OCR error (C:\Conversion\JobRoot\000007PA\tiff\2000SUKT.tif): Unspecified error

-------
                           DISCLAIMER

     report was  furnished to the U.S. Environmental  Protection
     . >  by  the  student identified on the cover page, under a  National
^Network for  Environmental Management Studies fellowship.

The  contents  are essentially as  received  from the author.   The
opinions, findings,  and conclusions  expressed  are those  of the author
and  not necessarily  those  of the  U.S.  Environmental Protection
Agency.  Mention,  if any, of company, process,  or product  names is
not to  be considered as an endorsement  by the U.S. Environmental
Protection  Agency.

-------
Applicability of Receptor Modeling to the
 EPA Great Lakes Toxic Deposition Study
                   by
             Michael A. Ling
              May 1, 1991

-------
ABSTRACT

     This paper reports the results of a preliminary receptor modeling
study which was set up to resemble an air toxics apportionment study
which might be a component of the upcoming Great Lakes Toxic Deposition
Study conducted by the United States Environmental Protection Agency
(EPA).  The key element of this study was the Chemical Mass Balance
computer model, Version 7 (CMB7), which was applied to an existing data
set from the Detroit area. During this application, possible data gaps and
existing needs relevant to the Great Lakes study were identified, and
relevant recommendations were made.
     The CMB7 model was applied to six 24-hour samples,  and a 'best fit'
was obtained for each sample. In general, the source apportionment
results suffered from three problems.  First, they were unable to account
for large portions of the particulate mass. Second, the precision of the
models was probably too low to resolve low-mass toxic species. Third, the
models omitted some sources because  profiles were either missing or of
poor quality, or key tracer species were missing from the ambient data.
     The problems encountered point to the need for better receptor data
including more species, better precision (especially for key tracer species),
and quality assurance  In addition, better source profile data are needed,
including new profiles for missing industries and quality assurance  for
existing profiles. In the best case, source-specific data would be used in
the model in place  of 'representative' profiles from the literature. Finally,
the problems point out the limitations of this approach when applied  to air
toxics  The results are limited by the detection limits of the analytical
equipment, and by the fact that some important sources of toxics are  at the
'noise level' of the model  resolution. Other modeling techniques are
needed  to complement the chemical mass balance, and the role of chemical
mass balance must be clarified before  a long-term data collection effort is
undertaken

-------
INTRODUCTION

     The ultimate goal of this study was to assess the applicability of
receptor modeling to the Environmental Protection Agency (EPA) Great
Lakes Toxic Deposition Study (referred to hereafter at the 'Great Lakes
Study').  This study, mandated by the  1990 Clean Air Act Amendments, is
intended to quantify the portion of toxic pollutants in the Great Lakes
which comes  from the deposition of airborne toxics. The majority of this  -
toxic deposition comes from atmospheric particulates which are washed
out during  rainfall events, or which settle directly into the water.  Thus,
characterization of sources of airborne toxic particulates is essential to the
Great Lakes Study
     Of potential use in this characterization are the techniques of
receptor modeling Receptor modeling is a tool which is used to quantify
the impacts of various sources of atmospheric particulates on the total
particulate load and on the loads of individual species. The calculations are
based on information collected at ambient samplers, or receptors, rather
than at the source (which is the more common approach). This receptor
information may include species concentrations, particle morphology,
chemistry,  and variability information.
     To assess the applicability  of receptor techniques to the Great Lakes
study, a simple preliminary study was set up (referred to hereafter as the
'preliminary study')  A study was designed which would reflect conditions
similar to those m the Great Lakes study, which would not require
additional expensive monitoring and sampling programs and which could
be accomplished within a relatively short time frame (i.e. the summer of

-------
1991)  This study was built around the EPA's Chemical Mass Balance
software, Version 7  (CMB7)3, and used already-existing ambient air
quality data from a long term monitoring program in the Detroit-Windsor
metropolitan area 6
     As an additional component to the preliminary study, a literature
search was conducted on receptor modeling techniques. This search was
conducted to find other successful receptor modeling studies and similar
reports. This information was intended to aid in both the implementation
of the preliminary study, and in the process of making recommendations - _
for the Great Lakes study
     Based on the results from the preliminary study, the applicability of
receptor techniques to the Great Lakes Study was examined. When
problems were encountered during the  preliminary study, their relevance
to the Great Lakes study was discussed, and, where possible,
recommendations for overcoming these problems were made   It should be
noted that, in comparison to this preliminary study, the Great Lakes study
will have a much larger scope, will be much more data-intensive, and will
have a much longer  time frame  The preliminary study is not intended to
reflect the full scope of the Great Lakes study, but is intended to identify
its relevant problems  and potential solutions.

CMB 7 Model
     The CMB7 model is the primary tool used in this project. It is a
computer model which is used to identify and quantify source
contributions to the particulate mass at a given receptor. This is done
using an effective variance weighted least squares solution to a set of
linear equations.2 This set of equations represents the receptor

-------
concentration of each given chemical species as the sum of the products of
source profile species and source contributions  To do this, the model
requires two assumptions  (1) that the total particulate mass at the
receptor is a direct linear sum of the mass contribution from a number of
specific sources, and (2) that the mass and chemical composition of source
emissions is preserved in transport and on the sample filter.2  The
advantages of using this  model in the preliminary study are that it uses
data which are  readily available, and that it costs relatively little to
implement
      The data requirements for this model include the following: (Da
source profile for each source category in the airshed, consisting of the
fractional amount of each chemical species emitted by that source  type,
and (2) the receptor concentrations of each species. Uncertainty estimates
are required for both sets of data. Given these data, the model  generates
estimates for the amount of each species which is contributed by each
source type, and for the total mass contributed by each  source type
      To apply the model, the user starts with the list of available  source
profiles and the available receptor data.  A subset of the receptor chemical
species is chosen to be the  set of fitting species (i.e., the  species used in the
least squares calculation) and a subset of the available source profiles is
chosen to be the set of fitting sources (i.e., the sources used in the least
squares calculation)   The fitting sources chosen are the sources which are
expected to comprise the bulk of the particulate mass in a given sample,
and the fitting species are the species which are expected  to be significant
in the emissions of the fitting sources. The CMB7 model then uses the
fitting species to calculate the source contributions for the fitting sources
included  The other species are then calculated based on the knowlege of

-------
the mass fraction of each species in each source (from the source profile).
Thus, through the choice of fitting species and fitting sources, the user
plays an active role in identifying and modeling the pollutants and sources
of concern

Data Set for Preliminary Study
      For the receptor concentrations, a previously existing data set from
the Detroit-Windsor metropolitan area was chosen.  This data set was
available from a long-term ambient air monitoring program6, during
which 24-hour samples were taken  (starting at midnight) every six days
from July 20,  1967 to September 2,  1990.  The data were taken at two
different sites, one at Windsor, Canada, and the other at Walpole Island,
Canada  These data were taken from a dichotomous sampler, so that
coarse (2.5-10 pm) and fine (0 1 -2.5 |im) diameter particle fractions could
be analyzed separately. Thus, the complete data set consisted of four sets
of 166 samples each: coarse and fine mass fractions from both the Windsor
and Walpole Island monitoring stations.
      The species included in this data set are those species which could be
analyzed by the X-ray fluorescence method (XRF).  These species include
S04, N03, Al, Si, P, S, Cl, K, Ca, Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, Ge, As,
Se, Br, Rb, Sr, Y, Zr, Nb, Mo, Pd, Ag, Cd, In, Sn, Sb, Te, I, Cs, Ba, La, and Pb.
It should be noted that several chemical species cannot be analyzed by
XRF, and are, hence, not included in  this data set.  These include, among
others, Na, Mg, Hg, and organic and elemental carbon.
      For the source profiles, the EPA source composition library data
base5 was used. This data base contains a large number of profiles for
many different source categories, and was easily accessible.

-------
PROCEDURE

Sample Selection
      In order to perform CMB calculations, a small number of samples
was selected from all the existing data. Since this study had no direct
management goal, the criteria for sample selection were based on their
relevance to the objectives of the preliminary study, namely, to create a
situation which resembles the Great Lakes study. Therefore, sample
selection was intended to match the modeling exercise to the Great Lakes
study where possible.  However, this selection was limited at times by data
availability. In general, samples were also selected to allow for flexibility
and ease in  modeling.  Thus, if problems were encountered in modeling for
these selected cases, it is Jikely  that similar, or even more severe, problems
would be encountered  for the 'more difficult' samples.
      The first step of the sample selection process was to identify, from
among all the  reported species, pollutants of concern. Since this project did
not have an explicitly stated management goal, the choice of "pollutants of
concern" was largely arbitrary.  Since the Great Lakes study is concerned
with toxics,  the pollutants of concern for this study were taken from the
list of toxics in title III of the 1990 Clean Air Act Amendments. Given the
short time frame of the project, it was decided to focus on three pollutants
from this list  Lead  (Pb), Chromium (Cr), and Selenium  (Se) were chosen
because of their toxicity and the availability of data for these species.
      In addition to focusing on three pollutants, it was decided to focus
only on the  fine fraction of the Windsor data. The Walpole site is
somewhat removed from the major sources in the Detroit-Windsor
                               7

