PB89-181101
United States Office of Air Quality EPA-450/4-83-014R
Environmental Protection Planning and Standards May 1987
Agency Research Triangle Park NC 27711
SERA Receptor Model
Technical Series,
Volume III (Revised)
CMB User's Manual
(Version 6.0)
-------
I
EPA-450/4-83-014R
RECEPTOR MODEL TECHNICAL SERIES,
VOLUME III (REVISED)
CMB USER'S MANUAL (VERSION 6.0)
By
Kenneth Axetel1
PEI Associates, Inc.
Cincinnati, Ohio 45246
and
Or. John G. Watson
Desert Researcn Institute
Reno, Nevada 89b06
Contract No. 68-02-3890
EPA Project Officer: Thompson G. Pace
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office Of Air And Radiation
Office Of Air Quality Planning And Standards
Research Triangle Park NC 27711
May 1987
-------
This report has been reviewed by the Office Of Air Quality Planning And Standards, U.S. Environmental Protection
Agency, and approved for publication as received from the contractor. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, neither does mention of trade names or commercial
products constitute endorsement or recommendation for use.
EPA-450/4-83-014R
ii
-------
CONTENTS
Figures iv
Tables v
1. Introduction 1
What the model does 1
How the model works 3
Purpose of user's manual 3
Organization of user's manual 4
2. Getting the CMB Model Up and Running 5
Hardware requirements 5
Starting 5
Saving results 7
Trial runs with test data 8
3. Building CMB Input Files 10
File types and formats 10
File generation methods 21
4. Moving Through an Interactive Session 23
Calling up data files 23
Using available commands to set up the CMB runs 24
Selecting initial sources 26
Selecting initial species 27
Selecting test site and day 28
Selecting size fraction 29
Initiating a CMB run 29
Recovery summary information 30
5. Evaluating CMB Model Outputs 32
Source contribution estimates display 32
Uncertainty/similarity display 35
Species concentrations display 37
Additional diagnostics--SSC0NT command 39
References 41
Appendices
A. Theory of chemical mass balance model A-l
B. Diagnosis of error messages B-l
C. Printout of test data set for PACS1 C-l
i i i
-------
FIGURES
Number Page
1 CMB model inputs and outputs 2
2 Example source names file 12
3 Example chemical species codes file 14
4 Recommended species codes based on atomic numbers 15
5 Example fine particulate source profile file 17
6 Example ambient concentration data file 20
7 Source contribution estimates 33
8 Uncertainty/similarity display 36
9 Species concentrations 38
10 Species-source contributions display 40
iv
-------
TABLES
Number Page
1 Format for Fine and Coarse Source Profile Files 16
2 Format for Ambient Concentration Data File 19
3 Commands in CMB Program 25
v
-------
SECTION 1
INTRODUCTION
1.1 WHAT THE MODEL DOES
The Chemical Mass Balance (CMB) Model uses the chemical composition of an
ambient particulate sample to estimate the relative contributions of different
source categories to the measured particulate concentration. The chemical
composition of each source category's emissions (source profile) must also be
known in order to run the model. The information required by and provided by
the model are shown in Figure 1.
The CMB model identifies chemically distinct source categories, not
individual emission sources. For example, one run might indicate that 6.7
3
yg/m are from residual oil combustion, but this contribution cannot be
further resolved into concentrations attributable to Power Plant 2, Industrial
Boiler 3, Hospital Heating Plant 6, etc. Also, sources that have nearly
identical chemical compositions (e.g., unpaved roads and construction activi-
ties) may not be distinguishable in the model.
The model can be run for size fractions of the particulate sample, such
as PM^q, if ambient and source chemical composition data are both available
for one or more size fractions. Ideally, separate runs are made for fine and
coarse size fractions.
The CMB model provides source contribution estimates only for the day of
sampling. Exact sources to be included in the model should be specific to
that date and must be determined by the user from meteorological and emission
inventory data separate from the model's required input. Because a CMB run
is applicable to a single location, a single date, and possibly a single size
fraction, most users perform a series of model runs in order to generate a
comprehensive source-receptor analysis.
1
-------
li iEL
:ene.nt
MARINE
QQAO DUST
JCCO STOVE
RESIDUAL OIL
SEC SO.
SEC HO,
IDENTIFY
LIKELY
SOURCES
Section 3
1
SEC SOA
RESIDUAL OIL
M STOVE
rqao pus:
MARINE "
SOURCE
PROFILES
CEMENT
STEEL
F .0021
V
.0006
1* .0126
Cr
.0210
Mq .0050
Mn
.0870
A1 .0065
Pe
.3200
S1 .0510
N1
.0070
5 .0196
Cu
.0028
CI .0125
Zn
.0120
K .0092
Pb
.0076
Ca .0620
S°4
.0250
T1 .0020
*4
.0000
INPUTS
chemical
COMPOSITION OF
AMBIENT SAMPLE
(ONE DAY)
I
Section 3
3.01*45
* .307
Ha .300
Mg .343
A1 .377
Si 2.769
3 .602
CI .695
< .597
Zi .949
T1 .087
V .078
Zr .216
PINE
*i
Fe
Mi
Cu
in
8r
Pb
oc
EC
SO,
NO
1.229
3.901
.178
.033
.1*7
.007
.116
11.016
2.323
2.950
.240
TOT 50.000
MODEL
Section 4
(to get model
running, see
Section 2)
POSSIBLE
CHANGES TO
IMPROVE
RESULTS
OUTPUTS
SOURCE
CONTRIBUTION
ESTIMATES
(SCE) .
STEEL
9.797
CEMENT
2.026
MARINE
l.ooa
ROAO OUST
10.076
WOOD STOVE
19.619
RESIDUAL OIL
2.013
SEC SO.
1.997
SEC MO*
3.463
DIAGNOSTICS
Section 5
EVALUATION
OF MODEL
RESULTS
Figure 1. CMB model--inputs and outputs.
2
-------
1.2 HOW THE MODEL WORKS
CMB is a statistical model in which a separate mass balance equation is
developed for each chemical species. The equations are of the form:
Ci = Fi1S1 + Fi2 S2 + ' " ' + FiJSJ
where = Particulate concentration of chemical species i
measured in the ambient air, ug/m3
F.. = Fraction of particulate emissions from source j that is
J of chemical i
S. = Ambient particulate concentration resulting from source
J j; also referred to as source contribution estimate
(SCE); this is the unknown in the model
J = Total number of sources.
The number of equations is equal to the number of chemical species; the number
of unknowns (S values) is equal to the number of sources (J). In order to
solve equations in CMB, there must be more chemical species than source
categories.
The simultaneous equations are solved by a procedure known as least
squares estimation. Actual calculations of the least square fit are done in
the computer program by matrix algebra.
1.3 PURPOSE OF USER'S MANUAL
This manual describes Version 6.0 of the Chemical Mass Balance Receptor
Model. It is designed to allow persons to constructively use the CMB model
with only a few hours learning time. Because of this emphasis on rapid
command of modeling procedures, the manual covers primarily the mechanical
aspects of running the model. Users seeking more information on the statis-
tical or theoretical bases of chemical mass balance receptor modeling are
referred to Appendix A, Theory of Chemical Mass Balance Model.
The manual is intended for wide use by State and local air pollution
control agency personnel in developing State Implementation Plans (SIP's) for
PMjq. The U.S. Environmental Protection Agency (EPA) has published two
3
-------
companion documents to this manual that are also SlP-oriented and that should
be consulted for this application. The first, Protocol for Applying and
Validating the CMB Model (USEPA 1987A), provides guidance on applicabi 1 i ty,
assumptions, and interpretation of results. The protocol provides a practical
strategy for obtaining valid results. The second document, Reconciling CMB
and Dispersion Model Results (USEPA 1987B), recommends a procedure for examin-
ing and reconciling differences between dispersion and receptor models' results.
Changes made to Version 6 of CMB subsequent to this user's manual will be
described briefly in a README.DOC file located on the same diskette as the
updated CMB program. You may read this file by using either the TYPE or
PRINT commands in DOS.
1.4 ORGANIZATION OF USER'S MANUAL
Version 6 of the CMB model is interactive. It provides prompts and
explanations in the sequence of displays during a run. However, a new user
will still need to refer to this manual for selecting options and format-
ting responses until he/she is thoroughly familiar with the model's capabili-
ties.
The four sections of the manual following this introduction and their
respective contents are:
Section Title Contents
2
Getting the CMB Model
Up and Running
Basic instructions to get
the computer program operating
3
Building CMB Input Files Formats and instructions for
ambient and source profile
data entry
4
Moving Through an
Interactive Session
Possible responses and con-
sequences for all interactive
queries
5
Evaluating CMB Model
outputs
Descriptions and interpreta-
tions of output statistics
4
-------
SECTION 2
GETTING THE CMB MODEL UP AND RUNNING
2.1 HARDWARE REQUIREMENTS
The CMB program runs on an IBM PC or IBM-compatible personal computer.
A minimum memory capacity of 320 K is needed for the program.
Version 6 of CMB is compiled to take advantage of an 8087 coprocessor.
The coprocessor reduces calculation times by a factor of 10 to 20. If an
8087 coprocessor is not available for the PC on which the program is to be
run * the program will automatically run without the 8087. Run times may be
unacceptably long (on the order of several minutes in some cases) without the
8087 coprocessor.
The program software and test data files are contained on one 5-ls-inch
double density floppy diskette formatted by the MS/DOS or PC/DOS operating
system. The CMB program is named CMB6.EXE on this diskette. The CMB soft-
ware may be run from this diskette, but its limited space for input and out-
put files makes this inconvenient. It is recommended that the program and
data files be transferred to a hard disk directory, if available, for routine
modeling.
