United States
Environmental Protection
Agency
Air
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/4-81-016a
July 1981
                   RECEPTOR MODEL TECHNICAL SERIES
                               VOLUME  I

                     OVERVIEW OF RECEPTOR MODEL
                     APPLICATION TO PARTICIPATE
                     SOURCE APPORTIONMENT

-------
                                  EPA-450/4-81-016a
        Receptor Model Technical Series

                   Volume I
          Overview of Receptor Model
Application to Participate Source Apportionment
       Monitoring and Data Analysis Division
    Office of Air Quality Planning and Standards
          \< Q  '-:  • - •    •- ••-  "• -''-y.'ioa Agcnc
     U.S. ENVIRONMENTAL PROTECTION AGENCY
      Office of Air, Noise and Radiation
 Office of Air Quality Planning and Standards
 Research Triangle Park, North Carolina  27711
                   July 1981

-------
     This document is issued by the Environmental Protection Agency to report



technical methods of interest to a limited number of readers.  Copies are avail-



able free of charge to Federal employees, current contractors and grantees, and



non-profit organizations (.in limited quantities) from the Library Services Office



(MD-35), Research Triangle Park, North Carolina 27711 •, or, for a fee, from the



National Technical Information Service, 5285 Royal Road, Springfield, Virginia



22161.





     This document has been reviewed by the Office of Air Quality Planning and



Standards, U.S. Environmental Protection Agency and approved for publication.







                                        Publication No. EPA-450/4-81-016a
   13..C
                                        n

-------
TABLE OF CONTENTS

Abstract	v

Executi ve Summary	vi

Acknowledgements	xii

List of Tables	xiii

List of Figures	,	xiv


1.0  Introduction	 1

2.0  Overview of Source Apportionment Methods	 7

     2.1  Source Model s	10
     2.2  Receptor Model s	10

          2.2.1  Microscopic Techniques	10
          2.2.2  Chemical Methods	13
          2.2.3  Physical Methods	24
          2.2.4  Hybrid Models	25

     2.3  Emission Inventory Methods	27
     2.4  Selecting the Appropriate Techniques	29

3.0  Program Design and Management	33

     3.1  Experimental Design	33

          3.1.1  Sampling and Analytical  Design	34
          3.1.2  Development of Source Composition Data	41
          3.1.3  Data Interpretation	44

     3.2  Typical  Program Costs	44
     3.3  Program Management	50

4.0  Applications of Receptor Models	53

     4.1  Development of Control  Strategies	53
     4.2  Validation of Dispersion Models	55
          4.2.1  Study Design	56
          4.2.2  Comparison of Model Results	59
          4.2.3  Dispersion Model Improvements	60

5.0  References	,	61

Appendix 1	1-1
                                         m

-------
                                    ABSTRACT
                         Receptor Model Technical Series
                                    Volume I
                           Overview of Receptor Model
                Applications to Participate Source Apportionment

     Volume I of the Receptor Model Technical Series presents an overview of
current particulate source apportionment methods and their applications to
control strategy development programs.  It is the first in a series of documents
describing methods which can be used to identify source impacts, using data
collected at the receptor.  This is unlike source (dispersion) models that esti-
mate source impacts based on emission factors, plume behavior and meteorology.
Volume II of the series describes the Chemical Mass Balance Receptor Model in
detail.  Future volumes will describe other receptor model techniques.

     Information presented in this series is directed to regulatory profes-
sionals responsible for particulate control strategy development or related
programs requiring source apportionment analysis.  Major receptor methods are
discussed, applications to control strategy development are presented and areas
in which receptor models compliment source (dispersion) models are explored.

     Properly applied, and with supportive evidence developed through inde-
pendent approaches, receptor models can be used independently or in concert with
dispersion models, to provide important new information to regulatory agencies.

-------
EXECUTIVE SUMMARY
     Air pollution professionals responsible for the development of air parti-
culate control strategies and other particle impact analysis programs, have long
been faced with the basic need to understand the relative importance of specific
source impacts within their airsheds.   Control  officers responsible for developing
particulate attainment strategies must be able to provide convincing evidence
that (a) the relative importance of emission sources is understood and that (b)
the control programs proposed are cost effective and can be adopted by the
community with confidence.

     Until quite recently, traditional approaches to the problem of apportioning
source impacts have been limited to dispersion, or source, models which use
emission inventory data (gathered at the emission source) with meteorological
data to estimate impacts at the receptor.  Unlike source models, receptor models
deduce source impacts based on ambient particulate morphology, chemistry and
variability information collected at the receptor.  The increased interest in
receptor models has resulted from the inability of dispersion models to assess
short-term source impacts or identify sources which collectively account for all
of the measured mass.  These shortcomings are largely the result of the diffi-
culty in developing accurate 24-hour particulate inventories and meteorological
data bases.

     In spite of their limitations, dispersion models will remain as an important
tool in the regulatory effort since they are the only means through which parti-
culate impact assessments can be made for proposed, new sources or at locations
without measured data.  Projections of future air quality and control strategy
scenario analysis also require dispersion model approaches.  Many receptor models

-------
are also limited in their ability to identify specific sources (e.g., paved road
dust) from among a group of chemically similar sources, such as soil dust, rock
crusher emissions or building demolition dust.  They do, however, provide
information on the entire aerosol mass, not just those sources within the air-
shed.  Used properly, however, receptor models can provide important insights
into the nature of a community's particulate problem.  Used in association with
dispersion models, receptor-oriented techniques can be used to improve the
performance of dispersion models and add confidence to the ability of the dis-
persion model  to simulate impacts from specific sources.

     The purpose of this document is to provide a brief overview of current
receptor models, their application, advantages and disadvantages.  Subsequent
Volumes in this series will discuss the individual models in greater detail.
The information in this document is directed to local, State and Federal  air
pollution regulatory professionals responsible for assessing the impact of
particulate sources or in developing and evaluating State implementation plans.
Others interested in the origin of ambient aerosols and in improving confidence
in particulate dispersion model results will also find this information of
interest.

     Receptor  models discussed herein can be placed in one of the eight cate-
gories, each of which is described below:

Chemical Mass  Balance
     This method matches source particle size and chemical "fingerprints" to
those measured at the receptor to back-calculate the impact of specific sources
or source classes of similar chemical composition.  Given data on the ambient
concentrations of several chemical species and the percent by weight of these
                                        V1T

-------
species in the emissions from the sources, a set of equations is prepared and
solved to determine the source impacts.

Enrichment Factors
     Ambient aerosol composition data are used in association with a reference
element (usually a crustal  element such as Fe, Al or Si) to provide an estimate
of the degree to which a specific ambient aerosol element has been "enriched" by
an anthropogenic source.  If the "enriched" element is known to be a unique
tracer for a specific source, and the concentration of the tracer in the source
emissions is known, a crude estimate of the source's impact can be made.

Microscopic Techniques
     Particle identification by optical  microscopy was one of the first, and the
most widely used method of source apportionment of coarse mode particles.  Current
technology has been expanded to include computer-driven scanning electron micros-
copy coupled with x-ray fluorescence analysis to provide a particle-by-particle
analysis of ambient coarse particulate filters.  As a consequence, particle
identification methods traditionally founded on particle size shape, color,
bifringence, and surface properties has been expanded to include elemental  compo-
sition and rapid, computer assisted analysis permitting large numbers of particles
to be analyzed at minimum cost.

Multivariate Methods
     Statistical methods include factor analysis, regression methods, principal
component and cluster analysis techniques.  These methods deduce information on
source impacts on the basis of the variability of chemical species measured
within a large set of particulate samples.  Given the premise that chemical
                                        vm

-------
species emitted from a specific source will vary in time ias measured at the
receptor) in the same manner, multivariate methods detect the common variability
of the chemical species.  The analyst then identifies the contributing source by
comparing those species with similar variability to the chemical composition of
sources within the airshed.

Radioisotope Analysis
     Measurement of carbon-14/total carbon ratios have recently been used to
distinguish "modern" from fossil fuel carbon.  Using this method, carbon emitted
from contemporary sources (wood burning, fireplaces, leaf fragments) has been
distinguished from particles released by the combustion of fossil fuels, e.g.,
auto and diesel exhaust, coal and fuel oil carbonaceous aerosols.

Spatial Series Analysis
     Spatial relationships between aerosol chemistry measured at numerous
receptors can provide important clues to likely contributing sources when viewed
in relation to emission density maps and given a basic understanding of the
chemical composition of source emissions.

Time Series Analysis
     Qualitative indications of source impacts based on temporal variations in
aerosol mass and chemistry can be used,  in association with source emission
activity and transport data, to gain insight into likely source impacts over
time.

X-Ray Diffraction (XRD)
     Quantitative identification of crystalline substances  in the coarse mode
(>2.5 ym) by XRD has enabled analysts to determine impacts  from fugitive emission
                                        IX

-------
sources with reasonable accuracy (+25%)  for moderate to heavily loaded  filters.

     Each of these methods has its own associated advantages  and disadvantages.
While no single technique is best used in all  situations,  two or more receptor
models are often used in association with one  another (or  dispersion models) to
provide increased confidence in program results.   Hybrid models, which  combine
the best elements of two or more receptor or source-oriented  models frequently
offer the best approach to identifying source  impacts.

     Rather than viewing receptor models as a  group of mathematical or analytical
techniques, the analyst should recognize that  the success  of  a receptor model
study is highly dependent upon the adequacy of the air sampling, analytical,
source characterization and data interpretation programs.   Project management and
quality assurance tasks are also important elements.  The  overall source appor-
tionment program design, of which the receptor model is only  one part,  should
therefore be of prime importance to those considering such a  study.  Analytical
and air sampling techniques are inexorably linked to the receptor model to be
used.

     Size resolution of the ambient (and source emissions) aerosol into fine
(<2.5ym) and coarse (_2.5-10 ym) modes provides a simple, yet  highly effective
means of apportioning particles associated with combustion sources, secondary
aerosols and controlled emission process losses from fugitive dusts and grinding
and crushing operations.  Commercially available sampling  equipment typically
collects the aerosol on inert filter substrates that are compatible with most of
the analysis techniques used  in receptor model studies.

-------
     Typical program costs can range from a few thousand dollars for a one year
study of ISP source impacts at a single location to one hundred thousand dollars
for a four station program designed to include fine and coarse mode analysis.
More costly programs may be required in those situations which program results
will heavily impact the communities.

     Receptor models, principally optical microscopy, have been widely applied to
particle source identification studies within the past 10 years.  Studies assoc-
iated with control strategy development programs, demonstration control strategy
analysis, dispersion model validation, PSD and "bubble" analysis programs are
typical.  In each case, the user must select the most appropriate receptor (or
source) model(s) and program design to provide the level of confidence needed to
support the regulatory action required.  In the case of control strategy develop-
ment programs, receptor models can be used in association with dispersion models
to (a) identify inconsistencies among source impact estimates derived from each
method, (b) alert the analyst to emission inventory and dispersion modeling
assumption errors, and (c) lend confidence to the dispersions model's ability to
reasonably predict the impact of specific sources targeted for control.

     Receptor-oriented methods of particulate source apportionment have evolved
in recent years into a new air pollution science which is distinctly different
than dispersion modeling.   Several of these methods have demonstrated their
                                                                       v
ability to quantitatively identify source impact, often in situations which are
not amenable to dispersion modeling.  Total, fine, coarse and inhalable parti-
culate fractions have been studied using these methods.   Properly applied and
with sufficient supportive evidence, receptor models can provide an important new
tool to regulatory agencies.

-------
                                ACKNOWLEDGEMENTS

     The rapid growth in receptor model  technology has been accomplished through
the efforts of many researchers and regulatory professionals.   While the current
technology of receptor modeling is still  in an early stage of development, those
associated with its development see these methods as a tool of growing importance
and potential to regulatory agencies.

     The principal  author of this document is John E.  Core of the Monitoring and
Data Analysis Division, Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency.   The comments and assistance of those involved
in the development of receptor models, as well as the Control  Programs Develop-
ment Staff, and Neil Berg, Jr., of OAQPS were important to the preparation of
this document.
                                        xii

-------
                                 List of Tables



1.  Participate Impact Assessment Senarios:  Source and Receptor
     Model Applications	6

2.  Source Apportionment - Receptor Model Advantages and  Disadvantages	11

3.  Typical Results of Coarse Mode Analysis by SEM-XRF	14

4.  Chemical Mass Balance:  Advantages and Disadvantages	21

5.  Source Apportionment Technique Appl ications	31

6.  Analytical Techniques for Urban Aerosols	35

7.  Chemical Receptor Model Aerosol Analysis Requirements	37

8.  Source Apportionment Study Costs:  Component Estimates	46
                                         xm

-------
                                 List of Figures

1.   Source-Receptor Model Pathways to Source Apportionment	8
2.   Source Apporti onment Methods	9
3.   Source-Receptor Model Hybrids	26
4.   Project Management Organization	51
5.   Source-Receptor Model Pathways to Control Strategy Analysis	54
6.   Source-Receptor Model Impact Comparisons	60
                                        xiv

-------
1.0  INTRODUCTION
     The Clean Air Act, as amended in August 1977, requires States to develop
and implement State Implementation Plans (SIPs) designed to bring their airsheds
into compliance with National Ambient Air Quality Standards (NAAQS).   Faced with
limited airshed capacity, emission growth, public concern and regulatory require-
ments, many communities are confronted with the difficult task of determining
which particulate emission sources are contributing to these nonattainment
problems and what the relative source impacts are.  Although traditional tech-
niques using dispersion modeling for source impact apportionment will remain an
important tool in airshed management, recent advances in receptor oriented
techniques are now beginning to offer an additional useful  tool.  Receptor
models (techniques that estimate the contributions of emission sources based on
data obtained at the receptor) include a variety of methods, each of which are
briefly discussed in Chapter 2 of this document.  In-depth discussion of these
methods will be published as subsequent volumes in this series.   Volume II
(EPA-450/4-81-016b) describes the Chemical Mass Balance Method.

