o
V
*
Protocol for Applying and Validating the
CMS Model for PM2.5 and VOC
-------
-------
EPA-451/R-04-001
December 2004
Protocol for Applying and Validating the CMB Model for PM2.5 and VOC
By:
John G. Watson et al.
Desert Research Institute
University and Community College System of Nevada
Reno,NV 89512
Prepared for:
C. Thomas Coulter and Charles W. Lewis, Project Officers
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Contract No. 5D1808NAEX
US. Environmental Protection Agency
Office of Air Quality Planning & Standards
Emissions, Monitoring & Analysis Division
Air Quality Modeling Group
-------
ACKNOWLEDGMENTS
This revised protocol for applying and validating the Chemical Mass Balance Model
(CMB) was originally developed by Desert Research Institute (DRI) of trhe University and
Community College System of Nevada under Contract 5D1808NAEX with EPA's Office of Air
Quality Planning & Standards. The Project Officers were C. Thomas Coulter1 and Charles W.
Lewis.2 Substantial contributions to the initial draft of this protocol were made by DRI staff
members John G. Watson, Judith C. Chow, and Eric M. Fujita. Tom Coulter spent considerable
time reviewing and reformatting the protocol, and harmonizing it with the latest version of
CMB: EPA-CMB8.2. He also developed and produced its Appendixes A, B and G.
DISCLAIMER
This protocol was reviewed by EPA for publication. The information presented here
does not necessarily express the views or policies of the U.S. Environmental Protection Agency
or the State of Nevada. The mention of commercial hardware and software in this document
does not constitute endorsement of these products. No explicit or implied warranties are given
for the software and data sets described in this document.
'Air Quality Modeling Group, Office of Air Quality Planning & Standards; Research Triangle Park, NC 27711
National Exposure Research Laboratory, Office of Research & Development; Research Triangle Park, NC 27711
ii
-------
Table of Contents
Page
1. Introduction 1-1
1.1 Protocol Objectives 1-1
1.2 CMB Model Development and History 1-2
1.3 Protocol Overview 1-3
2. CMB Relationships with Other Air Quality Models 2-1
2.1 Conceptual Models 2-1
2.2 Emissions Models 2-1
2.3 Meteorological Models 2-2
2.4 Chemical Models 2-3
2.5 Source Dispersion Model 2-3
2.6 Receptor Models 2-4
3. CMB Source and Receptor Input Data 3-1
3.1 Source Profiles 3-1
3.1.1 Common Emissions Sources 3-1
3.1.2 Source Profile Normalization Options 3-2
3.1.3 PM2.5 Source Characteristics 3-6
3.1.4 VOC Source Characteristics 3-11
3.1.5 Source Characterization Methods 3-14
3.1.6 Source Profile Data Bases 3-16
3.2 Receptor Measurements 3-17
3.2.1 Physical and Chemical Characteristics of Receptor Concentrations 3-17
3.2.2 Receptor Characterization Methods 3-19
3.2.3 Sampler Siting 3-28
3.2.4 Temporal Variability 3-30
3.2.5 Receptor Measurement Data Bases 3-31
3.3 CMB Application Levels 3-31
4. Assumptions, Performance Measures, and Validation Procedures 4-1
4.1 Fundamental Assumptions and Potential Deviations 4-1
4.2 CMB Performance Measures 4-4
4.3 Protocol Steps 4-14
4.3.1 Determine the Applicability of CMB 4-14
4.3.2 Format Input Files and Perform Initial Model Runs 4-14
4.3.3 Evaluate Outputs and Performance Measures 4-17
4.3.4 Evaluate Deviations from Model Assumptions 4-17
4.3.5 Modify Model Inputs to Remediate Problems 4-17
4.3.6 Evaluate the Consistency and Stability of the Model Results 4-19
4.3.7 Corroborate CMB Results with Other Modeling and Analyses 4-20
in
-------
Table of Contents (continued)
Page
5. Example of Application and Validation for PM2.5 5-1
5.1 Model Applicability 5-1
5.2 Initial Source Contribution Estimates 5-6
5.3 Model Outputs and Performance Measures 5-15
5.4 Deviations from Model Assumptions 5-15
5.5 Identification and Correction of Model Input Errors 5-17
5.6 Consistency and Stability of Source Contributions 5-17
5.7 Consistency with Other Simulations and Data Analyses 5-18
6. Example of Application and Validation for VOC 6-1
6.1 Model Applicability 6-1
6.2 Initial Source Contribution Estimates 6-1
6.3 Examine Model Outputs and Performance Measures 6-4
6.4 Test Deviations from Model Assumptions 6-5
6.5 Identify and Correct Model Input Errors 6-7
6.6 Evaluate Consistency and Stability of Source Contributions 6-7
6.7 Determine Consistency with Other Simulations and Data Analyses 6-7
7. Summary, Conclusions, and Future Prospects 7-1
8. References 8-1
APPENDIX A. 54 PAMS target compounds (hydrocarbons) listed in their elution sequence A-l
APPENDIX B. Normalization for the VOC Source Profile B-l
APPENDIX C. Internet Links to Modeling Software and Data Sets C-l
APPENDIX D. CMB Mathematics D-l
APPENDIX E. Summary of CMB PMio Source Apportionment Studies E-l
APPENDIX F. Summary of CMB VOC Source Apportionment Studies F-l
APPENDIX G. Procedures for Treating Secondary Particles G-l
IV
-------
List of Tables
Page
Table 3.1-1 Chemicals from Particles in Different Emissions Sources 3-8
Table 3.1-2 Organic Compounds Found in Different Source Emissions and in Ambient Air3-9
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods for Particle and
VOC Receptor Modeling 3-20
Table 4.2-1 CMB8 Outputs and Performance Measures 4-9
Table 5.1-1 Wintertime Emissions Inventory for Denver Metro Area 5-2
Table 5.1-2 Source Composition Profiles from NFRAQS 5-3
Table 5.1-3 Source Composition Profiles from the 1987 Scenic Denver Study and Other
Studies 5-5
Table 5.2-la Sensitivity of Total Carbon Apportionment to Alternative Wood Combustion
Profiles (Welby, 01/17/97 at 0600 to 1200 MST) 5-7
Table 5.2-lb Sensitivity of Total Carbon Apportionment to Alternative Meat Cooking Profiles
(Welby, 01/17/97 at 0600 to 1200 MST) 5-8
Table 5.2-2a Sensitivity of Total Carbon Apportionment to Alternative Cold-Start Profiles
(Welby, 01/17/97 at 0600 to 1200 MST) 5-9
Table 5.2-2b Sensitivity of Total Carbon Apportionment to Alternative Hot-Stabilized and
High Particle Emitter Profiles (Welby, 01/17/97 at 0600 to 1200 MST) .... 5-10
Table 5.2-2c Sensitivity of Total Carbon Apportionment to Fitting Species (Welby, 01/17/97 at
0600 to 1200 MST) 5-11
Table 6.2-1 VOC Source Profiles for NARSTO-NE CMB 6-2
Table 6.2-2 PAMS Measured Species and CMB Fitting Species 6-3
Table 6.3-1 CMB Sensitivity Tests for Vehicle Exhaust Profiles 6-4
Table 6.3-2 CMB Sensitivity Tests for Different Profiles 6-5
-------
List of Figures
Page
Figure 3.2-1 Spatial distribution of average PM2.5 source contributions from gasoline exhaust,
diesel exhaust, suspended dust, vegetative burning, secondary ammonium sulfate,
secondary ammonium nitrate, and primary coal-fired power station fly ash in and
near Denver, CO during winter, 1996-97 3-29
Figure 3.2-2 PM2.5 source contributions at the Welby site north of Denver, CO during winter of
1996-97 3-30
Figure 4.2-1 CMB8 source contribution display 4-5
Figure 4.2-2 Eligible space collinearity display 4-6
Figure 4.2-3 Species concentration display 4-7
Figure 5.7-1 Average PM2.5 source contributions at the Welby site near Denver, CO during the
winter of 1996-97 5-18
Figure 6.7-1 Hourly average VOC source contributions by day of week at Lynn, MA 6-8
Figure 6.7-2 Wind direction dependence of VOC source contributions at Lynn, MA 6-9
VI
-------
1. INTRODUCTION
The Chemical Mass Balance (CMB) air quality model is one of several models that have
been applied to air resources management. Receptor models use the chemical and physical
characteristics of gases and particles measured at source and receptor to both identify the
presence of and to quantify source contributions to receptor concentrations. Receptor models are
generally contrasted with dispersion models that use pollutant emissions rate estimates,
meteorological transport, and chemical transformation mechanisms to estimate the contribution
of each source to receptor concentrations. The two types of models are complementary, with
each type having strengths that compensate for the weaknesses of the other.
The CMB receptor model consists of a solution to linear equations that express each
receptor chemical concentration as a linear sum of products of source profile abundances and
source contributions. The source profile abundances (i.e., the mass fraction of a chemical or
other property in the emissions from each source type) and the receptor concentrations, with
appropriate uncertainty estimates, serve as input data to CMB. In order to distinguish among
source type contributions, the measured chemical and physical characteristics must be such that
they are present in different proportions in different source emissions and changes in these
proportions between source and receptor are negligible or can be approximated. The CMB
model calculates values for the contributions from each source and the uncertainties of those
values.
The CMB model is applicable to multi-species data sets, the most common of which are
chemically characterized particulate matter (PM) and volatile organic compounds (VOC). PM2.5
and PMio (mass of particles with aerodynamic diameters less than 2.5 and 10|im, respectively)
are regulated by National Ambient Air Quality Standards (NAAQS, EPA, 1997a). VOC are not
specifically regulated, but they are precursors for ozone, which is subject to NAAQS (EPA,
1997a).
CMB model results are used to determine how much different sources contribute to
ambient concentrations. This knowledge is usually used with source attributions determined by
other models to justify emissions reduction strategies.
1.1 Protocol Objectives
This protocol describes how to use the CMB model in practical applications to determine
the contributions of different sources to PM2.5 and VOC. Its objectives are to:
• Document measurement approaches and data sources for source and receptor
input data.
• Describe the seven step applications and validation protocol to be followed when
using the CMB model for source apportionment.
• Present examples for PM2.5 and VOC apportionment using contemporary data sets
and source types.
1 - 1
-------
1.2 CMB Model Development and History
This protocol supplements and expands on the earlier protocol for applying and
validating the CMB model (EPA, 1987a; Watson et al., 1991) that was widely used to develop
State Implementation Plans for the previous PMio NAAQS (EPA 1987b, 1987c). With the
subsequent adoption of a PM2.5 NAAQS, as well as greater interest in apportioning VOC for its
role in local formation of photochemical oxidants, EPA decided to produce this enhanced
protocol, which supersedes the earlier edition.
The CMB software has evolved over more than two decades to facilitate model
application and validation. The Chemical Mass Balance receptor model was first applied by
Hidy and Friedlander (1971), Winchester and Nifong (1971), and Kneip et al. (1972). The
original applications used unique chemical species associated with each source-type, the so-
called "tracer" solution. Friedlander (1973) introduced the ordinary weighted least-squares
solution to CMB equations, and this had the advantages of relaxing the constraint of a unique
species in each source type and of providing estimates of uncertainties associated with the source
contributions. Gordon (1980, 1988) and Kowalkzyk et al. (1978) subsequently applied this
method to elemental concentrations measured in source and receptor samples. The ordinary
weighted least squares solution was limited in that only the uncertainties of the receptor
concentrations were considered; the uncertainties of the source profiles, which are typically
much higher than the uncertainties of the receptor concentrations, were neglected.
The first interactive user-oriented software for the CMB model was programmed in 1978
in FORTRAN IV on a PRIME 300 minicomputer (Watson, 1979). The PRIME 300 was limited
to 3 megabytes of storage and 64 kilobytes of random access memory. CMB versions 1 through
6 updated this original version and were subject to many of the limitations dictated by the
original computing system. CMB7 was written in a combination of the C and FORTRAN
languages for the DOS operating system. With Windows® 3.1, 95, and NT becoming the most
widely used operating systems, CMB8 created a user interface for CMB7 calculations using the
Borland Delphi object oriented language.
CMB1 was used in the Portland Aerosol Characterization Study (PACS) to develop a
State Implementation Plan for the control of Total Suspended Particulate Matter (Watson, 1979).
This modeling was the first to identify and quantify residential wood combustion as a major
contributor to particulate levels in a U.S. urban area. CMB2 was installed on EPA's UNIVAC
system in 1980 from which it could be operated by direct dial-up from a remote terminal. CMB3
streamlined the computer code in FORTRAN 77 for the EPA UNIVAC and added a ridge
regression solution to the effective variance least-squares estimation method for solving the
CMB equations (Williamson and DuBose, 1983). The ridge regression algorithm was thought to
reduce the effects of collinearity (i.e., two or more source profiles which are too similar to be
separated from each other by the model) on source contribution estimates. Henry (1982)
showed, however, that the ridge regression solution was equivalent to changing the source
profiles from their measured values until the collinearity disappeared. Henry (1982) determined
that the source contribution estimates given by the ridge regression solution did not represent
reality, and its use for air quality modeling was abandoned.
1 -2
-------
CMB4, created in 1984, ported the CMB3 software to an IBM/XT microcomputer and
added the original effective variance solution of CMB1. CMB5 was an experimental version
that contained several solution methods, performance diagnostics, and output displays. CMB5
was used as a test bed for evaluating model performance measures, and it was revised nine times
in response to recommendations and findings of these scientists and regulators. These revisions
resulted in CMB6 (EPA, 1987d) and the original protocol for applying and validating the CMB
model (EPA, 1987a). A protocol for reconciling CMB source contribution estimates with those
determined by dispersion modeling (EPA, 1987e) was also published.
While CMB7 (Watson etal, 1990a; EPA, 1990) improved the ease of use, it did not
appreciably modify the model validation performance measures. CMB8 (Watson etal., 1997a)
has major changes in the collinearity measures (Henry, 1992) that have resulted from more than
ten years of experience in using the CMB6 and CMB7 methods for model evaluation. EPA-
CMB8.2 (EPA, 2004; Coulter and Scalco, 2005) incorporates the upgrade features that CMB8
has over CMB7, but also corrects errors/problems identified with CMB8, updates the linear
algebra library, adds enhancements for a more robust and user-friendly system, and employs
code that has been reorganized, refactored, and is well-documented.
Although the protocol is applicable to earlier versions of the software (e.g., CMB7), the
examples given are specific to Windows®-based versions developed since CMB7. Throughout
this protocol, reference is made to CMB as well as to CMB8. For practical purposes, the
protocol applies ideally to either CMB8 or EPA-CMB8.2. As implied in this protocol's title, our
intention is to make this protocol "generic", applicable to any of the latest Windows® versions
that have evolved since CMB7.
1.3 Protocol Overview
The CMB modeling procedure requires: 1) identification of the contributing sources
types; 2) selection of chemical species or other properties to be included in the calculation; 3)
estimation of the fraction of each of the chemical species which is contained in each source type
(source profiles); 4) estimation of the uncertainty in both ambient concentrations and source
profiles; and 5) solution of the chemical mass balance equations. The CMB model is implicit in
all factor analysis and multiple linear regression models that intend to quantitatively estimate
source contributions (Watson, 1984). These models attempt to derive source profiles from the
covariation in space and/or time of many different samples of atmospheric constituents that
originate in different sources. These profiles are then used in a CMB calculation to quantify
source contributions to each ambient sample. Section 3 describes the types of data needed to
apply and validate the CMB model.
The CMB model is intended to complement rather than replace other data analysis and
modeling methods. The CMB model explains observations that have already been taken, but it
does not predict the future. When source contributions are proportional to emissions, as they
often are for PM and VOC, then a source-specific proportional rollback is used to estimate the
effects of emissions reductions. Similarly, when a secondary compound (a substance formed in
the atmosphere rather than directly emitted by sources) apportioned by CMB is known to be
limited by a certain precursor, a proportional rollback is used on the controlling precursor.
1 -3
-------
The most widespread use of CMB over the past decade has been to justify emissions
reduction measures in PMio non-attainment areas. More recently, CMB has been coupled with
extinction efficiency receptor models to estimate source contributions to light extinction and
with aerosol equilibrium models to estimate the effects of ammonia and oxides of nitrogen
emissions reductions on secondary nitrates. Section 2 describes how CMB relates to other air
quality models and Appendix C identifies Internet web sites where more information about these
models may be obtained.
Several solution methods have been proposed for the CMB equations: 1) single unique
species to represent each source (tracer solution) (Miller etal., 1972); 2) linear programming
solution (Hougland, 1983); 3) ordinary weighted least squares, weighting only by precision
estimates of ambient measurements (Friedlander, 1973; Guardrail and Friedlander, 1975); 4)
ridge regression weighted least squares (Williamson and Daboecia, 1983); 5) partial least
squares (Larson and Long, 1989; Long etal., 1988); 6) neural networks (Song and Hopke,
1996); 7) Britt and Luecke (1973) least squares; and 8) effective variance weighted least squares
(Watson etal., 1984). CMB8 software allows solutions 1, 3, 7, and 8 to be implemented, and
this facilitates tests of the effect of solution method on model results. Appendix D shows how
these solution methods relate to each other and documents the mathematical basis for CMB
performance measures.
The effective variance weighted least squares solution is almost universally applied
because it: 1) theoretically yields the most likely solutions to the CMB equations, providing
model assumptions are met; 2) uses all available chemical measurements, not just so-called
"tracer" species; 3) analytically estimates the uncertainty of the source contributions based on
the uncertainty of both the ambient concentrations and source profiles; and 4) gives greater
influence to chemical species with lower uncertainty estimates in both the source and receptor
measurements than to species with higher uncertainty estimates. The effective variance is a
simplification of a more mathematically exact, but less practical, generalized least squares
solution proposed by Britt and Luecke (1973).
CMB model assumptions are: 1) compositions of source emissions are constant over the
period of ambient and source sampling; 2) chemical species do not react with each other (i.e.,
they add linearly); 3) all sources with a potential for contributing to the receptor have been
identified and have had their emissions characterized; 4) the number of sources or source
categories is less than or equal to the number of species; 5) the source profiles are linearly
independent of each other; and 6) measurement uncertainties are random, uncorrelated, and
normally distributed.
The degree to which these assumptions are met in applications depends to a large extent
on the particle and gas properties measured at source and receptor. CMB model performance is
examined generically, by applying analytical and randomized testing methods, and specifically
for each application by following an applications and validation protocol. The six assumptions
are fairly restrictive and they will never be totally complied with in actual practice. Fortunately,
the CMB model can tolerate reasonable deviations from these assumptions, though these
deviations increase the stated uncertainties of the source contribution estimates. Section 4
explains these assumptions and summarizes the results of tests that evaluate deviations from
them.
1 -4
-------
The seven-step applications and validation protocol: 1) determines model applicability;
2) selects a variety of profiles to represent identified contributors; 3) evaluates model outputs
and performance measures; 4) identifies and evaluates deviations from model assumptions;
5) identifies and corrects of model input deficiencies; 6) verifies consistency and stability of
source contribution estimates; and 7) evaluates CMB results with respect to other data analysis
and source assessment methods. This protocol is illustrated for a PMzs example in Section 5 and
for a Volatile Organic Compound (VOC) example in Section 6. These examples contain
sufficient detail that the protocol can be followed for other source apportionment studies.
Appendix A lists the 54 PAMS (Photochemical Assessment Monitoring Sites) target
compounds (hydrocarbons) listed in their elution sequence. Appendix B describes the
normalization procedure for the VOC source profile. Appendix C lists Internet links (URLs) for
modeling software and data sets. Appendix D describes the CMB mathematical formulations.
Appendices E & F summarize applications of CMB to PM and VOC source apportionment.
These are related to a comprehensive bibliography of methodological and application examples
that can be consulted for greater detail. Appendix G explains procedures for treating secondary
particles within the constraints of CMB.
1 -5
-------
-------
2. CMB RELATIONSHIPS WITH OTHER Am QUALITY MODELS
Most excessive pollutant to which large populations are exposed result from various
source emissions that are transported and transformed by the atmosphere. In some cases, the
emissions emit visible plumes that can be seen to traveling toward a receptor. It is more often
the case, however, that plumes are invisible, or that many slightly visible plumes mix together
and disperse over wide areas. Different models of emissions and the atmosphere are used to
integrate science and measurements to determine the contributions from specific sources or
source types. These models are imperfect representations of reality, making many assumptions
and operating on limited data bases.
As much effort is needed to evaluate their veracity as to apply them. For this reason,
several different and independent models are commonly applied, linked to one another and
independent of each other, to quantify source. Discrepancies between model results helps to
identify and improve their weakness and to apply uncertainty bounds that should be used when
designing control strategies. Commonly used air quality models are: 1) conceptual models; 2)
emissions models; 3) meteorological models; 4) chemical models; 5) source-oriented models;
and 6) receptor models.
2.1 Conceptual Models
Conceptual models describe the relevant physical and chemical processes that affect
emissions, transport, and transformation. They are the starting point for any source
apportionment process. Conceptual models take advantage of the large body of scientific
knowledge already acquired. They identify the sources that are likely to be present and eliminate
those that are not. They examine meteorological conditions that affect concentrations and focus
further modeling on the conditions conducive to the high concentrations. Although the
conceptual models described earlier in this chapter are consistent with current information, they
are not yet verified. Field study measurements are designed to test them as hypotheses, and they
will likely change.
A conceptual model should be formulated prior to designing a CMB source
apportionment study. This should include a conception of the sources, their zones of influence,
transport from distant areas, timing of emissions throughout the day, and meteorology that
affects transport, dispersion, and transformation. This conceptual model should be used to guide
the location of monitoring sites, the time of samples, the selection of samples for laboratory
analysis, and the species that are quantified in those samples.
2.2 Emissions Models
Emissions models estimate temporal and spatial emission rates based on activity level,
emission rate per unit of activity, and meteorology (EPA, 1996). Emissions models are often
empirically derived from tests on representative source types, such as paved and unpaved roads,
motor vehicle exhaust, biota, and industries. Emissions models are used to construct emissions
inventories that are used as the basis for control strategy.
2- 1
-------
Emissions models and their results are used to identify initial sources types for inclusion
in a CMB analysis. When emissions rates are chemically speciated, the same profiles used for
that speciation might also be applicable to the CMB apportionment. The CMB is often used to
evaluate emissions models and to identify areas where they need improvement (e.g., Fujita etal.,
1994, 1995a, 1995b).
Emissions inventory models are often used to develop control strategies by linear
rollback (Earth, 1970; deNevers and Morris, 1975; Cass, 1981; Cass and McRae, 1981, 1983).
Rollback assumes that atmospheric concentrations in excess of background are proportional to
aggregate emission rates. Reducing excessive concentrations of a pollutant to levels below a
pre-set standard requires emissions reductions that are proportionally equal to the relative
amount by which the standard is exceeded.
Linear rollback does not consider the effects of meteorological transport between source
and receptor or the differences in gas-to-particle conversion for different precursor emitters. It is
most valid for spatial and temporal averages of ambient concentrations that represent the entire
airshed containing urban-scale sources. The effect of transport from distant sources located
outside the airshed is compensated by subtracting background concentrations, measured nearby
but outside the airshed, from ambient levels prior to determining needed emissions reductions.
Linear rollback also assumes for secondary particles, such as ammonium nitrate and ammonium
sulfate, that one of the precursors limits particle formation.
CMB is often used in conjunction with linear rollback to determine the contribution of
source categories to excessive concentrations. The linear rollback is then performed on a
category specific basis, starting with the largest contributors. This is often considered to be a
more accurate method of justifying emissions reductions because the relative emissions from
individual sources within a category are believed to be more accurate than the absolute emissions
within the category or the relative emissions between categories.
2.3 Meteorological Models
Meteorological models describe transport, dispersion, vertical mixing, and moisture in
time and space. Meteorological models consist of straight line, interpolation (termed
diagnostic), and first principle (termed prognostic) formulations, with increasing levels of
complexity and requirements for computational and data resources.
The straight line model is applied to hourly wind directions from a single monitor,
assuming an air mass travels a distance equal to the wind velocity in the measured direction,
regardless of the distance from the monitoring site. This model is applicable for a few hours of
transport in flat terrain, typically for evaluating a single emission source. Interpolation models
integrate wind speed and directions from multiple measurement locations, including upper air
measurements provide by remote sensors or balloon launches. The more advanced of these
models allow barriers, such as mountains, to be placed between monitors. Wind fields,
therefore, show different directions and velocities at different horizontal and vertical positions.
Interpolation wind models are applicable to domains with a large number of well-placed
monitors and for estimating the movement of air masses from many sources over transport times
of more than half a day. The number and placement of monitors, especially upper air monitors,
is especially important in mountainous terrain and in coastal areas where winds are unusual.
2-2
-------
First principle models (Stauffer and Seaman, 1994; Seaman et al., 1995; Koracin et al.,
1993, Koracin and Enger, 1994) embody scientists' best knowledge of atmospheric physics and
thermodynamics, employing basic equations for conservation and transfer of energy and
momentum. Also known as "prognostic models," first principle models purport to need no data
other than values from a sparse upper air network for interpolation. They are computationally
intensive, often requiring supercomputers but are becoming more practical and cost-effective as
workstation and desktop computers become more powerful. Modern versions use "four-
dimensional data assimilation" that compare model-calculated wind, humidity, and temperature
fields with measurements and "nudge" model outputs toward observations. A more complex
meteorological model is not necessarily a better model for a specific application. The MM5
meteorological model has been adopted as the platform for central California air quality studies
(Seaman etal., 1995).
Meteorological models are useful in conjunction with a CMB analysis to determine
where contributions might have come from. These models can often be used to determine the
relative contributions from individual sources within a source category to better focus control
strategies. These models are also useful adjunct analyses applicable to the seventh step of the
applications and validation protocol.
2.4 Chemical Models
Chemical models describe transformation of directly emitted particles and gases to
secondary particles and gases. Chemical models also estimate the equilibrium between gas and
particle phases for volatile species. Chemical models have been or are being developed for: 1)
photochemical formation of ozone, sulfate, nitrate, and organic particles in clear air (Seinfeld
and Pandis, 1998); 2) sulfate and nitrate formation in fogs and clouds (Seinfeld and Pandis,
1998); 3) inorganic aerosol equilibrium (Kim etal., 1993a, 1993b); and 4) organic aerosol
equilibrium (Pankow, 1994a, 1994b). Chemical models are reasonably well developed for ozone
and inorganic particles, but they are still under development for organic particles and gases.
Chemical models can be embedded in source-oriented dispersion models, or they can be
applied to infer source contributions or limiting precursors as a receptor model using
measurements from a monitoring site. Chemical equilibrium models, for example, are used to
determine the extent to which ammonia or nitric acid reductions will reduce secondary
ammonium nitrate concentrations estimated by CMB (Watson etal., 1994a).
Chemical models have also been used to simulate changes between source and receptor
(Friedlander, 1981; Lin and Milford, 1994; Venkatraman and Friedlander, 1994). These models
are often overly simplified, and require additional assumptions regarding chemical mechanisms,
relative transformation and deposition rates, mixing volumes, and transport times.
2.5 Source Dispersion Models
Source-oriented dispersion models use the outputs from emissions, meteorological, and
chemical models to estimate concentrations measured at receptors. They include mathematical
simulations of transport, dispersion, vertical mixing, deposition, and chemical models to
represent transformation. The most common source dispersion models are Gaussian plume, puff,
and grid formulations. Gaussian plume models (Schulze, 1990; Freeman et al., 1986; Schwede
2O
- 3
-------
and Paumier, 1997) are most often associated with the straight line wind model and estimate a
bell-shaped concentration field in the vertical and horizontal directions from the wind direction.
These models are commonly used to evaluate potential effects of primary emissions from ducted
sources, such as industrial stacks. Puff, or trajectory, models treat emissions from a variety of
sources as independent entities that are moved in a curvilinear wind field generated by a
diagnostic or prognostic wind model. Grid models place transfer pollutants between boxes with
pre-defined vertical and horizontal dimensions (Bowman etal., 1995; Byun and Dennis, 1995;
Yamartino etal., 1992). The 3-D grid-based photochemical SAQM-AERO model is the main
platform that has been developed for central California studies.
2.6 Receptor Models
Receptor models (Cooper and Watson, 1980; Watson, 1984; Javitz etal, 1988a, 1988b)
infer contributions from different primary source emissions or precursors from multivariate
measurements taken at one or more receptor sites. Receptor models are based on the same
scientific principles as source dispersion models, but they are inferential rather than predictive of
source contributions. They include CMB, factor analysis (and other forms of principal
component analysis), empirical orthogonal functions, multiple linear regression, enrichment
factors, neural networks, cluster analysis, Fourier Transform time series, and a number of other
multivariate methods. In each case these other receptor models are used to identify patterns in
chemical composition, time, or space.
Several of the model types described above can be used as either source-oriented or
receptor-oriented models. An ammonium nitrate chemical equilibrium model, for example, can
be used as a source model within the context of an air quality model. It can also be used as a
receptor model when ammonia, nitric acid, ammonium nitrate, temperature, and relative
humidity measurements are available at a receptor. Wind models have source-oriented forward
trajectory modes and receptor-oriented back-trajectory modes. Each of these formulations is
useful and of value in any source apportionment effort.
Analysis methods are often termed receptor models, but they serve as inputs to models.
Carbon-14 (14C), microscopic analysis, gas chromatograms, x-ray spectra, and many other
analytical outputs are analogous to source profiles in that they represent a pattern that might
allow a source contribution to be identified and quantified. Without the receptor model
mathematics and applications framework, however, these methods cannot provide valid
quantifiable source apportionments.
2-4
-------
3. CMB SOURCE AND RECEPTOR INPUT DATA
This section describes the types of measurements that are useful and available for both
source and receptor input data to CMB. It provides references to publications and data bases that
contain greater detail on these topics.
3.2 Source Profiles
Source profiles are the mass abundances (fraction of total mass) of a chemical species in
source emissions. Source profiles are intended to represent a category of source rather than
individual emitters. The number and meaning of these categories is limited by the degree of
similarity between the profiles. Mathematically, this similarity is termed "collinearity," which
means that two or more of the CMB equations are redundant and the set of equations cannot be
solved. Owing to measurement error, however, CMB equations are never completely collinear
in a mathematical sense. When two or more source profiles are "collinear" in a CMB solution,
standard errors on source contributions are often very high. Some source contributions may be
outlandishly high, while others may be negative. Determining the degree of collinearity is one
of the main objectives of CMB validation.
3.1.1 Common Emissions Sources
Emissions inventories need to be examined before a CMB source apportionment to
determine which source profiles will be needed and which chemical components must be
measured in local source emissions and ambient air. Emissions inventories include thousands of
individual emitters and dozens of source categories. To be useful for receptor modeling, the
categories must be grouped into more generalized categories with similar source profiles. For
example, an inventory will often contain separate entries for power generation, industrial, and
institutional coal combustion. Since these combustion processes, and often the coal, are similar
in a given airshed, it is unlikely that their contributions can be distinguished by CMB and they
must be combined into a "coal-burning category." The actual combinations depend on the
profiles available or that are likely to be acquired for a CMB study. Other categories that are
often combined for particulate and/or VOC are:
• Vegetative burning and cooking: Fireplaces, wood stoves, prescribed burns,
wildfires, char-broiling, and meat cooking. Some of these subcategories may be
separated when appropriate organic compounds are measured.
• Diesel exhaust: Heavy and light duty cars and trucks, off-road equipment,
stationary engines for pumps and generators, and locomotives.
• Gasoline exhaust: Heavy and light duty cars and trucks, and small engines.
Emissions inventories do not usually contain breakdowns by cold-starts and
visibly smoking vehicles, although these might be discriminated by certain
organic compounds in a profile. Since leaded fuels are no longer used in the U.S.,
there is no need to seek this separation.
• Gasoline evaporative emissions: Fueling stations, hot-soak vehicles.
3- 1
-------
• Fugitive dust: Paved roads, unpaved roads, agricultural tilling, construction, wind
erosion, and industrial aggregate. These can sometimes be divided into
subcategories based on single particle profiles or the measurement of specific
mineral composition.
• Solvents and coatings: Paints, degreasers, and solvents. These can also be
broken down into subcategories, not usually identified in emissions inventories,
when the specific types of solvents have been determined.
• Metals: Copper smelters, lead smelters, steel mills, and aluminum mills. These
often have similar metal emissions but in different abundances depending on the
process.
• Aggregate handling: Cement, quarrying, and mining. Ores, in particular, are
often enriched in the materials being extracted and subcategories may be defined
for these cases. When low level measurements of trace elements such as copper,
zinc, and lead are made, metal processing operations that use these materials can
be classified into separate categories.
Most emissions inventories show 80% to 90% of suspended particles originating from
suspended dust. This does not imply that other particle sources can or should be ignored.
Appendix E shows that previous PMio source apportionment studies reported substantial
contributions from other particle emitters.
VOC emissions inventories typically show stationary sources and on-road mobile sources
contributing equally to total Reactive Organic Gases (ROG) emissions in an area. The summary
of VOC source apportionment studies in Appendix F shows that source contributions from
different vehicle components typically contribute the largest, and often the vast majority, of
ambient VOC concentrations. Vehicle-related emissions, including exhaust, evaporated fuel,
and liquid fuel are ubiquitous in all urban areas. Architectural (i.e., paints) and industrial
solvents (i.e., cleaning and process solvents, as in printing) are also common to, but highly
variable in, most urban areas. Petrochemical production and oil refining are more specific to
certain urban settings, such as the Texas coast, where these activities are numerous. Biogenic
emissions are larger in the eastern U.S., where forests are lush, in contrast to the arid west. VOC
emissions in inventories are often reported in equivalent units of methane or propane.
Comparisons of relative CMB source attributions to emissions inventories requires appropriate
reconciliation between the inventory units and source contribution units.
3.1.2 Source Profile Normalization Options
Source profiles are created by sampling emissions from a variety of single emitters or
small groups of emitters. These samples are then subjected to a variety of chemical and physical
analyses to determine those properties that will allow contributions from the sources they
represent to be distinguished at receptors. Each of these properties must be normalized (scaled)
to some common property in the emissions from all sources. The two most widely used
normalization properties are total particle mass or total volatile organic compound emissions that
accompany the chemical components. The normalization procedure is one in which the
measured concentrations are expressed as ratios (fractional abundances), and is necessary to
construct source profile input files needed by CMB.
3 -2
-------
In a PM2.5 source apportionment study, the logical normalization factor is the PM2.5 mass
emission, while in a VOC source apportionment study the total VOC is the logical
normalization. One of the difficulties in combining PM2.5 and VOC source apportionment is that
there are some particle sources (e.g., suspended dust) that have negligible VOC components and
some VOC sources (e.g., solvents, evaporated gasoline, biogenics) that have negligible particle
components. However, there are many sources, such as vehicle exhaust, cooking, and wood
combustion, that have both large VOC and PM components, and profiles that are respectively
normalized to both should be considered to increase the utility of the profiles for both VOC and
PM source apportionment studies.
Individual profiles are formed from individual samples, and the uncertainty estimates of
the numerator and denominator are propagated (Watson et a/., 1995) to obtain the individual
profile uncertainties. These individual profiles are further composited to obtain the source
profiles used for CMB source apportionment. The simplest composite consists of the average
and standard deviation of abundances for all individual profiles within a group. For example, if
ten tests of diesel vehicle exhaust are taken, each abundance is an average of the ten individual
abundances and the uncertainty is the standard deviation of that average. Outlier tests are often
applied to reject individual profiles that unduly bias the standard deviation of the composite
(average) profile. In general, abundances that exceed two standard deviations calculated without
the inclusion of that abundance should be omitted from a profile. There are always some outliers
in any series of source tests, usually for reasons that can never be determined. For this reason it
is important to obtain ten or more samples that run the range of operating conditions and fuels in
an area to estimate source profiles.
Particle mass is well-defined and easy to measure, so most particle profiles for a stated
size fraction are reasonably comparable, regardless of how they were measured. This is not the
case for VOC profiles, where a wide variety of normalization factors and measurement units
have been applied. Inventories employ different conventions for defining VOC. Many
published VOC profiles are not comparable to each other, or with the ambient measurements, in
terms of their normalization.
Several terms are used inconsistently but interchangeably to describe different fractions
of atmospheric organic material. Common definitions and units must be used for ambient
concentrations, source profiles, and emissions rates. The following terms are defined as they are
used throughout this protocol, and these definitions are recommended for future CMB source
apportionment projects:
• Cx: Molecules containing x carbon atoms (e.g., C7 means the molecule contains
seven carbon atoms). This notation is useful since many sampling and analysis
techniques respond to different numbers of carbon atoms rather than to specific
compounds.
• Organic carbon: Gases and particles containing carbon and hydrogen atoms in
various ratios. Organic compounds found in ambient air may also be associated
with other elements and compounds, particularly oxygen, nitrogen, sulfur,
halogens, and metals. Various operational definitions based on measurement
method are applied to different subsets of organic compounds.
-3
-------
• Inorganic carbon: Carbon dioxide and carbon monoxide are the most abundant
inorganic gases found in the atmosphere, while amorphous graphite is the most
common particulate component. Particulate elemental carbon is operationally
defined by optical and combustion methods (Chow etal., 1993), and it contains
heavy organic material as well as inorganic carbon.
• Hydrocarbons: Organic compounds that consist only of carbon and hydrogen
atoms.
• Reactive organic gases (ROG): Organic gases with potential to react (<30 day
half-life) with the hydroxyl radical and other chemicals, resulting in ozone and
secondary organic aerosol. The most reactive chemicals are not necessarily the
largest contributors to undesirable end-products, however, as this depends on the
magnitude of their emissions as well as on their reactivity (Carter, 1990; Carter
andLurmann, 1991).
• Total organic gases (TOG): Organic gases with and without high hydroxyl
reactivity. TOG typically includes ROG plus methane and halocarbons.
• Non-methane hydrocarbons (NMHC, also termed "light" hydrocarbons): C2
through C12 (light) hydrocarbons collected in stainless steel canisters and
measured by gas chromatography with flame ionization detection (GC-FID) by
EPA method TO-14 (EPA, 1997b). NMHC excludes carbonyls, halocarbons,
carbon dioxide, and carbon monoxide even though some of these may be
quantified by the same method. NMHC is most often used to quantify ozone
precursors.
• Halocarbons: NMHC with chlorine, fluorine, or bromine compounds attached,
quantified from canisters by gas chromatography with electron capture detection
(GC-ECD) (Farwell and Rasmussen, 1976). Methyl chloride, methylchloroform,
methylbromide, and various refrigerants (Freon-12, Freon-22, SUVA) are most
commonly measured (Rasmussen etal., 1980; Khalil etal., 1985; Wang etal.,
1997). These compounds have long lifetimes and are not reactive enough to
cause major changes in tropospheric ozone and secondary organic aerosol.
Halocarbons have been implicated in the long-term depletion of stratospheric
ozone (Lovelock eta/., 1973).
• Heavy hydrocarbons: C10 through C20 hydrocarbons collected on Tenax
absorbing substrates and analyzed by thermal desorption and gas chromatography
(Pellizzari etal., 1984; Hawthorne and Miller, 1986; Walling etal., 1986;
Kamense^a/., 1988, 1989; Riba etal., 1988; Zielinska and Fujita, 1994a;
Zielinska and Fung, 1994; Zielinska et a/., 1996; Clausen and Wolkoff, 1997).
These are sometimes termed "semi-volatile" compounds because the >C15
compounds are often found as both gases and particles (Hampton etal., 1982,
1983). Most of the total hydrocarbon mass is measured in the gas phase.
• Carbonyls: Aldehydes and ketones, the most common being formaldehyde,
acetone, and acetylaldehyde (Carlier etal., 1986; Altshuller, 1993). Carbonyls
are operationally defined as Cx through C7 oxygenated compounds measured by
collection on acidified 2,4-dinitrophenylhydrazine (DNPH)-impregnated C18
3-4
-------
cartridges and analyzed by high performance liquid chromatography with UV
detection (HPLC/UV) (Cofer and Edahl, 1986; Zielinska and Fujita, 1994b;
Grosjean and Grosjean, 1996; Kleindienst etal., 1998).
• Non-methane organic gases (NMOG): NMHC plus carbonyls.
• Semi-volatile organic compounds (SVOC): Particles and gases collected on
filters backed with solid absorbent such as polyurethane foam (PUF) and XAD,
extracted in a variety of solvents, and analyzed by gas chromatography/mass
spectrometry or HPLC/UV (Greaves et al, 1985; Chuang et al, 1987). This class
includes compounds such as polycyclic aromatic hydrocarbons (PAHs),
methoxyphenols and lactones, pesticides, and other polar and non-polar organic
compounds. The heavy hydrocarbons are often classified as SVOC, but they are
given a separate identity here for precision and clarity.
• Volatile organic compounds (VOC): NMHC plus heavy hydrocarbons plus
carbonyls plus halocarbons, typically
-------
Note (however) that in many available source profiles, not all of the PAMS target
compounds will be represented. They are usually best represented in vehicle exhaust profiles.
Note also that most source profiles available in the literature (e.g., SPECIATE; see Appendix C)
will not list uncertainties. Because CMB requires that uncertainties be input for its computation,
nominal values of 5-10% may be assumed if there is no better information.
3.1.3 PMzs Source Characteristics
Table 3.1-1 identifies typical abundances of elements, ions, and carbon in different
source emissions that have been found useful for CMB. Table 3.1-2 shows several of the
organic aerosol compounds that are present in ambient aerosol and that are believed to originate
in different source emissions. Note that many of these organic compounds are semi-volatile and
may be predominantly in the gas or particle phase, depending on ambient temperature and other
factors that affect equilibrium.
In geological material, aluminum (Al), silicon (Si), potassium (K), calcium (Ca), and iron
(Fe) have large abundances with low variabilities. The total potassium (K) abundance is 15 to
30 times the abundance of soluble potassium (K+). Aluminum (Al), potassium (K), calcium
(Ca), and iron (Fe) abundances are similar among the profiles, but the silicon (Si) abundances
range from 14% in unpaved road dust to 20% in paved road dust. Lead (Pb) is sometimes
abundant in paved road dust, but it is as low as 0.004% in the other geological profiles, probably
due to deposition from previously emitted leaded-gasoline vehicle exhaust or remnants of lead
from the exhaust trains of older vehicles. Elemental carbon (EC) abundances are highly variable
in geological material, and are often negligible in natural soil samples. Organic carbon (OC) is
typically 5% to 15% in geological emitters. It is most abundant in paved road and agricultural
dusts, although the specific compounds are probably quite different for these two sources (Chow
et a/., 1994). Motor vehicle emissions (e.g., brake and tire wear, oil drips) could result in greater
abundances of Pb, EC, and OC in paved road dust. Soluble sulfate, nitrate, and ammonium
abundances are low, in the range of 0 to 0.3%. Sodium and chloride are also low, with less than
0.5% in abundance. Larger abundances of these materials may be found temporarily soon after
roadway de-icing, however.
Organic and elemental carbon are the most abundant species in motor vehicle exhaust,
accounting for over 95% of the total mass. Watson et al. (1996a) found the lead (Pb) abundance
is negligible and highly variable (0.024 ± 0.036%) in 1995 motor vehicle exhaust profiles from
northwestern Colorado. The abundance of bromine (Br) was also low, in the range of 0.01% to
0.05%. Zinc was present in most exhaust profiles, usually at levels of 0.05% or less. The
abundances of organic and total carbon can be quite variable in motor vehicle exhaust profiles.
Organic carbon abundances ranged from 36% in highway vehicle emissions to 70% in local
traffic emissions.
The ratio of organic to total carbon (OC/TC) was 0.58 in the composite vehicle profile
for northwestern Colorado. This OC/TC ratio is similar to those reported by Watson et al.
(1994b) in Phoenix, AZ, with 0.69 for gasoline-fueled vehicle exhaust, 0.55 for diesel-fueled
vehicle exhaust, and 0.52 for a mixture of vehicle types in roadside tests. Earlier measurements
in Denver, CO (Watson et al., 1990b) reported an OC/TC ratio of 0.39 for the cold transient
cycle and 0.81 for the cold stabilized cycle.
3 -6
-------
Watson et al. (1996a) also compared residential wood combustion (RWC), residential
coal combustion (RCC), and forest fire PM2.5 profiles. Average OC abundances ranged from
-50% in RWC and the forest fire profiles to -70% in the RCC profile. EC averaged 3% in forest
fire, 12% in RWC, and 26% in RCC. The OC/TC ratio was highest in the forest fire profile
(OC/TC = 0.94) and similar for the two residential combustion profiles, with 0.73 in RCC and
0.81 in RWC. Chow and Watson (1997c) measured profiles for asparagus field burning in
California's Imperial Valley with OC/TC ratios of 0.93, similar to the 0.94 ratio found in the
forest fire emissions. A similar observation was made for charbroil cooking emissions, with
60% to 70% OC abundances and high (>0.95) OC/TC ratios.
The K+/K ratios of 0.80 to 0.90 in burning profiles (Galloway etal., 1989) are in large
contrast to the low soluble to total potassium ratios found in geological material. Sulfate, nitrate,
and silicon abundances in RCC are 2 to 4 times higher than those in the RWC and forest fire
profiles. The ammonium abundance is highly variable, with an average of 1.4% in RCC and
0.1% in the RWC and forest fire profiles.
3 -7
-------
Table 3.1-1 Chemicals from Particles in Different Emissions Sources
Source Type
Paved Road Dust
Unpaved Road Dust
Construction
Agricultural Soil
Natural Soil
Lake Bed
Motor Vehicle
Vegetative Burning
Residual Oil
Combustion
Incinerator
Dominant
Particle Size
Coarse
Coarse
Coarse
Coarse
Coarse
Coarse
Fine
Fine
Fine
Fine
< 0.1%
Cr, Sr, Pb, Zr
NOf. NH4+, P, Zn, Sr, Ba
Cr, Mn, Zn, Sr, Ba
NOf. NH4+, Cr, Zn, Sr
Cr, Mn, Sr, Zn, Ba
Mn, Sr, Ba
Cr, Ni, Y
Ca, Mn, Fe, Zn, Br, Rb, Pb
K+, OC, Cl, Ti, Cr, Co, Ga,
Se
V, Mn, Cu, Ag, Sn
Chemical Abundances
0.1 to 1%
SO4=, Na+, K+, P, S, Cl, Mn, Zn,
Ba, Ti
SO4=, Na+, K+, P, S, Cl, Mn, Ba,
Ti
SO4=, K+, S, Ti,
SO4=, Na+, K+, S, Cl, Mn, Ba, Ti
Cr, NA+, EC, P, S, Cl, Ti
K+, Ti
NH4+, Si, Cl, Al, Si, P, Ca, Mn,
Fe, Zn, Br, Pb
NO3 . SO4=, NH4+, Na+, S
NH4+, Na+, Zn, Fe, Si
K+, Al, Ti, Zn, Hg
in Percent Mass
1 to 10%
Elemental Carbon (EC),
Al, K, Ca, Fe
OC, Al, K, Ca, Fe
OC, Al, K, Ca, Fe
OC, Al, K, Ca, Fe
OC, Al, Mg, K, Ca, Fe
S04=, Na+, OC, Al, S, Cl,
K, Ca, Fe
Cl- NO3 . SO4=, NH4+, S
Cl- K+, Cl, K
V, OC, EC, Ni
NOr. Na+, EC, Si, S, Ca,
> 10%
Organic Carbon
(OC), Si
Si
Si
Si
Si
Si
OC,EC
OC,EC
S, S04=
SO4=, NH+ OC, Cl
Coal-Fired Boiler
Oil-Fired Power Plant
Steel Blast Furnace
Smelter Fire
Antimony Roaster
Marine
Fine Cl, Cr, Mn, Ga, As, Se, Br,
Rb,Zr
Fine V, Ni, Se, As, Br, Ba
Fine V, Ni, Se,
Fine V, Mn, Sb, Cr, Ti
Fine V, Cl, Ni, Mn
Fine Ti, V, Ni, Sr, Zr, Pd, Ag, Sn,
and Coarse Sb, Pb
NH4+, P, K, Ti, V, Ni, Zn, Sr,
Ba,Pb
Al, Si, P, K, Zn
Al, Si, P, K, Zn
Cd, Zn, Mg, Na, Ca, K, Se
SO4=, Sb, Pb
Al, Si, K, Ca, Fe, Cu, Zn, Ba,
La
Fe, Br, La, Pb
SO4=, OC, EC, Al, S, Ca,
Fe
NH4+, OC, EC, Na, Ca,
Pb
Mn, OC, EC
Fe, Cu, As, Pb
S
NO3 . SO4=, OC, EC
Si
S, S04=
Fe
S
None reported
Cr, Na+, Na, Cl
-------
Table 3.1-2 Organic Compounds Found in Different Source Emissions
and in Ambient Air
Predominant Sources
Particle-Gas Phase
Distribution
PAH, for example
naphthalene
methylnaphthalenes
dimethylnaphthalenes
biphenyl
acenaphthylene
acenaphthene
fluorene
phenanthrene
anthracene
fluoranthene
pyrene
retene
benzo [b] naphtho [2,1 ] thiophene
benz [a] anthracene
chrysene
benzo [b+j+k] fluoranthene
benzo [ejpyrene
benzo [ajpyrene
indene[123-cd]pyrene
dibenzo[ah+ac]anthracene
benzo [ghi] pery lene
coronene
Hopanes and Sterenes
Cholestanes
Trisnorhopanes
Norhopanes
Hopanes
Guaiacols, for example
4-methylguaiacol
4-allylguaiacol
isouegenol
Acetovanillone
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Wood smoke -softwood
Motor vehicles
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles, wood smoke
Motor vehicles
Motor vehicles
Motor vehicles
Motor vehicles
Wood smoke
Wood smoke
Wood smoke
Wood smoke
Gas Phase
Gas Phase
Gas Phase
Gas Phase
Gas Phase
Gas Phase
Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Particle Phase
Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle Phase
Syringols, for example
Syringol
4-methylsyringol
Syringaldehyde
Lactons, for example
Caprolactone
Decanolactone
Undecanoic-G-Lactone
Sterols, for example
Cholesterol
Wood smoke, mostly hardwood
Wood smoke, mostly hardwood
Wood smoke, mostly hardwood
Meat cooking
Meat cooking
Meat cooking
Meat cooking
Particle-Gas Phase
Particle-Gas Phase
Particle Phase
Gas Phase
Particle-Gas Phase
Particle-Gas Phase
Particle Phase
Sitosterol
Meat cooKing, wood smoKe
^article fhase
3 -9
-------
Coal-fired power generation profiles differ substantially from residential coal burning,
even though the fuels are similar, owing to the different emissions control technologies. Sulfate
is one of the most abundant constituents in the particle phase and sulfur dioxide can be hundreds
to thousands of time higher than the particle mass. Sulfur dioxide is a good indicator of
contributions from nearby coal-fired power stations for which it has not reacted or deposited
significantly during transport to a receptor. Crustal elements such as silicon (Si), calcium (Ca),
and iron (Fe) in the coal-fired boiler profiles are present at 30% to 50% of the corresponding
levels in geological material with the exception of aluminum (Al) which is present at similar or
higher levels than those found in geological material. Other elements such as phosphorus (P),
potassium (K), titanium (Ti), chromium (Cr), manganese (Mn), strontium (Sr), zirconium (Zr),
and barium (Ba) are present at less than 1% levels.
Watson et al. (1996a) detected selenium (Se) at the level of 0.2% to 0.4% in coal-fired
power station emissions with no scrubbers or wet scrubbers, but not in emissions from a unit
with a dry limestone scrubber. Selenium is usually in the gaseous phase within hot stack
emissions, and it condenses on particles when air is cooled in the dilution chamber. Abundances
of calcium (15%), chloride (1%), and nitrate (1%) in the limestone-scrubbed unit were a few
times higher than in the other units. These differences may have resulted from the dry lime
scrubber, which added some calcium and absorbed the selenium in the vapor phase.
Sulfate, nitrate, and ammonium abundances in directly emitted particles are not sufficient
to account for the concentrations of these species measured in the atmosphere. Ambient mass
concentrations contain both primary and secondary particles. Primary particles are those which
are directly emitted by sources; these particles often undergo few changes between source and
receptor. Atmospheric concentrations of primary particles are, on average, proportional to the
quantities that are emitted. Secondary particles are those that form in the atmosphere from gases
that are directly emitted by sources.
Sulfur dioxide, ammonia, and oxides of nitrogen are the precursors for sulfuric acid,
ammonium bisulfate, ammonium sulfate, and ammonium nitrate particles (Seinfeld, 1986;
Watson et al., 1994a). Several VOCs may also change into particles; the majority of these
transformations result from intense photochemical reactions that also create high ozone levels
(Grosjean and Seinfeld, 1989). Several of these particles, notably those containing ammonium
nitrate, are volatile and transfer mass between the gas and particle phase to maintain a chemical
equilibrium (Stelson and Seinfeld, 1982a-c). This volatility has implications for ambient
concentration measurements as well as for gas and particle concentrations in the atmosphere.
