United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S3-85/014 Apr. 1985
&EPA Project Summary
Receptor Models for Airborne
Organic Species
J. M. Daisey, P. J. Lioy, and T. J. Kneip
The purposes of this study were (1)
to critically review existing source
emissions data to determine if varia-
tions in the organic composition of
such emissions can be used to dis-
tinguish sources of airborne particu-
late matter and particulate organic
matter, (2) to attempt to develop
receptor source apportionment
models for three fractions of par-
ticulate organic matter and selected
polycyclic aromatic hydrocarbons, us-
ing an existing set of ambient aerosol
measurements made at the site in
New York, and (3) to define critical
needs for the development of receptor
models for airborne particulate
organic matter.
The literature search focused on
polycyclic aromatic hydrocarbons
(PAH), aliphatic hydrocarbons, car-
boxylic acids, aza-arenes, sulfur
heterocyclic PAH, and nitro-PAH. As
most of the reported data were PAH
measurements, ratios of PAH to a
reference compound were calculated
and compiled for comparisons of
source emissions profiles.
Receptor source apportionment
models were successfully developed
for three fractions of respirable (Dn =
3.5 urn) particulate organic matter and
two PAH using the factor analysis-
multiple regression modeling tech-
nique. The models included adjust-
ments for shifts in the vapor-particle
distribution due to temperature
changes.
The lack of adequate organic com-
position data for source emissions
was found to be a critical limitation
for receptor model validation and also
for development of chemical mass
balance models. Recommendations
were made for further development of
receptor models for particulate
organic species.
This Project Summary was
developed by EPA's Atmospheric
Sciences Research Laboratory,
Research Triangle Park, NC, to an-
nounce key findings of the research
project that is fully documented in a
separate report of the same title (see
Project Report ordering information at
back).
Introduction
There has been increasing interest in
and development of receptor source ap-
portionment models to assist in defining
control strategies for particulate
pollutants. Such models attempt to iden-
tify the principal sources of airborne par-
ticles and to determine their contributions
to ambient aerosol mass concentrations
using measurements made at sampling,
i.e., "receptor" sites. Most of the work in
this field has focused on particle mass
and elemental composition; little work has
been done on receptor models for par-
ticulate organic matter.
The organic fraction of the aerosol con-
stitutes 10 to 40% of the airborne par-
ticles that can penetrate the human
respiratory system. Extractable organic
matter (EOM) and many of the subtrac-
tions and compounds within this fraction
have been shown to be biologically active
in both mammalian and bacterial
bioassays. Thus, there is reason to
suspect that the organic fraction may be
of significance to human health and that
control of the sources of these materials
may be required.
The ultimate source of much of the
primary and secondary particulate organic
-------
matter in urban and suburban aerosols is
combustion of fuels for transportation,
heating, power production, and industrial
processes. Recent trends toward in-
creased utilization of diesel engines in
light-duty trucks and autos and of coal
and wood as fuels for home heating are
of particular concern, since the emissions
from these sources are rich in organic
compounds and can impact large popula-
tion centers with severe impacts in
specific neighborhoods. At present, we
lack the tools that would enable us to
determine the amounts and classes of
organic compounds (as they exist in the
atmosphere) contributed to urban at-
mospheres by various types of sources.
The purpose of this study was to in-
vestigate the potential of receptor model-
ing techniques as tools for determining
sources of airborne organic compounds or
paniculate organic mass. Specific objec-
tives were (1) to review existing data and
to determine if these data indicate that
variations in the organic composition of
source emissions could be used to
distinguish sources of airborne paniculate
matter and particulate organic matter, (2)
to empirically determine if receptor model-
ing techniques could be used to identify
the sources of airborne particulate organic
pollutants and estimate their contribu-
tions, and (3) to define critical needs in
the development of receptor source ap-
portionment models for particulate
organic species. In addition, samples of
inhalable (DM £ 15 pm) particulate matter
were collected during the Philadelphia
Model Evaluations Study of 1982 and
were analyzed for EOM and PAH.
Critical Review of Existing
Particulate Organic Source
Emissions Data
Available literature on the organic com-
position of emissions from major sources-
of organic particulate pollutants were
critically reviewed to determine if the ex-
isting data show significant differences
among sources that would be useful in
developing "fingerprints" for receptor
source apportionment models and if so,
for which classes of compounds and what
source types. Data were also sought on
unique organic or carbon species that
could be used as source tracers. The ade-
quacy of the existing data for receptor
modeling was then evaluated.
