United States
                     Environmental Protection
                     Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
                            >.-,
                                 •v
                    Research and Development
EPA/600/S3-88/026 Sept. 1988
&EPA          Project  Summary
                     Development  and  Evaluation  of
                     Composite   Receptor  Methods
                     Thomas G. Dzubay
                      A  composite  receptor  method for
                    PM-10 apportionment was evaluated to
                    determine the stability of  its solutions
                    and to devise cost-effective measure-
                    ment strategies. Aerosol samples used
                    in the evaluation were collected during
                    summer,  1982, by dichotomous  sam-
                    plers at three sites in the vicinity of
                    Philadelphia, PA. The composite recep-
                    tor method consisted   of a  wind-
                    trajectory method,  chemical mass
                    balance (CMB), and multiple linear
                    regression  (MLR). Several  industrial
                    sources were determined by CMB, and
                    vehicle exhaust and a sulfur-component
                    were determined by MLR. Measured Pb
                    minus  a  CMB-derived correction for
                    other sources was used as an indepen-
                    dent variable in MLR, and MLR results for
                    the Pb abundance in vehicle exhaust
                    agreed with predictions of a model for
                    vehicle emission factors. In resolving PM
                    10 into 11 components, scanning elec-
                    tron microscopy was essential for coal-
                    fly  ash and  botanical matter,  x-ray
                    fluorescence (XRF) was needed for the
                    sulfur-component and vehicle exhaust,
                    and  instrumental  neutron  activation
                    analysis (INAA) was essential for flui-
                    dized catalytic crackers at refineries. The
                    remaining components were determined
                    well by either INAA or XRF. Ten to twen-
                    ty samples were sufficient to determine
                    average source contributions by CMB. At
                    least 50 samples were needed to deter-
                    mine vehicle exhaust and the sulfur-
                    component by MLR.
                      This Project Summary was developed
                    by ERA's Atmospheric Sciences Research
                    Laboratory, Research Triangle Park, NC,
                    to announce 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
  Receptor models are  techniques for
using  data on  the composition of at-
mospheric aerosol to resolve paniculate
mass  concentration into  components
related to emission sources. The U. S. En-
vironmental  Protection  Agency has
specified that receptor models may be used
with dispersion models to develop plans to
meet  its new PM-10 standard  and has
developed guidelines on how the chemical
mass balance (CMB) receptor model is to
be used. The present report explores how
the resolution and accuracy of  the CMB
model can be  improved by using it in con-
junction with other receptor methods and
by including measurement data by a variety
of methods including x-ray fluorescence
(XRF),  instrumental  neutron  activation
analysis (INAA), scanning  electron
microscopy (SEM), ion chromatography (1C)
and x-ray powder diffraction (XPD).
  This report is based on data obtained in
a study intended to develop and improve
the capabilities of receptor models. Am-
bient aerosol was collected in the PM-10
size range (particle diam < 10 ^m) at three
sites in the Philadelphia area during sum-
mer, 1982.  Source emissions were col-
lected using a dilution-cooling technique at
seven  major stationary sources. Surface
soil and dust were collected at multiple sites
and suspended by an aerosol generator.
Ambient, source, and soil samples were
deposited on  filters  by  dichotomous
samplers and analyzed by XRF, INAA, 1C,
pyrolysis, SEM and XPD.
  The  Philadelphia study led to the
development  of a  composite  receptor
method that uses several different recep-
tor methods in concert. It consists of a wind-
trajectory method to identify significant
sources of trace elements, CMB to deter-
mine their contributions to PM-10, and CMB