2.2 STARTING
2.2.1 Disk Operating System (DOS)
After the PC and screen units are turned on, the computer is "booted"
with a DOS operating system by placing the DOS diskette into drive A. DOS
Version 2.0 or later is required. The standard DOS configuration must be
modified so that the CMB program can open up to 14 files simultaneously. This
is accomplished by modifying or creating a file entitled CONFIG.SYS and placing
5
-------
it on the DOS diskette. If the PC has a hard disk, the CONFIG.SYS file should
be placed in the root directory.
An existing CONFIG.SYS file is modified by adding the line 'FILES=14'
(see instructions below). If none exists, a new CONFIG.SYS is created by
moving to the appropriate drive (A> for DOS on diskette, usually C> for a
hard disk) and using the text editor EDLIN (supplied with the operating system)
to add the following:
A> or C> EDLIN CONFIG.SYS (ENTER)
I (ENTER)
FILES=14 (ENTER)
(CONTROL)(BREAK)
E (ENTER)
The sequence above activates EDLIN, names the new file CONFIG.SYS, enters
the Insert mode (I), increases the number of files in memory to 14, exits the
Insert mode, exits EDLIN, and leaves the new file where it will be part of
the DOS operating system when the computer is subsequently booted.
If the CMB program is to be used immediately, the computer must be re-
started by simultaneously pressing (CONTROL)(ALT)(DEL) so that the new
CONFIG.SYS file can expand the maximum number of files to 14. In subsequent
starts, no special action is necessary. However, always be sure to use the
modified DOS diskette when running CMB.
2.2.2 The CMB and Data File Diskettes
The CMB executable program and test data files are on a single diskette,
but they may be copied onto separate diskettes to free more space for output.
The placement of these diskettes in drives depends on the configuration of
the system being usedhard disk-drive, dual disk-drive, or single disk-drive.
If the PC is equipped with a hard disk, the CMB program should be copied
onto the hard disk (into a specified subdirectory) from the floppy disk drive
with the COPY command:
C:
CD/
MKDIR CMB
CD CMB
COPY A:*.*
6
-------
This command will also transfer the input data files. Be sure that the desired
subdirectory has been accessed before using the copy command.
With a dual drive, CMB and data file diskettes can be placed in separate
drives. If the PC has only a single disk drive, the data files to be used
in a run (six files) must normally reside on the same diskette as the CMB pro-
gram. However, with output files, the capacity of the diskette will soon be
exceeded.
2.3 SAVING RESULTS
2.3.1 Printouts
A permanent record of each CMB run can be obtained in one of two differ-
ent forms:
o A printout of the entire interactive session, including all
routine CMB program outputs
o Printouts of just the information from each run that the user
wants.
The complete printout is obtained by simultaneously pressing (C0NTR0L)(PRT SC)
at the beginning of the run; the printer must be connected to the computer
and its power switch turned on. Pressing (CONTROL)(PRT SC) a second time
switches off this feature. Selected information (second method above) is
printed by pressing (SHIFT)(PRT SC) when desired data from the run is
displayed on the screen.
If several runs are anticipated, the volume of printout generated by the
first method may be so great that analysis becomes unwieldy. Also, the time
required to complete a run is much longer if all input and output are printed.
It then becomes more expedient to use the second method, typing (SHIFT){PRT
SC) whenever useful information is on the screen.
2.3.2 Storage of Outputs
The results of any CMB run can be permanently stored in a disk file by
using the WRITE command at the completion of the run. With multiple runs,
the output files should be appropriately named so they can be distinguished
from one another at a later date.
7
-------
The two disk files generated by the WRITE command--CMB0UT and SUMMRY--are
1n text format. Therefore, these files can be Input to another software pro-
gram (e.g., a word processor or text editor) for editing and enhancement.
They can be read using either the TYPE or PRINT commands in DOS.
2.4 TRIAL RUNS WITH TEST DATA
A test data set has been distributed with the CMB program to assist new
users in gaining familiarity with 1t prior to entering their own data. The
CMB program and test data files should be loaded into the computer as de-
scribed in Section 2.2. The program is activated by addressing the drive
(and directory, if any) with CMB 1n it, typing CMB6, and pressing the ENTER
key UJ). The CAPS LOCK key must be on for the interactive session.
The program begins an interactive format at this point by asking:
DISK FILE FOR INITIAL INPUT?
IF NOT ENTER CARRIAGE RETURN
IF SO ENTER NAME OF DISK FILE
The appropriate response to bring up the test data is 'INPORT.DAT', which
should be followed by pressing the ENTER key again.
Another query, 'DO YOU WISH TO RENAME CMBOUT?', follows. Press ENTER
again. The title screen for CMB appears next, followed by the statement:
THE 'HELP1 COMMAND LISTS COMMANDS.
USE COMMANDS AE, DE, AS, DS FOR CHANGES.
ENTER COMMAND
Enter the command 'SELECT'. This allows the user to specify the sampling
site and date of ambient data for the run. Three days of ambient data are
included in the test set:
Site Year Date
PACS1 77 0813
PACS2 78 0124
PACS3 77 0807
Any of these three dates can be entered for the initial run in response to
the program queries:
8
-------
ENTER DESIRED CMB SITE CODE: XXXXXXXXXXXX
ENTER YEAR: YY
ENTER DATE: MMDD
The user is also asked to select the size fraction for analysis, and can
respond with either 'F' for fine or 'C' for coarse.
The next user command should be 'CMB', which initiates the least square
calculations and presents the results. The program produces three displays of
output statistics and then permits entry of another command to modify input
data or obtain additional output for the completed run.
Eight source categories have been preselected as the fitting sources for
the initial run. They are listed in the first output display, Source Contri-
bution Estimates. The names (and source numbers) of 12 additional sources
with complete input data in the test files can be found by using the command
PINFO. The third output screenSpecies Concentrations--!ists up to 35
chemical species for which complete input data are available and notes the 21
fitting species used in the run with asterisks.
Detailed instructions for the interactive portion of the CMB program
are in Section 4 of this user's manual. These instructions should be followed
closely during practice sessions with the test data.
A printout of the initial run with the test data set for PACS1 is
presented in Appendix C. If any difficulties are encountered following the
instructions in Section 4, Appendix C should be consulted. Several additional
runs for the same site and sampling date should be performed to determine
whether predictions can be improved by adding or deleting some source(s) or
species(s).
9
-------
SECTION 3
BUILDING CMB INPUT FILES
3.1 FILE TYPES AND FORMATS
Six different files are required as input to CMB as listed below:
File
List of Input Data File Names
List of Source Names
List of Chemical Species Codes
Source Profiles, Fine Particle
Source Profiles, Coarse Particle
Ambient Concentration Data
Usual File Name
IN .DAT
SO .DAT
PO .DAT
FS .DAT
CS .DAT
DA- .DAT
Each file except the Ambient Concentration Data file contains a single
line format (i.e., entries on sequential lines all have the same format). The
formats for the six files are presented in separate subsections below. The
notations A, I, and F refer to alphanumeric, integer, and floating-point for-
mats, respectively. A floating-point format designated F8.6, for example,
indicates that the field is 8 characters wide and if no decimal point appears
in the field it is assumed that six characters are to the right of the implied
decimal. Frequently, this decimal point is included in the data in a float-
ing point field; in that case, the actual decimal location overrides the
implied location in the "F" format. Data entries in the I format should
always be right justified (i.e., aligned to the right side of the field).
The six files can be created and stored in any order on the diskette or
hard disk because they are called individually by name into the CMB program.
In general, only one SO, PO, FS, and CS file is required for a CMB analysis
of a site or regional set of sites. However, if different FS or CS files are
desired for separate runs, this can easily be accoumodated by naming the al-
ternate files differently. The two-letter prefixes for file names shown in
10
-------
the table above are recommended to avoid confusion, but are not required for
the program to run. Up to six additional alphanumeric characters can be used
in the file names to distinguish different data sets.
3.1.1 List of Input Data File Names (IN .DAT)
This file contains the names of the other files used by the CMB program
in the exact order that they are called into the program. The order of files
in the listing is as follows:
FS .DAT
CS .DAT
DA .DAT
SO .DAT
PO .DAT
The names must correspond to those of the user's five files, including any
file name extensions.
The line format for this brief file is a 12 or fewer character alpha-
numeric name left oriented in columns 1-12 of lines 1-5.
The purpose of this file is to save the effort of keying in the individual
file names in response to program queries each time CMB is run. If this file
is not present, the program will request the five file names in a series of
interactive queries.
3.1.2 List of Source Names (50 .DAT)
This file has a 2-digit integer source code number in columns 1-2 and
an 8-character alphanumeric source name in columns 5-12. If the source is to
be included as an initial fitting source (see discussion in Section 4.2), an
asterisk should be placed in column 15. All other columns on each line are
usually blank, but columns 16-80 can be used for comments, which will appear
only if the Source Names file is printed or displayed on the screen.
The maximum number of sources (lines) that can be listed in this file is
35. The maximum number of sources in the file that can be included in a
single CMB run is 16. Source names need not be listed in numeric order.
The format for the Source Names file and example input for 15 sources are
shown in Figure 2.
11
-------
ro
12
SOURCE
CODE
AS
SOURCE
NAME
COMMENT AREA
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
0
1
R
E
s
I
D
0
I
L
*
0
2
W
0
0
D
S
T
0
V
0
4
F
I
E
L
D
B
R
N
*
0
7
R
0
A
D
D
U
S
T
0
8
A
3
P
H
A
L
T
P
1
0
T
R
A
F
F
1
C
1
1
M
A
R
I
N
E
*
1
2
E
L
E
V
A
T
0
R
1
5
F
E
R
T
I
L
I
Z
1
8
C
O
A
L
P
W
R
P
*
2
1
S
M
E
L
T
E
R
2
3
C
0
N
S
T
R
U
c
2
4
c
E
M
E
N
T
*
2
6
s
E
C
S
O
4
*
2
9
s
E
C
N
0
3
Figure 2. Example source names file.