-------
metropolitan area, and hence does not reflect the participate composition
as accurately as the Windsor data. In addition, the coarse fraction of the
Windsor data did not have very high masses of the selected pollutants of
concern. It was assumed that higher masses would lead to better results,
so the fine fraction was used for this analysis  Thus, this study focused on
the fine fraction of the Windsor data set. Time permitting, a similar study
could be done for the other three  data sets.
     Next, the actual sample days were chosen. Two days of concern were
chosen for each pollutant, in order to examine differences between runs.
First, days were targeted which had high concentrations of the pollutants
of concern selected above.  The samples were ranked according to
concentrations of Pb, Cr, and Se so that high incidences of these pollutants
could be easily identified.
     Another criteria in selecting sample days was wind direction  First,
days were selected with prevailing winds from the south or west This is
because most of the major sources are in this direction from the collector.
In addition, days were chosen during which the wind direction had low
variability. This was done in an effort to create 'stable' model runs, where
a particular group of sources (the upwind group) would be expected to be
more easily identified. Meteorological data  were evaluated for the days
having the highest concentrations of Pb, Cr, and Se, and the days with the
most stable wind speed and direction were chosen. For each pollutant of
concern, two days were chosen which had high mass and low wind
variability, for a total subset of six days. These days are listed in Table  1,
along with wind data.
                               6

-------
               Table 1 Sample Days in Preliminary Study
   Including Concentration, Rank, Wind Direction, and Wind Variability
Species
Cr
Cr
Pb
Pb
Se
Se
Sample
Date
7/20/67
6/19/67
9/12/66
5/27/66
12/23/66
1/10/69
Species
Cone.
(Mg/m3)
0.0107
0.0067
0.1230
0.1 190
0 0074
0 0069
Rank of
Species
cone.
*2 Cr day
*4 Cr day
* 1 Pb day
*2 Pb day
* 1 Se day
*3 Se day
Wind Dir.
(degrees)
243
272
160
219
212
216
Wind Var
(degrees)
49
30
30
15
36
17
Source Profile Selection
     The other piece of data required by the CMB7 model is the source
profiles.  These profiles list, for each source category, the relative
contribution of each chemical species to the total particulate emissions.  For
the purposes of  this preliminary study, these profiles were obtained from
literature values compiled in the EPA's source composition library data
base, using the Volatile Organic Compound (VOC)/Particulate Matter (PM)
Speciation Data System program, Version 1.4 5. This data base contains
profiles obtained through a variety of methods, and includes profiles for
many different source categories.
     Selection of source  profiles from the EPA data base is not a clear-cut
procedure Profiles in the EPA database are grouped according to industry
                               9

-------
categories. However, for some industries, many profiles exist. For
example, among the steel industry profiles in the data base are profiles for
coke oven emissions, fugitive coke dust, basic oxygen furnace, sinter plant,
and several others, including some 'industry average' profiles. On the
other hand, for some industries, profiles do not exist. In neither case is it
clear which profiles are the best ones to use in the model, and profile
selection is often a trial-and-error process.
      It should also be pointed out that there is great variability among
source profiles.  The data base includes profiles developed from a variety --
of different sampling and analytical methods, and taken from a variety of
'representative' industry sites. To account for this, the EPA data base
includes a measure of data quality, with each profile assigned a rating
from A (sound sampling and analytical methodology, good size
fractionation, recommended for use) to E (poor fractionation and  poor or
undocumented methodology, recommended for use only if other data are
unavailable) 5
      In order to decide which source profiles to include in the model runs,
  an emissions inventory and a map of the area were examined. The map,
 which displays the major sources in the Detroit-Windsor area, is shown in
   Figure 1.  From this map and the emissions inventory, major sources
  within a 10km radius of the Windsor receptor were identified, and the
   most closely matching profile(s) were selected. The 10km radius was
  chosen in order to keep the number of source types manageable.  It is
reasonable to assume that sources outside this radius are either redundant
 source types, or do not contribute significantly to the total particulate load
 at the receptor because they are  so far away.  (An exception is sulfate, for
                              10

-------
                               Figure  1   Major Sources - Detroit Area
                 WINDSOR  - DETROIT POINT SOURCES
A41
  «"
       tee*
~*i     "
Wer Paper Co
                                                   176 BASF Corporation C&l
                                                   18< Detrwt Mjnapai
Type o' Source Distance Beamg
Base ton & Ste*!
AutoMfg
SteeiMI
Auto Body A«mb
Auto MTg
Power Ptent
Heatng Rant
Auto HTg
Cement Rant
Blast Furnace
Power Rant
fewer Rant
Chem M*g
MRS Rant
Cement Rant
Power Rant
Power Rant
Heatng Rant
Assembly
Paper Ml
Gtesj Marxrfa
Salt Rant
MTg Rant'
Paper Mfl
Oen-, MTV
hcir^rztor
9 7
5
9 3
9 9
8 7
7 4
2 6
65
8 3
6 1
7 7
9 3
7.4
10
85
4.2
S 3
4 3
6 2
6 2
98
4 9
6
8 2
7 8
7 8
227
288
26C
266
47
231
1
2se
24S
24C
353
ze:
28E
4<
24<
2s;
24(
33!
34.
25
26
24
23
23
29
35
                                                1  1

-------
which a special profile was included, to reflect background sulfate
concentrations arising from long-range transport.)  The legend of Figure 1
contains data for the potential sources identified within 10km, including
their distances and directions from the Windsor monitoring site.6
     For each source identified above, a corresponding source profile or
group of profiles was located  When a profile did not exist for an industry,
a profile from the most closely matching industry was used. When
multiple profiles for  an industry were available, several were chosen to
represent the available variety, with emphasis placed on those with a
better data quality rating, if available.  Examples of two typical profiles are
included in Appendix A

Model Runs
     Once the source profiles and the receptor concentrations were
obtained, the CMB model was run for each of the six sample days. While
the actual model application process depends on the skill and experience of
the user, two EPA documents provide detailed guidance on  CMB
application.  Detailed instructions for running the model are contained in
the CMB7 users manual 3 and an EPA document furnishes guidance on
interpreting the diagnostics included with the  CMB output 4.  These
documents were the main guidance used in this application of the CMB7
model.  They were used in obtaining best estimates for the  relative
contributions of the different source types to the total particulate load.
These documents were also used to estimate the validity and uncertainty
of these estimates.
     It should be noted  that the aforementioned manuals do not provide
step-by-step instructions for model application. Rather, they provide
                              12

-------
guidance which the modeler must interpret and tailor to the case at hand.
For this reason, it is not possible to outline a specific methodology to use
for each case  However the general approach can be described. For each
case, the model was run with a large  number of fitting sources and fitting
species First, the source categories were chosen for the initial run.  The
sources included in the initial run represent ubiquitous sources (auto
exhaust, paved road dust, sulfate) as  well as likely upwind point sources in
the Detroit-Windsor area. The fitting species were then chosen for the run,
based on the species which are known to contribute to the source
categories used
     Starting from the initial run, the results were then 'fine tuned' using
the criteria outlined in the guidance documents.  Table 2 defines the key
terms and variables included in the CMB output. Table 2 also includes the
recommended ranges for-the diagnostic variables. The guidance document
also recommends strategies for correcting these variables when they fall
out of range. When diagnostic variables were out of range, corrective
strategies were applied so as to bring as many of these variables as
possible to within the recommended  target ranges. This usually involved
eliminating uncertain or unimportant profiles and fitting species.
It should be noted that, in cases with large data gaps or other flaws, it may
not be  possible to correct all diagnostic variables simultaneously.  In some
cases it may even be impossible to correct one or more of them (such as total
mass) at all.  In these cases, a subjective judgment must be made about which
run is truly 'best', and sometimes a reasonable result may be possible to
obtain.  Different modelers may arrive at different results.  This situation is
clearly not ideal, and represents a case where some additional refinements
                              13

-------
                  Table 2:  Terms Used in CMB7 Outputs
              Including Target Values for Diagnostic Variables

Term                      Description

Source Type           The source category, abbreviated as follows:

ESPCOAL              ESP controlled coal-fired power plant.
INCIN                 Municipal Waste incinerator.
ESPKRAFT             ESP controlled Kraft recovery furnace (paper mill).
STLFND               Steel Foundry.
SALTMINE            Mineral processing plant
RESOIL                Residual oil combustion
LTVEHIC              Leaded light-duty vehicles.
COKEDUST             Fugitive coke dust.
SULFATE              Background sulfate.

(Other sources were used, but eliminated during the model runs.)

Source Contribution      Estimate (in ug/m3) of source contribution to
Estimate (SCE)           fitted ambient data.

Standard Error (STDERR)  Uncertainty in the SCE.  (Target- STDERR « SCE).

t-statistic (T-STAT)      Ratio of SCE to STDERR.  (Target: t > 2.0).

R-square (R-SQUARE)     Variance in ambient species concentration which is
                        explained by calculated species concentrations.
                        (Target : 0.6 to  1.0).

Chi-square (CHI-SQUARE) Like R-SQUARE, but takes into account uncertainty
                        in calc. species concentrations. (Target: 0 to 4.0).

Percent Mass            Percent of total ambient mass explained by the
                        sum of the SCE's. (Target: 100% +/- 208).

Degrees of Freedom (DF)  Number of species in fit minus number of sources
                        in fit. (Target: DF > 5).

Uncertainty/Similarity    Shows source types which model cannot easily
Clusters                 distinguish between. (Target: none).

-------
Fitting Species (I)
    Table 2 Continued

A * in this column means that the species was
used in the fit
Measured Species Cone.    Measured species concentration from ambient
(MEAS)                  data. A value of -99.0 indicates a value below
                         the detection limit.
Calculated Species Cone.   Calculated species concentration, with
(CALC)                   uncertainty.
Ratio of CALC to MEAS
(RATIO C/M)
Simple ratio of CALC to MEAS, and its uncertainty.
Identifies species over- or under- accounted for
by the model. (Target: 0.5 to 2.0)
Ratio of-residual to        Difference between CALC-MEAS, and its uncertainty.
uncertainty (RATIO R/U)  Like RATIO C/M, identifies species over- or under-
                       •  accounted for by the model (Target: R/U <|2.0|).