     In the early phases of most control programs, point and area sources that
were major contributors within the nation's particulate nonattainment areas were
obvious and regulatory programs focused on emission reductions from these sources.
Today, however, many communities that have seen benefits from the first and even
second rounds of control still face continuing particulate nonattainment problems
from aerosols of unknown origin despite the high level of control applied to
many point sources.  It is in the latter case that an improved understanding of
source contributions are especially needed if cost-effective emission reductions
are to be achieved.

-------
     The importance of an accurate assessment of source contributions during


development of strategies is hard to overemphasize considering the cost of


control.  An appropriate example is that of a small , western nonattaimnent area


located in a confined river valley surrounded by actively cultivated agricultural


land.  A large industrial point source located within the valley was believed to


be the largest source contribution based on an emission inventory rollback model


analysis.  Construction of a new bag house to meet the required emission level


was estimated to cost nearly $3 million.  A more comprehensive source apportion-


ment study using receptor methods, however, identified uninventoried soil dust as


the major contributing source and estimated the annual geometric mean impact of

                                      3
the point source at less than 1.0 yg/m .  Had the point source controls been


installed, the industry's impact, as estimated by the source apportionment study,


would have decreased by about 0.3 yg/m , reducing the measured TSP annual mean

                     •3                                                     O
from 100 to 99.7 yg/m , for an overall cost benefit of $10 million per yg/m


reduction.



     Results from source apportionment studies in Portland, Oregon, have shown

                                                                                3
cost benefit control alternatives to range from an overall cost savings per yg/m

                                   o
to as high as $4.2 million per ^yg/m .   While the cost benefits of alternative
strategies will vary widely between communities, it is clear that considering the

costs, health implications and the importance of implementing a successful regu-

latory program, accurate and comprehensive source apportionment studies are


critical to the completion of State Implementation Plan revisions.



     Faced with the need to identify and quantify contributions of different

source categories to airshed particulate problems, air pollution agencies have,

in the past, turned to traditional emission inventory and dispersion modeling

-------
approaches to provide valuable clues to the likely ppint of origin of the particle
mass.  Other information, however, exists which can be used to identify the
origin of particles:

          - particle size, shape, color density or optical properties
          - chemical composition (organic and inorganic)
          - radioisotope chemistry
          - particle size distribution
          - concentration variability with space, time and meteorology

     Because of the limited ability of some of the above characteristics to
provide quantitative estimates of source impacts, there has been increasing
interest in the development of receptor models.   A receptor model is one which
estimates source contributions given ambient observations made at the sampling,
or receptor, site.  The growth in receptor model technology has followed the
development of rapid, low cost multi-element analytical techniques which provide
extensive "fingerprint" information on the ambient aerosol.  With accurate
ambient aerosol "fingerprints" at hand, receptor models provide the keys neces-
sary to draw inferences as to which sources, or source categories, contribute to
                                  1 2
the observed ambient aerosol  mass.  '
     Determining the sources  of airborne particulate matter is a very difficult
problem because of the complexity of urban source mix.  The problem is often
confounded by the predominance of nonducted and  widely distributed area (fugitive)
sources and the lack of understanding of the sources of secondary aerosol, their
formation and transport.   The advent of receptor methodology and recent develop-
ments in the areas of trace element analysis now permit a much more detailed
analysis of ambient aerosol  samples.   By providing detailed information on the

-------
sources of the total, fine and inhalable particles,  receptor models will  likely
play a major role daring the development of strategies for controlling airborne
participate matter.
     As promising as receptor models are, they cannot be applied to problems in
which (a) the source does not exist (e.g., proposed  new stationary sources), (b)
impact increases associated with changes in stack dispersion parameters (plume
temperature, volume  or height},  (c) identification of sources with chemically
nondistinct emissions is needed, or (d) identification of sources contributing to
secondary particulates (.e.g., sulfates) that often dominate the fine particle
mass is necessary.    Receptor models can be used to  forecast future source impacts
or to analyze alternative control  scenarios if no changes in plume buoyancy,
spatial or temporal  characteristics of the emissions are assumed.  For these
reasons, receptor models cannot replace dispersion models.   They can, however, be
applied to situations not conducive to dispersion models analysis.  Identifica-
tion of source impacts during actual,  short term particulate episodes, discovery
of "new" uninventoried sources within  the airshed and source apportionment of the
background, as well  as urban aerosol are a few of the most important advantages
offered by receptor  models.
     Air pollution programs have traditionally used  source oriented (dispersion)
models to estimate the impact of emission sources in their airsheds.  While
numerous dispersion  models (Gaussian,  box models, conservation of mass grid
models, particle-in-cell, etc.)  have been developed, all require emission inven-
tory, meteorological and dispersion parameter data to provide predictions of the
ambient particulate  concentrations.  All of the source oriented models suffer

-------
from an inherent inability to provide estimates of the source contributions to
the aerosol mass transported into urban areas.  Dispersion models cannot, there-
fore, provide a complete apportionment of the total mass of particulate measured
within urban nonattainment areas.

     Perhaps the most important deficiency of the dispersion modeling approach to
source apportionment is in the accuracy and lack of completeness of the emission
inventories used in the models.  Complete fugitive dust inventories are extremely
difficult to develop, considering the importance of transient sources, such as
building demolition, road maintenance and construction.  Source operating schedules
are often not accurately known, emission factors are imprecise and those sources
that are not inventoried are simply excluded from the analysis.   Dispersion
models are further limited by our ability to develop accurate regional emission
inventories for input to models used to estimate source impacts during episodes
in which the 24-hour particulate NAAQS is violated.  As imperfect as dispersion
models are, however, they remain a key element in control  strategy development
because they are the only mechanism through which the air quality benefits of
alternative emission control programs can be evaluated, potential new source
impacts can be assessed and future particulate air quality predictions can be
prepared.   Table 1  lists several particulate impact assessment scenarios commonly
encountered by regulatory authorities.   Those situations most appropriate to
source (dispersion) and receptor modeling are listed.   The analysts selection of
the most appropriate model, or mix of source apportionment methods, will depend
upon the specific problem and resources at hand.

-------
     The objective of this  guideline  is  to  provide  those  responsible  for

control  strategy development with  a better  understanding  of  (a)  the  limitations

and advantages of receptor  models,  (b) how  receptor models can  be  applied  to

practical  problems of source impact identification, (c) their use  in  control

strategy development, and (d) the  use of these  techniques to compliment informa-

tion gained from other methods.  Since the  most complete  and accurate understanding

of the source contributions will likely  come  from the  application  of  several

source apportionment techniques, no single  receptor method should  be  used  to

the exclusion of other information.   Experience has shown that  as  many different

independent approaches as possible should be  used in source  apportionment

studies to ensure that complete, valid conclusions  can be drawn.
                                     Table 1

                 Particulate Matter Impact Assessment  Scenarios:
                     Source and Receptor Model  Applications
     Source (Dispersion) Models	

    Predictions of future air quality
    Analysis of alternative control
     strategies
    Identification of secondary
     aerosol sources
    Impact predictions associated
     with changes in stack height,
     temperature or volume
    Identification of impacts from
     one specific source from a group
     of sources, all with similar
     emission characteristics
    Receptor Models
- Fugitive emission impacts
- Analysis of actual, worst
   case 24-hour impacts
- Identification of new,
   uninventoried sources
- Complex terrain/meteorology
   unsuitable for dispersion
   modeling
- Impact analysis of sources
   with unique emission
   chemistry, morphology or
   variability
- Whenever source apportionment
   must be independent of emis-
   sion inventory and meteorology.

-------
2.0  OVERVIEW OF SOURCE APPORTIONMENT METHODS
     Receptor models are one of several  source apportionment methods used to
identify source contributions to the ambient levels of particulate mass.   Although
receptor models have been investigated as a method of source apportionment over
the past ten years, only recently have the techniques been recognized as  a
distinct discipline and, in some applications, as an alternative and supplement
to dispersion modeling.  Dispersion models rely on measurements of emission
rates, stack parameters and meteorology to estimate impacts.  Hence, dispersion
models are source-oriented models.   In contrast, receptor models use actual
ambient aerosol measurements to identify source contributions.   Figure 1  illus-
trates the relationship between source and receptor models.

     Since the urban atmosphere contains a complex and varied mixture of  both
organic compounds and inorganic compounds emitted by natural and anthropogenic
sources, receptor models must sort out source emissions using combinations of
properties that are uniquely associated with each source.  Aerosol chemical
composition, particle size, concentration variability and particle morphology
are the four most important parameters used in receptor modeling.   Unfortunately,
each source's unique combination of "fingerprints" are often superimposed on one
another such that no single source can be easily identified and quantified.  The
receptor model's task is to separate the "fingerprints," often using source
emission composition information as a point of comparison.

     Receptor models are typically classified as microscopic, physical or chemical
methods, with each class subdivided into one or more specific areas of technology.
Figure 2 shows the relationship among these disciplines.  As the technology
develops, the distinction between these three classes becomes less distinct, as

-------
                          M  O
         o.

         ee
r—       O.
o:
         oc
          I
         cc
         a
         o
                                                                                             «  JS  O  -

                                                                                             M  V  >  B
                                                         e
                                                        £
                                                         D
                                                         e

-------
















*
C\J

CJ
r
u_























in
o

UJ
E
UJ
•i
^
^_
c;
o
c.
UJ
cr


— '









t


































































—
Emission
Inventory














O tn


u O
U E
C£













O) in
u — _
L- CJ
3 -0
O O
f> Z
Micro-
Inventory





U 0
tn *->







1C "O


QJ OJ




U
C. in
U O
in -C
U 3u
s:




[Dispersion
Models


























—










U Ul
o —
*-» Ul
QJ ^
L- «£
t—




"-*O
X











•e -0
u o
C. OJ


n —

UJ
10


















*









•o£
4-> Li.
rl^
1 =s
J «£l^








«»>































OJ

OJ —
t/*> t/l
— • c
ft)
a.
to




VI
QJ in
11 to
QJ >i

QJ C
E «E




CJ L.
E 0
U U
L- Li.




Q. in
O —•
O >%

o c
•o
en








QJ r*""1



^ *fO
*-• c

ai






i —
in

U
.— c
tC fC
. M U t— -
E co
QJ












c
O v
£^
» C




i/t
U —


u_ c

4_}
c
CJ
E
1 E



CD

tJ CT
i. •*-
t— *J
3
0



















JT


h-

-1













           c
           o
u    c    
-------
in the case of computer assisted scanning electron microscopy *-ray fluorescence
(SEM-XRF) methods which classify individual particles by physical  shape and
elemental composition.  Table 2 lists several source apportionment methods with
their relative advantages and disadvantages.   Each of the major methods are
reviewed below in greater detail.
2.1  Source Models
     Dispersion, or source models, estimate source impacts given information on
emission rates, plume buoyancy, source activity (e.g., emission rates, operating
hours and production schedule), and local meteorology.  They represent the
traditional means through which source impacts have been identified in recent
years.  The models attempt to mathematically simulate plume dispersion in the
atmosphere and aerosol physics downwind of the source to estimate source impacts
at a specified receptor.  Numerous approaches to the complexities involved in
dispersion modeling have been suggested:  box models, Gaussian and grid (numerical)
models being the most common.  Special constraints dictated by complex terrain,
particle deposition, secondary particle formation and short-term predictive
capabilities have been developed as have adaptation to provide regional and
long-range transport predictive capability.  Since a thorough review of source
modeling is outside of the scope of this document, the reader is referred to
other publications for a more thorough discussion. '
2.2  Receptor Models

     2.2.1  Microscopic Methods
          Particle identification by microscopic methods include optical,
scanning electron microscope (SEM) and automated SEM-Xray fluorescence (XRF)
                                        10

-------
Table 2.  Source Apportionment - Receptor Model  Advantages and Disadvantages
Source Apportionment
       Tools	
         Techniques
         Advantages
      Disadvantages
Microscopy

   Optical



   SEM



   Automated SEM-XRF
Chemical

   Enrichment Factors
   Time Series Analysis
   Spatial Series
     Analysis
Use of color, surface texture and
optical properties for particle
identification

