLE MANIFOLD
NEEDLE VALVE
Approx. 18m 1/4
Teflon tubing
t
LMA-3
TEI MODEL 42
Figure E-1. LMA-3 field setup for SOS '99, Nashville, TN.
-------
Sample Line
from Mo
Converter
Sample Line
from Direct
OFF
ON
OFF
OFF OFF NO2
OFF ON NOX
ON ON NOY
Out to
Sampling
Manifold Pump
Figure E-2. LMA-3 NOY multipath converter.
------- |