Source Pace and Watson, 1967 4

-------
in the data (source profiles and/or receptor concentrations) are needed.
     For each of the six cases (three pollutants, two days each) the model
was run and a fit was obtained. The final fit for each case represents a fit
which met as many of the guidance criteria as possible, and is referred to
as a 'best fit' (even though this judgment has a subjective component.) The
results for the six cases varied considerably, and are discussed in the
following section.

RESULTS AND DISCUSSION

     This section describes results from each of the six cases, and also
reports some general observations across cases. The discussion section
that follows explains some of the reasons for  the inadequacy of the
analysis, and comments on the relevance of these problems to the Great
Lakes study.  Where possible, suggestions for improvement are made.

Case 1: Chromium
Case la:  7/20/S7, High Chromium Day

     The  results from the best run are given in Table 3-  This run had a
high R-square (0.99) and a low Chi-square (0.56)  with six degrees of
freedom.  In addition, the six sources included in  the final run had fairly
low standard errors and correspondingly high t-statistics (>2.0). These
values meet the criteria for an acceptable run. However, the result only
explained 54.3 percent of the mass, which is unacceptably low.
                             16

-------
                        Table 3 CMB output for Case la
SOURCE  CONTRIBUTION ESTIMATES -  SITE: WIN1
SAMPLE  DURATION        24       START HOUR
        R  SQUARE       .99     PERCENT MASS
     CHI  SQUARE       .56               DE
           DATE:  07/20/87
        00         SIZE:
      54. 3
         6
CMB7 3388
SOUR
it
11201
17105
23103
90011
90013
9999S
CE
TYPE
ESPCOAL
INCIN
ESPKRAFT
STLFND
SALTMINE
SULFATE
SCE(UG/M3)
5.4071
.7039
14.5379
.4819
2.0616
2.3079
STD ERR
.6818
.0949
2.9904
.1210
.5021
.6908
TSTAT
7.9306
7.4179
4 .8616
3.9841
4.1061
3.3408
MEASURED CONCENTRATION  FOR  SIZE:  F
      47 .0+-      4.7
       UNCERTAINTY/SIMILARITY  CLUSTERS
CMB7 33889
                                                          SUM OF CLUSTER SOURCES
SPECIES CONCENTRATIONS  -   SITE: WIN1
SAMPLE DURATION         24      START  HOUR
       R SQUARE        .99    PERCENT  MASS
     CHI SQUARE        .56               DF
   DATE: 07/20/87    CMB7  33889
       00        SIZE:     F
     54.3
        6
1 MASS
203
204
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
45
37
38
39
40
SO4
NC3
AL
SI
P
s
CL
K
CA
sc
TI
V
CR
MN
FE
CO
NI
CU
ZN
GA
GE
AS
SE
BR
RB
SR
V
Zr
	 j. 	 nLAS 	
T 47.00000+- 4
14.60000+- 1
.30000+-
.94100+-
.84600+-
.08160+-
* 4.71000+-
.02390+-
* .45400+-
* .30200+-
-99. 00000+ — 99
* .06410+-
.01100+-
* .01070+-
* .02830+-
* .55700+-
.00490+-
.00270+-
* .00490+-
* .11100+-
-99.00000+ — 99
-99.00000+ — 99
. 00090+-
.00320+-
.01410+-
. 00400+-
.00340+-
-99. 00000+ — 99
. 00090+-
.70000
.50000
.05000
.14110
.12690
.01220
.47100
.00360
.04540
.03020
.00000
.00960
.00160
.00160
.00420
.05570
.00020
.00040
.00070
.01670
.00000
.00000
.00010
.00050
.00210
.00010
.00010
.00000
.00010
	 CALC 	
25.50024+-
10.10071+-
.02160+-
.96739+-
1.52504+-
.05296+-
4.71^30+-
.5Zf72+-
.44700+-
.21704+-
.C0300+-
.C^663+-
.00479+-
.01339+-
.02633+-
.59672+-
.00361+-
.01427+-
.00494+-
.09237+-
.00172+-
.00000+-
.00566+-
.00240+-
.01251+-
.00234+-
.00185+-
.00000+-
.00542+-
	 RATIO C/M 	 RATIO R/U
2.55581 .54+- .08 -4 n
1.31920
.02840
.07617
.09472
.02549
.26999
.06072
.03698
.04006
.00158
.00288
.00088
.00509
.00357
.03992
.00253
.00759
.00056
.00458
.00039
.00158
.00083
.00057
.00181
.00067
.00128
.00069
.00164
.69 + -
.07+-
1.03+-
1.80+-
. 65+-
1.00+-
21.95+-
. 98+-
.98+-
.00+-
.88+-
. 44+-
1.25+-
. 93+-
1.07+-
.74+-
5.29+-
1.01+-
. 83+-
.00+-
.00+-
6.29+-
.75+-
.89 + -
.59 + -
. 5 *• J •-
.0 • -:- -
6.r-^-
. 11
. 10
. 17
. 29
. 33
. 12
4.17
. 13
. 17
. 00
. 14
. 10
. 51
. 19
. 13
. 52
2 . 92
. 18
. 13
. 00
. 00
1. 16
. 21
. 18
. 17
. 38
. 00
1 . 94
-2 . 3
-4 . 8
. 2
4 . 3
-1.0
. 0
8 .2
— i
. ^
- , i
1 . 0
- . 7
-3 4
J • *t
. 5
— A
• *t
f.
• V
— R
• — '
1 5
X • J
. 0
— 1 1
A • J.
1 0
A » V
1 0
X • W
5. 7
-1 1
•*. . X
- . 6
-2 . 4
-1 2
A • £.
1 . 0
2.7

-------
41
<1
46
47
48
49
50
51
52
53
55
56
57
82
tfB
HO
PD
AG
CD
IN
SN
SB
TE
I
CS
BA
LA
PB
-99.
-99.
-99.
-99.
-99.
-99.
*
*
.
-99.
-99.
•
.
*
00000+ — 99
00000+ — 99
00000+ — 99
00000+ — 99
00000+ — 99
00000+ — 99
00570+-
00720+-
01010+-
00000+--99
00000+ — 99
01580+-
01840+-
06160+-
.00000
.00000
.00000
.00000
.00000
.00000
.00090
.00110
.00030
.00000
.00000
.00040
.00050
.00920
.00000+-
.00142+-
.00218+-
.00381+-
.00569+-
.00189+-
.00831+-
.00920+-
.00000+-
.00000+-
.00000+-
.02711+-
.01657+-
.06849+-
.00158
.00158
.00283
.00310
.00295
.00441
.00397
.00969
.00158
.00158
.00158
.02158
.01964
.00406
.00+-
.00 + -
.00 + -
.00+-
.00 + -
.00 + -
1. 46+-
1.28+-
.00 + -
.00+-
.00 + -
1.72+-
.90+-
1. 11+-
.00
.00
.00
.00
.00
.00
.73
1.36
.16
.00
.00
1.37
1.07
.18
1
1
1
1
1
1


-6
1
1

-

-C
.C
.C
.£.
.C
.C
.€
. 2
i
.C
.C
C
. 1
. /

-------
     This run indicated that, of the explained mass, the paper mill profile
(Kraft recovery furnace) accounted for the most particulate matter,
responsible for $1% of the particle mass. The next most important profile
was the coal-fired power plant, accounting for 1 1.5%.  The salt mine and
sulfate profiles each accounted for about 5^ of the mass. The steel
foundry and incinerator profiles accounted for less than one percent.
     Although most of the mass came from the paper mill and the power
plant, apportionment results indicate that most of the chromium came
from the steel foundry (60%) and the coal-fired power plant (30%).
Smaller fractions came from  the salt plant (24%), the paper mill (11%), and
the incinerator (1%)  The foundry had to be included in the run because
of the high fraction of chromium in its emissions, even though they
comprise only a small fraction of the total mass. Note that these values
add up to over  100%, indicating  that more chromium was predicted than
was measured *

Case Ib: d/19/67, High Chromium Day

     The results for this run are given in Table 4  This  run was similar to
Case la  It had high R-square (0.96) and low Chi-square (2 30) with five
degrees of freedom   The seven sources included had low standard errors
and high t-statistics (t >2.0)  Again, however, the percent mass was
unacceptably low (36.5)
     The source apportionment was also similar to Case la.  The paper
mill profile comprised the highest fraction of the mass (11.6%), followed by
the coal-fired power plant (7.1%), and sulfate (6.0%). The model reported
                             19

-------
                         Table  4: CMS output for Case  Ib
SOURCE CONTRIBUTION ESTIMATES -  SITE: KIN1
SAMPLE DURATION        24      START HOUR
       R  SQUARE       .98    PERCENT MASS
     CHI  SQUARE      2.30              DF
          D?TE:  08/19/87
       00        SIZE:
     37.5
        5
             CMB7 3388
SOURCE
* TYPE SCE(UG/M3)
11201
17105
21203
23103
90011
90013
9999S
ESPCOAL
INCIN
COKEDUST
ESPKRAFT
STLFND
SALTMINE
SULFATE
1.7861
.8863
.9150
3.0884
.5782
.6171
1.4972
STD ERR
.2721
.1100
.2813
1.0244
.1058
.2083
.2591
TSTAT
6.
8.
3.
3.
5.
2.
5.
5636
0585
2533
0149
4663
9628
7792
MEASURED  CONCENTRATION FOR SIZE: F
       25.0-t—     2.5