Can be used with particles <1 m
Classifies particles by size, shape
and elemental composition.  Analyti-
cal speed, ability to count large
numbers of particles
Provides evidence of a source's
impact by changes in aerosol
composition.  Simple

Provides clues to sources; simple,
inexpensive
Provides clues to sources, simple,
inexpensive
   Chemical Mass Balance  Provides quantitative estimates
                          based on real data.   Impact
                          uncertainties provided
   Multivariate
     Analysis
   Radioisotope
     Analysis

Physical

   X-Ray Diffraction


Other

   Dispersion Modeling



   Trajectory Analysis


   Emission Inventory
   Microinventory
No prior knowledge of sources
needed to resolve element patterns,
compositon required to identify
sources by common names

Direct, quantitative measure of
fossil carbon
Direct quantifaction of crystalline
particles
Estimates impact.from future and
hypothetical sources
Helps identify approximate
source location

Traditional method of source
contribution analysis.  Simple
to use
Qualitative estimate of nearby
fugitive dust and point source
impacts
Limited to particles >2 pm,
semi-quantitative, highly
dependent on operator skill

Costly to use on large numbers
of particles, not useful for
amorphous species

Still in early stage of develop-
ment.  Costly.  Not reliable for
organics, coarse fraction
only
Semi-quantitative method;
requires source composition
data.  Often not specific

Generally does not provide
specific source impact
information

Does not provide source impact
information

Source composition data
required; chemically non-
descriptive sources cannot
be identified

Large data sets required,
cannot provide short-
apportionment
Costly.  Limited to fossil-
"modern" carbon apportionment
Coarse particles only.  Not
useful for amorphous aerosols
Difficulty in preparing accurate
emission inventory and transport
input data

Cannot quantitatively estimate
specific source impacts

Fugitive sources impossible
to inventory, background
aerosol not known; source
impacts incorrectly assumed
to be proportional to
emissions

Does not provide specific
source short term periods
                                            n

-------
methods.  These general groupings include specialized techniques, such as the
use of polarized light optical techniques.  Optical microscopy was one of the
first widely used methods of particle identification and remains a valuable
tool to qualitatively identify morphologically unique particles larger than
about 2 ym.  Particle identification is based on particle size, shape, color,
and other characteristics, often with reference to a "library" of particles
from known emission sources.  In the case of optical microscopy, particle color
and optical properties alone can often be used to assign particles to specific
sources based on the operator's experience, extent of his particle reference
library and associations with other particles found in the sample.  Optical
microscopy methods are limited by their relatively poor precision among analysts
and the high cost associated with analyzing a sufficiently large number of
particles to adequately represent the entire population of particles collected
on a filter.

          Although optical techniques are limited to particles greater than 1-
2 pro, electron microscopy can be used to identify particles in smaller sizes
using particle-by-particle analysis methods automated by computer.  Individual
particle identification by SEM is being supplemented with elemental analysis
using energy dispersive x-ray fluorescence analysis.  Once particle morphology,
elemental composition and density data from numerous "reference" sources are
developed, the SEM-XRF methods automated by computer can allow the research to
scan the large number of particles required to conduct a representative analysis
of the sample.  While automated SEM-XRF methods appear promising, current tech-
nology is limited by the XRF sensitivity to particles greater than 1 pm in
diameter.  Analytical costs are also relatively high.  Recent advances in this
technique have, for the first time, successfully differentiated between coal fly
                                         12

-------
ash components and the lithophillic (crustal} participate fraction captyred on
coarse mode dichotomous filters.
          Table 3 lists typical source categories identified by the current SEM-
XRF methods.   Note that although the mass less than 2 vm is estimated to be
quite low, gravimetric analysis of the filters showed a significant fine particle
mass model.  This technology is quite new, however, and more research and method
development is needed before wide-spread application of the method becomes
possible.

     2.2.2  Chemical  Methods
          Chemical  methods of receptor modeling are the most recent approach to
particulate mass source apportionment.  Many different methods have evolved from
different origins and are often perceived as distinctly different models when,
in many cases, their only differences are in terminology and approach.

          Chemical  methods, in general, identify sources by comparing ambient
chemical patterns or fingerprints (inter-elemental patterns, spatial  or time
variant patterns} with source chemical patterns.   Source contributions  are
quantified by a least squares multiple regression analysis on either the parti-
culate mass on different filters or the mass of individual chemical species on a
single filter.  Although similarities in the different chemical  approaches are
greater than their differences, they have been historically divided into two
categories:  Ca) chemical  mass balance methods,  which attempt to define the most
probable linear combination of sources to explain the chemical  pattern  on a
single filter, and (b\ multivariate methods, which attempt to define  the most
probable linear combination of sources to explain either the time or  spatial
variability in ambient chemical patterns.7
                                        13

-------
                                       Table  3
Typical Automated SEM-XRF Analysis
Coarse Particle Fraction Source Apportioment
Average of Ten Samples - Durham, NC
(Weight % of Total Sample and Average Particle Diameter* urn)
Chemical Category
Low Counts (1 )
Marine Aerosol
Ca and S Particles
Wood Combustion (2)
Refuse Combustion
Auto Exhaust t ^am-ii •>•
Fuel Oil Combustion'
Coal Fly Ash
Industrial Combustion (3)
Dolomite (4)
Limestone (4)
Construction f1*)
Asbestos/Talc
mite (4)
Kaolinite (4)
Feldspar (4)
Mica (Biotite) (4)
Mica (Muscovite) (4)
Al Rich Particles
Si Rich Particles
Lithophilic (4)
Other
Note:
<1 . 0 ym
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.01
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Q..-Q1
1-2 urn
0.0
0.0
0.0
0.0
0.0
0.01
0.0
0.07
0.0
0.0
0.0
0.0
0.0
0.11
0.01
0.0
0.0
0.01
0.0
0.02
0.06
0.0
0.29
2-4 urn
0.13
0.08
0.10
0.02
0.03
0.60
0.01
4.36
0.11
0.03
0.08
0.13
0.12
0.72
0.60 •
0.06
0.04
0.28
0.05
0.66
2.52
0.27
11.00

4O m
— Q yiTl
0.27
0.27
0.21
0.07
0.10
1.56
0.01
19.53
0.39
0.03
0.31
0.42
0.31
2.20
1.20
0.06
0.09
1.33
0.07
1.83
7.81
0.50
38.57
8-16 urn
0.15
0.26
0.0
0.13
0.10
0.95
0.0
24.07
0.17
0.0
0.13
0.11
0.20
2.08
0.97
0.06
0.02
1.14
0.08
1.18
5.28
0.43
37.51

> 1 6 urn
0.0
0.07
0.16
0.0
0.0
0.73
0.0
9.64
0.0
0.0
0.0
0.0
0.12
0.64
0.19
' 0.08
0.10
0.25
0.0
0.06
0.56
0.05
. 12.65

Total
0.55
0.68
0.47
0.22
0.22
3.85
0.02
57,. 67
0.67-
0.06
0.53
0.66
0.75
5.75
2.97
0.26
0.25
3.01
0.20
3.76
16.23
1.25
100.02

1)  Amorphous Compounds.
2)  Charred Wood  Fiber
3)  Combustion Sources Other Than Process  Emissions
4)  Usually Associated With Fuoitivp r>u<;t  Sourrps (e.f)..
*  Aerodynamic diameter

   SOURCE:  D. L.  Johnson (Reference  6)
                                          14

-------
            RadioIsptppe Analysis
          Measurements of radioisotope carbon-14/carbon-12 ratios have recently
been used by Currie and Klouda (National Bureau of Standards) and Cooper (Oregon
Graduate Center) to distinguish between "modern" and fossil fuel derived
                      o
carbonaceous aerosols.   The ability of radiocarbon measurements to resolve
fossil from "modern" carbon is based on the fact that the isotopic ratio of
14  12
  C/  C in plant and animal tissue is in equilibrium with the same ratio in
atmospheric CC^j whereas fossil carbon contains essentially no radiocarbon (  C)
because of its old age compared to the half life of the   C isotope (5730 years).
Consequently, carbon-14 is a unique tracer of carbonaceous aerosols emitted from
contemporary sources (.such as woodburning) and can be used to apportion the total
carbon content of the aerosol into two distinct classifications.  Limitations
include the inability of the method to identify the specific carbon source, the
high cost of analysis and analytical requirements for at least 5 milligrams of
carbon (e.g., 50 yg/m3 of carbon, 24-hour average at 60 liters per minute) to
obtain 10-20% precision in the analysis.  Source activity information on trans-
port, coupled with size resolved samples, have been used to limit the number of
possible emission sources.

          As the carbon-14 analytical method is improved, the speed, cost and
precision of the method will likely improve to the point where this tool may
become available for analysis of large numbers of samples.   Automated sample
preparation equipment and low level radiocarbon counters could eventually process
10,000 samples per year at a cost of $200 to $400 per sample compared to current
costs approaching $800. per sample (in groups of less than 4).
                                        15

-------
            Enrichment Factor Models
          The method uses ambient aerosol elemental composition data in associa-
tion with a reference element to provide an estimate of the degree to which a
                                                             Q
given element has been "enriched" by an anthropogenic source.

          The generalized format for this application is:
                               CCi/CER)_ aerosol                  /^
               EF(n aerosol = (C./CER) reference"

Where EF/^% aerosol is the enrichment of the tracer element "i" in the aerosol,
relative to some source material  (usually crustal} and a reference element
(usually Fe, Al, Si or Ca)_; CI/CER is the ratio of the concentrations of tracer
element (C..) and the reference element (CER) in the aerosol and the reference
material.  The reference elements (C£R) is often associated with crustal material
because of (a) their abundance in ambient aerosol samples,  and (b) referencing
the element enrichment to a crustal  (soil) source tends to minimize the effects
of local meteorology on the enrichment factor EF,. s.  Factors close to 1.0
indicate that the reference material (or source of similar composition) is the
primary source of the aerosol whereas factors significantly greater than 1.0 can
normally be interpreted as resulting from the impact of other, anthropogenic
sources.

          The enrichment factor concept can be extended to provide a semi-
quantitative estimate of source impacts if a unique elemental tracer can be
assigned to a source and the weight percent concentration of the element in the
source emission mass is known.  In this case, the formula below is used:
                                        16

-------
          r    (C,/CPD Impact) - CC./CPP No Impact 1
          Ls = 	L_M	J	bD	    %  C.     (2)
                       (C1/CER Impact 1 CF..J)               '

Where C  is the source impact, in yg/m3; C^ is the ambient concentration of the
tracer element in yg/m3; FI . is the fraction of tracer element "i" in source "j
(weight %) and C"- is the mean ambient concentration of element "i" (C-) (pg/m3)
during the study period.  Used in this form, enrichment factor analysis is quite
similar to a simple chemical mass balance model.  For example, the investigator
may be interested in determining the relative enrichment in the fine particulate
potassium concentration CCi) relative to the crustal reference element iron (C^)
during periods of suspected source impact.  Lyons   used fine particle potassium
                                                            K/
as a tracer for smoke from vegetative burning.  The average   Fe ratio during
smoke impact periods was found to be 2.0, compared to an average ratio of 0.6
during periods of no smoke impact.  If the average fine particle potassium concen-
tration found during the study was 0.2 yg/m3 and the smoke is known to contain
6.0% by weight potassium, equation 1 can be used to estimate the average fine
particle smoke impact:  (Cs):

                    r      2.0-0.6      n „
                    us    (2.0) (.060)  x u'^
                    Cs  =  2.3 yg/m3

          As useful and flexible as enrichment techniques are, they are only
semi-quantitative at best.   The method's best application is in identifying the
presence of abnormal sources at rural or remote sites and in providing an elemental
"index" relatively free of meteorological influence.  By examining enrichment
factors over time, and in relation to source activity and transport, inferences
                                        17

-------
can often be made as to the likely source of the tracer element.   Since the
technique is not applicable in situations in which-multiple sources are emitting
the same element, it will therefore not find wide application in source appor-
tionment studies.
            Time Series Analysis
          These methods are based on the assumption that the chemical species
emitted by a source will have the same time dependence at the receptor, resulting
in a high degree of correlation between the species.  To minimize meteorological
effects, the species concentrations may be expressed as a percent of the aerosol
mass or normalized in relation to another specie.  Once highly correlated specie
groups are identified, the contributing source can be inferred from (a) a quali-
tative understanding of the airshed source emissions and composition, and (b) the
diurnal, weekly and seasonal trends in the data and source emissions.  Long term
trends in pollutant concentrations are one example of time series analysis.

          These methods are simple, yet provide strong evidence as to possible
source influences.  They do not, however, provide quantitative estimates of
specific source impacts.  Their principal use is in trend analysis to evaluate
control strategy effectiveness and track air quality trends.
            Spatial Series Analysis
          The spatial relationships between aerosol mass and chemical species can
provide important clues as to likely sources if a sufficiently dense network of
sampling sites has been included in the project design.
                                        18

-------
          Measured concentrations of the aerosol mass and/Or specific chemical
species from several sampling sites obtained for identical time periods can be
examined in relationship to airshed emission points.   Implications can then be
made between the sources and spatial patterns, if some prior knowledge of the
chemical nature of the emission composition is at hand.   In some cases, in which
a single element is unique to a source, this kind of simple analysis has quali-
tatively identified the presence of previously unknown sources or served to
illustrate the geographical extent of a source's impact.