Dust suspended from bare land, roadways, agricultural fields, and construction sites is
predominantly a primary pollutant, but it does play a role in secondary particle formation (Chow
and Watson, 1992; Chow etal., 1994). Some components of dust, such as ammonium nitrate
fertilizer, may volatilize into ammonia and nitric acid gases, thereby contributing to secondary
aerosol. Alkaline particles, such as calcium carbonate, may react with nitric and hydrochloric
acid gases while on the ground, in the atmosphere, or on filter samples to form coarse particle
nitrates and chlorides. Ammonium sulfate fertilizers and minerals such as gypsum (calcium
sulfate) may be mistaken for secondary sulfates when particle samples are chemically analyzed.
3- 10
-------
These examples show that although there are similarities in chemical compositions for
different sources, using source profiles from one airshed or time period may not provide a valid
CMB apportionment for ambient samples in another airshed or in another time period. Source
emissions of precursor gaseous and primary particles are highly variable due to differences in
fuel use, operating conditions, and sampling methods. Source and ambient measurements must
be paired in time to establish reasonable estimates of source/receptor relationships. Trace metals
acquired from elemental analysis of Teflon-membrane filters are only abundant in the geological
and some industrial profiles. Elemental measurements by themselves are necessary, but
insufficient, for a receptor modeling study. Chemical speciation must also include ammonium,
sulfate, nitrate, organic carbon, and elemental carbon. Simultaneous gas measurements as well
as other characteristics of suspended particles will be needed as more refined control strategies
are developed using CMB.
3.1.4 VOC Source Characteristics
The largest body of knowledge about organic gas source compositions is related to
mobile source emissions (Sampson and Springer, 1973; Black etal, 1980; Carey and Cohen,
1980; Hampton etal., 1982, 1983; Jensen and Kites, 1983; Nelson and Quigley, 1983, 1984;
Kawamuraera/., 1985; Booker et al., 1986; Sigsbyetal., 1987; Hlavinka and Bullin, 1988;
Zweidingere^a/., 1988, 1990; McClenny etal., 1989; Snow etal., 1989; Stamp etal., 1989,
1990a, 1990b, 1992, 1996; Bailey etal., 1990a, 1990b; Japar etal., 1990, 1991; Trier et al,
1990; Williams etal, 1990; Chan etal, 1991; Kaiser et al., 1991; Wallington and Japar, 1991,
1993; Chock and Winkler, 1992; Corchnoy etal, 1992; Hoekman, 1992; McCabe etal, 1992;
Siegl etal, 1992; Stedman, 1992; Bailey and Eggleston, 1993; Diehl etal, 1993; Chock etal,
1994; Haszpra and Szilagyi, 1994; Zielinska and Fung, 1994; Conner etal, 1995; Duffy and
Nelson, 1996; Pierson etal, 1996; Sagebiel etal, 1996, 1997; Sjorene^a/., 1996; Zielinska et
al, 1996; Fujita etal., 1997a, 1997b; Gelencsar etal, 1997; Gertler etal, 1997a, 1997b;
Guicherit, 1997; Simo etal, 1997). These tests include emissions from spark-ignition (gasoline-
fueled) vehicle exhaust, compression ignition (diesel-fueled) vehicle exhaust, liquid gasoline,
and evaporative gasoline emissions from fuel handling and vehicle operation.
With only the light hydrocarbons measured, the heavy-duty diesel and light-duty gasoline
exhaust profiles are similar, and are often collinear in CMB calculations. Ethene, acetylene, 1-
butene, iso-butene, propane, propene, isopentane, n-pentane, 2,2 dimethylbutane, 2-
methylpentane, n-hexane, benzene, 3-methyhexane, toluene, ethyl benzene, m- &/>-xylene, m-
ethyltoluene, and 1,2,4-trimethylbenzene, are the most abundant compounds in either or both of
these emissions. Several of these are short-lived and are only used in CMB calculations where
fresh emissions are expected, as during early morning. Major differences between diesel and
gasoline exhaust profiles are evident for acetylene, iso-butene, isopentane, n-hexane, and 2-
methylhexane, which are most abundant in gasoline exhaust and for propene, propane, 2,2
dimethylbutane, n-decane, and n-undecane which are more abundant in diesel exhaust. Gertler
et al. (1995) show that the CMB discrimination between diesel and gasoline exhaust is
distinctive when the heavy hydrocarbons are included. Most of these compounds are highly
enriched in diesel exhaust while having negligible abundances in normal-running gasoline
vehicle exhaust.
3- 11
-------
Liquid gasoline contains many compounds in common with gasoline-vehicle exhaust. It
is depleted in combustion products such as ethane, ethene, and acetylene. Evaporated gasoline is
also depleted in these combustion compounds, as well as heavier hydrocarbons that volatilize
more slowly from liquid fuels. Isobutane, -butane, t-2 butene, and especially isopentane are
enriched in evaporated gasoline. Methyl tertiary butyl ether (MTBE) stands out as a large
constituent of all gasoline-related emissions that clearly separates these from diesel in areas
where it is used as an additive. These differences are sufficient for CMB separation of gasoline
exhaust from liquid and evaporated gasoline, and often from diesel exhaust, in ambient air.
Gasoline compositions vary with location and time of year. Liquid gasoline and headspace
evaporated gasoline samples should be analyzed at times and places consistent with ambient
VOC measurements.
Petrochemical production, especially the refining of gasoline and other fuel oils (Sexton
and Westberg, 1979, 1983; Fujita etal., 1995a), can be a large contributor in areas such as
Houston, TX. Ethane, propene, propane, n-pentane, t-2 hexene, benzene, n-heptane, toluene, and
n-octane are abundant species. Most of these overlap with liquid and evaporated gasoline
vapors. Refinery VOC measurements often contain a large fraction of unidentified NMHC that
includes real, but unreported, chemical compounds that are not in the other profiles. If properly
quantified, these could probably assist CMB resolution of refinery and other petrochemical
sources.
Although solvents from paints and industrial uses are large components of all ROG
inventories, their reported profiles are few (Kitto et a/., 1997; Guo et a/., 1998). Censullo et al.
(1996) recently evaluated a large number of different solvent uses in southern California. These
profiles are depleted in the species common to fuel use and production, with larger abundances
of styrene, n-decane, and especially "other" compounds. The "other" VOCs are quantified and
differ substantially among the different coatings tested. These are sufficient to separate various
coating and solvent emissions from other contributors. California requires special solvent and
coating formulations to comply with air quality emissions requirements, so these profiles are
likely to be very specific to a particular area.
Printing ink solvents from offset (Wadden etal., 1995a, 1995b) and rotogravure are
commonly identified in emissions inventories. Most of these emissions are captured, condensed,
and re-used by modern printing facilities, especially the toluene used for thin rotogravure inks.
These may be enriched in styrene, n-nonane, and 1,2,4 trimethylbenzene, similar to the other
solvents. Again, there is a large "other" fraction of identified compounds that allow the
separation of solvent contributions to ambient VOC.
In addition to these common emissions sources, landfills are sometimes identified as
large TOG emitters owing to their prodigious production of methane (Brosseau and Heitz, 1994;
Eitzer, 1995). A variety of reactive organic gases may accompany the methane, depending of
the nature of the landfill wastes and disposal practices. Brosseau and Heitz (1994) summarize
measurements from many landfills, finding acetone, alpha terpinene, benzene, butyl alcohol,
dichlorobenzene, dichloromethane, ethylbenzene, ethyl mercaptan, limonene, furans, terpenes,
toluene, vinyl acetate, vinyl chloride, and xylene to be among the most abundant components of
ROG.
3- 12
-------
Several of these compounds, such as vinyl chloride, are not common to widespread area
sources and might be used to determine landfill source contributions by CMB. Kalman (1986)
identifies several VOCs outgassed by plastics when they are heated. Acetone was consistently
the most abundant ROG found in emissions from the surveyed landfills, probably resulting from
the anaerobic decay of discarded organic material. Similar reactions in dumpsters and trash
cans, as well as in the natural environment, may account for a portion of the unexplained acetone
observed by Fujita et al. (1994) in Los Angeles and by Singh et al. (1994) at more remote
locations. Acetone is also a product of photochemistry. Shonnard and Bell (1993) document
substantial quantities of benzene emanating from contaminated soil, a situation that will
presumably improve as modern amelioration methods are applied to these dumpsites (Fox,
1996).
Garcia et al. (1992) found small quantities of VOC emitted by several French coal-fired
power stations, with benzene, toluene, ethylbenzene, xylenes, tetrachloroethane, benzaldehyde,
and phenol being the most abundant compounds. Abundances of these compounds were
substantially enriched over their abundances in the fuel, indicating that these compounds do not
combust as well as other fuel components or that they form as part of the combustion process.
Some data have also been reported for petroleum fires (Booher and Janke, 1997), food and
beverage production (Passant et al, 1993), household products and indoor building materials
(Sack etal, 1992, Sanchez et al, 1987), ferry boats (Cooper etal, 1996), hot asphalt application
(Kitto et al, 1997), fish rendering (Ohira et al, 1976), and phytoplankton in the ocean (McKay
etal, 1996).
Biogenic VOC emissions from trees and shrubs (Tingey etal., 1978, 1981; Arnts and
Meeks, 1981; Tingey, 1981; Arnts etal., 1982; Altshuller, 1983; Hov etal., 1983; Shaw etal,
1983; Lamb etal, 1984, 1985, 1986, 1987, 1993; Oliver ef al, 1984; Roberts et al, 1985; Gay,
1987; Riba etal, 1987; Chameides etal, 1988; Juttner, 1988; Yokouchi and Ambe, 1988; Das,
1992; Hewitt and Street, 1992; Khalil and Rasmussen, 1992; Nondek et al, 1992; Winer et al,
1992; Zhang etal, 1992; Grosjean etal, 1993a, 1993b, 1993c; Guenther et al, 1993, 1994,
1996; Jobsonetal., 1994; Tanner and Zielinska, 1994, Ciccioli etal., 1995, 1997a, 1997b;
Fuentes etal, 1996; Kempf etal, 1996; Benjamin etal, 1997; Berlin etal, 1997; Cao etal,
1997; Owen etal, 1997; Pier etal, 1997; Schuhetal., 1997; Street et al., 1997; Young et al,
1997) are typically reported for isoprene and monoterpenes such as «-pinene and /?-pinene.
These compounds are very reactive and are usually detected only in forested areas. Isidorov et
al (1985) found a wide variety of heavy hydrocarbons in air dominated by different types of
plants and trees that might be more stable indicators of biogenic contributions to ambient VOC.
Variations in biogenic emissions source profiles are difficult to quantify due to the
variability in vegetation types, ambient temperature, seasonal growth cycles, and degree of
drought. Despite its high reactivity, isoprene is commonly used as marker for biogenic
emissions. Terpenes are not often quantified in ambient samples owing to measurement
difficulties. Although the effects of photochemical reactions on the source contributions can be
minimized for other major hydrocarbon sources by using fitting species with lifetimes
comparable to air mass residence times, this is not possible for a single-species biogenic profile
based upon isoprene with input data from conventional VOC measurement methods.
Fujita and Lu (1997) estimated an adjustment to biogenic contributions based on changes
in the ratios of reactive hydrocarbons (e.g., isomers of xylene) to a relatively unreactive
3- 13
-------
hydrocarbon (e.g., benzene) between morning and afternoon samples to account for the loss of
isoprene due to photochemical reactions. The average ratios of afternoon to morning
xylenes/benzene ratios reflect the net fractional loss of xylenes due to atmospheric reactions.
This fractional loss is adjusted to isoprene by applying the ratio of the OH* radical reaction rate
constants for xylenes and isoprene. Adjustment factors of 6.6 to 10.0 were derived by this
method for the biogenic contribution of ambient hydrocarbon in Phoenix, AZ (Fujita and Lu,
1997).
Biogenic contributions can be distinguished from fossil fuel contributions to ambient
VOC by the 14C isotope which is much more abundant in recently-living organisms than in
ancient coal, oil, and natural gas fuels (Conny and Currie, 1996; Klouda etal., 1996; Rasmussen
et a/., 1996; Lewis et a/., 1999). 14C is conserved with chemical transformations, thereby
enabling the participation of biogenic emissions in photochemistry to be quantified by analysis
of VOC end-products. Vegetative burning (Darley etal., 1966; Rahmdal etal., 1982; Khalil et
al, 1983; Rahmdahl, 1983;Ramdahl andMoller, 1983; Edgerton etal, 1984, 1985, 1986;
Edgerton, 1985; Isidorov etal, 1985; Hawthorne etal, 1988, 1989; Rau and Khalil, 1989; Ward
and Hardy, 1989; Hurst et a/., 1994; Koppmann et a/., 1997) has also been identified by its
contributions of methyl chloride and retene in ambient air, but the compounds in its NMHC and
NMOG emissions are poorly characterized.
3.1.5 Source Characterization Methods
Several methods have been devised to extract samples from sources which will have
chemical and physical properties similar to those found at a receptor (Gordon et a/., 1986; Chow
et a/., 1988). In each of these methods, emitted parti culate matter or gases are collected on
substrates or in containers that are subsequently analyzed for chemical content in a laboratory.
The ideal source sampling method would allow for chemical and physical
transformations of source emissions to occur prior to sample collection. Lacking this ideal, the
sampling would at least quantify the precursors of the receptor profile so that a theoretically or
empirically derived transformation could be applied. Methods used to sample source emissions
in receptor model studies include: 1) hot exhaust sampling; 2) diluted exhaust sampling; 3)
plume sampling from airborne platforms; 4) ground-based sampling of single-source dominated
air; and 5) grab sampling and resuspension.
Hot exhaust sampling is well established for determining the emission rates of criteria
pollutants, including primary particulate matter and some VOCs. Hot exhaust does not permit
the condensation of vapors into particles prior to sampling, and it sometimes interferes with the
sampling substrate or container. In vegetative burning, for example, many of the vapors do not
condense until they are near ambient temperatures. In coal-fired station emissions, the selenium
does not condense on other particles until temperatures approach ambient. Hot exhaust samples
are not often taken on substrates or in containers amenable to extensive chemical analysis.
Components of these compliance-oriented methods have been incorporated into other exhaust
sampling procedures. Although most commonly applied, hot exhaust sampling rarely yields
profiles that represent profiles as detected at receptors because it does not account for
transformations which take place when the emissions cool. Hot exhaust sampling is not
appropriate for receptor modeling studies.
3- 14
-------
Several dilution samplers have been developed to bring hot exhaust effluents to ambient
temperature by mixing with clean, cool air (Cooper et al, 1988, 1989; Heinsohn and Davis,
1980; Hildemann et al., 1989; Houck et al., 1982; Hueglin et al. ,1997; McCain and Williamson,
1984; McDonald etal, 1998; Merrill and Harris, 1987; Sousa et al, 1985; Westerholm etal,
1988; Williamson and Smith, 1979 ). Dilution samplers draw hot exhaust gases into a chamber
where they are mixed with filtered ambient air. After an aging period, the particles are drawn
through a size-selective inlet and onto substrates or into sample containers. Multiple samples for
different chemical analyses are obtained simultaneously or via sequential sampling of the same
gas stream. Stainless steel or Teflon-coated chambers are used where species might be reactive.
Recent sampling systems acquire gaseous as well as particulate samples that can be used to
apportion both particles and VOC (McDonald et al., 1998; Zielinska et al., 1998) and measure
emission rates as well as source profiles.
Diluted exhaust samplers lend themselves to laboratory simulations of emissions from
individual sources. Dynamometer simulations of motor vehicle driving with exhaust sampled
from a dilution tunnel can provide examples of aggregate emissions for a large number of
separate vehicles. Similarly, wood stoves and fireplaces can be operated under different burning
cycles with emissions sampled from a dilution tunnel.
Source sampling from airborne platforms to characterize the chemical and physical
properties of emissions has been performed from airplanes (Small et al., 1981; Richards etal.,
1981, 1985), tethered balloons (Armstrong etal, 1980; Shah etal, 1989) and helicopters.
Sampling components of appropriate weight and packaging are elevated above the emissions,
usually on the order of 100 to 500 meters, to draw samples of the effluent.
The major advantage of airborne sampling for source characterization is that source
profile fractionation might be determined if the sample can be taken at a time after emission (i.e.,
distance) sufficient to have allowed transformations to take place. The drawbacks of airborne
plume sampling are: 1) it is difficult to know when the sampler is in the plume and when it is in
ambient air; 2) it is difficult to stay in the plume long enough to obtain a sample; and 3) ambient
air mixes with the plume, so the source profile is really a combination of emissions and ambient
air.
Ground-based source sampling is identical to receptor sampling, but it is applied in
situations for which the air being sampled is known to be dominated by emissions from a given
source. The requirements of this method are: 1) meteorological conditions and sampling times
conducive to domination by a particular source; 2) samples short enough to take advantage of
those conditions; and 3) a minimum of other interfering source contributions.
Tunnels, parking garages, vehicle staging areas, and isolated but heavily traveled
roadways are often used to obtain samples for motor vehicle exhaust. Tunnels are especially
useful for this because a large number of vehicles can be evaluated with little interference from
sources other than suspended road dust (Benner et al, 1989; Bishop et al, 1996; Chang et al,
1981; Dannecker et al, 1990; Duffy and Nelson, 1996; Fraser et al, 1998; Gertler and Pierson,
1996; Gertler etal, 1997a; Gillies etal, 1998; Hering etal, 1984; Ingalls, 1989; Khalili etal,
1995; Barrefors, 1996; Lonneman etal, 1986; Miguel, 1984; Moeckli etal, 1996; Pierson and
Brachaczek, 1976, 1983; Pierson et al, 1990, 1996; Rogak etal, 1998; Staehelin etal, 1998;
Weingartner et al., 1997; Zielinska and Fung, 1994).
3- 15
-------
Using source-dominated samples, Rheingrover and Gordon (1980) and Annergarn et al.
(1992) characterized several point sources using ambient virtual impactor measurements when
the sampling was downwind of the source. Chow (1985) examined the effects of an elevated
coal-fired power plant emission on ground-based samples in a rural environment. The presence of
the plume from corresponding SO2 and wind direction measurements could be discerned, but it was
not possible to discern other chemical concentrations contributed by the power plant owing to an
overwhelming abundance of geological material in her 24-hour sample. This method may be much
better for fugitive and area sources, however, because their influence is more constant over time.
The advantages of ground-based sampling are: 1) it is representative of fractionated
(presuming transformations are complete) and composite (for area sources such as home heating,
motor vehicles, and resuspended dust) source profiles; 2) it is relatively economical; and 3) it is
compatible with other receptor samples. The disadvantages are: 1) sampling times may be too
short to obtain an adequate deposit; and 2) contributions from other source types interfere with
the source profile.
Grab sampling and resuspension in the laboratory (Chow et a/., 1994) is most often
applied to fugitive dust sources that are usually not ducted and require numerous samples to
represent a large population. Grab sampling and resuspension involves: 1) removal of a
precipitated residue of the emissions; 2) resuspension and sampling onto substrates through size-
selective inlets; and 3) analysis for the selected species. A simple sample swept, shoveled, or
vacuumed from a storage pile, transfer system, or roadbed can be taken to represent these source
types. Five to ten different samples from the same source are averaged to obtain a representative
source profile. This method is semi-established, or at least as established as the chemical and
physical analyses applied to it, because procedures are widely accepted and results are
reproducible within a method, though not necessarily among methods. The main advantages of
grab sampling and resuspension are simplicity, reliability, and low cost.
3.1.6 Source Profile Data Bases
Several compilations of particle profiles have been produced that might be applicable to a
Level 1 source assessment described in Section 3.3 (Watson, 1979; EPA, 1988; Sheffield and
Gordon, 1985; Core and Houck, 1987; Cooper etal, 1987; Houck etal, 1989; Chow and
Watson, 1994a; Watson etal., 1994a, 1996a, 1996b; Chow and Watson, 1997a, 1997b; Chow et
a/., 1997). These include chemical abundances of elements, ions, and carbon for geological
material (e.g., paved and unpaved road dust, soil dust, storage pile), motor vehicle exhaust (e.g.,
diesel-, leaded-gasoline-, and unleaded-gasoline-fueled vehicles), vegetative burning (e.g., wood
stoves, fireplaces, forest fires, prescribed burning), industrial boiler emissions, and other aerosol
sources. More modern, research-oriented profiles include specific organic compounds or
functional groups, elemental isotopes, and microscopic characteristics of single particles.
As fuels, technologies, and use patterns have changed from 1970 to the present, so have
the chemical profiles for many emissions sources. Lead has been phased out of U.S. and
Canadian fuels, but it is still used in some Mexican gasolines that might affect PM2.5 in border
areas. Catalytic converters on spark-ignition vehicles, improved compression-ignition engines
(Pierson etal., 1996), and newly-designed wood combustion appliances have substantially
reduced carbon abundances in emissions from these small but numerous sources.
3- 16
-------
Similarly, process improvements and new source performance standards have resulted in
changes in chemical component emissions from large industrial emitters. Source profiles must
be paired in time with ambient PM2.5 chemical species measurements to establish a reasonable
estimate of what is expected in ambient air.
Several compilations of VOC source profiles have also been assembled (Shah and Singh,
1988; EPA, 1988; Scheff etal, 1989a, 1989b; Shah etal, 1989; Doskey etal, 1992; Harley et
al., 1992; Fujita etal., 1997a) from original measurements and a combination of published and
unpublished test results. Most of these profiles are limited for CMB use because: 1) they
represent older technology and fuels that are different today; 2) documentation is lacking or
insufficient; 3) compound abundances are normalized to different definitions of NMOG or
NMHC and are derived from a variety of measurement units; and 4) reported VOCs are not the
same among profiles.
The most complete and available compilation of organic speciation profiles are those
associated with the example in Section 6. These are available with the CMB8 software.
3.2 Receptor Measurements
Receptor measurements need to be a subset of the source profile measurements. They
must include at least those species in the source profiles that allow sources to be separated.
3.2.1 Physical and Chemical Characteristics of Receptor Concentrations
Several characteristics of VOC and particle emissions were discussed above. Major
chemical components of PM2.5 or PMio mass in urban and non-urban areas consist of geological
material, carbon, nitrate, sulfate, ammonium, sodium chloride, and liquid water:
• Geological Material: Suspended dust consists mainly of oxides of aluminum,
silicon, calcium, titanium, iron, and other metal oxides (Chow and Watson, 1992).
The precise combination of these minerals depends on the geology of the area and
industrial processes such as steel-making, smelting, mining, and cement
production. Geological material is mostly in the coarse particle fraction, and
typically constitutes -50% of PMio while only contributing 5 to 15% of PM2.5
(Chow etal., 1992a; Watson etal., 1994b).
• Organic Carbon: Particulate organic carbon consists of hundreds, possibly
thousands, of separate compounds. The mass concentration of organic carbon can
be accurately measured, as can carbonate carbon, but only about 10% of specific
organic compounds that it contains have been measured. Vehicle exhaust (Rogge
et al., 1993a; Zielinska et al., 1998), residential and agricultural burning (Rogge
et al., 1998; Zielinska etal., 1998), meat cooking (Rogge etal., 1991; Zielinska et
al, 1998), fuel combustion (Rogge etal., 1993b, 1997), road dust (Rogge etal.,
1993c), and particle formation from heavy hydrocarbon (C8 to C20) gases (Pandis
et al., 1992) are the major sources of organic carbon in PM2.5. Because of this
lack of molecular specificity, and owing to the semi-volatile nature of many
carbon compounds, particulate "organic carbon" is operationally defined by the
sampling and analysis method (Chow etal., 1993; Hering etal., 1985).
3 - 17
-------
• Elemental Carbon: Elemental carbon is black, often called "soot." Elemental
carbon contains pure, graphitic carbon, but it also contains high molecular weight,
dark-colored, non-volatile organic materials such as tar, biogenics, and coke.
Elemental carbon usually accompanies organic carbon in combustion emissions
with diesel exhaust (Watson et a/., 1994c) being the largest contributor.
• Nitrate: Ammonium nitrate (NH4NO3) is the most abundant nitrate compound,
resulting from a reversible gas/particle equilibrium between ammonia gas (NH3),
nitric acid gas (HNO3), and particulate ammonium nitrate. Because this
equilibrium is reversible, ammonium nitrate particles can easily evaporate in the
atmosphere, or after they have been collected on a filter, owing to changes in
temperature and relative humidity (Stelson and Seinfeld, 1982a, 1982b; Allen et
a/., 1989). Sodium nitrate (NaNO3) is found in the PM2.5 and coarse fractions
near sea coasts and salt playas (e.g., Watson et a/., 1994b) where nitric acid vapor
irreversibly reacts with sea salt (NaCl).
• Sulfate: Ammonium sulfate ((NH4)2SO4), ammonium bisulfate ((NH4HSO4), and
sulfuric acid (H2SO4) are the most common forms of sulfate found in atmospheric
particles, resulting from conversion of gases to particles. These compounds are
water-soluble and reside almost exclusively in the PM2.5 size fraction. Sodium
sulfate (Na2SO4) may be found in coastal areas where sulfuric acid has been
neutralized by sodium chloride (NaCl) in sea salt. Though gypsum (Ca2SO4) and
some other geological compounds contain sulfate, these are not easily dissolved
in water for chemical analysis. They are more abundant in the coarse fraction
than in PMzs, and are usually classified in the geological fraction.
• Ammonium: Ammonium sulfate ((NH4)2SO4), ammonium bisulfate (NH4HSO4),
and ammonium nitrate (NH4NO3) are the most common compounds. The sulfate
compounds result from irreversible reactions between sulfuric acid and ammonia
gas, while the ammonium nitrate can migrate between gases and particle phases
(Watson etal., 1994a). Ammonium ions may coexist with sulfate, nitrate, and
hydrogen ions in small water droplets. While most of the sulfur dioxide and
oxides of nitrogen precursors of these compounds originate from fuel combustion
in stationary and mobile sources, most of the ammonia derives from living beings,
especially animal husbandry practiced in dairies and feedlots.
• Sodium Chloride: Salt is found in suspended particles near sea coasts, open
playas, and after de-icing materials are applied. Bulk sea water contains 57±7%
chloride, 32±4% sodium, 8±1% sulfate, 1.1±0.1% soluble potassium, and
1.2±0.2% calcium (Pytkowicz and Kester, 1971). In its raw form (e.g., deicing
sand), salt is usually in the coarse particle fraction and classified as a geological
material (Chow et a/., 1996a, 1996b). After evaporating from a suspended water
droplet (as in sea salt or when resuspended from melting snow), it is abundant in
the PM2.5 fraction. Sodium chloride is often neutralized by nitric or sulfuric acid
in urban air where it is often encountered as sodium nitrate or sodium sulfate
ra/., 1987).
Liquid Water: Soluble nitrates, sulfates, ammonium, sodium, other inorganic
3 - 18
-------
ions, and some organic material (Saxena and Hildemann, 1997) absorb water
vapor from the atmosphere, especially when relative humidity exceeds 70% (Tang
and Munkelwitz, 1993). Sulfuric acid absorbs some water at all humidities.
Particles containing these compounds grow into the droplet mode as they take on
liquid water. Some of this water is retained when particles are sampled and
weighed for mass concentration. The precise amount of water quantified in a
PM2.5 depends on its ionic composition and the equilibration relative humidity
applied prior to laboratory weighing.
3.2.2 Receptor Characterization Methods
A variety of sampling and analysis methods have been applied to acquire measurements
at source and receptor for both particles (Chow, 1995; Chow and Watson, 1994b, 1998) and
VOC (Zielinska etal., 1994, 1996). Table 3.2-1 specifies gas and particle chemical compounds
that are quantified by these methods and are being reported in source profiles.
A mnemonic is given for each chemical species that is used by CMB8 to identify the
compound. As can be seen in Table 3.2-1, most of these mnemonics bear a resemblance to the
chemical compound names. These mnemonics are reasonably straightforward for elemental
species, but they can be complex for organic species.
Several compounds can be measured by different methods, and it is a good idea to
designate these mnemonics differently. For example, the elements in Table 3.2-1 might also be
quantified by proton induced x-ray emission spectroscopy (PIXE), instrumental neutron
activation analysis (INAA), inductively couple plasma emission spectroscopy (ICP/ES) in
addition to or in place of x-ray fluorescence (XRF). The "X" in the third place of the mnemonic
could be replaced with another identifier to designate these methods. As noted above, water
soluble potassium (KPA) and total potassium (KPX) are measured by different methods, but also
represent different characteristics that distinguish among source contributions. These need to be
designated by different mnemonics.
3- 19
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
MSG
CO
HNO3
N02
SO2
NH3
CLI
N3I
S4I
N4C
KPA
TCT
OCT
ECT
NAX
MGX
ALX
SIX
PHX
SUX
CLX
KPX
CAX
TIX
VAX
CRX
MNX
FEX
COX
NIX
CUX
ZNX
GAX
ASX
SEX
BRX
RBX
SRX
YTX
ZRX
MOX
PDX
AGX
CDX
INX
SNX
SBX
BAX
Species
Mass
Carbon monoxide
Nitric Acid
Nitrogen Dioxide
Sulfur dioxide
Ammonia
Chloride
Nitrate
Sulfate
Ammonium
Soluble Potassium
Total Carbon
Organic Carbon
Elemental Carbon
Sodium
Magnesium
Aluminum
Silicon
Phosphorus
Sulfur
Chlorine
Potassium
Calcium
Titanium
Vanadium
Chromium
Manganese
Iron
Cobalt
Nickel
Copper
Zinc
Gallium
Arsenic
Selenium
Bromine
Rubidium
Strontium
Yttrium
Zirconium
Molybdenum
Palladium
Silver
Cadmium
Induium
Tin
Antimony
Barium
Method3
GRAV
NDIR
NACL/IC
TEA/AC
KOH/IC
CA/AC
Q/IC
Q/IC
Q/IC
Q/AC
Q/AA
Q/TOR
Q/TOR
Q/TOR
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
Group"
N
G
G
G
G
G
IP
IP
IP
IP
IP
OP
OP
OP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
IP
3-20
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
LAX
AUX
HGX
TLX
PBX
URX
NAPHTH
MNAPH2
MNAPH1
DMN267
DM1367
D14523
DMN12
DMN18
BIPHEN
M_2BPH
M_3BPH
M_4BPH
ATMNAP
EM_12N
BTMNAP
CTMNAP
EM_21N
ETMNAP
FTMNAP
GTMNAP
HTMNAP
TM128N
ACNAPY
ACNAPE
PHENAN
FLUORE
A_MFLU
M_1FLU
B_MFLU
C_MFLU
A_MPHT
M_2PHT
B_MPHT
C_MPHT
M_1PHT
DM36PH
A_DMPH
B_DMPH
C_DMPH
DM1 7PH
D_DMPH
E_DMPH
Species
Lanthanum
Gold
Mercury
Thallium
Lead
Uranium
Naphthalene
2-menaphthalene
1 -menaphthalene
2,6+2,7-dimenaphthalene
1 ,7+ 1,3+1 ,6-dimenaphthalene
2,3+1 ,4+ 1 ,5-dimenaphthalene
1 ,2-dimenaphthalene
1 ,8-dimenapthalene
Biphenyl
2-Methylbiphenyl
3 -Methy Ibipheny 1
4-Methylbiphenyl
A-Trimethylnaphthalene
1 -Ethyl-2-methy hiaphthalene
B-Trimethylnaphthalene
C-Trimethylnaphthalene
2-Ethyl- 1 -methy hiaphthalene
E-Trimethylnaphthalene
F-Trimethy hiaphthalene
G-Trimethylnaphthalene
H-Trimethylnaphthalene
1 ,2,8-Trimethylnaphthalene
Acenaphthylene
Acenaphthene
Phenanthrene
Fluorene
A-Methylfluorene
1 -Methy Ifluorene
B-Methylfluorene
C-Methylfluorene
A-Methylphenanthrene
2-Methylphenanthrene
B-Methylphenanthrene
C-Methylphenanthrene
1 -Methy Iphenanthrene
3 ,6-Dimethy Iphenanthrene
A-Dimethy Iphenanthrene
B-Dimethy Iphenanthrene
C-Dimethy Iphenanthrene
1 ,7-Dimethy Iphenanthrene
D-Dimethy Iphenanthrene
E-Dimethy Iphenanthrene
Method3
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
T/XRF
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
Group"
IP
IP
IP
IP
IP
IP
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OG
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
3-21
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
ANTHRA
M_9ANT
FLUORA
PYRENE
A_MPYR
B_MPYR
C_MPYR
D_MPYR
E_MPYR
F_MPYR
RETENE
BNTIOP
BAANTH
M_7BAA
CHRYSN
BBJKFL
BEPYRN
BAPYRN
M_7BPY
INCDPY
DBANTH
BBCHRN
BGHIPE
CORONE
GCAPLA
GUACOL
M4GUCL
E4GUCL
SYRGOL
PPGUCL
A4GUCL
GNONLA
F4GUCL
M4SYRG
E4SYRG
ISOEUG
GDECLA
ACETVA
UNGLAC
SYRALD
ERGOS
SITOS
C27SDS
C27RDS
C27RCH
C27SBC
C27RAC
AABTNH
Species
Anthracene
9-Methylanthracene
Fluoranthene
Pyrene
A-Methylpyrene
B-Methylpyrene
C-Methylpyrene
D-Methylpyrene
E-Methylpyrene
F-Methylpyrene
Retene
Benzonaphthothiophene
Benz(a)anthracene
7-Methylbenz[a]anthracene
Chrysene
Benzo(b+j+k)FL
BeP
BaP
7-Methylbenzo[a]pyrene
Indeno[123-cd]Pyrene
Dibenz(ah+ac)anthracene
Benzo(b)chrysene
Benzo(ghi)Perylene
Coronene
A-Caprolactone
Guaiacol
4-Methylguaiacol
4-Ethylguaiacol
Syringol
Propylguaiacol
4-Allylguaiacol
G-Nonanoic Lactone
4-Formylguaiacol
4-Methylsyringol
4-Ethylsyringol
Isoeugenol
G-Decanolactone
Acetovanillone
Undecanoic-G-Lactone
Syringaldehyde
Ergostane
Sitostane
Diasterane-1
Diasterane-2
Cholestane-1
Cholestane-2
Cholestane-3
Trisnorhopane-1
Method3
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
Group"
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
3-22
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
AB_TNH
AB30NH
CHLSRL
BA30NH
AB_HOP
STEROW
BA_HOP
SABHHP
RABHHP
SITOST
BB_HOP
STEROM
SABBHH
RABBHH
IDNMHC
UNID
METHAN
ACETYL
CO_PPM
ETHENE
MEACRO
ETHANE
METOH
FORMAL
PROPE
C02PPM
ACETAL
N_PROP
ETHOH
BUDI13
BUTYN
ACETO
ACROLN
BEABYL
C2BUTE
LBUT1E
LIBUTE
T2BUTE
PROAL
I_BUTA
N_BUTA
CPENTE
I_PREN
CROTON
B1E2M
B1E3ME
B2E2M
CPENTA
Species
Trisnorhopane-2
Norhopane-1
Cholesterol
Norhopane-2
Hopane-1
Steroid-w
Hopane-2
Homohopane-1
Homohopane-2
Sitosterol
Hopane-3
Steroid-m
Bishomohopane- 1
Bishomohopane-2
Total Identified NMHC
Unidentified1
methane
acetylene
carbon monoxide
ethene
methacrolein
ethane
methanol
formaldehyde
propene
carbon dioxide
acetaldehyde
propane
ethanol
1,3-butadiene
l&2-butyne
acetone
acrolein
1 -butene&i-butene
c-2-butene
1-butene
iso-butene
t-2-butene
propionaldehy de
isobutane
-butane
cyclopentene
isoprene
crotonaldehyde
2-methy 1- 1 -butene
3-methyl- 1 -butene
2-methy 1-2-butene
cyclopentane
Method3
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
GC/MS
F
F
F
F
F
F
F,D
F
F
D
F
F
F
F
F
F
F
F,D
D
F
F
F
F
F
D
F
F
F
F
D
F
F
F
F
Group"
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
OP
N
p
Y
0
AL
P
OH
AL
O
AL
P
OH
0
Y
K
AL
0
O
0
O
0
AL
P
P
0
O
AL
O
0
O
p
3-23
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
PENTE1
C2PENE
T2PENE
PRAL2M
BUAL
BUONE
IPENTA
N_PENT
BENZE
CPENE1
CYHEXE
C2HEXE
C3HEXE
C6OLE1
CYHEXA
HEX1E
MCYPNA
P1E2ME
P1E3ME
P1E4ME
P2E2ME
P2E3MC
P2E3ME
P2E3MT
T2HEXE
T3HEXE
MECL2
VALAL
BU22DM
BU23DM
N_HEX
PENA2M
PENA3M
MTBE
TOLUE
PHENOL
MEBR
C12DCE
T12DCE
VINECL
C70LE1
CPA13M
MECYHX
T3HEPE
ETDC12
HEXAL
BU223M
Species
1-pentene
c-2-pentene
t-2-pentene
2-methylpropanal
butanal
butanone
isopentane
n-pentane
benzene
1 -methylcy clopentene
cyclohexene
c-2-hexene
c-3-hexene
C6 olefin
cyclohexane
1-hexene
methylcyclopentane
2-methy 1- 1 -pentene
3-methyl- 1 -pentene
4-methy 1- 1 -pentene
2-methyl-2-pentene
cw-3-methy 1-2 -pentene
3-methyl-2-pentene
trans-3-methyl-2-pentene
t-2-hexene
t-3-hexene
methylene chloride
valeraldehyde
2,2-dimethylbutane
2,3-dimethylbutane
n-hexane
2-methy Ipentane
3-methylpentane
methyl tertiary butyl ether (ppbv)
toluene
phenol
methylbromide
cis- 1 ,2,-dichloroethylene
trans- 1 ,2-dichloroethy lene
viny lidenechloride
C7 olefin
1 ,3-dimethy Icyclopentane
methylcyclohexane
t-3-heptene
1 ,2-dichloroethane
hexanal
2,2,3-trimethylbutane
Method3
F
F
F
F
F,D
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
E
D
F
F
F
F
F
F
F,T
T
E
E
E
E
F
F
F
F
E
F,D
F
Group"
0
O
0
AL
AL
K
P
P
A
O
0
O
0
O
P
O
P
O
0
O
0
O
0
O
0
O
X
AL
P
P
P
P
P
E
A
AL
X
X
X
X
0
A
P
O
X
AL
A
3-24
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
HEXA2M
HEXA3M
HEXE4M
N_HEPT
PEN22M
PEN23M
PEN24M
PEN33M
PA3ET
STYR
HEPAL
BENZAL
ETBZ
MP_XYL
0_XYL
CHX11M
OCT1E
P1E244
N_OCT
HEP2ME
HEP3ME
HEX24M
HEX25M
HX23DM
PA224M
PA234M
INDENE
INDAN
CCL3
ACPHONE
TOLUAL
BZ123M
BZ124M
BZ135M
IPRBZ
M_ETOL
MEOCT
N_PRBZ
0_ETOL
P_ETOL
F12
IPCYHX
NONE1
OCTAL
NAPHTH
HEP24D
HEP25D
Species
2-methylhexane
3-methylhexane
4-methylhexene
n-heptane
2,2-dimethylpentane
2,3-dimethylpentane
2,4-dimethylpentane
3,3-dimethylpentane
3-ethylpentane
styrene
heptanal
benzaldehyde
ethylbenzene
m - & />-xylene
o-xylene
1 , 1 -dimethy Icy clohexane
octene-1
2,4,4-trimethyl- 1 -pentene
n-octane
2-methylheptane
3-methylheptane
2,4-diemthylhexane
2,5-diemthylhexane
2,3-dimethylhexane
2,2,4-trimethylpentane
2,3,4-trimethylpentane
indene
indan
chlorofonn
acetophenone
tolualdehyde
1 ,2,3-trimethylbenzene
1 ,2,4-trimethylbenzene
1 ,3,5-trimethylbenzene
isopropylbenzene
m-ethyltoluene
methyloctane
n-propylbenzene
o-ethyltoluene
/>-ethyltoluene
Freon 12
isopropy Icy clohexane
1-nonene
octanal
naphthalene
2,4-dimethylheptane
2,5-dimethylheptane
Method3
F
F
F
F
F
F
F
F
F
F,T
F
F,D,T
F,T
F,T
F,T
F
F
F
F,T
F
F
F
F
F
F
F
F,T
F,T
E
T
D
F,T
F,T
F,T
F,T
F,T
T
F,T
F,T
F,T
E
F
T
F
F
F
F
Group"
P
P
P
P
P
P
P
P
P
A
A
AL
A
A
A
P
0
O
P
P
P
P
P
P
P
P
A
A
X
K
AL
A
A
A
A
A
P
A
A
A
X
P
0
AL
A
P
P
3-25
-------
Table 3.2-1 Chemical Compounds, Mnemonics, and Measurement Methods
for Particle and VOC Receptor Modeling
Mnemonic
HEP26D
HEPS 3D
HEP44D
HEP4ME
HEX225
HEX235
N_NON
OCT2ME
OCT3ME
TCENE
IND_1M
IND_2M
TCE112
MECCL3
BZ1234
BZ1235
BZ1245
BZDME
DETBZ1
DETBZ2
DETBZ3
DMETBZ
I_BUBZ
IPRTOL
N_BUBZ
S_BUBZ
A_PINE
B_PINE
LIMONE
Fll
NONAL
NAPJM
NAP_2M
DMOCT
N_DEC
OCT26D
OCT36M
INDDM1
MDCBZ
ODCBZ
PDCBZ
DETMBZ
ACNAPY
CCL4
ACENPE
DMN12
DMN13
Species
2,6-dimethylheptane
3,3-dimethylheptane
4,4-dimethylheptane
4-methylheptane
2 ,2 , 5 -trimethy Ihexane
2,3,5 -trimethy Ihexane
n-nonane
2-methyloctane
3-methyloctane
trichloroethylene
1-methylindan
2-methylindan
1 , 1 ,2-trichloroethane
methyl chlorofonn
1 ,2,3,4-tetramethylbenzene
1 ,2,3,5-tetramethylbenzene
1 ,2,4,5-tetramethy Ibenzene
1 ,3-dimethyl-4-ethy Ibenzene
m-diethy Ibenzene
/>-diethylbenzene
o-diethy Ibenzene
dimethylethy Ibenzene
isobuty Ibenzene
isopropyltoluene
n-buty Ibenzene
sec-butylbenzene
alpha-pinene
beta-pinene
limonene
Freon 1 1
nonanal
1 -methy Inaphthalene
2-methylnaphthalene
dimethyloctane
n-decane
2,6-dimethyloctane
3 ,6-dimethy loctane
dimethylindan
m-dichlorobenzene
o-dichlorobenzene
para-dichlorobenzene
diethyhnethy Ibenzene
acenaphthylene
carbon tetrachloride
acenaphthene
1 ,2-dimethy Inaphthalene
1 ,3-dimethy hiaphthalene
Method3
F
F
F
F
F
F
F,T
F
F
E
F
F
E
E
T
T
F
F
F
F
F
T
F
F
F
F
F
F
T
E
F
T
T
T
F
F
F
T
E
E
T
T
T
E
T
T
T
Group"
P
P
P
P
P
P
P
P
P
X
A
A
X
X
A
A
A
A
A
A
A
A
A
A
A
A
O
0
O
X
AL
A
A
P
P
P
P
A
X
X
X
A
A
X
A
A
A
3-26
-------
Mnemonic
DMN14
DMN15
DMN18
DMN23
DMN26
DMN27
NAP1ET
NAP2ET
N_UNDE
PERC
N_DODE
DBRME
PHENA
N_TRID
F113
F114
ETDB12
N_TETD
CLDBRM
N_PEND
N_HEXD
N_HEPD
N_OCTD
N_NOND
N_EICO
N_HENE
Species
1 ,4-dimethy Inaphthalene
1 , 5 -dimethy Inaphthalene
1 ,8-dimethy hiaphthalene
2 , 3 -dimethy hiaphthalene
2,6-dimethy hiaphthalene
2,7-dimethy hiaphthalene
1 -ethy hiaphthalene
2-ethylnaphthalene
n-undecane
perchloroethy lene
n-dodecane
1 ,3-dibromomethane
phenanthrene
n-tridecane
FreonllS
Freon 114
1 ,2-dibromoethane
n-tetradecane
chlorodibromomethane
n-pentadecane
n-hexadecane
n-heptadecane
n-octadecane
n-nonadecane
n-eicosane
n-heneicosane
Method3
T
T
T
T
T
T
T
T
T,F
E
F
E
T
T
E
E
E
T
E
T
T
T
T
T
T
T
Group"
A
A
A
A
A
A
A
A
P
X
P
X
A
P
X
X
X
P
X
P
P
P
P
P
P
P
AC=Automated colorimetry
CA/AC=Citric acid filter and automated
colorimetry
D=DNPH with HPLC/UV
E=Canister with GC/ECD
F=Canister with GC/FID
GC/MS=Gas chromatography mass
spectrometry
GRAV=Gravimetric,
IC=Ion chromatography
KOH/IC=Potassium hydroxide filter & ion
chromatography
NACL/IC=Sodium chloride filter & ion
chromatography
NDIR=Non-Dispersive Infrared
T=Tenax with GC/FID,
TEA/IC=Triethanolamine filter & automated
colorimetry
XRF= X-ray fluorescence
Group codes:
A = aromatic VOC
AL = aldehyde VOC
E = ether VOC
IG=inorganic gas
IP=inorganic particle
K = ketone VOC
O = alkene (olefin) VOC
OG=organic gas
OH = alcoholVOC
OP=organic particle
P = parafm VOC
X = haogenatedVOC
Y = alkyne VOC
1 Sum of unidentified hydrocarbons. Excludes halogenated and oxygenated compounds.
-------
3.2.3 Sampler Siting
The chemical dimension can be supplemented by spatial separation of receptors to further
define the source categories or the specific emitters represented by different source types. These
sites are classified as background, transport, gradient, and source sites that are intended to
measure the following (Watson etal., 1997b):
• Community Representative (CORE): CORE sites are intended to represent
concentrations of large populations that live, work, and play within 5 to 10 km
surrounding the site. These sites are most affected by regional and urban scale
contributions with relatively small neighborhood scale and smaller contributions.
• Background: Background sites intend to measure concentrations that are not
influenced by emissions from the regulated study area. These are located in
pristine areas, away from local or urban sources. Few background locations are
completely devoid of anthropogenic emissions.
• Interbasin transport: These sites are intended to evaluate concentrations along
established or potential transport pathways. In mountainous terrain, these are
typically located at the mountain passes through which inflows and outflows have
been documented. In flat terrain they are located between urban areas or
industrial source areas and urban areas.
• Intrabasin gradient: These sites are located in large regional areas, such as the
Great Lakes region, the northeast corridor, the Los Angeles area, and within
California's San Joaquin Valley where urban complexes are in non-urban areas
between core sites. They are intended to evaluate the extent to which one
urbanized area in an airshed affects concentrations in another urban area, as well
as the extent to which urban contributions arrive at non-urban locations within an
airshed.
• Source: Source sites are located right next to, and downwind of, representative
and identifiable emitters. Where practical, these are located within 1 km of
gradient or core sites to further evaluate the zone of influence of these source
emissions.
Figure 3.2-1 shows how sampler siting within and between urban areas can assist in
determining which components are regional and which are nearby contributors. In this example
it is apparent that most of the primary contributions from carbon and geological material are
from urban and neighborhood sources, while secondary nitrate and sulfate are contributed from
outside the urban area. This would not be discernible from a single sampling location in the city
center. The source contributions in Figure 3.2-1 were determined by CMB applied to elemental,
ionic, and carbon measurements without use of specific organic compounds.
3-28
-------
Masters
LD GV (f)
Diesel (b)
Road dust/geological (c)
Meat & wood
Ammonium Sulfate
Ammonium Nitrate
Coal power station
Unexplained
ID
10
20 30 Kilometers
Figure 3.2-1. Spatial distribution of average PM2.5 source contributions from gasoline exhaust
(LDGV), diesel exhaust (diesel), suspended dust (road dust/geological), vegetative burning
(meat & wood), secondary ammonium sulfate, secondary ammonium nitrate, and primary coal-
fired power station fly ash in and near Denver, CO during winter, 1996-97 (Watson et a/., 1998).
3-29
-------
PM2.5 - Welby
nLDGV cold start
Q meat cooking
D ammonium sulfate
• LDGV hot stabilized
• RWC (softwood)
nammonium nitrate
nLDGV high emitter
nRWC (hardwood)
ncoal power station
idiesel exhaust
Hroad dust/geological
nUnexplained
Figure 3.2-2. PM2.5 source contributions at the Welby site north of Denver, CO, during winter
of 1996-97. Organic compounds were used in these apportionments, with resulting addition of
source categories for gasoline exhaust for cold starts (LDGV cold start), normal running (LDGV
hot stabilized), and poorly maintained (LDGV high emitter) vehicles. Vegetative burning is
separated into meat cooking and residential wood combustion (RWC) for softwood and
hardwood. Samples were taken from 0600-1200, 1200-1800, and 1800-0600 MST, with the
morning sample directly over the date.
3.2.4 Temporal Variability
Temporal variability in concentrations is important because it helps to confirm source
contributions by bracketing their emissions in time. Seasonal variations often allow vegetative
burning contributions to be attributed to prescribed burning and wildfires during summer, when
residential burning is at a minimum, and to woodstoves and fireplaces that are used during cool
weather.
Figure 3.2-2 shows the temporal variation of source contributions at a site near Denver,
CO. Motor vehicle exhaust contributions are typically largest during morning samples, and
residential wood combustion is abundant in nighttime samples, especially near New Years Day.
3-30
-------
3.2.5 Receptor Measurement Data Bases
Ambient chemical concentrations are not commonly available for CMB source
apportionment. Special studies have been conducted to acquire the needed data at representative
receptors during periods where PM or VOC concentrations have been found excessive.
Appendices E and F identify many of these studies that have adequate data bases. Lioy et al.
(1980); Chow and Watson (1989); and Watson and Chow (1992) summarize other chemically
speciated data sets for suspended particles.
The most complete chemical data base to which CMB can be applied is the Interagency
Monitoring of Protected Visual Environments (IMPROVE) network that has acquire elemental,
ionic, and carbon measurements at National Parks and Wilderness areas since 1987. The most
comprehensive VOC data base derives from the Photochemical Assessment Monitoring Sites
(PAMS) that takes canister or continuous gas chromatographic measurements at urban and
suburban sites during the summer.
New networks in support of the PM2.5 NAAQS will acquire speciated measurements at
several hundred sites throughout the United States. One of the specific purposes of these
measurements is to obtain source contributions via CMB modeling. Many of them will be
collocated with PAMS sites, thereby offering the opportunity to use VOC and PMzs chemical
components together in the source apportionment. Appendix C provides Internet links to data
bases containing measurements useful for source apportionment studies.
3.3 CMB Application Levels
There is no single sampling and analysis design that will permit successful CMB source
apportionment in every urban area. Since measurements can be costly, it is useful to examine
existing samples and existing data to assist in forming a conceptual model prior to designing a
full-scale source apportionment study. Three sequential levels of complexity (EPA, 1984) can
be applied, with each level being more costly, but supplying more accurate and precise
information than the previous level. The levels are useful as a shorthand notation of the general
level of comprehensiveness of a CMB study but have no regulatory significance. A given level
may not provide valid results because of data limitations. In such cases, the next higher level
may need to be undertaken to complete the CMB analysis.
The basic level of CMB application (Level I) uses existing data or data that can be
readily obtained from analyses of existing samples (Gordon et al., 1984). Source profiles that
were measured elsewhere, but that can be related to local sources, are also used. This effort
confirms the selection of contributing sources from the preliminary analysis and eliminates
minor contributors from further scrutiny. If the sources contributing to the high concentrations
of PMio are apparent and sufficiently certain, no further work will be needed. Otherwise, this
effort serves to reduce the areas to be studied in greater detail under an intermediate (Level II)
analysis.
The intermediate (or Level II) analysis involves additional chemical analyses on existing
samples or the acquisition of additional samples from existing sampling sites. It is intended to
fill the gaps in model input data which may have been discovered in Level I so as to reduce
uncertainty in results of the Level I source apportionment. A comprehensive CMB analysis
(Level II) involves the acquisition of new data from new source and ambient sampling activities.
3-31
-------
Local dust samples are obtained and analyzed, at a minimum. Ground-based vehicle exhaust and
vegetative burning profiles are also often acquired. Industrial source profiles are usually adapted
from other studies. Light hydrocarbons are measured for a VOC apportionment study and
elements, ions, and carbon are quantified for PM2.5 or PMio. Where new sampling is possible,
sampling locations and times are selected to bracket suspected contributors.
A Level III analysis is only applied in the most complex airsheds where the costs of
emissions reduction are high and their effectiveness is uncertain. A Level III study involves
original source testing and measurements beyond the basic particulate or VOC species. Heavy
hydrocarbons and organic particles are measured at source and receptor. A Level III study
usually involves a complex and detailed application of all model types specified in Section 2.