        UNCERTAINTY/SIMILARITY CLUSTERS
CMB7 33889
SUM OF CLUSTER SOURCES
SPECIES CONCENTRATIONS - SITE: WIN1 DATE: 08/19/87
SAMPLE DURATION 24 START HOUR 00
R SQUARE .98 PERCENT MASS 37.5
CHI SQUARE 2.30
or £.v_ X r*o
1 MASS
203
204
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
45
37
38
39
S04
N03
AL
SI
P
S
CL
K
CA
SC
TI
V
CR
MN
FE
CO
NI
CU
ZN
GA
GE
AS
SE
BR
RB
SR
Y
T 25.00000+- 2
5.70000+-
.80000+-
.44800+-
.14400+-
* .04490+-
* 2.06000+-
.05710+-
* .16200+-
* .13400+-
-99.00000+ — 99
* .01880+-
.01530+-
* .00870+-
* .02620+-
* .30600+-
.00330+-
.00320+-
* .00800+-
* .14100+-
-99.00000+ — 99
.00110+-
-99.00000+--99
.00620+-
.01900+-
.00180+-
.00220+-
-99 . 00000+--99
.50000
.60000
.10000
.06720
.02160
.00670
.20600
.00860
.01620
.01340
.00000
.00280
.00230
.00130
.00390
.03060
.00020
.00050
.00120
.02110
.00000
.00010
.00000
.00090
.00280
.00010
.00010
.00000
DF
9.36838-r-
3.25999-r-
.00638+-
.35377+-
.58867+-
.02054+-
2.06000+-
.35017+-
.16693+-
.13139+-
.00000+-
.02116+-
.00187+-
.01000+-
.02683+-
.29528+-
.00108+-
.01011+-
.00795+-
.11167+-
.00065+-
.00000+-
.00217+-
.00076+-
.00708-1—
.00067+-
.00067+-
.00000+-
CMB7
SIZE:
33889
F

5
T^ » rn T ^\ f*> f \j •n>mT-/-\ Y-I /TI
.87776 .37+- .05 -5.9
.28285
.00871
.02014
.03072
.00572
.06060
.01663
.00917
.01228
.00042
.00099
.00091
.00574
.00372
.01600
.00072
.00165
.00067
.00474
.00023
.00042
.00045
.00024
.00048
.00023
.00036
.00029
.57+-
.01+-
.79 + -
4.09+-
.46+-
1.00+-
6.13+-
1.03+-
.98+-
.00+-
1.13+-
.12 + -
1.15+-
1.02+-
.96+-
.33+-
3.16+-
.99+-
.79+-
.00+-
.00+-
.00+-
.12 + -
.37 + -
.37 + -
.30 + -
. 00 + -
.08
.01
.13
.65
.14
.10
.97
.12
.13
.00
.18
.06
.68
.21
.11
.22
.71
.17
.12
.00
.38
.00
.04
.06
.13
.16
.00
-3.7
-7.9
-1.3
11.8
-2.8
.0
15.7
.3
-.1
1.0
.8
-5.4
.2
.1
-.3
-3.0
4.0
-.0
-1.4
1.0
-2.6
1.0
-5.8
-4.2
-4.4
-4.1
1.0

-------
40
41
42
46
47
48
49
50
51
52
53
55
56
57
82
ZR
NB
MO
PD
AG
CD
IN
SN
SB
TE
I
CS
BA
LA
PB
-99
-99

-99
-99
-99
-99
-99
*


-99
-99

*
.00000+ — 99.
.00000+ — 99.
.00200+-
.00000+ — 99.
.00000+ — 99.
.00000+ — 99.
.00000+ — 99.
.00000+ — 99.
.00660+-
.00760+-
.00720+-
. 00000+ — 99.
.00000+--99.
.02020+-
.06840+-
00000
00000
00010
00000
00000
00000
00000
00000
00100
00020
00020
00000
00000
00050
01030
.00164+-
.00000+-
.00057+-
.00046+-
.00095+-
.00404+-
.00040+-
.00787+-
.00767+-
.00000+-
.00000+-
.00000+-
.01334+-
.00352+-
.07923+-
.00041
.00042
.00041
.00065
.00094
.00113
.00097
.00155
.00293
.00042
.00042
.00042
.00782
.00418
.00341
.00 + -
.00+-
.28 + -
.00+-
.00+-
.00+-
.00+-
.00 + -
1.16+-
.00 + -
.00 + -
.00 + -
.00 + -
.17+-
1. 16+-
.00
.00
.20
.00
.00
.00
.00
.00
.48
,05
.06
.00
.00
.21
.18
1
1
-3
1
1
1
1
1

-16
-15
1
1
-4
1
.0
.0
.4
.0
.0
.0
.0
.0
.3
.5
.6
.0
.0
.0
.0
2 1

-------
that the incinerator, coke dust, steel foundry, and salt mine profiles
explained less than  5% of the mass.
     The source contributions to chromium are similar to those in case  la.
Again the steel foundry dominated (65%), followed by the coal-fired power
plant (12*), the salt plant (9%), the paper mill (3*), coke dust (2*), and the
incinerator (2%)  Again, more chromium was predicted by the model than
was observed.

Case 2:  Lead

Case 2a-5/2 7/66, High Lead Day

     The results for this run are given in Table 5. The R-square  (.92) and
Chi-square (3-06) were not as good as for Case  1, but still fell within the
target range specified by the guidance document, as did the degrees of
freedom (7).  All the sources included had low standard errors and high
t-statistics (t >2.0), except for light vehicles. However, the light vehicles
profile was left in the final run because it represents an important source
of lead, which should not be omitted. Unlike the analyses for Chromium,
the run for this case explained close to 100% of the mass (actually 107.6
<£}
<* I.
     According to the source apportionment, the overwhelming majority
of the calculated particulate mass is from the paper mill profile (66.6%).
The residual oil combustion profile was the next most important (32.0%).
The incinerator profile made up 3.4% of the mass. The steel foundry
contributed less than one percent of the mass, as did the light vehicle
profile (despite its high relative contribution to lead.)
                             22

-------
                         Table 5. CMB output for Case 2a
SOURCE CONTRIBUTION ESTIMATES  -   SITE:  WIN1
SAMPLE DURATION        24      START HOUR
       R  SQUARE       .92    PERCENT MASS
     CHI  SQUARE      3.06               DF
          DATE:  05/27/88
       00        SIZE:
    107.8
        7
             CMB7 3388
SOURCE
* TYPE SCE(UG/M3)
13501
17105
23103
31103
90011
RESOIL
INCIN
ESPKRAFT
LTVEHIC
STLFND
9
1
19


.2730
.0122
.8841
.1838
.9077
STD ERR
2.1761
.2616
2.8831
.1030
.2117
TSTAT
4.2613
3.8690
6.8968
1.7837
4.2883
MEASURED  CONCENTRATION FOR SIZE:  F
      29.0+-      2.9

       UNCERTAINTY/SIMILARITY  CLUSTERS
CMB7 33889
SUM OF CLUSTER SOURCES

"SPECI
SAMPL
SPECI
1
203
204
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
45
37
38
39
17105
17105
31103
31103

ES CONCENTRATIONS - SITE: WIN1
E DURATION 24 START
R SQUARE .92 PERCENT
CHI SQUARE 3.06
rc_ _ T 	 vrrxc 	
.CO
MASS
SO4
N03
AL
SI
P
S
CL
K
CA
SC
TI
V
CR
MN
FE
CO
NI
CU
ZN
GA
GE
AS
SE
BR
RB
SR
Y

T 29.00000*- 2
8.70000*-
.20000+-
.04150+-
.16200+-
* .02660+-
* 3.44000*-
.03150*-
* .60600+-
* .31300+-
.00340+-
.00800+-
-99.00000+ — 99
.00360+-
* .04340+-
* .58300+-
.00390+-
.00230+-
* .00540+-
* .25100+-
.00300+-
-99.00000+--99
* .00220+-
.00420+-
* .03670+-
.00290+-
.00070*-
-99 . 00000* — 99

.90000
.90000
.05000
.00620
.02430
.00400
.34400
.00470
.06060
.03130
.00010
.00120
.00000
.00050
.00650
.05830
.00020
.00040
.00080
.03770
.00010
.00000
.00030
.00060
.00550
.00100
.00010
.00000

DATE
HOUR
MASS 107
DF

31.26084+-
14 .76535+-
.08285+-
.15562+-
.17130+-
.02589+-
4.13277+-
.70004+-
.51349+-
.19210+-
.00000+-
.01263+-
.31967+-
.01813+-
.04773+-
.49294+-
.00000+-
.51460+-
.01020+-
.16226+-
.00042+-
.00000+-
.00061+-
.00063+-
.02963+-
.00256+-
.00128+-
-00000+-