          Other spatial models such as pollution rose and cluster analysis are
useful, but rely on constant wind directions and emission rates to identify the
contributing sources.   These constraints severely limit the method's ability to
                          g
identify specific sources.

            Chemical .Mass Balance
          The purpose of the chemical mass balance technique (also known as
chemical element or chemical species balance) is to permit the investigator to
use available ambient aerosol composition data (elemental, ion, carbon, etc.) in
association with similar ambient aerosol data, to back-calculate the impact of
specific source, and source classes at the receptor.   The method essentially
matches the source's particle size and chemical  "fingerprint" to those measured
at the receptor.

          The CMB method is based on the assumption that  the mass of the material
deposited on a filter at the sampling site is a  linear combination of the mass
contributed from each of a number of sources and that the mass and chemical
composition of source emissions are conserved from the time of emission to the
                                        19

-------
time the sample is taken.  With these assumptions, it can be shown that the
concentration of a given chemical  specie measured on the ambient filter is
equal to the sum of the same chemical specie contributed by each source.   More
specifically:
          c, - IF)jSj

          where  Ci  = fraction of the measured aerosol mass, attributed to
                       specie i, at the sampling site, dimensionless
                 F.. = fraction of specie "i" in the emissions from
                       source j (e.g., 2% of soil dust mass is Na), dimensionless
                 S.  = fraction of the aerosol mass collected at the sampler
                       from source j (e.g., soil impact, expressed as a ratio
                       to mass), dimensionless
          Since C. and F-. can be measured by chemical analysis of ambient and
source samples, the S. source contribution can be determined mathematically by a
                     \J
least squares solution to a set of simultaneous equations.  '  '    Advantages
and disadvantages of the CMB method are summarized in Table 4.  Further detail on
                                                        15
the CMB method is presented in Volume II of this series.
            Multivariate Methods
          Statistical and multivariate methods include factor analysis, regres-
sion methods, target transformation factor analysis (JTFA), cluster and principal
component analysis techniques, and other methods.  These methods extract informa-
tion about the source's contribution on the basis of the variability of chemical
species concentrations measured within a large set of particulate samples.  Given
the premise that the chemical species concentrations measured at the receptor
                                        20

-------
Table 4.   Chemical Mass Balance Advantages and Disadvantages
  Advantages

0 Quantitative method

0 Can be applied to data from a single
  sample.   Can use small data sets

0 Can be applied to data collected
  from short term sample (<24 hours)

° Provides an estimate of source
  impact uncertainty


0 Internal consistency checks of
  analysis quality can be made

0 Identifies sources/source classes
  by common name
  Disadvantages

0 Requires source composition data

D Cannot resolve sources of secondary
  aerosols

0 Typically unable to account for
  all  measured Ca, Cu, Zn and carbon

0 Assumes that all species emitted by
  a source, within a given size range,
  have the same residence time

0 Cannot distinguish sources of
  similar chemical composition
                                       21

-------
attributed to a specific source will all vary in time or space in the same
manner, multivariate methods detect the common variability among the species.
Source identity is then implied by comparing species of similar variability to
the composition of specific sources.  Although the methods require no prior
knowledge of the number or composition of the source components to identify major
source groupings, the correct labeling of the groupings or factors in terms of
common source names does require a knowledge of the chemical composition of
possible sources.  Important advantage of these methods include:

               0 the ability to include non-chemical measurements, such as light
scattering, gaseous pollutant measurements and meteorology in the data set;

               0 no prior assumptions regarding the number of sources impacting
the receptor.

          These advantages allow factor analysis methods to associate primary
particles with secondary species formed between their release and measurement at
the receptor.  For example, secondary sulfate formed in the plume of a residual
oil fired boiler could be identified by factor analysis based on similar variation
of Ni and V (fuel oil trace elements) with sulfate.  If the plume becomes suffi-
ciently well dispersed downwind to become homogeneous, however, the association
may not be detected.

          Multivariate methods have been found effective in identifying sources
of secondary aerosols, fugitive emissions, motor vehicle exhaust, soil, residual
oil combustion and other source components.  A recent review of receptor model
technology  has summarized results from source apportionment studies using multi-
variate methods as follows:
                                        22

-------
               0 Hopke (University of Illinois)  and co-workers applied factor
analysis to 18 elements measured from a set of 90 total  suspended participate
samples from Boston.  Six source classes (factors) which collectively accounted
for 77 percent of the system variability were identified.   The three major
components identified were a soil component possibly containing coal fly ash,
marine aerosol, residual  oil combustion.

               0 Moyers and co-workers (University of Arizona) completed factor
analysis of 24 elements from a TSP hi-vol network of 11  sites in Tucson, Arizona.
The major impacts found were soil and secondary species  such as ammonium and
sulfate ion.  Auto exhaust and mining related emissions  were also found as was
a marine aerosol component.

               0 Lewis (EPA-ESRL) and Macias (Washington University) used
factor analysis on fine and coarse mode dichotomous samples taken in Charleston,
West Virginia in 1976.  Nineteen elements were used to resolve the soil, ammonium
sulfate and automotive components as well as an unknown  component composed of
C, N, Si, K, Ca, Fe, Zn,  Se and Sr.  No coal fly ash component was found despite
                                          1 o
its importance as a source in the airshed.

          As helpful as multivariate methods are in source resolution studies,
factor analysis methods have several weaknesses.  Very strong, consistent
source impacts which co-exist amidst weaker, more variable sources may not be
identified because the method depends on source variability, rather than absolute
values, to identify the source component.  The methods do  not provide informa-
tion on the composition of the source components but indicate only the fraction
of an element's variability associated with a given source.
                                        23

-------
          Unlike factor analysis which apportions system variance, Target Trans-
formation Factor Analysis (TTFA) permits an analyst to apportion aerosol mass in
a manner similar in nature to the chemical mass balance method, but without the
presumption as to the number or nature of the sources in the system.  The method
appears to hold promise as a useful  tool, but will require additional testing
before its full value to source apportionment analysis can be evaluated.  Recent
TTFA studies by D.  Alpert and Hopke  (University of Illinois) on St. Louis, Missouri
                                                                    19
size resolved aerosol data was completed using data for 27 elements.    The study
identified fireworks smoke (detected in July 4th samples), motor vehicle exhaust,
a soil coal fly ash component, limestone, paint pigments, refuse incineration and
sulfate ion as contributing sources, the sum of which estimated the fine and
coarse mass measured to within an average deviation of about 16 and 12 percent,
respectively.

     2.2.3  Physical Methods
            X-Ray Diffraction
          X-ray diffraction (.XRD) has been successfully used by several investi-
      20  21
gators  '    to quantitatively identify concentrations of crystalline substances
in ambient samples.  Road salt, lime, calcium carbonate, gypsum, cement dust and
numerous crustal minerals (calcite quartz, biotite, etc.) associated with fugitive
emissions have been successfully identified.
          Recent work by Davis (.South Dakota School of Mines) using dichotomous
samplers with Teflon filters indicates that XRD analysis of ambient samples with
high mass loadings (200 .yg/cm2 of filter area)* can be accomplished within an
*  Equivalent to 60 vg/m3 at 15 1pm (24 hours) using a dichotomous sampler.
                                        24

-------
accuracy range of 1 25% for most components.  Since crystalline components most
suitable to XRD analyses are found primarily in the coarse particle fraction, the
method's principal application has been in coarse particle source apportionment.
X-ray diffraction cannot be used to measure non-crystalline aerosols (organics,
acid aerosols, etc.).   Moderately heavy filter loadings are usually required for
analysis.  New instruments now under development promise to greatly shorten
current analysis time.

            Trajectory Analysis
          This technique attempts to identify the approximate origin of particles
by recreating the trajectory of the air parcel  sampled.  In the case of 24-hour
measurements, the analysis typically assumes only a single trajectory, arrival  of
the air parcel at the  site either at the end or midpoint of the sampling period
and transport of only  small particles.   Trajectory analysis has been used to
estimate the portion of the aerosol mass associated with long range transport,
                                                                     Q
but it cannot quantitatively estimate the impact of specific sources.

     2.2.4  Hybrid Models
          Since a major limitation of receptor models is their inability to
distinguish among sources of similar chemical/physical  composition, hybrid
models which combine the advantages of two or more receptor methods (or receptor
and source models} have been seen as a promising approach.   Figure 3 describes
the hybrid source/receptor model  process,  using receptor models to calculate
contributions of the source types and the source model  to quantify contributions
from specific sources.   Application of chemical  mass balance/dispersion model  is
discussed in Volume II  of this series,  while Chapter 4  presents an overview of
the receptor/dispersion model  approach.
                                        25

-------
        Figure 3.   Source-receptor Model Hybrids

                                                                                Receptor  Model
                                                                                Apportionment of
                                                                                Source  Classes
                                                        Coal  Combustion
                                                   Unpaved Road Oust
                                     Specific
                                     Source
                                     Impacts
                                    Paved Road Oust
After Watson (Reference 20)
26

-------
                                                                  22
          The microinventory-regression approach developed by Pace   has been
suggested as a useful  method of apportioning chemically similar sources  (e.g.,
soil  dust, paved road  dust,  rock crusher emissions) identified by chemical  mass
balance or factor analysis.   The current state of the art requires that  the
analyst conduct the microinventory-regression analysis as a separate task
following completion of the  receptor model  work.  No combined software package
currently exists which joins the two methods.

2.3  Emission Inventory Methods
     Airshed emission  inventories were the  basis of most first attempts  to
determine the relative contribution of specific sources to community particulate
problems.  Experience, however, has shown that this approach is confounded  by
numerous difficulties  that severely limit the direct use of inventories  for
source apportionment.   Perhaps the most important sources of error inherent in
the required assumptions are that (a) the inventory accurately includes  all
sources affecting the  receptor, (b) that the ratio of emission rates within the
inventory is the same  as the ratio of the source impacts, and (c) the method
does  not account for transported pollutants.

     Emission inventory methods are, however, useful to identify major sources  or
source classes that may contribute to measured particulate matter concentrations.
The inventory's value  in source apportionment work is greatly enhanced if each
source class is spatially resolved into grids of sufficient resolution (2 km x
2 km) to be useful  in  trajectory analysis.   A spatially resolved fugitive dust
inventory, when viewed in relation to a known dust source can often  identify
inventory deficiencies while providing guidance on needed improvements.   As  an
example, application of a uniform paved road dust emission factor throughout a
                                        27

-------
community of mixed land uses may, when examined as gridded emission densities,
conflict with "real world" impressions of the amount of street dust found in
industrial as compared to residential areas.  Correction of the deficiency may
required use of land-use specific emission factors which, when included in the
dispersion model data base, would greatly improve modeling results.

     Emission inventory scaling techniques can occasionally be used in associa-
tion with other source apportionment methods to identify the impact of an unknown
source.  If, for example, chemical mass balance methods can identify the impact
from a residual oil-fired boiler located adjacent to a distillate oil-fired
stack, the impact from the distillate fuel combustion emissions can reasonably be
assumed to be in the same proportion as the emission rates of the two boilers if
(a) the plume height and dispersion characteristics of both sources are similar,
(b) the operating schedule for both sources is identical, and (c) the particle
size distribution and deposition rates of the emission are the same.

     This approach is occasionally used to identify the impact of diesel truck
emissions.  Given the impact from leaded automotive exhaust (often based on a Pb
tracer) and the inventory ratio of diesel/leaded exhaust emission, the impact of
diesel emissions can be calculated if the diesel and automotive emissions are
assumed to be spatially and temporally similar.  This may, or may not, be defen-
sible based on the location of the sampling sites to nearby roadways and the
local traffic mix.

     Microinventory methods are procedures by which annual particulate matter
emissions are estimated in relatively small areas surrounding the monitoring
     22
site.    Area sources are typically  inventoried within a one-mile radius while
                                        28

-------
point sources within a five-mile radius are usually  included.   Inventories
developed for each of nine sectors surrounding  the site  are  then  related to  site
ISP mass data by regression analysis techniques.  While  this method does not
assist in aerosol characterization, it can qualitatively estimate relative
fugitive dust and point source impacts by performing  intersite  comparison of
emissions for sampling locations with similar particulate matter concentrations.
Microinventory data can be used to produce a multivariable, empirically-based
predictive equation or can be used with dispersion models to identify relative
source impacts.

2.4  Selecting the Appropriate Technique
     Selection of the most appropriate technique from those discussed above, will
largely depend on the analyst's objectives, resources and the available technology.
For the purposes of this Section, howevers we will assume that  the objective of
the source apportionment program is to identify the sources contributing to
inhalable particulate mass loading during actual air pollution episodes.  Regard-
less of the intent of the analysis, it is highly important that the analyst:

          1.   Identify the nature of the problem and review information already
on hand.   Familiarity with the airshed's sources (magnitude, location, operating
characteristics  of sources, fuels and fugitive emissions), the chemical composi-
tion of the aerosol,  it's size distribution,  dispersion  model results, and
meteorology are  of key importance.