-------
3.1.3 List of Chemical Species Codes (PO .DAT)
This file has an identical format to the Source Names file described
abovea 2-digit species code in columns 1-2, an 8-character alphanumeric
species name in columns 5-12, an asterisk in column 15 if the species is to
be an initial fitting species, and all other columns blank. No species should
be given the code 01; this code is reserved for particulate matter concentration
measured at the sampler, and is termed TOTAL. TOTAL could be included in
the species listing (with species code 01), but generally it is not used as
a fitting species because this would tend to force the model to make the
sum of source contribution estimates (SCE's) equal the measured concentration.
The maximum number of chemical species (lines) that can be listed in
this file is 35. The maximum number of species that can be included in a
single CMB run is 21. Chemical species do not have to be listed in numeric
order.
In order to maintain consistency with the species coding in the EPA
source profile library, atomic numbers should be used as the species codes
for chemical elements. This is not required by the program, but strongly
recommended. The example format in Figure 3 provides some of these numbers.
Codes for all species can be found in Figure 4. The three-digit species codes
in the source library must be changed to two-digit codes to conform to the
input format of CMB. Change codes 201-204 to codes 91-94.
3.1.4 Fine Particle Source Profile (FS .DAT)
The line format for this file has four required input fields plus an un-
structured comment area that is frequently used to indicate what data are. re-
corded in those four fields. The columns and formats for the fields are as
shown in Table 1. All source code numbers and species codes used in this file
must be listed in the previous two files, the List of Source Names and List of
Chemical Species.
The number of lines in the Fine Particle Source Profile (FS) file should
be equal to the number of source profiles times the number of species per
profile. For example, if 20 source Drofiles are needed for the CMB analysis
and each has an average of 22 chemical species, the FS file would have 440
13
-------
12
AS
SPECIES
SPECIES
CODE
NAME
COMMENT" AREA
1
2
3
4
s
6
7
8
3
10
11
12
13
14
15
18
17
18
IB
20
21
22
23
24
25
26
27
28
30
9
F
1
1
N
A
1
2
M
G
1
3
A
I
1
4
S
1
1
6
S
1
7
C
L
1
9
K
2
0
C
A
2
2
T
1
«
2
3
V
2
5
M
N
2
6
F
6
2
8
N
1
3
0
Z
N
3
5
B
R
8
2
P
B
9
1
0
C
9
2
E
C
9
3
S
o
4
9
4
N
o
3
2
4
C
p
2
9
C
u
3
4
s
E
9
0
T
C
1
T
0
T
A
l
Figure 3. Example chemical species codes file.
14
-------
Species
Speci es
code
name
4
BE
5
B
9
F
11
NA
12
MG
13
AL
14
SI
15
P
16
S
17
CL
19
K
20
CA
21
SC
22
TI
23
V
24
CR
25
MN
26
FE
27
CO
28
NI
29
CU
30
ZN
31
6A
32
GE
33
AS
34
SE
35
' BR
37
RB
38
SR
40
ZR
47
AG
48
CD
50
SN
51
SB
55
CS
56
BA
58
CE
80
HG
82
PB
90
Total car
91
OC
92
EC
93
S04
94
N03
SUM (%)
TOTAL
Notes: OC = organic carbon, EC = elemental carbon, NA = not analyzed, and
NR = not reported.
Figure 4. Recommended species codes based on atomic numbers.
15
-------
TABLE 1. FORMAT FOR FINE AND COARSE SOURCE PROFILE FILES
Column
Format
' * ¦ ¦ ¦ ' i 1 i S
Contents
1-2
12
Source code number
5-6
12
Species code
9-16
F8.X
Fraction of fine (or coarse) source
emissions from indicated species
19-26
F8.X
Uncertainty of species emission
estimate (standard error of
fraction)
27-80
Can be used for comments; usually
to identify size fraction, source
name, and species name
lines. This file, along with the corresponding Coarse Particle Source Pro-
file (CS) file, constitutes the bulk of the input data.
The source profile data are often obtained from published references,
such as the Receptor Model Source Composition Library (USEPA 1984), and others
(Hopke, 1985 and Gordon, 1987). However, profile data should be selected
carefully. Source profiles from published references may be adequate for
screening runs, but airshed-specific profiles are strongly recommended for
final CMB runs.
If the emissions of a chemical species from a source are zero, that line
should still be included in the source profile file. Also, every species must
have an uncertainty value specified in columns 19-26, and that value cannot
be zero. Estimated values for uncertainties can be conservatively assumed to
be the same as the species fractions in columns 9-16 (unless the species frac-
tion is zero).
All source profiles should include data for a species if it is to be in-
cluded as a fitting species. The CMB analysis might yield biased results if,
for instance, some source profiles measured carbon but carbon data were not
reported (but actually present) for other source profiles.
Some sources emit only one or a few chemical species. Single constituent
source types, such as those used to apportion secondary sulfate and nitrate,
have only one species each and should be entered in the FS file as shown in
Figure 5. A fine source profile for one other source, RESIDOIL, is also
shown in Figure 5.
-------
SOURCE
COOENO.
! SPECES
| COOENO.
FRACTONOF
EMISSIONS FBOM
SPECIES
8TAN0AA0
ERROR OF
FRACTION
COMMENT AREA
1
2
3
4
ft
7
9
10
11
12
13
14
16
18
17
18
19
20
21
22
23
24
28
24
27
28
29
30
31
32
»
34
38
38
37
38
38
40
41
42
43
44
48
48
47
0
1
9
0
0
0
S
3
0
0
0
1
9
F
1
N
E
R
E
8
1
0
O
1
L
F
0
1
1
1
0
3
5
0
0
0
1
7
0
0
F
|
N
E
R
E
S
1
D
o
1
L
N
A
0
1
1
2
0
0
0
0
0
0
a
0
0
0
F
\
N
E
R
E
S
1
0
o
1
L
M
Q
0
1
1
3
0
0
5
a
0
0
0
2
4
0
F
1
N
E
R
E
S
1
D
0
1
L
A
L
0
1
X
4
0
0
9
6
0
0
0
4
8
0
F
1
N
E
R
E
S
1
D
o
1
L
S
1
0
1
\
6
1
3
3
0
0
0
2
4
0
0
F
|
N
E
R
E
S
1
D
0
1
L
S
0
1
1
7
0
0
0
0
0
0
0
1
0
0
F
1
N
E
R
E
s
1
0
0
1
L
c
L
0
1
1
9
0
0
2
8
0
0
0
1
0
0
F
1
N
E
R
E
s
1
0
o
1
L
K
0
1
2
0
0
1
5
a
0
0
0
6
4
0
F
1
N
E
R
E
s
1
D
o
1
L
C
A
0
1
2
2
0
0
1
i
0
0
0
0
a
8
F
1
N
E
R
E
s
1
D
o
1
L
T
I
0
1
2
3
0
3
4
4
0
0
0
7
s
' 0
F
1
N
E
R
E
s
1
0
o
1
L
V
'
0
\
2
5
0
0
0
4
6
0
0
0
i
3
F
1
N
E
R
E
s
1
D
0
1
L
M
N
0
1
2
6
0
2
9
7
0
0
0
6
1
0
F
I
N
E
R
E
s
1
D
0
1
L
F
E
0
1
2
0
5
3
6
0
0
1
2
i
0
F
1
N
E
R
E
s
1
D
0
1
L
N
1
0
1
3
0
0
0
4
0
0
0
0
1
8
0
F
1
N
E
R
E
s
1
0
o
1
L
Z
N
0
1
3
S
0
0
0
1
3
0
0
0
2
1
F
1
N
E
R
E
s
1
D
o
1
L
B
R
0
1
e
2
0
0
1
1
0
0
0
0
6
4
F
1
N
E
R
E
s
1
D
o
1
L
P
B
0
1
9
1
0
7
a
0
0
0
2
0
0
F
\
N
E
R
E
s
1
D
o
1
L
O
C
0
1
9
2
0
2
4
2
0
0
2
S
0
0
F
1
N
E
R
E
s
1
D
o
1
L
E
C
0
1
9
3
4
9
i
0
0
1
1
9
0
0
F
i
N
E
R
E
s
1
D
o
1
L
S
0
4
0
1
9
4
0
0
6
S
0
0
0
4
4
0
F
1
N
E
R
E
s
1
D
o
1
L
N
o
3
0
1
2
4
0
0
0
4
7
0
0
0
1
s
F
1
H
E
R
E
s
1
0
0
1
L
C
R
2
6
9
3
7
2
7
0
0
0
5
0
0
0
F
1
H
E
S
E
c
S
o
4
S
0
4
2
4
9
4
7
7
S
0
0
0
5
0
0
0
F
1
N
E
S
E
c
N
0
3
N
O
3
Figure 5. Example.fine particulate source profile file.
-------
3.1.5 Coarse Particle Source Profile
The format and data requirements for this file are the same as those for
the Fine Particle Source Profile file (Table 1 and Figure 5).
If only PM^q size range of source profile data are available, they should
be entered as coarse and the fine file left blank. However, a named file
(containing no data) should still be created for the Fine Particle Source
Profile file to prevent a file opening error when CMB is run.
Source profiles for the fine size fraction are for particulate 0-2.5 urn
and for the coarse size fraction are for particulate 2.5-10 um. If a single
source apportionment for PM^q is desired from the CMB model, the fine and
coarse source profiles can be combined into a PM^q source profile. The pro-
cedure for calculating the PM^q species fractions and uncertainties from
coarse and fine profile data is described on pages 11-33 to 11-36 of the
Receptor Model Source Composition Library publication (USEPA 1984).
3.1.6 Ambient Concentration Data (DA .DAT)
This file contains two line formats for each day of sampling data. The
header format (identified by the numbers 03 in columns 1 and 2) provides in-
formation on the receptor site and date and is limited to a single line. The
second type of format (identified by the numbers 30 in columns 1 and 2) re-
cords ambient concentrations for the above site and date, with a separate
line for each chemical species. The detailed formats are shown in Table 2.
An example Ambient Concentration Data file for one sampling date is presented
in Figure 6. Note that the uncertainty of air ambient species measurement
cannot be equal to zero. It should be set equal to the minimum analytical
detection limit for this species.