: 05/27/8
00
.8
7

2.81794
2.10246
.05374
.08837
.05592
.03383
.42148
.07639
.04853
.06233
.00220
.00372
.06956
.00940
.00649
.07290
.00220
.11268
.00234
.01790
.00096
.00220
.00172
.00111
.00496
.00123
.00184
.00102
1. 196*
1.196*
-
.216
.216
8 CMB7 3 3 889
SIZE: F
ATIO C/M 	 RATIO R/U
1.08+-
1.70+-
.41 + -
3.75+- 2
1.06+-
.97+- 1
1.20+-
22.22+- 4
.85+-
.61+-
.00+-
1.58+-
.00+-
5. 04+- 2
1.10+-
.85 + -
.00+-
*****+_62
1.89+-
.65+-
.14+-
.00+-
.28 + -
.15 + -
.81 + -
.88 + -
1.83+- 2
.00 + -
.15
.30
.29
.20
. 38
.28
. 17
. 11
.12
.21
.65
.52
.00
.70
.22
. 15
.56
. 56
.52
.12
.32
.00
.78
.27
. 18
.52
. 65
. 00
.6
2.7
-1.6
1.3
.2
-.0
1.3
8.7
-1.2
-1.7
-1.5
1.2
1.0
1.5
.5
-1.0
-1.8
4.5
1.9
-2.1
-2.7
1.0
-.9
-2.8
-1.0
-. 2
.3
1 . 0

-------
t ^
41
42
46
47
48
49
50
51
52
53
55
56
57
82
ZR
NB
MO
PD
AG
CD
IN
SN
SB
TE
I
CS
BA
LA
PB
-99
-99
-99
-99
-99

-99
*
-99
-99
-99
-99
-99
-99
*
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00320+-
.00000+ — 99
.00400+-
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00000+ — 99
.00000+ — 99
.11900+-
.00000
.00000
.00000
. 00000
.00000
.00010
.00000
.00060
.00000
.00000
.00000
.00000
.00000
.00000
.01790
.00179+-
.00000+-
.00196+-
.00298+-
.00358+-
.00569+-
.00258+-
.01177+-
.00646+-
.00000+-
.00000+-
.00000+-
.02199+-
.02267+-
.13900+-
.00238
.00220
.00220
.00389
.00389
.00313
.00604
.00411
.01197
.00220
.00220
.00220
.02289
.02686
.01622
.00 + -
.00 + -
.00 + -
.00+-
.00+-
1.78+-
.00+-
2.94+-
.00+-
.00+-
.00+-
.00+-
.00 + -
.00+-
1.17+-
.00
.00
.00
.00
.00
.98
.00
1.12
.00
.00
.00
.00
.00
.00
.22
1.
1-.
i _
1.
1.
.
1.
1.
X .
1.
ii _
^
1 .
1 .
•

-------
     The source apportionment of lead predicted that, for this case, most
of the lead  (698) came from the municipal waste incinerator and from
light vehicles (33%)  Smaller contributions were from residual oil
combustion (9%), the paper mill (3%), and the steel foundry  (3$). As with
chromium,  these percentages added up to over 100%. Note that  although
the incinerator and vehicle profiles are a very small component of the total
mass, they  make up nearly the entire lead contribution.

Case 2b- 9/12/66, High Lead Day

     The results for this run are given in table 6.  The R-square (0.99) and
Chi-square  (0.46) for this case are better than in case 2a, although there
were fewer degrees of freedom (5).  Several sources yielded low standard
errors and  high t-statistics and were included in the run. Despite this
inclusion, the percent mass predicted here is much lower than in case 2a
(26 2%).
     For this case, the sulfate profile (9.4%) and the paper mill profile
(6.6%) shared the bulk of the explained mass. Coke dust explained 4.3% of
the mass, and the other profiles explained small (less than 3^) amounts.
     As in  case 2a, the lead apportionment was dominated by the
incinerator  (42%) and vehicle (33^)  profiles.  Smaller portions were
explained by coke dust (2%), steel foundry (1%), paper mill (< 1%), and coal
fired power plant (< 1%) profiles. For this case, lead was not overpredicted,
as the calculated lead value was only 60% of the measured value.

-------
                         Table 6: CMB output for Case 2b
SOURCE CONTRIBUTION  ESTIMATES  -   SITE:  WIN1
SAMPLE DURATION        24      START HOUR
       R  SQUARE       .99    PERCENT MASS
     CHI  SQUARE       .48               DF
          DATE:  09/12/88
       OC        SIZE:
     28.2
        5
             CMB7 3388
SOUR
*
11201
17105
21203
23103
31103
90011
9999S
CE
TYPE SC
ESPCOAL
INCIN
COKEDUST
ESPKRAFT
LTVEHIC
STLFND
SULFATE
:E(UG/M3)
.4050
.6418
1.1745
2.3701
.1886
.3006
2.5447
STD ERR
.1261
.0953
.1974
.6931
.0623
.0583
.3123
TSTAT
3.2117
6.7375
5.9486
3.4193
3.0266
5.1576
8.1492
MEASURED  CONCENTRATION  FOR  SIZE:  F
       27.0+-     2.7

        UNCERTAINTY/SIMILARITY  CLUSTERS
CMB7 33889
SUM OF CLUSTER SOURCES
SPECI
SAMPL
SPECI
1
203
204
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
45
37
38
39
ES CONCENTRATIONS - SITE: WIN1 DATE: 09/12/88
,E DURATION 24 START HOUR 00
R SQUARE .99 PERCENT MASS 28.2
CHI SQUARE .48 DF 5

MASS
SO4
N03
AL
SI
P
S
CL
K
CA
SC
TI
V
CR
MN
FE
CO
NI
CU
ZN
GA
GE
AS
SE
BR
RB
SR
Y

T 27.00000+- 2.70000
6.70000+- .70000
.50000+- .10000
.15800+- .02370
.15100+- .02260
.02780+- .00420
* 2.92000+- .29200
-99.00000+ — 99.00000
* .10500+- .01050
* .05480+- .00550
.00510+- .00010
* .00720+- .00110
-99.00000+ — 99.00000
-99 . 00000+ — 99 . 00000
* .01530+- .00230
* .11700+- .01170
-99.00000+ — 99.00000
-99 . 00000+ — 99 . 00000
* .00700+- .00100
* .07930+- .01190
.00140+- .00010
-99.00000+ — 99.00000
* .00080+- .00010
.00380+- .00060
* .01370+- .00210
-99 . 00000+ — 99 .00000
-99.00000+ — 99.00000
-99.00000+ — 99.00000

7.62531+-
3.80621+-
.00322+-
.12960+-
.21549+-
.01031+-
2.92000+-
.22782+-
.10444+-
.05040+-
.00000+-
.00724+-
.00072+-
.00469+-
.01547+-
.11861+-
.00000+-
.00508+-
.00770+-
.08128+-
.00029+-
.00000+-
.00090+-
.00024+-
.01684+-
.00043+-
.00039+-
.00000+-
	 — — — — n^-
.64951
.21484
.00502
.01239
.01342
.00415
.0497S
.01116
.00618
.00477
.00038
.00054
.00069
.00300
.00228
.00857
.00038
.00127
.00084
.00347
.00030
.00038
.00139
.00030
.00401
.00031
.00036
.00031
! CMB7 33889
SIZE: F
iTIO C/M 	 RATIO R/U
.28 + -
.57 + -
.01 + -
.82 + -
1.43+-
.37 + -
1.00+-
.00 + -
.99+-
.92 + -
.00+-
1.01+-
.00 + -
.00+-
1.01+-
1.01+-
.00+-
.00+-
1.10+-
1.02+-
.21+-
.00+-
1.12+-
.06 + -
1.23+-
.00 + -
.00 + -
.00 + -
.04
.07
.01
.15
.23
.16
.10
.00
.12
.13
.07
.17
.00
.00
.21
.13
.00
.00
.20
.16
.22
.00
1.75
.08
.35
.00
.00
.00
-7.0
-4.0
-5.0
-1.1
2.5
-3.0
.0
1.0
-.0
-.6
-13.1
.0
1.0
1.0
.1
.1
1.0
1.0
.5
.2
-3.5
1.0
.1
-5.3
.7
1.0
1.0
1.0

-------
40
41
42
46
47
48
49
50
51
52
53
55
56
57
82
ZR
NB
MO
PD
AG
CD
IN
SN
SB
TE
I
CS
BA
LA
PB
-99
-99
-99
-99
-99
-99
-99
*

-99
-99
-99
-99
-99
*
.00000+ — 99.
. 00000+ — 99 .
.00000+ — 99.
. 00000 + --99 .
.00000+ — 99.
. 00000 + — 99 .
.00000+ — 99.
.00530+-
.00500+-
.00000+ — 99.
.00000+ — 99.
.00000 + — 99.
.00000+ — 99.
.00000+ — 99.
.12300+-
00000
00000
00000
00000
00000
00000
00000
00080
00080
00000
00000
00000
00000
00000
01850
.00052+-
.00000+-
.00042+-
.00036+-
.00052+-
.00284+-
.00031+-
.00572+-
.00572+-
.00000+-
.00000+-
.00000+-
.00983+-
.00270+-
.09758+-
.00040
.00038
.00037
.00054
.00090
.00114
.00077
.00152
.00280
.00038
.00038
.00038
.00769
.00324
.01464
.00 + -
.00 + -
.00 + -
.00 + -
.00+-
.00+-
.00 + -
1.08+-
1.14+-
.00+-
.00 + -
.00+-
.00 + -
.00 + -
.79 + -
.00
. 00
.00
.00
.00
.00
.00
.33
.59
.00
.00
.00
.00
.00
. 17
1
1
1
1
1
1
1


1
1
1
. 1
1
-1
.0
.0
.0
.0
.0
.0
.0
.2
.2
.0
.0
.0
.0
.0
.1
27

-------
Case 3- Selenium

Case 3a  12/2 3/6 d, High Selenium Day

     The results for this run are given in Table 7.  The R-square (0.99)
and Chi-square (2.44) are acceptable for this run. However, the run-had
only 4 degrees of freedom. The sources which are included had  low
standard errors and high t-statistics (t > 5-2).  The percent mass is also
good (97.5*)
     As usual, the mass was dominated by the paper mill profile (66.5*).
The coal-fired power plant (15-9*) and the salt mine (10.4%) profiles were
also significant  The incinerator and steel foundry profiles were also
included, but each made up less than 3% of the mass
     The apportionment for selenium attributed most of it to the salt
plant (12%} and the coal-fired power plant (12%).  Smaller portions were
attributed to the paper mill (4%) and the incinerator (<  1%). Only 26% of
the measured selenium was explained by the above sources  This suggests
that a major source or sources of selenium are missing from the fitting
sources.