          2.   Develop specific objectives, resources and the available time frame
within which  the analysis must be conducted.   These restraints are often critical
to the proper technique selection.
                                        29

-------
          3.  Insure that tasks included in the program design are consistent
with the objectives, resources, sampling and analytical techniques to be used.

     In the example case described above, dispersion modeling and microinventory
methods are not likely to be helpful given the extreme difficulty in accurately
assessing the emissions for a single day in history.  Multivariate methods, which
require large data sets, could only be used if the sampling program were designed
to capture a large number of time-resolved samples from a large number of sampling
sites.   Time series analysis also requires data captured over long time periods.

     The available options, listed in Table 5, provide the analyst with several
choices of coarse and fine mode techniques which provide qualitative, semiquan-
titative or quantitative results depending on the analysis requirements and
resources at hand.  Methods 1 and 2 can provide semiquantitative data at less
cost than method 7's (.CMB) quantitative results, if it is assumed that no CMB
source composition data are on hand.  Considering the wide range of possible
applications, no single technique can be considered the most effective.

     Note that the relative resource requirement ratings in Table 5 are by
necessity subjective.  It is, however, clear that the resource requirements of
methods 2, 3, 10 and 12 are beyond those available, on an in-house basis, to most
agencies.

     Each source apportionment tool has its unique strengths and limitations, and
each can provide valuable insight into sources contributing to air particulate
levels.  The most cost effective tool or set of tools, however, will depend on
the nature of the airshed, potential sources and the accuracy and precision of
source apportionment required.
                                        30

-------
                   i—   i—  CO  OJ   CM   OJ  CM  OJ   CO   i—
                            LO  oj   oj   oj  in
                                                              LO  OJ  LO
                            LO   r—   r—
                                                    CO   OJ
                                                                  CO  LO
cations*









LO

cu
r—
_Q
ro
I—











Q.
Q.
ct

Ol
3
cr
•r-
E
J=
O
CU
1—

-l->
E
 O
i. to
0 •!-
^•^ t— Q.
r — CO LU
N^*»

to
0)
•r- 0) O)
4-> E X3
•r- ••- 0
r- LU. S
•r—
J3
10
Q.
ro
O QJ
to
i- 

C_>
to
-a
0
O)
ro
•r*
i.
ro
•r—
r—
3

spersion modeling
•r-
Q
O
O
ro
•i—
•o
i.
1
X
croinventory
*r—
2:
dioisotope Analysis
re
Cd
                                                                           CU -Q
                                                                           +-) ro
                                                                           X D.I
                                                                           «a: ro
                                                                              o •
                                                                           T3    i
                                                                           £Z I ^ '
                                                                           ro E •(

                                                                           E^
                                                                           O E
                                                                           -U O
                                                                           i. •!-

                                                                           E S-
                                                                           cn o
                                                                           O O.
                                                                           i. Q.
ensive
source
-t->
to
O


O)
.E
•I-)

cn
E

(U
                                    LO


                                      A

                                    LO


                                     O
                                 CU  O

                           -13 •> -r-  OJ
                           E •»-> 4-> r—
                           (O +-> ro  ro
                           3 ro 4->  O
                                                                                          (1)

                                                                                       ££
                                                                                       4-> 3

                                                                                       < CO
                            I  -r- 4->
                           •i— i— C  ro
                           E fO fO
                           O! 3 3  E
                           to cr cr  o

                           II  II  II  T3
                                     Ol
                           cr_i E  cn
                           coo-cr-d
                                     3
                                                                                                      CM
            •O
             O
                  i—  OJ  CO
                                                    oo   cn  O
                                                                       CM
                                                31

-------
     The results from a single interpretive approach may be insufficient to
develop the level of confidence required in those cases where strong action is to
be taken.  Ultimately, the decision as to the likely adequacy of the analysis
must be judged on a case by case basis by the program manager.  The objective of
source apportionment studies must be to build a strong enough bridge of circum-
stantial information to quantitatively relate a source to an impact.  Thus, the
entire information base must support, and be internally consistent with, the
study's conclusions if decision makers are to have confidence that their actions
will result in improved air quality.
                                        32

-------
3.0  PROGRAM DESIGN AND MANAGEMENT
     The success of source apportionment programs  is highly dependent upon the
thoroughness with which the program is designed.   Emphasis must be placed on a
closely integrated approach which brings together  the program objectives, field
sampling program, analytical, quality assurance, data management, and data analysis
tasks.  Accomplishment of these tasks requires a well organized system for program
management encompassing technical, fiscal, administrative, and community concerns.

     As a first step in program design, specific objectives must be drafted and
extensively discussed to ensure that the end result of the program provide
useful information to the regulatory process.  Following identification of
specific program objectives,  program constraints (time and resource considera-
tions) important to program design must be known.  A detailed survey of existing
manpower, fiscal, equipment,  and data resources should be completed to identify
data gaps that need to be included in the program  plan, equipment needs, personnel
shortages and/or fiscal shortfalls.   These shortcomings must be resolved at an
early stage of the program planning process to insure that an overall, viable
program can be implemented.

3.1  Experimental Design
     There are four main phases of a source apportionment study:   ambient sampling
and analysis,  program design, development of source composition data, and data
interpretation.  Each must be optimized to attain the highest level of source
resolution and most accurate  apportionment results.
                                        33

-------
     3.1.1  Sampling and Analysis Design
          The first phase, sampling, is as critical as the other two in contri-
buting to the overall resolving power of a study.  Sampling parameters which
exert the greatest influence on source resolution include selection of ambient
samplers, filter substrates, sampling frequency and sampling duration.  These
parameters, however, are not independent and their choice will represent a com-
promise between maximizing analytical sensitivity, precisely defining the vari-
ability of the source strengths and available resources.

          Although the specific ambient sampling equipment used in the study will
depend on the receptor model(s) to be used, a dichotomous sampler with a fine to
coarse cut point of about 2 -pm is usually preferred for use with chemical receptor
models because of the bimodal  nature of the ambient aerosol and its sources.
Physical  separation of the aerosol according to particle size is the first stage
of source resolution since it divides the particles into two broad classes,
separates the basic (coarse) particles from the acidic (fine) aerosol, and increases
analytical sensitivity.

          The filter material  of choice for chemical and microscopic methods is
a thin Teflon membrane, since it minimizes artifact formation, maximizes analy-
tical sensitivity by I-ray fluorescence analysis and is compatible with micro-
scopic analysis protocol.  Although X-ray fluorescence (XRF) may not be the only
analytical technique used, it is generally accepted as being the most cost
effective analysis for chemical models.  XRF analysis cannot, however, be used to
obtain data on Na and other potentially valuable trace elements.  Table 6 lists
analytical techniques useful for ambient aerosol measurement.  Filter surface
density and blank filter trace element impurities are very low, thereby improving
                                        34

-------

























en
r-l
O
en
o
>-i
CU
>
r— 1
td
3
.
vO
CU
rH
^a
ctj
H














c
o
rt
i
J

o
*





CJ
[easurab!
Elements V


>,
~
D
O



(~-
o .-;
u 0.
a. E
,>* rt
H in




_c
S
z" s
- c
(J «
- E
BC ^

U "~>
^~ (U

C ^- ^

c >, ~
•3 S c

til
>^ c 
U -^ t—
> i- 3
c ° o
0 •? ;
2 Z =
-3 •- "^
10t5
zx 3
£ £

O CQ
£ £
1 -"
P 55
TJ "" X
0 - <
K q z"
O o ^" c
"i »• o J
a. < - >
._- .0 - w -
1/5 ^ z a." H
< ca > t/5 5


b c
0 & 0 &
o J2 o "2
O SOS
— . 	 , —


uT E E
- _^ a a
_2 ^ "*> *o
'_) E (^J fN
6 5 A V




OS
X


E »>
^33^
O, — —
kr O.
- ff C
f> £ ^
J 0 0
§ "
-^ "O O
2 C J l—
c
. . _" c/1
-3 •?• -^
-7 ^ 0)
^ •- 0
s ^ ^
0 „ ^
^ i ci
^ J tu
g if
E - -a -

1 1 §-s
Z £ § g
£«3
c « >
s J
• - u
°." -
> u >•_
t— ~- -2
_r-i
u in T3
c/i . O
, \-,
rt ffl DO
u. -• £ "
^. ^- rf <*"
C < f" J3
" ft7
— c ~ K
< M = T
j? = a °
S 0 -> -'
--?£-.
rt o fc
Z O oo uo


n
c
1 5
c3 S



m
2 S .1




<^
<;
Z
u
E
o o

O _ fc-
— 	 Q
C 3 ,2^~.
S 5 a «
^sS.'S
2 = c
'-> 0 U
§i E « E
^ _O aj _U
kXi u *O U
3"
u

Z
o"
O
c"
S
o""5
• -D
r", £
U c
^=°N.
u
M yf U
su -
» " ^
-3 T3
zo^
»' ^ ™
—* -cT _*
j etf <


"rt
C
"S E°
O ^J
0 S



O :-> ^
5^ ^
=, y s-
£ 0 E


c
o
- d,
5 o
— ' yi
< -s


fli C
H o

(J rt
ra -d
rt _
u x
t) C1
w ^
S
(^ n
y ..-
"5. c
E =
rj •_)
' £
O -j
> 3=,
0 S
^ CL
-5 c
2 -S
U j-
g -g ^
x S 5

i
0
1
m
i
u.
i ^
C
Z
1 *~'
en
CU
iH
Partic
CU
p
^
0
ja
^
-H
<;

b,
C
n
i —
O
1-4
4-1
P
C

w
a
B
PA
VJ
35

-------
analytical sensitivity.  The analytical sensitivity of JKRF for aerosols deposited
on a stretched Teflon membrane with a density of about 0.3 to 0.4 milligrams per
                        o
square centimeter (mg/cm }, for example, is about three times greater than for an
aerosol deposited on a cellulose based filter with a surface density of about
       2 7
4 mg/cm •    This difference can be translated into either more information for
the same analytical costs or the same information for a lower analysis cost.

          Selection of sampling frequency and duration will be determined primarily
by the relevant standard, potential sources and resources.  Since the current
particulate matter standards are based on a 24-hour and annual averages, sampling
duration should probably not exceed 24 hours.  Sampling for less than 24 hours
will usually substantially improve the ability of receptor models to resolve
source impacts because of the more precisely defined time dependence.  Sampling
for less than 24 hours, however, increases cost substantially and may not be
necessary to achieve the objective of a program designed to provide an initial
suryey of source impacts.

          The primary objective of the ambient aerosol chemical analysis step is
to accurately measure the major chemical components and key indicating species
(e.g., Pb for automotive exhaust, Ni and V for residual oil, Al and Si for road
dust) for input to chemical receptor models.  Although it is not essential that
all of the major chemical species be measured, it greatly improves the credibility
and confidence in the final results if most of the aerosol mass is explained.
Carbon and silicon should be measured because they represent two of the most
abundant chemical species present in a typical urban aerosol, and because they
are useful source indicating elements.  In general, Na, Mg, Al, S, K, Ca, Fe and
Pb should be measured because of their abundance and their roles in source fitting.
                                        36

-------
Other elements such as P, C1, Ti, M, Cr, Mn, Ni, Cu, Zn, Br, Rb, Sr, Zr, Cd,  In,
Sn, Ba, rare earths, etc., explain smaller portions of the total mass but may,
nevertheless, be key indicating elements.  Chemical species such as sulfate,
ammonium and nitrate ions, as well as specific organic compounds, may also be
useful.  Table 7 lists chemical species that should preferably be measured in the
fine and coarse mode ambient (and source) aerosol  if chemical receptor models are
to be used.  Fewer species have been successfully  used, however.
                                     Table 7
                         Chemical Receptor Model Aerosol
                              Analysis Requirements
                        (size resolved samples preferred)
                                                            Pb
                                                            N03
                                                            Organic Carbon*
                                                            Elemental Carbon*
mass
Na
Mg
Al
Si
S
As
Sn
Cl
K
Ca
V
Cr
Mn
Sr
Sb
Fe
Ni
Cu
Se
Br
Ba
Cd

          Sample analysis requirements of other receptor models, such as optical
microscopy, X-ray diffraction or radioisotope methods, is an integral part of the
technique itself, thereby largely bypassing the requirements described above.
     Organic and elemental carbon are measured by volatilization of the sample at
     650°C in a helium atmosphere to detect the organic carbon component followed
     by oxidation in a helium-oxygen atmosphere to measure the elemental carbon
     fraction.  Measurement of these two components have proven useful in appor-
     tioning carbon sources (wood burning, diesel exhaust, etc.)15
                                        37

-------
          Measurement of specific organic compounds  should  be  undertaken  only
after a thorough review of the expected source,  transport,  and analysis chemis-
tries has shown the candidate compounds to have  reasonably  constant production
rate, a reasonable aerosol lifetime and ability  to be measured by a relatively
low cost analysis technique.   The set of essential chemicals and preferred
analytical techniques will be determined by the  specific sources in each  air-
shed.  Selection of an optimum sampling and analysis protocol, based on a
thorough understanding of the chemistry of potential sources,  can result  in
substantial cost savings and  greatly improved source resolution.
          Important considerations influencing the success  and cost of the
program include:

               0 The Number and Location of Sampling Sites
               Sampling location in "hot-spot" nonattainment areas, coupled
with a similarly equipped site at a background location form the basis for most
studies.  Other locations in  areas of potential  industrial  growth or in areas
of differing land use may also be necessary to meet the program design requirements,
               0 Sampling Scheme
               The sampling frequency may be based on a random schedule  (usually
patterned after the SLAMS/NAMS network schedule) or on synoptic meteorological
conditions.  Other schemes designed to collect data during  air pollution  epi-
sodes, periods of visibility impairment, specific plume transport conditions or
periods of source activity may also be selected.  Diurnal or weekend-weekday
sampling schemes may occasionally prove useful in situations where the source
under study exhibits major day/night or weekend emission changes.  Regardless  of
                                        38

-------
the scheme selected, however, the relationship between the collected data and
the 24-hour worst case, seasonal or annual  averaging periods are also likely to
be of major interest.