The numbers of sites/days of ambient data in this file are not limited.
The CMB program uses each day's data in a separate run when the unique com-
bination of receptor site identification and date are specified in the inter-
active portion of the program. Thus, several dates for both size fractions
can be included behind each other in thf same file.
During data entry, the same species codes as in the List of Chemical
Species Codes (P0) file must be used. Also, concentration uncertainties
(standard errors in columns 48-56 and 70-78) must be greater than zero.
18
-------
TABLE 2. FORMAT FOR AMBIENT CONCENTRATION DATA FILE
Line type
Column
Format3
Contents
Header
1-2
12
'03'
4-15
A12
Receptor identification
17-18
A2
Year, YY
19-22
A4
Month and day, MMDD
24-25
12
Duration of sample, hours
27-28
12
Starting hour of sample
33-34
12
Size fractions on next lines
12 = fine and coarse
13 = fine and total
35-80
Must be blank
Concentrations
1-2
12
'30'
4-15
A12
Receptor identification
17-18
A2
Year, YY
! 19-22
A4
Month and day, MMDD
24-25
12
Duration of sample, hours
27-28
12
Starting hour of sample
33-34
12
Species code
37-45
F9.X
Concentration of fine
fraction
48-56
F9.X
Standard error of fine
fraction (measurement un-
.
certainty)
59-67
F9.X
Concentration of coarse
'fraction or total
70-78
F9.X
Standard error of coarse
fraction or total (measure-
ment uncertainty)
I - Integer (12 ^ 0 through 99)
A - Alphanumeric (A12 up to 12 characters, A-Z, 0-9, any keyboard symbol)
F - Floating point (F9.X ^ up to 8 integers and a decimal point). Note: I
is very helpful in reading the file to have the decimal points aligned
vertically. A decimal point must be included in the field.
19
-------
it
jgyro* |
MMTMMT
vm
3,
a
' *
1
$
m» i ¦
~j «
'1
11
«
!
ff
V
»
as
XI
If
n-
a
71 I »
r
a
9
SI 31
V
3$
*
4t
1
»J
«
I 3
f
N
?
f
a
c h j r
₯
I
f
f
f
a
7
4
f
q
|
f
|
1 l
|
j
1
it
**
IMMQr
r
1
9*m
I
h
r
C9ce*n»roi
9t«K
mCTtM
3
4
11
If
*3
14
tl
!«
IT
a
n
a
B
it
n
if
»l "
*
91
»
9*
»
r«
E
¦m
&
49
M
«c
3
ft[
C
n
FT
Ll
T
3
)
.
ft
7
9
M
' 1
U^-
Ij
t
ft
J
u
n
4
ft
ft
|«
JL
r
»
«
«
9
a
4
~
i*
t
n"
7
2.
ft
3
f
£_
1.
I
n
4
rr
»
3
t
3
_0_
3
«_
n
>
JM
>
-ft_
}
ft_
a
4
ft
ti
«
3
«
ft
i
3
ft
n
4
|_i
«
3
_
n
__
l_t
«
[I
u.
1
E
[Z
ft_
a
4
ft
t
3
4
»
<
3
i
mm
M
P
H
1
r
¦
«
M
M
_L
M
. 1 !
0
9N»
MlD
mm
aw
TO
A
at
<*
at*
*cn
tm
9*
«M«
Q3M
MB
Km
DM
ten
M
CM
1 [
»
t
t
»
3
0
»
" 1
r
«
t
1
7
ft
9
9
t
1
«
i
4
3
ft
ft
>
|
c
ft
ft
1
J
ft
¦
|
3
1
ft
3
1
ft
ft
ft
2
ft
1
t
ft
t
»
4
ft
f
1
ft
4
ft
4
1
1
i
0
«
ft
ft
4
t
9
9
s
ft
4
4
1
ft
1
ft
ft
1
ft
i j
7
ft
3
ft
«
ft
ft
i i
i
t
3
ft
3 1
3
I |
9
ft
.
3
ft
ft
3
ft
3
ft
3
ft
3
9
4
ft
4
4
ft
ft
3
3
ft
ft
t
4
|
4
3
i
ft
9
f
t
9
7
S
ft
ft
t
ft
3
ft
ft
1
3
ft
4
ft
9
1
ft
S
3
ft
?
1
3
9
9
;
ft
2
3
ft
3
ft
3
1
ft
3
ft
ft
ft
3
t 1
ft
3
3
ft
a
ft
i
|
|
Figure 6. Example ambient concentration data file.
20
-------
If only PM^q data are available, all fine particle concentrations should
be set to zero and the "fine and total" designation (13} should be entered
in columns 33-34 of the header record. CMB analyses should be performed only
on the total size fraction.
3.2 FILE GENERATION METHODS
There are three common methods of creating the CMB input files: manually
entering the data in the correct format using a text editor or word processing
program; editing existing input files with a text editor or word processing
program; or transferring files from the EPA source library.
3.2.1 Keyboard Input
A word processing program can be used to type entire input files.
Figures 2 through 6 should be followed in establishing the formats for the
files. Tabs should not be set to facilitate spacing between fields. Completed
files should be appropriately named and then copied onto a diskette or hard
disk as described in Section 2.2.2.
Word processing programs generally must be used in the text mode (with
a carriage return after each line). EDLIN, the line editor supplied with
MD/DOS or PC/DOS, may also be used.
3.2.2 Editing Existing Files
It is often preferable to start with existing input files such as the
test data set and make changes in them with a word processing program or a text
editing program such as EDLIN to obtain the desired files. This method usually
reduces format errors and saves time if only a few changes are needed. Because
of differences in individual editing programs, the user is advised to consult
the manual for the editing program to be used for directions on opening a cooy
of the existing file, deleting and adding material, saving the changes, and
renaming the file.
3.2.3 Transferring Files from EPA Library
More than 100 source profiles for fine and coarse particulate (components
of FS and CS files, the files with most number of lines) are available on
21
-------
diskette from EPA. These source profiles can be copied directly onto the
diskette or hard disk being used to assemble input data files for Version 6
with a software utility program supplied with the diskette version of the EPA
source composition library. Separate documentation accompanies the source
composition library diskette.
Whenever the source profile files are transferred from EPA library files
and merged with Source Name, Chemical Species Code, and Ambient Concentration
Data files generated elsewhere, make sure that source and species codes are
consistent throughout the five files. The diskette version EPA library pro-
files all use atomic numbers for elements, 01 for mass concentration, 91 for
organic carbon, 92 for elemental carbon, 93 for SO^, and 94 for NC^-
22
-------
SECTION 4
MOVING THROUGH AN INTERACTIVE SESSION
The CMB program and its data files are loaded as described in Section 2.
The discussion in this section follows the order of prompts in the inter-
active session exactly. For additional clarification, the user can review
CMB program queries and example responses shown in Appendix C.
4.1 CALLING UP DATA FILES
After the user types CMB6, the program responds with 'MAKE SURE YOUR
CAPS LOCK IS ONI!!' if CMB is in the directory under that name. Be sure to
check that the CAPS LOCK key is on at this time. The program only recognizes
upper case instructions.
The first interactive statement requiring a response is:
DISK FILE FOR INITIAL INPUT?
IF NOT ENTER CARRIAGE RETURN
IF SO ENTER THE FILE NAME
Assuming that the six input files have been prepared as described in Section
3, the user should type in the name of the file that contains the other input
file names (e.g., IN .DAT) at this time. Include the three-character
extender (.DAT) in the response or the program w.ill print an error message
indicating that the file cannot be found.
The program then searches for each of the five input files. If they are
all located and properly named, the program automatically continues with the
next prompt. If an error is found, an appropriate error message is displayed
and the user must correct the files before performing the run. The session
has to be initiated again (by typing in CMB6) after correcting the error.
The second prompt sets up the output file:
DO YOU WISH TO RENAME CMBOUT?
IF NOT ENTER A CARRIAGE RETURN.
IF SO ENTER THE FILE NAME.
23
-------
A permanent record of the last CMB run is created under CMBOUT or the
specified file name (if one is specified) by using the 'WRITE' command. The
results in CMBOUT are lost when a new session is started; printouts become
the only record.
After the second response, the title screen appears. This display re-
mains for almost one minute before the third prompt is shown. The title
screen shows revision number and release date of the Version 6 CMB program
being used. The availability of later editions can be determined by contact-
ing the principal authors or the USEPA Project Officer identified on the title
screen. CMB results will not differ in later editions, but convenience,
efficiency and error trapping may be added in response to user suggestions.
4.2 USING AVAILABLE COMMANDS TO SET UP THE CMB RUNS
Several alternatives can be exercised beginning at this point. The
prompt is:
THE 'HELP' COMMAND LISTS COMMANDS
USE COMMANDS AE, DE, AS, DS FOR CHANGES.
ENTER COMMAND
Available commands and their actions are presented in Table 3.
If the user is uncertain of the complete lists of source categories and
chemical species for which data have been input, or which have been designated
for the initial fit, he or she may want these shown and/or printed out before
proceeding further. This can be accomplished with the command 'PINFO'.
A second option at this time (either before or after using the PINFO
command) would be to add or delete sources or species from the fit designated
in the input files. This can be done prior to making a CMB run with the
commands AE, OE, AS, and DS, as described in the following two sections.
In most cases, the appropriate response to the third prompt (ENTER
COMMAND) is 'SELECT'. This command allows the user to select the ambient data
set (site and date) for the run, as described in Section 4.6.