Case 3t>:  1/10/69, High Selenium Day

     An adequate model run for this case could not be achieved  The
percent mass was always extremely low, and very few of the results could
be brought in line with the recommended values.  In addition, inclusion or
deletion of source categories and/or species produces large and erratic

-------
                        Table 7- CMB output for Case 3a
SOURCE  CONTRIBUTION ESTIMATES -  SITE:  WIN1
SAMPLE  DURATION         24      START HOUR
        R  SQUARE        .99    PERCENT MASS
     CHI  SQUARE       2.44              DF
     DATE: 12/23/88
  00        SIZE:
97.5
   4
CMB7 3388
SOURCE
* TYPE SCE(UG/M3)
11201
17105
23103
90011
90013
ESPCOAL
INCIN
ESPKRAFT
STLFND
SALTMINE
2.
.
9.
.
1.
3857
3401
9863
3486
5640
STD ERR
.3340
.0231
.9307
.0611
.2973

7
14
10
5
5
TSTAT
. 1426
.7308
.7296
.7015
.2605
MEASURED CONCENTRATION  FOR  SIZE:  F
      15.0+-       .4
       UNCERTAINTY/SIMILARITY CLUSTERS     CMB7  33889     SUM OF CLUSTER SOURCES
SPECIES CONCENTRATIONS - SITE: WIN1 DATE: 12/23/88 CMB7 33889
SAMPLE DURATION 24 START HOUR 00 SIZE: F
R SQUARE .99 PERCENT MASS 97.5
CHI SQUARE 2.44 DF 4
SPECTTC! 	 T 	 MFSC 	 ^.»T^.
1
203
204
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
45
37
38
39
40
41
MASS
SO4
N03
AL
SI
P
s
CL
K
CA
sc
TI
V
CR
MN
FE
CO
NI
CU
ZN
GA
GE
AS
SE
BR
RB
SR
Y
ZR
NB
T 15.00000+-
3 .75000+-
.28000+-
.33200+-
. 11600+-
* .01990+-
* 1.21000+-
.27600+-
* .33900+-
* .19800+-
.00920+-
.00000+-
.00600+-
.00000+-
* .01830+-
* .29800+-
* .00290+-
.00000+-
.00000+-
* .04600+-
.00000+-
.00000+-
.00000+-
.00740+-
.01500+-
.00280+-
. 00090^-
. 00000+-
. 00000-^-
. ooooc*-
.40000
.08000
. 02000
.00220
.00080
.00020
.00130
.00070
.00080
.00060
.00010
.00000
. 00040
. 00000
.00050
.00180
.00020
.00000
.00000
.00050
.00000
.00000
.00000
.00020
.00020
.00010
.00010
.00000
. 00000
. ccooo
	 V_AJ_,C 	
14. 62479+-
5.37687+-
.01528+-
.45823+-
.74234+-
.02743+-
1.61121+-
.33465+-
.28282+-
.20249+-
.00000+-
.02625+-
.00225+-
.00882+-
.01836+-
.30145+-
.00274+-
.00943+-
.00289+-
.04634+-
.00080+-
.00000+-
.00287+-
.00117+-
.00781+-
.00147+-
.00111+-
.00000+-
. 00271+-
. 00000^-
	 RATIO C/M 	 RATIO R/U
.92030 . <57 + - m _ A
.90763
.01999
,04734
.05931
.01720
.18577
.04227
.02561
.02995
.00104
.00137
.00043
.00366
.00254
.02472
.00182
.00522
.00037
.00257
.00021
.00104
.00054
.00035
.00121
.00043
.00085
. 00035
.00111
.00104
1. 43 + -
. 05 + -
1. 38+-
6. 40+-
1. 38+-
1. 33+-
1 . 21 + -
.83+-
1 . 02 + -
.00+-
. 00 + -
.38 + -
. 00 + -
1.00+-
1. 01+-
.94+-
.00+-
.00+-
1.01+-
.00+-
.00+-
. 00+-
• .16+-
.52+-
. 52+-
1 . 24+-
. 00 + -
. 00 + -
.00-r-
24
• £. **
. 07
. 14
51
• -> X
86
• W V
1 C
• JL ~J
1 5
f ± -J
.08
15
• J> -mt
.11
00
• V W
.08
. 00
.14
f\0
• WO
f.'i
• D J
.00
.00
.06
.00
00
• V w
00
• \J w
.05
08
• w O
1 6
• J- U
. 9 5
00
• V \J
no
m \S \S
.00
IP
• c
-Q A
J • *t
? 7
£. • /
in &
A U . D
2-)
* ^
1 A
X . **
-2.2
-8.8
1 Q O
.1 " . £.
-6.4
") A
£• • ^
.0
— T
.1
1.8
7.7
.1
3.9
5 A
• *l
-15.5
_c p
-> • O
-i n
J * W
-1
• J
7 A
£. • *t
.0

-------
42
46
47
48
49
50
51
52
53
55
56
57
82
MO
PD
AG
CD
IN
SN
SB
TE
I
CS
BA
LA
PB
. OOCOO + -
. 00000+-
.00000+-
. 00000+-
. 00430+-
* .00490+-
.01540+-
.01280+-
.00000+-
.00000+-
.00000+-
. 00000+-
. 02440+-
.00000
.00000
.00000
.00000
.00020
.00020
.00030
.00030
.00000
.00000
.00000
.00000
.00040
. 00092+-
.00150+-
.00232+-
.00342+-
.00130+-
.00452+-
.00602+-
.00000+-
.00000+-
.00000+-
.01947+-
. 01138+-
.03537+-
.00104
.00192
.00214
.00207
.00301
.00278
.00679
.00104
.00104
.00104
.01511
.01348
.00249
.00 + -
.00 + -
.00+-
.00 + -
.30+-
.92+-
.39 + -
.00+-
.00+-
.00 + -
.00 + -
.00 + -
1.45+-
.00
.00
.00
.00
.70
.57
.44
.08
.00
.00
.00
.00
. 10
.9
.8
1.1
1.7
-1.0
-.1
-1.4
-11.8
.0
.0
1. 3
.8
4. 3

-------
changes m the results, making a good, stable, 'best fit' impossible. A more
detailed review of this day showed that an exceptionally high percentage
(76%) of the receptor mass came from species whose concentrations were
either below detection limit or not reported at all. This probably
accounted for the difficulty in obtaining a result on this day.
     In addition  to the tables showing the model output for each day,
Table 6 summarizes important results from the five viable runs in a single
table

Discussion of problems.
     The results  above, together with other information obtained during
the process of generating these results, indicated several problems which,
if they carry over to the Great Lakes Study, will limit the usefulness of
chemical mass balance receptor modeling to the study  In general, several
problems were persistent  The problems were more serious in some cases
than others, but, m general, they prevented adequate results from being
obtained The major problems are described here.
     First, the percent mass explained was generally low. Often, the
analysis was unable to account for over half the total mass. This implies
missing source profiles and/or missing species. Certainly, a run which
could not account for over half the mass must be considered inadequate,
and may indicate that much of the toxic mass was unaccountable as well.
     Second, profiles for key sources of toxics were likely missing. Often,
sources which are known to contribute significantly to the concentration of
a toxic species did not show up  as contributing to the total mass. If one is
interested in total mass estimates, this is not a problem. However for a

-------
Table 6:  Summary of Results from Preliminary Study
Sample Date Toxic Meas Species
Species Concentration
7/20/67 Cr 0.0107
0 0107
00107
0.0107
00107
00107
6/19/67 Cr 0.0067
0 0067
0 0067
0 0067
0 0067
00067
0 0067
5/27/66 Pb 0.119
0 1 19
0.119
0.119
0.119
9/12/66 Pb 0.123
0.123
0.123
0.123
0.123
0.123
0.123
12/23/6S Se 0.0074
0.0074
00074
0.0074
0.0074
Source type Source Fraction of
Contribution Toxic Species
ESPCOAL
INCIN
ESPKRAFT
STLFND
SALTMINE
SULFATE
ESPCOAL
INCIN
ESPKRAFT
STLFND
SALTMINE
SULFATE
COKEDUST
RESOIL
INCIN
ESPKRAFT
LTVEH1C
STLFND
ESPCOAL
INCIN
ESPKRAFT
STLFND
LTVEHIC
SULFATE
COKEDUST
ESPCOAL
INCIN
ESPKRAFT
SALTMINE
STLFND
54071
07039
145379
0.4619
2.0616
2.3079
1.7661
06663
30664
0.5762
06171
1 4972
0915
9.273
1.0122
196641
0 1636
0.9077
0405
0.6416
2 3701
o 3006
0.1666
2 5447
1.1745
2.3o57
0.3401
99663
1.564
0.3466
29*
-IX
11X
60X
24X
OX
12X
2X
3X
66X
9X
OX
2X
9X
69X
3X
33X
3X
ox
42X
OX
IX
33X
ox
2X
12X
OX
4X
12X
OX

-------
study like the Great Lakes Study, which focuses on toxics, this is a serious
limitation
      Third, known contributors to the total mass were missing  This
reflects the model's inability, in some cases, to account for not just toxics,
but also for total mass.  If these missing sources also contribute to the toxic
mass, then their contributions will be missing.
      Fourth, the source apportionment was often  unrealistic. Typically,  -
the model's outcome suggested results which were not consistent with
previous source apportionment studies in similar airsheds, or with
emissions inventories for the Detroit area  No strict model validation was
undertaken  during this study, but the results  generated are unlikely to be
valid.  As a result, they probably will not reflect the true  source
apportionment of the total mass or its toxic fraction
      Fifth, the model runs tended to  produce unstable results. The
outputs proved to be extremely sensitive to certain source categories or
species.  Although some sensitivity is expected, results often fluctuated
wildly, and produced erratic diagnostics which made it very difficult to
obtain a 'best run'
      Finally, the degrees of freedom  were often at or below recommended
limits   Since degrees of freedom is defined to be the number of fitting of
species minus the number of fitting sources, this implies that there were
not enough fitting species (i.e., their data were missing).  This causes the
model to 'force a fit' without enough receptor  species data.  This forced fit
is likely to be an inaccurate reflection of the true source apportionment.