-------
combined with multiple linear regression
(MLR) to determine a sulfur-component and
vehicle exhaust.  The composite method
was developed to overcome limitations that
those receptor models  have when  used
individually.
  In developing the composite method, it
was recognized that three criteria must be
met for a chemical species to be included
in a CMB:
(1) All sources of the species must be in-
   cluded in the CMB.
(2) The  species' abundance  in  parti-
   culate emissions from  all  of  those
   sources must be known.
(3) Sources with  collinear signatures
   (i. e. similar species abundances) can-
   not be determined individually and must
   be included in a broadly defined source
   category.
  Although signature data have been com-
piled for more than 120 sources, collinearity
becomes severe  in a typical CMB if more
than only 6 to 12 components are included.
Thus, the wind  trajectory  method was
valuable for selecting sources to include in
CMB analyses. Components in CMBs  for
fine particles were crustal matter and emis-
sions from municipal incinerators, oil-fired
power plants, fluidized catalytic cracker, an-
timony roaster, and paint pigment manufac-
turing. Because signatures for coal-fly ash
and soil were collinear, criterion 3 required
that they be included in a broadly defined
crustal source category- Elements included
in CMB analyses were Si, Ti, Fe, Ni, Zn, Cd,
Sn and Sb measured by XRF and Al, V, La,
Ce and Sm measured by INAA.
  Criterion 2 led to the exclusion of Pb and
S from the CMB because their abundances
were not known  well for vehicle exhaust
and a sulfur-component. The latter con-
sisted of sulfate, cations and possibly water
or organic matter. Multiple linear regression
(MLR) can be used to estimate S- and Pb-
abundances, but measured  Pb could not
be used as an independent variable in MLR
because vehicle exhaust was not the only
significant source of Pb. Thus, the  above
described CMB was applied to each am-
bient sample  to determine non vehicular
Pb, which was subtracted from measured
Pb to yield a vehicle-exhaust tracer. This
work  describes  application  of  MLR  to
various data subsets chosen to reveal any
dependence on sampling site, sample size
and measurement method.
  Apportionments presented here include
SEM results for botanical matter and coal-
fly ash in coarse particles. SEM provided
estimates for several other components in-
cluding  emissions  from municipal  in-
cinerators. For samples heavily impacted
by municipal incinerators, SEM confirmed
CMB results. However, cases were iden-
tified  where  CMB  results  indicated
municipal incinerator emissions when none
were detected by SEM. Evaluation of those
cases suggested possible departure from
criterion 1 because Zn, an important tracer
for incinerators, had other possible sources
not included in the CMB. Thus, this work
includes an evaluation to determine the ef-
fect of cases with high Zn concentrations
on the results of the  composite receptor
method.

Results
  The composite method enabled PM-10 to
be  resolved  into  11  components. The
method provided  an estimate  of  the
average Pb abundance in vehicle exhaust
when incinerators and oil combustion con-
tributed a portion of the Pb. Final results
indicate that the vehicle exhaust portion of
PM-10 ranged from 5% to 10%. Stationary
sources included in the CMB contributed
less than  5%  to PM-10.  The  sulfur-
component contributed 54 + 10% of PM
10. Wind-stratified data indicated that 80 +
2O% of particulate S was from regional
sources beyond the  Philadelphia area.
Multiple linear regression of S vs.  tracers
Se and either V or Ni attributed 72 + 8 and
16 + 5% of particulate S to coal- and oil-
fired power plants,  respectively.
  The combined use of XRF and INAA
enabled more sources to be resolved than
was possible when only one method was
used. XRF data were essential for obtain-
ing accurate determinations of the large
sulfur-component and vehicle exhaust.  IN-
AA was essential for determining a 0.2
fiQlm3  contribution  by fluid  catalytic
crackers at refineries. SEM was essential
for determining  0.2 and 2 /tg/m3 contribu-
tions of coal-fly  ash and botanical matter,
respectively. Remaining components were
determined well by either  INAA or XRF
data.