The CMB applications and validation protocol described here is appropriate to all three
levels of PM and VOC assessment. It provides estimates of precision and validity that serve to
define the measurement requirements for the next level of analysis. These estimates can also be
used to determine whether or not the model results at a given level of PM and VOC assessment
are certain enough to eliminate the need for more extensive assessment.
3-32
-------
4. ASSUMPTIONS, PERFORMANCE MEASURES, AND VALIDATION PROCEDURES
4.1 Fundamental Assumptions and Potential Deviations
The CMB model assumptions are:
1. Compositions of source emissions are constant over the period of ambient and source
sampling.
2. Chemical species do not react with each other, i.e., they add linearly.
3. All sources with a potential for significantly contributing to the receptor have been
identified and have had their emissions characterized.
4. The source compositions are linearly independent of each other.
5. The number of sources or source categories is less than or equal to the number of
chemical species.
6. Measurement uncertainties are random, uncorrelated, and normally distributed.
Assumptions 1 through 6 are fairly restrictive and will never be totally complied with in
actual practice. Fortunately, CMB can tolerate deviations from these assumptions, though these
deviations increase the stated uncertainties of the source contribution estimates.
The CMB model has been subjected to a number of tests to determine its abilities to
tolerate deviations from model assumptions (Watson, 1979; Gordon etal., 1981; deCesar and
Cooper, 1982; Henry, 1982, 1992; Currie etal, 1984; Dzubay etal, 1984; deCesar etal, 1985a,
1985b, 1986; Javitz and Watson, 1986; Lowenthal and Rahn, 1988a, 1088b, Lowenthal etal,
1987, 1988, 1992, 1994; Javitz et a/., 1988a, 1988b; Cheng and Hopke, 1989; Kim and Henry,
1989; Henry and Kim, 1990; White and Macias, 1991). These studies all point to the same basic
conclusions regarding deviations from the above-stated assumptions.
With regard to Assumption 1, source compositions, as seen at the receptor, are known to
vary substantially among sources, and even within a single source over an extended period of
time. These variations are both systematic and random and are caused by three phenomena: 1)
transformation and deposition between the emissions point and the receptor; 2) differences in
fuel type and operating processes between similar sources or the same source in time; and 3)
uncertainties or differences between the source profile measurement methods. Evaluation
studies have generally compared CMB results from several tests using randomly perturbed input
data and from substitutions of different source profiles for the same source type. These tests
consistently demonstrate that the error in the estimated source contributions due to biases in all
of the elements of a source profile is in direct proportion to the magnitude of the biases. For
random errors, the magnitude of the source contribution errors decreases as the difference
between the number of species and sources increases.
Javitz etal. (1988b), for example, examined a simple 4-source urban airshed and a
complex 10-source urban airshed using randomly perturbed source profiles and receptor
concentrations with known source contributions. These tests with 17 commonly measured
chemical species showed that primary mobile, geological, coal-fired power plant, and vegetative
burning source types can be apportioned with uncertainties of approximately 30% when
coefficients of variation in the source profiles are as high as 50%. This performance was
demonstrated even without the presence of unique "tracer" species such as selenium for coal-
fired power plants or soluble potassium for vegetative burning.
4- 1
-------
In a complex urban airshed, which added residual oil combustion, marine aerosol, steel
production, lead smelting, municipal incineration, and a continental background aerosol, it was
found that the geological, coal-fired power plant, and background source profiles were collinear
with the measured species. At coefficients of variation in the source profiles as low as 25%,
average absolute errors were on the order of 60%, 50%, and 130% for the geological, coal-
burning, and background sources, respectively. All other sources were apportioned with average
absolute errors of approximately 30% even when coefficients of variation in the source profiles
reached 50%. These tests were performed with commonly measured chemical species, and
results would improve with a greater number of species that are uniquely emitted by the different
source types.
With regard to the nonlinear summation of species, Assumption 2, it is necessary to
measure source profiles, or modify them by some objective method, to account for changes in
the character between source and receptor. The conversion of gases to particles and reactions
between particles are not inherently linear processes. This assumption is especially applicable to
the end products of photochemical reactions and their apportionment to the sources of the
precursors. Further model evaluation is necessary to determine the tolerance of CMB to
deviations from this assumption.
The current practice is to apportion the primary material that has not changed between
source and receptor. The remaining quantities of reactive species such as ammonium, nitrate,
sulfate, and organic carbon are then apportioned to chemical compounds (single constituent
source type) rather than directly to sources. While this approach is not as satisfying as a direct
apportionment, it at least separates primary from secondary emitters, and the types of compounds
apportioned give some insight into the chemical pathways that formed them. As noted in
Section 3, when profiles are coupled with chemical reaction mechanisms and rates, deposition
velocities, atmospheric equilibrium, and methods to estimate transport and aging time, it is
possible to produce "aged" source profiles which will allow this direct attribution of reactive
species to sources. This apportionment requires measurements of gaseous as well as particulate
species at receptor sites, and is one of the main arguments for combining PMzs and VOC source
apportionment studies together.
A major challenge to the application of CMB is the identification of the primary
contributing sources for inclusion in the model, Assumption 3. Watson (1979) systematically
increased the number of sources contributing to his simulated data from four to eight
contributors while solving the CMB equations assuming only four sources. More sources were
included in the least squares solution than those that were actually contributors.
These studies found that underrepresenting the number of sources had little effect on the
calculated source contributions if the prominent species contributed by the missing sources were
excluded from the solution. When the number of sources was underrepresented, and when
prominent species of the omitted sources were included in the calculation of source
contributions, the contributions of sources with properties in common with the omitted sources
were overestimated. When source types actually present were excluded from the solution, ratios
of calculated to measured concentrations were often outside of the 0.5 to 2.0 range, and the sum
of the source contributions was much less than the total measured mass. The low
calculated/measured ratios indicated which source compositions should be included. When the
number of sources was overrepresented, the sources not actually present yielded contributions
4-2
-------
less than their standard errors if their source profiles were significantly distinct from those of
other sources. The over-specification of sources decreased the standard errors of the source
contribution estimates.
Determining deviations from Assumption 4, the linear independence of source
compositions, is one of the main goals of CMB validation. The degree of collinearity depends
on the number of source categories contributing to influential fitting species, the relative
contributions from source types with similar (but not identical) profiles, the variability of species
abundances in the profiles, and the relative contribution from each category. These conditions
vary from sample to sample, so it is not possible to state that two or more profiles are overly
collinear prior to applying them to a specific sample. Similarly, the presence or absence of a
"unique" or "tracer" species does not guarantee that collinearity is eliminated, especially if the
"tracer" is at a very low abundance (e.g., <0.1%) and is highly variable. The variability of the
profile abundances is more influential than the distinctness of the chemical species, in many
cases.
Lowenthal etal. (1992), for example, showed that diesel and gasoline vehicle exhaust
were non-collinear in a simple airshed where they were the major source of carbon. When a
vegetative burning contribution was present, however, the diesel and gasoline exhaust profiles
were too collinear to allow discrimination of their contributions, and only a composite "motor
vehicle exhaust" contribution could be estimated.
Gordon etal. (1981) found instabilities in the ordinary weighted least square solutions to
the CMB equations when species presumed to be "unique" to a certain source type were
removed from the solution. Using simulated data with known perturbations ranging from 0 to
20%, Watson (1979) found that in the presence of likely uncertainties, sources such as urban
dust and continental background dust cannot be adequately resolved by least squares fitting, even
though their compositions are not identical. Several nearly unique ratios must exist for good
separation.
Several "regression diagnostics" have been proposed for least squares estimation
methods similar to the CMB effective variance solution (e.g., Belsley et a/., 1980; DeCesar et
a/., 1985a, 1985b). Kim and Henry (1989) show that most of these diagnostics are not
meaningful because they are based on the assumption of zero uncertainty in the source profiles.
Kim and Henry (1989) demonstrate, through the examination of randomly perturbed model input
data, that the values for these diagnostics vary substantially with typical random changes in the
source profiles. Tests performed on simulated data with obviously collinear source compositions
typically result in positive and negative values for the collinear source types as well as large
standard errors in the collinear source contribution estimates. Unless the source compositions
are nearly identical, the sum of these large positive and negative values very closely
approximates the sum of the true contributions.
CMB8 makes the collinearity measures proposed by Henry (1992) more transparent to
identify the degree of collinearity. These measure the degree of overlap among source profiles
as if they were vectors in a multi-dimensional space. The user can set the overlap he or she is
willing to tolerate for a selected maximum uncertainty in the quantity being apportioned (i.e.,
total VOC or PM mass to which the profiles are normalized). Little guidance is given in this
protocol or elsewhere on how to select these overlaps and uncertainties, or on what the
implication of that selection might be. By having these options available in CMB 8, however, it
4-3
-------
is hoped that such a body of knowledge can be acquired as more source apportionment studies
are completed and the collinearity issue is studied in greater detail.
With most commonly measured species for particles (e.g., ions, elements, carbon) and
common source types (e.g., motor vehicle, geological, residual oil, sea salt, steel production,
wood burning, various industrial processes, secondary sulfuric acid, secondary ammonium
bisulfate, secondary ammonium sulfate, secondary ammonium nitrate, secondary sodium
nitrate), approximately five to seven source types are linearly independent of each other. About
the same number of VOC source types (e.g., motor vehicle exhaust, liquid gasoline, evaporated
gasoline, degreasers and coatings, graphic arts, biogenics) can be distinguished with most
commonly measured species for VOC (e.g., C2 - Cio hydrocarbons in canisters). The degree of
resolution and number of source types can be enhanced substantially, as will be shown in Section
5, when more detailed particle and gaseous organic compounds are measured at source and
receptor, and when gas and particles are measured in conjunction with each other.
With regard to Assumption 5, the true number of individual sources contributing to
receptor concentrations is generally much larger than the number of species that can be
measured. It is therefore necessary to group sources into source types of similar compositions so
that this assumption is met. For the most commonly measured species, meeting Assumption 4
practically defines these groupings.
With respect to Assumption 6 (the randomness, normality, and the uncorrelated nature of
measurement uncertainties), there are few results available from verification or evaluation
studies. Every least squares solution to the CMB equations requires this assumption, as
demonstrated by the derivation of Watson et al. (1984). In reality, very little is known about the
distribution of errors for the source compositions and the ambient concentrations. If anything,
the distribution probably follows a log-normal rather than a normal distribution. Ambient
concentrations can never be negative, and a normal distribution allows a substantial proportion
of negative values, while a log-normal distribution allows no negative values. For small errors
(e.g., less than 20%), the actual distribution may not be important, but for large errors it probably
is important. A symmetric distribution becomes less probable as the coefficient of variation of
the measurement increases. This assumption still requires further evaluation to determine the
effects of its deviations.
4.2 CMB Performance Measures
Figures 4.2-1, 4.2-2, and 4.2-3 show the three segments of CMB model output that are
displayed each time the model is applied to a set of source profiles and chemical species. These
outputs accompany each application. Table 4.2-1 describes the model outputs and performance
measures in these displays. The use of these measures to evaluate CMB solutions is explained in
subsequent sections.
4-4
-------
Chemical Mass Balance Version EPA-CMB8.2
Report Date: 11/8/2004
SAMPLE: OPTIONS: I
SITE: WELBY BRITT & LUECKE: No
SAMPLE DATE: 01/1
DURATION: 6
START HOUR: 06
SIZE: Fine
Species Array: 5
Sources Array: 1
FITTING STATISTICS:
R SQUARE
CHI SQUARE
SOURCE CONTRIBUTION
SOURCE
EST CODE NAME
YES N001 NVNSP
YES N007 NVNSP2
YES N010 NVSM
YES N013 NWHD
YES N050 NMc
YES N055 NWFSc
YES N067 NWSHc2
YES N074 NRDC
YES N082 AMSUL
YES N084 AMNIT
YES N124 PCHCLC1
7/97 SOURCE ELIMINATION: No
BEST FIT:
0.92
0.61
ESTIMATES:
SCE((jg/m3)
3.32689
0.92216
6.50609
7.23091
1.86174
0.81836
2.45221
6.37413
6.84817
13.68479
-0.25354
49.77192
MEASURED CONCENTRATION FOR SIZE:
53.1+- 2.
7
No
% MASS 93.8
NPUT FILES:
INnf raqs . i n8
PRnf raqs . sel
SPnf raqs . sel
.XXX
ADnf raqs . txt
PRnfraqs.txt
DEGREES FREEDOM 72
Std Err
2.06033
0.41334
3.73659
1.83963
2.03008
0.47345
1.30867
1.65625
0.83248
1.33685
1.24129
Fi ne
Tstat
1.61474
2.23096
1.74118
3.93062
0.91708
1.72850
1.87383
3.84852
8.22628
10.23660
-0.20425
Figure 4.2-1. CMB8 source contribution display.
4-5
-------
ELIGIBLE SPACE DIM.
1 / Singular Value
Eligible Space Display
11 FOR MAX. UNC. = 10.61606 (20.% OF TOTAL MEAS. MASS)
0.25514 0.37985
1.86357 2.18762
0.75550 0.86353 0.93826 1.34737 1.36735 1.59420
4.20631
NUMBER ESTIMABLE SOURCES = 11 FOR MIN. PROJ. = 0.95
PROJ. SOURCE PROJ. SOURCE PROJ. SOURCE PROJ. SOURCE PROJ. SOURCE
1.0000 N001
1.0000 N055
1.0000 N124
1.0000 N007
1.0000 N067
1.0000 N010
1.0000 N074
1.0000 N013
1.0000 N082
1.0000 N050
1.0000 N084
ESTIMABLE LINEAR COMBINATIONS OF INESTIMABLE SOURCES
COEFF. SOURCE COEFF. SOURCE COEFF. SOURCE COEFF. SOURCE SCE Std Err
Figure 4.2-2. Eligible space collinearity display.
4-6
-------
Figure 4.2-3. Species concentration display.
SPECIES CONCENTRATIONS:
SPECIES FIT MEASURED
MSGC
CLIC
N3IC
S4IC
N4CC
KPAC
TCTC
OCTC
ECTC
NAXC
MGXC
ALXC
SIXC
PHXC
SUXC
CLXC
KPXC
CAXC
TIXC
VAXC
CRXC
MNXC
FEXC
NIXC
CUXC
ZNXC
ASXC
SEXC
BRXC
RBXC
SRXC
ZRXC
HGXC
PBXC
NAPHTH
MNAPH2
MNAPH1
DMN267
DM1367
D14523
DMN12
BIPHEN
M 2BPH
M 3BPH
M 4BPH
ATMNAP
EM 12N
BTMNAP
CTMNAP
EM 21N
ETMNAP
FTMNAP
GTMNAP
HTMNAP
TM128N
ACNAPY
ACNAPE
PHENAN
MSGU
CLIU *
N3IU *
S4IU *
N4CU *
KPAU *
TCTU
OCTU *
ECTU *
NAXU *
MGXU *
ALXU *
SIXU *
PHXU *
SUXU
CLXU *
KPXU *
CAXU *
TIXU *
VAXU *
CRXU *
MNXU *
FEXU *
NIXU *
CUXU
ZNXU
ASXU *
SEXU *
BRXU *
RBXU *
SRXU *
ZRXU *
HGXU *
PBXU *
NAPHTH
MNAPH2
MNAPH1
DMN267
DM1367
D14523
DMN12U
BIPHEN
M 2BPH
M 3BPH
M 4BPH
ATMNAP
EM 12N
BTMNAP
CTMNAP
EM 21N
ETMNAP
FTMNAP
GTMNAP
HTMNAP
TM128N
ACNAPY
ACNAPE
PHENAN *
53
0
10
5
5
0
21
13
8
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.08030+-
.45540+-
.48780+-
.03660+-
.04460+-
.06260+-
.55080+-
.42380+-
.12700+-
.07980+-
.01450<
.13630+-
.52050+-
.00000<
.05200+-
.34080+-
.12110+-
.21120+-
.00000<
.00000<
.00000<
.01120+-
.34700+-
.00010<
.01190+-
.09250+-
.00250<
.00110+-
.00490+-
.00020<
.00170+-
.00010<
.00000<
.01120+-
.15685+-
.54554+-
.32017+-
.13343+-
.21683+-
.06834+-
.02676+-
.08488+-
.01973+-
.07953+-
.04232+-
.06545+-
.02026+-
.06678+-
.07318+-
.00257+-
.05283+-
.04885+-
.02637+-
.00561+-
.00283+-
.02183+-
.02829+-
.04598+-
2.70112
0.05500
0.69455
0.39970
0.26760
0.00540
1.38265
1.14530
0.77460
0.04150
0.04340
0.01320
0.02830
0.01640
0.10310
0.02290
0.00820
0.01240
0.02990
0.01660
0.00510
0.00180
0.01770
0.00140
0.00120
0.00480
0.00340
0.00100
0.00090
0.00120
0.00090
0.00190
0.00300
0.00280
0.06375
0.03148
0.01963
0.00974
0.01504
0.00613
0.00310
0.00494
0.00147
0.00443
0.00275
0.00444
0.00192
0.00496
0.00484
0.00056
0.00473
0.00378
0.00279
0.00100
0.00097
0.00300
0.00216
0.00297
49
0
10
5
4
0
21
13
8
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
0
-0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
CALCULATED
CALCULATED
.77192+-
.19340+-
.66432+-
.09540+-
.96436+-
.02193+-
.47394+-
.44418+-
.02976+-
.11505+-
.05238<
.34546+-
.13702+-
.00529<
.74683+-
.20754+-
.16756+-
.17604+-
.01525<
.00097<
.0015K
.00385+-
.24112+-
.00035<
.00551+-
.01933+-
.00020<
.00004+-
.00141+-
.00077<
.00150+-
.00079<
.00010<
.00576+-
.10813+-
.47372+-
.25244+-
.05868+-
.08922+-
.02964+-
.01277+-
.02298+-
.00207+-
.01531+-
.00831+-
.02170+-
.00753+-
.02225+-
.02223+-
.00155+-
.01536+-
.01568+-
.00904+-
.00399+-
.00128+-
.09141+-
.01991+-
.10035+-
2.80479
0.16807
1.06124
0.50542
0.37570
0.10357
0.57545
0.86062
0.93415
0.10432
0.06781
0.21019
0.52088
0.06631
0.17037
0.15976
0.06941
0.11150
0.03837
0.01616
0.00388
0.00311
0.14516
0.00226
0.00513
0.01895
0.00464
0.00226
0.00383
0.00213
0.00234
0.00277
0.00404
0.01204
0.70144
0.23977
0.11945
0.02370
0.03454
0.01144
0.00481
0.00887
0.00210
0.00639
0.00350
0.00843
0.00304
0.00791
0.00744
0.00174
0.00514
0.00527
0.00316
0.00197
0.00170
0.04484
0.02033
0.03346
MEASURED
0
0
1
1
0
0
1
1
0
1
3
2
2
0
0
0
1
0
0
0
0
0
0
3
0
0
0
-0
0
3
0
7
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4
0
2
.94+- 0.
.42+- 0.
.02+- 0.
.01+- 0.
.98+- 0.
.35+- 1.
.00+- 0.
.00+- 0.
.99+- 0.
.44+- 1.
.61< 11.
.53+- 1.
.18+- 1.
.00< 0.
.85+- 0.
.61+- 0.
.38+- 0.
.83+- 0.
.00< 0.
.00< 0.
.00< 0.
.34+- 0.
.69+- 0.
.51< 54.
.46+- 0.
.21+- 0.
.08< 1.
.04+- 2.
.29+- 0.
.87< 25.
.88+- 1.
RESIDUAL
UNCERTAINTY
07
37
12
13
09
65
07
11
15
51
78
56
01
00
09
47
58
53
00
00
00
28
42
03
43
21
86
06
78
55
45
.88< *****
!oo< o.
.51+- 1.
.96+- 0.
.87+- 0.
.79+- 0.
.44+- 0.
.41+- 0.
.43+- 0.
.48+- 0.
.27+- 0.
.10+- 0.
.19+- 0.
.20+- 0.
.33+- 0.
.37+- 0.
.33+- 0.
.30+- 0.
.60+- 0.
.29+- 0.
.32+- 0.
.34+- 0.
.71+- 0.
.45+- 0.
.19+- 2.
.70+- 0.
.18+- 0.
00
08
61
44
38
18
16
17
19
11
11
08
08
13
15
12
10
69
10
11
13
37
62
13
72
74
-0.8
-1.5
0.1
0.1
-0.2
-0.4
-0.1
0.0
-0.1
0.3
0.5
1.0
1.2
0.1
-1.5
-0.8
0.7
-0.3
0.3
0.0
0.2
-2.0
-0.7
0.1
-1.2
-3.7
-0.4
-0.5
-0.9
0.2
-0.1
0.2
0.0
-0.4
-0.1
-0.3
-0.6
-2.9
-3.4
-3.0
-2.4
-6.1
-6.9
-8.3
-7.7
-4.6
-3.5
-4.8
-5.7
-0.6
-5.4
-5.1
-4.1
-0.7
-0.8
1.5
-0.4
1.6
4-7
-------
Figure 4.2-3 (continued). Species concentration display.
FLUORE
A MFLU
M 1FLU
B MFLU
C MFLU
A MPHT
M 2PHT
B MPHT
C MPHT
M 1PHT
DM36PH
A DMPH
B DMPH
C DMPH
DM17PH
D DMPH
E DMPH
ANTHRA
FLUORA
PYRENE
B MPYR
D MPYR
F MPYR
RETENE
BAANTH
CHRYSN
BBJKFL
BEPYRN
BAPYRN
INCDPY
DBANTH
BGHIPE
CORONE
GUACOL
M4GUCL
E4GUCL
SYRGOL
PPGUCL
A4GUCL
GNONLA
F4GUCL
M4SYRG
E4SYRG
ISOEUG
GDECLA
ACETVA
UNGLAC
SYRALD
C27SDS
C27RDS
C27RAC
AB30NH
CHLSRL
BA30NH
AB HOP
BA HOP
SABHHP
RABHHP
STEROM
SABBHH
RABBHH
CO
NOX
S02
FLUORE *
A MFLU *
M 1FLU *
B MFLU *
C MFLU *
A MPHT *
M 2PHT *
B MPHT *
C MPHT *
M 1PHT *
DM36PH *
A DMPH *
B DMPH *
C DMPH *
DM17PH *
D DMPH *
E DMPH *
ANTHRA
FLUORA
PYRENE
B MPYR *
D MPYR *
F MPYR *
RETENE *
BAANTH *
CHRYSN *
BBJKFL *
BEPYRN *
BAPYRN *
INCDPY *
DBANTH *
BGHIPE *
CORONE *
GUACOL
M4GUCL
E4GUCL *
SYRGOL
PPGUCL *
A4GUCL *
GNONLA *
F4GUCL *
M4SYRG *
E4SYRG *
ISOEUG *
GDECLA *
ACETVA *
UNGLAC *
SYRALD
C27SDS *
C27RDS *
C27RAC *
AB30NH *
CHLSRL *
BA30NH *
AB HOP *
BA HOP *
SABHHP *
RABHHP *
STEROM
SABBHH *
RABBHH *
COU
NOXU
S02U
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
3.
0.
9.
03263+-
01556+-
00697+-
00351+-
01187+-
01087+-
01164+-
00065+-
00797+-
00694+-
00269+-
00325+-
00204+-
00650+-
00286+-
00236+-
00222+-
00408+-
00579+-
00694+-
00118+-
00077+-
00094+-
00074+-
00162+-
00177+-
00183+-
00157+-
00127<
00115+-
00027<
00260+-
00162<
04846+-
00080<
00286+-
00000<
00000<
00000<
00487+-
01388+-
00552<
02555+-
02425+-
00227+-
00035<
00313+-
06167+-
00092+-
00080+-
00000<
00139+-
00000<
00092+-
00000<
00012<
00062+-
00035<
00000<
00038<
00027<
29642+-
60373+-
15143+-
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
.00278
.00144
.00094
.00057
.00109
.00100
.00105
.00046
.00078
.00079
.00064
.00069
.00047
.00067
.00064
.00052
.00064
.00098
.00084
.00064
.00042
.00042
.00042
.00059
.00126
.00067
.00076
.00059
.00138
.00113
.00163
.00147
.00292
.01833
.00109
.00105
.00171
.00046
.00054
.00123
.00709
.00602
.00566
.00359
.00113
.00196
.00122
.01311
.00067
.00050
.00063
.00075
.00447
.00063
.00063
.00042
.00050
.00042
.00731
.00046
.00042
.34448
.06120
.08306
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7
0
-9
.03197+-
.01204+-
.00559+-
.00287+-
.02067+-
.00989+-
.01089+-
.00330+-
.00680+-
.00729+-
.00207+-
.00268+-
.00134+-
.00426+-
.00211+-
.00171+-
.00185+-
.02458+-
.03360+-
.04160+-
.00079+-
.00107+-
.00085+-
.00035+-
.00152+-
.00111+-
.00214+-
.00075+-
.00093<
.00055+-
.00006<
.00169+-
.00122<
.02527+-
.02785<
.00656+-
.04130<
.00090<
.00417<
.00444+-
.01459+-
.01250<
.00551+-
.01139+-
.00176+-
.00408<
.00706+-
.00321+-
.00052+-
.00041+-
.0008K
.00120+-
.00115<
.00074+-
.00095<
.0003K
.00021+-
.00015<
.00035<
.00015<
.00009<
.22289+-
.32596+-
.95237+-
0.01276
0.00448
0.00243
0.00189
0.00713
0.00409
0.00442
0.00254
0.00308
0.00321
0.00187
0.00206
0.00176
0.00252
0.00189
0.00178
0.00184
0.00835
0.01209
0.01530
0.00169
0.00169
0.00170
0.00167
0.00176
0.00169
0.00182
0.00171
0.00177
0.00172
0.00176
0.00220
0.00216
0.01222
0.01598
0.00394
0.02025
0.00172
0.00266
0.00253
0.00660
0.00437
0.00194
0.00457
0.00190
0.00246
0.00665
0.00207
0.00168
0.00167
0.00172
0.00174
0.00190
0.00173
0.00170
0.00172
0.00168
0.00168
0.00178
0.00168
0.00422
6.74482
0.21088
5.06539
0.98+-
0.77+-
0.80+-
0.82+-
1.74+-
0.91+-
0.94+-
5.08+-
0.85+-
1.05+-
0.77+-
0.83+-
0.66+-
0.66+-
0.74+-
0.72+-
0.83+-
6.03+-
5.80+-
5.99+-
0.67+-
1.39+-
0.90+-
0.47+-
0.94+-
0.62+-
1.17+-
0.48+-
0.73<
0.48+-
0.22<
0.65+-
0.75<
0.52+-
34.94<
2.29+-
0.00<
0.00<
0.00<
0.91+-
1.05+-
2.26<
0.22+-
0.47+-
0.78+-
11 . 52<
2i26+-
0.05+-
0.57+-
0.51+-
0.00<
0.86+-
0.00<
0.81+-
0.00<
2.6K
0.34+-
0.43<
0.00<
0.40<
0.36<
2.19+-
0.54+-
-1.09+-
0.40
0.30
0.37
0.55
0.62
0.39
0.39
5.31
0.39
0.48
0.72
0.66
0.88
0.39
0.68
0.77
0.86
2.51
2.25
2.27
1.45
2.33
1.84
2.30
1.30
0.98
1.11
1.11
1.60
1.56
6.74
0.92
1.90
0.32
51.70
1.61
0.00
0.00
0.00
0.57
0.72
2.59
0.09
0.20
0.92
64.23
2.30
0.04
1.88
2.12
0.00
1.34
0.00
1.97
0.00
17.28
2.72
4.76
0.00
4.39
15.86
2.06
0.35
0.82
-0.1
-0.7
-0.5
-0.3
1.2
-0.2
-0.2
1.0
-0.4
0.1
-0.3
-0.3
-0.4
-0.9
-0.4
-0.4
-0.2
2.4
2.3
2.3
-0.2
0.2
-0.1
-0.2
0.0
-0.4
0.2
-0.5
-0.2
-0.3
-0.1
-0.3
-0.1
-1.1
1.7
0.9
2.0
0.5
1.5
-0.2
0.1
0.9
-3.4
-2.2
-0.2
1.2
0.6
-4.4
-0.2
-0.2
0.4
-0.1
0.2
-0.1
0.5
0.1
-0.2
-0.1
0.0
-0.1
0.0
0.6
-1.3
-2.7
4-8
-------
Table 4.2-1
CMB8 Outputs and Performance Measures
Output/ Stati sti c/ (J ooe
Abbreviation
Description
Source Contribution Display
Source Contribution Estimate SCE
Contribution from the source type designated by the profile under NAME to
the profile normalizing component (usually PM2.5 mass or total VOC).
Units can be specified in the options menu of CMB8.
Standard Error
Std Err The uncertainty of the source contribution estimate (SCE), expressed as one
standard deviation of the most probable SCE. This is an indicator of the
precision or certainty of each SCE. The STD ERR is estimated by
propagating the uncertainty estimates of the receptor data and source
profiles through the effective variance least-squares calculations. Its
magnitude is a function of the uncertainties in the input data and the amount
of collinearity (i.e., degree of similarity) among source profiles. Two or
three times the standard error may be taken as an upper limit of the source
contribution. [Target Std Err « SCE]
Tstat Ratio of the SCE to its Std Err. A high Tstat suggests a nonzero SCE.
[Target > 2.0]
R-square R-SQUARE Variance in ambient species concentrations explained by the calculated
species concentrations. A low R SQUARE (<0.8) indicates that the selected
source profiles have not accounted for the variance in the selected receptor
concentrations. Ranges from 0 to 1.0. [Target 0.8 to 1.0.]
t-Statistic
PERCENT The sum of SCE divided by the total mass or VOC concentration. A value
MASS or approaching 100% is desired. A %MASS near 100% can be misleading
%MASS because a poor fit can force a high %MASS. [Target 100% ± 20%.]
DF The number of species in fit minus number of sources in fit. Solutions with
larger degrees of freedom are typically more stable and robust than ones
with small degrees of freedom. [Target > 5]
Percent Mass Accounted For
Degrees of Freedom
Chi-square
CHI SQUARE
Similar to R-SQUARE except that it also considers the uncertainties of the
calculated species concentrations. A large CHI SQUARE (>4.0) means that
one or more of the calculated species concentrations differs from the
measured concentrations by several uncertainty intervals. The values for
these statistics exceed their targets when: 1) contributing sources have been
omitted from the CMB calculation; 2) one or more source profiles have
been selected which do not represent the contributing source types; 3)
uncertainty estimates of receptor or source profile data are underestimated;
and/or 4) source or receptor data are inaccurate. CHI SQUARE is the
square root of the sum of the squares of the RATIO R/U that correspond to
fitting species divided by the DF. [Target 0.0 to 4.0]
4-9
-------
Table 4.2-1
CMB8 Outputs and Performance Measures
Output/ Stati sti c/C ode
Abbreviation
Description
Source Contribution Display (cont.)
Site, sample duration, date, start
hour, size
Describes the sample being modeled by location, time, and length of
sample. Size refers to different particle size fractions, typically PM10 or
PM2.5 (sometimes called "fine" particles).
Britt and Luecke Solution
Source Elimination
B and L A "Yes" flag indicates that the complex Britt and Luecke (1973) solution
has been applied. A "No" flag means the default effective variance solution
has been applied.
SRC ELEVI The source elimination option automatically removes negative SCE or SCE
less than the corresponding Std Err before printing the solution. A "Yes"
flag means that the option is on and the default "No" means it is off. It is
recommended that negative and negligible source types be removed
manually as they may be indicators of collinearity that should be considered
when interpreting the source categories represented by source profiles.
Allows the "best" solution to be obtained automatically among up to ten
combinations of source profiles based on a relative weighting of the chi-
square, R-Square, Percent Mass, and Estimable Sources performance
measures. The weights can be set in the options menu.
Weights for CHI SQR, R SQR,
PCMASS, andFRCEST
Estimable Source Profile
EST A "Yes" flag in this column indicates that the source is estimable within the
uncertainty parameters defined in the options menu. A "No" flag indicates
that the source is not estimable within the uncertainty parameters.
Code and Name
The source code matches the profile with the source combinations in the
source selection file. The name corresponds to a short mnemonic that
designates the source profile.
4- 10
-------
Table 4.2-1
CMB8 Outputs and Performance Measures
Output/ Stati sti c/C ode
Abbreviation Description
Estimable Space Display
Eligible Space Dimension and
Maximum Uncertainty
Singular Value
Number of Estimable Sources
Estimable Linear Combinations
Replaces U/S CLUSTERS and SUM OF CLUSTER SOURCES in CMB7
This treatment (Henry, 1992) uses two parameters, maximum source
uncertainty and minimum source projection on the eligible space. These are
set to default values of 1.0 and 0.95, respectively, in CMB8. The maximum
source uncertainty determines the eligible space to be spanned by the
eigenvectors whose inverse singular values are less than or equal to the
maximum source uncertainty. Estimable sources are defined to be those
whose projection on the eligible space is at least the minimum source
projection. Inestimable sources are sources that are not estimable. To
modify these values click in the edit boxes and edit with keyboard entry.
The singular value decomposition of the source transfer matrix.
The sources that are estimable given their source contributions and
propagated uncertainties. This changes with the acceptable uncertainty
specified in the options menu.
COEFF. Show clusters of sources which the model cannot easily distinguish between
SOURCE and that are likely to be interfering with the model's ability to provide a
good set of SCE's. [Target - No clusters.]
SCE Estimates the sum of SCE's of the sources in a cluster and the standard error
Std Err of the sum. Not needed if source profiles of cluster sources can be
improved. The standard error of the SCE follows the ± in the display.
4- 11
-------
Table 4.2-1
CMB8 Outputs and Performance Measures
Output/ Stati sti c/C ode
Abbreviation
Description
Species Concentration Display
Selected Species In the Fit
Missing Measurement for Species
Measured Species Concentration
Calculated Species Concentration
Ratio of Calculated to Measured
Species
M
MEAS
CALC
CALCULATED
MEASURED
Ratio of Residual to Its Uncertainty RESIDUAL
UNCERTAINTY
A "*" in this column indicates the species is included in the calculation of
the source contribution estimate.
Status: M in column indicates missing measurement. These are indicated
by -99 in the input data set.
Ambient species concentrations (measurements and uncertainties)
Calculated chemical concentrations and propagated uncertainties based on
the selected profiles and the source contribution estimates. These are
reported both for fitting and non-fitting species.
Ratio of CALC/MEAS and its uncertainty Used to identify species that are
over/under accounted for by the model. The ratios should be near 1.00 if
the model has accurately explained the measured concentrations. Ratios
that deviate from unity by more than two uncertainty intervals indicate that
an incorrect set of profiles is being used to explain the measured
concentrations. [Target 0.5 to 2.O.]
Ratio of the signed difference between the calculated and measured
concentration (i.e., the residual) divided by the uncertainty of that residual
(i.e., square root of the sum of the squares of the uncertainty in the
calculated and measured concentrations). Used to identify species that are
over- or under-accounted for by the model. The RATIO R/U specifies the
number of uncertainty intervals by which the calculated and measured
concentrations differ. When the absolute value of the RATIO R/U exceeds
2, the residual is significant. If it is positive, then one or more of the
profiles is contributing too much to that species. If it is negative, then there
is an insufficient contribution to that species and a source may be missing.
The sum of the squared RATIO R/U for fitting species divided by the
degrees of freedom yields the CHI SQUARE. The highest RATIO R/U
values for fitting species are the cause of high CHI SQUARE values.
[Target |<2.0|.]
4- 12
-------
Output/Statistic/Code Abbreviation Description
Command Display
Modified Psuedo Inverse Matrix MPIN Shows which species most influence the source contribution estimate
corresponding to each profile. It is examined to determine that the logical
marker species are having the most influence on the apportionment.
Species - Source Contribution SSCONT Shows the fraction of each measured species concentration that is
accounted for by the calculated species for each source or source category.
This can be > 1.0 for a particular source if that species is over-accounted
for by the fit. It is used to identify the sources which are accounting for
particular species.
4- i:
-------
4.3 Protocol Steps
Each of the seven steps in the application and validation protocol is described below with
respect to their general application. They are illustrated in greater detail for specific examples in
Sections 5 and 6.
4.3.1 Determine the Applicability of CMB
The following conditions must be met for CMB to be applicable:
1. A sufficient number of PM or VOC receptor samples have been taken with accepted
sampling methods to fulfill study objectives. If objectives are to determine how to
attain NAAQS, samples should represent annual average and maximum
concentrations for PM2.5 and PMio and correspond to maximum 8-hour average ozone
concentrations for VOC.
2. Samples are amenable to or have been analyzed for a variety of chemical species. As
noted above, elements, ions, and carbon are the minimal needs for PM apportionment
and light hydrocarbons in canisters or automatic gas chromatographs are the minimal
requirements for VOC apportionment.
3. Potential source contributors can be identified and grouped into source categories of
distinct chemical compositions with respect to the receptor species available from
requirement 2.
4. Source profiles are available, from the study area or from similar sources, that
represent the source compositions as they would appear at the receptors. Changes in
source composition between source and receptor must be accommodated in order for
the model to be physically meaningful.
5. The number of source types in a single application of CMB must be fewer than the
number of chemical species measured above lower quantifiable limits at the receptor.
Unless all five of the above requirements are met, the Chemical Mass Balance receptor
model is not applicable to the situation under study. These are necessary, but not sufficient,
requirements, and it may still be found that even though these requirements are met, the
precision and validity of CMB results are not adequate for control strategy decisions. The
remaining steps in the applications and validation protocol must be completed to arrive at this
conclusion.
4.3.2 Format Input Files and Perform Initial Model Runs
CMB8 allows input data files to be prepared in spreadsheet formats and, with
contemporary computer memories, has no practical limit on the number of source profiles,
chemical species, or individual samples that can be included in a single file. It is convenient,
however, to divide input data into groups by site or season when data sets are large.
The initial model runs usually contain many more profiles than are used in production
runs to determine how different composites might affect the precision and stability of the source
contribution estimates. One or more initial arrays (combinations) of source profiles are usually
examined during this step. Various arrays of fitting species are also examined.
4- 14
-------
During this step, it may be necessary to modify ambient data or source profiles by
making additional assumptions. This is very often the case when some profiles are obtained
from another study and may not report all of the species available in the other profiles or in the
receptor samples. A default value of zero with a standard deviation equal to an analytical
detection limit may be assigned to a species in a source profile if that species is known to be
absent from that source type from previous tests of similar sources.
When selecting fitting species, only one of the different measurements of the same
species (such as elemental carbon and total carbon or sulfur and sulfate) should be included in
the fit. If more than one measurement of the same species is included in the CMB solution, then
that species influences the source contribution estimates more than it should. This does not
apply to soluble and insoluble species (such as potassium), which are really different species that
distinguish among source types.
Concentrations with values below detection limits may be included only if their
uncertainty is also included. Minimum detection limits may be used to estimate this uncertainty
if it is not otherwise reported. If the uncertainty is underestimated or is not specified (and given
a default value of zero), then these very imprecise measurements will have an excessive
influence on the source contribution estimates.
Secondary components can be represented by their chemical form. In the simplest case, a
single constituent source type (Section 4.1), such as secondary organic carbon that contains only
an entry in the organic carbon column, may be used. These should be used sparingly, however,
because a single constituent type effectively removes the influence of that source profile species
on the source contribution estimates. A discussion and example application for PMio is
presented in Appendix G.
Uncertainties assigned to the measurements for use in the CMB application should be
reviewed to ensure that they are realistic estimates (see Appendix B). Measurement
uncertainties should be provided as part of the measurement process. Typical measurement
uncertainties are on the order of ±5% to ±20%, with some species being more uncertain than
others because of analytical interferences and proximity to detection limits. Uncertainties in
source profiles could be much greater. The model considers these uncertainties when it develops
the "fit". Species with high uncertainties are unlikely to be very influential in the fit.
Chemical measurements are usually reported with their measurement uncertainties
determined from error propagation of chemical analysis and flow rate uncertainties (e.g., Watson
et a/., 1995). These uncertainties are determined from periodic performance tests and replicate
analyses. The reporting of these uncertainties should be specified when the measurements are
made. If chemical concentrations are available without uncertainties, typical uncertainties may
be assigned based on those reported in previous analyses. The value of the diagnostics provided
by the CMB software is substantially decreased without an adequate and accurate definition of
measurement uncertainties in receptor data.
In most cases, the individual samples should be run separately in CMB. Compositing or
combining the data from several samples will usually decrease the number of sources that CMB
can resolve. Likewise, separate analysis of different PM size fractions is preferable to a "total"
sample that combines the two size fractions. The sources contributing to these two size fractions
are generally quite different.
4- 15
-------
Several source profiles for each source type may be included in the source profile input
files, but only one profile from each type should be included in a fit. The set of profiles that best
explains the measurements may differ from sample to sample, both because the profiles are
different and because the source contributions change in magnitude. Several sources will nearly
always contribute, and profiles should be included to represent vehicle exhaust, suspended dust,
secondary sulfate, and possibly secondary nitrate or vegetative burning. Natural sources, such as
sea salt or wind blown dust, should be included if these are in the proximity of the receptors.
Point sources, such as coal-fired power stations, steel mills, cement production facilities, and
other industrial sources in an emissions inventory are next in priority. These may be very
directional, depending on which way winds are oriented between source and receptor. Finally,
single constituent source types can be added as a last resort when there is no other explanation
for a chemical species. This is sometimes done for zinc and copper, which are often in excess
owing to nearby plating or metal handling operations.
In selecting source profiles for inclusion in a fit, it is helpful to review wind direction
data and eliminate sources that have virtually no chance of contributing a detectable
concentration because they are downwind of the receptor. Source types that are unlikely to be
emitting during the period of time being studied (e.g., wood smoke emissions during hot summer
months) can be omitted, or their profiles should be replaced with ones that represent wildfires or
prescribed burning that might occur during that period.
The final selection of the most appropriate source types and the profiles to represent
those source types results from interactive applications of CMB with an evaluation of the
diagnostic measures. It is possible that more than one subset of source types and source profiles
will fit the receptor data equally well. The interactive application of the model to different
source subsets will identify these cases.
Some sources have emissions that are chemically similar or consistent over time, i.e.,
although the absolute magnitude of the emissions may vary, the relative composition of many of
the measured species present in a source may be sufficiently stable. However, the chemistry of
some species could be variable if the source changes its operating conditions, feedstock, or fuel.
This variability must be reflected in the uncertainties that are assigned to each species in the
profile. (These concerns about source profile variability are analogous to those faced by the
dispersion modeler when estimating emission rates or dispersion parameters.)
Because the CMB model uses the information provided by all species included in the fit,
mis-estimation of a single species, even so-called "tracer" species, may not appreciably affect
the source contribution estimates. This is especially true if these species have been assigned
uncertainties which reflect their variability. When these uncertainties are adequately estimated,
other, less variable species provide a larger influence on the source contribution estimates.
4- 16
-------
4.3.3 Evaluate Outputs and Performance Measures
Model outputs and performance measures are described in Table 4.3-1. These are
examined for different combinations of fitting profiles and fitting species to determine the
optimal fit to the data. This process will become more evident when applied to the specific
examples in Sections 5 and 6.
4.3.4 Evaluate Deviations from Model Assumptions
The CMB performance measures and tests using different profiles and fitting species can
often indicate when deviations from model assumptions may have occurred. These deviations
do not necessarily invalidate the CMB results - they merely indicate the potential for invalidity.
This is why a separate step is necessary in the applications and validation protocol which
evaluates the effects of these deviations from assumptions and determines whether or not these
effects can be tolerated.
4.3.5 Modify Model Inputs to Remediate Problems
There are four main categories of problems which, once they have been identified, can be
addressed to improve the performance of the model. The problem categories are: 1) insufficient
receptor measurements; 2) insufficient source measurements; 3) incorrect profile combinations;
and 4) source profile collinearity. Not all "indications" must persist for a problem to be present.
The more "indications" that persist, the more evidence of a problem. Because of the complex
interactions of all of the data in a least squares estimate, the statistics or diagnostics may not
always be adequate to conclusively isolate a problem with model input. Additional physical
evidence is also very helpful.
There may be inaccuracies in the receptor measurements that have not been uncovered in
the routine data validation. If the data are "suspect" and there are no apparent data entry or
analytical errors, the next step would be to eliminate the suspect species from the fit and rerun
the model. Examine the changes in the estimates for each source. If the estimate changes by
more than one standard error, and if the receptor concentration or a source profile value for the
removed species is suspect, then either remeasure the species or use the SCE calculated without
that species in the fit. A RATIO R/U « -2.0 for a species suggests either the ambient data are
high or the profile data are low for that species while a RATIO R/U » 2.0 for a species would
imply that the ambient data are low or the profile data for that species are high. In this case, it is
prudent to:
1. Review the uncertainty assigned to the species with the high residuals. Make any
justifiable and appropriate changes and rerun CMB. If this improves the RATIO
R/U, Step 2 is not necessary.
2. Delete the suspect species from the list of fitting species and rerun. If the SCE
changes by at least one standard error, do not use this species in the fit until it has
been remeasured.
An unacceptable RATIO R/U can also indicate that the set of profiles is not optimized or
that the uncertainty for that species is underestimated in the receptor measurements or source
profiles.
4- 17
-------
A gross error in the value of one or more species in a profile might result in a high
standard error in the SCE and a high residual for those species. Therefore, one or more high
residual values suggests that the uncertain source profile (and the associated species in
particular) be checked and remeasured if necessary. This condition may be indicated by a SCE
that is inconsistent with preliminary analyses or physical evidence. If one or more species has a
"high" (positive or negative) residual which cannot be attributed to incorrect ambient data, one
should examine the SSCONT to see if one source contribution dominates that species. In this
case, review profile data for the suspect species carefully. Correct or remeasure profile if
necessary.
Missing source types are identified by a low percent mass explained (e.g., less than 80%)
and/or a RATIO R/U « -2.0 for chemical species which are in the missing source. A "high
negative" residual for one or more species and a high Chi-Square are also indicative of missing
sources. The key to identifying these sources resides in the calculated to measured chemical
concentrations listed in the species concentration display. "High negative" residuals imply that a
source is needed which will supply a larger quantity of that species. The source profiles may be
listed from CMB8 to assure that they have been properly formatted and read into the software.
These profiles can be examined to determine which ones would supply sufficient quantities of
the missing concentrations if they were added to the set of fitting sources. The CMB model can
be reapplied as many times as is necessary to determine which source types and source profiles
best account for the underestimated receptor concentrations. A source should not be included in
the final fit just because it "explains" the data, however; there must be a physical justification for
the source's contribution at a receptor if it is to be included in the fit.
Noncontributing source types, i.e., source types with contributions lower than detection
limits, are identified by Tstat values below 2. Such source types may be eliminated from the fit
if the source contribution is indeed small. If the source is present but with a very small
contribution to total mass, it should only be removed from the fit if the SSCONT shows that
none of the species in the source account for more than 5% to 10% of the ambient concentration
for those species.
Estimable linear combinations (of inestimable sources) may occur owing to high profile
uncertainty or excessive collinearity with low profile uncertainty. To determine if the
uncertainty in the SCE is due to high profile uncertainty, reduce the uncertainties in the profile to
levels that might be reasonable to achieve if the source profiles were measured more precisely;
then, rerun CMB - if the clusters containing those sources are no longer listed, it is likely that
collinearity per se is not significant. Remeasurement of the profile will probably improve the
uncertainties of the source contribution estimates. It is possible that reducing the uncertainty
will not eliminate the clusters but the SCE uncertainty will likely be improved somewhat. This
would suggest that collinearity is also present.
Remedies for unacceptably high uncertainties due to collinearity can take five forms,
ranked from most to least desirable.
1. The profile of one or more of the cluster sources could be improved by measuring
additional species.
2. Reduce the uncertainties in the source profiles of the cluster sources. If the Tstat
becomes > 2.0, and if these profile uncertainties are realistically achievable by
remeasurement, then the "apparent" collinearity can be improved in large part by
4- 18
-------
improving the uncertainty in the profiles. Ideally, the cluster for that group of sources
would disappear. Remeasure and rerun CMB with the improved measurements.
More precise source profile measurements must be obtained before reapplying the
model.
3. The estimate of the SCE of the source categories that are estimable linear
combinations of inestimable sources. Obtain independent estimates of the
contributions of the individual source categories and use them to apportion the SCEs
into the source categories.
4. Combine the profiles of the collinear source profiles into a single profile of a
"composite source category" that chemically represents the source categories
identified by the estimable linear combinations of inestimable sources. For example,
resuspended road dust and windblown soil dust are chemically similar, and some
modelers include a single term to represent "crustal material" instead of the two
individual source types. This would result in improved source estimates of the crustal
component, which can then serve as an estimate of the combined impact of the two
sources. This aggregate source contribution estimate might then be partitioned into
its components by another method (e.g., dispersion modeling, microscopy, or wind
trajectory analysis).
5. Species that are causing the similarity in source profiles might be deleted from the fit.
These species can often be determined from the display produced by the
Contributions by Species tab in the Main Report. Often one of the inestimable
sources will be » 100% for that species and the other will be negative.
Unfortunately, eliminating too many species from the fit may cause the model to fail
the applicability requirements. Also, the results should acknowledge that the deleted
source may be present.
4.3.6 Evaluate the Consistency and Stability of the Model Results
The CMB estimates should be tested to see how sensitive they are to the various input
data. Unstable estimates (source contribution estimates that change by more than one standard
error) are an indication that the model may not be providing stable results. For CMB validation,
the term "model stability tests" is usually taken to mean the evaluation of model estimates to
changes in input parameters, such as the selected sources and their profiles, as well as selection
of fitting species used to reach a solution with the CMB model.
The CMB model's effective variance fitting procedure uses estimates of the source
profile and receptor concentration uncertainties to "weight" their effect in arriving at source
contribution estimates. It is helpful to explore how sensitive the source contribution estimates
are to changes in the source profiles and these uncertainties. This can be done by introducing
changes into the source profiles and rerunning the model for each change.
The model user can select several species from a source(s) of particular regulatory
interest and assign worst case values to those species in the profile. The model can then be rerun
with the worst case profile(s). A practical way to accomplish this sensitivity analysis is to
include a "worst case" source profile along with the "best estimate" profile in the data file. The
resulting source estimate(s) can be considered "brackets" to the source contribution estimates
and can be compared to the uncertainty intervals calculated for each run. If the bracketing
4- 19
-------
interval is greater than the calculated uncertainty interval, then the model may be sensitive to
changes in the source profiles.
The stability of source contribution estimates with respect to receptor concentrations is
best tested with collocated chemical measurements from one of the sampling sites. These
collocated measurements are usually included as part of the quality assurance plan for a subset of
all samples. If nearly equivalent source contribution estimates are derived from these two
independent measurements of the same ambient air, then the receptor data are not likely causing
instabilities in the CMB results.
Lacking these collocated data, portions of the input data may be perturbed randomly or
systematically in proportion to their uncertainty. The source contribution estimates for the
sources of regulatory interest should not change by more than one standard error in response to
small perturbations if the results are stable. (A "small" perturbation is defined as one standard
error of the ambient species concentrations.) If the results are not stable, the validity of the CMB
result for that particular data set are questionable.
The stability of CMB model results to the fitting species can be evaluated by identifying
a species which SSCONT attributes in large part to a single source. Eliminate this species from
the fit and examine how much the corresponding source contribution changes. If this change is
greater than the Std Err, then that species must be greatly influencing the "fit". Review the
quality of both the source and ambient measurements for that species carefully because of its
influence on the model estimates.
4.3.7 Corroborate CMB Results with Other Modeling and Analyses
If the CMB model is determined to be applicable, the summary statistics and diagnostics
are generally within target ranges, there are no significant deviations from model assumptions,
and the sensitivity tests reveal no unacceptable instability or consistency problems, the CMB
analysis is considered valid. If uncertainties associated with source estimates are too high for
decision-making purposes even after taking the steps recommended in this protocol, then the
source compositions being used are not representative of the sources in the airshed, or they
contain too much uncertainty associated with the influential species.
It is recommended that both a dispersion model and receptor model be used in a
collaborative manner to perform an apportionment, provided that the dispersion model is
applicable and the receptor model is valid for the particular application. Spatial and time series
distributions, similar to the examples in Section 3, should be examined to establish that source
contribution magnitudes are consistent with the locations and timing expected from those
sources.
4-20
-------
5. EXAMPLE OF APPLICATION AND VALIDATION FOR PM? 5
This example demonstrates how the CMB applications and validation protocol is applied
to PM2.5 measurements from the Northern Front Range Air Quality Study (NFRAQS, Watson et
a/., 1998). NFRAQS is a preview of future PM2.5 source apportionment studies that have high
stakes in terms of control strategy development. NFRAQS used a variety of the models
described in Section 2, including CMB, to determine the source categories and individual
emitters contributing to excessive contributions of primary suspended dust, carbon, and
secondary ammonium, sulfate, and nitrate.