-------
     The problems listed above can loosely be grouped into three major
categories  problems associated with low explained mass (1, 4, 5, and 6),
problems associated with missing sources (1, 2, 3, and 4), and problems
associated with inherent limits on the model's ability to resolve toxics from
among the total mass (2).

Causes and Corrective Actions

The probable causes of the problems described above can be grouped into
three major categories, detailed below.  Where possible, corrective actions
are suggested

1  The receptor data were inadequate
     The receptor data severely limited the usefulness of the study for
several reasons.
     First, there were many values which were reported as either missing
or below the  detection limit. In several cases, these values were key
tracer species. For example, to resolve the residual oil combustion source
category, nickel and vanadium are often used as tracer species. Yet, a
large proportion of the days had unreported values for nickel, vanadium,
or both. In addition, a large number of missing species limits the number
of source categories which can be resolved.  The number of degrees of
freedom (that is, the number of species in the fit minus the number of
sources in the fit), according to the guidance documents, should be greater
than five.  However, the analysis was hampered by missing values,
resulting in a low number of species to include, and hence, a low number
of degrees of freedom.

-------
     Second, there were several species which were completely missing
because they could not be analyzed using X-ray fluorescence  Not only is
this important to some tracer species, but the missing species (most
notably organic and elemental carbon) comprise a large fraction of the
total mass. The probable reason for the difficulty encountered in case 3b
is that 76% of the total mass on that day was accounted for by species
other than the ones which were measured. This is also probably the
reason why the calculated percent mass was low in almost every case.
     Third, the receptor data were of poor quality. It was discovered,
during EPA's  verification of the receptor data that there were
discrepancies between measured and reported values. This discrepancy
may be serious enough to significantly alter the numerical results obtained
in the preliminary study.  (On the other hand, it should not change the
qualitative nature  of the problems encountered, though it may serve to
either mitigate or exacerbate them)  This problem was discovered too late
in the study to redo the entire source apportionment.
     If chemical mass balance analyses are to be useful to the Great Lakes
Study,  the receptor data must be greatly improved. To whatever extent
possible, better data (i.e., more precise, with less uncertainty) must be
obtained for trace elements, especially those which are likely to be used as
tracer species In addition, key missing species must be identified and
analyzed, especially organic and elemental carbon. Finally, quality
assurance must be performed to be sure that the analysis yields consistent
and accurate  results.
                             35

-------
2  The source profiles were inadequate for this airshed
     The nature of profiles available from the source profile data base
also limited the effectiveness of this study.  The major difficulty was
choosing source profiles which accurately reflected the sources in the area
of study  For example, in the emissions inventory, some sources are
described vaguely as "steel mill." Given, the number of different source
types that could exist in a "steel mill/' the choice of a profile is not clear-
cut. Clearly, knowlege of specific processes  at each site would help with
this selection.
     On the other hand, even if it is known which processes are operating,
at a specific site the profiles provided in the data base are probably not
accurate representations of site-specific emissions. Many profiles are
based on national or international industry averages, reflecting widely
varying processes, even among the same source category. Extrapolation to
specific sources may thus be inaccurate. This inaccuracy may be
acceptable in a small airshed with a few chemically distinct sources, but
for a large, complex airshed such as the one in this study, this inaccuracy
limits the ability to resolve chemically similar sources.
      In addition, there were no profiles available for some sources in the
airshed. For example, the profile most closely matching the salt plant was
one for "Mineral Products - Average." Similar data gaps exist for several
other sources  in the area, some of which certainly make significant
contributions  to the total measured particulate load. True source
contribution estimates for these sources will differ if the profile used
differs significantly from the source's actual emissions.

-------
     Finally, the available profiles were, in general, of poor quality  For
the most part, the profiles were of data quality C, D, or E (the three lowest
ratings)  In general, these data are of below average quality and thus, may
have poor estimates of percent mass, uncertainty, or both, and may also
have poor mass fraction data  It is also generally the case, that profiles
often do not include all compounds of interest because they may have
been developed from limited chemical analyses or were, at the time of
their development,  only concerned with a fewspecies5.
     For the reasons above, the source apportionment in this preliminary
study is likely to be inaccurate, especially for the purposes of the EPA
Great Lakes Study   Since the emissions of toxic species are often only a
small portion of total emissions, the problems in  choosing profiles, and the
resulting errors in total source apportionment estimates, are likely to
obscure any meaningful results in the toxic fraction
     These problems point to the need for better source profiles if
chemical mass balance analyses are to be useful  to the Great Lakes Study.
At the  very least, the quality of existing profiles  should be improved,
especially with regard to toxic species, and new profiles should be
developed for sources in the area which do not currently have profiles.
Even if these improvements are made, the profile data may not be
representative of the Great Lakes area.  Specific profiles for each particular
source in the Great  Lakes area should be developed if resolution is to be
high enough to distinguish among many sources  of airborne toxics. The
profiles presently m the data base are likely good enough for screening
purposes, but source-specific data are probably needed to get the level of
detail necessary for the Great Lakes study and any subsequent regulatory
actions based on its results

-------
3  Sulfate overwhemlmgly dominated the mass fraction
     The majority of measured participate mass was sulfate, and sulfate-
nch sources dominated the source apportionment results.  Normally, this
would not be a problem, and it fact, would suggest that the receptor model
is correctly apportioning the mass to sulfate-rich sources such as paper
mills and coal-fired power plants. However, the goal of the Great Lakes
Study is to determine source contributions to the particulate air toxics load,
which is often only a small portion of the total particulate load.  This can
lead to problems in resolving the sources of the toxics becasue they are
obscured by the total mass.
CONCLUSIONS
     During the preliminary study conducted in this paper, several
problems were encountered which would have direct bearing on the
ability of Chemical Mass Balance techniques to contribute to the Great
Lakes Study.  Because of these problems, quantitative source
apportionment of air toxics was not possible with the existing data, even
for a screening-level study. Some of these problems could be eliminated
with better data, but some problems stem from inherent limits to the
receptor modeling approach
     The existing data must substantially improved in order to obtain
useful results. Better source profile data are needed. These data should be
site-specific, if at all possible. If not, they should at least reflect all the
major sources in the  area, and should incorporate knowledge  about specific
industrial processes in the airshed.

-------
      In addition, better ambient data are needed, especially for potential
tracer species These data should include all key species, including those
which cannot be analyzed by X-ray fluorescence.  These data should be as
precise as possible, and should be subject to rigorous quality assurance.
     However, even if the source profile and ambient data are improved,
source apportionment of toxics may not be feasible using Chemical Mass
Balance techniques.  Clearly, the required data improvements will be very
costly, and the cost to obtain all the needed data at the required
resolutions may be prohibitive, especially in the case of site-specific profile
data  These data are further restricted by the limits on analytical
equipment, which  may prohibit obtaining the necessary precision,
particularly for tracer species. Finally, even with perfect data, some
sources  of toxics are  at the "noise level" of the model,  particularly in a
complicated airshed like the Great Lakes area.
     These conclusions highlight the need  to clarify the role of Chemical
Mass Balance receptor modeling in the Great Lakes study  By itself, this
technique will likely be unable to aid in characterizing the particulate mass
in the Great Lakes area  Even if  it could, the costs would probably  be
prohibitive. However, in conjunction with other receptor modeling
techniques, as well as non-receptor based techniques, Chemical Mass
Balance can provide some useful information.  This preliminary study has
identified some of the problems  and shortcomings of the CMB7 model.
With these in mind, an appropriate role for the model can be found which
will certainly assist with the Great Lakes Study.
                             39

-------
ACKNOWLEDGEMENT
     The author would like to thank David E Gumnup of the EPA Office of
Air Quality Planning and Standards for his sponsorship of this project, and
for his guidance throughout its development.
     The author would also like to thank Robert K. Stevens and Teri L.
Conner of the EPA's Atmospheric Research Laboratory for their technical
assistance
                             40

-------
                            REFERENCES

1    Core, John E  "Receptor Model Technical Series, Volume I: Overview of
    Receptor Model Application to Participate Source Apportionment" U.S.
    Environmental Protection Agency Publication No. EPA-450/4-6 1-0 I6a,
    July  1961

2    Core, John E  "Receptor Model Technical Series, Volume II: Chemical
    Mass Balance." U.S. Environmental Protection Agency Publication No..
    EPA-450/4-61-Ol6b, July  1961.

3    Watson, John G., et. al. "Receptor Model Technical Series, Volume III:
    CMB7  Users Manual " U.S. Environmental Protection Agency
    Publication No    EPA-450/4090-004, January 1990.