Based on existing emissions inventories,
emphasis was placed on obtaining par-
ticulate organic source emission data for
motor vehicles, home heating (gas, oil,
wood, and coal), industrial boilers, power
2
plants (oil, coal, gas), large incinerators,
coke production, petroleum refining, and
soil.
The classes of organic compounds
selected for the literature search were
those known to be present in the ambient
aerosol or in source emissions that might
be useful in distinguishing sources of par-
ticulate matter: PAH, aliphatic hydrocar-
bons, carboxylic acids, aza-arenes
(nitrogen heterocyclics), sulfur hetero-
cyclic PAH, and nitro-PAH.
Copies of approximately 180 publica-
tions were obtained and reviewed, but
few of these reported data that were
suitable for the purpose of this review.
The majority of those that were suitable
reported PAH measurements. For each
source sample, the ratios of individual
PAH to a reference compound,
benzo(e)pyrene, were calculated and
these data were compiled for comparisons
among sources. Major factors to be con-
sidered when comparing the existing
organic source composition data were
reviewed and discussed. These include
the effects of combustion conditions,
sampling protocols, sample preservation,
extraction, and analytical methods on the
reported PAH or other organic compound
profiles.
Despite their many limitations, the ex-
isting data were judged to provide certain
useful information for receptor modeling.
First, the PAH profiles of sources that
have been repeatedly sampled and anal-
yzed by the same investigators appear to
be quite reproducible. Second, the ex-
isting data indicate that there are organic
compositional differences that can be ex-
ploited in differentiating certain sources.
Furthermore, the data provide indications
of compounds that should be more fully
investigated. The PAH, in particular, ap-
pear most promising. Their proportions in
emissions (gas plus particle) from a given
source type frequently vary over several
orders of magnitude, which enhances
their potential usefulness for source
discrimination. There also appear to be
several unique or almost unique PAH
tracers that may be useful. The existing
data provide a basis for selecting those
compounds within this class that are likely
to be most stable in the atmosphere. In
addition, good sampling and analytical
methods already exist for this class of
compounds.
There are, however, many deficiencies
in the existing data on the organic com-
position of source emissions. The data
have generally been collected for the pur-
pose of determining emission rates and
are consequently inadequate for use in
receptor source apportionment modeling.
Specific deficiencies are the following:
1. Data generally exist only for the PAH,
and even these have not been ade-
quately measured in many sources of
interest. Other composition variables,
e.g., other classes of organics, trace
elements, carbon, and particle mass
have rarely been simultaneously
measured for the same source.
2. Existing organic composition data are
rarely representative of the average
emissions for a source type. Usually,
they represent a single source and a
single set of operating conditions for
that source.
3. Sampling methods used for source
emissions are frequently incompatible
with those used for ambient samples
collected at a receptor site. In the case
of source sampling, organic vapors and
emitted particles are both collected.
For ambient samples, organic vapors
are rarely collected with the particles.
In addition, sampling duration and filter
loading differ for the two types of
samples; this may also affect composi-
tion.
4. Differences in the particle size cuts of
collected samples affect composition.
5. The analytical methods and quality
assurance practices used to collect ex-
isting source composition data vary
widely. For receptor modeling, both
accuracy and precision of the measure-
ment are critical.
Finally, it should be recognized that
changes in the composition of source
emissions can occur as a consequence of
regulation and technology; consequently,
receptor and source sampling and analysis
methods and measurements must be con-
tinuously updated for receptor modeling.
Development of Receptor
Source Apportionment
Models for Particulate
Organic Matter
To evaluate the feasibility of developing
receptor source apportionment models for
particulate organic matter, an existing set
of ambient aerosol measurements was
analyzed using a factor analysis-multiple
regression modeling method that had
been previously developed in this
laboratory. This method was selected
because the lack of adequate organic
source emissions composition data
precluded the development of chemical
mass balance models.
Weekly samples of respirable (Deo =
3.5 jtm) particulate matter that had been |
collected on the roof of a 15-story "
-------
building in the Manhattan section of New
York City during 1979 and 1980 had been
analyzed for three fractions -of EOM,
PAH, several trace elements (V, Pb, Cu,
Mn), and water-soluble sulfate. This data
base was extended by X-ray fluorescence
(XRF) analysis for additional trace
elements (Fe, Mo, P, Br, Ti and Ni) and a
second analysis (XRF) for Pb, V, and S.
The concentrations of three organic frac-
tions of increasing polarity were deter-
mined by sequentially extracting the par-
ticulate samples with cyclohexane (CX),
dichloromethane (DCM), and acetone
(ACE).