Most of the corranands in Table 3 are for use after at least one run has
been completed, to change some of the input data for a subsequent run or to
obtain specific results of the completed run. When the program returns to
this point for an additional run or an alternate command, all of the input
24
-------
TABLE 3. COMMANDS IN CMB PROGRAM
General purpose
of command
Command
Specific action
Runs CMB
calculations
CMB
Performs CMB calculations
Changes input
data
SELECT
Allows user to select ambient data set
(site and date) for CMB run
AE
Adds one or more chemical species to the
fit
DE
Deletes one or more chemical soecies from
the fit
AS
Adds one or more sources to the fit
DS
Deletes one or more sources from the fit
SIZE
Changes size fraction
Displays results
PINFO
Shows current and available input data
ambient data set, fitting species, fitting
sources
PMATRIX
Displays requested data from input files-
Source Profile, Ambient Concentration,
Source Names, or Chemical Species Codes
SSCONT
Shows calculated contribution of each
source to each species concentration
PCOMP
Writes computed averages for a series of
CMB runs at a site to an output file
WRITE
Sends current CMB results to diskette
(CMBOUT)
Miscellaneous
HELP
Displays information on available commands
similar to this table
EXIT
1
i
Terminates interactive session; user is
returned to operating system
data (fitting sources, fitting chemical species, ambient data set, and size
fraction) remain at their previous values unless changed by the user. This
feature eliminates the need for repetitive entries.
25
-------
There are six possible commands to initate changes in input data for
additional runs--AE, DE, AS, DS, SELECT, and SIZE. The commands AE or DE
require that the user specify the species to be added or deleted; similarly,
AS or DS require that the user specify the source(s) to be added or deleted.
The command SELECT allows the user to specify a different sampling site/date.
The command SIZE changes the size fraction (from fine to coarse or coarse to
fine) without any further user response.
4.3 SELECTING INITIAL SOURCES
A group of sources has already been designated as initial fitting sources
in the List of Source Names input file (SO .DAT). These should be reviewed
for their applicability to the specific run to be made using the criteria
discussed below.
Selection of source types is limited to the source code numbers in the
Source Name file and to 16 entries. The user should evaluate information
other than the CMB input data to select these sources. Recommended informa-
tion sources are point and area source emission inventories of the area sur-
rounding the sampling site, meteorological data for the sampling day, and
microscopic analysis of the sampled particulate.
All source categories in the Source Names file that could reasonably be
expected to contribute to ambient concentrations on the sampling day should
be included in the initial run. This approach is theoretically correct be-
cause overspecification (inclusion of sources with no impact) results in un-
biased SCE's and larger standard errors than a correctly specified model. In
contrast, underspecification (omission of sources that actually have an
impact) results in biased estimates.
Sources are added to the fit with the command AS and deleted with the
command DS. The program responds to these commands with:
SIZE IS FINE
INPUT CODE OF ADDED (or DELETED) SOURCE
Type the source code of the source to be added or deleted and press ENTER.
The program repeats its prompt to provide an opportunity to add or delete
another source. This interaction is repeated until all source codes
26
-------
associated with the changes have been entered. Enter a blank line after the
last prompt to signal that the changes are complete.
In subsequent runs, sources may appropriately be deleted if prior-round
model results indicate a negligible contribution. However, sources should
not be added just to improve model diagnostics, unless some new external
information indicates a possible impact from the source(s). One proper
reason for adding a source is that there was inadequate capacity to include
it in the first run as one of the initial 16 sources.
If sulfate is to be specified as a fitting species, it should definitely
be included in the fit as a single species source. If it is removed as either
a source or species in a subsequent run, sulfate should be completely excluded
from the fit (both source and species removed).
4.4 SELECTING INITIAL SPECIES
The initial fitting species are designated in the List of Chemical
Species Codes input file (P .DAT). Changes can be made in these prior to
the initial run with the AE or DE commands in a manner exactly analogous to
adding or deleting sources. The fitting species are the only ones for which
mass balance equations are developed during the CMB run.
In general, 21 chemical species are used in a model run. By entering
the maximum number of species (21), the amount of information available for
the least squares solution is maximized.
The user has responsibility for assuring consistency between the fitting
soecies specified at this point in the program, the species available for
each source in its source profile, and the species analyzed in the ambient
sample. For example, fluoride (F) should only be selected as a fitting
species if a ^luoride concentration has been measured for the ambient sample
and all of the source profiles include a fluoride component.
If a species is selected for fitting but no source profile data are
available for that species in the profile data file, the model assumes a "zero"
value and no error message appears. However, serious errors can occur in the
CMB model results. As a consequence of potential problems associated with
missing species data, data availability is often the limiting factor in deter-
mining which species will be selected for inclusion in a CMB run.
27
-------
The species CI, Br, NH^, and N03 should be used with caution because of
atmospheric reactions that have been shown to change their solid phase con-
centrations .
4.5 SELECTING TEST SITE AND DAY
After the command SELECT has been entered, the program responds with the
request:
ENTER DESIRED CMB SITE CODE: XXXXXXXXXXXX
The user should then type and enter the receptor identification name for the
desired site exactly as it appears in columns 4-15 of the Ambient Concentra-
tion Data file. The program then requests:
ENTER YEAR: YY
and
ENTER DATE: MMDD
to which the user should respond (in two steps) with the 2-digit year, month,
and date corresponding to the sampling data set desired. Again, the date
specified should be exactly the same as it appears in columns 17-18 and
19-22 of the Ambient Concentration Data file.
The program then asks for desired size fraction (see Section 4.6 below)
and states:
DATA SEARCH BEGUN FOR
SITE: XXXXXXXXXXXX YEAR: YY DATE: MMDD
If the requested Ambient Concentration Data file is found, the program imme-
diately responds with another ENTER COMMAND prompt. If the specified site
code and date do not match those of a sampling data set in the Ambient Concen-
tration Data file, the program lists all sampling data sets on file and
requests:
28
-------
DATA SET NOT FOUND FOR
SITE: XXXXXXXXXXXX YEAR: YY DATE: MMDD
USE ONE OF THOSE LISTED ABOVE
ENTER COMMAND
The user must then enter 'SELECT' again and the site code and date for one of
the listed sampling data sets.
If the contents or exact names of data sets in the Ambient Concentration
Data file are unknown when the interactive session begins, any names and
dates can be input in response to the orogram's prompts. After the program
fails to find the spurious data set, it provides a complete listing from
which the desired data sets can be chosen.
4.6 SELECTING SIZE FRACTION
In order to completely define the desired ambient data set, the user
must specify whether the fine or coarse size fraction is to be analyzed.
This is done by typing 'F' or 'C' in response to the program query:
INPUT DESIRED SIZE FRACTION: (FINE OR COARSE)
If PM1q data are substituted for coarse fraction in all files, 'C is still
specified at this point. Records of the size change must be kept outside
the model.
Check that data for the desired size fraction are available in the
source profile and ambient data files. If data files are named but contain
no dSta, the program will run through the least squares analysis and report
mainly zeros in the summary tables.
4.7 INITIATING A CMB RUN
At any time after the ambient data set has been selected, a CMB least
squares solution procedure can be obtained by typing 'CMB' in response to ENTER
COMMAND. The program makes no response while the calculations are being per-
formed (usually about one minute), but displays three screens of results after
the calculations are complete with no additional commands.
29
-------
To stop the displays from scrolling, the keys (CTRL) and (NUMLOCK) can
be pressed simultaneously. To restart, press any key or the space bar.-
Another ENTER COMMAND prompt is displayed after the three screens have
been viewed. The user then has the following options:
1. Enter one of the six commands that change input data in preparation
for an additional run (AE, DE, AS, DS, SELECT, or SIZE)
2. Obtain additional statistics on the run just completed
3. Display results or input data in different formats
4. Store the results in a disk report file with the WRITE command
5. Exit the CMB program.
Commands under Option 1 have been discussed in this section. Option 4 is
discussed in the following subsection; Options 2 and 3 are discussed in
Section 5.
4.8 RECOVERING SUMMARY INFORMATION
Printouts of data from CMB runs are normally obtained with the PRT'SC
key on an intermittent or continuous basis during program execution, as
described in Section 2.3.1.
The results of each run may also be placed in permanent storage for
later use or display by employing the WRITE command following the CMB command.
A permanent record of each run in a session is maintained separately. Results
from different sessions can be distinguished by renaming the CMBOUT file at
the beginning of each session. The WRITE command must be repeated after every
run to save the results of that run. Storage of data is confirmed by the
program response 'WRITTEN'.
To retrieve these permanent output files, the CMB program must be exited.
Then, with the computer still in the same subdirectory (with hard disk), the
printer must be put in condensed (17.5 characters per inch or 132 column)
mode. This is accomplished by following instructions in your printer manual.
A printout of an expanded version of the screen displays 'PRINT CMBOUT' or
the renamed file for CMBOUT.
30
-------
If both a fine fraction and coarse fraction run were performed for the
same site/date, an additional report can be obtained. By typing the command
'PRINT SUMMRV , a summary of the fine, coarse, and total source contributions
for that date is generated and printed.
31
-------
SECTION 5
EVALUATING CMB MODEL OUTPUTS
Three different tabular summaries of model outputs and performance
measures are automatically produced after the CMB command is entered:
1. Source Contribution Display
2. Uncertainty/Similarity Display
3. Species Concentration Display.
The model outputs and guidelines for their evaluation are summarized
below. Interpretation is discussed elsewhere (USEPA 1987A). Mathematical
formulae and significance levels are presented with a more detailed explana-
tion in Appendix A.
5.1 SOURCE CONTRIBUTION ESTIMATES DISPLAY
An example Source Contribution Table is shown in Figure 7. Six outputs
and diagnostics plus a restatement of the ambient sampling data and conditions
are included in this display.
5.1.1 Reduced Chi Square
The reduced chi square, which will be called a chi square in this manual,
is a measure of the goodness-of-fit for a CMB run. The value of chi square
is inversely proportional to the squares of the uncertainty terms for species
in the source profiles and ambient data. Therefore, a data set with relatively
high uncertainties would have a lower chi square than a corresponding data
set with relatively low uncertainties.
Chi square for a sample can be any positive value, with values approach-
ing 1.0 indicating the best fit. A value greater than 4.0 suggests that the
model has not explained ambient measurements well.