The PM2.5 CMB data set is used for this example because it is one of the few data sets
that contains specific organic compounds in both source and receptor measurements. It is
anticipated that, as organic aerosol measurement methods are standardized and the cost of their
application decreases, these measurements will become more standard in future source
apportionment studies.
Each of the seven steps in the protocol is illustrated using examples from this data set.
The CMB input files are available with the other test data sets on EPA's website
(www.epa.gov/scram001 ; Appendix C). The NFRAQS source profiles may be useful for Level
I or Level II assessments in other airsheds in preparation for a more comprehensive source
apportionment study.
5.1 Model Applicability
The requirements for CMB model applicability are: 1) a sufficient number of receptor
samples is taken with an accepted method to evaluate temporal and spatial variations; 2) samples
are analyzed for chemical species which are also present in source emissions; 3) potential source
contributors have been identified and chemically characterized; and 4) the number of
non-collinear source types is less than or equal to the number of measured species.
In NFRAQS, aerosol samples were taken by well-characterized methods and
measurements were fully evaluated (Chow et a/., 1998). Samples were analyzed for 20 days
throughout the winter of 1996-97. Two (i.e., Welby and Brighton) of the nine sampling sites
acquired samples for organic aerosol analysis.
Table 5.1-1 shows an inventory that was especially compiled for NFRAQS, using
published emission factors (not specific to the Denver area) and different activity estimates.
According to this inventory, the major sources were: 1) normal hot-stabilized gasoline-powered
vehicle exhaust; 2) gasoline-powered vehicle emitting visible smoke; 3) diesel exhaust; 4) meat
cooking; 5) wood combustion; 6) road dust and sanding; 7) secondary ammonium sulfate; 8)
secondary ammonium nitrate; and 9) industrial point sources, including coal-fired power
stations, refineries, etc. Owing to previous source testing (Watson et a/., 1990b), it was believed
that cold starts of gasoline vehicles might be a significant contributor and that soft wood (used
mostly in fireplaces because it offers a nice flame) and hard wood (used in wood stoves because
it heats more efficiently) might be distinguishable if the appropriate organic compounds were
measured. Samples were acquired by dilution sampling of vehicles on dynamometers, wood
burning on laboratory stoves, meat cooking in a laboratory kitchen, and grab sampling of
suspendable dust. These samples were analyzed in the laboratory for elements, ions, carbon,
organic aerosol compounds, and 14C using the same methods applied at the receptors. Profiles
from coal-fired electrical generation were available from a previous study in the area (Watson et
al, 1988).
5- 1
-------
Table 5.1-1 Wintertime Emissions Inventory for Denver Metro Area
Source emission rate estimates (tons/day)"
Gas Exhaust
Visibly Smoking Gas Exhaust
Diesel Exhaust
Off-Road Exhaust
Wood Burning
Road Dust & Sand
Coal Power Stations
Other Industries
Natural Gas
Unpaved Road Dust
Restaurant Cooking
Construction Dust
Biogenic
Industrial
Area Sources
Total
PMio
1.7
0.2
5.0
1.8
1.8
49.6
1.3
7.8
0.5
28.2
1.4
2.2
101.6
PMl.5
1.6
0.2
4.9
1.8
1.8
7.4
0.7
2.6
0.5
4.2
1.4
0.3
27.5
SO,
3.3
1.5
1.7
0.0
0.0
62.1
16.7
0.0
0.0
0.0
0.0
85.2
NOV
137.7
36.1
27.4
0.0
0.0
64.3
47.8
28.4
0.0
0.0
0.0
3.0
1.0
345.7
NH, VOC
157.6
8.4
14.3
31.3
34.8
89.8
336.2
CO
1340.8
30.9
111.7
0.0
22.3
72.3
1578.0
"Regional Air Quality Council "Review of Blueprint for Clean Air Emissions Inventories"; April 8, 1998.
Tables 5.1-2 and 5.1-3 identify and describe the profiles available in theNFRAQS data
base. A subset of these profiles was used for testing. The number of fitting species used in
CMB (about 80-85 species with organic species, about 20-25 species with conventional element,
ion, and carbon species) exceeds the number of source types (up to 11 source types). The CMB
model is applicable to source apportionment of this PM2.5 data base.
5-2
-------
Table 5.1-2 Source Composition Profiles from NFRAQS
PNO
N001
N002
N003
N004
N005
N006
N007
N008
N009
N010
N011
N012
N013
N014
N015
N016
N017
N018
N019
N020
N021
N022
N023
N024
N025
N026
N027
N028
N029
N030
N031
N032
N033
N034
N035
N036
N037
N038
N039
N040
N041
N042
N043
N044
N045
N046
N047
N048
N049
N050
N051
Mnemonic
NWNSP
NWLP
NWHP
NWNSPI
NWNSPlmC
NWNSPIpC
NWNSP2
NWNSP2mC
NWNSP2pC
NWSM
NWSMmC
NWSMpC
NWHD
NWHDmC
NWHDpC
NWLCPI
NWLCP2
NWLCP3
NWLCPC
NWL2PI
NWL2P2
NWL2P3
NWL2C
NWHCPI
NWHCP2
NWHCP3
NWHCPC
NWHIPI
NWHIP2
NWHIP3
NWfflC
NWnSPI
NWnSP2
NWnSP3
NWnSPC
NWSCPI
NWSCP2
NWSCP3
NWSCPC
NWSaPI
NWSaP2
NWSaP3
NWSaPC
NWLDCPI
NWLDCP2
NWLDCP3
NWLDCPC
NWHDc
NWHDOc
NMc
NMAHa
Size
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
Type
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Individual
Individual
Individual
Individual
Composite
Composite
Composite
Composite
Individual
Individual
Individual
Individual
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Average
Description
Winter, non-smoker, Phase 1 minus Phase 2, L2, ML1, Ml, M2, M3, HI
Winter, non-smoker, Phase 1 - Phase 2, L2, ML1, ML2
Winter, non-smoker, Phase 1 - Phase 2, Ml, M3, HI
Winter, non-smoker, Phase 1 -UPI, MLIP1, MIP1, M2P1, M3P1, fflPl
Winter, non-smoker, Phase 1 - UPI, MLIP1, MIP1, M2P1, M3P1, fflPl minus backup carbon
Winter, non-smoker, Phase 1 - L2P1, MLIP1, MIP1, M2P1, M3P1, fflPl plus backup carbon
Winter, non-smoker, Phase 2-L1P2, L2P2, MLIP2, MIP2, M2P2, M3P2, HIP2, H2P2
Winter, non-smoker, Phase 2 - L1P2, L2P2, MLIP2, MIP2, M2P2, M3P2, HIP2, H2P2 minus backup carbon
Winter, non-smoker, Phase 2 - LIPI, L2P2, MLIP2, MIP2, M2P2, M3P2, HIP2, H2P2 plus backup carbon
Winter, smokers, S2P1, S2P2, S2P3, S3P1, S3P2, S3P3
Winter, smokers, S2P1, S2P2, S2P3, S3P1, S3P2, S3P3 minus backup carbon
Winter, smokers, S2P1, S2P2, S2P3, S3P1, S3P2, S3P3 plus backup carbon
Winter, heavy-duty diesel, runs 2-15
Winter, heavy-duty diesel, runs 2-15 minus backup carbon
Winter, heavy-duty diesel, runs 2-15 plus backup carbon
Winter, low emitter, phase 1 - UPI, MLIP1, M2P1
Winter, low emitter, phase 2 - L2P2, MLIP2, M2P2
Winter, low emitter, phase 3 - L2P3, MLIP3, M2P3
Winter, low emitter, FTP composite - L2PC, MLIPC, M2PC
Winter, low emitter, phase 1 - L2PI
Winter, low emitter, phase 2 - L2P2
Winter, low emitter, phase 3 - L2P3
Winter, low emitter, FTP composite - L2PC
Winter, high emitter, phasel - M1P1, M3P1, FflPI
Winter, high emitter, phase 2 - MIP2, M3P2, HIP2
Winter, high emitter, phase 3 - Ml P3, M3P3, HI P3
Winter, high emitter, FTP composite - MIPC, M3PC, HI PC
Winter, high emitter, phase 1 , HTPI
Winter, high emitter, phase 2, HIP2
Winter, high emitter, phase 3, HTP3
Winter, high emitter, FTP composite, FflPC
Winter, non-smoker, phase 1, UPI, MLIP1, ML2P1, M2P1, MIP1, M3P1, fflPl
Winter, non-smoker, phase 2, L2P2, MLIP2, ML2P2, M2P2, MIP2, M3P2, HIP2
Winter, non-smoker, phase 3, L2P3, MLIP3, ML2P3, M2P3, MIP3, M3P3, HIP3
Winter, non-smoker, FT? composite, L2PC, MLIPC, ML2PC, M2PC, MIPC, M3PC, FflPC
Winter, smoker, phase 1, SIP1, S2P1, S3P1
Winter, smoker, phase 2, SIP2, S2P2, S3P2
Winter, smoker, phase 3, SIP3, S2P3, S3P3
Winter, smoker, FTP composite, SIPC, S2PC, S3PC
Winter, smoker, phase 1, S2P1, S3P1
Winter, smoker, phase 2, S2P2, S3P2
Winter, smoker, phase 3, S2P3, S3P3
Winter, smoker, FTP composite, S2PC, S3PC
Winter, light-duty diesel, phase 1, LDIPI, LD2P1, LD3P1, LD4P1, LD5P1
Winter, light-duty diesel, phase 2, LDIP2, LD2P2, LD3P2, LD4P2, LD5P2
Winter, light-duty diesel, phase 3, LDIP3, LD2P3, LD3P3, LD4P3, LD5P3
Winter, light-duty diesel, FTP composite, LDIPC, LD2PC, LD3PC, LD4PC, LD5PC
Winter, heavy-duty diesel, runs 2-15 (all)
Winter, heavy-duty diesel, runs 2,5,8,10,16,17,24,32 (others)
Composite of NMAHa, NMCH, NMCCa, and NMCK
3 replicate samples, automated charbroiler, hamburger, samples MAHI, 2, and 3
5-3
-------
Table 5.1-2 (cont.) Source Composition Profiles from NFRAQS
PNO
N052
N053
N054
N055
N056
N057
N058
N059
N060
N061
N062
N063
N064
N065
N066
N067
N068
N069
N070
N071
N072
N073
N074
N075
N076
N077
N078
N079
N080
N081
N135
N136
N137
N138
N139
N140
N141
N142
N143
N144
N145
N146
N147
N148
N149
N150
N151
N152
N153
N154
Mnemonic
NMCH
NMCCa
NMCK
NWFSc
NWFGPDa
NWFEND
NWFGAMD
NWFGBD
NWFHc
NWFEHD
NWFGHD
NWFGOD
NWFGOW
NWFGDD
NWSHc
NWSHDHH
NWSHDHL
NWSHDLH
NWSHDLL
NWSODHLa
NWSOWHL
NRDC
NRDOI
NRD02
NRD03
NRD04
NRD05
AMSUL
AMBSUL
AMNIT
NSLCPI
NSLCP2
NSLCP3
NSLCPC
NSMCP1
NSMCP2
NSMCP3
NSMCPC
NSHCPI
NSHCP2
NSHCP3
NSHCPC
NSSCPI
NSSCP2
NSSCP3
NSSCPC
NSLDCP1
NSLDCP2
NSLDCP3
NSLDCPC
Size
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
Type
Individual
Average
Individual
Composite
Average
Individual
Individual
Individual
Composite
Individual
Individual
Individual
Individual
Individual
Composite
Individual
Individual
Individual
Individual
Individual
Individual
Composite
Individual
Individual
Individual
Individual
Individual
Calculated
Calculated
Calculated
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Description
Charbroiled hamburger, sample MCLH
Charbroiled chicken w/skin, samples MCC1 and MCC2
Charbroiled steak, sample MCKI
Fireplace burning soft woods
Fireplace, Pine, samples WFGPD1, 2 and 3 with Grate, Dry
Fireplace, pinon, sample WFENDI, Empty, Dry
Fireplace, Apple/Mesquite, sample WFGAMDI, with Grate, Dry
Fireplace, Bundled wood, sample WFGBDI, with Grate, Dry
Fireplace burning hard woods
Fireplace, mixed Hardwood, sample WFEHDI, Empty, Dry
Fireplace, mixed Hardwood, sample WFGHD2, with Grate, Dry
Fireplace, Oak, sample WFGODI, with Grate, Dry
Fireplace, Oak, sample WFGOWI, with Grate, Wet
Fireplace, Duraflame, sample WFGDDI, with Grate, Dry
Woodstove burning hardwood
Woodstove, mixed hardwood, sample WSHDHHI, Dry, High fuel, High burn
Woodstove, mixed hardwood, sample WSHDHLI, Dry, High fuel,
Woodstove, mixed hardwood, sample WSHDLLH, Dry, Low fuel,
Woodstove, mixed hardwood, sample WSHDLLI, Dry, Low fuel,
Woodstove, Oak, sample WSODHLI, Dry, High fuel, Low burn
Woodstove, Oak, WSOWHL2, Wet, High fuel, Low bum
Composite roaddust, NRDOI to 05
Jewell, w of Kendall on 10/25/96, sample 717
Kipling at Federal on 3/3/97, sample 818
Kipling at Federal on 3/2/97, sample 819
Speer, Bannock to 1 1th on 12/21/96, sample 831
Jewell, w of Kendall on 12/24/96, sample 800
Secondary ammonium sulfate
Secondary ammonium bisulfate
Secondary ammonium nitrate
Summer, Light-Duty, GasoSummer, LineL2Pl
Summer, Light-Duty, GasoSummer, LineL2P2
Summer, Light-Duty, GasoSummer, LineL2P3
Summer, Light-Duty, GasoSummer, LineL2PC
Summer, Light-Duty, GasoSummer, LineMIPl
Summer, Light-Duty, Gasosummer, LineM I P2
Summer, Light-Duty, GasoSummer, LineMIPS
Summer, Light-Duty, GasoSummer, LineMIPC
Summer, Light-Duty, GasoSummer, LineLUPI
Summer, Light-Duty, GasoSummer, LineHIP2
Summer, Light-Duty, GasoSummer, LineLHP3
Summer, Light-Duty, GasoSummer, LineLUPC
Summer, Light-Duty, GasoSummer, LineSIPl
Summer, Light-Duty, GasoSummer, LineSIP2
Summer, Light-Duty, GasoSummer, LlneSIPS
Summer, Light-Duty, GasoSummer, LlneSIPC
Summer, Light-Duty, DieselLDIPl
Summer, Light-Duty, DieseILDIP2
Summer, Light-Duty, DleseILDIP3
Summer, Light-Duty, DieselLDIPC
Low burn
High burn
Low bum
5-4
-------
Table 5.1-3
Source Composition Profiles from the 1987 Scenic Denver Study and Other Studies
PNO
N082
N083
N084
N085
N086
N087
N088
N089
N090
N091
N092
N093
N094
N095
N096
N097
N098
N099
N100
N101
N102
N103
N104
N105
N106
N107
N108
N109
N110
Nil I
N112
N113
N114
N115
N116
N117
N118
N119
N120
N121
N122
N123
N124
N125
N126
N127
N128
N129
N130
N131
N132
N133
N134
Mnemonic
BRAKE
TRDST
GPHWYCI
GCYSTC3
GSCRDC4
GRSDMC5
GHWYTC6
GHWSTC7
GPRDVC8
GUADVC9
GAGSLC2
MNDCC
MNDCS
MNDCH
MLCC
MLCS
MLCH
MUCCC
MUCCS
MUCCH
MUOCC
MUOCS
MUOCH
MUCC
MUCS
MUCH
MD50U50S
MD75U25S
MD95U5S
ML50U50S
ML25U75S
ML5U95S
MD5L2U3S
MD7515US
MD851OUS
MD3035US
MD305OUS
PCHKC04
PCOOROI
PCHKG03
PCHKC03
PCHCLCI
PCOALC2
CCRCSCI
WFPLLOI
WFPLH02
WWSIN03
WWSTC04
WWSLF05
WWSSF06
WFIRECI
WSTOVC2
WRWCBC3
Size
B
B
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
Type
Individual
Individual
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Description
radial tire (195/608-15 Toyo, 7200 miles
3 samples. Interstate highway.
2 samples. City street near civic center.
2 samples. Secondary paved roads.
2 samples. Road sanding material.
2 samples. Stapleton Tunnel.
3 samples. Stapleton Tunnel and adjacent dirt.
6 samples. Paved roads (general).
5 samples. Unpaved roads (general).
2 samples. Agricultural soil.
9 samples. Dynamometer diesel, cold start.
9 samples. Dynamometer diesel, cold stabilized.
9 samples. Dynamometer diesel, hot transient.
10 samples. Dynamometer leaded, cold start.
10 samples. Dynamometer leaded, cold stabilized.
9 samples. Dynamometer leaded, hot transient.
8 samples. Dynamometer unleaded, closed - loop catalyst, cold start.
4 samples. Dynamometer unleaded, closed - loop catalyst, cold stabilized.
3 samples. Dynamometer unleaded, closed - loop catalyst, hot transient.
14 samples. Dynamometer unleaded, oxidation catalyst, cold start.
8 samples. Dynamometer unleaded, oxidation catalyst, cold stabilized.
5 samples. Dynamometer unleaded, oxidation catalyst, hot transient.
22 samples. Dynamometer unleaded, both catalyst types, cold start.
12 samples. Dynamometer unleaded, both catalyst types, cold stabilized.
8 samples. Dynamometer unleaded, both catalyst types, hot transient.
21 samples. Dynamometer, 50% diesel, 50% unleaded, cold stabilized
21 samples. Dynamometer, 75% diesel, 25% unleaded, cold stabilized
21 samples. Dynamometer, 95% diesel, 5% unleaded, cold stabilized
22 samples. Dynamometer, 50% leaded, 50% unleaded, cold stabilized
22 samples. Dynamometer, 25% leaded, 75% unleaded, cold stabilized
22 samples. Dynamometer, 5 % leaded, 95% unleaded, cold stabilized
31 samples. Dynamometer, 50% diesel, 20% leaded, 30% unleaded, cold stabilized.
31 samples. Dynamometer, 75% diesel, 15% leaded, 10% unleaded, cold stabilized.
31 samples. Dynamometer, 85% diesel, 10% leaded, 5% unleaded, cold stabilized.
31 samples. Dynamometer, 30% diesel, 35% leaded, 35% unleaded, cold stabilized.
31 samples. Dynamometer, 30% diesel, 50% leaded, 20% unleaded, cold stabilized.
Cherokee Pow. Pt, #4 boiler burning coal, mech. collector., elec. precipitator, wet scrubber
Adolph Coors Co., #5 boiler burning coal and brewery sludge, wet scrubber.
Cherokee Pow. Pt., #3 boiler burning natural gas, no control equipment.
Cherokee Pow. Pt., #3 boiler burning coal, bag house.
Composite, PCHKC03 & PCHKC04, boilers burning coal.
Composite, PCHKC03, PCHKC04 & PCOOROI, boilers burning coal.
Composite, 3 samples. Fluidized Catalytic Cracker (FCC) regenerator.
Fireplace, low burn rate.
Fireplace, high bum rate.
Fireplace insert (treated as woodstove).
Woodstove, thermostatically controlled.
Woodstove, large firebox.
Woodstove, small firebox.
Composite of two fireplace tests.
Composite of four woodstove tests.
Composite of all six tests.
5-5
-------
5.2 Initial Source Contribution Estimates
Initial CMB tests were performed to select a default combination of source profiles and
fitting species for the ambient data. These tests were performed to apportion carbon, as this was
a major focus of theNFRAQS. A preliminary set of source profiles consisting of at least one
source profile from each source category was applied to NFRAQS Winter 97 organic speciated
measurements. These initial apportionments were calculated in CMB's batch mode. No attempt
was made to manually improve the apportionment. These tests were used to select a default set
of fitting species and examine the sensitivity of total carbon and PM2.5 apportionment to
alternative source profiles within source categories. Tables 5.2-1 and 5.2-2 show the results of
the sensitivity tests for meat cooking and wood combustion profiles, and motor vehicle profiles,
respectively, for the sample collected from the Welby sampling site on 01/17/97 at 0600 to 1200
MST. This sample had the highest PM2.5 and total carbon concentration during the study and
showed detectable contributions from all major sources in the initial CMB tests.
The range of the carbonaceous fraction explained by alternative individual wood stove
profiles (using hardwoods) was 5.7% to 21.7% of parti culate carbon. Moisture content is the
primary reason for the variability in apportionment. The composite profile that was used in the
NFRAQS CMB as the default profile for this source category (NWSHc2) gives an average
contribution of 14.0%.
Fireplace combustion profiles (using softwoods) yield contributions of 1.2% carbon for
NWFGPDa and 4.4% for NWFEND. The composite of these two profiles (WFSc) gives a
carbon contribution that is equal to the variability in the apportionments for individual softwood
profiles.
With the exception of NWFGFID, the individual fireplace/hardwood profiles yield similar
apportionments in the range of 0.7% to 0.8% carbon. The NWFGFID apportionment is a
statistical outlier, and is excluded from the composite profile (NWFHc) for this source category.
This composite profile gives an average apportionment of 0.8% carbon, which is similar to all
other profiles within the composite. Apple and mesquite are both hardwoods, and their profile
yields a carbon apportionment consistent with the majority of the hardwood profiles.
Synthetic log yields the highest apportionment among the fireplace profiles, with an
average of 47% carbon. This profile is composed of an abundance of elemental carbon, with a
lack of other key "marker" compounds. The high elemental carbon content of the synthetic log
causes it to be collinear with the diesel profiles, which is why the apportionment for this profile
is so high.
Several wood combustion profiles were used together in order to examine the potential
for collinearity among subcategories of wood combustion. The softwood composite (NWFSc)
gives similar apportionments with all alternative hardwood profiles regardless of the type of
appliance, wood stove or fireplace. This softwood composite profile is not collinear with any
other wood combustion profiles. Using the fireplace/hardwood composite profile with the wood
stove/hardwood composite profile results in negative source contributions for the
fireplace/hardwood profile. This fireplace/hardwood profile also causes an overall increase in
the average predicted apportionment for wood stove. This indicates collinearity between the
profiles for hardwood combustion in fireplaces and wood stoves. Because the contribution to
5-6
-------
Table 5.2-la Sensitivity of Total Carbon Apportionment to Alternative Wood Combustion Profiles (Welby,
01/17/97 at 0600 to 1200 MST)
Test
Concentration of TC (ug/m3)
R-squared
Chi-squared
Percent mass
Absolute Contribution (ugTC/m3)
LDGV, cold start
LDGV, hot stabilized
LDGV, high particle emitter
Diesel Exhaust
Meat composite
Road or geologic dust
Coal power stations
Fireplace, softwood composite
Wood stove hardwood composite
Wood stove hardwood
Wood stove oak
Wood stove wet oak
Fireplace, pine average
Fireplace, pinion-no grate
Fireplace, apple/mesquite
Fireplace, bundled wood
Fireplace, hardwood composite
Test
Concentration of TC (ug/m3)
R-squared
Chi-squared
Percent mass
Absolute Contribution (uz£C/m3)
LDGV, cold start
LDGV, hot stabilized
LDGV, high particle emitter
Diesel Exhaust
Meat composite
Road or geologic dust
Coal power stations
Fireplace, softwood composite
Wood stove hardwood composite
Fireplace, apple/mesquite
Fireplace, bundled wood
Fireplace, hardwood composite
Fireplace, hardwood-no grate
Fireplace, hardwood
Fireplace, oak
Fireplace, wet oak
Fireplace, synthetic log
Profile
nvnsp
nvnsp2
nvsm
nwhdc
nine
nrdc
pchclcl
nwfsc
nwshc2
nwshdhl
nwsodhla
nwsowhl
nwfgpda
nwfend
nwfgamd
nwfgbd
nwfhc
Profile
nvnsp
nvnsp2
nvsm
nwhdc
nmc
nrdc
pchclcl
nwfsc
nwshc2
nwfgamd
nwfgbd
nwfhc
nwfehd
nwfghd
nwfgod
nwfgow
nwfgdd
Base
21.55±1.38
0.93469
0.48714
99.56697663
3.21 ±0.28
0.80 ±0.05
5.23 ±0.21
7. 03 ±0.28
1.94 ±0.1 5
0.56±0.18
0.00 ±0.00
-0.33 ±0.11
3.01 ±0.14
Base
21.55±1.38
0.93469
0.48714
99.56697663
3.21 ±0.28
0.80 ±0.05
5.23 ±0.21
7. 03 ±0.28
1.94 ±0.1 5
0.56±0.18
0.00 ±0.00
-0.33 ±0.11
3.01 ±0.14
W01
21. 55 ±1.38
0.93289
0.49703
99.40744659
3.24 ± 0.28
0.76 ±0.05
5.65 ± 0.23
6.92 ± 0.27
1.98 ±0.16
0.57 ±0.18
0.00 ± 0.00
2.32 ±0.11
W10
21. 55 ±1.38
0.9299
0.51275
98.91832322
3. 47 ±0.30
0.67 ± 0.04
7. 36 ±0.30
6.88 ± 0.27
2.18±0.17
0.58 ±0.19
0.00 ± 0.00
0.17 ±0.01
W02
21.55±1.38
0.92899
0.52859
99.32327338
3.14±0.27
0.71 ±0.05
6.72 ±0.27
7. 02 ±0.28
2.02±0.16
0.57 ±0.18
0.00 ±0.00
1.22 ±0.08
Wll
21.55±1.38
0.93048
0.51143
98.90180411
3.34 ±0.29
0.68 ± 0.04
7. 50 ±0.30
6.93 ± 0.27
2.15±0.17
0.58 ±0.18
0.00 ±0.00
0.13 ±0.01
W03
21 .55 ±1.38
0.93299
0.49595
99.14931232
3. 46 ±0.30
0.71 ±0.05
5. 53 ±0.22
6.87 ±0.27
2.03 ±0.16
0.57 ±0.18
0.00 ±0.00
2.20 ±0.15
W12
21 .55 ±1.38
0.93243
0.49398
98.81614604
3. 47 ±0.30
0.67 ±0.04
7. 30 ±0.30
6.88 ±0.27
2.22 ±0.18
0.58 ±0.19
0.00 ±0.00
0.17 ±0.01
W04
21. 55 ±1.38
0.93631
0.46708
99.51277911
3.82 ±0.33
0.81 ±0.05
3.07 ±0.12
6.45 ± 0.26
2. 07 ±0.16
0.55 ±0.18
0.00 ±0.00
4.68 ±0.30
W13
21. 55 ±1.38
0.93403
0.48274
98.79851328
3.44 ±0.30
0.67 ± 0.04
7. 37 ±0.30
6.89 ± 0.27
2.19±0.17
0.59±0.19
0.00 ±0.00
0.15 ±0.01
W05
21.55±1.38
0.92211
0.57092
99.08973217
3. 44 ±0.30
0.67 ± 0.04
7.35 ±0.30
6.86 ±0.27
2.19±0.17
0.59±0.19
0.00 ±0.00
0.26 ±0.02
W14
21.55±1.38
0.93187
0.54105
101.3780927
2.96 ±0.26
0.82 ±0.05
4.49 ±0.18
-0.50 ±0.02
3.31 ±0.26
0.60 ±0.19
0.00 ±0.00
10.18 ±0.60
W06
21. 55 ±1.38
0.92392
0.56866
99.23200067
3.31 ±0.29
0.64 ± 0.04
7. 33 ±0.30
6.41 ±0.25
2.17 ±0.17
0.59 ±0.19
0.00 ± 0.00
0.94 ± 0.05
W15
21. 55 ±1.38
0.93396
0.49936
99.12935947
3. 39 ±0.29
0.72 ±0.05
6.34 ±0.26
6.97 ±0.28
2.09 ±0.17
0.57 ±0.18
0.00 ± 0.00
-0.24 ±0.08
1.43 ±0.07
0.10 ±0.01
W07
21.55±1.38
0.93393
0.48505
98.82333834
3.43 ±0.30
0.67 ± 0.04
7.41 ±0.30
6.89 ±0.27
2.18±0.17
0.57 ±0.18
0.00 ±0.00
0.14±0.01
W16
21.55±1.38
0.93392
0.50149
98.99405126
3.42 ±0.30
0.69 ±0.05
6.69 ±0.27
6.91 ±0.27
2.12±0.17
0.56±0.18
0.00 ±0.00
-0.09 ±0.03
0.92 ±0.04
O.lOiO.Ol
W08
21. 55 ±1.38
0.93264
0.49967
98.85841825
3. 24 ±0.28
0.69 ±0.04
7. 18 ±0.29
6.86 ±0.27
2.10±0.17
0.58 ±0.18
0.00 ±0.00
0.66 ±0.04
W17
21. 55 ±1.38
0.93938
0.49558
100.7689274
3. 04 ±0.26
0.87 ±0.06
3. 50 ±0.14
1.12±0.04
2.92 ±0.23
0.58 ±0.18
0.00 ±0.00
-0.52±0.17
2.20 ±0.10
8. 00 ±0.47
W09
21. 55 ±1.38
0.93451
0.47904
98.82203909
3.40 ±0.30
0.68 ± 0.04
7.38 ±0.30
6.91 ±0.27
2.17±0.17
0.58 ±0.19
0.00 ±0.00
0.18 ±0.01
W18
21. 55 ±1.38
0.93508
0.48035
98.81582122
3.42 ±0.30
0.68 ± 0.04
7.41 ±0.30
6. 94 ±0.27
2.18±0.17
0.58 ±0.19
0.00 ±0.00
-0.10 ±0.03
0.19±0.01
5-7
-------
Table 5.2-lb Sensitivity of Total Carbon Apportionment to Alternative Meat Cooking Profiles
(Welby, 01/17/97 at 0600 to 1200 MST)
Test
Concentration of TC (ug/m3)
R-squared
Chi- squared
Percent mass
Absolute Contribution (ugTC/m3)
LDGV, cold start
LDGV, hot stabilized
LDGV, high particle emitter
Diesel Exhaust
Fireplace, softwood composite
Wood stove hardwood composite
Road and geologic dust
Coal power station
Meat composite
Hamburger, automated charbroiler
Hamburger, under-fired charbroiler
Chicken, under-fired charbroiler
Steak, under-fired charbroiler
Test
cone
rsquar
chisquar
pcmass
nvnsp
nvnsp2
nvsm
nwhdc
nwfsc
nwshc2
nrdc
pchclcl
nmc
nmaha
nmch
nmcca
nmck
Base
21.55± 1.38
0.93
0.49
99.6
3.21 ±0.28
0.80 ±0.05
5. 23 ±0.21
7.03 ±0.28
-0.33 ±0.11
3.01 ±0.14
0.56 ±0.18
0.00 ±0.00
1.94±0.15
M01
21.55± 1.38
0.93
0.50
99.9
3.23 ±0.28
0.82 ±0.05
4.51 ±0.18
7.08 ±0.28
-0.32±0.11
2. 93 ±0.14
0.55 ±0.17
0.00 ±0.00
2.72±0.18
M02
21.55± 1.38
0.93
0.49
99.6
3.40 ±0.29
0.82 ±0.05
4. 45 ±0.18
6.96 ±0.28
-0.33 ±0.11
3.04 ±0.14
0.55±0.18
0.00 ±0.00
2. 58 ±0.17
M03
21.55± 1.38
0.94
0.48
99.5
3.09 ±0.27
0.78 ±0.05
5. 99 ±0.24
7.09 ±0.28
-0.33 ±0.11
3. 02 ±0.14
0.57 ±0.18
0.00 ±0.00
1.24 ±0.09
M04
21.55± 1.38
0.93
0.49
99.5
3.19±0.28
0.80 ±0.05
5. 33 ±0.22
7.01 ±0.28
-0.34±0.11
3. 05 ±0.14
0.57 ±0.18
0.00 ±0.00
1.84±0.12
5-
-------
Table 5.2-2a Sensitivity of Total Carbon Apportionment to Alternative Cold-Start Profiles
(Welby, 01/17/97 at 0600 to 1200 MST)
Concentration ofTC(ug/m3)
R-square
Chi-square
Percent mass
Absolute Contribution fug/m3*)
LDGV (cold start, 1)
LDGV (cold start, l,ml)
LDGV (cold start, l,ml,m)
LDGV (cold start, l,ml,m)
LDGV (cold start, l,ml,m,h)
LDGV (cold start, m,h)
LDGV (hot stabilized, l,ml,m,h)
LDGV (phase 1.2&3, s)
HD diesel
Meat cooking
Wood (fireplace, softwood)
Wood (woodstove, hardwood)
Road dust/geological
Coal-fired power station
Concentration ofTC(ug/m3)
R-square
Chi-square
Percent mass
Absolute Contribution ('ug/mS'l
LDGV (cold start, h)
LDGV (cold start, s,h)
LDGV (cold start, s)
LDGV (phase l,l,ml,m,h)
LDGV (phase l,s,h)
LDGV (phase l,s)
LDGV (hot stabilized, l,ml,m,h)
LDGV (phase 1,2&3, s)
HD diesel
Meat cooking
Wood (fireplace, softwood)
Wood (woodstove, hardwood)
Road dust/geological
Coal-fired power station
Profile
nvcs!2
nvlp
nvcslm
nvcslml
nvnsp
nvhp
nvnsp2
nvsm
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
Profile
nvcshl
nvcssh
nvcss
nvnsp 1
nvplsh
nvpls
nvnsp2
nvsm
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
Base
21. 6± 1.4
0.96
0.27
99.1
3. 50 ±0.30
0.68 ±0.04
6. 54 ±0.27
6.80 ±0.27
2.17±0.17
0.07 ±0.02
1.08 ±0.05
0.50±0.16
0.00 ±0.00
CS6
21. 6± 1.4
0.96
0.34
102.9
3. 08 ±0.12
0.91 ±0.06
10. 02 ±0.41
5. 92 ±0.23
0.83 ±0.07
0.23 ±0.07
0.61 ±0.03
0.56±0.18
0.01 ±0.01
CS1
21. 6± 1.4
0.95
0.47
102.1
1.37±0.11
0.58 ±0.04
9. 78 ±0.40
7.80 ±0.31
0.91 ±0.07
0.12 ±0.04
0.86 ±0.04
0.56±0.18
0.02 ±0.02
CS7
21. 6± 1.4
0.96
0.28
102.8
8. 99 ±0.44
0.63 ±0.04
6. 58 ±0.27
5. 26 ±0.21
-0.51 ±0.04
0.25 ±0.08
0.41 ±0.02
0.54±0.17
0.00 ±0.00
CS2
21.6±1.4
0.95
0.38
100.3
1.61 ±0.18
0.67 ±0.04
8. 88 ±0.36
7.75 ±0.31
1.19±0.09
0.16 ±0.05
0.81 ±0.04
0.55 ±0.17
0.01 ±0.01
CSS
21.6±1.4
0.96
0.28
100.0
12.61 ±0.48
0.54 ±0.04
2. 05 ±0.08
6. 50 ±0.26
-1.32±0.10
0.29 ±0.10
0.30 ±0.01
0.55 ±0.18
0.01 ±0.01
CS3
21.6±1.4
0.96
0.31
99.5
2. 67 ±0.25
0.69 ±0.04
7.09 ±0.29
7.40 ±0.29
1.93 ±0.15
0.09 ±0.03
1.06 ±0.05
0.51 ±0.16
0.01 ±0.01
CS9
21.6±1.4
0.97
0.21
98.6
7.49 ±0.31
4.99 ±0.20
4.62 ±0.18
2. 52 ±0.20
0.03 ±0.01
1.14±0.05
0.47 ±0.15
0.00 ±0.00
CS4
21. 6± 1.4
0.96
0.31
99.6
2. 13 ±0.24
0.66 ±0.04
7.61 ±0.31
7.45 ± 0.30
1.97±0.16
0.10 ±0.03
1.05 ±0.05
0.51 ±0.16
0.01 ±0.01
CS10
21. 6± 1.4
0.96
0.28
103.8
9.71 ±0.43
0.66 ±0.04
8. 36 ±0.34
4.86±0.19
-2.04±0.16
0.40 ±0.13
-0.15 ±0.01
0.57±0.18
0.00 ±0.00
CSS
21. 6± 1.4
0.96
0.29
99.0
5. 37 ±0.24
0.80 ±0.05
5.61 ±0.23
5. 99 ±0.24
1.99±0.16
0.09 ±0.03
0.95 ±0.04
0.54±0.17
0.01 ±0.01
CS11
21. 6± 1.4
0.96
0.25
101.4
16. 83 ±0.62
0.67 ±0.04
4.14±0.17
4.93 ±0.20
.4.43 ± 0.35
0.76 ±0.25
-1.64±0.08
0.59±0.19
0.00 ±0.00
5-9
-------
Table 5.2-2b Sensitivity of Total Carbon Apportionment to Alternative Hot-Stabilized and High
Particle Emitter Profiles (Welby, 01/17/97 at 0600 to 1200 MST)
Concentration ofTC(ug/m3)
R-square
Chi-square
Percent mass
Absolute Contribution fug/m3)
LDGV (cold start, l,ml,m,h)
LDGV (phase2, 1)
LDGV (phase2, l,ml,m,h)
LDGV (phase2, h)
LDGV (phase2, h,s)
LDGV (phase2, s)
LDGV (cold & phase2, s)
LDGV(phasel23, s)
HD diesel
Meat cooking
Wood (fireplace, softwood)
Wood (woodstove, hardwood)
Road dust/geological
Coal-fired power station
Concentration ofTC(ug/m3)
R-square
Chi-square
Percent mass
Absolute Contribution (ue/m31
LDGV (cold start, l,ml,m,h)
LDGV (phase2, l,ml,m,h)
LDGV (cold & phase3, s)
LDGV (cold & phase23, s)
LDGV(phasel2, s)
LDGV(phasel23, h,s)
LDGV(phasel23, h,s)
LDGV(phasel23, s)
HD diesel
Meat cooking
Wood (fireplace, softwood)
Wood (woodstove, hardwood)
Road dust/geological
Coal-fired power station
Profile
nvnsp
nvhslc
nvnsp2
nvhshc
nvp2sh
nvp2s
nvcshs
nvsm
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
Profile
nvnsp
nvnsp2
nvscsp2
nvcshsa
nvplhs
nvsh
nvplhsa
nvs
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
Base
21.6±1.4
0.96
0.27
99.1
3. 50 ±0.30
0.68 ± 0.04
6.54 ±0.27
6. 80 ±0.27
2. 17 ±0.17
0.07 ±0.02
1.08 ±0.05
0.50 ±0.16
0.00 ± 0.00
HE4
21.6±1.4
0.96
0.27
93.3
4.95 ±0.43
0.75 ±0.05
1.03 ±0.04
6.20 ±0.25
5. 22 ±0.41
-0.10 ±0.03
1.59 ±0.07
0.47 ±0.15
0.00 ± 0.00
HS1
21.6±1.4
0.96
0.29
100.9
5.60 ±0.49
0.87 ±0.07
6.36 ±0.26
5. 32 ±0.21
2.03 ±0.16
0.07 ±0.02
1.01 ±0.05
0.48 ±0.15
0.00 ±0.00
HE5
21.6±1.4
0.97
0.18
95.8
0.12 ±0.01
0.66 ± 0.04
11.76 ±0.45
5.39 ±0.21
1.21 ±0.10
0.12 ±0.04
0.87 ±0.04
0.51 ±0.16
0.00 ±0.00
HS2
21.6± 1.4
0.97
0.20
99.2
2. 74 ±0.24
2.42 ±0.15
6.39 ±0.26
5. 99 ±0.24
2.17 ±0.17
0.09 ±0.03
1.06 ±0.05
0.51 ±0.16
0.01 ±0.01
HE6
21.6 ± 1.4
0.97
0.16
94.6
0.13 ±0.01
0.71 ± 0.05
11.27 ±0.43
5.50 ±0.22
1.26 ±0.10
0.14 ±0.05
0.81 ± 0.04
0.56 ±0.18
0.01 ±0.01
HE1
21.6± 1.4
0.97
0.18
92.6
2.90 ±0.25
-0.94 ± 0.06
8. 83 ±0.39
4.67 ±0.19
2.76 ±0.22
0.03 ± 0.01
1.19 ±0.06
0.50 ±0.16
0.00 ±0.00
HE7
21.6 ± 1.4
0.97
0.14
94.6
0.20 ±0.02
-0.60 ± 0.04
13. 51 ±0.62
4.01 ±0.16
1.71 ±0.14
0.07 ± 0.02
1.01 ±0.05
0.48 ±0.15
0.00 ±0.00
HE2
21.6 ± 1.4
0.96
0.26
93.2
5.60 ± 0.49
0.78 ±0.05
0.13 ±0.01
5.97 ± 0.24
5.61 ±0.44
-0.07 ± 0.02
1.59 ±0.07
0.47 ±0.15
0.00 ±0.00
HE8
21.6 ± 1.4
0.97
0.15
97.0
-0.04 ±0.00
0.75 ±0.05
12.29 ±0.47
5.49 ± 0.22
0.91 ±0.07
0.14 ±0.05
0.81 ±0.04
0.54 ±0.17
0.00 ±0.00
HE3
21.6 ± 1.4
0.96
0.19
93.0
0.34 ±0.03
0.60 ±0.04
10.65 ±0.43
5. 43 ±0.22
1.49 ±0.12
0.12 ±0.04
0.89 ±0.04
0.53 ±0.17
0.00 ± 0.00
HE9
21.6 ± 1.4
0.97
0.18
97.9
1.44 ±0.12
0.70 ±0.05
9.81 ±0.38
6.31 ±0.25
1.31 ±0.10
0.12 ±0.04
0.90 ±0.04
0.52 ±0.17
0.01 ± 0.01
5- 10
-------
Table 5.2-2c Sensitivity of Total Carbon Apportionment to Fitting Species (Welby, 01/17/97 at 0600
to 1200 MST)
Concentration ofTC ((ig/m3)
R-square
Chi-square
Percent mass
Absolute Contribution (ng/m3)
LDGV, cold start
LDGV, hot stabilized
LDGV, high particle emitter
Diesel Exhaust
Meat Cooking
Wood combustion, softwood
Wood combustion, hardwood
Road dust/geological
Coal-fired power station
Test
Concentration of TC ((ig/m3)
R-squared
Chi-squared
Percent mass
Absolute Contribution CngTC/m3)
LDGV, cold start
LDGV, hot stabilized
LDGV, high particle emitter
Diesel Exhaust
Meat composite
Fireplace, softwood composite
Wood stove hardwood composite
Road and geologic dust
Coal power station
Brake wear
Tire wear
Profile
nvnsp
nvnsp2
nvsm
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
Profile
nvnsp
nvnsp2
nvsm
nwhdc
nmc
nwfsc
nwshc2
nrdc
pchclcl
brake
trdst
Base
21.6 ± 1.4
0.96
0.27
99.1
3.50 ±0.30
0.68 ±0.04
6.54 ±0.27
6.80 ±0.27
2.17 ±0.17
0.07 ±0.02
1.08 ±0.05
0.50 ±0.16
0.00 ± 0.00
no MeO-Phenols
21.6 ±1.4
0.97
0.27
99.2
3.29 ±0.29
0.65 ± 0.04
7.20 ± 0.29
6.31 ±0.25
2.15±0.17
1.73 ±0.57
-0.45 ± 0.02
0.51 ±0.16
0.00 ±0.00
noOC
21.6 ± 1.4
0.96
0.25
80.5
4.62 ± 0.40
0.82 ±0.05
1.49 ±0.06
6.55 ±0.26
2.22 ±0.18
0.06 ±0.02
1.12 ±0.05
0.48 ±0.15
0.00 ±0.00
no spec organics
21.6 ±1.4
1.00
0.04
-10.65 ±0.93
11.67 ±0.76
47.54 ± 1.93
-73.95 ±2.93
-31. 50 ±2.49
223.95 ±73. 84
1.7977E+308
0.57±0.18
-0.03 ±0.03
no EC
21. 6 ± 1.4
0.96
0.27
83.2
3.38 ±0.29
0.70 ±0.05
6.67 ±0.27
2. 50 ±0.10
3.03 ± 0.24
0.08 ±0.03
1.06 ±0.05
0.51 ±0.16
0.00 ±0.00
with PAH (g)
21.6 ± 1.4
0.88
0.58
108.8
12.19 ± 1.06
-0.34 ±0.02
5.61 ±0.23
3.33 ±0.13
2.06 ±0.16
0.09 ±0.03
0.01 ±0.00
0.51 ±0.16
-0.01 ±0.01
no OC & EC
21. 6 ± 1.4
0.95
0.26
72.9
4.14 ±0.36
0.79 ±0.05
3.28 ±0.13
2.83 ±0.11
3.01 ±0.24
0.07 ±0.02
1.11 ±0.05
0.49 ±0.16
0.00 ±0.00
with brake
21.55 ± 1.38
0.93502
0.5028
99.54957589
3.09 ±0.27
0.79 ±0.05
5.77 ±0.23
7.04 ±0.28
1.45 ±0.12
-0.33 ±0.11
3.04 ±0.14
0.52 ±0.17
-0.01 ±0.01
0.10 ±0.02
no Hopa & Stera
21.6 ± 1.4
0.97
0.31
99.4
3.41 ±0.30
0.70 ±0.05
6.56 ±0.27
6.88 ±0.27
2.23 ±0.18
0.07 ±0.02
1.06 ±0.05
0.50 ±0.16
0.00 ±0.00
with trdst
21.55 ± 1.38
0.93493
0.49487
99.36452475
3.61 ±0.31
0.82 ±0.05
3. 83 ±0.16
6.08 ±0.24
1.64±0.13
-0.34±0.11
3.08 ±0.14
0.54±0.17
0.00 ±0.00
2.14 ±0.43
no Lact & Stero
21.6 ± 1.4
0.96
0.27
99.9
3.73 ± 0.32
0.74 ±0.05
4. 72 ±0.19
6.92 ±0.27
3.80 ±0.30
0.06 ±0.02
1.06 ±0.05
0.48 ±0.15
0.00 ±0.00
5- 11
-------
ambient carbon predicted from the fireplace/hardwood profile is negligible (0.5%), the wood
stove/hardwood profile was selected as the default hardwood profile.
The fractions of ambient fine particles attributed to five alternative meat cooking profiles
range from 5.7% to 12.6% of total ambient PM25 carbon with an average of 9.7%. This average
is consistent with the apportionment results from the composite of all of these meat profiles
(nmc), which was 9.0%. Because each of the alternative meat cooking profiles is collinear with
each other, the composite profile, NMc, was selected as the default meat cooking profile.
Sensitivity tests were performed to examine the effect of alternative LDGV cold start
profiles (Table 5.2-2a) and alternative LDGV hot stabilized and high particle emitter profiles
(Tale 5.2-2b) on the apportionment using a common set of default profiles for non-vehicular
sources. The base case represents the set of default profiles used in the NFRAQS "extended
species" CMB runs.
Each of the alternative LDGV cold start profiles was used individually with the default
set of profiles. Tests CS1 to CSS in Table 5.2-2 show the results for alternative incremental cold
start profiles (i.e., FTP Phase 1 minus Phase 3), and are arranged according to increasing
composite emission rates. The apportionment of total PM2.5 carbon ranges from 6.2% for
incremental cold starts for low emitters to a high of 59% for visible smokers. The increase in
contribution of cold starts with the inclusion of visible smokers is mostly at the expense of the
high particle emitter category, which is comprised of visible smoking vehicles in both cold start
and hot stabilized modes. Among the alternative incremental cold start profiles for non-smoking
vehicles, the range in apportionment is 6.2% to 25.2%. The default profile for this category
(nvnsp), which is an average of all non-smoking vehicles, gives an apportionment of 16.4% for
cold start emissions.
In general, the apportionment (SCE) for other carbon sources varies less with alternative
cold start profiles. The corresponding ranges in LDGV hot stabilized emissions, LDGV high
particle emitters, diesel exhaust, meat cooking and wood combustion are 2.7% to 4.1%, 26.3% to
45.2%, 27% to 36%, 3.7% to 10.2%, 3.8% to 5.3%, respectively. In comparison to the default
incremental cold start profile, the corresponding Phase 1 profile yields 35.2% contribution with
zero hot stabilized emissions and lower contributions for LDGV high particle emitters and diesel
exhaust. Compared to the profile for Phase 1, the incremental cold start is more distinguishable
(chemically) from high emitters and non-smoking hot stabilized emissions.
All non-smoker Phase 2 profiles give lower carbon contributions than any of the Phase 1
profiles, regardless of the emitter category. With the exception of one sample, all of the
alternative smoker profiles give about the same apportionment regardless of phase. The
differences in apportionment between Phases 1 and 2 are within 15% and even lower with Phase
3. While the relative apportionment among the three spark-ignition profiles (i.e., Phase 1, Phase
2, and smoker) vary with alternative profiles for smokers, the amount apportioned to the default
meat cooking profile is relatively insensitive to the use of alternative smoker profiles.
The initial CMB tests done for heavy-duty diesel exhaust showed that apportionment
(SCE) of diesel exhaust is relatively insensitive to the differences in elemental carbon (EC)
abundances in the profile. Diesel exhaust with relative abundances of EC of 86% and 63% show
differences in apportionment (SCE) of total carbon of 10%. In addition to EC, the CMB
5- 12
-------
sensitivity matrix shows that particulate PAHs, especially methyl- and dimethyl-phenanthrene,
have strong influence on the apportionment.
Table 5.2-2c shows the effects of using alternative sets of fitting species on the
apportionment. The default fitting species included inorganic species and particle-phase organic
species (particulate PAH, methoxy phenols, lactones, sterols, hopanes, and steranes) with R/U
ratios between -2 and +2. The tests included the following changes to the default set of fitting
species: 1) no organic carbon (OC); 2) no EC; 3) no OC and EC; 4) no hopanes and steranes; 5)
no lactones and sterols; 6) no phenols; 7) no organic species; and 8) addition of gas-phase PAHs.
Removing OC from the set of fitting species reduces the apportionment of LDGV high
particle emitter from 30.6% to 8.6%. However the PERCENT MASS is reduced from 99.1% to
80.5%, and a slight increase results in the other three vehicle exhaust profiles. Although organic
carbon is the major component in the LDGV high emitter profile, it is incorrect to assume that
the apportionment of this source is keyed simply on organic carbon. This statement would be
true if the smoker profile is the only organic carbon source used in the fit. There are seven other
sources of organic carbon that are used in the apportionment for NFRAQS.
The CMB sensitivity matrices show that organic carbon has the greatest influence on
apportionment of LDGV high particle emitter and has little effect on the other sources of organic
carbon. However, contributions of LDGV cold starts are strongly influenced by phenanthrene,
fluorene, methylfluorene isomers, and heavy PAHs. Hot-stabilized particulate emissions from
non-smokers are influenced by methylfluorene isomers, and methylphenanthrene and
dimethylphenanthrene isomers. Lactones influence the apportionment for meat cooking and
guiacols and syringols affect apportionments of softwood and hardwood combustion,
respectively. Diesel exhaust and road dust are predominantly influenced by elemental carbon
and crustal elements, respectively. The attributions of carbon to the other seven sources are all
influenced by species other than organic carbon.
Removing EC from the set of fitting species reduces the apportionment of diesel exhaust
from 31.9% to 13.9%. As with the previous case with OC, the PERCENT MASS is reduced (to
83.2%) and contributions of LDGV high particle emitter and meat cooking increase by small
margins. Removing both OC and EC cause decreases in contributions of both LDGV high
particle emitter and diesel exhaust and correspondingly larger decrease in the percent of mass
attributed (72.9%).
Removing hopanes and steranes has negligible effect on the apportionment. The
effective variance weighted solutions in CMB8 gives greater influence to chemical species with
lower uncertainty in both source and ambient measurements. As ambient levels approach
detection levels for marker compounds, as was the case for these species, their influence on the
CMB fit decreases.
According to the CMB sensitivity matrix, the expected marker species (i.e., lactones and
sterols) have the greatest influence on the apportionment of meat cooking. Removing them
resulted in an unexpected increase in apportionment for meat cooking and offsetting decrease in
the prediction for LDGV high particle emitters. The long-chain g-lactones and cholesterol are
considered "marker" species for meat cooking. However, motor vehicles were also found to
emit "lactones", which raised questions regarding their proper identification. Since some of the
light-duty gasoline and heavy-duty diesel vehicle exhaust samples, when analyzed by electron
5- 13
-------
impact/mass spectrometry (EI/MS), showed an m/z 85 ion (characteristic of lactones) at the same
gas chromatographic retention time that correspond to some of the lactones, they were
re-analyzed by chemical ionization/mass spectrometry (CI/MS) in order to confirm their identity.