               0 Sample Analysis Criteria
               Since wholesale analysis of  all  of the samples collected during
a study can be extremely costly, a set of sample analysis criteria are often
developed as a basis upon which samples may be selected for chemical analysis.
Whatever the criteria used, they must be consistent with the program objectives,
resources and data analysis requirements.  A basic criterion used in many
studies, for example, is to conduct analysis of all samples which exceed the
ambient air standard.

               0 Ambient Sampling Equipment
               Sampling instrumentation is  of key importance to the credibility
and overall cost of most source apportionment studies.   Selection of instru-
mentation must include many considerations:  compatibility with analytical
requirements, reliability, serviceability,  automated filter change capability
and power requirements to name a few.

               0 Analytical Methods
               The analytical methods to be used are selected after the chemical
specie to be measured, analytical precision and the detection limit requirements
are known.   The speed and cost of the analysis, data handling, and personnel
requirements and quality assurance aspects  all  have a major impact on the
program cost.
                                        39

-------
               0 Data Management
               Systems for data capture, entry and merging are often required
to manage the large volume of data generated during chemical  receptor model
studies.  Because of the large volume of data generated and the tendency to
minimize the importance of data management, these tasks often become the
single, most complex and time consuming aspect of large scale programs.   Hard
experience has emphasized the importance of a well designed data capture system
at an early stage of the project.

               0 Quality Assurance
               The likely success  of any source apportionment program is directly
tied to the quality of the field program protocols, sample handling procedures,
instrument siting, laboratory QA and data validation procedures.  Failure of
the program design to adequately address quality assurance needs tends to
compromise the validity of the results.

               0 Sources To Be Characterized
               The number and nature of the sources to be chemically (or mor-
phologically) characterized, the number of samples to be collected for analysis
and the analytical work to be conducted may potentially be important considera-
tions in developing the program budget.  In addition to the data capture
activities, resources must be allocated to the interpretation and documentation
of the source composition data.

               0 Meteorological Data Requirements
               The number, location and instrumentation to be included in surface
meteorological stations can be an important cost, especially when data capture,
                                        40

-------
instrument servicing and personnel costs are considered.   The inclusion of a
program to capture upper air data and the staff time required to process and
interpret the data, can be extremely costly unless upper air soundings are
restricted to specially selected periods.

               0 Data Analysis and Reporting Requirements
               The scope and timing of the analysis and reporting requirements,
level of technical support required and degree of dependence on consultant
assistance will play a major role in program design and funding requirements.

          During development of the program design, a few key concepts should be
upheld:  [a), the most cost effective approach to sample analysis has proven to
rest on selection of a few, critical samples rather than wholesale analysis of
all collected filters, (b] program emphasis on quality assurance, and (c) alloca-
tion of sufficient resources Cfunding and time) to assure that data analysis and
reporting tasks can be adequately completed.

     3.1.2  Development of Source Composition Data
          Since nearly all of the receptor models require a knowledge of the
chemistry or morphology of source emissions, the development of a library of
emission chemistry data or particle reference samples (or both) are often required.
Although source emission chemistry is often obtained from the literature, experience
has demonstrated the value in collection and analysis of size-resolved samples
from local  sources, just as locally derived source emission data is preferable to
published emission factors.
                                        41

-------
          Source composition data reported in the literature can be successfully
used if care is taken to insure that reported chemical  composition of a source
emission is reasonably representative of the airshed being studied.  For example,
motor vehicle particulate emission composition reported for St.  Louis (1978) may
not be representative of 1981  auto emissions in Los Angeles.  Adjustments to the
composite transportation emission composition must often be made to adjust for
differences in the leaded/unleaded fuel  use or the relative ratio of diesel  to
light duty motor vehicles in the two cities.  Since the emission composition will
vary as a function of the chemical composition of the process input materials,
the chemical "fingerprints" of particles emitted from two different steel mills,
for example, may be very dissimilar.
          Until such time that a comprehensive, size resolved particulate emission
characterization data base can be developed, analysts interested in applying
chemical receptor models will  have to research (and adapt) source data reported
in the literature or conduct tests of local sources.  Although programs are  now
underway within EPA to expand the receptor model source characterization library,
the need to adapt emission chemistry to reflect local sources will still remain.
Default emission chemistry values need to be developed to meet the needs of those
that wish to use chemical receptor models as a screening tool.
          In many studies, measurements of local sources may be required to
supplement information obtained from the literature.  In these cases, a source
sampling program must be designed which considers four key issues:

               9 The source characteristics measured should reflect those per-
ceived at the receptor.  Because of condensation of vapors, deposition of large
                                        42

-------
particles and the loss of volatile species as the plume is transported downwind,
samples taken from stacks (especially those at high temperature) may not provide
data that are chemically or morphologically similar to the aerosol measured at
the receptor several kilometers away.  The tendency to neglect fugitive emissions
that are difficult to source test is a further problem.  Fortunately, many of
these difficulties can be minimized through the use of size resolved source
testing trains that employ stack gas dilution systems.  Plume sampling from
aircraft or tethered balloon have also been used.   Attempts to deduce source
emission characteristics using ambient samplers located in expected areas of
                                       24
maximum impact have also been reported.

               0 The program design must be targeted at facilities which are
believed to be representative of similar source within the study area.  Profes-
sionals that are familiar with local manufacturing processes within the airshed
should be consulted during the selection process.   Ideally, the facilities tested
should be those that have been shown, through dispersion modeling analysis, to
have significant impacts at the receptor(s) of interest.

               0 All major emission sources should be chemically characterized,
either through a source testing program or by adaptation of literature values.
Existing emission inventory, microinventory, and dispersion modeling data are
typically used as a guide in evaluating the adequacy of the emission charac-
terization data base,

               0 The source testing protocol must  be consistent with the require-
ments of the analytical  techniques to be used.   Unfortunately, stack gas dilution
sampling designed to capture fine and coarse mode  particles on chemically inert
                                        43

-------
filters are not well documented in the open literature or commercially available.
Several sampling trains have been developed by receptor model  researchers,  however.
Further information on sampling train configurations can be obtained by contacting
EPA's Technical Support Staff,  Industrial  and Environmental Research Laboratory
(IERL), Research Triangle Park, North Carolina 27711.

          Volume II of this series discusses the application of data reported in
                                                                    15
the literature and source sampling program design in greater detail.

     3.1.3  Data Interpretation
          The interpretation stage consists of applying one or more of the
receptor model approaches to interpreting the chemical and physical data generated.
If the objective of a source apportionment study is the support of effective
control action, the level of confidence required to initiate action may be
established with a single receptor model interpretive approach or it may require
information from additional interpretive approaches such as wind sector analysis,
dispersion models, microscopy,  etc.  The first step of any source apportionment
study should therefore include  a thorough review of the potential sources,  their
chemistry and time dependence,  if the highest level of confidence is to be
established in the final results.

3.2  Typical Program Costs
     Although much is heard of the large scale source apportionment studies
typically costing hundreds of thousands of dollars in funding, the application of
receptor model techniques need not necessarily be an expensive, resource intensive
effort.  While the actual cost of a source apportionment study will depend  in
large measure on the scope of the program, a crude estimate of likely costs have
                                        44

-------
                                      25
been prepared for three example cases.     Component costs,  based on 1980 prices,



used in these examples are shown in Table 8.
                                        45

-------
Table 8.  Source Apportionment Study Cpsts;   Program Component Estimates
Ijem
Ambient Sampler
Filter Media
Filter Mass
Measurements and
Logistics
XRF Elemental Analysis
Total Carbon Analysis
Elemental and Organic
Carbon
"Short" INAA*
"Intermediate" and
"Long" INAA**
Microscopy
Cost Range
$200 to $10,000 each
$0.25 to $3.00 each
$1.00 to $10.00 per
filter

$15 to $30 per filter
$10 to $20 per filter
$25 to $50 per filter

$20 to $50 per filter
$50 to $150 per filter

$200 to $500 per filter
Estimate Used
$5,000
$2
$5

$25
$15
$35

$30
$70

$300
     * "Short" Instrumental Neutron Activation Analysis (INAA) analysis
       includes Na, Mg, Al, Br, V, Ti, K, Mn.
    ** "Intermediate" and "Long" INAA analysis provides data on As, Sb,
       Co, Cr, Fe, Hg, Sb, Se, Zn and numerous other trace elements.
                                        46

-------
Example Case 1:  Minimal Program (_$5,QOQ)_
     A minimal program costing no more than $5,000 would consist of a one-year
study of source impacts at a single sampling location at which TSP NAAQS violations
occur.  By sampling every sixth day, concurrent with routine hi-vol measurements,
samples can be collected on an inexpensive TSP sampler in a form suitable for
direct mass and chemical analysis.   Appendix I describes a low volume TSP sampler
that has been used successfully in several receptor model studies.  A separate
sampler is required to insure that the aerosol is collected on a filter substrate
that is compatible with XRF analysis requirements.  Source composition data from
literature sources Cadapted to the airshed under study) can be supplemented with
source characterization samples (ten filters) from a few key sources.  Typical
costs for such a program include:

          0 Low-volume ambient TSP sampler*       $  900
          e Teflon-substrate filters (80)         $  300
          0 Mass measurements and logistics       $  400
          9 W analysis (poor detection limits}  $  400
          p Chemical  Mass Balance calculations    $1,200
          0 Data analysis and reporting           $1.000
                                        Total     $5,000
     * See Appendix I

These costs assume that (a) field measurements are conducted by agency or company
air monitoring staff and that site rental, power and maintenance expenses are
absorbed by the agency, (b) source apportionment for ambient TSP is required to
identify annual and 24-hour impact, (c) that source testing equipment and associated
costs are paid by the agency, and (d) that agency staff develop the source matrix
data.
                                        47

-------
Example Case 2:   Moderate Program Costs ($26,000)
     An expanded, four station program sampling on an every sixth day schedule
for TSP only can be implemented for about $25,000.  Such  a  program could pro-
vide for improved analytical  sensitivity, ion and  carbon  analysis, improved
source characterization and data interpretation.   A key element in improving
the cost-effectiveness of the program is selection of 30  days  for sample
analysis based on a review of annual  meteorological patterns that can be
reflected in the days sampled.  Assuming 240 ambient filters,  30 source samples
and 30 samples analyzed for quality assurance purposes, cost can be cut by
analyzing only 180 of the 240 ambient filters.   In this example, component
costs are as follows:

          0 Low volume ambient TSP samplers*      $ 4,000
          0 Teflon substrate filters              $   900
          0 XRF elemental analysis (moderate      $ 3,600
            detection limits)
          0 Elemental and organic carbon analysis  $ 4,200
          0 Ion analysis                          $ 5,400
          0 CMB calculations                      $ 2,700
          0 Data interpretation and reporting     $4,000
                                        Total     $26,300
     * See Appendix I

     As in the first example, many program costs related  to air monitoring,
source testing and source emission characterization tasks are not reflected in
these figures.  These tasks are, however, well  within the capabilities of most
regulatory agencies using staff professionals.
                                        48

-------
Example Case 3:  Expanded Program Cost L$10.Q,QQQl
     Further expansion of the program objectives noted in Case 2 to include TSP,
fine and coarse mode particle source apportionment, more comprehensive source
characterization, and improved airshed input analysis are illustrated below.  By
sampling on an every third day basis to improve opportunities for episode analysis,
this program would provide for the capture of 1444 ambient and 100 source filters,
150 samples analyzed for quality assurance and blanks, all 1700 samples measured
for mass and 700 samples to be chemically analyzed (40 days selected for analysis).
Component costs are itemized below:

          0 Dichotomous samplers (leased)         $10,000
          9 Teflon-substrate filters              $ 5,100
          0 Mass and logistics                    $ 8,500
          0 XRF Cbest detection limits)           $18,400
          0 Elejnental and organic carbon analysis $ 5,600
          0 Ion analysis                          $22,080
          0 "Short" INAA analysis                 $22,080
          p CMB calculations                      $ 7,360
          0 Data interpretation and reporting     $10,000
                                        Total     $99,130

     More ambitious programs designed to apportion light extinction, provide
greater spatial coverage and develop data required for meteorological  data base
improvements would, of course, expand program costs greatly.   The key point to be
made here is that the application of receptor model techniques need not be an
extremely costly, research program.   Experience in the use of these methods to
resolve routine problems in air resource management has shown receptor techniques
to be practical tools.
                                   49

-------
3.3  Program Management
     The degree of program management required will depend to a great extent on
the likely consequences of the results to the community and those sources likely
to be regulated.  The following discussion is directed to those who intend to
conduct receptor model studies in support of the control  strategy development
process.  Regardless of the potential importance of the study results to the
community, however, capable program management is of central  importance to the
successful completion of receptor model studies.  Key elements of the program
management are discussed below.