32
-------
SOURCE CONTRIBUTION ESTIMATES - SITE: PACS3 DATE: 0807 77 VERSION: 6.0
SAMPLE DURATION 24 START HOUR 0 SIZE: COARSE
R SQUARE .97 PERCENT MASS 102.2
CHI SQUARE 1.50 DF 13
SOURCE
* TYPE
1 MARIN
3 UDUST
4 AUTPB
5 RDOIL
8 KRAFT
11 ALPRO
12 STEEL
13 FERMN
SCE(UG/M3)
15.5726
61.8332
35.1579
13.2395
8.4592
4.4485
-2.1408
1.3753
STD ERR
2.4487
4.9872
4.4520
2.0769
8.8923
2.7329
1.6307
.8020
TSTAT
6.3595
12.3984
7.8972
6.3746
.9513
1.6278
-1.3128
1.7149
MEASURED CONCENTRATION FINE/COARSE/TOTAL:
135.0+- 13.5/ 135.0+- 13.5/ 270.0+- 19.1
9
Figure 7. Source contribution estimates.
33
-------
5.1.2 R-Square
R-square is the variance in the ambient species data as explained by the
calculated species concentrations. It is determined by a linear regression of
measured versus model-calculated values for each of the fitting species.
R-square values can range from 0.0 to 1.0. The closer the value is to
1.0, the better the model has explained ambient species concentrations for
that data set. If R-square is less than 0.8, the model is considered tc have
poorly predicted species concentrations and, therefore, source contributions.
R-square is mathematically related to the reduced chi square as shown in
Appendix A. It will generally approach a value of 1.0 when the reduced chi
square indicates a "good fit."
5.1.3 Percent Mass
Percent mass is the ratio of the sum of model-calculated species concen-
trations (for the particle size fraction being analyzed) to the measured mass
concentration. The model should account for essentially all of the measured
mass (percent mass = 100.0). If the model under- or overpredicts mass by more
than 20 percent, it has not been well constructed. Percent mass values less
than 80 percent may indicate a missing source type or species not included in
the fit. Values over 120 percent may indicate inclusion in the fit of a
source type that actually has no impact or collinearity problems, as discussed
in Section 5.2.
Even though all three statistics that measure overall CMB model per-
formance with a given data set (R-square, chi square, and percent mass) have
values in the "acceptable" range, the fit may still be improved by justifiable
changes in the input data. Improvement is indicated by change in these sta-
tistics toward their optimum values of 1.0 for R-square, 1.0 for chi square,
and 100.0 for percent mass.
5.1.4 Degrees of Freedom (DF)
Degrees of freedom is the number of species in the fit (maximum 21)
minus the number of source types (maximum 16). The model does not run if DF
is less than one. Many of the model statistics may deteriorate if DF is less
than five. Number of degrees of freedom must be known for statistical inter-
pretation of the chi square value.
-------
5.1.5 Source Contribution Estimates (SCE)
Source contribution estimates'are the main product of the CMB model.
SCE's are the concentrations in ug/m^ attributed to each of the specified
sources. The sum of these concentrations should approximate the measured
mass concentration.
Negative SCE values are not physically meaningful, but tney can occur as
a result of collinear source profiles or when their absolute values are less
than their corresponding standard errors. These latter negative values are
of no concern. If a negative SCE is reported, source combinations (in the
next display) including this source should be examined to identify the
collinear sources and properly quantify their combined contribution.
5.1.6 Standard Error (STD ERR)
The standard error of estimate is the uncertainty in the associated SCE.
The standard error is interpreted by use of the t-statistic described below.
3
As an approximation, as the standard error in yg/m approaches the value of
the SCE, the SCE may no longer be significantly different from zero and is
therefore not meaningful.
5.1.7 T-Statistic (TSTAT)
The TSTAT measures the statistical significance of an SCE by comparing
the SCE to its standard error. A TSTAT value of less than about 2.0 indi-.
cates that the SCE is not significantly different than zero. The presence
of low TSTAT values for several sources within a run may be due to col lin-
earities among source profiles or to uncertainties in the input data.
5.2 UNCERTAINTY/SIMILARITY DISPLAY
The display following the source contribution estimates presents uncer-
tainty/similarity clusters. Figure 8 shows an example of this output.
The first column of this display contains the clusters, one cluster on
each row. Eaeh cluster is identified by the code numbers associated wjth a
source profile. The clusters are formed if two criteria are met: 1) two or
more source types in an eigenvector derived from the singular value
35
-------
UNCERTAINITY/SIMILARITY CLUSTERS
SUM OF CLUSTER SOURCES
1 11
12 13
12 13
20.02+-
-.76+-
-.76+-
3.67
.86
.86
Figure 8. Uncertainty/similarity display.
decomposition exceed 0.25 (these terms are explained in Appendix A), and
2) the t-statistic for any one of these source types is less than or equal to
2.0. These uncertainty/similarity clusters are caused by excessive similarity
(col 1inearity) among the source profiles in the cluster or by high uncertainties
in the individual source profiles. The standard errors associated with the
source contribution estimates of one or more sources identified in a cluster
are usually very large, often too large to allow an adequate decision to be
made.
If collinearity is the cause of these excessive uncertainties, then the
uncertainty of the sum of the source contributions for a cluster may be smaller
than the uncertainty of any single source contribution in the cluster. The
sum of source contributions and the standard error of the sum are shown in
the final two columns of this display. This sum may be more useful than the
individual source contribution estimates if the standard error of the sum is
substantially lower than the standard errors of each SCE. The sum does not
allow differentiation among the contribution estimates of sources contained
in the cluster.
An example of the uncertainty/similarity clusters display is shown in
Figure 8. In it, sources 1 and 11 form one cluster. The estimated sum of
the SCE's- for these two sources is 20.02 +_ 3.67. A second cluster is formed
by sources 12 and 13. The repeat of a cluster should be ignored.
In an actual example, the cluster numbers will not appear if the two
criteria above are not met. This would be interpreted as meaning that there
are no source clusters in that model run that are sufficiently similar to
cause the model estimates to have high uncertainties.
36
-------
5.3 SPECIES CONCENTRATIONS DISPLAY
An example Species Concentrations display is shown in Figure 9. The data
presented in the table heading are the same as in the heading of the SCE dis-
play (Figure 7). The body of the table shows measured and model-calculated
concentrations for each chemical species so that the model's fit by species
can be examined. Species used in the fit are marked with an asterisk in the
column labeled "I." The column labe'led "M" contains an "M" if data are miss-
ing from the receptor file. This "M" allows the user to distinguish missing
data from nondetectable levels of a species.
There are two diagnostics included in the Species Concentrations display.
The column labeled RATIO R/U contains the ratio of the signed difference be-
tween the calculated and measured concentration (the residual) divided by the
uncertainty of that residual. The R/U ratio is calculated for all species,
even those not included in the fit. The lower the absolute value of the
ratio, the better the model has explained the species (e.g., a low residual
with a high uncertainty will provide a low RATIO R/U). A high RATIO R/U (e.g.,
>2.0) indicates that the residual is significant. If the RATIO R/U is <-2.0,
a missing source may contain a significant quantity of that species.
The column marked RATIO C/M shows the ratio of calculated to measured
concentration and the standard error of that ratio for every chemical species
with measured data (including some that are not fitting species). The ratios
should all be near 1.00 if the model has predicted observed species concen-
trations well.
If the C/M ratio for a species is considerably lower than 1.00, the model
is unable to find sources with emissions containing that species to fully
explain its measured presence. This may indicate a missing source that is
rich in the species with the low ratio. If the C/M ratio for a species is
considerably greater than 1.00, the best fit for all species has forced con-
tributions of the overpredicted species to the high level. There are several
possible explanations for this occurrence (USEPA 1987A).
37
-------
SPECIES CONCENTRATIONS - SITE: PACS3 DATE: 08G7 77 VERSION: 6.0
SAMPLE DURATION 24 START HOUR 0 SIZE: COARSE
R SQUARE .97 PERCENT MASS 102.2
CHI SQUARE 1.50 DF 13
SPECIES-
¦I-
M MEAS-
CALC-
RATIO
C/M
RATI
0 R/U
1
TOT
135.00000+-
13.50000
137.94540+-
8.53619
1.02+-
.12
TOT
.2
9
F
.01000+-
.00100
.20279+-
.09532
20.28+-
9.75
F
2.0
11
NA
*
7.59000+-
.75900
8.34527+-
.66827
1.10+-
.14
NA
.7
12
MG
*
1.46000+-
.14600
1.68068+-
.47624
1.15+-
.35
MG
.4
13
AL
*
6.29000+-
.62900
5.94886+-
.51657
.95+-
.13
AL
-.4
14
SI
*
17.10000+-
1.71000
17.65057+-
.93566
1.03+-
.12
SI
.3
16
S
2.12000+-
.21200
2.67595+-
.38234
1.26+-
.22
S
1.3
17
CL
*
9.97000+-
.99700
7.54866+-
1.59989
. 76+-
.18
CL
-1.3
19
K
*
1.37000+-
.13700
1.07583+
.07430
.79+-
.10
K
-1.9
20
CA
*
2.14000+-
.21400
2.67331+-
.24078
1.25+-
.17
CA
1.7
22
TI
*
.48200+-
.04800
.63881+-
.11806
1.33+-
.28
TI
1.2
23
V
*
.51900+-
.05200
.47296+-
.09936
.91+-
.21
V
-.4
24
CR
*
.02700+-
.00300
.03099+-
.03546
1.15+-
1.32
CR
.1
25
MN
~
.11900+-
.01200
.12400+-
.02528
1.04+-
.24
MN
.2
26
FE
~
5.85000+-
.58500
4.19096+-
.35283
.72+-
.09
FE
-2.4
28
NI
*
.78800+-
.07900
.73140+-
.16032
.93+-
.22
NI
-.3
29
CU
*
.04900+-
.00500
.05995+-
.01355
1.22+-
.30
CU
.8
30
ZN
*
.19700+-
.02000
.22672+-
.05656
1.15+-
.31
ZN
.5
35
BR
*
1.38000+-
.13800
1.80798+-
.59785
1.31+-
.45
BR
.7
82
PB
~
7.85000+-
.78500
7.25928+-
1.05885
.92+-
.16
PB
-.4
91
OC
*
24.60000+-
2.46000
22.03127+-
3.68929
.90+-
.17
OC
-.5
92
EC
*
2.31000+-
.23100
2.94891+-
.78235
1.28+-
.36
EC
.8
93
S04
*
8.47000+-
.84700
9.50631+-
1.70589
1.12+-
.23
S04
.5
94
N03
*
.44000+-
.04400
.49675+-
.19718
1.13+-
.46
N03
.3
Figure 9. Species concentrations.