None of the light-duty gasoline vehicle exhaust samples contains detectable amounts of
lactones. All four lactones were found in heavy-duty diesel exhaust samples, but in much lower
amounts than quantified by EI/MS. Six ambient samples (three from the Welby site and three
from the Brighton site) were also re-analyzed for lactones using the CI/MS technique. All
lactones, previously quantified by EI/MS in the ambient samples, were also identified and
quantified by the CI/MS technique. Since all ambient samples were quantified by the EI/MS
technique using the m/z 85 ion, some of the compounds emitted by motor vehicles could
"artifactually" contribute to the intensity of this ion. Thus, lactones were retained in all motor
vehicle profiles, whether or not they were truly lactones. This situation is analogous to the
application of "organic" carbon in CMB which contain a variety of unidentified organic
compounds. The lactones are also imprecisely quantified, as reflected in measurement
uncertainty estimates that approach or exceed 30% of their concentrations. Owing to these large
uncertainties, the CMB8 effective variance solution reduces their influence on the apportionment
relative to more precisely measured components. When lactones are removed as fitting species,
however, the standard error of source contribution estimates for meat cooking increases.
Removing methoxyphenols from the fit results in a shift of the attribution from hardwood
to softwood. This is expected since there are other markers for softwood (e.g., retene and 1,7
dimethylphenanthrene) while syringols are the primary markers for hardwoods. Including
gas-phase PAH in the set of fitting species causes significant increase in the cold start
contribution and an overestimation of mass because gas-phase organic species are generally not
correlated with particulate mass.
Removing all organic species from the fit results in complete breakdown of CMB fit due
to significant collinearity that results among the subcategories of motor vehicle as well as wood
and meat combustion profiles. Tests were conducted using "conventional" species, which
include only total organic carbon, elemental carbon, inorganic ions (nitrate, sulfate, ammonium),
and elements. LDGV profiles were combined into one composite profile and meat cooking and
wood combustion were combined for the conventional CMB. The effective variance weighted
solutions in CMB8 uses all available chemical measurements, not just "tracer" species, and gives
greater influence to chemical species with lower uncertainty in both source and ambient
measurements.
As ambient levels approach detection levels for marker compounds, as was the case for
many of the samples from Brighton, their influence on the CMB fit decreases. This situation
could lead to overestimation of LDGV high particle emitter and underestimation of other sources
of organic carbon, and could explain the differences that exist between the "extended" and
"conventional" CMB results for Brighton. The low-concentration samples from Brighton
contain less than 1 or 2 |ig/m3 of total carbon and are associated with transport from the north
rather than from the Denver area, and are not representative of the urban source mix. In contrast,
the samples from Welby are more representative of the urban source mix, and typically contain
levels of total particulate carbon that allow for quantitative determination of organic markers.
The "extended" and "conventional" CMB results are consistent with each other, and
comparisons with isotopic carbon measurements are more consistent for this site than for
5- 14
-------
Brighton. The "conventional" CMB results for CAMP and Highlands indicate that results for
Welby are likely representative of the Denver urban area.
Because tire wear and brake wear were not tested as part the NFRAQS study, chemical
composition profiles were developed for these source from data published by Hildemann et al.
(1991) and Rogge et al. (1993c). Hildemann et al. (1991) reports elemental data for tire wear
and brake wear. Organic data were obtained from Rogge et al (1993b). The profiles were
derived by converting the emission data into weight fractions normalized to total measured fine
particle mass. A nominal uncertainty of 20% was applied to the weight fractions. Table 5.2-2c
shows the average source contribution estimates for tire dust and brake wear for the sample that
was examined as part of the sensitivity tests of alternative source profiles. The contributions of
brake wear is negligible (<1%). Tire wear is about 10% of the total carbon. However, the
apportionment of PM2.5 has high standard errors, which indicates that there is a high degree of
uncertainty and/or collinearity in these profiles. These profiles were not used in the final
NFRAQS apportionment.
5.3 Model Outputs and Performance Measures
Nearly 1,000 individual CMB calculations were performed for NFRAQS in various
sensitivity tests. Apportionment of the NFRAQS ambient data included 132 apportionments
using the "extended" data sets that include specific organic compounds measured at the Welby
and Brighton sites of 6-hour or 12-hour durations for the Winter 97 samples. The CMB model
was also applied to 150 24-hour average "conventional" data sets from all seven NFRAQS
Winter 97 sites that included the elemental, ionic, and elemental/organic carbon concentrations
that are most commonly measured on source and receptor samples. This allowed for comparison
of source contribution estimates derived from the "extended" and "conventional" CMB
calculations for the Welby and Brighton data.
For these source apportionments, R-SQUARE typically exceeded 0.9 and CHI-SQUARE
values typically ranged from 0.3 and 0.6. PERCENT MASS values for organic carbon,
elemental carbon, total carbon, and PM2.5 were within one standard deviation of 100% most of
the time.
5.4 Deviations from Model Assumptions
Assumptions 1 and 2 of the CMB model specify that the compositions of source
emissions are constant over the period of ambient and source sampling, and that chemical
species do not react with one another. Once released into the atmosphere, primary emissions are
subjected to dispersion and transport and, at the same time, to various physical and chemical
processes that determine their ultimate environmental fate. Primary emissions from motor
vehicles, residential wood combustion, meat cooking, etc., are complex mixtures containing
thousands of organic and inorganic constituents in the gas and particulate phases.
These compounds have different chemical reactivities and are removed by dry and wet
deposition processes at varying rates. Some of the gaseous species, by a series of chemical
transformations, are converted into particles, forming secondary aerosol. Sulfates and nitrates
are the most common secondary particles, though a fraction of organic carbon can also result
from VOC via atmospheric reactions.
5- 15
-------
While the mechanisms and pathways for inorganic secondary particles are fairly well
known, those for secondary organic aerosols are not well understood. Hundreds of precursors
are involved in these reactions, and the rates at which these particles form are highly dependent
on the concentrations of other pollutants and meteorological variables. Organic compounds
present in the gas phase undergo atmospheric transformation through reactions with reactive
gaseous species such as OH* radicals, NO3* radicals, or O3.
Atmospheric lifetimes can be estimated for several organic compounds in direct
gas-phase emissions due to known tropospheric chemical removal reactions (Atkinson, 1988).
These lifetimes (i.e., the time for the compound to decay to 1/e or 37% of its original
concentration) are calculated from the corresponding measured reaction rate constants and the
average ambient concentration of the tropospheric species involved. Although the individual
rate constants are known to a reasonable degree of accuracy (in general, to within a factor of
two), the tropospheric concentrations of these key reactive species are much more uncertain.
For example, the ambient concentrations of OH* radicals at any given time and/or
location are uncertain to a factor of at least five, and more likely ten (Atkinson, 1988). The
tropospheric diurnally and annually averaged OH* radical concentrations are more certain, to
possibly a factor of two. For this reason, calculated lifetimes are approximate only for those
reactive species concentrations that are listed in the footnotes. However, these data permit one
to estimate the contribution of each of these atmospheric reactions to the overall removal rates of
most pollutants from the atmosphere. The major atmospheric loss process for most of the direct
emission constituents is by daytime reaction with OH* radicals.
For some pollutants, photolysis, reactions with ozone, and reactions with NO3 radicals
during nighttime hours are also important removal routes. For alkanes, the atmospheric lifetimes
calculated from the corresponding measured reaction rate constant and the average ambient
concentration of OH* radicals, ranges from -19 days for propane (C3H8) to ~1 day for
n-pentadecane (C15H32). For aromatic hydrocarbons, lifetimes range from 18 days for benzene to
a few hours for methylnaphthalenes (assuming average 12-hour daylight OH* radical
concentration of 1 x 106 molecule/cm3).
Secondary organic compounds in particulate matter include aliphatic acids, aromatic
acids, nitro aromatics, carbonyls, esters, phenols, and aliphatic nitrates (Grosjean, 1992;
Grosjean and Seinfeld, 1989). However, these compounds can also be present in primary
emissions (e.g., Rogge, 1993), thus they are not unique tracers for atmospheric transformation
processes.
It has been reported that, in the presence of NOX, the OH* radical reactions with
fluoranthene and pyrene present in the gas phase lead to the formation of specific nitroarene
isomers different from those present in the direct emissions (Arey etal., 1986, 1989; Atkinson et
a/., 1990; Zielinska et a/., 1990). The nighttime reactions with NO3 radicals lead to the same
product as OH* radical reactions which form nitro-fluoranthene and nitro-pyrene isomers
(Zielinska etal., 1986). In contrast, the electrophilic nitration reaction of fluoranthene, or
pyrene, involving an NO2+ ion, produces mainly 3-nitrofluoranthene from fluoranthene and
1-nitropyrene from pyrene and these isomers are present in direct emissions from combustion
sources.
5- 16
-------
In order to assess the importance of atmospheric formation of secondary aerosol, the
concentration of 2-nitrofluoranthene and 2-nitropyrene was measured during the NFRAQS at the
Welby site (Watson etal., 1998). Although 2-nitrofluoranthene and 2-nitropyrene are present in
low amounts in daytime samples collected at the Welby site, their concentrations are not
significantly different during sunny and cloudy days. Secondary organic aerosol formation was
therefore considered negligible during the NFRAQS Winter 97 intensive operating period.
With respect to Assumption 3 involving the inclusion of all source types, it appears from
the PERCENT MASS performance measures that all of the significant contributors have been
included in most of the CMBs.
With respect to Assumption 4 concerning number of species and number of sources, 85
species and up to 11 source profiles were used in each calculation. The number of chemical
species always exceeded the number of source types.
With respect to Assumption 5 concerning collinearity, this was largely eliminated by the
inclusion of specific organic species in the extended data sets. These were sufficient to separate
contributions from fireplaces, wood stoves, meat cooking, diesel exhaust, gasoline smoker
exhaust, gasoline cold-start exhaust, and gasoline hot-stabilized exhaust. For the conventional
data sets, wood burning and meat cooking were collinear and the gasoline exhaust contributors
were collinear. Source-types for suspended dust, secondary ammonium sulfate and ammonium
nitrate, and specific coal-fired power station contributions could not be resolved. Profiles for
other industrial point sources were lacking, and their primary particle contributions could not be
explicitly estimated by CMB.
The effects of deviations from Assumption 6 on the randomness and normality of
measurement errors remain to be studied. For this study, all of the CMB assumptions are met to
the extent that the source contribution estimates can be considered valid.
5.5 Identification and Correction of Model Input Errors
Many Level III validation deficiencies in the processing, formatting, compositing, and
reporting of ambient concentration and source profile measurements were identified and corrected or
flagged as a result of CMBS source apportionment. Corrections and flags have been incorporated
into the NFRAQS data base, and the results presented by Watson et al. (1998) reflect these changes.
Some chemical species concentrations were physically unreasonable, as indicated by large
CHI-SQUARE values with a large R/U value for the related species. The trimethylnaphthalenes and
biphenyls consistently showed large R/U values. The reason is not readily apparent. In these cases
the suspect species was removed from the fit. In general, the CMB modeling was robust enough
that, when performance measures were within acceptable ranges around target values, there was
little effect of suspect concentrations on the source contribution estimates.
5.6 Consistency and Stability of Source Contributions
The source contribution estimates and the statistics and diagnostic information were
reviewed to determine the validity of the initial model results. The analysis was repeated by
eliminating source profiles that gave negative source contribution estimates or standard errors that
exceed the source contribution estimates. The good agreement between the calculated source
5- 17
-------
contributions and the measured ambient concentrations indicate that all major source types were
included in the calculations, that ambient and source profile measurements are reasonably accurate,
and that the source profiles are reasonably representative of actual emissions.
5.7 Consistency with Other Simulations and Data Analyses
Watson et al. (1998) demonstrate that the CMB8 source contribution estimates for carbon
and PM2.5 are consistent with other NFRAQS data analysis and simulations. Figure 5.7-1 shows the
CMB source apportionment at the Welby site. This chart shows substantial discrepancies with
respect to the inventory5 in Table 5.1-1, in which diesel exhaust emissions are four times gasoline
exhaust emissions. The cold start and high emitter portions of the gasoline exhaust reverse these
proportions. These discrepancies should be further investigated.
Ammonium
Nitrate
25%
Coal Power
Station
2%
Gas
Cold Start
12%
Ammonium
Sulfate
10%
Dust
16%
Hardwood
Burning
3%
Gas
Normal
3%
Gas
High Emitter
13%
Diesel
10%
Meat Cooking
4%
Softwood Burning
2%
Figure 5.7-1. Average PM2.5 source contributions at the Welby site near Denver, CO during the
winter of 1996-97.
5This SIP-planning inventory was compiled externally to the NFRAQS.
5- 18
-------
6. EXAMPLE OF APPLICATION AND VALIDATION FOR VOC
This example of VOC source apportionment is taken from the NARSTO-NE ozone study
that took place between June 1 and August 31, 1995 in the region between Washington DC and
Boston, MA (Fujita et a/., 1998). While several different types of VOC samples were taken, this
example focuses on hourly measurements acquired with automated gas chromatographs at PAMS.
6.1 Model Applicability
The data set includes hydrocarbon measurements for 55 species6 measured hourly at eight
PAMS sites in six source areas (E. Hartford, CT; McMillan Reservoir, DC; Chicopee, MA; Lynn,
MA; Lake Clifton, MD; Bronx, NY) and two downwind (Type 1, 3, or 4) PAMS sites (Lums Pond,
DE and Rider College, NJ). The speciated hydrocarbon data are hourly measurements by automated
gas chromatographs providing more than 15,000 hydrocarbon samples covering the period from
June 1, 1995 to August 31, 1995. The sites, number of samples, and number of species measured
are sufficient to perform a CMB source apportionment.
6.2 Initial Source Contribution Estimates
Table 6.2-1 lists the mnemonic of the profiles that were considered in this study with short
descriptions. The actual profiles are reported by Fujita et al. (1998) and are included as one of the
CMB8 test data sets. As discussed in Section 3.1.2, the profiles are expressed as mass fractions of
total NMHC. Compounds other than the PAMS target VOCs (Appendix A) that are in the profiles
have been grouped into a category named "others". The 28 species that were used as fitting species
in the CMB analysis are identified in Table 6.2-2 with asterisks. Compounds reported as
"unknown" were grouped into a category named "UNTD".
The PAMS target compounds typically account for about 80% of the ambient hydrocarbons
in urban areas. Although MTBE is a major component in reformulated gasoline and in the exhaust
of vehicles using reformulated gasoline, it was not included in the profiles because MTBE is not
measured in the PAMS program. The source profile data reported in units of ppbC were converted
to g/m3 prior to calculating the mass fractions (expressed as percentages) using species-specific
conversion factors (Section 3.1.2). One-sigma uncertainties were derived from variations among
multiple measurements for a particular source type or a nominal analytical uncertainty of ±15%.
The assigned uncertainties are the larger of the two values.
In urban locations, motor vehicle exhaust and evaporative emissions of gasoline are often the
major sources of hydrocarbon emissions. Composites of dynamometer measurements of vehicles of
varying age and mileage are commonly used to represent fleet-averaged exhaust profiles. For these
profiles to represent the actual fleet-average exhaust near ambient monitoring sites, the fuels in the
dynamometer tests should resemble the fuels used in the study region and the mix of test vehicles
should reflect the relative influence of non-catalyst vehicles or high emitters and catalyst-equipped
normal emitters.
6Note that as of March 1998, the 2-Methyl-l-Pentene was removed; the list is now 54 species (see Appendix AA).
6- 1
-------
Table 6.2-1 VOC Source Profiles for NARSTO-NE CMB
No. Mnemonic Description
1 Tu_TusHD Tuscarora Tunnel, Heavy duty emissions
2 Tu_MchHD Ft. McHenry Tunnel, Heavy duty emissions
3 Exh_CalO Callahan Tunnel emissions with diesel contributions removed.
4 Exh_LinO Lincoln Tunnel emissions with diesel contributions removed.
5 Exh_Call Callahan Tunnel emissions with diesel and 5-10% of running loss contributions removed.
6 Exh_Linl Lincoln Tunnel emissions with diesel and 5—10% of running loss contributions removed.
7 Exh_Cal2 Callahan Tunnel emissions with diesel and 15—25% of running loss contributions removed.
8 Exh_Lin2 Lincoln Tunnel emissions with diesel and 15—25% of running loss contributions removed.
9 WA_Tul Mt. Baker Tunnel emissions with diesel and 5—10% of running loss contributions removed.
10 Tu_Calla Callahan Tunnel emissions
11 Tu_Lin Lincoln Tunnel emissions
12 Tu_TusLD Tuscarora Tunnel, Light duty emissions
13 Tu_MchLD Ft. McHenry Tunnel, Light duty emissions
14 BoCSJTip Tip O'Neill Garage emissions, Boston, cold start
15 ExhSOla Derived from the FTP tests of Sigsby et al.
16 BoglOl Boston liquid gasoline composite.
17 LA_liqGs LA liquid gasoline composite.
18 WA_Liq Washington liquid gasoline composite of 15 samples, weighted by brands and grades.
19 BogvOl Boston headspace vapor composite
20 LA_Hsvap LA headspace vapor composite
21 WA_Vap Washington headspace composite of 15 samples, weighted by brands and grades.
22 COATcomp Composite of various coating emissions, weighted by total emissions.
23 CNG Commercial natural gas
24 GNG Geogenic natural gas
25 LPG Liquified petroleum gas
26 Biogenic Constructed biogenic profile
27 Unid Sum of unidentified species.
Previous studies showed that source attributions between tailpipe and evaporative emissions
from receptor modeling can vary greatly depending on the particular profile chosen for tailpipe
emissions (Harley et al., 1992; Fujita et al., 1994; Pierson et al., 1996). This is because tailpipe
emissions are a mixture of hydrocarbons produced during combustion along with unburned gasoline
resulting from incomplete combustion. In the CMB calculation, liquid gasoline represents the
additional unburned gasoline (due to misfiring and other engine malfunctions) that is not included in
the exhaust profile, plus evaporative emissions from gasoline spillage, hot soaks, and some portion
of resting losses (leaks, permeation). The profile for gasoline headspace vapor is taken to represent
fuel tank vapor losses (e.g., migration of fuel vapor from the canister). Measuring exhaust in on-
road tunnels is one way to obtain a composite profile for a larger mix of vehicles.
While tunnel measurements are reasonable approximations for exhaust profiles of the light-
duty fleet, they also include varying amounts of diesel exhaust and running evaporative losses. The
composite light-duty exhaust profiles that were derived by Fujita etal. (1997a) from measurements
by Gertler et al. (1997a) in the Lincoln and Callahan Tunnels were used in this study.
-2
-------
Table 6.2-2 Measured PAMS Species and CMB Fitting Species
PAMS Species CMB Fit
ETHANE *
ETHENE
ACETYL *
LBUT1E
PROPE
N_PROP *
I_BUTA *
N_BUTA *
T2BUTE
C2BUTE
IPENTA *
PENTE1
N_PENT *
I_PREN *
T2PENE
C2PENE
B2E2M
BU22DM *
CPENTE
P1E4ME
CPENTA *
BU23DM *
PENA2M *
PENA3M *
P1E2ME
N_HEX *
T2HEXE
C2HEXE
MCYPNA *
PEN24M *
BENZE *
CYHEXA *
HEXA2M *
PEN23M *
HEXA3M *
PA224M *
PAMS Species CMB Fit
N_HEPT *
MECYHX *
PA234M *
TOLUE *
HEP2ME *
HEP3ME *
N_OCT *
ETBZ
MP_XYL
STYR
O_XYL
N_NON *
IPRBZ
N_PRBZ
M_ETOL
P_ETOL
BZ135M
O_ETOL
BZ124M
N_DEC *
BZ123M
DETBZ1
DETBZ2
N_UNDE *
UNTO
-------
6.3 Examine Model Outputs and Performance Measures
The tunnel-derived exhaust profiles, uncorrected tunnel profiles, and dynamometer-derived
exhaust profiles were applied to the same ambient samples to determine the sensitivity of the CMB
model to alternative exhaust profiles. Table 6.3-1 shows the effect of alternative vehicle exhaust
profiles on the average source contributions for a set of 65 ambient samples from the PAMS site at
Bronx, NY. Samples for this test were collected during the 0700 to 0800 EDT in the summer of
1995. Each of the ambient samples were apportioned with the diesel profile, TUJVICHHD, plus
twelve alternative gasoline vehicle exhaust profiles (ExhSOla, Exh_CalO, Exh_Call, Exh_Cal2,
Exh_LinO, Exh_Linl, Exh_Lin2, Tu_Calla, Tu_Lin, Tu_Mchld, Tu_Tusld, and Wa_Tul) using
only fitting species that are enriched in diesel and spark-ignition vehicle exhaust (ethene, acetylene,
propene, benzene, nonane, decane, and undecane).
Table 6.3-1 CMB Sensitivity Tests for Vehicle Exhaust Profiles
#of
samples3
65
65
65
65
65
65
65
65
65
65
65
65
TNMOC
(ug/m3)
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
r7
0.88
0.92
0.92
0.92
0.93
0.93
0.93
0.92
0.93
0.88
0.86
0.91
x2
3.27
2.86
2.81
2.75
2.40
2.40
2.28
2.87
2.77
4.83
5.04
3.46
%of
NMHC
65.2
70.5
66.6
61.2
78.8
75.1
68.7
70.5
73.0
66.6
57.5
52.2
HD Profiles
Tu_Mchhd
Tu Mchhd
Tu_Mchhd
Tu_Mchhd
Tu_Mchhd
Tu Mchhd
Tu Mchhd
Tu_Mchhd
Tu Mchhd
Tu Mchhd
Tu_Mchhd
Tu Mchhd
LD Profiles
EXH801A
EXH_CALO
EXH_CAL1
EXH CAL2
EXH_LINO
EXH_LIN1
EXH_LIN2
TU_CALLA
TU_LIN
TU_MCHLD
TUJTUSLD
WA TU1
HD (%)
22.2
22.0
21.8
21.9
21.8
21.6
21.6
19.3
14.5
22.7
24.3
22.8
LD (%)
43.0
48.6
44.8
39.3
57.0
53.5
47.1
51.2
58.5
43.9
33.2
29.3
a Samples collected between 0700 and 0800 EDT at Bronx, NY were used in the test.
Source contribution estimates using alternative gasoline vehicle exhaust profiles range from
50% to 70% of total NMHC. Exhaust profiles for relatively cleaner fleets (e.g., Tuscarora and
Mount Baker Tunnels) yield lower contributions. Exhaust contributions varied by no more than
10% for the three levels of assumed headspace vapor contributions for both Lincoln and Callahan
Tunnels profiles. The profiles corresponding to the maximum level of evaporative correction gave
exhaust contributions about 5% to 6% greater than profiles corresponding to averages between no
correction and maximum correction. Profiles derived from the tunnel measurements at the Lincoln
Tunnel consistently yielded the best model performance.
Table 6.3-2 shows the effect of alternative gasoline profiles on the average source
contributions for the same set of 65 ambient samples from the PAMS sites in Bronx, NY during the
0700 to 0800 EDT sampling period. Use of the vapor profiles for gasoline samples from either
Boston or Los Angeles results in large overestimation of total NMHC. In contrast, the vapor profile
for the Washington samples yield total predicted NMHC contributions that are, on average, about
90% of the observed ambient NMHC. Less than 100% is expected as only vehicle-related source
profiles were included in these sensitivity tests. Adding the other default source profiles does not
-4
-------
Table 6.3-2 CMB Sensitivity Tests for Different Profiles
No.
samples3
65
65
65
65
65
65
65
65
65
65
TNMOX
(Hg/m3)
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
144.2
r2
0.
0.
0.
0.
0.
0.
0.
0.
0.
0,
,80
,86
,82
,82
,86
,82
,80
,86
,82
,89
x2
6.63
4.61
6.03
6.52
4.54
6.30
6.63
4.61
6.03
2,8]
% of
NMHC
124.0
115.7
90.1
139.4
114.6
94.8
124.0
115.7
90.1
109.2
Liq. Gaso.
BOglOl
BOglOl
BOglOl
LA liqGS
LA liqGS
LA liqGS
WA liq
WA liq
WA liq
WA lia
Gaso. Vapor
BOgvOl
LA HsVap
WAJVap
BOgvOl
LA HsVap
WAJVap
BOgvOl
LA HsVap
WA Vap
WA Van
HDD
(%)
21.6
21.6
22.0
19.7
22.1
21.7
21.6
21.6
22.0
7.7
LDD Liquid Gaso.
(%) Gaso. Vapor Coating CNG GNG LPG Biogenic
44.9
43.3
44.6
44.7
42.7
44.8
44.9
43.3
44.6
50.9
0.0
0.0
0.0
53.7
0.0
6.5
0.0
0.0
0.0
5.0
57.4
50.8
23.5
18.5
51.1
21.8
57.4
50.8
23.5
18.0 13.5 8.8 2.9 1.4 0.9
Samples collected between 0700 and 0800 EDT at Bronx, NY were used in the test.
Tu_Mchhd and Exh_Lin2 were used to represent HD and LD, respectively.
significantly alter the contributions among the tailpipe and evaporative emissions for gasoline
vehicles, but reduces the contribution of diesel exhaust from 22.0% to 7.7%. The difference is
assigned to surface coating because decane and undecane are major components of both diesel
exhaust and surface coatings. Because the sensitivity tests shown in Table 6.3-1 indicate that diesel
exhaust is the correct source of the higher molecular weight species at the Bronx site, the surface
coating profiles were not used in the default set of profiles in order to avoid potential for collinearity
between these two profiles.
6.4 Test Deviations from Model Assumptions
Assumptions 1 and 2 of the CMB model specify that the compositions of source emissions
are constant over the period of ambient and source sampling, and that chemical species do not react
with one another. The CMB model was applied to the ten alternative diesel and evaporative
emissions-corrected samples for each tunnel run with diesel exhaust and evaporative emissions as
source profile. The model performance parameters and comparisons of calculated and measured
amounts of total NMHC, isobutane, -butane, and isopentane were examined to determine the level
of evaporative corrections that yield the best fit. These tests showed that the fit deteriorates rapidly
beyond a certain level of assumed headspace vapor contribution. This level is typically 15% to
25%. The predicted vapor contributions do not increase above these levels of assumed vapor
contribution. This is consistent with the expectation since there is a limit to the fractional
contribution of running losses to hydrocarbons mixing ratios in roadway tunnels. Because the
performance parameters for various levels of assumed headspace vapor contributions are similar up
to the level at which the fit deteriorates, three sets of corrected profiles were derived for each tunnel
run. One profile corresponded to no evaporative correction (i.e., only diesel correction), and a
second set of profiles corresponded to the maximum level of evaporative correction before the fit
begins to deteriorate (15% to 20%). In the 3rd profile, a composite of the best fitting diesel corrected
profiles was made for the uncorrected tunnel measurements.
For PAMS sites, the significant contributors to the average 24-hour ambient NMHC are
gasoline vehicle exhaust (40%), gasoline vapor (17%), and diesel exhaust (17%). Natural gas leak
(9%), liquid gasoline (7%), liquefied petroleum gas (4%), and biogenic emissions (4%) are minor
6-5
-------
contributors to NMHC. On average, only 4% of the identified NMHC are unexplained. Three of
the Type 2 sites (Chicopee and Lynn, MA and Bronx, NY) account for the relatively high average
contribution of diesel exhaust. Higher diesel contributions are possible at the Bronx site due to
diesel buses.
However, there are no obvious sources of the high levels of heavy hydrocarbon at Chicopee
and Lynn that result in high diesel contributions at those sites. Contributions of liquid gasoline and
gasoline vapors are also much higher at Chicopee than for a typical Type 2 site. Removing
Chicopee and Lynn from the average, decreases the average contribution of diesel exhaust for the
remaining sites to 12% and increases gasoline vehicle exhaust to 46%.
The significant contributors to the average 24-hour ambient NMHC at downwind PAMS
sites (Lums Pond, DE and Rider College, NJ) are gasoline vehicle exhaust (30%), gasoline vapor
(18%), liquefied petroleum gas (18%), and natural gas (11%). Diesel exhaust (9%), liquid gasoline
(6%), biogenic emissions (5%) are minor contributors to NMHC.
Biogenic emissions are a significantly larger fraction of total NMHC in suburban and rural
areas than in urban area. The contributions of isoprene over a 24-hour average range from 4% to
12% in suburban and rural areas and are less than 2% in urban areas. Because isoprene is emitted
only during daylight hours with peak emission rates occurring at midday, it is one of the larger
sources of NMHC during the day in suburban and rural areas. For the CMB calculations performed
in this study, only species with summertime lifetimes greater than that of toluene (~9 hours) were
used as fitting species. An exception to this is isoprene. It was included as a fitting species despite
its high reactivity because it serves as a marker for biogenic emissions. The source contribution
estimates underestimated the actual source contributions of biogenic emissions, i.e., they provide a
lower limit to biogenic contributions.
The actual contributions of isoprene may be estimated by examining changes between
morning and afternoon samples in the ratios of reactive hydrocarbons (e.g., isomers of xylene) to a
relatively unreactive hydrocarbon (e.g., benzene) from a common source (i.e., vehicle exhaust) as an
indicator of the net fractional loss of reactive hydrocarbon between the two sampling periods. These
ratios are invariant to atmospheric dispersion and include continuous injections of fresh emissions
into the air parcel during its transport to the sampling site. The ratio of afternoon to morning
xylenes^enzene ratios reflects the net fractional loss of xylenes due to atmospheric reactions. This
fractional loss is adjusted to isoprene by applying the ratio of the OH* radical reaction rate constants
for xylenes (18.8) and isoprene (101.0). Based on this approach, Fujita et al. (1997b) estimated that
the actual daytime contributions of isoprene to total NMHC emissions may be 5 to 10 times greater
than CMB estimates.
Unidentified compounds are not considered in the apportionment because a large fraction of
these compounds are not quantified in the PAMS program. These compounds include terpenes and
higher molecular weight aldehydes, which are relatively more abundant in non-urban areas.
6-6
-------
6.5 Identify and Correct Model Input Errors
Substantial validation was done on the profiles and ambient data sets, and no major
discrepancies were found as part of the source apportionment.
6.6 Evaluate Consistency and Stability of Source Contributions
The source contribution estimates and the statistics and diagnostic information were
reviewed to determine the validity of the initial model results. The analysis was repeated by
eliminating source profiles that gave negative source contribution estimates or standard errors that
exceed the source contribution estimates. The good agreement between the calculated source
contributions and the measured ambient concentrations indicate that all major source types were
included in the calculations, that ambient and source profile measurements are reasonably
accurate, and that the source profiles are reasonably representative of actual emissions.
6.7 Determine Consistency with Other Simulations and Data Analyses
Hourly data offers substantial opportunities to evaluate consistency with other analyses,
especially expected diurnal, weekly, and spatial variations in source emissions. Figure 6.7-1
shows the diurnal variations of the absolute source contributions for each source category by day
of the week at the Lynn, MA site. While motor vehicle exhaust contributions generally peak
during morning and afternoon commute periods on weekday, the average contributions are
significantly lower during weekend mornings. These patterns provide confidence in the proper
apportionment of vehicle emissions. The diurnal and day-of-the-week patterns in the liquid
gasoline contributions are essentially identical to motor vehicle exhaust, which suggests that a
large fraction of the liquid gasoline contribution may be associated with tailpipe emissions rather
than evaporative emissions from either vehicle or industrial sources.
The diurnal variations in the contribution of natural gas correlate with diurnal variations in
vertical mixing. This diurnal pattern and lack of day-of-the-week variations are consistent with
constant leakage of natural gas. Liquefied petroleum gas (LPG) generally shows the same
diurnal variations. However, lower contributions for LPG during weekend mornings suggests
some correlation with the vehicle exhaust profile since the latter profile is derived from roadside
ambient measurements.
-7
-------
50
40 -
ma 30 -
^ 20 -
10 -
0
Lynn - HD Exhaust
^^25^^^S^iS3^^^
; —/K :
I
50
40 -
30 -
20 -
10 -
0
00 01 02 03 04 05 06 07 08 09 10 II 12 13 14 15 16 11 18 19 20 21 22 23
Lynn - LD Exhaust
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
25
20 -
m 15 -
a
]? 10 -I
Lynn - Gasoline Vapor
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Lynn - Biogenic
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
H Thursday
• Monday
• Friday
-Tuesday
Saturday
—A Wednesday
• Sunday
Figure 6.7-1. Hourly average VOC source contributions by day of week at Lynn, MA.
-------
Figure 6.7-2 shows the average source contributions to NMHC by wind sector (centered
on N, NE, E, SE, S, SW, W, NW) at the Lynn, MA site for the evening period. Contributions of
gasoline vehicle exhaust are predominantly from the southeast, south, and southwest. In
contrast, the contribution of diesel exhaust is more or less uniform from all directions. This
pattern suggests a very strong local source that dominates the ambient VOC composition near
the sampling site. It also indicates that the source of the heavy hydrocarbons that are ascribed to
diesel exhaust is some source other than diesel vehicles.
Lynn (18-24EDT)
NW
NE
W
• HD Exhaust
LD Exhaust
. . . Liq. Gas.
Gaso. Vap.
CNG
Biogenic
SW
SE
Figure 6.7-2. Wind direction dependence of VOC source contributions at Lynn, MA.
6-9
-------
-------
7. SUMMARY, CONCLUSIONS, AND FUTURE PROSPECTS
This applications and validation protocol has summarized knowledge on using the
Chemical Mass Balance receptor model to determine source contributions to suspended particles
and VOC measured at receptors. It describes new performance measures incorporated into
CMB8 modeling software that facilitate the evaluation of similarity among different profiles.
Examples are given for PM2.5 source apportionment in Denver, CO, and for VOC source
apportionment in the northeast corridor of the U.S. These examples demonstrate how the
applications and validation steps can be used to build confidence in the source apportionment
results.
New measurement methods, especially for organic aerosol and heavy hydrocarbons, will
expand the ability of CMB to better distinguish sources from each other. Organic aerosol
measurements allow cold-starts and high emitting gasoline exhaust to be distinguished from
normal running vehicle exhaust. Initial indications are that emissions inventories do not
adequately account for these emissions. Hourly gas chromatographic data allow the diurnal
cycles and dominant wind directions of VOC sources to be estimated. These add confidence to
the CMB apportionments, as well as elucidating temporal and spatial relationship between
emissions and ambient concentrations.
Collocation of PMzs speciation sites with PAMS VOC sites will allow gas and particle
properties to be used together in a single CMB apportionment. This holds the potential to
provide more accurate source apportionments for a wider variety of chemical components.
CMB8 provides a myriad of options that can be applied to better understand the CMB
source apportionment method. The new collinearity measures need to be better characterized to
provide more specific guidance for their use in practical situations. Furthermore, the
Britt-Luecke algorithm, as implemented in CMB8, has not undergone comprehensive testing. It
is therefore not recommended for inexperienced users. Its inclusion as an option is mainly
intended to provide the opportunity for interested advanced users to perform research
investigations needed to establish its future viability.
This protocol will surely be revised as more and better data become available, and we
gain more experience and skill in applying CMB to source apportionment studies.
7- 1
-------
-------
8. REFERENCES
Allen, A.G., R.M. Harrison and J.W. Erisman (1989). Field measurements of the dissociation of
ammonium nitrate and ammonium chloride aerosols. Atmos. Environ., 23(7): 1591-1599.
Altshuller, A.P. (1983). Natural volatile organic substances and their effect on air quality in the
United States. Atmos. Environ., 17: 2131-2166.
Altshuller, A.P. (1993). Production of aldehydes as primary emissions and from secondary
atmospheric reactions of alkenes and alkanes during the night and early morning hours.
Atmos. Environ., 27A(1): 21-32.
Annegarn, H.J., G.M. Braga Marcazzan, E. Cereda, M. Marchinonni and A. Zucchiatti (1992).
Source Profiles by Unique Ratios (SPUR) Analysis: Determination of source profiles
from receptor-site streaker samples. Atmos. Environ., 26(2): 333-343.
Arey, J.B., B. Zielinska, R. Atkinson, A.M. Winer, T. Ramdahl and J.N. Pitts, Jr. (1986). The
formation of nitro-PAH from the gas-phase reactions of fluoranthene and pyrene with the
OH radical in the presence of NOX. Atmos. Environ., 20: 2339-2345.
Arey, J.B., B. Zielinska, R. Atkinson and S.M. Aschmann, (1989). Nitroarene products from the
gas-phase reactions of volatile poly cyclic aromatic hydrocarbons with OH radical and
N2O5. Int. J. Chem. Kinet., 21: 775-799.
Armstrong, J.A., Russell, P.A., L.E. Sparks and D.C. Drehmel (1980). Tethered balloon
sampling systems for monitoring air pollution. JAPCA, 31: 735-743.
Arnts, R.R. and S.A. Meeks (1981). Biogenic hydrocarbon contribution to the ambient air of
selected areas. Atmos. Environ., 15: 1643-1651.
Arnts, R.R., W.B. Peterson, R.L. Seila and B.W. Gay, Jr. (1982). Estimates of alpha-pinene
emissions from a loblolly pine forest using an atmospheric diffusion model. Atmos.
Environ., 16: 2127-2138.
Aronian, P.P., P.A. Scheff and R.A. Wadden (1989). Wintertime source-reconciliation of
ambient organics. Atmos. Environ., 23(5): 911-920.
Atkinson, R. (1988). Atmospheric transformations of automotive emissions. In Air Pollution,
the Automobile, and Public Health, A.Y. Watson, R.R. Bates, andD. Kennedy, Eds.
National Academy Press (sponsored by The Health Effects Institute), Washington, DC,
pp. 99-132.
Atkinson, R., J.B. Arey, B. Zielinska and S.M. Aschmann (1990). Kinetics and nitro-products of
the gas-phase OH and NO3 radical-initiated reactions of naphthalene-d8 fluoranthene- d10
and pyrene. Int. J. Chem. Kinet., 22: 999-1014.
- 1
-------
Bailey, J.C., K. Gunary, B. Schmidl and M.L. Williams (1990a). Speciated hydrocarbon
emissions from a sample of UK vehicles on the road over a range of Speeds. Sci. Total
Environ., 93: 199-206.
Bailey, J.C., B. Schmidl and M.L. Williams (1990b). Speciated hydrocarbon emissions from
vehicles operated over the normal speed range on the road. Atmos. Environ., 24A: 43-52.
Bailey, J.C. and S. Eggleston. (1993). The contribution of gasoline fueled vehicle exhaust to the
UK speciated hydrocarbon inventory. Sci. Total Environ., 134: 263-271.
Barrefors, G. (1996). Air pollutants in road tunnels. Sci. Total Environ., 189-190:431-436.
Barth, D. (1970). Federal motor vehicle emissions goals for CO, HC, and NOX based on desired
air quality levels. JAPCA, 20(8): 519-527.
Belsley, D.A., E.D. Kuh and R.E. Welsch (1980). Regression Diagnostics: Identifying
Influential Data and Sources of Collinearity. John Wiley and Sons, New York, NY.
Benarie, M.M. (1976). Urban Air Pollution Modeling Without Computers. Prepared by U.S.
Environmental Protection Agency, Research Triangle Park, NC.
Benjamin, M.T., M. Sudol, L. Bloch and A.M. Winer (1996). Low-emitting urban forests: A
taxonomic methodology for assigning isoprene and monoterpene emission rates. Atmos.
Environ., 30(9): 1437-1451.
Benjamin, M.T. and A.M. Winer (1998). Estimating the ozone-forming potential of urban trees
and shrubs. Atmos. Environ., 32(1): 53-68.
Benner, B.A., G.E. Gordon and S.A. Wise (1989). Mobile sources of atmospheric PAH: A
roadway tunnel study. Environ. Sci. Technol, 23: 1269-1278.
Berlin, N., M. Staudt, U. Hansen, G. Seufert, P. Ciccioli, P. Foster, J.L. Fugit and L. Torres
(1997). Diurnal and seasonal course of monterpene emissions from Quercus ilex (L.)
under natural conditions - application of light and temperature algorithms. Atmos.
Environ., 31(S1): 135-144.
Bishop, G.A., S.E. McLaren, D.H. Stedman, W.R. Pierson, R.B. Zweidinger and W.D. Ray
(1996). Method comparisons of vehicle emissions measurements in the Fort McHenry
and Tuscarora mountain tunnels. Atmos. Environ., 30(2): 2307-2316.
Black, F.M., L.E. High and J.M. Lang (1980). Composition of automobile evaporative and
tailpipe hydrocarbon emissions. JAPCA, 30(11): 1216-1221.
Booher, L.E. and B. Janke (1997). Air emissions from petroleum hydrocarbon fires during
controlled burning. J. Am. Ind. Hyg. Assoc., 58(5): 359-365.
-2
-------
Booker, R.L., P.M. Boone and E.S. Macias (1986). Methyl alkanes: A tracer for motor vehicle
emissions? In Transactions, Receptor Methods for Source Apportionment: Real World
Issues and Applications., T.G. Pace, Ed. Air Pollution Control Association, Pittsburgh,
PA, pp. 268-270.
Bowman, P.M., C. Pilinis and J.H. Seinfeld (1995). Ozone and aerosol productivity of reactive
organics. Atmos. Environ., 29(5): 579-590.
Britt, H.I. and R.H. Luecke (1973). The estimation of parameters in nonlinear, implicit models.
Technometrics, 15(2): 233-247.
Brosseau, J. and M. Heitz (1994). Trace gas compound emissions from municipal landfill
sanitary sites. Atmos. Environ., 28(2): 285-294.
Byun, D.W. and R. Dennis (1995). Design artifacts in Eulerian air quality models: Evaluation
of the effects of layer thickness and vertical profile correction on surface ozone
concentrations. Atmos. Environ., 29(1): 105-126.
Calloway, C.P., S.M. Li, J.W. Buchanan and R.K. Stevens (1989). A refinement of the
potassium tracer method for residential wood smoke. Atmos. Environ., 23(1): 67-69.
Cao, X.L., C. Boissard, AJ. Juan, C.N. Hewitt and M. Gallagher (1997). Biogenic emissions of
volatile organic compounds from gorse (Ulex europaeus}: Diurnal emission fluxes at
Kelling Heath, England. J. Geophys. Res., 102(D15): 18903-18916.
Carey, P., and J. Cohen (1980). Comparison of gas phase hydrocarbon emissions from light-
duty gasoline vehicles and light-duty vehicles equipped with diesel engines. Report No.
CTAB/PA/80-5. Prepared by Office of Air, Noise and Radiation, U.S. Environmental
Protection Agency, Ann Arbor, MI.
Carlier, P., H. Hannachi and G. Mouvier (1986). The chemistry of carbonyl compounds in the
atmosphere - A review. Atmos. Environ., 20: 2079-2099.
Carter, W.P.L. (1990). A detailed mechanism for the gas-phase atmospheric reactions of organic
compounds. Atmos. Environ., 24A(3): 481-518.
Carter, W.P.L. and F.W. Lurmann (1991). Evaluation of a detailed gas-phase atmospheric
reaction mechanism using environmental chamber data. Atmos. Environ., 25A(12):
2771-2806.
Cass, G.R. (1981). Sulfate air quality control strategy design. Atmos. Environ., 15(7): 1227-
1249.
Cass, G.R. and GJ. McRae (1981). Source-receptor reconciliation of South Coast Air Basin
particulate air quality data. Report No. ARB-R-82-156. Prepared by California Air
Resources Board, Sacramento, CA.
-------
Cass, G.R. and GJ. McRae (1983). Source-receptor reconciliation of routine air monitoring data
for trace metals: An emission inventory assisted approach. Environ. Sci. Technol,
17(3): 129-139.
Censullo, A.C., D.R. Jones and M.T. Wills (1996). Improvement of speciation profiles for
architectural and industrial maintenance coating operations. Final Report. Prepared for
California Air Resources Board, Sacramento, CA, by California Polytechnic State
University, San Luis Obispo, CA.
Chameides, W.L., R.W. Lindsay, J. Richardson and C.S. Kiang (1988). The role of biogenic
hydrocarbons in urban photochemical smog: Atlanta as a case study. Science, 241:
1473-1475.
Chan, C.C., J.D. Spengler, H. Ozkaynak and M. Lefkopoulou (1991). Commuter exposures to
VOCs in Boston, Massachusetts. JAWMA, 41(12): 1594-1600.
Chang, T.Y., S.W. Modzelewski, J.M. Norbeck and W.R. Pierson (1981). Tunnel air quality and
vehicle emissions. Atmos. Environ., 15: 1011.
Cheng, M.D. and P.K. Hopke (1989). Identification of markers for chemical mass balance
receptor model. Atmos. Environ., 23: 1373-1384.
Chock, D.P. and S.L. Winkler (1992). Exhaust emissions of reformulated gasolines and
methanol fuels based on the working data sets of the Auto/Oil Air Quality Improvement
Research Program. In Transactions, PM10 Standards andNontraditional Particulate
Source Control, J.C. Chow and D.M. Ono, Eds. Air & Waste Management Association,
Pittsburgh, PA, pp. 899-923.
Chock, D.P., S.L. Winkler, T.Y. Chang, SJ. Rudy and Z.K. Shen (1994). Urban ozone air
quality impact of emissions from vehicles using reformulated gasolines and M85. Atmos.
Environ., 28(17): 2777-2788.
Chow, J.C. (1985). A composite modeling approach to assess air pollution source/receptor
relationships. Sc.D.Dissertation, Harvard University, Cambridge, MA.
Chow, J.C., J.G. Watson, C.A. Frazier, R.T. Egami, A. Goodrich and C. Ralph (1988). Spatial
and temporal source contributions to PM10 and PM2 5 in Reno, NV. In Transactions,
PM10: Implementation of Standards, C.V. Mathai and D.H. Stonefield, Eds. Air
Pollution Control Association, Pittsburgh, PA, pp. 438-457.
Chow, J.C. and J.G. Watson (1989). Summary of particulate data bases for receptor modeling in
the United States. In Transactions, Receptor Models in Air Resources Management, J.G.
Watson, Ed. Air & Waste Management Association, Pittsburgh, PA, pp. 108-133.
Chow, J.C., J.G. Watson, R.T. Egami, C.A. Frazier, Z. Lu, A. Goodrich and A. Bird (1990).
Evaluation of regenerative-air vacuum street sweeping on geological contributions to
PM10. JAWMA, 40(8): 1134-1142.
8-4
-------
Chow, J.C., J.G. Watson, L.W. Richards, D.L. Haase, C. McDade, D.L. Dietrich, D. Moon and
C.S. Sloane (1991). The 1989-90 Phoenix PM10 Study, Volume II: Source
apportionment. Report No. DRI 8931.6F2. Prepared for Arizona Department of
Environmental Quality, Phoenix, AZ, by Desert Research Institute, Reno, NV.
Chow, J.C. and J.G. Watson (1992). Fugitive emissions add to air pollution. Environ. Protect.,
3:26-31.
Chow, J.C., J.G. Watson, D.H. Lowenthal, P.A. Solomon, K.L. Magliano, S.D. Ziman and L.W.
Richards (1992a). PM10 source apportionment in California's San Joaquin Valley.
Atmos. Environ., 26A(18): 3335-3354.
Chow, J.C., J.G. Watson, D.H. Lowenthal, C.A. Frazier, B.A. Hinsvark, L.C. Pritchett and G.R.
Neuroth (1992b). Wintertime PM10 and PM2 5 mass and chemical compositions in
Tucson, Arizona. In Transactions, PM10 Standards and Non-Traditional Particulate
Source Controls, J.C. Chow and D.M. Ono, Eds. Air and Waste Management
Association, Pittsburgh, PA, pp. 231-243.
Chow, J.C., C.S. Liu, J.C. Cassmassi, J.G. Watson, Z. Lu and L.C. Pritchett (1992c). A
neighborhood-scale study of PM10 source contributions in Rubidoux, California. Atmos.
Environ., 26A(4): 693-706.
Chow, J.C., J.G. Watson, L.C. Pritchett, W.R. Pierson, C.A. Frazier and R.G. Purcell (1993).
The DRI Thermal/Optical Reflectance carbon analysis system: Description, evaluation
and applications in U.S. air quality studies. Atmos. Environ., 27A(8): 1185-1201.
Chow, J.C. and J.G. Watson (1994a). Contemporary source profiles for geological material and
motor vehicle emissions. Report No. DRI 2625.2F. Prepared for U.S. EPA, Office of
Air Quality Planning and Standards, Research Triangle Park, NC, by Desert Research
Institute, Reno, NV.
Chow, J.C. and J.G. Watson (1994b). Guidelines for PM10 sampling and analysis applicable to
receptor modeling. Report No. EPA-452/R-94-009. Prepared for U.S. EPA, Office of
Air Quality Planning and Standards, Research Triangle Park, NC, by Desert Research
Institute, Reno, NV.
Chow, J.C., J.G. Watson, J.E. Houck, L.C. Pritchett, C.F. Rogers, C.A. Frazier, R.T. Egami and
B.M. Ball (1994). A laboratory resuspension chamber to measure fugitive dust size
distributions and chemical compositions. Atmos. Environ., 28(21): 3463-3481.
Chow, J.C. (1995). Critical review: Measurement methods to determine compliance with
ambient air quality standards for suspended particles. JAWMA, 45(5): 320-382.
Chow, J.C., D. Fairley, J.G. Watson, R. de Mandel, E.M. Fujita, D.H. Lowenthal, Z. Lu, C.A.
Frazier, G. Long and J. Cordova (1995). Source apportionment of wintertime PM10 at
San Jose, CA. J. Environ. Eng., 21: 378-387.
-5
-------
Chow, J.C., J.G. Watson, D.H. Lowenthal and RJ. Countess (1996a). Sources and chemistry of
PM10 aerosol in Santa Barbara County, CA. Atmos. Environ., 30(9): 1489-1499.
Chow, J.C., J.G. Watson, Z. Lu, D.H. Lowenthal, C.A. Frazier, P.A. Solomon, R.H.Thuillier and
K.L. Magliano (1996b). Descriptive analysis of PM2 5 and PM10 at regionally
representative locations during SJVAQS/AUSPEX. Atmos. Environ., 30(12): 2079-2112.
Chow, J.C. and J.G. Watson (1997a). Imperial Valley/Mexicali Cross Border PM10 Transport
Study. Report No. 4692.1D1. Prepared for U.S. Environmental Protection Agency,
Region IX, San Francisco, CA, by Desert Research Institute, Reno, NV.
Chow, J.C. and J.G. Watson (1997b). Fugitive dust and other source contributions to PM10 in
Nevada's Las Vegas Valley. Report No. 4039. IF. Prepared for Clark County
Department of Comprehensive Planning, Las Vegas, NV, by Desert Research Institute,
Reno, NV.
Chow, J.C., J.G. Watson, F. Divita, Jr., M.C. Green, B.A. Bates, W. Jones, G. Torres, R. Fischer
and D. Lam (1997). Cross-border transport of PM10 between California and Mexico.
Presentation at the AWMA international conference on Particulate Matter: Health and
Regulatory Issues in Pittsburgh, PA, Apr. 4, 1995. Air & Waste Management
Association, Pittsburgh, PA.
Chow, J.C. and J.G. Watson (1998). Guideline on speciated particulate monitoring. Prepared
for U.S. EPA, Research Triangle Park, NC, by Desert Research Institute, Reno, NV.
www.dri.edu.
Chow, J.C., B. Zielinska, J.G. Watson, E.M. Fujita, H.W. Richards, W.D. Neff, D. Dietrich and
S.V. Hering (1998). Northern Front Range Air Quality Study. Volume A: Ambient
measurements. Prepared for Colorado State University, Cooperative Institute for
Research in the Atmosphere, Fort Collins, CO, by Desert Research Institute, Reno, NV.
http: //charon.cira.colostate.edu/DRIFinal/ZipFiles/.
Chuang, J.C., S.V. Hannan andN.K. Wilson (1987). Field comparison of polyurethane foam
and XAD-2 resin for air sampling for polynuclear aromatic hydrocarbons. Environ. Sci.
Technol, 21(8): 798-804.
Chung, J., R.A. Wadden and P.A. Scheff (1996). Development of ozone-precursor relationships
using VOC receptor modeling. Atmos. Environ., 30(18): 3167-3179.
Ciccioli, P., A. Cecinato, E.M. Brancaleoni, A. Bracchetti and M. Frattoni (1995). Polar volatile
organic compounds (VOC) of natural origin as precursors of ozone. Environmental
Monitoring and Assessment, 31(1-2): 211-218.
Ciccioli, P., C. Fabozzi, E.M. Brancaleoni, A. Cecinato, M. Frattoni, S. Cieslik, D. Kotzias, G.
Seufert, P. Foster and R. Steinbrecher (1997a). Biogenic emissions from the
Mediterranean pseudosteppe ecosystem present in Castelporziano. Atmos. Environ.,
31(S1): 167-176.
8-6
-------
Ciccioli, P., C. Fabozzi, E.M. Brancaleoni, A. Cecinato, M. Frattoni, F. Loreto, J. Kesselmeier,
L. Schafer, K. Bode, L. Torres and J.L. Fugit (1997b). Use of the isoprene algorithm for
predicting the monoterpene emission from the Mediterranean holm oak Quercus ilex L.:
performance and limits of this approach. J. Geophys. Res., 102(D19): 23319-23328.
Clausen, P.A. and P. Wolkoff (1997). Degradation products of Tenax TA formed during
sampling and thermal desorption analysis: indicators of reactive species indoors. Atmos.
Environ., 31(5): 715-726.