Factor analysis was used to identify ma-
jor sources of airborne particulate organic
matter and select source emissions
tracers. Six to nine factors were obtained,
depending on the number of variables in-
cluded in the factor analysis. A consistent
pattern of factors was observed; the
following major sources were identified
through the high factor loadings for cer-
tain source tracer elements: (1) oil burning
(V); (2) motor vehicles (Pb, Br); (3)
resuspended soil (Fe, Ti, Mn); (4) sulfate-
related aerosol (S04°, S); and (5) incinera-
tion (Cu). These factors were consistent
with the known sources of airborne par-
ticulate matter in New York City, where
residual oil is the major fuel used for
power generation and space heating.
Neither wood nor coal are currently used
as fuels.
Multiple regression models of the form
[Organic fraction] = k,T, + k2T2 + —
~r~ Kj I j + n
were developed, where the T, are concen-
trations of source tracer elements selected
from the factor analysis, the k, are multi-
ple regression coefficients, and R is the
portion of the concentration of organic
matter that cannot be attributed to any of
the sources in the model. The coeffici-
ents, which are determined from the am-
bient measurements, are proportional to
the ratios of particulate organic matter to
tracer elements in the source emissions.
For model verification these coefficients
were compared to available source emis-
sion data. The multiple regression models
were then used to estimate the average
contributions of various source types to
each organic fraction.
The coefficients of the selected source
emissions tracers and the values of R for
the models are summarized in Table 1.
The contributions of each source type
estimated from the models are presented
in Table 2. The models indicated that
esidual oil burning and resuspended soil
ire the major sources of respirable par-
Table 1. Multiple Regression Coefficients of Extractable Organic Matter Models
Coefficients of tracer elements of models ± S.E.a
Organic fraction
(dependent variable) V
Ti
Pb
Cu
CX
DCM
ACE
25
29
± 3
b
± 6
20±
55±
42 ±
12
13
24
1.1
± 0.4
b
b
b
0.11 ±
0.28 ±
0.06
0.10
b
b
22±
10
1.2 ± 0.3
0.28 ± 0.29
1.4 ± 0.6
'Coefficients ± the standard errors of the coefficients are for independent variables expressed in units
of micrograms per cubic meter.
hNo statistically significant coefficient obtained.
Table 2. Summary of Estimated Contributions of Various Sources to the Average Ambient
Concentrations of Extractable Organic Matter in New York City for 1979 and 1980"
Source Type CX" DCMb ACE11 Total EOM c
Residual Oil
Motor Vehicles
Resuspended Soil
Incineration
Sulfate- Related
Unattributed
1.06 ± 0. 13
0.55 ± 0.21
0.33 ± 0.20
d
d
1. 17 ± 0.33
d
d
0.92 ± 0.22
d
0.44 ± 0.22
0.28 ± 0.29
1.26 ± 0.27
d
0.71 ± 0.41
0.54 ± 0.25
1.08 ± 0.39
1.36 ±0.63
2.3 ± 0.4 (24%)
0.55 ± 0.21(6%)
2.0 ± 0.8 (21%)
0.54 ± 0.25(6%)
1.5 ± 0.6 (15%)
2.8 ± 1.2 (29%)
'Contributions reported in units of fig/m3; the last column of the Table presents the contributions of
each source type as a percentage of total EOM.
bCX = cyclohexane-soluble organic matter; DCM = dichloromethane-soluble organic matter; ACE =
acetone-soluble organic matter.
'Total EOM = CX + DCM + ACE.
dNo statistically significant coefficient obtained.
ticulate organics at this site during 1979
and 1980. These two sources accounted
for 24 ± 4% and 21 ± 8% of the total
EOM (CX + DCM + ACE), respectively.
Sulfate-related organic aerosol accounted
for an additional 15 ± 6% of EOM, while
motor vehicles and incineration each con-
tributed 6% at this rooftop site in
Manhattan.
The models indicated that the sources
of organic aerosol can be different for
fractions of different chemical composi-
tion. For example, sulfate-related organic
aerosol was found only for the more polar
DCM and ACE fractions, which would
contain any secondary organic aerosol.
Twenty-nine percent of the total EOM
could not be attributed to known sources
but is suspected to originate in part from
upwind urban and regional background
sources. Unidentified sources and filter
sampling artifacts may also contribute to
the residual. The samples used for model
development were collected over periods
of one week in order to maintain continui-
ty with total suspended paniculate matter
(TSP) sampling that was begun in 1967.