Conclusions
  Evaluation of data for 156 samples from
the Philadelphia study led to the design of
a cost-effective  plan for PM-10 apportion-
ment by receptor models. Because INAA
and SEM-EDX data were not directly used
in MLR, it was only necessary to analyze
enough samples  to provide  adequate
estimates of the means. Standard errors
computed from the standard deviations of
CMB-derived component concentrations
suggest that 10 to 20 randomly selected
samples analyzed by INAA and SEM-EDX
should be sufficient for major species. Both
fine and coarse fractions should be ana-
lyzed by INAA rather than only the fine frac-
tion as was done in the Philadelphia study.
It  is appropriate  that all samples  were
analyzed by XRF because XRF analysis is
less  expensive  and  many  analyzed
samples were needed for the composite of
CMB and MLR. Analysis of data subsets by
MLR indicates that the sulfur-component
was determined well with as few as 31
samples, but the  vehicle component re-
quired more than  50 samples.
  Analysis  of  sulfate  collected   in
dichotomous samplers was of little value in
the present study other than to confirm that
XRF could be used to measure S in sulfate.
In future  studies, it would be more useful
to collect samples with annular denuder
systems so that the nitrate  volatilization
problem can be avoided and both gaseous
and particulate S  and N compounds can
be analyzed.
  In the Philadelphia study, two dichoto-
mous samplers  were  used to collect
samples for XRF and SEM analyses. The
use of two dichotomous samplers is  not
only expensive, but the resulting fine frac-
tion samples are unsuitable for analysis by
SEM. A modified dichotomous sampler was
developed which enables one device to col-
lect both fine and coarse particles that can
be analyzed by XRF, INAA and SEM.
  The  abundance  of  S in the sulfur-
component derived  by CMB-MLR was at
least 19% lower than that of S in pure
sulfate salts. Measurement error can  ac-
count for  no more  than  half of the  dif-
ference. Much of the difference may be due
to water  or  carbon  bound to sulfate
particles.
  If Pb is used to determine vehicle ex-
haust by CMB analysis, the Pb abundance
needs to be determined. In  the present
study, it was shown that the abundance of
Pb in vehicle exhaust could be derived by
CMB-MLR, and the results  agreed with
predictions of a comprehensive emissions
model. If the model and its parameters
were fully validated, it could  be used in-
stead of an  expensive set  of  field
measurements  to derive the Pb  abun-
dance. However, not all parameters in that
model have been validated, and the model
does not include particulate mass resulting
from condensation  of exhaust gasses  in the
atmosphere. Until the model is validated for
use with  CMB, a composite of CMB and
MLR will  be  needed to  determine the Pb
abundance.
  Decreasing use of Pb in fuel will  make
it difficult to use Pb as a tracer for vehicle
exhaust in future years. The vehicle model
predicts  & factor of 20 decrease in Pb-
emissions per unit of VMT between 1982
and 1990. Such a decrease implies that if

-------
the measurements described in this report
were repeated in 1990, vehicles would con-
tribute only 20% of ambient Pb. Such a low
vehicle  contribution  will  be  a  difficult
challenge even for the composite of MLR
and CMB. One way to overcome this dif-
ficulty might be to use the composite of
CMB and MLR to deduce the Pb abun-
dance in vehicle exhaust sampled near a
roadway. Because Pb emission rates differ
widely among various classes of vehicles,
it is important that the roadway sampling
site represents the vehicle mix and driving
speeds of areas where PM-10 is to be ap-
portioned. An alternative would be to use
another tracer for vehicles. A tendency for
Br to volatilize in the presence of  acid
aerosol makes Br a tracer of questionable
quality. Carbon monoxide may be a better
one, but, in future years, as the proportion
of vehicles with control devices for CO in-
creases, non vehicular sources of CO are
likely to interfere with CO from vehicles.
The EPA author Thomas G. Dzubay is with the Atmospheric Sciences Research
  Laboratory, Research Triangle Park, NC 27711.
The complete  report, entitled "Development and Evaluation  of Composite
  Receptor Methods," (Order No.  PB 88-234 0677AS; Cost: $14.95, 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 authors can be contacted at:
        Atmospheric Sciences Research Laboratory
        U.S. Environmental Protection Agency
        Research Triangle Park, NC 27711

-------
United States
Environmental Protection
Center for Environmental Research
Information
Cincinnati OH 45268
us.OFFICIAL WAIL;
       f  .!'-"* n ii r- r i f> c"~^

Official Business
,-.'.-,-= ;
/ ' * t
"^••;,'.V '"r :
-7r P'
'-- v>
U v ; /' •,' [
« r
t, vJ .
• r
?
*s*
" <'**
Penalty for Private Use $300
EPA/600/S3-88/026


           OOOQ329

-------