Cofer, W.R., III and R.A. Edahl, Jr. (1986). A new technique for collection, concentration and
determination of gaseous tropospheric formaldehyde. Atmos. Environ., 20: 979-984.
Cohen, M.A., P.B. Ryan, J.D. Spengler, H. Ozkaynak and C. Hayes (1991a). Source-receptor
study of volatile organic compounds and particulate matter in the Kanawha Valley, WV -
I. Methods and descriptive statistics. Atmos. Environ., 25B(1): 79-93.
Cohen, M.A., P.B. Ryan, J.D. Spengler, H. Ozkaynak and C. Hayes (1991b). Source-receptor
study of volatile organic compounds and parti culate matter in the Kanawha Valley, WV -
II. Analysis of factors contributing to VOC and particle exposures. Atmos. Environ.,
25B(1): 95-107.
Conner, T.L., J.F. Collins, W.A. Lonneman and R.L. Seila (1994). Comparison of Atlanta
emissions inventory with ambient data using chemical mass balance receptor modelling..
EPA/A&WMA, Pittsburgh, PA, pp. 147-158.
Conner, T.L., W.A. Lonneman and R.L. Seila (1995). Transportation-related volatile
hydrocarbon source profiles measured in Atlanta. JAWMA, 45(5): 383-394.
Conny, J.M. and L.A. Currie (1996). The isotopic characterization of methane, non-methane
hydrocarbons and formaldehyde in the troposphere. Atmos. Environ., 30(4): 621-628.
Cooper, D.A., K.R. Peterson and D.W. Simpson (1996). Hydrocarbon, PAH and PCB emissions
from ferries: a case study in the Skagerak-Kattegatt-Oresund region. Atmos. Environ.,
30: 2463-2474.
Cooper, J.A. and J.G. Watson (1980). Receptor oriented methods of air particulate source
apportionment. JAPCA, 30(10): 1116-1125.
Cooper, J.A., D.C. Redline, J.R. Sherman, L.M. Valdovinos, W.L. Pollard, L.C. Scavone and
C.R. West (1987). PM10 source composition library for the South Coast Air Basin.
Volume I: Source profile development documentation. Final report. Prepared by South
Coast Air Quality Management District, El Monte, CA.
-7
-------
Cooper, J.A., J.R. Sherman, E. Miller, D.C. Redline, L. Valdonovinos and W.L. Pollard (1988).
CMB source apportionment of PM10 downwind of an oil-fired power plant in Chula
Vista, California. In Transactions, PM10: Implementation of Standards., C.V. Mathai and
D.H. Stonefield, Eds. Air and Waste Management Association, Pittsburgh, PA, pp. 495-
507.
Cooper, J.A., E.A. Miller, D.C. Redline, R.L. Spidell, L.M. Caldwell, R.H. Sarver and B.L.
Tansyy (1989). PM10 source apportionment of Utah Valley winter episodes before,
during, and after closure of the West Orem steel plant. Prepared for Kimball, Parr,
Crokcett and Waddops, Salt Lake City, UT, by NEA, Inc., Beaverton, OR.
Corchnoy, S.B., J.B. Arey and R. Atkinson (1992). Hydrocarbon emissions from twelve urban
shade trees of the Los Angeles, California, air basin. Atmos. Environ., 26B(3): 339-348.
Core, I.E. and I.E. Houck (1987). Pacific Northwest Source Profile Library Sampling and
Analytical Protocols. Prepared by Oregon Department of Environmental Quality,
Portland, OR.
Coulter, C.T. and J.V. Scalco (2005). Chemical Mass Balance Software: EPA-CMB8.2.
Proceedings of A&WMA's 98th Conference & Exhibition; Minneapolis, MN, June 21-
24, 2005.
Currie, L.A., G.A. Klouda and KJ. Voorhees (1984). Atmospheric carbon: The importance of
accelerator mass spectrometry. Nuclear Instruments and Methods in Physics Research,
233 [B5](2): 371-379.
Dannecker, W., B. Schroder and H. Stechmann (1990). Organic and inorganic substances in
highway tunnel exhaust air. Sci. Total Environ., 93: 293-300.
Darley, E.F., F.R. Burleson, E.H. Mateer, J.T. Middleton and V.P. Osterli (1966). Contribution
of burning of agricultural wastes to photochemical air pollution. JAPCA, 16: 685-690.
Das, T.N. (1992). Short communication, Terpenes in a conifer forest: Detection at subparts per
billion level without sample preconcentration. Atmos. Environ., 26A(15): 2853-2858.
deCesar, R.T. and J.A. Cooper (1982). Evaluation of multivariate and chemical mass balance
approaches to aerosol source apportionments using synthetic data and an expanded PACS
data set. In Receptor Models Applied to Contemporary Pollution Problems., P.K. Hopke
and S.L. Dattner, Eds. Air Pollution Control Association, Pittsburgh, PA, pp. 127-140.
deCesar, R.T., S.A. Edgerton, M.A.K. Khalil and R.A. Rasmussen (1985a). Sensitivity analysis
of mass emission inventory for the Southern California Air Quality Study. Prepared for
Electric Power Research Institute, Palo Alto, CA.
deCesar, R.T., S.A. Edgerton, M.A.K. Khalil and R.A. Rasmussen (1985b). Sensitivity analysis
of mass balance receptor modeling: methyl chloride as an indicator of wood smoke.
Chemosphere, 14(10): 1495-1501.
-------
deCesar, R.T., S.A. Edgerton, M.A.K. Khalil and R.A. Rasmussen (1986). A tool for designing
receptor model studies to apportion source impacts with specified precisions. In
Transactions, Receptor Methods for Source Apportionment: Real World Issues and
Applications, T.G. Pace, Ed. Air Pollution Control Association, Pittsburgh, PA, pp. 56-
67.
deNevers, N. and J.R. Morris (1975). Rollback modeling: Basic and modified. JAPCA,25(9):
943-947.
Diehl, J.W., F.W. Finkbeiner and P.P. Di Sanzo (1993). Determination of benzene, toulene,
ethylbenzene, and xylenes in gasolines by gas chromatography/deuterium isotope
dilution Fourier transform infrared spectroscopy. Analytical Chemistry, 65(18): 2493-
2496.
Doskey, P.V., J.A. Porter and P.A. Scheff (1992). Source fingerprints for volatile non-methane
hydrocarbons. JAWMA, 42(11): 1437-1445.
Dresser, A.L. and B.K. Baird (1988). A dispersion and receptor model analysis of the
wintertime PM10 problem in Telluride, Colorado. In Transactions, PM10:
Implementation of Standards, C.V. Mathai and D.H. Stonefield, Eds. Air Pollution
Control Association, Pittsburgh, PA, pp. 458-471.
Duffy, B.L. and P.F. Nelson (1996). Non-methane exhaust composition in the Sydney Harbour
Tunnel: A focus on benzene and 1,3-Butadiene. Atmos. Environ., 30(15): 2759-2768.
Dzubay, T.G., R.K. Stevens, W.D. Balfour, HJ. Williamson, J.A. Cooper, I.E. Core, R.T.
deCesar, E.R. Crutcher, S.L. Dattner, B.L. Davis, S.L. Heisler, JJ. Shah, P.K. Hopke and
D.L. Johnson (1984). Interlaboratory comparison of receptor model results for Houston
aerosol. Atmos. Environ., 18(8): 1555-1566.
Edgerton, S.A., M.A.K. Khalil and R.A. Rasmussen (1984). Estimates of air pollution from
backyard burning. JAPCA, 34: 661.
Edgerton, S.A. (1985). Gaseous tracers in receptor modeling: Methyl chloride emissions from
wood combustion. Ph.D. Dissertation, Oregon Graduate Center, Beaverton, OR.
Edgerton, S.A., M.A.K. Khalil and R.A. Rasmussen (1985). Methodology for collecting short-
period integrated gas samples: Estimating acute exposure to woodburning pollution. J.
Environ. Sci. Health, A20: 563.
Edgerton, S.A., M.A.K. Khalil and R.A. Rasmussen (1986). Source emission characterization of
residential woodburning stoves and fireplaces: Fine particle/methyl chloride ratios for
use in chemical mass balance modeling. Environ. Sci. Technol, 20: 803-807.
Eitzer, B.D. (1995). Emissions of volatile organic chemicals from municipal solid waste
composting facilities. Environ. Sci. Technol., 29(4): 896-902.
-9
-------
EPA (1984). Receptor Model Technical Series: Source Apportionment Techniques and
Considerations in Combining their Use. Report No. EPA-450/4-84-020; U. S.
Environmental Protection Agency, Research Triangle Park, NC.
EPA (1985). A Cryogenic Preconcentration - Direct FID (PDFID) Method for Measurement of
NMOC in the Ambient Air. Report No. EPA-600/4-85-063; U.S. Environmental
Protection Agency; Research Triangle Park, NC; August 1985.
EPA (1987a). Protocol for Applying and Validating the CMB Model. Report No. EPA 450/4-
87-010; U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA (1987b). Revisions to the national ambient air quality standards for particulate matter: 40
CFR parts 51 and 52. Federal Register, 52 (July 1): 24634-24669.
EPA (1987c). PM-10 State Implementation Plan Development Guideline. Report No. EPA
450/2-86-001; U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA (1987d). Receptor model technical series, Volume III (Revised). CMB user's manual
(Version 6.0). Report No. EPA-450/4-83-014R; U.S. Environmental Protection Agency,
Research Triangle Park, NC.
EPA (1987e). Protocol for reconciling differences among receptor and dispersion models.
Report No. EPA-450/4-87-008. U.S. Environmental Protection Agency, Research
Triangle Park, NC.
EPA (1988). Air Emissions Species Manual, Volume II Particulate Matter Species Profiles.
Report No. EPA-450/2-88-003b. U.S. Environmental Protection Agency, Research
Triangle Park, NC.
EPA (1990). Receptor model technical series, Volume III (1989 Revision). CMB7 User's
Manual. Report No. EPA-450/4-90-004; U.S. Environmental Protection Agency,
Research Triangle Park, NC.
EPA (1996). Air quality criteria for particulate matter. Report No. EPA/600/P-95/001abcF.
U.S. EPA, National Center for Environmental Assessment, Research Triangle Park, NC.
EPA(1997a). National ambient air quality standards for particulate matter: Final Rule. Federal
Register, 62(138): 38651-38701.
http: //www.epa.gov/ttn/amtic/files/cfr/recent/pmnaaqs.pdf
EPA (1997b). Compendium method TO-14 A: Determination of volatile organic compounds
(VOCs) in ambient air using specially prepared canisters with subsequent analysis by gas
chromatography. Report No. EPA/625/R-96/010b. U.S. Environmental Protection
Agency, Cincinnati, OH.
EPA (2004). EPA-CMB8.2 Users Manual. Report No. EPA-452/R-04-011. December 2004.
U.S. Environmental Protection Agency, Research Triangle Park, NC.
Farwell, S.O. and R.A. Rasmussen (1976). Limitations of the FPD and ECD in atmospheric
analysis: A review. J. Chromatogr. Sci., 14: 224-234.
8- 10
-------
Fox, R.D. (1996). Critical review - Physical/chemical treatment of organically contaminated
soils and sediments. JAWMA, 46(5): 391-413.
Fraser, M.P., G.R. Cass and B.R.T. Simoneit (1998). Gas-phase and particle-phase organic
compounds emitted from motor vehicle traffic in a Los Angeles roadway tunnel.
Environ. Sci. Technol, 32(14): 2051-2060.
Freeman, D.L., R.T. Egami, N.F. Robinson, and J.G. Watson (1986). A method for propagating
measurement uncertainties through dispersion models. JAPCA, 36: 246-253.
Friedlander, S.K. (1973). Chemical element balances and identification of air pollution sources.
Environ. Sci. Technol., 7(3): 235-240.
Friedlander, S.K. (1981). New developments in receptor modeling theory. In Atmospheric
Aerosol: Source/Air Quality Relationships, E.S. Macias and P.K. Hopke, Eds.
American Chemical Society, Washington, DC, pp. 1-19.
Fuentes, J.D., D. Wang, H.H. Neumann, TJ. Gillespie, G. den Hartog and T.F. Dann (1996).
Ambient biogenic hydrocarbons and isoprene emissions from a mixed deciduous forest.
Journal of Atmospheric Chemistry, 25(1): 67-95.
Fujita, E.M., J.G. Watson, J.C. Chow and Z. Lu (1994). Validation of the chemical mass balance
receptor model applied to hydrocarbon source apportionment in the Southern California
Air Quality Study. Environ. Sci. Technol., 28(9): 1633-1649.
Fujita, E.M., Z. Lu, J.C. Sagebiel, N.F. Robinson and J.G. Watson (1995a). VOC source
apportionment for the Coastal Oxidant Assessment for Southeast Texas. Prepared for
Texas Natural Resource Conservation Commission, Austin, TX, by Desert Research
Institute, Reno, NV.
Fujita, E.M., J.G. Watson, J.C. Chow and K.L. Magliano (1995b). Receptor model and
emissions inventory source apportionments of nonmethane organic gases in California's
San Joaquin Valley and San Francisco Bay Area. Atmos. Environ., 29(21): 3019-3035.
Fujita, E.M. and Z. Lu (1997). Hydrocarbon receptor modeling for the 1996 Phoenix Ozone
Study. Prepared for ENSR Corporation, Camarillo, CA, by Desert Research Institute,
Reno, NV.
Fujita, E.M., Z. Lu, L. Sheetz, G. Harshfield, T. Hayes and B. Zielinska (1997a). Hydrocarbon
source apportionment in western Washington. Prepared for State of Washington Dept. of
Ecology, Lacy, WA, by Desert Research Institute, Reno, NV.
Fujita, E.M., Z. Lu, L. Sheetz, G. Harshfield and B. Zielinska (1997b). Determination of mobile
source emission source fraction using ambient field measurements. Prepared for
Coordinating Research Council, Atlanta, GA, by Desert Research Institute, Reno, NV.
- 11
-------
Fujita, E.M. and Z. Lu (1998). Analysis of data from the 1995 NARSTO-Northeast Study.
Volume III: Chemical mass balance receptor modeling. Draft final report. Prepared for
Coordinating Research Council, Atlanta, GA, by Desert Research Institute, Reno, NV.
Fujita, E.M., J.G. Watson, J.C. Chow, N.F. Robinson, L.W. Richards andN. Kumar (1998).
Northern Front Range Air Quality Study. Volume C: Source apportionment and
simulation methods and evaluation. Prepared for Colorado State University, Cooperative
Institute for Research in the Atmosphere, Ft. Collins, CO, by Desert Research Institute,
Reno, NV. http: //charon.cira.colostate.edu/DRIFinal/ZipFiles/.
Garcia, J.P., S. Beyne-Masclet and G. Mouvier (1992). Emissions of volatile organic
compounds by coal-fired power stations. Atmos. Environ., 26A(9): 1589-1597.
Gartrell, G., Jr. and S.K. Friedlander (1975). Relating particulate pollution to sources: The 1972
California Aerosol Characterization Study. Atmos. Environ., 9: 279-299.
Gay, D.A. (1987). A national inventory of biogenic hydrocarbon emissions based on a simple
forest canopy model. M.S. Thesis, Washington State University, Pullman, WA.
Gelencser, A., K. Siszler, K. and J. Hlavay (1997). Toluene-benzene concentration ratio as a
tool for characterizing the distance from vehicular emission sources. Environ. Sci.
Technol, 31(10): 2869-2871.
Gertler, A.W., D.N. Wittorff, B. Zielinska and J.C. Chow (1995). Determination of mobile
source particulate emission factors in tunnels. In Particulate Matter: Health and
Regulatory Issues. Air and Waste Management Association, Pittsburgh, PA, pp. 608-
622.
Gertler, A.W. and W.R. Pierson (1996). Recent measurements of mobile source emission factors
in North American tunnels. Sci. Total Environ., 189/190: 107-113.
Gertler, A.W., J.C. Sagebiel and W.A. Dippel (1997a). On road vehicle emissions: Results of
five urban tunnel experiments and comparison to emission factor models. In Air & Waste
Management Association Emissions Inventory Conference, Research Triangle Park, NC.
Gertler, A.W., D.N. Wittorff, R. McLaren, W. Belzer, and T. Dann (1997b). Characterization of
vehicle emissions in Vancouver, BC during the 1993 Lower Fraser Valley Oxidants
Study. Atmos. Environ., 31(14): 2107-2112.
Gillies, J.A., A.W. Gertler, J.C. Sagebiel and W.A. Dippel (1998). On-road PM2 5 and PM10
emissions in the Sepulveda Tunnel, Los Angeles, California. In Proceedings, PM25: A
Fine Particle Standard, J.C. Chow and P. Koutrakis, Eds. Air & Waste Management
Association, Pittsburgh, PA, pp. 677-698.
- 12
-------
Goldan, P.O., M. Trainer, W.C. Kuster, D.D. Parrish, J. Carpenter, J.M. Roberts, I.E. Yee and
F.C. Fehsenfeld (1995). Measurements of hydrocarbons, oxygenated hydrocarbons,
carbon monoxide, and nitrogen oxides in an urban basin in Colorado: Implications for
emission inventories. J. Geophys. Res., 100(D11): 22,771-22,785.
Gordon, G.E. (1980). Receptor models. Environ. Sci. Technol, 14: 792-800.
Gordon, G.E., W.H. Zoller, G.S. Kowalczyk and S.W. Rheingrover (1981). Composition of
source components needed for aerosol receptor models. In Atmospheric Aerosol:
Source Air/Quality Relationships, E.S. Macias and P.K. Hopke, Eds. American
Chemical Society, Washington, D.C, pp. 51-74.
Gordon, G.E., W.R. Pierson, J.M. Daisey, PJ. Lioy, J.A. Cooper, J.G. Watson and G.R. Cass
(1984). Considerations for design of source apportionment studies. Atmos. Environ.,
18(8): 1567-1582.
Gordon, G.E. (1988). Receptor models. Environ. Sci. Technol., 22(10): 1132-1142.
Gray, H.A., B. Landry, C.S. Liu, R.C. Henry, J.A. Cooper and J.R. Sherman (1988). Receptor
modeling for PM10 source apportionment in the South Coast Air Basin of California. In
Transactions, PM10: Implementation of Standards, C. V. Mathai and D.H. Stonefield,
Eds. Air Pollution Control Association, Philadelphia, PA, pp. 399-418.
Greaves, R.C., R.M. Barkley and R.E. Sievers (1985). Rapid sampling and analysis of volatile
constituents of airborne particulate matter. Analytical Chemistry, 57: 2807-2815.
Grosjean, D. and J.H. Seinfeld (1989). Parameterization of the formation potential of secondary
organic aerosols. Atmos. Environ., 23(8): 1733-1747.
Grosjean, D. (1992). In-situ organic aerosol formation during a smog episode: Estimated
production and chemical functionality. Atmos. Environ., 26A(6): 953-963.
Grosjean, D., E. Grosjean and E.L. Williams, II (1993a). Atmospheric chemistry of unsaturated
alcohols. Environ. Sci. Technol., 27(11): 2478-2485.
Grosjean, D., E.L. Williams, II and E. Grosjean (1993b). Atmospheric chemistry of isoprene
and its carbonyl products. Environ. Sci. Technol., 27(5): 830-840.
Grosjean, D., E.L. Williams, II, E. Grosjean, J.M. Andino and J.H. Seinfeld (1993c).
Atmospheric oxidation of biogenic hydrocarbons: Reaction of ozone with B-pinene, D-
limonene and trans-caryophyllene. Environ. Sci. Technol., 27(13): 2754-2758.
Grosjean, E. and D. Grosjean (1996). Carbonyl collection efficiency of the DNPH-coated C18
cartridge in dry air and in humid air. Environ. Sci. Technol., 30(3): 859-863.
- 13
-------
Guenther, A., P.R. Zimmerman, P.C. Harley, R.K. Monson and R. Fall (1993). Isoprene and
monoterpene emission rate variability: Model evaluations and sensitivity analyses. J.
Geophys. Res., 98(D7): 12609-12618.
Guenther, A., P.R. Zimmerman and M. Wildermuth (1994). Natural volatile organic compound
emission rate estimates for U.S. woodland landscapes. Atmos. Environ., 28(6): 1197-
1210.
Guenther, A., P.R. Zimmerman, L. Klinger, J. Greenberg, C. Ennis, K. Davis, W. Pollock, H.H.
Westberg, G. Allwine and C. Geron (1996). Estimates of regional natural volatile
organic compound fluxes from enclosure and ambient measurements. J. Geophys. Res.,
101(D1): 1345-1360.
Guicherit, R. (1997). Traffic as a source of volatile hydrocarbons in ambient air. Sci. Total
Environ., 205(2,3): 201-214.
Guo, Z., L.E. Sparks, B.A. Tichenor and J.C.S. Chang (1998). Predicting the emissions of
individual VOCs from petroleum-based indoor coatings. Atmos. Environ., 32(2): 231-
238.
Hampton, C.V., W.R. Pierson, T.M. Harvey, W.S. Updegrove and R.S. Marano (1982).
Hydrocarbon gases emitted from vehicles on the road. 1. A qualitative gas
chromatography/mass spectrometry survey. Environ. Sci. Technol, 16: 287-298.
Hampton, C.V., W.R. Pierson, D. Schuetzle and T.M. Harvey (1983). Hydrocarbon gases
emitted from vehicles on the road. 2. Determination of emission rates from diesel and
spark-ignition vehicles. Environ. Sci. Technol., 17(12): 699-708.
Harley, R.A., M.P. Hannigan and G.R. Cass (1992). Respeciation of organic gas emissions and
the detection of excess unburned gasoline in the atmosphere. Environ. Sci. Technol.,
26(12): 2395-2408.
Haszpra, L. and I. Szilagyi (1994). Non-methane hydrocarbon composition of car exhaust in
Hungary. Atmos. Environ., 28(16): 2609-2614.
Hawthorne, S.B. and DJ. Miller (1986). Extraction and recovery of organic pollutants from
environmental solids and Tenax GG using supercritical CO2. J. Chromatogr. Sci., 24:
258-264.
Hawthorne, S.B., DJ. Miller, R.M. Barkley and M.S. Krieger (1988). Identification of
methoxylated phenols as candidate tracers for atmospheric wood smoke pollution.
Environ. Sci. Technol., 22(10): 1191-1196.
Hawthorne, S.B., M.S. Krieger, DJ. Miller and M.B. Mathiason (1989). Collection and
quantitation of methoxylated phenol tracers for atmospheric pollution from residential
wood stoves. Environ. Sci. Technol., 23(4): 470-475.
- 14
-------
Heinsohn, R.J., J.W. Davis and K.T. Knapp (1980). Dilution source sampling systems. Environ.
Sci. Technol, 14: 1205.
Henry, R.C. (1982). Stability analysis of receptor models that use least squares fitting. In
Proceedings: Receptor Models Applied to Contemporary Air Pollution Problems, P.K.
Hopke and S.L. Dattner, Eds. Air Pollution Control Association, Pittsburgh, PA, pp.
141-162.
Henry, R.C. and B.M. Kim (1990). Extension of self-modeling curve resolution to mixtures of
more than three components Part 1. Finding the basic feasible region. Chemom. Intell.
Lab. Sys., 8: 205-216.
Henry, R.C. (1992). Dealing with near collinearity in chemical mass balance receptor models.
Atmos. Environ., 26A(5): 933-938.
Henry, R.C., C.W. Lewis and J.F. Collins (1994). Vehicle-related hydrocarbon source
compositions from ambient data: The GRACE/SAFER method. Environ. Sci. Technol.,
28(5): 823-832.
Hering, S.V., A.H. Miguel and R.L. Dod (1984). Tunnel measurements of the PAH, carbon
thermogram and elemental source signature for vehicular exhaust. Sci. Total Environ.,
36: 39.
Hewitt, C.N. and R.A. Street (1992). A qualitative assessment of the emission of non-methane
hydrocarbon compounds from the biosphere to the atmosphere in the U.K.: Present
knowledge and uncertainties. Atmos. Environ., 26A(17): 3069-3078.
Hidy, G.M. and S.K. Friedlander (1971). The nature of the Los Angeles aerosol. In Proceedings
of the Second International Clean Air Congress., H.M. Englund and W.T. Beery, Eds.
Academic Press, New York, pp. 391-404.
Hildemann, L.M., G.R. Cass and G.R. Markowski (1989). A dilution stack sampler for
collection of organic aerosol emissions: Design, characterization and field tests. Aerosol
Sci. Technol., 10: 193-204.
Hildemann, L.M., G.R. Markowski, M.C. Jones and G.R. Cass (1991). Submicrometer aerosol
mass distributions of emissions from boilers, fireplaces, automobiles, diesel trucks, and
meat-cooking operations. Aerosol Sci. Technol., 14(1): 138-152.
Hlavinka, M.W. and J.A. Bullin, J.A. (1988). Validation of mobile source emission estimates
using mass balance techniques. JAPCA, 38(8): 1035-1039.
Hoekman, S.K. (1992). Speciated measurements and calculated reactivities of vehicle exhaust
emissions from conventional and reformulated gasolines. Environ. Sci. Technol., 26(6):
1206-1216.
- 15
-------
Hopke, P.K., W. Wlaschin, S. Landsberger, C.W. Sweet and S.J. Vermette (1988). The source
apportionment of PM10 in South Chicago. In Transactions, PM10: Implementation of
Standards., C. V. Mathai and D.H. Stonefield, Eds. Air Pollution Control Association,
Pittsburgh, PA, pp. 484-494.
Houck, I.E., J.A. Cooper and E.R. Larson (1982). Dilution sampling for chemical receptor
source fingerprinting. In 75th Annual Meeting of the APCA, New Orleans, LA.
Houck, I.E., J.C. Chow, J.G. Watson, C.A. Simons, L.C. Pritchett, J.M. Goulet and C.A. Frazier
(1989). Determination of particle size distribution and chemical composition of
particulate matter from selected sources in California. Report No. A6-175-32. Prepared
for California Air Resources Board, Sacramento, CA, by OMNI Environmental Services
Inc. and Desert Research Institute, Beaverton, OR and Reno, NV.
Houck, I.E., J.A. Rau, S. Body and J.C. Chow (1992). Source apportionment - Pocatello, Idaho
PM10 nonattainment area. In Transactions, PM10 Standards and Nontraditional
Particulate Source Controls, J.C. Chow and D.M. Ono, Eds. Air and Waste
Management Association, Pittsburgh, PA, pp. 219-230.
Hougland, E.S. (1983). Chemical element balance by linear programming. In 76th Annual
Meeting of the Air Pollution Control Association, Atlanta, GA.
Hov, O.R., J. Schjoldager and B.M. Wathne (1983). Measurement and modeling of the
concentration of terpenes in coniferous forest air. J. Geophys. Res., 88: 10679-10688.
Hueglin, C., L. Scherrer and H. Burtscher (1997). An accurate continuously adjustable dilution
system (1: 10 to 1: 104) for submicron aerosols. J. Aerosol Sci., 28 (6): 1049-1055.
Hurst, D.F., D.W.T. Griffith and G.D. Cook (1994). Trace gas emissions from biomass burning
in tropical Australian savannas. J. Geophys. Res., 99(D8): 16441-16456.
Ingalls, M.N. (1989). On-road vehicle emissions factors from measurements in a Los Angeles
area tunnel. In 82nd Annual Meeting of the Air And Waste Management Assoication,
Anaheim, CA.
Isidorov, V.A., I.G. Zenkevich and B.V. loffe (1985). Volatile organic compounds in the
atmosphere of forests. Atmos. Environ., 19: 1-8.
Japar, S.M., T.J. Wallington, J.F.O. Richert and J.C. Ball (1990). The atmospheric chemistry of
oxygenated fuel additives: t-butyl alcohol, dimethyl ether and methyl t-butyl ether. Int.
J. Chem. Kinet., 22: 1257.
Japar, S.M., T.J. Wallington, S.J. Rudy and T.Y. Chang (1991). Ozone-forming potential of a
series of oxygenated organic compounds. Environ. Sci. Technol, 25: 415.
Javitz, H.S. and J.G. Watson (1986). Methods for testing receptor-oriented and source-oriented
models, T.G. Pace, Ed. Air Pollution Control Association, Pittsburgh, PA, pp. 161-174.
8- 16
-------
Javitz, H.S., J.G. Watson and N.F. Robinson (1988a). Performance of the chemical mass
balance model with simulated local-scale aerosols. Atmos. Environ., 22(10): 2309-2322.
Javitz, H.S., J.G. Watson, J.P. Guertin and P.K. Mueller (1988b). Results of a receptor modeling
feasibility study. JAPCA, 38(5): 661-667.
Jensen, T.E. and R.A. Kites (1983). Aromatic diesel emissions as a function of engine
conditions. Analytical Chemistry, 55: 594-599.
Jiang, W., D.L. Singleton, M. Hedley, R. McLaren, T. Dann and D. Wang (1997). Comparison
of organic compound compositions in the emissions inventory and ambient data for the
LowerFraser Valley. JAWMA, 47(8): 851-860.
Jobson, B.T., Z.W. Niki and L.A. Barrie (1994). Seasonal trends of isoprene. C2 - C5 Alkanes,
and acetylene at a remote boreal site in Canada. J. Geophys. Res., 99(D1): 1589-1600.
Juttner, F. (1988). Quantitative analysis of monsterpenes of volatile organic pollution products
(VOC) in forest air of the southern Black Forest. Chemosphere, 17: 309.
Kaiser, E.W., W.O. Siegl, Y.I. Henig, R.W. Anderson and F.H. Trinker (1991). Effect of fuel
structure on emissions from a spark-ignited engine. Environ. Sci. Technol, 25: 12.
Kalman, D. (1986). Survey analysis of volatile organics released from plastics under thermal
stress. J. Am. Ind. Hyg. Assoc., 47(5): 270-275.
Kamens, R.M., Z. Guo, J.N. Fulcher and D.A. Bell (1988). Influence of humidity, sunlight and
temperature on the daytime decay of polyaromatic hydrocarbons on atmospheric soot
particles. Environ. Sci. Technol., 22: 103-108.
Kamens, R.M., H. Karam, J. Guo, J.M. Perry and L. Stockburger (1989). The behavior of
oxygenated polycyclic aromatic hydrocarbons on atmospheric soot particles. Environ.
Sci. Technol, 23 (7): 801-806.
Kawamura, K., L.L. Ng and I.R. Kaplan (1985). Determination of organic acids (Ct -C10) in the
atmosphere, motor exhausts, and engine oil. Environ. Sci. Technol., 19: 1082-1086.
Kempf, K., E.J. Allwine, H.H. Westberg, C.S. Claiborn and B. Lamb (1996). Hydrocarbon
emissions from spruce species using environmental chamber and branch enclosure
methods. Atmos. Environ., 30(9): 1381-1389.
Kenski, D.M., R.A. Wadden, P.A. Scheff and W.A. Lonneman (1995). Receptor modeling
approach to VOC emission inventory validation. J. Environ. Eng., 121(7): 483-491.
Khalil, M.A.K., S.A. Edgerton and R.A. Rasmussen (1983). A gaseous tracer model for air
pollution from residential wood burning. Environ. Sci. Technol., 17(9): 555-559.
- 17
-------
Khalil, M.A.K., R.A. Rasmussen and S.A. Edgerton (1985). Gaseous tracers for sources of
regional scale pollution. JAPCA: 838.
Khalil, M.A.K. and R.A. Rasmussen (1992). Forest hydrocarbon emissions: Relationships
between fluxes and ambient concentrations. JAWMA, 42(6): 810-813.
Khalili, N.R., P.A. Scheff and T.M. Holsen (1995). PAH source fingerprints for coke ovens,
diesel and gasoline engines, highway tunnels, and wood combustion emissions. Atmos.
Environ., 29(4): 533-542.
Kim, B.M. and R.C. Henry (1989). Analysis of multicollinearity indicators and influential
species for chemical mass balance receptor model. In Transactions, Receptor Models in
Air Resources Management., J.G. Watson, Ed. Air & Waste Management Association,
Pittsburgh, PA, pp. 379-390.
Kim, B.M., R. Lewis and H. Hogo (1992). Source apportionment by chemical mass balance: A
comparison between measured source profiles and SAFER model estimated source
profiles. In Transactions, PM10 Standards andNontraditionalParticulate Source
Controls, J.C. Chow and D.M. Ono, Eds. Air & Waste Management Association,
Pittsburgh, PA, pp. 146-162.
Kim, Y.P., J.H. Seinfeld and P. Saxena (1993a). Atmospheric gas-aerosol equilibrium I.
Thermodynamic model. AerosolSci. Technol, 19(2): 157-181.
Kim, Y.P., J.H. Seinfeld and P. Saxena (1993b). Atmospheric gas-aerosol equilibrium II.
Analysis of common approximations and activity coefficient calculation methods.
Aerosol Sci. Technol., 19(2): 182-198.
Kitto, A.M., M. Pirbazari, B.N. Badriyha, V. Ravindran, R. Tyner and C.E. Synolakis (1997).
Emissions of volatile and semi-volatile organic compounds and parti culate matter from
hot asphalts. Environmental Technology, 18(2): 121-138.
Kleindienst, T.E., E.W. Corse, F.T. Blanchard and W.A. Lonneman (1998). Evaluation of the
performance of DNPH-coated silica gel and C18 cartridges in the measurement of
formaldehyde in the presence and absence of ozone. Environ. Sci. Technol.: 124-130.
Klouda, G.A., C.W. Lewis, R.A. Rasmussen, G.C. Rhoderick, R.L Sams, R.K. Stevens, L.A.
Currie, DJ. Donahue, A.J.T. lull and R.L. Seila (1996). Radiocarbon measurements of
atmospheric volatile organic compounds: quantifying the biogenic contribution.
Environ. Sci. Technol., 30(4): 1098-1105.
Kneip, T.J., M.T. Kleinman and M. Eisenbud (1972). Relative contribution of emission sources
to the total airborne particulates in New York City. In 3rdIUAPPA Clean Air Congress.
8- 18
-------
Koppmann, R., A. Khedim, J. Rudolph, D. Poppe, M.O. Andreae, G. Helas, M. Welling and T.
Zenker (1997). Emissions of organic trace gases from savanna fires in southern Africa
during the 1992 Southern African Fire Atmosphere Research Initiative and their impact
on the formation of tropospheric ozone. J. Geophys. Res., 102(D15): 18879-18888.
Koracin, D., L. Enger and X. Yang (1993). A numerical study of the boundary-layer dynamics
in a mountain valley. Part 1: Model validation and sensitivity experiments. Boundary
Layer Meteorology, 66: 357-394.
Koracin, D. and L. Enger (1994). A numerical study of boundary-layer dynamics in a mountain
valley. Part 2: Observed and simulated characteristics of atmospheric stability and local
flows. Boundary Layer Meteorology, 69(3): 249-283.
Kowalczyk, G.S., C.E. Choquette and G.E. Gordon (1978). Chemical element balances and
identification of air pollution sources in Washington, DC. Atmos. Environ., 12: 1143-
1153.
Lamb, B., H.H. Westberg, T. Quarles and D.L. Flyckt (1984). Natural hydrocarbon emission
rate measurements from selected forest sites. Report No. EPA-600/S3-84-001. Prepared
by U.S. Environmental Protection Agency, Research Triangle Park, NC.
Lamb, B., H.H. Westberg and G. Allwine (1985). Biogenic hydrocarbon emissions from
deciduous and coniferous species in the United States. J. Geophys. Res., 90: 2380-2390.
Lamb, B., H.H. Westberg and G. Allwine (1986). Isoprene emission fluxes determined by an
atmospheric tracer technique. Atmos. Environ., 21: 1-8.
Lamb, B., A. Guenther, D.A. Gay and H.H. Westberg (1987). A national inventory of natural
nonmethane hydrocarbons. Atmos. Environ., 21: 1695-1705.
Lamb, B., D.A. Gay, H.H. Westberg and T.E. Pierce (1993). A biogenic hydrocarbon emission
inventory for the USA using a simple forest canopy model. Atmos. Environ., 27A(11):
1673-1690.
Larson, T.V. and RJ. Vong (1989). Partial least squares regression methodology: Application
to source receptor modeling. In Transactions, Receptor Models in Air Resources
Management, J.G. Watson, Ed. Air & Waste Management Association, Pittsburgh, PA,
pp. 391-403.
Lewis, C.W., T.L. Conner, R.K. Stevens, J.F. Collins andR.C. Henry (1993). Receptor
modeling of volatile hydrocarbons measured in the 1990 Atlanta ozone precursor study.
In 86th Annual Meeting of the Air and Waste Management Association, Denver, CO.
Lewis, C.W., R.K. Stevens, R.A. Rasmussen, C.A. Cardelino and T.E. Pierce (1999). Biogenic
fraction of ambient VOC: Comparison of radiocarbon, chromatographic, and emissions
inventory estimates for Atlanta, Georgia. JAWMA, 49(3): 299-307.
8- 19
-------
Lin, C. and J.B. Milford (1994). Decay-adjusted chemical mass balance receptor modeling for
volatile organic compounds. Atmos. Environ., 28(20): 3261-3276.
Lioy, P.J., J.G. Watson and J.D. Spengler (1980). APCA specialty conference workshop on
baseline data for inhalable particulate matter. JAPCA, 30(10): 1126-1130.
Lonneman, W.A., R.L. Sella and S.A. Meeks (1986). Non-methane organic composition in the
Lincoln Tunnel. Environ. Sci. Technol, 20: 790-796.
Lovelock, I.E., RJ. Maggs and R. Wade (1973). Halogenated hydrocarbons in and over the
Atlantic. Nature, 241: 194.
Lowenthal, D.H., R.C. Hanumara, K.A. Rahn and L.A. Currie (1987). Effects of systematic
error, estimates and uncertainties in chemical mass balance apportionments: Quail Roost
II revisited. Atmos. Environ., 21(3): 501-510.
Lowenthal, D.H. and K.A. Rahn (1988a). Tests of regional elemental tracers of pollution
aerosols. 2. Sensitivity of signatures and apportionments to variations in operating
parameters. Atmos. Environ., 22: 420-426.
Lowenthal, D.H. and K.A. Rahn (1988b). Reproducibility of regional apportionments of
pollution aerosol in the northeastern United States. Atmos. Environ., 22(9): 1829-1833.
Lowenthal, D.H., K.R. Wunschel and K.A. Rahn (1988). Tests of regional elemental tracers of
pollution aerosols. 1. Distinctness of regional signatures, stability during transport, and
empirical validation. Environ. Sci. Technol., 22: 413-420.
Lowenthal, D.H., J.C. Chow, J.G. Watson, G.R. Neuroth, R.B. Robbins, B.P. Shafritz and RJ.
Countess (1992). The effects of collinearity on the ability to determine aerosol
contributions from diesel- and gasoline-powered vehicles using the chemical mass
balance model. Atmos. Environ., 26A(13): 2341-2351.
Lowenthal, D.H., B. Zielinska, J.C. Chow, J.G. Watson, M. Gautam, D.H. Ferguson, G.R.
Neuroth and K.D. Stevens (1994). Characterization of heavy-duty diesel vehicle
emissions. Atmos. Environ., 28(4): 731-743.
Lu, Z. (1996). Temporal and spatial analysis of VOC source contributions for Southeast Texas.
Ph.D. Dissertation, University of Nevada, Reno.
Magliano, K.L. (1988). Level 1 PM10 assessment in a California air basin. In Transactions,
PM10: Implementation of Standards, C.V. Mathai and D.H. Stonefield, Eds. Air
Pollution Control Association, Pittsburgh, PA, pp. 508-517.
Mayrsohn, H. and J.H. Crabtree (1976). Source reconciliation of atmospheric hydrocarbons.
Atmos. Environ., 10: 137-143.
8-20
-------
Mayrsohn, H., J.H. Crabtree, M. Kuramoto, R.D. Sothern and S.H. Mano (1977). Source
reconciliation of atmospheric hydrocarbons 1974. Atmos. Environ., 11: 189-192.
McCabe, R.W., W.O. Siegl, W. Chun and J. Perry (1992). Speciated hydrocarbon emissions
from the combustion of single component fuels II. Catalyst effects. JAWMA, 42(8):
1071-1077.
McCain, J.D. and A.D. Williamson (1984). Development and evaluation of dilution probes used
for sampling to determine source signatures. Report No. EPA 600/S3-84-045. Prepared
by U.S. Environmental Protection Agency, Research Triangle Park, NC.
McClenny, W.A., K.D. Oliver and J.D. Pleil (1989). A field strategy for sorting volatile
organics into source-related groups. Environ. Sci. Technol, 23: 1373-1379.
McDonald, J.D., B. Zielinska, E.M. Fujita, J.C. Chow, J.G. Watson and J.C. Sagebiel (1998).
The development of chemical emission profiles for residential wood combustion and
meat cooking for use in the apportionment of atmospheric carbonaceous aerosol. In
Proceedings, PM25: A Fine Particle Standard., J.C. Chow and P. Koutrakis, Eds. Air &
Waste Management Association, Pittsburgh, PA, pp. 645-660.
McKay, W.A., M.F. Turner, B.M.R. Jones and C.M. Halliwell (1996). Emissions of
hydrocarbons from marine phytoplankton—Some results from controlled laboratory
experiments. Atmos. Environ., 30(14): 2583-2594.
Merrill, R.G. and D.B. Harris (1987). Field and laboratory evaluation of a woodstove dilution
sampling system. In 80th Annual Meeting of the Air Pollution Control Association, New
York, NY.
Miguel, A.H. (1984). "Atmospheric" reactivity of parti culate poly cyclic aromatic hydrocarbons
collected in an urban tunnel. Sci. Total Environ., 36: 305-311.
Miller, M.S., S.K. Friedlander and G.M. Hidy (1972). A chemical element balance for the
Pasadena aerosol. J. Colloid Interface Sci., 39(1): 165-176.
Moeckli, M.A., M. Fierz and M.W. Sigrist (1996). Emission factors for ethene and ammonia
from a tunnel study with a photoacoustic trace gas detection system. Environ. Sci.
Technol., 30(9): 2864-2867.
Nelson, P.F. and S.M. Quigley (1982). Non-methane hydrocarbons in the atmosphere of
Sydney, Australia. Environ. Sci. Technol., 16: 650-655.
Nelson, P.F., S.M. Quigley and M.Y. Smith (1983). Sources of atmospheric hydrocarbons in
Sydney: A quantitative determination using a source reconciliation technique. Atmos.
Environ., 17(3): 439-449.
Nelson, P.F. and S.M. Quigley (1984). The hydrocarbon composition of exhaust emitted from
gasoline fueled vehicles. Atmos. Environ., 18:79.
8-21
-------
Nondek, L., D.R. Rodler and J.W. Birks (1992). Measurement of sub-ppbv concentrations of
aldehydes in a forest atmosphere using a new HPLC technique. Environ. Sci. Technol,
26(6): 1174-1178.
O'Shea, WJ. and P.A. Scheff (1988). A chemical mass balance for volatile organics in Chicago.
JAPCA, 38(8): 1020-1026.
Ohira, T., M.T. Ishuguro and O. Koyama (1976). Gas chromatographic analysis of offensive
odors at plants handling fish entrails and bones. Japan Soc. Air Pollut., 2: 58-59.
Oliver, W.R., G.W. Lundberg and J.D. Banks (1984). Biogenic hydrocarbon emissions in the
southern San Joaquin Valley air basin. Report No. SYSAPP-84/195. Prepared by
Systems Applications, Inc., San Rafael, CA.
Owen, S., C. Boissard, R.A. Street, S.C. Duckham, O. Csiky and C.N. Hewitt (1997). Screening
of 18 Mediterranean plant species for volatile organic compound emissions. Atmos.
Environ., 31(S1): 101-118.
Pandis, S.N., S.E. Paulson, U. Baltensperger, J.H. Seinfeld, R.C. Flagan, EJ. Palen and D.T.
Allen (1992). Aerosol production during the photooxidation and natural hydrocarbons.
In Precipitation Scavenging and Atmosphere-Surface Exchange. Vol. 3 - The Summers
Volume: Applications and Appraisals, S.E. Schwartz and W.G.N. Slinn, Eds.
Hemisphere Publishing Corp., Washington, D.C., pp. 1641-1652.
Pankow, J.F. (1994a). An absorption model of gas/particle partitioning of organic compunds in
the atmosphere. Atmos. Environ., 28(2): 185-188.
Pankow, J.F. (1994b). An absorption model of the gas/aerosol partitioning involved in the
formation of secondary organic aerosol. Atmos. Environ., 28(2): 189-194.
Pasquill, F. (1974). Atmospheric Diffusion, 2nd ed. Ellis Horwood, Chichester, England.
Passant, N.R., SJ. Richardson, R.P.J. Swannell, N. Gibson, MJ. Woodfield, J.P. van der Lugt,
J.H. Wolsink and P.G.M. Hesselink (1993). Emissions of volatile organic compounds
(VOCs) from the food and drink industries of the European community. Atmos.
Environ., 27A(16): 2555-2566.
Pellizzari, E.D., B. Demian and K.J. Krost (1984). Sampling of organic compounds in the
presence of reactive inorganic gases with Tenax GC. Analytical Chemistry, 56: 793-798.
Pier, P.A. and C. McDuffie, Jr. (1997). Seasonal isoprene emission rates and model comparisons
using whole-tree emissions from white oak. J. Geophys. Res., 102(D20): 23963-23972.
Pierson, W.R. and W.W. Brachaczek (1976). Particulate matter associated with vehicles on the
road, Paper 760039. Transactions of the Society of Automotive Engineers, 85: 209-227.
8-22
-------
Pierson, W.R. and W.W. Brachaczek (1983). Particulate matter associated with vehicles on the
road II. Aerosol Sci. Technol., 2: 1-40.
Pierson, W.R., A.W. Gertler and R.L. Bradow (1990). Comparison of the SCAQS tunnel study
with other on-road vehicle emission data. JAWMA, 40(11): 1495-1504.
Pierson, W.R., A.W. Gertler, N.F. Robinson, J.C. Sagebiel, B. Zielinska, G.A. Bishop, D.H.
Stedman, R.B. Zweidinger and W.D. Ray (1996). Real-world automotive emissions -
Summary of studies in the Fort McHenry and Tuscarora Mountain tunnels. Atmos.
Environ., 30(12): 2233-2256.
Pilinis, C., J.H. Seinfeld and C. Seigneur (1987). Mathematical modeling of the dynamics of
multicomponent atmospheric aerosols. Atmos. Environ., 21(4): 943-955.
Pytkowicz, R.M. and D.R. Kester (1971). The physical chemistry of sea water. Oceanogr. Mar.
Biol, 9 11-60.
Ramdahl, T., I. Alfheim, S. Rustad and T. Olsen (1982). Chemical and biological
characterization of emissions from small residential stoves burning wood and charcoal.
Chemosphere, 11(6): 601-611.
Ramdahl, T. (1983). Retene - a molecular marker of wood combustion in ambient air. Nature,
306: 580-582.
Ramdahl, T. and M. Moller (1983). Chemical and biological characterization of emissions from
a cereal straw burning furnace. Chemosphere, 12: 23.
Rasmussen, R.A., L.E. Rasmussen, M.A.K. Khalil and R.W. Dalluge (1980). Concentration
distribution of methyl chloride in the atmosphere. J. Geophys. Res., 85(C12): 7350.
Rasmussen, R.A., C.W. Lewis, R.K. Stevens, W.D. Ellenson and S.L. Dattner (1996).
Removing CO2 from atmospheric samples for radiocarbon measurements of volatile
organic compounds. Environ. Sci. Technol., 30(4): 1092-1097.
Rau, J.A. and M.A.K. Khalil (1989). Tracers of pollution from wood burning and receptor
modeling. In Transactions, Receptor Models in Air Resources Management., J.G.
Watson, Ed. Air & Waste Management Association, Pittsburgh, PA, pp. 353-364.
Rheingrover, S.W. and G.E. Gordon (1980). Identifying locations of dominant point sources of
elements in urban atmospheres from large multi-element data sets. In 4th International
Conference on Nuclear Methods in Environmental and Energy Research, Columbia, MO.
Riba, M.L., J.P. Tathy, N. Tsiropoulos, B. Monsarrat and L. Torres (1987). Diurnal variation in
the concentration of alpha- and beta-pinene in the Landes Forest (France). Atmos.
Environ., 21: 191-193.
-23
-------
Riba, M.L., N. Tsiropoulos and L. Torres (1988). Efficiency of absorption desorption of volatile
compounds: Application to isoprene on Tenax GC. J. Chromatogr., 437: 139.
Richards, L.W., J.A. Anderson, D.L. Blumenthal, A.A. Brandt, J.A. McDonald, N. Waters, E.S.
Macias and P.S. Bhardwaja (1981). The chemistry, aerosol physics, and optical
properties of a western coal-fired power plant plume. Atmos. Environ., 15: 2111-2134.
Richards, L.W., J.A. Anderson, D.L. Blumenthal, J.A. McDonald, E.S. Macias, S.V. Hering and
W.E. Wilson (1985). Chemical, aerosol, and optical measurements in the plumes of three
midwestern coal-fired power plants. Atmos. Environ., 19: 1685-1704.
Roberts, J.M., C.J. Hahn, F.C. Fehsenfeld, J.M. Warnock, D.L. Albritton and R.E. Sievers
(1985). Monoterpene hydrocarbons in the nighttime troposphere. Environ. Sci. Technol,
19: 364-369.
Rogak, S.N., S.I. Green and U. Pott (1998a). Use of tracer gas for direct calibration of emission-
factor measurements in a traffic tunnel. JAWMA, 48(6): 545-552.
Rogak, S.N., U. Pott, T. Dann and D. Wang (1998b). Gaseous emissions from vehicles in a
traffic tunnel in Vancouver, BC. JAWMA, 48(7): 604-615.
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1991). Sources
of fine organic aerosol - 1. Charbroilers and meat cooking operations. Environ. Sci.
Technol, 25(6): 1112-1125.
Rogge, W.F. (1993). Molecular tracers for sources of atmospheric carbon particles:
Measurements and model predictions. Ph.D. Dissertation, California Institute of
Technology, Pasadena, CA.
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1993a).
Sources of fine organic aerosol - 2. Noncatalyst and catalyst-equipped automobiles and
heavy-duty diesel trucks. Environ. Sci. Technol., 27(4): 636-651.
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1993b).
Sources of fine organic aerosol - 5. Natural gas home appliances. Environ. Sci. Technol.,
27(13): 2736-2744.
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1993c).
Sources of fine organic aerosol - 3. Road dust, tire debris, and organometallic brake
lining dust: Roads as sources and sinks. Environ. Sci. Technol., 27: 1892-1904.
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1997). Sources
of fine organic aerosol - 8. Boilers burning No. 2 distillate fuel oil. Environ. Sci.
Technol., 31(10): 2731-2737.
8-24
-------
Rogge, W.F., L.M. Hildemann, M.A. Mazurek, G.R. Cass and B.R.T. Simoneit (1998). Sources
of fine organic aerosol - 9. Pine, oak, and synthetic log combustion in residential
fireplaces. Environ. Sci. Technol, 32(1): 13-22.
Ryan, W.M., C.R. West, D.R. Holtz, T.A. Peters, J.A. Cooper and D.M. Ono (1988).
Reconciliation of receptor and dispersion modeling impacts of PM10 in Hay den, Arizona.
In Transactions, PM10: Implementation of Standards., C. V. Mathai and D.H. Stonefield,
Eds. Air Pollution Control Association, Pittsburgh, PA, pp. 419-429.
Sack, T.M., D.H. Steele, K. Hammerstrom and J. Remmers (1992). A survey of household
products for volatile organic compounds. Atmos. Environ., 26A(6): 1063-1070.
Sagebiel, J.C., B. Zielinska, W.R. Pierson and A.W. Gertler (1996). Real-world emissions and
calculated reactivities of organic species from motor vehicles. Atmos. Environ., 30(12):
2287-2296.
Sagebiel, J.C., B. Zielinska, P.A. Walsh, J.C. Chow, S.H. Cadle, P.A. Mulawa, K.T. Knapp, R.B.
Zweidinger and R. Snow (1997). PM10 exhaust samples collected during IM-240
dynamometer tests of in-service vehicles in Nevada. Environ. Sci. Technol., 31(1): 75-
83.
Sampson, R.E. and G.S. Springer (1973). Effects of exhaust gas temperature and fuel
composition on particulate emissions from spark-ignition engines. Environ. Sci.
Technol., 7: 55.
Sanchez, D.C., M.A. Mason and C.H. Norris (1987). Methods and results of characterization of
organic emissions from an indoor material. Atmos. Environ., 21(2): 337-345.
Saxena, P. and L.M. Hildemann (1997). Water absorption by organics: Survey of laboratory
evidence and evaluation of UNIFAC for estimating water activity. Environ. Sci.
Technol, 31(11): 3318-3324.
Scheff, P.A. and M. Klevs (1987). Source-receptor analysis of volatile hydrocarbons. J.
Environ. Eng., 113(5): 994-1005.
Scheff, P.A., R.A. Wadden, B.A. Bates and P.P. Aronian (1989a). Source fingerprints for
receptor modeling of volatile organics. JAPCA, 39(4): 469-478.