For organic sampling, 12- or 14-h samples
would be more appropriate as volatiliza-
tion losses and filter reactions would be
minimized.
Seasonal variations in temperature and
aerosol surface area can affect the
distribution of organics between the vapor
and particulate phases and, thus, might
account for some of the residual, R, of
the model for CX. A first-order approx-
imation was used to correct for such ef-
fects and account for some of the
residual. The results indicated that only
3-4% of the residual could be explained in
this way.
The coefficients of the model for the
CX-soluble organic fraction were in good
agreement with the ratios of this fraction
to the tracer elements in source emission
samples in those instances in which
source emissions data were available. No
source emission data were available for
such comparisons for the two other
organic fractions.
Some exploratory work has been done
on developing similar source apportion-
ment models for individual PAH. Using a
limited set of data, statistically significant
models were developed for benzo(a)-
pyrene and chrysene. The coefficients of
the models were in good agreement with
the few data available for source emis-
sions. Since the distribution of chrysene
between particulate and vapor phases is
strongly temperature dependent, a
Langmuir model was used to determine
total chrysene concentrations based on
particulate concentrations, ambient
temperature, and the heat of sublimation
of this compound. A second chrysene
model was then developed for total
chrysene (vapor plus particulate). Work
on PAH models is continuing with a much
3
-------
larger set of PAH measurements. In addi-
tion, the use of target transformation fac-
tor analysis to develop source emissions
profiles for paniculate organics based on
ambient measurements is under investiga-
tion.
The Philadelphia Study
As part of the Philadelphia Model
Evaluations Study of 1982, samples of in-
halable (DM < 15 ptn) paniculate matter
were collected for organic analyses. The
purpose of this work was to provide infor-
mation on the concentrations of EOM and
PAH and, if possible, use these data for
further model development.
Samples were collected every 24 h be-
tween July 25 and August 14, 1982 at the
Fireboat Station located at the intersec-
tion of Delaware and Allegheny Streets;
the Delaware River is to the east and In-
terstate Highway 95 is two blocks to the
west of the site. Heavy diesel truck traffic
was observed on Delaware and Allegheny
Streets.
The geometric mean concentrations of
the CX, DCM, and ACE fractions during
this period were 2.4, 1.9, and 7.6 /*g/m3,
respectively; daily variations in concentra-
tions are shown in Figure 1. The average
concentrations for the CX and DCM frac-
tions at this site were both 0.4
Philadelphia • Fireboat Station
7/25 7/29 8/2 8/6 8/10 8/14
Date. 1982
Figure 1. Variations in concentrations of
cyclohexane(CX)-, dichlorometh-
ane(DCM)-. and acetone (ACE)-
soluble organic matter at the
Philadelphia Fireboat Station.
higher than those observed at our site
located in Camden, NJ, to the south of
this site; the ACE concentration at the
Philadelphia site was 1.6 pg/m3 higher.
There was evidence of weekday-weekend
variations in the concentrations of CX-
soluble organic fraction; minima were
observed on the weekends. Some inter-
site (Philadelphia and Camden) correlation
was apparent for all three fractions, but
the strongest correlation (r = 0.70, p =
0.01) was observed for the polar ACE
fraction, which contains oxidized
hydrocarbons and secondary organic
aerosol formed during summertime smog
episodes.
The concentrations of 12 PAH (fluoran-
thene, pyrene, benz(a)anthracene,
chrysene, benzo(b)fluoranthene, benzolk)-
fluoranthene, and benzo(j)fluoranthene,
benzo(e)pyrene, and benzo(a)pyrene,
perylene, benzo(ghi)perylene and
indeno(1,2,3-cd)pyrene were measured
daily. The average concentrations of each
compound during weekdays and
weekends of the sampling period are
presented in Table 3. Concentrations of
individual compounds were generally less
than 1 ng/m3. The geometric mean con-
centration of benzo(a)pyrene during this
period was 0.13 ng/m3, and values ranged
from 0.03 to 1.07 ng/m3. The average
concentration of all PAH was lower on
the weekends than on weekdays.
Less motor vehicle traffic (diesels in
particular) in this area on the weekends
was hypothesized as the reason for lower
concentrations of both CX-soluble
organics and PAH compounds. Both PAH
compounds and the CX fraction materials
are directly emitted from combustion
sources, i.e., they are primary aerosol
species. Source apportionment modeling
was not possible due to the lack of ap-
propriate organic source emission data.