-------
5.4 ADDITIONAL DIAGNOSTICSSSCONT COMMAND
The command SSCONT produces a table that shows the fraction of each
species' calculated ambient concentration contributed by each source in the
fit. An example SSCONTS display is shown in Figure 10.
By pointing out major contributing sources according to the model's fit,
some potentially incorrect source profiles or incorrect extra sources respon-
sible for a large positive residual for a species may be identified.
39
-------
ENTER COMMAND
SSCONT
CALC SPECIES(PER SOURCE)
INDIVIDUAL RATIO =
MEAS SPECIES(ALL SOURCES)
SPEC\SOURCE
1
3
4
5
8
11
12
13
1 TOTAL
.115
.458
.260
.098
.063
.033
-.016
.010
9 F
.000
.495
.000
.702
.000
18.684
.000
.399
11 NA
.821
.143
.000
.061
.059
.014
-.004
.006
12 MG
.512
.652
.000
.000
.000
.082
-.095
.000
13 AL
.000
.649
.061
.011
.004
.221
-.002
.001
14 SI
.000
1.012
.017
.007
.001
.000
-.006
.001
16 S
.242
.000
.066
.831
.132
.000
-.020
.011
17 CL
.625
.000
.106
.000
.025
.005
-.004
.001
19 K
.159
.465
.018
.027
.025
.000
-.014
.105
20 CA
.102
.867
.205
.098
.014
.017
-.062
.008
22 TI
.000
1.296
.000
.030
.000
.007
-.009
.001
23 V
.000
.032
.000
.878
.000
.003
-.002
.001
24 CR
.000
1.031
.000
.230
1.504
.026
-1.665
.021
25 MN
.000
.520
.000
.051
.037
.000
-1.565
1.999
26 FE
.000
.606
.126
.067
.027
.003
-.117
.005
28 NI
.000
.003
.008
.901
.024
.012
-.019
:ooo
29 CU
.000
.379
.524
.203
.104
.127
-.122
.010
30 ZN
.000
.345
.625
.269
.000
.002
-.130
.040
35 BR
.023
.004
1.274
.001
.003
.004
.000
.002
82 PB
.000
.029
.896
.002
.000
.000
-.002
.000
91 OC
.000
.084
.715
.038
.054
.000
.000
.005
92 EC
.000
.415
.578
.178
.066
.031
.000
.009
93 S04
.184
.005
.054
.752
.118
.009
-.006
.007
94 N03
.000
.023
.727
.196
.000
.000
.000
.178
Figure 10. Species-source contributions display.
40
-------
REFERENCES
Belsley, D. A., K. Edwin, and R. E. Welsch. (1980) Regression Diagnostics:
Identifying Influential Data and Sources of Collinearity. John Wiley, New
York.
Benarie, M. M. (1980) Urban Air Pollution Modeling. MIT Press, Cambridge,
Massachusetts.
Bevington, ?. R. (1969) Data Reduction and Error Analysis for the Physical
Sciences. McGraw-Hill, New York.
Britt, H. I., and R. H. Luecke. (1973) The Estimation of Parameters in Non-
linear, Implicit Models. Technometrics, 15:223-247.
Currie, L. A., R. W. Gerlach, C. W. Lewis, et al. (1984) Interlaboratory
Comparison of Source Apportionment Procedures: Results for Simulated Data
Sets. Atm. Env., 18(8):1517-1537.
DeCesar, R. T., and J. A. Cooper. (1982) Evaluation of Multivariate and
Chemical Mass Balance Approaches to Aerosol Source Apportionmnet, Using
Synthetic Data and an Expanded PACS Data Set. Proceedings, Receptor Models
Applied to Contemporary Pollution Problems, Specialty Conference, Air
Pollution Control Association, Denver, MA.
DeCesar, R. T., S. A. Edgerton, M. A. K. Khalil, and R. A. Rasumssen. (1985a)
Sensitivity Analysis of Mass Balance Receptor Modeling: Methyl Chloride as
an Indicator of Wood Smoke. Chemosphere, 14, 1495.
DeCesar, R. T., S. A. Edgerton, M. A. Khalil, and R. A. Rasmussen. (1985b)
A Tool for Designing Receptor Model Studies to Apportion Source Impacts
With Specified Precisions. Receptor Methods for Source Apportionment: Real
World Issues and Applications, Air Pollution Control Association, Pittsburgh.
Friedlander, S. K. (1973) Chemical Element Balances and Identification of
Air Pollution Sources. Environmental Science and Technology, 7:235.
Gordon, G. E., W. H. Zoller, G. S. Kowalczyk, and S. W. Rheingrover. (1981)
Composition of Source Components Needed for Aerosol Receptor Models.
Atmospheric Aerosol: 5ource/Air Quality Relationships (edited by Macias and
Hopke), ACS Symposium Series No. 167, American Chemical Society, Washington,
D.C. 1981, pp. 51-74.
41
-------
Gordon, G. E. (1987), and Ann Sheffield, University of Maryland, Chemistry
Building, College Park, Maryland. Personal communication to Tom Pace, U.S.
Environmental Protection Agency, April 1987.
Henry, R. C. (1982) Stability Analysis of Receptor Models That Use Lowest
Squares Fitting. Receptor Models Applied to Contemporary Pollution Problems,
edited by S. L. Dattner and P. K. Hopke, Air Pollution Control Association,
Pittsburgh, PA. 141.
Hidy, G. M., and S. K. Friedlander. (1972) The Nature of the Los Angeles
Aerosol. 2nd IUAPPA Clean Air Congress, Washington, O.C.
Hopke, P. K. (1985) Receptor Modeling in Environmental Chemistry, John
Wiley and Sons, New York, 1985.
Javitz, H. S., and J. G. Watson. (1986) Feasibility Study of Receptor
Modeling for Apportioning Utility Contributions to Air Constituents,
Deposition Quality and Light Extinction. Draft Report for Electric Power
Research Institute, prepared by SRI International, Menlo Park, CA.
Kneip, T. J., M. T. Kleinman, and M. Eisenbud. (1972) Relative Contribution
of Emission Sources to the Total Airborne Particulates in New York City. 3rd
IUAPPA Clean Air Congress.
Kowalczyk, G. S. , C. E. Choquete, and G. E. Gordon. (1978) Cehmical Element
Balances and Identification of Air Pollution Sources in Washington, D.C.
Atmospheric Environment, 12:1143.
Pasquill, F. (1974) Atmospheric Diffusion. 2nd Ed. Halsted Press, John
Wiley, New York.
Seinfeld, J. H. (1975) Air Pollution: Physical and Chemical Fundamentals.
McGraw-Hill, New York.
U.S. Environmental Protection Agency. (1983) Receptor Model Technical
Series, Vol. Ill, Chemical Mass Balance Receptor Model User's Manual.
EPA450/4-83014.
U.S. Environmental Protection Agency. (1984) Receptor Model Source Composi-
tion Library. EPA-450/4-85-Q02, November.
U.S. Environmental Protection Agency. (1987A) Protocol for Applying and
Validating the CMB Model. EPA-450/4-87-010. U.S. Environmental Protection
Agency, Research Triangle Park, North Carolina. May 1987.
U.S. Environmental Protection Agency. (1987B) Procedures for Reconciling
Differences in Receptor and Dispersion Models. EPA-450/4-87-008. U.S. En-
vironmental Protection Agency, Research Triangle Park, North Carolina. May
1987.
42
-------
Watson, J. 6. (1979) Chemical Element Balance Receptor Model Methodology
for Assessing the Sources of Fine and Total Suspended Particulate Matter in
Portland^ Oregon. Ph. D. dissertation, Oregon Graduate Center, Beaverton,
Oregon.
Watson, J. 6., and N. F. Robinson. (1984) A Method to Specify Measurements
for Receptor Models. Proceedings of the National Symposium on Recent Advances
in Pollutant Monitoring of Ambient Air and Stationary Sources. U.S. Environ-
mental Protection Agency, Research Triangle Park, N.C. April 1984.
Watson, J. G. (1984) Overview of Receptor Model Principles. Journal of the
Air Pollution Control Association, 34:620.
Watson, J. G., A. W. Gertler, G. H. Prowell, J. C. Chow, S. V. Hering, L. W.
Richards, and D. S. Blumenthal. (1985) Causes of Secondary Aerosol in the
San Joaquin Valley Determined from the WOGA Aerosol Data Base. DRI Document
No. 6687.402, prepared for the Western Oil and Gas Association, Reno, NV.
Energy and Environmental Engineering Center, Desert Research Institute.
Watson, J. G., and J. C. Chow. (1986) Volume 4 - An Evaluation of Ambient
Aerosol Chemistry in the Western United States. Draft report of Western
States and Deposition Proj'ect, Phase I of SYSAPP-86-129, prepared for Western
Governor's Association, Denver, Colorado.
Winchester, J. W., and G. D. Nifong. (1971) Water Pollution in Lake Michigan
by Trace Elements from Aerosol Fallout. Water, Air, and Soil Pollution, 1:50.
43
-------
APPENDIX A
THEORY OF CHEMICAL MASS BALANCE MODEL
A. 1 NOMENCLATURE
The following symbols will be used to describe the CMB model solutions,
diagnostics, and results.