Scheff, P.A., P.P. Aronian, R.A. Wadden and B.A. Bates (1989b). Development of volatile
organic compound source fingerprints for receptor modeling applications. In
Transactions, Receptor Models in Air Resources Management, J.G. Watson, Ed. Air &
Waste Management Association, Pittsburgh, PA, pp. 83-107.
Scheff, P.A. and R.A. Wadden (1993). Receptor modeling of volatile organic compounds: 1.
Emission inventory and validation. Environ. Sci. Technol., 27(4): 617-625.
-25
-------
Scheff, P.A., R.A. Wadden, D.M. Kenski, J. Chung and G. Wolff (1996). Receptor model
evaluation of the Southeast Michigan Ozone Study ambient NMOC measurements.
JAWMA, 46(11): 1048-1057.
Schuh, G., A.C. Heiden, T. Hoffmann, J. Kahl, P. Rockel, J. Rudolph and J. Wildt (1997).
Emissions of volatile organic compounds from sunflower and beech: dependence on
temperature and light intensity. Journal of Atmospheric Chemistry, 27(3): 291-318.
Schulze, R.H. (1990). Dispersion modeling using personal computers. Atmos. Environ., 24A(8):
2051-2057.
Schwede, D.B. and J.O. Paumier (1997). Sensitivity of the industrial source complex model to
input deposition parameters. J. Appl. Meteorol, 36(8): 1096-1106.
Seaman, N.L., D.R. Stauffer and A.M. Lario-Gibbs (1995). A multiscale four-dimensional data
assimilation system applied in the San Joaquin Valley during SARMAP. Part I:
Modeling design and basic performance characteristics. Appl. Meteorol., 34(8): 1739-
1761.
Seinfeld, J.H. (1986). Atmospheric Chemistry and Physics of Air Pollution. John Wiley & Sons,
New York, NY.
Seinfeld, J.H. and S.N. Pandis (1998). Atmospheric Chemistry and Physics: From Air
Pollution to Climate Change. John Wiley & Sons, New York, NY.
Sexton, K. and H.H. Westberg (1979). Ambient air measurements of petroleum refinery
emissions. JAPCA, 29(11): 1149-1152.
Sexton, K. and H.H. Westberg (1983). Photochemical ozone formation from petroleum refinery
emissions. Atmos. Environ., 17: 467-475.
Shah, J.J. and H.B. Singh (1988). Distribution of volatile organic chemicals in outdoor and
indoor air. Environ. Sci. Technol, 22(12): 1381-1388.
Shah, J.J., R.L. Johnson and J.E. Houck (1989). Source characterization using tethered balloons
and lightweight samplers. In Transactions: Receptor Models in Air Resources
Management, J.G. Watson, Ed. Air & Waste Management Association, Pittsburgh, PA,
pp. 334-345.
Shaw, G.E. (1983). X-ray spectrometry of polar aerosols. Atmos. Environ., 17(2): 329-339.
Sheffield, A.E. and G.E. Gordon (1986). Variability of particle composition from ubiquitous
sources: Results from a new source-composition library. In Transactions, Receptor
Methods for Source Apportionment: Real World Issues and Applications., T.G. Pace, Ed.
Air Pollution Control Association, Pittsburgh, PA, pp. 9-22.
8-26
-------
Shonnard, D.R. and R.L. Bell (1993). Benzene emissions from a contaminated air-dry soil with
fluctuations of soil temperature or relative humidity. Environ. Sci. Technol, 27(13):
2909-2913.
Siegl, W.O., R.W. McCabe, W. Chun, E.W. Kaiser, J. Perry, Y.I. Henig, F.H. Trinker and R.W.
Anderson (1992). Speciated hydrocarbon emissions from the combustion of single
component fuels I. Effect of fuel structure. JAWMA, 42(7): 912-920.
Sigsby, I.E., S.B. Tejada, W.D. Ray, J.M. Lang and J.W. Duncan (1987). Volatile organic
compound emissions from 46 in-use passenger cars. Environ. Sci. Technol., 21: 466-475.
Simo, R., J.O. Grimalt and J. Albaiges (1997). Loss of unburned-fuel hydrocarbons from
combustion aerosols during atmospheric transport. Environ. Sci. Technol., 31(9): 2697.
Singh, H.B., D. O'Hara, D. Herlth, W. Sachse, D.R. Blake, J.D. Bradshaw, M. Kanakidou and
PJ. Crutzen (1994). Acetone in the atmosphere: Distribution, sources, and sinks. J.
Geophys. Res., 99(D1): 1805-1820.
Sjogren, M., H. Li, U. Rannug and R. Westerholm (1996). Multivariate analysis of exhaust
emissions from heavy-duty diesel fuels. Environ. Sci. Technol., 30(1): 38-49.
Skidmore, L. and J.C. Chow (1992). PM10 air quality assessment for the Jefferson County, Ohio
air quality control region. In Transactions, PM10 Standards andNontraditional
Paniculate Source Controls, J.C. Chow and D.M. Ono, Eds., pp. 1016-1031.
Small, M., M.S. Germani, A.M. Small, W.H. Zoller and J.L. Moyers (1981). Airborne plume
study of emissions from the processing of copper ores in southeastern Arizona. Environ.
Sci. Technol., 15(3): 293-299.
Snow, R., L. Baker, W. Crews, C.O. Davis, J. Duncan, N. Perry, P. Sludak, F.D. Stump, W.D.
Ray and J.N. Braddock (1989). Characterization of emissions from a methanol fueled
motor vehicle. JAPCA, 39(1): 48-54.
Song, X.H. and P.K. Hopke (1996). Solving the chemical mass balance problem using an
artificial neural network. Environ. Sci. Technol., 30(2): 531-535.
Sousa, J.A., J.M. Daisey, J.E. Houck and J.A. Cooper (1985). The mutagenic activity of
particulate organic matter collected with a dilution sampler at coal-fired power plants. In
78th Annual Meeting of the Air Pollution Control Association, Detroit, MI.
Staehelin, J., C. Keller, W. Stahel, K. Schlapfer and S. Wunderli (1998). Emission factors from
road traffic from a tunnel study (G ubrist tunnel, Switzerland) Part III: Results of
organic compounds, SO2 and speciation of organic exhaust emission. Atmos. Environ.,
32(6): 999-1010.
Stauffer, D.R. andN.L. Seaman (1994). Multiscale four-dimensional data assimilation. J. Appl.
MeteoroL, 33(3): 416-434.
-------
Stedman, D.H. (1992). Comments on "Speciated measurements and calculated reactivities of
vehicle exhaust emissions from conventional and reformulated gasolines". Environ. Sci.
Technol, 26(10): 2036.
Stelson, A.W. and J.H. Seinfeld (1982a). Relative humidity and pH dependence of the vapor
pressure of ammonium nitrate-nitric acid solutions at 25° C. Atmos. Environ., 16(5): 993-
1000.
Stelson, A.W. and J.H. Seinfeld (1982b). Relative humidity and temperature dependence of the
ammonium nitrate dissociation constant. Atmos. Environ., 16(5): 983-992.
Stelson, A.W. and J.H. Seinfeld (1982c). Thermodynamic prediction of the water activity,
NH4NO3 dissociation constant, density and refractive index for the MJ^M^^NH^SO^
H2O system at 25° C. Atmos. Environ., 16(10): 2507-2514.
Street, R.A., S. Owen, S.C. Duckham, C. Boissard and C.N. Hewitt (1997). Effect of habitat and
age on variations in volatile organic compound (VOC) emissions from Quercus ilex and
Pinuspinea. Atmos. Environ., 31(S1): 89-100.
Stump, F.D., S.B. Tejada, W.D. Ray, D.L. Dropkin, F. Black, W.B. Crews, R. Snow, P. Siudak,
C.O. Davis, L. Baker and N. Perry (1989). The influence of ambient temperature on
tailpipe emissions from 1984-1987 model year light-duty gasoline motor vehicles.
Atmos. Environ., 23: 307-320.
Stump, F.D., K.T. Knapp and W.D. Ray (1990a). Seasonal impact of blending oxygenated
organics with gasoline on motor vehicle tailpipe and evaporative emissions. JAWMA,
40(6): 872-880.
Stump, F.D., S.B. Tejada, W.D. Ray, D.L. Dropkin, F. Black, R. Snow, W.B. Crews, P. Sivdak,
C.O. Davis and P.L. Carter (1990b). The influence of ambient temperature on tailpipe
emissions from 1985 to 1987 model year light-duty gasoline motor vehicles - II. Atmos.
Environ., 24A: 2105-2112.
Stump, F.D., K.T. Knapp, W.D. Ray, R. Snow and C.S. Burton (1992). The composition of
motor vehicle organic emissions under elevated temperature summer driving conditions
(75 to 105°F). JAWMA, 42(2): 152-158.
Stump, F.D., K.T. Knapp and W.D. Ray (1996). Influence of ethanol-blended fuels on the
emissions from three pre-1985 light-duty passenger vehicles. JAWMA, 46(12): 1149-
1161.
Tang, IN. and H.R. Munkelwitz (1993). Composition and temperature dependence of the
deliquescence properties of hygrosopic aerosols. Atmos. Environ., 27A(4): 467-473.
Tanner, R.L. and B. Zielinska (1994). Determination of the biogenic emission rates of species
contributing to VOC in the San Joaquin Valley of California. Atmos. Environ., 28(6):
1113-1120.
8-28
-------
Tingey, D.T., H.C. Ratsch, M. Manning, L.C. Grothaus, W.F. Burns and E.W. Peterson (1978).
Isoprene emission rates from live oak. Report No. EPS-904/9-78-004. Prepared by U.S.
Environmental Protection Agency, Washington, DC.
Tingey, D.T. (1981). The effect of environmental factors on the emission of biogenic
hydrocarbons from live oak and slash pine. In Atmospheric Biogenic Hydrocarbons., J. J.
Bufalini and R.R. Arnts, Eds. Ann Arbor Science, Ann Arbor, MI, pp. 53-72.
Tingey, D.T., R.A. Evans and M. Gumpertz (1981). Effects of environmental conditions on
isoprene emission from live oak. Planta, 152: 565-570.
Trier, C.J., D.S. Coates, D.E. Fussey and M.M. Rhead (1990). The contribution of lubricating
oil to diesel exhaust emissions. Sci. Total Environ.: 167-174.
Venkataraman, C. and S.K. Friedlander (1994). Source resolution of fine particulate polycyclic
aromatic hydrocarbons using a receptor model modified for reactivity. JAWMA, 44(9):
1103-1108.
Vermette, S.J., A.L. Williams and S. Landsberger (1992). PM10 source apportionment using
local surface dust profiles: Examples from Chicago. In Transactions, PM10 Standards
andNontraditionalParticulate Source Controls, J.C. Chow and D.M. Ono, Eds. Air and
Waste Management Association, Pittsburgh, PA, pp. 262-271.
Vong, R.J., P. Geladi, S. Wold and K. Esbensen (1988). Source contributions to ambient aerosol
calculated by discriminant partial least squares regression (PLS). J. Chemometrics, 2:
281-296.
Wadden, R.A., I. Uno and S. Wakamatsu (1986). Source discrimination of short-term
hydrocarbon samples measured aloft. Environ. Sci. Technol, 20(5): 473-483.
Wadden, R.A., P.A. Scheff, I.E. Franke, L.M. Conroy and C.B. Keil (1995a). Determination of
VOC emission rates and compositions for offset printing. JAWMA, 45(7): 547-555.
Wadden, R.A., P.A. Scheff, I.E. Franke, L.M. Conroy, M. Javor, C.B. Keil and S.A. Milz
(1995b). VOC emission rates and emission factors for a sheetfed offset printing shop. J.
Am. Ind. Hyg. Assoc., 56(4): 368-376.
Walling, J.F., I.E. Bumgarner, DJ. Driscoll, C.M. Morris, A.E. Riley and L.H. Wright (1986).
Apparent reaction products desorbed from Tenax used to sample ambient air. Atmos.
Environ., 20: 51-57.
Wallington, TJ. and S.M. Japar (1991). Atmospheric chemistry of diethyl ether and ethyl tert-
butyl ether. Environ. Sci. Technol., 25: 410.
Wallington, T.J., J.M. Andino, A.R. Potts, SJ. Rudy, W.O. Siegl, Z. Zhang, MJ. Kurylo and
R.E. Huie (1993). Atmospheric chemistry of automotive fuel additives: Diisopropyl
ether. Environ. Sci. Technol., 27(1): 98-104.
8-29
-------
Wang, T., K.S. Lam, L.Y. Chan, A.S.Y. Lee and M.A. Carroll (1997). Trace gas measurements
in coastal Hong Kong during the PEM-West B. J. Geophys. Res., 102(D23): 28575-
28588.
Ward, D.E. and C.C. Hardy (1989). Organic and elemental profiles for smoke from prescribed
fires. In Transactions, Receptor Models in Air Resources Management., J.G. Watson, Ed.
Air & Waste Management Association, Pittsburgh, PA, pp. 299-321.
Watson, J.G. (1979). Chemical element balance receptor model methodology for assessing the
sources of fine and total suspended particulate matter in Portland, Oregon. Ph.D.
Dissertation, Oregon Graduate Center, Beaverton, OR, Beaverton, OR.
Watson, J.G. (1984). Overview of receptor model principles. JAPCA, 34: 619-623.
Watson, J.G., J.A. Cooper and J.J. Huntzicker (1984). The effective variance weighting for least
squares calculations applied to the mass balance receptor model. Atmos. Environ., 18(7):
1347-1355.
Watson, J.G., J.C. Chow, L.W. Richards, S.R. Andersen, J.E. Houck and D.L. Dietrich (1988).
The 1987-88 Metro Denver Brown Cloud Air Pollution Study, Volume III: Data
interpretation. Report No. DRI 8810.1. Prepared for 1987-88 Metro Denver Brown
Cloud Study, Inc., Greater Denver Chamber of Commerce, Denver, CO, by Desert
Research Institute, Reno, NV.
Watson, J.G., N.F. Robinson, J.C. Chow, R.C. Henry, B.M. Kim, T.G. Pace, E.L. Meyer and Q.
Nguyen (1990a). The USEPA/DRI chemical mass balance receptor model, CMB 7.0.
Environ. Software, 5(1): 38-49.
Watson, J.G., J.C. Chow, L.C. Pritchett, J.E. Houck, S. Burns and R.A. Ragazzi (1990b).
Composite source profiles for particulate motor vehicle exhaust source apportionment in
Denver, CO. In Transactions, Visibility and Fine Particles, C.V. Mathai, Ed. Air&
Waste Management Association, Pittsburgh, PA, pp. 422-436.
Watson, J.G., J.C. Chow and T.G. Pace (1991). Chemical mass balance. In Receptor Modeling
for Air Quality Management, P.K. Hopke, Ed. Elsevier Press, New York, NY, pp. 83-
116.
Watson, J.G. and J.C. Chow (1992). Data bases for PM10 and PM25 chemical compositions and
source profiles. In PM10 Standards andNontraditionalParticulate Source Controls, J.C.
Chow and D.M. Ono, Eds. Air & Waste Management Association, Pittsburgh, PA, pp.
61-91.
Watson, J.G. and J.C. Chow (1994). Clear sky visibility as a challenge for society. Annual Rev.
Energy Environ., 19: 241-266.
Watson, J.G., J.C. Chow, F.W. Lurmann and S. Musarra (1994a). Ammonium nitrate, nitric
acid, and ammonia equilibrium in wintertime Phoenix, Arizona. JAWMA, 44(4): 405-
412.
8-30
-------
Watson, J.G., J.C. Chow, E.M. Fujita, C.A. Frazier, Z. Lu, S.L. Heisler and C.T. Moore (1994b).
Aerosol data validation for the 1992-93 Tucson Urban Haze Study. In 87th Annual
Meeting of the Air & Waste Management Association, Cincinnati, OH.
Watson, J.G., J.C. Chow, Z. Lu, E.M. Fujita, D.H. Lowenthal and D.R. Lawson (1994c).
Chemical mass balance source apportionment of PM10 during the Southern California Air
Quality Study. AerosolSci. Technol, 21: 1-36.
Watson, J.G., J.C. Chow, D.H. Lowenthal, L.C. Pritchett, C.A. Frazier, G.R. Neuroth and R.
Robbins (1994d). Differences in the carbon composition of source profiles for diesel-
and gasoline-powered vehicles. Atmos. Environ., 28(15): 2493-2505.
Watson, J.G., P.J. Lioy and P.K. Mueller (1995). The measurement process: Precision,
accuracy, and validity. In Air Sampling Instruments for Evaluation of Atmospheric
Contaminants, 8th ed., B. Cohen and S.V. Hering, Eds. American Conference of
Governmental Industrial Hygienists, Cincinnati, OH, pp. 187-194.
Watson, J.G., D.L. Blumenthal, J.C. Chow, C.F. Cahill, L.W. Richards, D. Dietrich, R. Morris,
J.E. Houck, R.J. Dickson and S.R. Andersen (1996a). Mt. Zirkel Wilderness Area
reasonable attribution study of visibility impairment - Vol. II: Results of data analysis
and modeling. Prepared for Colorado Department of Public Health and Environment,
Denver, CO, by Desert Research Institute, Reno, NV.
Watson, J.G., J.C. Chow, J.A. Gillies, H. Moosmuller, C.F. Rogers, D.W. DuBois and J.C.
Derby (1996b). Effectiveness demonstration of fugitive dust control methods for public
unpaved roads and unpaved shoulders on paved roads. Report No. 685-5200. IF.
Prepared for San Joaquin Valley Unified Air Pollution Control District, Fresno, CA, by
Desert Research Institute, Reno, NV.
Watson, J.G, N.F. Robinson, C.W. Lewis, C.T. Coulter, J.C. Chow, E.M. Fujita, D.H.
Lowenthal, T.L. Conner, R.C. Henry and R.D. Willis (1997a). Chemical mass balance
receptor model version 8 (CMB) user's manual. Prepared for U.S. Environmental
Protection Agency, Research Triangle Park, NC, by Desert Research Institute, Reno, NV.
ftp: //eafs.sage.dri.edu/cmb80/model/.
Watson, J.G, J.C. Chow, D.W. DuBois, M.C. Green, N.H. Frank and M.L. Pitchford (1997b).
Guidance for network design and optimal site exposure for PM2 5 and PM10. Prepared for
U.S. EPA, Office of Air Quality Planning and Standards, Research Triangle Park, NC by
Desert Research Institute, Reno, NV.
Watson, J.G., E.M. Fujita, J.C. Chow, B. Zielinska, L.W. Richards, W.D. Neff and D. Dietrich
(1998). Northern Front Range Air Quality Study. Final Report. Prepared for Colorado
State University, Fort Collins, CO, by Desert Research Institute, Reno, NV.
www.nfraqs.colostate.edu/nfraqs/index2.html
Watson, J., J.C. Chow and E. Fujita (2001). Review of Volatile Organic Compound Source
Apportionment by Chemical Mass Balance. Atmos. Environ. 35: 1567-1584.
-31
-------
Weingartner, E., C. Keller, W.A. Stahel, H. Burtscher and U. Baltensperger (1997). Aerosol
emission in a road tunnel. Atmos. Environ., 31(3): 451-462.
Westerholm, R., R. Stenberg and J. Alsberg (1988). Some aspects of the distribution of
polycyclic aromatic hydrocarbons (PAH) between particles and gas phase from diluted
gasoline exhausts, generated with the use of a dilution tunnel, and its validity for
measurement in ambient air. Atmos. Environ., 22: 1005-1010.
White, W.H. and E.S. Macias (1991). Chemical mass balancing with ill defined sources:
Regional apportionment in the California desert. Atmos. Environ., 25A(8): 1547-1557.
Williams, R.L., F. Lipari and R.A. Potter (1990). Formaldehyde, methanol and hydrocarbon
emissions from methanol-fueled cars. JAWMA, 40(5): 747-756.
Williamson, A.D. and W.B. Smith (1979). A dilution sampling system for condensation aerosol
measurement: Design specification. Report No. SORI-EAS-79-793. Prepared by
Southern Research Institute.
Williamson, HJ. and D.A. Dubose (1983). Receptor model technical series, Volume III: User's
manual for chemical mass balance model. Report No. EPA-450/4-83-014. Prepared by
U.S. Environmental Protection Agency, Research Triangle Park, NC.
Winchester, J.W. and G.D. Nifong (1971). Water pollution in Lake Michigan by trace elements
from aerosol fallout. Water Air and Soil Pollution, 1:50-64.
Winer, A.M., J.B. Arey, R. Atkinson, S.M. Aschmann, W.D. Long, C.L. Morrison and D.M.
Olszyk (1992). Emission rates of organics from vegetation in California's Central Valley.
Atmos. Environ., 26A(14): 2647.
Yamartino, R.J., J.S. Scire, G.R. Carmichael and Y.S. Chang (1992). The CALGRID Mesoscale
Photochemical Grid Model-I. Model Formulation. Atmos. Environ., 26A(8): 1493-1512.
Yokouchi, Y. and Y. Ambe (1988). Diurnal variations of atmospheric isoprene and monoterpene
hydrocarbons in an agricultural area in summertime. J. Geophys. Res., 93: 3751-3759.
Young, V.L., B.N. Kieser, S.P. Chen and H. Niki (1997). Seasonal trends and local influences
on nonmethane hydrocarbon concentrations in the Canadian boreal forest. J. Geophys.
Res., 102(D5): 5913-5918.
Zelenka, M.P., P.J. Lioy, H. Qing-Ci and W.E. Wilson (1992). Development of a new
methodology for characterizing air pollution sources in China. In Transactions, PM10
Standards andNontraditionalParticulate Source Controls., J.C. Chow and D.M. Ono,
Eds. Air and Waste Management Association, Pittsburgh, PA, pp. 163-182.
Zhang, S.H., M.A. Shaw, J.H. Seinfeld and R.C. Flagan (1992). Photochemical aerosol
formation from -pinene- and -pinene. J. Geophys. Res., 97(D18): 20717.
-32
-------
Zielinska, B., J.B. Arey, R. Atkinson, T. Ramdahl, A.M. Winer and J.N. Pitts, Jr. (1986).
Reaction of dinitrogen pentoxide with fluoranthene. J. Am. Chern. Soc., 108: 4126-4132.
Zielinska, B., J.B. Arey and R. Atkinson (1990). The atmospheric formation of nitroarenes and
their occurrence in ambient air. In Proceedings of the Fourth International Conference
on N- Substituted Aryl Compounds: Occurrence, Metabolism and Biological Impact of
Nitroarenes, Cleveland, OH, 15-19 July 1989, Howard et al., Ed. Plenum Press, New
York, pp. 73-84.
Zielinska, B. and K.K. Fung (1994). The composition and concentration of hydrocarbons in the
range of C2 to C18 emitted from motor vehicles. Sci. Total Environ., 146/147: 281-288.
Zielinska, B. and E.M. Fujita (1994a). The composition and concentration of hydrocarbons in
the range of C2 to C18 in downtown Los Angeles, CA. Res. Chem. Intermed., 20: 321-
334.
Zielinska, B. and E.M. Fujita (1994b). Organic gas sampling. In Environmental Sampling for
Trace Analysis, B. Markert, Ed. VCH Publishers, New York, NY, pp. 163-184.
Zielinska, B., J.C. Sagebiel, L.H. Sheetz, G. Harshfield, E. Uberna, W.J. Hauze and J.H. Price
(1994). Hydrocarbons in the C8 - C20 range measured during COAST Study in Texas. In
Proceedings, U.S. EPA/A&WMA International Symposium on Measurement of Toxic and
Related Air Pollutants. Air & Waste Management Association, Pittsburgh, PA, pp. 714-
719.
Zielinska, B., J.C. Sagebiel, G. Harshfield, A.W. Gertler and W.R. Pierson (1996). Volatile
organic compounds up to C20 emitted from motor vehicles: Measurement methods.
Aimos. Environ., 30(12): 2269-2286.
Zielinska, B., J.D. McDonald, T. Hayes, J.C. Chow, E.M. Fujita and J.G. Watson (1998).
Northern Front Range Air Quality Study. Volume B: Source measurements. Prepared
for Colorado State University, Fort Collins, CO, by Desert Research Institute, Reno, NV.
http: //charon.cira.colostate.edu/DRIFinal/ZipFiles/.
Zweidinger, R.B., J.E. Sigsby, Jr., S.B. Tejada, F.D. Stump, D.L. Dropkin, W.D. Ray and J.W.
Duncan (1988). Detailed hydrocarbon and aldehyde mobile source emissions from
roadway studies. Environ. Sci. Technol., 22: 956-962.
Zweidinger, R.B., R.K. Stevens, C.W. Lewis and H.H. Westberg (1990). Identification of
volatile hydrocarbons as mobile source tracers for fme-particulate organics. Environ. Sci.
Technol., 24: 538-542.
-33
-------
-------
APPENDIX A.
54 PAMS TARGET COMPOUNDS (HYDROCARBONS) LISTED IN
THEIR ELUTION SEQUENCE.
1.
2.
3.
4.
5
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54
Hydrocarbon
Ethylene
Acetylene
Ethane
Propylene
Propane
Isobutane
1-Butene1
n-Butane
/-2-Butene
c-2-Butene
Isopentane
1-Pentene
n-Pentane
Isoprene
/-2-Pentene
c-2-Pentene
2,2-Dimethylbutane
Cyclopentane
2,3 -Dimethylbutane
2-Methylpentane
3-Methylpentane
n-Hexane
Methylcyclopentane
2,4-Dimethylpentane
Benzene
Cyclohexane
2-Methylhexane
2,3 -Dimethylpentane
3 -Methy Ihexane
2,2,4-Trimethylpentane
n-Heptane
Methylcyclohexane
2,3 ,4-Tnmethylpentane
Toluene
2-Methylheptane
3-Methylheptane
n-Octane
Ethylbenzene
m &/>-Xylene2
Styrene
o-Xylene
n-Nonane
Isopropylbenzene
n-Propylbenzene
/w-Etnyltoluene
/>-Ethyltoluene
1 ,3 ,5-Trimethylbenzene
o-Ethyltoluene
1 ,2,4-Trimethylbenzene
n-Decane
1 ,2,3 -Trimethylbenzene
/w-Diethylbenzene3
/>-Diethylbenzene
n-Undecane
AIRS
43203
43206
43202
43205
43204
43214
43280
43212
43216
43217
43221
43224
43220
43243
43226
43227
43244
43242
43284
43285
43230
43231
43262
43247
45201
43248
43263
43291
43249
43250
43232
43261
43252
45202
43960
43253
43233
45203
45109
45220
45204
43235
45210
45209
45212
45213
45207
45211
45208
43238
45225
45218
45219
43954
CAS
74851
74862
74840
115071
74986
75285
106989
106978
624646
590181
78784
109671
109660
78795
646048
627203
75832
287923
79298
107835
96140
110543
96377
108087
71432
110827
591764
565593
589344
540841
142825
108872
565753
108883
592278
589811
111659
100414
108383/106423
100425
95476
111842
98828
103651
620144
622968
108678
611143
95636
124185
526738
141935
105055
1120214
'Note that because 1-Butene and Isobutene elute at about the same time, they are difficult to resolve. The coeluting isomers are
assigned AIRS Parameter Code 43127. Isobutene is assigned AIRS Parameter Code 43270 and CAS 115117.
2These isomers of xylene are also difficult to resolve. Individually, their AIRS Parameter Codes are 45205 & 45206, respectively.
Respective CAS numbers are provided in the table.
3Also named 1,3-Diethylbenzene.
A- 1
-------
-------
APPENDIX B. NORMALIZATION FOR THE VOC SOURCE PROFILE
As we stressed in Section 3.1.2, the different ways in which source profiles are
constructed make it difficult in practice to compare and use VOC profiles from different studies.
Furthermore, inventories employ different conventions for defining VOC. Meaningful
comparison of CMB results with emission inventories requires a common reference. In this
discussion we explain how normalized profiles are constructed for input to CMB, and also point
out some differences to be seen in different normalization approaches.
In Table B-l are shown data collected from the 1990 Atlanta Ozone Precursor
Monitoring Study, in which motor vehicle emissions were sampled in canisters beside a roadway
in a tunnel-like underpass during periods of heavy traffic. Roadway tunnels are usually good
integrators of the mobile source fleet. In the first column is the average measured concentration
for the top 83 NMOC appearing in the chromatogram, based on 9 samples. Two additional
species, n-Tridecane and n-Tetradecane, were included to be consistent with the airport and
aircraft profiles also collected in this study. The average sum of the 85 peaks is 1949.4, while
the sum of all quantified NMOC chromatogram peaks is 2335.7 ppbC. Note that the example
presented here - in which multiple samples were taken of a single source type for profile
characterization, and subjected to averaging - is ideal. In some cases, only a single sample may
be taken. It is important to understand exactly how the values appearing in Table B-l were
derived:
E cik
Average Concentration for species /' = c. = — - , where B-l
n
cik = measured concentration (ppbC) for species /' in the chromatogram for sample &, and
n = number of samples available for constructing this source profile.
Fractional abundance of species /' = f. = — - , where B-2
n
c.,
fik = - (where summationk = sum of the concentration peaks for sample k B-3
summation^ used in the normalization).
n
k=l B-4
E Vrti
Uncertainty for fractional abundance of species /' = o-f = .
(the standard deviation of the mean fractional abundance determined from n samples).
In applying Eqns B-2, B-3, and B-4, professional judgement should of course be used in
the matter of potential outliers. For example, if for a particular sample k, fik is greatly different
from f., especially for multiple species /', the sample k should probably be deleted from the
calculations. For the VOC source profile presented in Table B-l, the uncertainties reported for
B- 1
-------
the mean fractional abundances were statistically analyzed. The mean is 12.7% and the median
is 8.9%, with a minimum uncertainty of 2.6% (ethylbenzene) and a maximum uncertainty of
73.5% (2,3,3 trimethyl-1-butene). As indicated by these statistics, the routine practice of some
laboratories to provide only a single, overall assessment of uncertainty/precision for an analysis
is an oversimplification.
Ideally, the fractional abundances for each species / are determined individually for each
source sample, and then averaged for the suite of samples taken. Note that the mean fractional
abundance wasn't simply determined by taking the mean concentration for species /', and then
dividing by the mean summation for the suite of samples, i.e.,
1 "
- - £ cik
ft + ' = ^-ti B-5
summation 1 "
— £ summation,
n k=\
Equations B-2 and B-5 are not equivalent and the value of f. determined via Eqn. B-5 in fact
may be considerably biased. Note also that the uncertainty wasn't simply determined as the
standard deviation of the mean concentration of species /' for the suite of samples taken, i.e.,
Unc. * a- =
\
n - 1
Equations B-4 and B-6 are not equivalent. Eqn. B-6, patterned after Eqn. B-l, is inappropriate
because it isn't normalized (i.e., does not account for variations in summation^). And because
Eqn. B-5 is not a robust estimate of relative species abundance, the associated uncertainty (a- )
is not an accurate reflection of the uncertainty for species /'.
As stated in Section 3.1.2, speciated VOC data are sometimes normalized to all the
compounds in the chromatogram, as shown in the first set of paired columns in Table B-l. In
fact, the values appearing in this set of columns are exactly as reported in Table 1 of the cited
JAPCA article, except that the latter were expressed as percentages. Normalized fractional
abundances are often presented as percentages because they are easier to read (fewer leading
decimal places). However, CMB requires that abundances be in fractional form; the ratios
represented in Table B-l are appropriate for the source profile input file required by CMB. In
this case, the value for summationk in Eqn. B-3, with which the fractional abundances (fik) were
determined, included the sum of all peaks in the chromatogram.
The 2nd set of paired columns represents a normalization to only the PAMS
(Photochemical Assessment Monitoring Stations) species. Here, the ratio was constructed as
before, but the value for summationk in Eqn. B-3 included only the sum of the measured PAMS
peaks.
B-2
-------
Note that since the ratios appearing in column 2 include only 85 NMOC species, the sum
of their fractional abundances normalized to total NMOC is less than unity. Conversely, since
the ratios appearing in the 4th column are based on the sum of only the PAMS species, the sum of
fractional abundances is greater than unity. Of course, had only PAMS species been sampled
and measured, the sum would theoretically be unity in the 4th column. There is a 35% difference
in fractional abundances normalized via the two approaches shown here. The ratio of 1.123 /
0.833 is 1.35, which is precisely the same ratio as 2335.7 / 1735.7 (sum of all measured peaks /
sum of PAMS peaks). This relationship is shown most clearly in the last entry (total NMOC), in
which the respective normalized values clearly reflect the 35% difference. It is worth noting that
many published source profiles, including the source from which these data were reproduced,
include an entry for total NMOC. When published profiles include a normalized value for total
NMOC, the normalization approach becomes obvious. Beyond its practical purposes, it is useful
to include an entry for total NMOC in the source profile input file for CMB provided it is
normalized to the sum of the measured PAMS species. Ideally, the uncertainty for the
normalized total NMOC abundance should be calculated as:
= A/(°7.)2+ (o-f. )2+ (o-f )2+ ... + (o7 )2
y •* i -M+I "* i+ L J m
Unc-TNMOC = °2+ o- 2+ o- 2+ ... + o 2, where
o>= normalized uncertainly computed via Eqn. B-4, and m = total number of NMOC species
measured. For convenience, the value for this uncertainty reported in Table B-l was based on
m = 85 (only the 85 values in column 5 were used). If possible, all measured NMOC species
should be used in this calculation.
We mentioned that the VOC source type characterized for this example is in many ways
ideal. Roadway sources such as this one are likely to reflect most, if not all, of the PAMS
species. However, many VOC source types may be sampled in which all or even most of the
PAMS species will not be represented. So long as the sampling system is capable of collecting
all of the PAMS species, and so long as the analytical system is capable of detecting all of the
PAMS species, the normalization procedure described in Eqns. B-l through B-4 should be
followed. The value for n will reflect the number of samples taken, and the value for summationk
will include the sum of as many PAMS peaks measured in the chromatogram for any sample k.
B-
-------
Table B-l. Measured VOC concentrations and their normalized fractional abundances.1
(Species flagged with '•' are PAMS target compounds listed in Appendix A)
Average Concentration Ratio to
VOC Species (ppbC) Abundance
• Ethylene
• Acetylene
• Ethane
• Propene
• Propane
• Isobutane
• 1-Butene/Isobutene
1,3 -Butadiene
• n-Butane
/-2-Butene
• c-2-Butene
3-Methyl-l-Butene
• Isopentane
• 1-Pentene
2-Methyl-l-Butene
• n-Pentane
• Isoprene
• /-2 -Pentene
• c-2-Pentene
2-Methyl-2-Butene
• 2,2-Dimethylbutane
Cyclopentene
4-Methyl- 1 -Pentene
• Cyclopentane
• 2,3-Dimethylbutane
• 2-Methylpentane
• 3-Methylpentane
2 -Methyl- 1 -Pentene
1-Hexene
• n-Hexane
/-3-Hexene
/-2-Hexene
c-2-Hexene
• Methylcyclopentane
• 2,4-Dimethylpentane
2,3 ,3 -Trimethyl- 1 -Butene
• Benzene
• Cyclohexane
• 2-Methylhexane
• 2,3-Dimethylpentane
• 3-Methylhexane
c- 1 ,3 -Dimethy Icy clopentane
3-Etliylpentane
• 2,2,4-Trimethylpentane
• n-Heptane
• Methylcyclohexane
2,5-Dimethylhexane
2,4-Dimethylhexane
• 2,3,4-Trimethylpentane
• Toluene
99.77
86.24
36.34
45.11
24.94
27.16
27.67
9.34
98.49
9.70
8.02
3.74
204.13
7.78
15.26
63.64
7.89
17.78
9.79
23.16
12.07
4.06
4.10
5.86
20.19
57.25
33.32
4.32
3.13
25.68
6.11
5.85
2.99
18.37
16.62
8.49
63.13
3.83
20.44
21.29
20.69
4.49
6.98
57.90
12.68
7.08
7.58
12.82
22.06
153.15
0.043400
0.038000
0.015500
0.019600
0.010500
0.011200
0.011810
0.003800
0.041100
0.004080
0.003340
0.001583
0.086400
0.003200
0.006360
0.026600
0.003200
0.007500
0.004110
0.009300
0.004950
0.001720
0.001790
0.002508
0.008630
0.024340
0.014180
0.001830
0.001338
0.010880
0.002580
0.002490
0.001270
0.007830
0.007040
0.003400
0.027300
0.001660
0.008740
0.009010
0.008880
0.001920
0.002980
0.025100
0.005400
0.003010
0.003290
0.005490
0.009500
0.065900
TNMOC
Unc.2
0.004700
0.006400
0.001900
0.001800
0.001800
0.002500
0.000900
0.001100
0.007100
0.000280
0.000300
0.000091
0.008400
0.001600
0.000530
0.002700
0.001700
0.000450
0.000290
0.002900
0.000850
0.000084
0.000016
0.000077
0.000250
0.000990
0.000620
0.000180
0.000055
0.000580
0.000570
0.000210
0.000110
0.000380
0.000580
0.002500
0.001900
0.000110
0.000530
0.000600
0.000420
0.000130
0.000150
0.002300
0.000380
0.000170
0.000330
0.000440
0.001100
0.003600
Ratio to PAMS
Abundance Unc.2
0.058590
0.051300
0.020925
0.026460
0.014175
0.015120
0.015944
0.005130
0.055485
0.005508
0.004509
0.002137
0.116640
0.004320
0.008586
0.035910
0.004320
0.010125
0.005549
0.012555
0.006683
0.002322
0.002417
0.003386
0.011650
0.032859
0.019143
0.002471
0.001806
0.014688
0.003483
0.003362
0.001715
0.010571
0.009504
0.004590
0.036855
0.002241
0.011799
0.012164
0.011988
0.002592
0.004023
0.033885
0.007290
0.004064
0.004442
0.007411
0.012825
0.088965
0.006345
0.008640
0.002565
0.002430
0.002430
0.003375
0.001215
0.001485
0.009585
0.000378
0.000405
0.000123
0.011340
0.002160
0.000716
0.003645
0.002295
0.000608
0.000392
0.003915
0.001148
0.000113
0.000216
0.000104
0.000338
0.001336
0.000837
0.000243
0.000074
0.000783
0.000769
0.000284
0.000149
0.000513
0.000783
0.003375
0.002565
0.000149
0.000716
0.000810
0.000567
0.000175
0.000203
0.003105
0.000513
0.000230
0.000446
0.000594
0.001485
0.004860
B-4
-------
2,3 -Dimethylhexane
• 2-Methylheptane
• 3-Methylheptane
2,2,5-Trimethylhexane
• n-Octane
2,5-Dimethylheptane
• Etylbenzene
• m/p-Xylene
4-Methyloctane
3-Methyloctane
• Styrene
• o-Xylene
• n-Nonane
• Isopropylbenzene
• n-Propylbenzene
• ffj-Ethyltoluene
• />-Ethyltoluene
• 1,3,5-Trimethylbenzene
• o-Etyltoluene
• 1,2,4-Trimethylbenzene
• n-Decane
• 1,2,3-Trimethylbenzene
n-Butylcyclohehexane
• /w-Diethylbenzene (1,3-Diethylbenzene)
• />-Diethylbenzene (1,4-Diethylbenzene)
1 ,3 -Dimethyl-4-Ethylbenzene
• n-Undecane
1,2,4,5-Tetramethylbenzene
1 ,2,3 ,5-Tetramethylbenzene
ffj-Diisoproplylbenzene
1 ,2,3 ,4-Tetramethylbenzene
napthalene
n-Dodecane
n-Tridecane
n-Tetradecane
Total NMOC
6.57
7.51
8.30
8.51
6.68
4.37
29.93
101.75
5.70
4.29
9.89
38.73
5.07
3.60
8.10
31.85
14.68
17.21
11.40
49.49
6.05
12.23
5.55
4.37
17.25
5.52
5.57
6.53
7.99
4.20
4.69
18.53
5.64
2.11
1.01
2335.66
0.002830
0.003240
0.003610
0.003720
0.002860
0.001888
0.012800
0.043500
0.002460
0.001860
0.004370
0.016620
0.002140
0.001530
0.003530
0.013810
0.006330
0.007490
0.004990
0.021600
0.002550
0.005320
0.002270
0.001890
0.007550
0.002360
0.002420
0.002720
0.003800
0.001570
0.001720
0.008130
0.002470
0.000922
0.000420
S: (0.833)
1.000000
0.000350
0.000240
0.000440
0.000420
0.000120
0.000084
0.000330
0.001400
0.000220
0.000130
0.000530
0.000800
0.000250
0.000150
0.000320
0.000900
0.000340
0.000600
0.000450
0.002200
0.000400
0.000420
0.000330
0.000110
0.000650
0.000220
0.000240
0.000500
0.001000
0.000930
0.000890
0.000810
0.000280
0.000098
0.000110
0.003820
0.004374
0.004874
0.005022
0.003861
0.002549
0.017280
0.058725
0.003321
0.002511
0.005900
0.022437
0.002889
0.002066
0.004766
0.018644
0.008546
0.010112
0.006736
0.029160
0.003443
0.007182
0.003065
0.002551
0.010193
0.003186
0.003267
0.003672
0.005130
0.002119
0.002322
0.010976
0.003334
0.001245
0.000567
(1.12)
1.346
0.000472
0.000324
0.000594
0.000567
0.000162
0.000113
0.000446
0.001890
0.000297
0.000175
0.000716
0.001080
0.000338
0.000203
0.000432
0.001215
0.000459
0.000810
0.000608
0.002970
0.000540
0.000567
0.000446
0.000149
0.000878
0.000297
0.000324
0.000675
0.001350
0.001256
0.001202
0.001094
0.000378
0.000132
0.000149
Mean summation top 85 VOC peaks: 1949.35
Mean summation PAMS peaks: 1735.72
Difference, TNMOC normalization vs. PAMS normalization: 34.6%
1 Data used with permission from Conner, T.L., W.A. Lonneman, and R.L Seila, 1995. Transportation-related volatile hydrocarbon source
profiles measured in Atlanta. JAWA4A45: 383-394.
2 Statistical analysis of these uncertainties is presented and discussed in the text.
B-5
-------
-------
APPENDIX C. INTERNET LINKS TO MODELING SOFTWARE AND DATA SETS
C. 1 Receptor Model s
• EPA-CMB8.2: www.epa.gov/scram001/
C.2 Source Profiles
• SPECIATE: www.epa.gov/ttn/chief/software/speciate
• Northern Front Range Air Quality Study Source Profiles (Particle Organics):
www.nfraqs.colostate.edu/nfraqs/index2.html
C.3 Ambient Measurements
• Center for Air Pollution Impact and Trend Analysis: capita.wustl.edu/
• IMPROVE Particle Measurements: www.epa.gov/ttn/amtic/visdata.html
C.4 Emissions Models and Inventories
• AP-42 Emissions Factors: www.epa.gov/ttn/chief/ap42/
• National Emissions Trends Report: www.epa.gov/ttn/chief/trends/trends98/
• National Emissions Inventory Data: www.epa.gov/ttn/chief/net/
C.5 Meteorological Models
• CALMET: www.epa.gov/scram001/
• MM5 (5th-Generation Penn State/NCAR Mesoscale Model):
www.mmm.ucar.edu/mm5/mm5v3/v3model.html
• WRF (Weather Research and Forecasting model): www.wrf-model.org
• National Centers for Environmental Prediction (NCEP) Products:
http://nomads.ncdc.noaa.gov/data-access.html
GFS (Global Forecast System)
HR GFS (Hign Resolution Global Forecast System)
Eta (regional mesoscale model)
NAM (North American Mesoscale)
NARR (North American Regional Reanalysis)
RUC (Rapid Update Cycle)
C-l
-------
C.6 Dispersion Models
• AERMOD: www.epa.gov/scram001
• CALPUFF: www.epa.gov/scram001/
• CMAQ (Community Multi-Scale Air Quality Model):
www.epa.gov/asmdnerl/models3/
C-2
-------
APPENDIX D. CMB MATHEMATICS
The source contribution (Sj) present at a receptor during a sampling period of length T
due to a source] with constant emission rate Ej is
SJ = DJ * EJ
where:
f 4«(0, 0(0, Xj]dt (D-2)
is a dispersion factor depending on wind velocity (u), atmospheric stability (a), and the location
of source j with respect to the receptor (Xj). All parameters in Equation A-2 vary with time, so
the instantaneous dispersion factor, Dj, must be integrated over time period T (Watson, 1979).
Various forms for Dj have been proposed (Pasquill, 1974; Benarie, 1976; Seinfeld and
Pandis, 1998), some including provisions for chemical reactions, removal, and specialized
topography. None are completely adequate to describe the complicated, random nature of
dispersion in the atmosphere. The advantage of receptor models is that an exact knowledge of Dj
is unnecessary.
If a number of sources, J, exists and there is no interaction between their emissions to
cause mass removal, the total mass measured at the receptor, C, will be a linear sum of the
contributions from the individual sources.
j J
C = E Dj • Ej = E Sj (D-3)
j= i J= i
Similarly, the concentration of elemental component i, Q will be
C. = E FfJ • Sj i = 1, ... / (D-4)
y=i
where: F;j = the fraction of source contribution Sj composed of element i. The number of
chemical species (I) must be greater than or equal to the number of sources (J) for a unique
solution to these equations.
Solutions to the CMB equations consist of: (1) a tracer solution; (2) a linear
D- 1
-------
programming solution; (3) an ordinary weighted least squares solution with or without an
intercept; (4) a ridge regression weighted least squares solution with or without an intercept; and
(5) an effective variance least squares solution with or without an intercept. An estimate of the
uncertainty associated with the source contributions is an integral part of several of these
solution methods.
Weighted linear least squares solutions are preferable to the tracer and linear
programming solutions because: (1) theoretically they yield the most likely solution to the CMB
equations, providing model assumptions are met; (2) they can make use of all available chemical
measurements, not just the so-called tracer species; (3) they are capable of analytically
estimating the uncertainty of the source contributions; and (4) there is, in practice, no such thing
as a "tracer." The effective variance solution developed and tested by Watson et al. (1984): (1)
provides realistic estimates of the uncertainties of the source contributions (owing to its
incorporation of both source profile and receptor data uncertainties); and (2) gives greater
influence to chemical species with lower values for uncertainty in both the source and receptor
measurements than to species with higher values for uncertainty. The effective variance solution
is derived by minimizing the weighted sums of the squares of the differences between the
measured and calculated values of Q and Fy (Britt and Luecke, 1973; Watson etal., 1984). The
solution algorithm is an iterative procedure which calculates a new set of Sj based on the Sj
estimated from the previous iteration. It is carried out by the following steps expressed in matrix
notation. A superscript k is used to designate the value of a variable at the k* iteration.
1. Set initial estimate of the source contributions equal to zero.
7=1, ... J (D-5)
2. Calculate the diagonal components of the effective variance matrix, Ve. All
off-diagonal components of this matrix are equal to zero.
(D-6)
.
Calculate the k+1 value of S.
4. Test the (k+l)th iteration of the Sj against the kth iteration. If any one differs by
more than 1%, then perform the next iteration. If all differ by less than 1%, then terminate the
algorithm.
D-2
-------
if | (Sk+l - Sf) I Sk+l | > 0.01, go to step 2
if (Sjk+l - Sjk) I Sk+l | < 0.01, go to step 5
(D-8)
5. Assign the (k+l)th iteration to Sj and os . All other calculations are performed
with these final values.
°s2 = [FT (Vi+1rl Ffl j = 1, ... J (D-9)
] •"
where: C = (Q ... Q)1, a column vector with Q as the ith component
S = (S^.Sj)1, a column vector with Sj as the jth component
F = an I x J matrix of FiJ3 the source composition matrix
oc = one standard deviation uncertainty of the Q measurement
OF = one standard deviation uncertainty of the F;j measurement
Ve = diagonal matrix of effective variances
The effective variance solution algorithm is very general, and it reduces to most of the
solutions cited above with the following modifications:
When the op are set equal to zero, the solution reduces to the ordinary weighted least
squares solution.
When the OF are set equal to the same constant value, the solution reduces to the
unweighted least squares solution.
When a column is added to the Fy matrix with all values equal to 1, an intercept term is
computed for the variable corresponding to this column.
When the number of source profiles equals the number of species (I = J), and if the
selected species are present only in a single, exclusive source profile, the solution reduces
to the tracer solution.
When the expression (FT (V^)"1 F) is rewritten as (FT (V^)"1 F - 4>I), with $ equal to
some non-zero number, known as the smoothing parameter, and I equal to the identity
matrix, the solution becomes the ridge regression solution (Williamson and Daboecia,
1983 and Henry etal, 1984).
Formulas for the performance measures are:
1
= =
J
Reduced chi square = f = — — E [(C. - E F.. Sf I VJ (D-10)
1 ~ J i= 1 j = 1
D-3
-------
J
Percent Mass = 100 (^ S.) / Ct , where Ct is the total measured mass (D-ll)
y'=i
R square = 1 - [(/ - J) X2] / [ E Q2 / FJ (D-12)
i= i "
Modified Pseudo-Inverse Matrix = (FT (Ve)~l F)~l FT (Fe)~1/2 (D-13)
The Singular Value Decomposition of the weighted F matrix is given by (Henry, 1992):
V™ F = UDVT (D-14)
where U and V are IXI and J X J orthogonal matrices, respectively, and where D is a diagonal
matrix with J nonzero and positive elements called the singular values of the decomposition.
The columns of V are called the eigenvectors of the composition and their components are
associated with the source types.