Recommendations
The existing literature on the organic
composition of source emissions,
although somewhat limited, suggest that
it may be possible to distinguish certain
types of sources by differences in the
organic composition of their emissions,
e.g., emissions from diesel and spark-
ignition engines, resuspended soil and
coal combustion, residential oil, coal, and
wood combustion, and motor vehicles
and coke ovens. The results of the recep-
tor source apportionment modeling work
for three fractions of EOM and selected
PAH indicate that it is feasible to develop
such models to estimate the contributions
of various types of sources to airborne
paniculate organics. However, it is clear
from this study that further progress in
these areas will require suitable organic
composition data on important combus-
tion sources. The following recommenda-
tions are made:
1. A source emissions sampler suitable
for collecting particulate organics that
are in temperature and pressure
equilibrium with vapor phase organics
and the vapor phase organics should
be developed and field tested. Field
testing should involve comparisons of
such samples with those collected by
the modified EPA method 5 and those
collected downwind in plumes from
sources.
2. It is recommended that a carefully
designed winter (to avoid
photochemical changes) field study be
conducted using the PAH as model
compounds in order to confirm that
differences in organic source emissions
exist that can be used for
distinguishing sources of receptor
modeling. Ambient and source samples
should be collected at a site influenced
Table 3. Average Weekday and Weekend Concentrations of Polycyclic Aromatic Hydrocarbons*
Compound
Weekdays
Weekends
Fluoranthene
Pyrene
Benzlalanthracene
Chrysene
Benzolbtfluoranthene
Benzolklfluoranthene
Benzoljlfluoranthene
Benzoletpyrene
Benzol a Ipyrene
Perylene
Benzotghilperylene
Indenol1,2,3-cd)pyrene
Mean of Total PAH
0.74 ± 0.57
0.57 ± 0.45
0.35 ± 0.41
1.09 ± 1.29
0.40 ± 0.35
0.28 ± 0.36
0.50 ± 0.36
0.16 ± 0.79
0.36 ± 0.37
0.06 ± 0.08
0.63 ± 0.55
0.32 ± 0.36
6.28 ± 5.21
0.25 ±
0.16 ±
0.05 ±
0.35 ±
O.JO ±
0.05 ±
0.19 ±
0.20 ±
0.06 ±
0.003 ±
0.16 ±
0.06 +
1.69 ±
0.14
0.11
0.04
0.31
0.08
0.05
0.12
0.18
0.04
0.003
0.16
0.07
1.14
*/? = 13 for weekdays; n = 5 for weekends. In calculating the average concentrations ± one standard
deviation, one half of the detection limit was used for values less than the detection limit. Values in the i
Table are expressed in nanograms per cubic meter. *
-------
by just a few major source types such
as automobiles, diesels, and oil
heating. Source and ambient samples
would have to be collected in such a
way that the composition data would
be comparable. Vapor phase and par-
ticulate PAH should be collected for all
samples. The source samples must be
sufficiently representative of the
average conditions of each source
type.
3. Portions of the samples or matched
samples should be analyzed for trace
elements and other composition
variables. The best possible quality
assurance practices should be used for
all organic and inorganic analyses. In
view of the impending loss of Pb and
Br as tracers of motor vehicle emis-
sions, more extensive chemical
characterization of emissions from
catalyst-equipped vehicles and diesel
vehicles should be undertaken to try to
identify individual compounds or pat-
terns of organic compounds that might
be useful as future tracers of motor
vehicle emissions.
4. If the experiment outlined in recom-
mendation 2 is successful, more exten-
sive source emissions testing should be
undertaken to define the organic and
inorganic composition of important
sources emitting toxic substances.
Sampling equipment and protocols
suitable for organics such as dilution
samples should be used.
J. M. Daisey. P. J. Lioy. and T. J. Kneip are with the New York University Medical
Center. New York. NY 10016.
James L. Cheney is the EPA Project Officer (see below).
The complete report, entitled "Receptor Models for Airborne Organic Species,"
(Order No. PB 85-172 583/AS; Cost: $19.00, subject to change) will be
available only from:
National Technical Information Service
5285 Port Royal Road
Springfield. VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC27711
ifrU.S.Government Printing Office: 1985 — 559-111/10818
-------
United States Center for Environmental Research
Environmental Protection Information
Agency Cincinnati OH 45268
Official Business
Penalty for Private Use $300
OCOC329 PS
U S ENVIR PROTECTION AGENCY
REGION 5 LIBRARY
230 S DiARBCRN STREfT
CHICAGO IL <5Q«04
------- |