     A detailed work plan for accomplishing the general tasks of the study should
be developed by those individuals responsible for conducting them.   The work plan
should include an identification and description of specific tasks  to be performed
and a detailed schedule for their completion.  The schedule should  show the
projected start and end date of each task, significant milestones to be reached
and tasks that must be completed before other activities can commence.  A staffing
and management plan assigning persons responsible for each activity is also
useful.
     In addition to work plans, an itemized budget listing costs associated with
major program tasks should be developed for both projected and actual expendi-
tures overtime.  Standard operating procedures used in the field and laboratory
programs are important to the project's quality assurance program as are proper
procedures for instrument calibration, verification of analytical methods and
data processing protocol.
                                        5Q

-------
                                   ; Figure 4
                            Organizational Chart for
                 Receptor Oriented Source Apportionment Project
                               TECHNICAL STEERING COMMITTEE
           ADMINISTRATIVE
               SUPPORT	
                                                    CITIZEN AND INDUSTRY

                                                     ADVISORY COMMITTEE
PROJECT ADMINISTRATOR
            FIELD
         OPERATIONS
DATA
MANAGEMEN
T
         FISCAL SUPPORT
    PARTICIPATING
     CONSULTANTS
 QUALITY
ASSURANCE
LABORATORY
OPERATIONS
     It may be advisable to include in the planning and implementation of the

study,  representatives of organizations (and the general public) who must make

regulatory decisions based on the results of the study.  The two key ingredients

in the  organization of a receptor model study are a technical steering committee

and an  advisory committee with a broad base of public, government and industrial

representatives.
                                                                    •
     The technical steering committee should be composed of individuals responsi-

ble for the different phases of the project.  This might include field and source

sampling programs, sample analysis, data management, meteorological analysis, and

project administration.   These representatives should come primarily from the
                                        51

-------
parent organization conducting the study but could also draw on representatives
from other government agencies, environmental  organizations and participating
consultants.  Responsibilities of the technical  advisory committee include
preparation of the study design and review of contractor-prepared documents to
the committee must assure technical integrity of the study during its imple-
mentation by frequent updates of work status,  resolution of technical problems
and coordination of all components of the study.

     The project advisory committee should have  representatives from the public
at large, environmental interest groups, those interested in public health,
local industry, Chamber of Commerce, public transportation officials, and a
representative from the Mayor's office.   Figure  4 outlines a suggested organiza-
tional chart.  Advisory comraittee functions are  to provide input into the
development of program objectives and study design and to inform their consti-
tuency of program progress and results.   The most important objective is to
ensure that advisory committee participants appreciate the technical foundation
of the program.  Community confidence in potentially unpopular control action
can be greatly strengthened through an active advisory committee program.
                                        52

-------
4.0  APPLICATIONS OF RECEPTOR MODELS
     Receptor models have been used in a number of applications including control
strategy development, stationary source enforcement actions, demonstration con-
trol strategy analysis, source impacts related to "bubble" analysis, dispersion
model validation, and general airshed source impact studies.  Although optical
microscopy has been the most widely used receptor method, studies in at least 37
cities utilizing chemical mass balance have been completed.  Factor and regres-
sion analysis have also been widely used.  Other techniques, such as spatial and
temporal analysis, have been extensively used over the years because of their
simplicity and value.  In those cases where an impact analysis of an existing
source is required, receptor models may provide an impact estimates of comparable
(or better) accuracy than dispersion models, especially in areas of complex
terrain.

4.1  Development of Control  Strategies
     Control  programs designed to attain and maintain National Ambient Air
Quality Standards for total  suspended particulate usually require that control
strategies are designed to attain standards at some future date specified in the
Clean Air Act.  This requires regulatory agencies to analyze alternative control
scenarios using projections  of future air quality.  Since receptor models cannot
be used to project air quality conditions or analyze control alternatives in all
but the most simple of cases, dispersion models must be used to demonstrate the
effectiveness of future strategies.  Figure 5 illustrates the role of receptor
models in the control strategy analysis process.   Inclusion of receptor model
analysis in the dispersion model  validation program helps to insure that control
strategy scenario analysis is based on a dispersion model that has been proven to
reasonably predict impacts from specific sources.
                                        53

-------
                                FIGURE 5

                     SOURCE/RECEPTOR MODEL  PATHWAYS
                      TO CONTROL STRATEGY ANALYSIS
        CURRENT YEAR DATA BASE

         .  Emission Inventory

         .  Meteorology
         .  Stack Parameters

         .  Operating Schedules
                         CURRENT SOURCE  DATA  BAS
                          - SOURCE COMPOSITION
                                   4
                                    AEE
                  1OTBIENT AEROSOL
                   .  Chemistry
                   .  Morphology
           Dispersion Model
               Analysis
                             Comparative
                               Analysis
                            Receptor Model
                              Analysis
 Possible Control
Strategy Senarios
H
                             Dispersion Model
                                Analysis
Validated Dispersion
       Model
   Future Year
   Projections
   (all sourcesA_
          Analysis  of  Control
           Strategy Senarios
 Projections
Point & Area
Source Emission
                                                       .  Meteorological
                               Cost & Energy
                              Analysis/Senario
                                                II
                           Strategy Selection   I

                                   t
                           SIP Revision Process I
                               SIP Adoption/
                               Approval
                                54-

-------
     Receptor model studies of the aerosol mass transported into an air quality
maintenance area (AQMA) provides important information as to the source composi-
tion of the "background" aerosol.  If the analysis identifies an important
upwind source with an impact and projected growth rate critical to the AQMA's
ability to maintain NAAQS, action may be needed to regulate the source.  In
those cases where receptor model analysis of the background aerosol is not
conducted, regulatory agencies responsible for strategy development may have
little knowledge of the identity, impact or potential emission growth of sources
outside of the AQMA boundary.

     Once a strategy has been implemented, however, receptor model techniques
can be used to track the impact of important sources and source groups (such as
crustal components-unpaved and paved road dust, soil dust, etc.).   Other appli-
cations include studies of source impacts during periods of air stagnation as
input to Emergency Action Plans, analysis of the impact of wind-entrained soil
dust on NAAQS attainment and a host of others.

4.2  Validation of Dispersion Models
     Air quality data have long been used to assess the ability of dispersion
models to simulate measured air quality.  This process, called model  validation,
is intended to lend support to the dispersion model's predictions  or provide a
basis for re-evaluating the dispersion model's data base or assumptions.   In
particulate modeling, predictions have traditionally been compared to ambient
total suspended particulate (JSP) measurements.  Unfortunately, past  experience
has shown that particulate matter dispersion models frequently fail to predict  a
significant portion of the measured TSP mass, largely because of the  difficulty
in inventorying Cand accurately simulating), fugitive dust sources that frequently
dominate the TSP mass.
                                        55

-------
     In those cases where dispersion model estimates are reasonably well  matched
to the TSP mass, little evidence of the model's ability to accurately predict
the impact of a single source, or group of sources, exists without some form of
receptor model analysis.   In addition, "traditional" forms of dispersion model
validation, which rely on comparisons of measured TSP mass to dispersion model
predictions, cannot provide the level of information needed to identify errors
in the impact estimates for specific sources.  While the ability of current
receptor models to address these concerns is limited, experience has shown recep-
tor model results to be of great value in improving dispersion model performance.

     The source-receptor CS/R) model validation process includes three key
tasks Ca) experimental study design, (b) model comparisons, and (c) dispersion
model improvements:

     4.2.1  Study Design
          The design of the S/R model analysis should include many of the consi-
derations discussed in Chapter 3 of this document.  Although existing ambient
aerosol chemistry, source composition data, emission inventory and meteorological
data can be used for the S/R analysis, the level of success of the receptor
model analysis program in identifying the impact of sources of interest will be
greatly enhanced if the project design is tailored to the receptor method,
sources and airshed of interest.
          A basic requirement of the S/R model comparison is one of concurrent
data set development.  This requires that the dispersion model analysis be
conducted for the same days, at the same sampling sites, as the receptor model
analysis.  The project experimental design, then, ideally requires that all of
                                        56

-------
the input data required for both the dispersion and the receptor model be
developed for the period in time during which the ambient monitoring program is
operational.  Emission inventories, specifically tailored for the period of
field sampling, should be used.

          The first step in building the program design should be one of selecting
the "target" sources—those emission sources for which the S/R model analysis is
to be conducted.   If the dispersion model simulates both area and point sources,
two sources within each group should be selected.  By including two types of
area sources and two point sources in the comparison, a measure of the dispersion
model's ability to simulate both types of sources is gained.   "Target" sources
are those sources which ideally meet the following criteria:

               (a)  The chemical, morphological or variable nature of their
emissions allows their impact to be accurately identified by receptor models.

               (J>)  They are major emission sources thought to impact the study
area.

               (c}  Accurate, complete emission inventories,  stack parameter
data and source operating schedules are available.

               Cd)  The composition of the source emissions is relatively constant
and well characterized.

          The area source which best meets these criteria is  leaded auto exhaust.
Soil  and road dust have also been used because of (a) their importance as a
major source impacting TSP concentration, and (b) the ability of receptor models
to identify soil-like particles (as a source group).   Resolution of soil  and
                                       57

-------
road dust into distinct sources (payed roads, unpayed roads, etc.I is, however,
difficult in those areas where both sources are found.  Point sources used as
analysis targets have included residual oil combustion sources (by reason of
their distinct Ni and V tracers) and ferromanganese steel production which is
characterized by its rich fine particle Mn content.

          Since the dispersion model estimates only encompass those "local"
sources included in the emission inventory, a compatible receptor model  analysis
must be developed through a program of "background" and "urban" aerosol  analysis.
Auto exhaust particles, for example, detected by the receptor model analysis of
incoming air, must be subtracted from the auto exhaust component measured at the
urban receptor to provide an impact estimate consistent with the dispersion
model.    Selection of the "background" monitoring site location, and the repre-
sentativeness of the urban site, are most important to the S/R comparison.

          The sampling network design must also include sufficient spatial
coverage and temporal resolution to allow comparison of the S/R model results
under a number of different transport and dispersion conditions.  In addition,
sampling equipment and sample analysis techniques used in the network must be
compatible with the receptor model requirements.
          Following completion of the aerosol sampling program and development
of dispersion model input data, an independent analysis of source impacts,  using
both model approaches, should be completed.  Independent model validation pro-
cedures should then be used to judge the validity of each model's results.

-------
     4.2.2  Comparison of Model Results
          Following validation of results from each model, annual average
dispersion model simulations are prepared for each of the target sources for
those sampling locations for which receptor model results are available.
"Local" receptor model estimates are prepared for each target source and plotted
in relation to the dispersion model results.  Figure 6(.A) shows such a plot for
results of initial dispersion model-Chemical Mass Balance predicted road dust
                            pc
impacts in Portland, Oregon.    Further investigation into any discrepancies
between the two model estimates should include a general review of the strengths
and weaknesses of both models and their input data.

          Comparisons of S/R models results for other transport conditions,
seasons of the year, source emission activities and target sources can often
resolve discrepancies by leading the analyst to deficiencies in one of the two
models.  Analysis of road dust impacts in Portland, for example, indicated con-
sistent dispersion model underpredictions which were more severe at sampling
sites in areas subject to fugitive dust impact [Sites A and C) than at commercial
(Site B) and residential locations (Site D).  The degree of dispersion model
underprediction is indicated by the horizontal distance of the point to the 1:1
slope line, assuming that the receptor model results are more nearly correct.
In the example case, optical microscopy and tnicroinventory results for the
Portland sites clearly supported the CMB results rather than those of the dispersion
model.
                                        59

-------
                                     Figure 6
                         Port land, Or. Dispersion  Model -
                            Predicted  Road Dust Impact*
                 Initial Prediction (A)
        Corrected  Inventory(B)
         ov
          . 301
         a
         E
        2  20
         E  10
             0
o
E
2  20
 a
 I  10
B
                                                                          1:1
              0-     10     20    30
      '0      10    20    30
                               Dispersion  Model Predicted

                 L  . j    .            (JIG/M3)
             North wind regime

     4.2.3   Disperipn Model Improvement

          Following  an analysis of the possible reasons  for the underprediction,

corrective  action can often be taken, thus  improving the user's confidence that

impacts  of  individual sources can be predicted with reasonable accuracy.  In the

case of  the Portland analysis, land-use specific road dust emission factors were

adopted, replacing an emission factor which ignored the  differences in soil dust

trackout between  industrial, commercial and industrial areas.  As a result, the

emission inventory was increased by 600% and the systematic underpredictions in

                                             27
the dispersion  model estimates were corrected.    Figure 6(B) shows the S/R
                                                                  *
comparison  based  on  the corrected emission  inventory.