4    Pace, T.G  and Watson, J.G.  "Protocol for Applying and Validating the
    CMB    Model " U.S. Environmental Protection Agency Publication No.
    EPA-450/4-67-010, May 1967.

5    Lynch, Susan K. " VOC/PM  Speciation Data System User's Manual,
    Version 1.4 "  U.S. Environmental Protection Agency Publication No.
    EPA-450/4-9 1-027, October  1991.

6    "Receptor Modelling Project of the Windsor Detroit Area: Final Report"
    Environment Canada Report No. MT9 1 1 10, May 1991.
                            41

-------
                              Appendix A

                         Sample Source Profiles
The following pages contain examples of two of the source profiles used in
the model runs  The tables contain data for percent by weight of each
chemical species in the profile, and their associated uncertainties.  The
profiles are broken down into three particle size fractions and total
particulate fraction   Note that only the  0 to 2.5 micron fraction was used
in this study because it focused on the fine mass fraction. (Profiles from
VOC/PM speciation software, Version 1  4).5
                              42

-------
f'P Profile Speciation Report
Profile Name : Coal-cired Pone" Pls-t
Prof i le Number : 11201
Data Dual i ty : B/D
Control Device : ESP
Reference(s) : 152
Data Source : Dilution train samples.
nine.
SCC Assignments: 10100201
Mass Fraction Data - Size Interval (urn)
Mass Fraction
Total Particulate
CAS Number
7440-41-7
7440-42-8
7782-41-4
7440-23-5
7439-95-4
7429-90-5
7440-21-3
7723-14-0
770^-34-9
7782-50-5
7,40-09-7
7440-70-2
7440-22-2
7440-32-6
7UO-62-2
7440-47-3
7,39-96-5
7439-89-6
7440-48-4
7440-02-0
7440-50-8
7440-66-6
7440-55-3
7440-56-4
7440-38-2
7782-49-2
7726-95-6
7440-17-7
7440-24-6
7440-67-7
7440-22-4
7440-43-9
7440-31-5
7440-36-0
7440-46-2
c>ptH
NLKTI
4
5
9
11
12
13
14
15
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
37
38
40
47
48
50
51
55
Matter -
~ i es
Syry
Be
B
F
Na
Mg
Al
Si
P
S
Cl
K
Ca
Sc
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
Ga
Ge
As
Se
Br
Rb
Sr
Zr
Ag
Cd
Sn
Sb
Cs
0-2
%wt
NA
NA
NA
NA
NA
15.679
23.837
0.620
3.305
0.094
1.30,
1.235
<
0.967
0.079
0.058
0.0
-------
Profile Speciation Reoc-t   -   continued  (profile  11201)
Tot si Fart ic
CAS Number
7440-39-3
7440-45-1
7439-97-6
7439-92-1




u.ate
Nurr
56
58
80
82
2C1
202
203
204
Matte'
Syr
Ba
Ce
He
Pb
OC
EC
S04
NC3
0-2
.5
U"
2.5-
°iut unc /o»'t
0.000
<
0.000
0.036
NA
NA
NA
NA
0
0.
0
0
0
0.
0.
0,
.098
.000
.006
.006
.000
.000
.000
.000
0.098
<
<
0.025
NA
NA
NA
NA
10 urr
unc
0.040
0.000
0.003
0.003
0.000
0.000
0.000
0.000
0-1
%wt
0.056
NE
NE
0.030
NE
NE
NE
NE
0 um
unc
0.046
0.000
0.000
0.004
0.000
0.000
0.000
0.000
Total
Measurea
,.t
C.056
NE
NE
O.C30
NE
NE
NE
NE
unc
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
                                       52.267         54.008         54.008

-------
'••'.• , e K ,
Fc'e-encc C
Date Sojr:
SCC Assic"
Mass Frart
. "^ c -, ; . ;r.:irc'£*.cr (Philadelphia)
c
i^pr
• Cat;
Total Part i culate

7440-4'.-
7*40-42-
7782-4 ' -
7440-23-
7*39-95-
7*29-90-
7*40-21-
7723-14-
770* - 34 -
7782-5C-
7440-09-
7*-0-7:-
7*43-22-
7*40-32-
7*40-62-
7440-47-
7439-96-
7439-89-
7*40-46-
7*40-C2-
7*40-50-
7440-66-
7440-55-
7440-56-
7440-35-
7782-49-
7726-95-
7440-17-
7440-24-
7440-67-
7439-98-
7440-43-
7440-31-
7440-36-
7*40-46-
7443-39-
-
7
E
*
c;
4
^
7
C
9
5
7
2
2
^
2
3
c;
6
u
U
8
6
3
4
2
2
6
7
6
7
7
9
5
r
u
tL
3
Ko-
.
5
9

" 2
•7
1-
1 =
16

1 9
2:
2:
22
23
2-
25
2;
27
2£
29
3C
31
32
33
3-
35
37
3£
4C
42
*£
5:
5'
55
5t
S' ( i -,:D: "at i^
: i Lut i 0" sample' .
;01001C'
• - Size Interva
Mass Fractic
Matte-
' i es
S/rr,
Be
B
F
Na
Mg
Al
Si
P
s
Cl
K
Ca
Sc
Ti
\l
Cr
Mn
Fe
Cc
Ni
Cu
Zn
Ga
Ge
As
Se
Br
Rb
Sr
Zr
Mo
Cd
Sn
Sb
Cs
Ba
0-2
5*;
NA
NA
NA
NA
NA
0.819
1.442
0.543
1.9Cc
21.299
6.46-
C.223
<
0.012
o.oc:
D.0'5
0.0'£
C.2^
<
0.032
0.14£
11.5C3
0.016
<
o.oo:
O.CC3
0.563
0.016
O.OC9
0.000
0.036
0.2E-
0.829
0.4-2
<
O.C2:
Representative sample frcxn grcjo c*
1 (UT) : (0-2.5) (0-t)
n : 0.260 0.310
.5 UT
unc
0.000
C.OOO
0.000
o.ooc
0.000
0.117
0.067
C.09E
0.301
0.9S5
0.305
O.C30
0.030
0.033
C.C01
0.001
0.001
0.013
0.000
0.032
0.007
0.525
0 . C 1 1
0.030
0.031
0.011
0.027
0.002
0.002
0.014
0.003
0.016
0.041
0.028
0.000
C.C29
2.5-
%.t
NA
NA
NA
NA
NA
5.827
8.570
0.894
3.487
16.678
4.491
5.870
<
1.097
0.024
0.111
0.091
1.874
<
0.085
0.053
4.012
0.000
<
0.000
0.003
0.375
0.005
0.023
0.000
0.003
0.024
0.004
0.086
<
0.000
10 urn
unc
0.000
O.OOC
0.000
0.000
0.000
0.466
0.581
0.299
0.601
2.218
0.640
0.381
0.000
0.069
0.008
0.009
0.008
0.129
C.OOO
0.010
0.015
0.918
0.022
0.000
0.071
0.020
0.057
0.008
0.008
0.039
0.023
0.049
0.089
0.113
0.000
0.242
0-10 urr,
%wt
NE
NE
NE
NE
NE
2.400
3.693
0.654
2.405
19.840
5.841
2.006
NE
0.355
0.008
0.045
0.041
0.783
NE
0.049
0.118
9.137
0.011
NE
0.000
0.003
0.504
0.013
0.013
0.000
0.026
0.202
0.568
0.330
NE
0.014
unc
0.000
0.000
0.000
c.occ
C.OOO
0.657
0.931
0.238
C.595
3.100
0.903
0.626
0.000
0.116
0.006
0.012
0.011
C.204
O.OCC
0.011
0.019
1.383
0.017
O.OCO
0.053
0.016
0.078
0.006
0.006
0.028
0.016
0.045
0.105
0.091
C.OOO
C.I 67
1C.
(C-10)
0.380
Total
Measured
%wt
NE
NE
NE
NE
NE
2.400
3.693
C.654
2.4C5
19.840
5.841
2.006
NE
0.355
0.008
0.045
0.041
C.783
NE
0.049
0.118
9.137
0.011
NE
0.000
O.OC3
0.504
C.013
0.013
0.000
0.026
0.202
0.568
0.330
NE
C.C1*
unc
C.OOO
C.OOO
0.000
0.000
C.OOO
0.000
C.OOO
0.000
C.OOO
C.OOO
0.000
o.coo
0.000
0.000
0.000
0.000
0.000
C.OOO
0.000
0 . OOC
o.ooc
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000

-------
Profile Speciation Report   -   continued (profile  17105)
"etc Fa-ticulate Matter
Species
CAS Mumbc- Mum
7440-45-1 58
7439-97-6 80
7439-92-1 82
201
202
203
204
Sy^
Ce
Hg
Pb
OC
EC
S04
N03
0-2

%wt
<
0.009
8.116
NA
NA
NA
NA
.5


0
0
0
0
0
0
0
urn

unc
.000
.004
.371
.000
.000
.000
.000
2.5-

%wt
<
0.011
2.428
NA
MA
MA
NA
10 urn

unc
0.000
0.012
0.633
0.000
0.000
0.000
0.000
0-10

%wt
NE
0.010
6.320
ME
NE
NE
NE


0
0
0
0
0
0
0
U"

unc
.000
.009
.961
.000
.000
.000
.000
Total
Measured

%wt
NE
0.010
6.320
NE
NE
NE
ME

unc
0.000
0.000
0.000
0.000
0.000
0.000
0.000
                        55.0n6         56.126         55.389         55.389

-------