D-4
-------
Appendix E. Summary of CMB PMio Source Apportionment Studies
Sampling Site
Time Period
Primary
Motor Primary Secondary Secondary Misc. Misc. Misc. Misc. Measured
Primary Primary Vehicle Vegetative Ammonium Ammonium Source Source Source Source PM-10
Geological Construction Exhaust Burning Sulfate Nitrate 1 2 3 4 Concentration
Central Phoenix, AZ (Chow et al, 1991)
Corona de Tucson, AZ (Chow et al, 1992c)
Craycroft, AZ (Chow et al, 1992c)
Downtown Tucson, AZ (Chowefa/., 1992c)
Hayden 1, AZ (Garfield) (Ryan et al, 1988)
Hayden 2, AZ (Jail) (Ryan et al, 1988)
Orange Grove, AZ (Chow et al, 1992a)
Phoenix, AZ (Estrella Park) (Chow et al, 1991)
Phoenix, AZ (Gunnery Rg.) (Chow et al, 1991)
Phoenix, AX (Pinnacle Pk.) (Chow et al, 1991)
Rillito, AZ (Thanukos et al, 1992)
Scottsdale, AZ (Chowef al, 1991)
West Phoenix, AZ (Chow et al, 1991)
Anacapa Island, CA (Chow et al, 1996b)
Anaheim, CA (Gray et al, 1988)
Anaheim, CA (Summer) (Watson et al, 1994a)
Anaheim, CA (Fall) (Watson et al, 1994c)
Azusa, CA (Summer) (Watson et al, 1994c)
Bakersfield, CA (Magliano, 1988)
Bakerfield, CA (Chow et al, 1992a)
Burbank, CA (Gray et al, 1988)
Burbank, CA (Summer) (Watson et al, 1994c)
Burbank, CA (Fall) (Watson et al, 1994c)
Chula Vista 1, CA (Bayside) (Cooper et al, 1988)
Chula Vista 2, CA (Del Ray) (Cooper et al, 1988)
Chula Vista 3, CA (Cooper et al, 1988)
Claremont, CA (Summer) (Watson et al, 1994c)
Crows Landing, CA (Chow et al, 1992a)
Downtown Los Angeles, CA (Gray et al, 1988)
Downtown Los Angeles, CA (Summer) (Watson et al, 1994c)
Downtown Los Angeles, CA (Fall) (Watson et al, 1994c)
Fellows, CA (Chow et al, 1992a)
Fresno, CA (Magliano, 1988)
Fresno, CA (Chow et al, 1992a)
Hawthorne, CA (Summer) (Watson et al, 1994c)
Hawthorne, CA (Fall) (Watson et al, 1994c)
Indio, CA (Kim et al, 1992)
Kern Wildlife Refuge, CA (Chow et al, 1992a)
Lennox, CA (Gray et al, 1988)
Long Beach, CA (Gray et al, 1988)
Winter 1989-90
Winter 1989-90
Winter 1989-90
Winter 1989-90
1986
1986
Winter 1989-90
Winter 1989-90
Winter 1989-90
Winter 1989-90
1988
Winter 1989-90
Winter 1989-90
1986
Summer 1987
Fall 1987
Summer 1987
1986
1988-89
1986
Summer 1987
Fall 1987
1986
1986
1986
Summer 1987
1988-89
1986
Summer 1987
Fall 1987
1988-89
1986
1988-89
Summer 1987
Fall 1987
1988-89
1986
1986
33.0
17.0
13.0
26.0
5.0
21.0
20.0
37.0
20.0
7.0
42.7
25.0
30.0
2.2
21.2
11.4
13.2
34.9
27.4
42.9
21.3
14.0
11.0
6.7
8.2
9.7
19.4
32.2
23.8
12.7
9.4
29.0
17.1
31.8
7.5
8.9
33.0
15.1
16.0
20.7
0.0
0.0
0.0
5.1
2.0b
4.0b
0.0
0.0
0.0
0.0
13.8b
0.0
0.0
0.0
0.0
0.0
0.0
0.0
3.0
1.6
0.0
0.0
0.0
0.0
0.3
0.3
0.0
0.0
0.0
0.0
0.0
1.4
0.7
0.0
0.0
0.0
3.0
2.0
0.1
0.0
25.0
1.6
8.3
14.0
0.0
0.0
15.0
10.0
5.5
2.9
1.2f
19.0
25.0
4.9
4.11
8.5
37.2
15.9
5.5
7.7
6. 11
17.0
39.1
0.8
1.5
1.4
14.4
2.2
6.41
16.2
41.1
2.1
4.0
6.8
5.6
35.1
4.4
2.2
4.61
5. 11
2.3
0.0
0.0
0.0
0.0
0.0
0.0
0.9
0.0
1.0
0.0
7.4
10.0
0.0
0.0
0.0
0.0
0.0
9.6'
6.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
3.4
0.0
0.0
0.0
3.4
9.2'
5.1
0.0
0.0
7.1
4.0
0.0
0.0
0.2
1.9
0.7
1.0
4.0
4.0
0.4
1.6
1.0
0.9
0.0
0.6
0.4
3.4
7.0
9.0
3.7
11.4
5.6
5.5
7.2
12.4
3.1
7.5
8.9
8.2
9.5
2.8
7.6
13.0
3.9
5.1
1.8
3.6
15.0
5.1
3.6
3.3
7.6
8.0
2.8
0.0
0.6
0.2
0.0
0.0
0.4
0.0
0.0
0.0
0.0
3.6
3.1
1.0
9.8
2.9
38.5
6.1
0.0
12.7
10.2
6.5
25.1
0.0
0.0
0.0
6.3
6.5
11.2
4.4
27.5
7.5
0.0
10.4
0.6
20.4
4.1
1.5
7.9
9.2
0.0
0.0
1.2"
1.3"
74.0C
28.0C
0.0
0.0
0.0
0.0
11.68
0.0
0.0
9.6h
0.4j
0.0>
0.0>
0.0"
0.5"
1.0m
O.lj
O.Qi
0.0>
0.4j
0.6>
0.6>
0.0>
0.5m
0.0
O.Qi
0.0>
7.0m
O.lj
0.3m
0.0>
0.0>
0.2
0.5m
0.2j
O.lj
0.0
0.0
0.0
0.0
5.0d
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.4"
6.5"
3.1"
5.7"
0.0
1.5"
0.911
5.7"
1.9"
2.7"
1.8h
1.7"
4.7h
1.5"
1.3"
6.5"
1.8"
1.4"
0.0
1.0"
7.0"
3.7h
1.0h
1.5"
3.1"
2.0h
0.0
0.0
0.0
0.0
1.0e
1.0e
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
8.2k
0.0
0.0
0.0
0.0
0.6k
9.8k
0.0
0.0
2.0k
0.0
0.0
0.0
1.2k
7.911
0.0
0.0
1.4k
0.0
O.lk
0.0
0.0
0.0
0.7k
7.6k
6.4k
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
64.0
19.1
23.4
48.0
105.0
59.0
34.2
55.0
27.0
12.0
79.5
55.0
69.0
26.0
52.1
51.3
104.0
92.1
67.6
79.6
56.6
72.3
94.8
28.8
31.1
29.6
70.0
52.5
60.2
67.6
98.6
54.6
48.1
71.5
45.9
85.1
58.0
47.8
46.9
51.9
E-l
-------
Appendix E. Summary of CMB PMio Source Apportionment Studies
Sampling Site
Long Beach, CA (Summer) (Watson et al, 1994c)
Long Beach, CA (Fall) (Watson et al, 1994c)
Magnolia, CA (Chow et al, 1992b)
Palm Springs, CA (Kim et al, 1992)
Riverside, CA (Chow et al, 1992b)
Rubidoux, CA (Gray et al, 1988)
Rubidoux, CA (Summer) (Watson et al, 1994b)
Rubidoux, CA (Fall) (Watson et al, 1994b)
Rubidoux, CA (Chow et al, 1992b)
San Jose, CA (4th St.) (Chow et al, 1995b)
San Jose, CA (San Carlos St.) (Chow et al, 1995b)
San Nicolas Island, CA (Summer) (Watson et al, 1994c)
Santa Barbara, CA (Chow et al, 1996b)
Santa Barbara, CA (GTC) (Chow et al, 1996b)
Santa Maria, CA (Chowetal., 1996b)
Santa Ynez, CA (Chow et al, 1996b)
Stockton, CA (Chow et al, 1992a)
Upland, CA (Gray et al, 1988)
Vandenberg AFB, CA (Watt Road) (Chow et al, 1996b)
Telluride 1, CO (Central) (Dresser and Baird, 1988)
Telluride 2, CO (Society Turn) (Dresser amd Baird, 1988)
Pocatello, ID (Houck et al, 1992)
S. Chicago, IL (Hope et al, 1988)
S.E. Chicago, IL (Vermette et al, 1992)
Reno, NV (Non-sweeping) (Chow et al, 1990)
Reno, NV (Sweeping) (Chow et al, 1990)
Reno, NV (Chow et al, 1988)
Sparks, NV (Chow et al, 1988)
Verdi, NV (Chow et al, 1988)
Follansbee, OH (Skidmore et al, 1992)
Mingo, OH (Skidmore et al, 1992)
Sewage Plant, OH (Skidmore et al, 1992)
Steubenville, OH (Skidmore et al, 1992)
WTOV Tower, OH (Skidmore et al, 1992)
Wuhan, China (Zelenka et al, 1992)
Time Period
Summer 1987
Fall 1987
1988
1988
1986
Summer 1987
Fall 1987
1988
Summer 1987
1989
1986
Winter 1986
Winter 1986
1990
1986
1988
Winter 1987
Winter 1987
1986-87
1986-87
1986-87
1991
1991
1991
1991
1991
Primary Primary
Geological Construction
11.1
11.3
31.7
16.4
32.6
43.1
34.9
19.2
48.0
13.1
11.8
1.6
9.5
3.2
7.4
4.6
34.4
25.4
4.5
32.0
12.1
8.3
27.2
14.7V
9.7
11.8
14.9
15.1
7.8
10.0
12.0
22.0
8.3
7.4
55.0
0.0
0.0
0.0
1.4
0.0
4.0"
4.5
16.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.5
0.4j
0.0
0.0
0.0
7.5"
2.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
21.4
Primary
Motor Primary Secondary Secondary Misc.
Vehicle Vegetative Ammonium Ammonium Source
Exhaust Burning Sulfate Nitrate 1
6.3
42.8
11.2
2.3
7.0
5.61
17.3
30.3
10.2
9.2
8.9
0.9
14.7
5.1
7.6
6.8
5.2
4.11
3.2
0.0
0.0
0.1
2.8
0.9f
8.7
11.0
10.0
11.6
4.0
35.0
14.0
12.0
14.0
16.0
1.2
0.0
0.0
0.0
5.1
0.0
0.0
0.0
0.0
0.0
31.3
31.3
0.0
0.0
0.0
0.0
0.0
4.8
0.0
0.0
98.7
7.3
0.0
0.0
0.0
0.1
1.2
1.9
13.4
1.1
0.0
4.1
0.0
0.8
0.2
49.21
10.9
3.8
4.9
3.7
4.8
6.4
9.5
2.1
5.3
2.3
2.1
3.7
3.2
2.8
3.1
2.2
3.1
6.4
1.9
0.0
0.0
0.0
15.4s
7.7
0.6
0.8
1.3
2.7
0.9
16.0
15.0
13.0
14.0
15.0
28.1
0.8
23.2
19.7
4.2
21.4
21.3
27.4
31.6
21.7
13.3
12.8
0.5
1.0
0.5
1.4
0.6
7.0
14.5
1.0
0.0
0.0
0.0
0.0
0.0
0.2
0.2
0.6
0.9
0.1
0.0
0.0
0.0
0.0
0.0
17.0
0.1J
0.0>
o.y
O.lj
o.y
o.y
O.Qi
O.Qi
0.4j
0.911
0.7h
0.0"
6.4h
6.3"
5.7"
4.0h
0.7°
0.6>
9.3h
61.3"
7.3"
0.0
15.1'
0.8'
0.0
0.0
0.0
0.0
0.0
9.3'
3.4'
6.6'
3.8'
3.4'
49. 5y
Misc. Misc.
Source Source
2 3
2.2"
2.7"
1.2h
0.5h
1.3"
1.0"
5.1"
1.1"
1.5"
0.0
0.0
4.3h
0.0
0.0
0.0
0.0
1.8"
0.6"
0.0
0.0
0.0
0.0
2.2°
0.3h
0.0
0.0
0.0
0.0
0.0
0.0
11.0"
8.7"
5.0"
7.9"
13.6Z
0.0
0.0
1.2°
0.0
1.1°
5.9"
0.0
0.0
5.7°
0.0
0.0
0.0
0.0
0.0
0.0
0.0
o.ok
7.8k
0.0
0.0
0.0
84. lr
0.0
1.1™
0.0
0.0
0.0
0.2k
0.0
0.0
0.0
0.0
0.0
0.0
1.2-
Misc. Measured
Source PM-10
4 Concentration
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
7.78
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
46.1
96.1
66.0
35.1
64.0
87.4
114.8
112.0
87.0
68.4
64.9
17.4
34.0
20.5
27.0
19.0
62.4
58.0
20.6
208.0
27.0
100.0
80.1
41.0
20.4
24.9
30.0
41.0
15.0
66.0
60.0
62.0
46.0
49.0
224.9
E-2
-------
Appendix E. Summary of CMB PMio Source Apportionment Studies
a Smelter background aerosol.
b Cement plant sources, including kiln stacks, gypsum pile, and kiln area.
c Copper ore.
d Copper tailings.
e Copper smelter building.
f Heavy-duty diesel exhaust emission.
g Background aerosol.
h Marine aerosol, road salt, and sea salt plus sodium nitrate.
1 Motor vehicle exhaust from diesel and leaded gasoline.
' Residual oil combustion.
k Secondary organic carbon.
1 Biomass burning.
m Primary crude oil.
11 NaCl + NaN03.
0 Lime.
p Road sanding material.
q Asphalt industry.
r Phosphorus/phosphate industry.
s Regional sulfate.
' Steel mills.
u Refuse incinerator.
v Local road dust, coal yard road dust, steel haul road dust.
w Incineration.
* Unexplained mass.
y Residential coal burning.
z Aluminum processing.
m Primary lead smelter.
E-3
-------
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Lower Fraser Valley, British
Columbia, Canada (7/89 to
8/91) (Jiang e/a/., 1997).
24-hour canister samples every
sixth day at 8 sites in Lower
Fraser Valley for non-methane
organic compounds (NMOC)a
with 558 species.
Emissions profiles are grouped based
on 74 SARVAPb or AIRSC and
converted to molar emissions of
emitted NMOC species for
hydrocarbons (e.g., aromatics, alkanes
[parafins], alkenes [olefins], alkynes,
oxygenated compounds [esters,
carboxylic acids, ketones, aldehydes,
alcohols, ethers], others [Si-, S-, N-,
or halogen-containing], and
unidentified mixtures [Cl to C12
mixtures and >C12 mixtures]).
Descriptive data analyses were made
to compile emissions and ambient
measurements.
Overall high degree of similarity was found between emissions inventory and
ambient NMOC measurements.
Discrepancies between emissions inventory and ambient NMOC were found for
biogenic compounds (isoprene, -pinene, -pinene) and some species related to
light-duty vehicle exhaust.
Emission profile is used to calculate rate constants and product yields of reactions
in photochemical modeling.
Light-duty gasoline vehicles account for -80% of NMOC in the inventory.
Speciated emission profiles for light-duty gasoline exhaust need to be updated
periodically in the emissions inventory.
Western Washington Study,
Seattle, WA (6/95 to 9/95, and
7/96 to 8/96) (Fujita et a/.,
1997b).
3-hour C2 - C10 canister and
carbonyl DNPHd samples
beginning at 0600 and 1200
PDT twice per week (Tuesdays
and Thursdays) at 7 sites during
mornings and at 6 sites during
afternoons during summer
1995. Hourly C2 - C10 canister
and carbonyl DNPH samples
beginning at 0900, 1200, and/or
1500 PDT at 8 sites during
ozone episodes in summer
1996.
TotalNMHC = sumof25
abundant hydrocarbons.
Effective variance weighted least
squares CMBe with composite motor
vehicle exhaust (including tunnel,
garage, roadside, and dynamometer
for light-duty gasoline and diesel
exhaust), gasoline evaporation, liquid
gasoline, gasoline vapor, CNGf,
GNG8, LPGh, architectural coatings,
industrial solvents and coatings, and
biogenic profiles.
14 to 27 stable species used in CMB
calculation with 56 to 69 other species
in profiles and ambient data for
validation.
Average source contributions to total measured NMHC in % NMHC for:
Vehicle exhaust
Liquid gasoline
Gasoline vapor
Gas (CNG)
Gas (GNG)
Gas (LPG)
Biogenics
Unexplained
Summer 1995
3 9% to 57%
(including 0%
to 38% diesel
exhaust and
19% to 54%
light-duty
vehicle exhaust)
2% to 23%
7% to 15%
0% to 7.7%
0.6% to 12%
0% to 3.1%
0.2% to 2.5%
0%to28%
Summer 1996
40% to 104%
(including 0.8%
to 66% diesel
exhaust and
16% to 65%
light-duty
vehicle exhaust)
0%to37%
7% to 36%
0%to5.7%
0% to 7.1%
0% to 2.4%
1.2% to 7.8%
-10% to-50%
Larger errors in source contribution estimates were reported in 1996 because of
uncertainties in quantifying unidentified sources and fewer samples collected.
F-l
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
San Joaquin Valley and San
Francisco Bay Area, CA (7/90
to8/90)(Fujitae/a/., 1995b).
2-hour C2 - C12 canister and
carbonyl DNPH samples at
0800, 1000,1200, and 1400
PDTat 34 sites in central
California representing urban,
nonurban, oilfield background,
and forested areas. NMOG =
mass equivalent sum of all GC
peaks from canister sample plus
carbonyls from DNPH.
Effective variance weighted least
squares CMB with motor vehicle
exhaust, gasoline evaporation, liquid
gasoline, CNG, GNG, LPG, oil
production, architectural coatings,
industrial solvents and coatings,
biogenics, and acetone profiles from
other studies.
26 stable species used in CMB
calculation with 57 other species in
profiles and ambient data for
validation.
Typical average contributions to NMOG:
3 5% to 70%
10% to 40%
I%to2%
30% to 50% (southern
sites near oilfields),
5% to 15% (other sites)
10% to 15% (Yosemite and Sequoia only, based on
measured isoprene only)
2% to 20% (morning),
20% to 60% (afternoon)
Most NMOG compounds were explained well by the profiles for the morning
samples. Reactive precursors and end-products diverged from CMB estimates in
afternoon samples, resulting in large contributions from "Others". Ambient motor
vehicle contributions were -twice proportions from inventory neighboring grid
squares.
Vehicle exhaust
Gasoline evaporation
Coatings and solvents
Oil production
Biogenic:
Others
Los Angeles, CA (southern
California and southeast desert
areas) (6/74 to 9/74) (Mayrsohn
and Crabtree, 1976; Mayrsohn
etal., 1977).
3-hour Cj - C10 samples were
acquired starting 0200, 0600,
1100, 1200, and 1400 PDT at 6
south coast air basin sites
(around Los Angeles, Long
Beach, and El Monte) and at 2
southeast desert air basin sites
(Banning and Palm Springs).
NMHC = isthesumof39
species plus Cn - C18
compounds.
Multi-regression analysis and ordinary
least square equations were used with
vehicle exhaust, liquid gasoline,
gasoline vapor, CNG, GNG, and LPG
profiles derived in 1973.
Average contributions to NMHC (Cj to C10):
CMB-
calculated Regression
Vehicle exhaust 53% 47%
Liquid gasoline 12% 31%
Gasoline vapor 10% —
CNG 5% 8%
GNG 19% 14%
LPG 1% —
F-2
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Los Angeles, CA (8/86) (Harley
etal, 1992).
Hourly, 4-, and 8-hour canister
samples collected at 0400,
0600, 0800, and 1100 PDT at 9
sites between 8/10/86 and
8/21/86.
Effective variance least squares CMB
with revised and respeciated organic
gas emissions profiles for gasoline
engine exhaust (non-catalyst,
catalyst), unburned gasoline (whole
liquid gasoline, headspace gasoline
vapor), commercial jet exhaust,
architectural coatings (solvent-borne,
water-borne, thinning solvents),
industrial coatings, and industrial
adhesives.
Average contributions to NMOC:
Gasoline engine exhaust 31% to 37%
Whole liquid gasoline 32% to 38%
Headspace gasoline vapor 5% to 13%
Waste and natural gas 10% to 15%
Dry cleaning 0% to 4%
Degreasing solvents 5% to 12%
Respeciation of organic gas emissions result in large changes in basin-wide
emission estimates for 1,3-butadiene, ethylene glycol, methanol, and cyclohexane.
Reactivity for surface coatings and thinning solvents are revised.
Key discrepancies between CMB and the emissions inventory were found for
unburned gasoline. Excess unburned gasoline is suspected to be a combination of
emissions from tailpipe, hot-soak evaporative, and fuel spillage.
Southern California Air Quality
Study (SCAQS), Los Angeles,
CA (7/87 to 9/97) (Fujita et al.,
1994)
Hourly C2 - C12 canister samples
at 0700, 1200, and 1600 PDT
(PST during fall) at 8 sites
during summer and at 6 sites
during fall. Additional samples
at 0300, 0900, and 1400 at two
sites during the summer and fall
in Southern California.
Total NMHC = all GC/FID
peaks except for oxygenated
compounds.
Effective variance weighted least
squares CMB with motor vehicle
exhaust, gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, GNG,
LPG, architectural coatings, industrial
solvents and coatings, and biogenic
profiles from CARB's modeling
emissions data system and auto/oil
program.
27 stable species used in CMB
calculation with 34 other species in
profiles and ambient data for
validation.
Average CMB-calculated source contributions to total NMHC in % NMHC in:
Vehicle exhaust
Liquid gasoline
Gasoline vapor
Other
Summer
49% to 54%
11% to 17%
10% to 11%
21% to 30%
Fall
54% to 68%
14% to 15%
7% to 11%
11% to 21%
CMB estimates for 0700 to 0800 PDT sampling periods are 2 to 3 times higher
than the emissions inventory, with even larger discrepancies during midday.
Non-motor-vehicle hydrocarbon emissions are overestimated in the inventory,
while on-road motor vehicle emissions are underestimated.
Photochemical modeling with adjusted (increased) on-road emissions improved
model performance.
F-3
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Los Angeles, CA (7/95 to
10/95) (Fujitae/a/., 1997b)
3-hour C2 - C12 canister samples
starting at 0700 and 1400 PDT
for six 7-day periods at 3 ARB
sites and for 3-day periods at 8
CRCk sites.
NMHC = sum of 25 species
plus MTBE1.
Effective variance weighted least
squares CMB with composite motor
vehicle exhaust (including tunnel,
garage, roadside, and dynamometer
for light-duty gasoline and diesel
exhaust), gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, GNG,
LPG, architectural coatings, industrial
solvents and coatings, and biogenic
profiles.
27 stable species used in CMB
calculation with 56 other species in
profiles and ambient data for
validation.
Average contributions to sum of NMHC and MTBE with 10 alternative gasoline
vehicle profiles are:
3 ARB sites 8 CRC sites
Vehicle exhaust 54% to 64% 56% to 68%
(including 11% (including 10%
to 15% diesel to 15% diesel
exhaust and exhaust and
38% to 50% 42% to 54%
light-duty light-duty
vehicle exhaust) vehicle exhaust)
0.6% to 11% I%tol4%
15% to 29%
3.1% to 3.7%
5.2% to 8.7%
2.6% to 3.7%
0.2% to 0.3%
Coatings (architectural) 3.2% to 5%
Coatings (industrial) 1.7% to 9.3%
Liquid gasoline
Gasoline vapor
Gas (CNG)
Gas (GNG)
Gas (LPG)
Biogenics
10% to 20%
2.2% to 2.7%
6.6% to 8.6%
1.9% to 3.0%
0.1% to 0.2%
0.3% to 1.1%
4.1% to 6.9%
Coatings (other) 1.7% to 10% 1.1% to 8.9%
Unexplained -11% to-3.5% -0.9% to 7.8%
Ratios of tailpipe to evaporative emissions are 2.4 in the morning and 1.7 in the
afternoon. Morning samples attribute 48% to vehicle exhaust and 20% to
evaporative emissions.
Emission rates of heavy-duty diesel hydrocarbons are twice those of light-duty
gasoline on a -per-mile basis. About 50% of diesel emissions are >C10.
Motor-vehicle-related emissions and sources of ethane and propane gas account
for >90% of ambient NMHC. Evidence of emissions from solvent use were
found, but not as significant as the 30% to 40% identified in the inventory.
Boulder, CO (2/91 to 1/91)
(Goldanetal., 1995).
Hourly C3 - C10 NMHC with
Auto-GCm, NO, NOX, NOy, CO,
and SO2 at one
traffic-dominated site between
Boulder and Denver, CO.
Calculated correlations and ratios
among measured components,
especially with respect to NOy. These
were compared with speciated
estimates from the 1985 National
Acid Precipitation Assessment
Program Emissions Inventory.
Did not explicitly calculate source contributions to NMHC.
F-4
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
1996 Phoenix Ozone Study,
Phoenix, AZ (summer, 1996)
(Fujita and Lu, 1997)
Four-hour canister samples
beginning at 0700 and 1130 at
two sites and at 0700 at one
site.
Effective variance weighted least
squares CMB with composite motor
vehicle exhaust (including tunnel,
garage, roadside, and dynamometer
for light-duty gasoline and diesel
exhaust), gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, GNG,
LPG, architectural coatings, industrial
solvents and coatings, and biogenic
profiles.
27 stable species used in CMB
calculation with 56 other species in
profiles and ambient data for
validation.
Average contributions to sum of NMHC at 3 sites are:
Vehicle exhaust 5.4% to 11.7% Gas (LPG)
51.5% to 59.0%
6.6% to 8.9%
2.8% to 7.3%
1.2% to 3.2%
5.6% to 8.0%
Gasoline exhaust
Liquid gasoline
Gasoline vapor
Gas (CNG)
Gas (GNG)
0.5% to 1.0%
Biogenics 0.2% to 2.6%
Coatings (architectural) 0.4% to 2.2%
Coatings (industrial) 2.2% to 3.1%
Unexplained 7% to 8%
Because of the reactivity of isoprene, biogenic contributions are lower limits.
Actual contributions may be 5 to 10 times higher than CMB estimates.
Coastal Oxidant Assessment for
Southeast Texas (COAST)
Study, Houston, TX (7/93 to
8/93) (Fujita et al, 1995a; Lu,
1996).
Hourly canister (C2 - C10 HC)
and DNPH cartridge (CrC7
carbonyl compounds) from 6
surface sites, 6 times per day,
during summer 1993.
Measurements aloft (aircraft)
starting -0600 and -1200 CDT
each day at 6 locations. Hourly
auto-GC measurements at 2
sites.
Total NMHC = sum of 25
abundant hydrocarbons.
Effective variance weighted least
squares CMB with motor vehicle
exhaust, gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, GNG,
LPG, architectural coatings, industrial
solvents and coatings, and biogenic
profiles from CARB's modeling
emissions data system and auto/oil
program.
27 stable species used in CMB
calculation with 34 other species in
profiles and ambient data for
validation.
Average CMB-calculated source contribution to total NMHC in % NMHC for:
2 auto- 6 surface 6 aloft
GC sites sites sites
Vehicle exhaust 19% to 27%
12% to 15%
13% to 14%
11% to 12%
9% to 18%
0.4% to 1.6%
22% to 27%
19% to 36%
7% to 15%
7% to 20%
3.5% to 6.4%
22% to 50%
0.5% to 1.8%
-2% to 16%
Profiles need to be developed in Houston area to separate gasoline vs. diesel
contribution. Vehicle exhaust and industrial sources (e.g., refinery) are the largest
NMHC contributors. Source contributions aloft were 20% of those found at the
surface. Emissions inventory overestimates biogenic emissions, but is comparable
for the sum of liquid gasoline, gasoline vapor, industrial, and gas (CNG).
Liquid gasoline
Gasoline vapor
CNG
Industrial
Biogenic
Unexplained
17% to 39%
l%to 12%
0%to 11%
8% to 16%
19% to 52%
0.1% to 7%
-I%to39%
F-5
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
1996 Paso del Norte Ozone
Study, El Paso, TX (8/96 to
9/96)(Fujitae/a/., 1998)
Hourly auto-GC measurements
at one site and two-hour
canister samples five times a
day at four sites. Additional
survey canister samples at five
sites.
Effective variance weighted least
squares CMB with diesel exhaust,
gasoline vehicle exhaust, propane
buses, gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, LPG,
industrial, solvent-based primers and
enamels and isoprene.
27 stable species used in CMB
calculation with 34 other species in
profiles and ambient data for
validation.
Average CMB-calculated source contribution to total NMHC in % NMHC for:
Diesel exhaust
Gasoline exhaust
Propane bus
Liquid gasoline
Gasoline vapor
LPG
CNG
Industrial
Surface coating
Biogenic
Unexplained
El Paso
GC sites
I%to6%
26% to 44%
0.3% to 2.8%
26% to 3 3%
0.3% to 3.7%
0.5% to 3.4%
1.7% to 5.8%
7% to 21%
1.8% to 5.8%
0% to 0.1%
0% to 1.3%
El Paso
sites
<2%
50% to 65%
<2%
negligible
0 to 3%
10% to 20%
l%to 1.5%
negligible
Juarez
sites
2% to 5%
50% to 65%
2% to 5%
2% to 8%
included with liquid
2% to 9%
10% to 30%
2% to 3%
negligible
Source contribution estimates for auto-GC samples are averaged by day-of-the-
week and by wind direction.
Chicago, IL (9/85 to 10/85)
(O'Shea and Scheff, 1988).
45-minute Teflon bag samples
at 1200 and 1300 LOT.
NMHC = sum of
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Chicago, IL( 11/86 to 2/87)
(Aronian etal, 1989).
4-hour C2 - C7 and Tenax trap
(aromatic and chlorinated)
samples beginning 0800 and
1200 LOT at 3 downtown
Chicago sites. 20-hour samples
starting 1200 LOT at one
central city site.
NMOC = sum of 23 organic
compounds (including C2 - C7,
alkanes, and aromatic and
chlorinated compounds)
analyzed by GC/FID and
GC/MS.
Ordinary least squares CMB with 8
profiles (vehicle exhaust, gasoline
vapor, architectural coatings, graphic
arts, vapor degreasing, dry cleaning,
wastewater treatment, and petroleum
refining).
23 stable species used in CMB
calculation.
Average contributions to NMOC:
Emission
NMOC Inventory
Vehicle exhaust 35% 34%
Gasoline vapor 8.4% 7.6%
Solvent (architectural coating) 1.5% 5.5%
Solvent (graphic arts) 1.3% 9.8%
Vapor degreasing 2.4% 3.1%
Dry cleaning 0.5% 0.1%
Industry (refining) 18% 1.3%
Others 33% 39%
Largest discrepancies were found for petroleum refining contributions between
the CMB-calculated and emissions inventory estimates.
Chicago, IL (7/87 to 9/87)
(Scheff andWadden, 1993).
4-hour C2 - C7 and Tenax trap
(aromatic and chlorinated)
samples beginning 0800 and
1200 LOT at 3 downtown
Chicago sites. 20-hour samples
starting 1200 LOT at one
central city site.
NMOC = sum of 23 organic
compounds (including C2 - C7,
alkanes, and aromatic and
chlorinated compounds)
analyzed by GC/FID and
GC/MS.
Ordinary least squares CMB with 8
profiles (vehicle exhaust, gasoline
vapor, architectural coatings, graphic
arts, vapor degreasing, dry cleaning,
wastewater treatment, and petroleum
refining).
23 stable species used in CMB
calculation.
Average contributions to NMOC:
Emission
NMOC Inventory
Vehicle exhaust 21% 39%
Gasoline vapor 7.1% 7.6%
Solvent (architectural coating) 3.1% 5.5%
Solvent (graphic arts) 1.0% 9.8%
Vapor degreasing 3.4% 3.1%
Dry cleaning 0.3% 0.1%
Industry (refining) 7.4% 1.3%
Largest discrepancies were found for petroleum refining contributions between
the CMB-calculated and emissions inventory estimates.
F-7
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Detroit, MI (7/88 to 8/88)
(Chang et a/., 1996).
1-hour canister samples at 0600,
0800, 1200, 1400, and 2200
LOT at 18 sites.
NMHC = sum of 24 abundant
hydrocarbons.
Ordinary weighted least squares CMB
with vehicle exhaust, gasoline vapor,
petrochemical, architectural coatings,
graphic arts, polyethylene, refinery,
and coke oven profiles from other
studies.
Sensitivity tests with 18 stable and 24
reactive species yielded similar source
contribution estimates.
Average contributions to NMHC:
Vehicle exhaust
Gasoline vapor
Solvent (architectural coating)
Solvent (graphic arts)
Industry (refinery)
Industry (coke oven)
Other
CMB-
calculated
28%
9%
2.5%
4.7%
17%
3.7%
35%
CMB model calculations can be used to verify the emissions inventory estimates
prior to ozone modeling to derive more reliable source/receptor relationships.
Southeast Michigan Ozone
Study (SEMOS), Detroit, MI
(7/93to8/93)(Scheffe/a/.,
1996).
2-hour canister data at 0600,
1000, and 1400 LOT at 4 sites
for 8 days during summer 1993.
NMOC = sum of
chromatographic peaks.
Ordinary weighted least squares CMB
with vehicle exhaust, gasoline vapor,
petrochemical, architectural coatings,
graphic arts, polyethylene, refinery,
and coke oven profiles from other
studies.
Sensitivity tests with 18 stable and 24
reactive species yielded similar source
contribution estimates.
Average CMB-calculated contributions to NMOC:
Vehicle exhaust:
Liquid gasoline
Gasoline vapor
Solvent (architectural coating)
Solvent (graphic arts)
Industry (refinery)
Industry (coke ovens)
37% to 40%
2% to 9%
I%to3%
2% to 5%
I%to4%
3% to 10%
I%to4%
Percent of total NMOC explained by CMB ranged from 54% during afternoon to
69% during early evening (i.e., 1800 LDT). Good agreement between CMB and
emissions inventory for sum of vehicle and gasoline, architectural coatings, and
coke ovens. CMB reports higher contributions from refinery and graphic arts
industry than does inventory. Study demonstrated the effectiveness of CMB for
development of emission control strategy.
F-8
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Beaumont, TX, Detroit, MI,
Chicago, IL, Washington, DC,
and Atlanta, GA (summers of
1984 to 1988) (Kenski et ai,
1995).
1-hour and 3-hour canister
samples at 0600 and 0900 LOT,
except at Detroit with 1-hour
samples at 0600, 0900, 1200,
1400, and 2200 LOT.
NMHC = sum of 24 abundant
hydrocarbons.
Ordinary weighted least squares CMB
with vehicle exhaust, gasoline vapor,
petrochemical, architectural coatings,
graphic arts, polyethylene, refinery,
and coke oven profiles from other
studies.
Sensitivity tests with 18 stable and 24
reactive species yielded similar source
contribution estimates.
Average CMB-calculated contributions to NMHC:
Vehicle exhaust:
Gasoline vapor
Solvent (architectural coating)
Solvent (graphic arts)
Industry (refinery)
Industry (coke ovens)
Polyethylene
28% to 55%
9% to 20%
2% to 6%
5% to 12%
9% to 17% (in cities with refineries)
4% (in Detroit and Washington only
7% (in Beaumont only)
Motor vehicle fractions agreed with inventories, but refinery and graphic arts
proportions were substantially different.
Boston, MA (8/95 to 9/95)
(Fujitae/a/., 1997b).
Hourly C2 - C12 canister samples
at 0700 and 1300 EST for 14
days at 3 main sites and 3
supplemental sites.
Total NMHC = 25 most
abundant NMHC species plus
MTBE.
Effective variance weighted least
squares CMB with composite motor
vehicle exhaust (including tunnel,
garage, roadside, and dynamometer
for light-duty gasoline and diesel
exhaust), gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, GNG,
LPG, architectural coatings, industrial
solvents and coatings, and biogenic
profiles.
27 stable species used in CMB
calculation with 56 other species in
profiles and ambient data for
validation.
Average source contributions to sum of NMHC and MTBE:
Vehicle exhaust:
Liquid gasoline
Gasoline vapor
Gas (CNG)
Gas (GNG)
Gas (LPG)
Gas (biogenic)
Solvent (architectural coating)
Solvent (industrial coating)
Unidentified
Unexplained
50% to 69% (including 17% to 26% diesel
exhaust and 27% to 48% light-duty vehicle
exhaust)
2% to 19%
3% to 10%
3.1% to 3.5%
2.0% to 4.4%
0% to 0.3%
1.4% to 1.5%
0.4% to 0.6%
1.5% to 3.0%
6% to 11%
2% to 5%
Ratios of tailpipe to evaporative emissions with and without East Boston sites are
4.0 and 7.9 in the morning, and 3.9 and 7.4 in the afternoon, respectively. These
ratios are 2 to 4 times those found in Southern California. Morning samples
attribute 50.4% and 50.1% to vehicle exhaust and 12.5% and 7.9% to evaporative
emissions with and without the East Boston sites, respectively.
Motor-vehicle-related emissions and sources of ethane and propane gas account
for >90% of ambient NMHC. Evidence of emissions from solvent use were
found, but not as significant as the 30% to 40% identified in the inventory.
F-9
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
1995 NARSTO-Northeast
Ozone Study (6/95 to 8/95)
(Fujita and Lu, 1998)
Four, 3-hr canister samples at
five sites and three, 3-hr
samples at three sites. Also 13,
hourly auto-GC and 7 canister-
based PAMS sites operated by
state and local air pollution
agencies in the northeastern
states.
Effective variance weighted least
squares CMB with composite motor
vehicle exhaust (including tunnel,
garage, roadside, and dynamometer
for light-duty gasoline and diesel
exhaust), gasoline evaporation, liquid
gasoline, gasoline vapor, CNG, LPG,
and biogenic profiles.
27 stable species used in CMB
calculation with 56 other species in
profiles and ambient data for
validation.
Average CMB-calculated source contribution to total NMHC in % NMHC for:
Diesel exhaust
Gasoline exhaust
Liquid gasoline
Gasoline vapor
CNG
LPG
Biogenic
Unidentified
Unexplained
10 surface
PAMS sites
3% to 44%
17% to 62%
2% to 19%
12% to 31%
7% to 12%
I%to30%
I%tol2%
6 surface
sites
7 aloft
sites
6% to 8% 6% to 9%
7% to 23% 9% to 25%
I%to7% 1.5% to 5.7%
2% to 6% 0.8% to 5.9%
3% to 34% 7% to 11%
0.6% to 3.2% 0.7% to 2.4%
2% to 10% 0% to 5.8%
46% to 72% 59% to 75%
-8% to 17% -4% to -8% -6% to -11%
Unidentified compounds are the largest fraction of total NMHC in NARSTO-
Northeast supplemental hydrocarbon measurement. These samples were collected
in area that reflect more regional air quality, rather than PAMS sites that reflect
more urban/suburban area. Because of the reactivity of isoprene, biogenic
contributions are lower limits. Actual contributions may be 5 to 10 times higher
than CMB estimates (Fujita, 1997).
Newark, NJ (7/80 to 8/90)
(ScheffandKlevs, 1987).
1-hour canister samples at 0600
and 0800 EOT at 2 sites: high
traffic and industrial.
NMHC = sum of 24 abundant
hydrocarbons.
Ordinary weighted least squares CMB
with vehicle exhaust, gasoline vapor,
petrochemical, paint, and refinery
profiles from other studies.
Sensitivity tests with 18 stable and 24
stable and reactive species yielded
similar source contribution estimates.
For average NMHC:
Vehicle Exhaust
Gasoline vapor
Petrochemical
Paint
Refinery
Unexplained
15% to 21%
18% to 33%
3% to 6%
I%to7%
24% to 27%
15% to 28%
Industrial site showed three times the petrochemical and refinery contributions
than the nearby urban site.
F-10
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Kanawha Valley, WV
(Charleston, WV) (4/87 to 3/88)
(Cohen et al., 1991a, 199 Ib).
12-hour VOC (Tenax-GC
downstream of glass-fiber filter)
and particles.
Samples starting 0600 and 1800
EST for 15 days per month for
4 months at 3 sites.
19 VOCs were measured.
A mobile van also collected
NO2, NO3, O3, light scattering,
and meteorological
measurements alone with VOC
and particles at one in-valley
site.
Univariate and multivariate
correlation, analysis of variance
(ANOVA), and factor analysis
methods were used to resolve source
contributions.
Factor analysis with % variance explained:
Vehicle-related exhaust
(including tailpipe emissions,
fuel evaporation, and road dust)
General VOCs
12% to 48% (e.g., high
correlations for decane,
/w-xylene, and benzene)
15% to 17% (e.g., high correlations for styrene,
methylchloroform, toluene)
5% to 6% (e.g., high correlations for chloroform,
carbon tetrachloride, methyl chloroform)
11% to 38% (e.g., high
correlations for
benzene, decane,
particle organic and
elemental carbon)
5% to 13% (e.g., high
high correlations for
H2SO4, S)
VOC is a good indicator for motor vehicle exhaust if the atmosphere is enriched
with alkylated-aromated compounds.
Local and regional sources contributed to pollution levels at Kanawha Valley,
WV, using a combination of gas and particle data.
Chlorinated VOCs
Forest fire and combustion
related emissions
Acid particles
1990 Atlanta Ozone Precursor
Study, Atlanta, GA (7/90 to
8/90) (Henry et al, 1994).
Hourly C2 - C10 canister samples
with auto-GC at 6 sites.
Total NMOC = 54 hydrocarbon
species with ethane excluded.
GRACE11 and SAFER0 statistical
methods used to derive vehicle-related
source compositions for vehicles in
motion (tailpipe plus running losses),
evaporative gasoline, and headspace
gasoline vapor. Statistically derived
compositions were compared with
measured source compositions. 37
species were used to derive source
profiles.
Unburned gasoline comprised -50% of tailpipe emissions. There are other
unaccounted sources for whole gasoline.
Most studies include the whole gasoline contribution with the headspace gasoline
vapor component.
1990 Atlanta data shows that the whole gasoline component is considerably larger
than the headspace gasoline vapor component. Relative source contributions were
62% roadway vehicle emissions, 15% whole gasoline, and 4% headspace vapor.
GRACE/SAFER provide cost-effective alternatives to estimate source profiles,
but require several hundred observations.
F-ll
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
1990 Atlanta Ozone Precursor
Study, Atlanta, GA (7/90 to
8/90) (Lin and Milford, 1994).
Hourly C2 - C10 canister samples
with auto-GC at 2 sites.
Total NMOC = 54 hydrocarbon
species with ethane excluded.
Source profiles were decay-adjusted
to account for different reaction rates
of VOCs for roadway vehicle and
gasoline vapor emissions. PCAP,
CMB, and decay-adjusted CMB were
applied to test synthetic and ambient
data. 13 stable species used in CMB
calculation with 36 other species in
profiles and ambient data for
validation.
Decay factors assumed for propylene were 0.78 to 0.92 (morning samples), 0.41
to 0.60 (afternoon samples), and 0.59 to 0.76 (average).
Adjusted CMB source contributions to NMHC in % NMHC were:
Roadway vehicle emissions
Whole gasoline
Headspace gasoline vapor
Total vehicle-related
sources in summer 1990
CMB-
calculated
61% to 65%
not calculated
4% to 8.2%
Lewis et
al. (19931
44%
15%
4%
66% to 73% 62%
CMB adjusted for decay factor gives small improvements to source contribution
estimates.
1990 Atlanta Ozone Precursor
Study, Atlanta, GA (7/90 to
8/90) (Conner et al., 1994).
Hourly C2 - C10 canister samples
with auto-GC at 6 sites.
Total NMOC = 54 hydrocarbon
species with ethane excluded.
Source profiles for roadway vehicle
emissions, whole gasoline, and
headspace gasoline vapor were
developed. Profiles for pure propane,
natural gas, and industrial coatings
(auto painting) were used. 18 stable
species used in CMB calculation with
35 other species in profiles and
ambient data for validation.
Average contribution to total NMOC in % of total NMOC are:
Roadway vehicle emissions
Whole gasoline
Headspace gasoline vapor
Propane and natural gas
Industrial coatings (auto painting)
69% to 79%
6% to 16%
4% to 15%
2% to 5%
0%to5%
Emissions inventory (based on MOBILE model for mobile sources) generally
underestimate actual motor vehicle source contributions based on CMB results by
10% to 30%, assuming 2% to 47% biogenic emissions.
Emissions inventory generally overestimate point and area sources by 10% to
20%.
F-12
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
1990 Atlanta Ozone Precursor
Study, Atlanta, GA (7/90 to
8/90) (Lewise/ al., 1996).
Hourly C2 - C10 canister samples
with auto-GC at 1 site.
Total NMOC = 54 hydrocarbon
species with ethane excluded,
using 5 weekday samples.
Source profiles for roadway vehicle
emissions, whole gasoline, and
headspace gasoline vapor were
developed. Profiles for pure propane,
natural gas, and industrial coatings
(auto painting) were used. 16 fitting
species used in CMB calculation with
31 remaining species in profiles and
ambient data for validation.
Average contribution to total NMOC in % of total NMOC are:
49%
10%
3.7%
4.9%
2.9%
2.2%
2.6%
24.4%
Uncertainties in source contribution estimates are 9% to 14% for roadway vehicle,
natural gas, propane-rich, and isoprene-rich emissions; 30% to 40% for
evaporative gasoline emissions; and 50% for industrial solvent emissions. Using
the VOC 14C abundance' attributed 9% to 17% VOC as biogenic emissions for the
mid-morning and late evening samples, respectively. Additional research is in
progress using linear programming-factor analysis (e.g., GRACE/SAFER) to
refine source profiles.
Roadway vehicle emissions
Whole gasoline
Headspace gasoline vapor
Propane
Natural gas
Isoprene
Industrial solvent
Unexplained
Tokyo, Japan (7/81) (Wadden et
al., 1986).
1- to 1.5-hour aloft Pyrex glass
vessel (aircraft, 300 to 1,500m)
samples from 2 days (7/16/81
and 7/18/81).
NMHC =18 compounds from
c, to a.
Ordinary least squares CMB with £
profiles (vehicle exhaust, gasoline
vapor, petrochemical plants, paint
solvents, degreasing, dry cleaning,
petroleum refinery, and rubber
production).
Average contributions to NMHC:
Vehicle exhaust 7%
Gasoline vapor 11%
Paint solvents 27%
Petroleum refinery 27%
Unexplained 29%
Short-term (~1 hour) samples are useful in providing diurnal and directional
information to delineate source/receptor relationships.
F-13
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
Study, Location, and Period
Measurements
Source Apportionment Method
Findings
Sydney, Australia (9/79 to 6/80)
Nelson and Quigley, 1982,
Nelson et al., 1983)
Hourly C2 - C10 samples
acquired on 400 ml glass gas
pipettes btween 0600 and 1200
on non-windy days at 3 sites.
NMHC = sum of 69 compounds
from C2 to C10
Comparison between ambient
measurements, source measurements
(i.e. vehicle exhaust, evaporative
gasoline, industrial solvents), and
emissions inventory were made.
Average contributions to NMHC:
Vehicle exhaust 36 ± 4%
Evaporative gasoline 32 ± 4%
Evaporative solvents 23 ± 4%
Gas leakage 3.5 ± 5%
Industrial Processes 5 ± 1%
Source contribution estimates vary significantly from the central business district
(high concentration of vehicle exhaust) to the sites near the industrial area (high
concentrations of evaporative and industrial process emissions).
Sensitivity of source contribution estimates in source compositions is -0.5 to
4.0%.
Study results are in reasonable agreement with hydrocarbon emissions inventory.
The study results indicate a somewhat greater contribution from evaporative
gasoline relative to vehicle exhaust and solvent evaporation in the inventory.
The Netherlands (1974 to 1994)
(cities of Delft, 1974 and 1977;
Hague, 1974 and 1977;
Kollumerwaard, winter and
summer 1994; andBrabart,
summer 1994) (Guicherit,
1997).
Hourly C2 - C5 samples
acquired with cold trap
(stainless steel loop packed with
glass beads) submerged into
liquid nitrogen.
Hourly C3 - C9 and
preconcentrated C6 - C16
samples acquired with carbon
trap (carbosieve, carbotrap,
carbotrap C) at 3 sites.
Comparisons between ambient
measurements, source measurements
(tunnel and dynamometer), and
emissions inventory were made for
data collected over the last 20 years.
No quantitative calculations were provided.
Using ambient measurements to validate emissions estimates, the study concluded
that major hydrocarbon emissions are parafins (46%), aromatics (30%), and
olefms (15%), with the remainder consisting of acetylene and aldehydes.
Toluene concentrations were consistently >20%, which closely resembled vehicle
emissions derived from tunnel and dynamometer measurements.
Diesel exhaust accounted for 14% to 21% of NMHC with high alkanes.
Elevated ambient temperature during summer resulted in higher evaporative
emissions and increased photochemical degradation.
F-14
-------
Appendix F. Summary of CMB VOC Source Apportionment Studies
a NMOC = non-methane organic compounds.
b SAROAD = U.S. EPA's Storage and Retrieval of Aerometric Data system.
c AIRS = U.S. EPA's Aerometric Information Retrieval System.
d DNPH = C18 cartridge impregnated with 2,4-dinitrophenylhydrazine.
e CMB = Chemical Mass Balance
f CNG = compressed natural gas.
g GNG = geogenic natural gas.
h LPG = liquefied petroleum gas.
1 California Air Resources Board, Sacramento, CA.
' NMHC = non-methane hydrocarbon.
k CRC = Coordinating Research Council, Atlanta, GA.
1 MTBE = methyl-t-butyl ether, a major component in reformulated gasoline.
m Auto-GC = automated gas chromatography with flame ionization detector (FID).
11 GRACE = Graphical Ratio Analysis for Composition Estimates, used to generate constraints (e.g., ratios) for SAFER model
input.
0 SAFER = Source Apportionment by Factors with Explicit Restriction.
p PCA = Principle Component Analysis.
q Kloudarfa/. (1996)
F- 15
-------
-------
APPENDIX G. PROCEDURES FOR TREATING SECONDARY PARTICLES
One of the key assumptions made by EPA-CMB8.2 is that chemical species do not react
with each other; Section 1.2 of the EPA-CMB8.2 Users Manual (EPA, 2004) and Section 4.1 of
this Protocol. This means that compositions for the source categories are obtainable which
represent the source profile as it is perceived at the receptor for the chemical species of interest
(Section 4.3.1 of this Protocol). Thus, EPA-CMB8.2 assumes no changes to the aerosol during
transport and ideally apportions the primary material that has not changed between source and
receptor. However, certain species, e.g., sulfur (S), that dominate polluted airsheds have both
primary and secondary sources.
In such airsheds, secondary aerosols may contribute significantly to the ambient loading
seen at receptors. These secondary materials are often in the form of reactive species such as
NH4+, SO4=, NO3", and organic carbon. If sources of such materials are not explicitly treated,
EPA-CMB8.2 will tend to underaccount for total particle mass (% MASS value in the Main
Report (Table 4.2-1 of this Protocol). If a compound which is secondarily formed or is normally
associated with regional scale pollution (such as sulfate) is included as a fitting species, a single
constituent source type must also be included in the fit. Use of the single constituent source
profile for secondary particles was initially suggested by Watson (1979) and is described briefly
in the Section 4.3.2 of this Protocol. With this technique, the secondary species are apportioned
to chemical compounds rather than directly to sources (Section 4.1 of this Protocol).
Table G-l illustrates an example of the way the technique was used in an actual application
for California's South Coast Air Basin (Watson et a/., 1994). Secondary source profiles
consisting of "pure" ammonium sulfate (AMSUL), ammonium bisulfate (AMBSUL), ammonium
nitrate (AMNIT), and organic carbon (OC) were used to apportion the remaining NH4+, SO4=,
NO3", and OC that would not be apportioned to the primary particle profiles. For some
secondary species thought to be significant (e.g., note the OC column), a source profile was
created which includes only that component, in which the percentage composition in the profile
is set to 100%. For other secondary species, only some chemical components may have been
measured. For instance, elemental S and/or sulfate ion (SO4=) may be measured rather than
ammonium sulfate, (NH4)2SO4. In such a case, the respective species abundances in the
(NH4)2SO4 would equal the mass % of each species in (NH4)2SO4. Thus, in the AMSUL profile
the abundance of S in pure (NH4)2SO4 is listed as 24.3% and the abundance of SO4= is listed as
72.7%. Examples are also given for other secondary species and their chemical components. In
all cases, the uncertainty was arbitrarily set to 10%. In the EPA-CMB8.2 calculations, the
portion of a measured secondary species not accounted for by other source types becomes
assigned to its corresponding single constituent source type, as represented by profiles such as
those described here.
The examples given above are described as profiles for secondary species. However, the
secondary profile may not represent secondary aerosol exclusively. For example, Watson et al.
(1994) indicated that the OC profile in Table G-l may account for contributions from fugitive
sources not included in the EPA-CMB8.2 calculation (e.g., cooking, plant parts, or tire wear) in
addition to secondary sources. In such a case, the technique may be considered as a means to get
an upper estimate of the amount of aerosol attributable to secondary formation.
G-l
-------
One of the advantages of using the single constituent source profile technique is that it can
account for that part of the ambient mass that is not accounted for by the primary sources
included in the EPA-CMB8.2 calculations. However, this technique cannot yield any
information on the specific source types contributing to the species in the single constituent
profiles. Furthermore, the ambient mass may still be underestimated in some cases. For
example, Conner et al. (1993) reported that fine particle mass may have been underaccounted for
in their CMB calculations because of the likelihood for some amount of water associated with
hygroscopic (or deliquescent) sulfates. The amount of mass due to this water depends of the
form of the sulfate and relative humidity factors.
G-2
-------
Table G-l. Secondary Aerosol Source Profiles (abundances are % of total mass)1
Species2
AMSUL
PM2.5 & Coarse3
Cone. ± Unc.4
Cl"
NO3-
scv
NH4+
Na+
TC
OC
EC
Na
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Ni
Cu
Zn
As
Se
Br
Sr
0.0
0.0
72.7
27.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
24.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.0
0.0
7.3
2.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
AMBSUL
PMi5 & Coarse3
Cone. ± Unc.4
0.0
0.0
83.5
15.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
27.9
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.0
0.0
8.3
1.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.8
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
AMNIT
PMis & Coarse3
Cone. ± Unc.4
0.0
77.5
0.0
22.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.0
7.8
0.0
2.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
OC
PMi5 & Coarse3
Cone. ± Unc.4
0.0
0.0
0.0
0.0
0.0
100
100
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.0
0.0
0.0
0.0
0.0
10
10
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
'Reproduced with permission from Watson et al. (1994)
2TC (Total Carbon) = OC + EC; sum does not include Na+, Cl', S, or TC.
3PMlO - PM2.5
4Conc. is the average abundance (% of total mass) for several samples of emissions from the same source type. Unc. is the
standard deviation of the abundances for these samples.
G-3
-------
Mo
Cd
Sn
Sb
Ba
Hg
Pb
Sum
100
99.1
8.5
100
100
10
G-4
-------
REFERENCES
EPA, 2004. EPA-CMB8.2 Users Manual. Report No. EPA-452/R-04-011. December
2004. U.S. Environmental Protection Agency, Research Triangle Park, NC.
Watson, J. G., 1979. Chemical Element Balance Receptor Model Methodology for
Assessing the Sources of Fine and Total Suspended Paniculate Matter in Portland,
Oregon. Ph.D. Dissertation, Oregon Graduate Center, Beaverton, OR; pp. 412-413.
Watson, J.G., J.C. Chow, Z. Lu, E.M. Fujita, D.H. Lowenthal, D.R. Lawson, and L.L.
Ashbaugh, 1994. Chemical Mass Balance Source Apportionment of PM10 during the
Southern California Air Quality Study. AerosolSci. Technol.,2l(l): 1-36.
Conner, T.L., J.L. Miller, R.D. Willis, R.B. Kellogg, and T.F. Dann, 1993. Source
Apportionment of Fine and Coarse Particles in Southern Ontario, Canada. Presented at
the 86th Annual Meeting and Exhibition, Air & Waste Management Association, Denver,
CO; 13- 18 June 1993.
G-5
------- |