         Upon  completion of the S/R model  validation process, the analyst can

have increased  confidence in the overall ability of the  dispersion model to

simulate both the impact of a large number  of sources and in  individual sources

as well.


                                       60

-------
5.0  REFERENCES

 1.  G. E. Gordon, "Receptor Models," Environmental Science and Technology,
     1980, 14, 792-800.

 2.  J. A. Cooper, J. G. Watson, "Receptor Oriented Methods of Air Participate
     Source Apportionment" Journal Air Pollution Control Association, October 1980.

 3.  J. E. Core, T. G. Pace, "Receptor Models - How Great Thou Art!"  Proceeding
     of the 74th Annual Meeting of the Air Pollution Control Association, June,
     1981.

 4.  "Guidelines on Air Quality Models," U.S. EPA, OAQPS, EPA-450/2-78-027,
     April 1978.

 5.  D. B. Turner, "Workbook of Atmospheric Dispersion Estimates."  PHS Publication
     No.  999-AP-26 (NTIS PB 191482), Environmental Protection Agency, Research
     Triangle Park, North Carolina, 1970.

 6.  D. L. Johnson, New York University - Syracuse, Personal Communication,
     October 1980.

 7.  J. A. Cooper, "Review of the Chemical Receptor Model of Aerosol Source
     Apportionment."  Proceedings of the ACS Symposium on Chemical Composition
     of Atmospheric Aerosols:  Source/Air Quality Relationships, 1981.

 8.  J. A. Cooper, L. A. Currie, G, A. Klouda, "Application of Carbon-14
     Measurements to Impact Assessment of Contemporary Carbon Sources on Urban
     Air Quality."  Environmental Science and Technology (in press) 1980.

 9.  J. A. Throgmorton, K. Axetell, "Digest of Ambient Particulate Analysis and
     Assessment Methods."  U.S.  Environmental Protection Agency Publication No.
     EPA-450/3-78-113, p. 108-114.

10.  C. E. Lyons, R. A. Eldred,  F. P. Terraglio, J. E. Core, "Relating Particulate
     Matter and Impacts in the Willamette Valley During Field and Slash Burning."
     Proceeding of the Pacific Northwest International Section, Air Pollution
     Control  Association, June 1979, Paper 79-46.3.

11.  "National Air Quality and Emissions Trend Report," U.S. Environmental
     Protection Agency, EPA-450/1-77-002, December 1977.

12.  G. S. Kowalczyk, C. E. Choquette and G.  E.  Gordon, "Chemical Element
     Balances Identification of Air Pollution Sources in Washington, D.C."
     Atmospheric Environment, Volume 12, p.  1143, 1978.

13.  S. K. Friedlander, "Chemical Element Balance and Identification of Air
     Pollution Sources," Environmental Science and Technology,  Vol. 7, No. 3,
     March 1973, p. 235-240.
                                        61

-------
14.  J. A. Cooper and J. G.  Watson, "Portland Aerosol Characterization Study,"
     Final Report to the State of Oregon Department of Environmental Quality,
     July 1979.

15.  "Receptor Model Guideline Series, Volume II, Chemical Mass Balance."
     U.S. Environmental Protection Agency, Office of Air Quality Planning and
     Standards, EPA-450/4-81-016b, July 1981.

16.  D. J. Alpert and P. K.  Hopke, "A Quantitative Determination of Sources in
     the Boston Urban Aerosol."  Atmospheric Environment, _H, 1137 (1980).

17.  P. 0. Gaarenstroom, S.  P. Perone and J. L. Meyer, "Application of Pattern
     Recognition and Factor Analysis for Characterization of Atmospheric Particulate
     Composition in the Southwest Desert Atmosphere."  Environmental Science and
     Technology, Volume 2,  1977.

18.  C. W. Lewis and E. S.  Macias, "Composition of Size-Fractionated Aerosol in
     Charleston, West Virginia."   Atmospheric Environment, Volume 14, 1980.

19.  D. J. Alpert, P. K. Hopke, "Modified Factor Analysis of Selected RAPS
     Aerosol  Data."  U.S.  Environmental  Protection Agency, Environmental Sciences
     Research Laboratory,  Contract Final Report "D600YNAEX: June 1980.

20.  P. 0. Warner, L. Saad,  J. 0. Jackson, "Identification and Quantitative
     Analysis of Particulate Air Contaminants by X-Ray Diffraction Spectrometry."
     Wayne County Department of Health,  Detroit, Michigan.  Presented at 64th
     Annual  Meeting of APCA, June 1971.

21.  B. L. Davis, "The Use of X-Ray Diffraction Quantitative Analysis in Air
     Quality Source Studies."  Proceedings of the 3rd Symposium on Electron
     Microscopy and X-Ray Applications to Environmental and Occupational
     Health Hazards, Colorado Springs, Colorado, September 1979.

22.  T. G. Pace, K. Axtell  and R. Zimmer, "Microinventories for TSP."  U.S.
     Environmental Protection Agency, Research Triangle Park, North Carolina,
     April 1978.

23.  Controlling Airborne Particles, National Academy of Sciences, 1979.

24.  J. G. Watson, et. al.  "The State of the Art of Receptor Models Relating
     Ambient Suspended Particulate Matter to Sources."  EPA-600/2-81-039,
     March 1981.

25.  J. A. Cooper, "Course Outline and Notes for Chemical and Physical Methods
     of Quantitative Source Apportionment."  Oregon Graduate Center, July 1980.

26.  J. E. Core, P. L. Hanrahan and J. A. Cooper, "Air Particulate Control
     Strategy Development:   A New Approach Using Chemical Mass Balance Methods."
     Proceedings of the ACS Nuclear and Environmental Division Symposium,
     August 1980.

27.  P. L. Hanrahan, "Improved Particulate Dispersion Modeling Results:  A New
     Approach Using Chemical Mass Balance."  Proceeding of the 74th Annual
     Meeting, Air Pollution Control Association, June 1981.


                                        62

-------
                                   APPENDIX I
                             Low-volume TSP Sampler

     The low-volume TSP sampler described herein was designed to capture, on an
inert substate suitable for x-ray fluorescence and ion analysis, the same total
particulate fraction that is measured by the hi-volume sampler.   Inlet geometry,
air flow velocity and cost were the three most important factors that influenced
the design.  To date, the sampler has been used in three Oregon  aerosol  studies
with good results.  A comparison of aerosol mass collected by the sampler and
collocated hi-vol federal reference method equipment is shown in Figure A.
Construction details are shown in Figure B.  Sampler specifications are as follow:
          Sample flow rate
          Sampling period
          Filter medium

          Filter holder


          Filter cap

          Filter assembly holding flange
80 1pm
24 hours (typical)
Cellulose acetate (47 mm), 1.2 micron
Millipore Corporation, RA
Nucleopore Corporation, No. 430400
aerosol holder, 47 mm (with
quick connect fitting)
4" PVC plastic pipe cap (Yardley
89644 or similar)
.05" acrylic plastic cemented to
filter cap
                                        1-1

-------
                                                                        80/03/11.     12.<'.!!.
SCATTERGRAN OF    iDOUX)   HV«aSS   TSP IWSS BI FRh
                (ACROSS)  LUBSSS   LOU-UOL TSP

                7.50     22.50     37.50     52.10     47.50     82.50     ?7.SO    111.50     117.SO    H2.30
'10.00 >
1
I * >
1 •
I •
!35.00 >
I • * »
I > «
I
I « *
120.50 t •
I » •
I 2 » «
I «
I « »
105.00 * • •
I
I • «
I **••**
I •
'0.00 > «
I . , .
I *•*••«
I » »• »
I • >
75.00 * •
1 • • * * •
1 9 » • *
1 * • * * t« t«
I «
60.00 * • » • • *
1 .3 « .
I • •
1 2 • «
I • «
45.00 • • •
I • » • 3
I «• « 2 «
I 3 «2 « •
I « 2 • »
30.00 * •• <»2« » »
I 2"
I *• 2 2>
I •• 2
I . ,.
15.00 * •••• •
I "2
I «2
I •
I
0 »
t 15.00 30.00 49.00 40.00 75.00 90.00 105.00 120.00 131.00
ST*TIS!ICS..
CORRELATION (II- .!44tl 1! SQUARE! - .8T28J SIGNIFICANCE R -
STB ERR OF ESt - 11.73373 INTERCEPT (A) - 7.4*834 ST1 ERROR OF A -
SIGNIFICANCE * - .00003 SLOPE (1) - 1.03934 STD ERROR OF B -
SICNIFICANCE 1 - .00001
Pinnvta A TCD Mace TrvmnA ri cnn l\ nu_t(rtl AVovcnc U-i — wnl T^P Ca mn 1 o v»c \
+
I
I
I
I
+
I
I
I
I
*
I
« I
I
I
+
I '
I
I
I
+
I
I
1
I
t
I
I
I
I
+
I
I
I
I
fr
I
I
I
I
*
I
I
I
I
+
I
I
I
I
+
150.00

.00001
1 .7»232
.02«0


150.00




US. 00




110.00




105. CO




?0.00




75.00




40.00




43.00




30.00




15.00




0





ft
r
                                                      1-2

-------
                               FILTER CAP
          ASSEMBLY
    HOLDING FLANGE
      A-A
                      A:
                      8:
                      C:
                      D:
                          A-A
                    3. 0 cm
                    6.2 cm
                    2.5 cm
                    0.5 cm
1  1/8"
2  13/16"
15/16"
3/16"
Scale:  I'.l'
  D:
  E:
  F-
11.5 cm 4 1/2"
 4.8 cm 1  7/8"
 1.7 cm 3/4"
 9.6 cm 3 7/8"
 6.0 cm 2 3/8"
 5.4 cm  7/8"
Figure B.  Low-volume  TSP Sampler Detail
                              1-3

-------
                                   TECHNICAL REPORT DATA
                            (Please read Instructions on the reverse before completing)
                                                    v

                                                     «
1  REPORT NO.
	EPA-450/4-81-ni6a-
4. TITLE AND SUBTITLE
                                                           3. RECIPIENT'S ACCESSION NO.
   Receptor Model Technical  Series,  Volume I
   Overview of Receptor Model  Application to Participate
   Source Apportionment
             5. REPORT DATE
                July 1981
             6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)

   John  E.  Core
                                                          8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING OR'ANIZATION NAME AND ADDRESS
   U.S.  Environmental Protecti9n  Agency
   Office of Air Quality Planning and  Standards
   Menttoring and DAta Analysis Division
   Research Triangle Park, North  Carolina  27711
                                                           10. PROGRAM ELEMENT NO.
             11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
                                                           13. TYPE OF REPORT AND PERIOD COVERED
                                                                 Final
   Same as  No.  9.
                                                           14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
        Volume I of the Receptor  Model  Technical Series presents  an  overview of
   current particulate source  apportionment methods and their  applications to
   control strategy development programs.   It is the first in  a series  of documents
   describing methods which can be  used to identify source impacts,  using data
   collected at the receptor.  This is  unlike source (dispersion)  models that esti-
   mate source strengths based on emission factors, plume behavior and  meteorology.
   Volume II of the series describes the Chemical Mass Balance Receptor Model in
   detail.  Future volumes will describe other receptor model  techniques.
        Information presented  in  this series is directed to  regulatory  professionals
   responsible for particulate control  strategy development  or related  programs
   requiring source apportionment analysis.  Major receptor  methods  are discussed,
   applications to control strategy development are presented  and areas in which
   receptor models compliment  source (dispersion) models are explored.
        Properly applied, and  with  supportive evidence developed  through independent
   approaches, receptor models can  be used independently or  in concert  with dispersion
   models, to provide important new information to regulatory  agencies.
17.
                               KEY WORDS AND DOCUMENT ANALYSIS
                 DESCRIPTORS
                                             b.IDENTIFIERS/OPEN ENDED TERMS
                          c.  COSATI Field/Group
   Aerosols
   Air Pollution
   Receptor models
   Particulate matter
   Data analysis
                                              Air Pollution Control
                                                 9

                                                 fr   «
13. DISTRIBUTION STATEMENT

   Public
19 SECURITY CLASS (ThisReport)
   Unclassified
21. NO. OF PAGES
    80
20. SECURITY CLASS (This page)
   Unclassified
                                                                        22. PRICE
EPA Form 2220-1 (9-73)

-------