C- = The measured ambient concentration of species i
C = (Ci...Ct)T, a column vector whose ith component is C.
-t 1 1 1
F. . = The fraction of source j composed of species i, the source
J composition coefficient
F = An I x J matrix of F.., the source composition matrix
- 1J
I = Number of species used in the least squares calculation
J = Number of sources used in the least squares calculation
crr = Standard deviation of the C- measurement
S' 1
Qr = Standard deviation of the F.. measurement
ij J
J
aD = Standard deviation of ( I F..S--C-)
K . , "111 1
1 j=i -J J .
V = Diagonal matrix of effective variances,
S. = Source contribution of source j
J
£ = (Sr..Sj)T, a column vector whose jth component is S.
2 rv,
;< = Chi square
A-l
-------
2
Xj j = Reduced chi square
D- = Dispersion parameter for source j
J
E- = Emission rate for source j
J
Ct = Total measured mass
A. 2 THE PHYSICAL BASIS FOR RECEPTOR MODELS
A receptor oriented model must always be a representation of reality and
not just a series of equations. The relationship between a source model and
the CMB receptor model is simple and is presented here.
In general, the aerosol mass concentration at a receptor during a sam-
pling period of length T due to a source j with constant emission rate E- is
J
where Dj = /J d (TT(t), a(t), )dt (2)
is a dispersion factor depending on wind velocity, "u(t), atmospheric stability,
a(t), and the location of source j with respect to the receptor, "x.. u and a,
«J
and possibly even If- in the case of a mobile source, will vary with time; the
«J
instantaneous dispersion factor, d, must be integrated over the sampling
period. When = -1, the source of emissions E,. is not in an upwind
J J
quadrant with respect to the receptor and d-0.
Various forms for d have been proposed (Pasqui 1 1, 1974, Benarie, 1976,
Seinfeld, 1975), some including provisions for chemical reactions, removal
and specialized topography. None are completely adequate to describe the
complicated, random nature of dispersion in the atmosphere. The advantage
of receptor models is that an exact knowledge of D. is unnecessary.
J
If a number of sources, J, exists and there is no interaction between
their aerosols to cause mass removal, the total aerosol mass measured at the
receDtor, C, will be a linear sum of the contributions from the individual
sources
A-2
-------
J J
C = Z D.E,; = Z S, (3)
j=l J J j=l J
Similarly, the concentration of elemental component i, C. will be
J J
C. = I F..S. » Z F-.D.E,, i - 1,1 (4)
i j=i ^ J j=l J J
where F.. is the fraction of source contribution S- composed of element i.
vJ
In the source model each S, is known from Equations 1 and 2, the F.; are
J J
determined by chemical analyses of representative samples from source j, and
Ci is calculated from Equation 4. The receptor model starts with the measure-
ments of C-, uses some knowledge about the chemical composition of the sources
and attempts to quantify S,, or at least to make a statement about its varia-
J
bility or significance as a contributor to the total mass concentration, C.
If the C. and the F^. . at the receptor for all J -of the source types
suspected of affecting the receptor are known, and J<_I, a set of simultaneous
equations exists from which the source type contributions, S., may be calcu-
lated. Several solution methods have been proposed and tested. The method
selected for Version 5 of the CMB receptor model is the effective variance
weighted least squares solution (Watson et a!., 1984).
The assumptions of the CMB with an effective variance solution are:
1. Compositions of source emissions are constant over the- period of
ambient and source sampling.
2. Chemical species do not react with each other, i.e., they add
linearly.
3. All sources with a potential for significantly contributing to the
receptor have been identified and have had their emissions
characterized.
4. The number of sources or source categories is less than or equal to
the number of components.
5. The source compositions are linearly independent of each other.
6. Measurement uncertainties are random, uncorrelated, and normally
distributed.
A-3
-------
A. 3 THE CMB MODEL AND MULTIPLE LINEAR REGRESSION
The effective variance solution of the CMB equations is an application
of multiple linear regression (MLR) in which a least squares solution to the
equations is obtained. Several different modifications of the least squares
method have been evaluated during development of versions, as explained in
Section A.8 below.
The measured concentrations of individual chemical species in an ambient
particulate sample constitute the array of response or dependent variables.
The fractions of these chemical species contained in the emissions from each
source type constitute the array of explanatory or regressor variables.
These source chemical composition descriptions are called source profiles,
source signatures, or source fingerprints.
The regression coefficients calculated by multiple linear regression are
the estimated contributions from each identified source type, or the Source
Contribution Estimates (SCE's). The main goal of the CMB modeling effort is
to obtain meaningful values for the SCE's.
MLR has been widely used and researched by statisticians for the past
half century. However, the emphasis has almost always been to fit the data
to a given set of variables rather than to estimate the unknown values in a
physical model. The CMB application is derivable from physical laws and is
not just an empirical fit to acquired data.
A.4 HISTORY OF CMB SOFTWARE DEVELOPMENT
The Chemical Mass Balance receptor model was first developed quasi-
independently by Hidy and Friedlander (1972), Kneip et al. (1972), and Win-
chester and Nifong (1971). The original applications used unique chemical
species associated with each source type, the so-called "tracer" solution.
Friedlander (1973) introduced the ordinary weighted least squares solution to
the CMB equations, and this had the advantages of relaxing the constraint of
a unique species in each source type and of providing estimates of uncertainties
to
associated with the source contributions.
Dr. G. E. Gordon and his students at the University of Maryland (e.g.,
Kowalkzyk et al., 1978) subsequently applied this method to a wider range of
A-4
-------
chemical species that they measured in source and receptor samples. The
ordinary weighted least squares solution was limited in that only the uncer-
tainties of the receptor concentrations were considered; the uncertainties
of the source profiles, which are typically much higher than the uncertain-
ties of the receptor concentrations, were neglected.
The first user-oriented software for the CMB model was programmed in
1979 by Dr. John Watson at the Oregon Graduate Center in FORTRAN IV on a PRIME
300 minicomputer. All subsequent versions have followed tne same basic pattern
of this original, and some of the constraints imposed by limited computer
power and storage at the time of initial development have not been modified.
Version 1 incorporated a variety of solutions, including the ordinary weighted
least squares solution used by Drs. Friedlander and Gordon, an effective
variance procedure which used the uncertainties of the source compositions as
well as the uncertainties of the receptor concentrations, and a theoretically
(but not necessarily practically) "exact" solution which would change the
source compositions toward more statistically correct values with each itera-
tion. Extensive tests with simulated ambient data verified the effective
variance solution's source contribution estimates and their uncertainties.
The Portland Aerosol Characterization Study (PACS) and ancillary studies
in the State of Oregon were ultimately used for the development of SIP's for
TSP. These studies led to the first enhancements of Version 1 by Mr. Pat
Hanrahan of the Oregon Department of Environmental Quality (DEQ).
Version 2 of the CMB Receptor Model was installed on EPA's UNIVAC system
by Mr. John Core of Oregon DEQ during an assignment with EPA in 1980. This
version was identical to Version 1 except that it operated on more versatile
input data files. Receptor model workshops were conducted in Washington,
D.C., Denver, CO, and Houston, TX, for State and local pollution control
agencies using remote tie-ins to the UNIVAC.-
Version 3 was implemented by Radian, Inc. by Messrs. Hugh Williamson and
Dennis DuBcse in 1982 on the EPA UNIVAC. The software was completely rewritten
in FORTRAN 77, though the command structure and features of Version 1 were
retained. A ridge regression option was added to address the issue of col-
linearity in addition to the effective variance solution. The collinearity
issue (i.e., two or more source types having source compositions which are
A-5
-------
too similar to be separated from each other by the model) was found to be a
major limitation of the CMB in practical applications. Excessive collinearity
in an application increases the uncertainty of CMB model results to intolera-
ble levels.
Dr. Ronald C. Henry of the University of Southern California theoretically
examined the collinearity issue in his presentation at the 1982 Receptor Model
Specialty Conference in Danvers, MA (Henry, 1982). He presented the singular
value decomposition of the weighted source matrix as a method of determining
the degree of collinearity in a single application. Dr. Henry also demon-
strated that a ridge regression solution could not return realistic solutions
in cases of excessive collinearity.
Version 4 of the CMB was created at the Desert Research Institute (DRI)
of the University of Nevada, under DRI and Oregon Department of Environmental
Quality (DEQ) sponsorship in 1984. This version transferred the Version 3
computer code"to the IBM/XT and made the modifications required for it to run
under Microsoft FORTRAN, added the original solution subroutines of Versions 1
and 2, and programmed Henry's singular value decomposition. Most of the data
file structures and reporting formats of Version 3 were retained to take
advantage of the documentation in the EPA user's manual prepared for that'
version. Various revisions to this version were made at DRI and by Mr. Pat
Hanrahan of Oregon DEQ in response to user feedback about "bugs" and con-
venience features.
Dr. Richard T. Decesar of Oregon Graduate Center and Dr. Luke Wijnberg
of PEI Associates, Inc. (PEI) began investigating theoretical methods of
identifying collinearity related to CMB applications in 1984, and they
identified additional diagnostics for multiple linear regression solutions
which are aDplicable to the CMB (Belsley et al., 1980). Dr. Wijnberg
specified these diagnostics, and they were implemented in the CMB program at
DRI during the latter half of 1985. The utility of these diagnostics for
practical applications is a topic of current research.
This Version 5 software was distributed to a number of scientists, and
regulators who applied it to both synthesized and real data sets. These
reviewers conversed at a workshop in May 1986, to exchange their results,
A-6
-------
recommend changes to the software, and to construct applications, validation,
and model reconciliation protocols. Their recommendations have been imple-
mented via a series of revisions to Version 5, resulting in the current
Version 6 described in this manual.
A.5 EFFECTIVE VARIANCE WEIGHTED LEAST SQUARES SOLUTION ALGORITHM
The effective variance weighted least squares solution to a set of CMB
equations requires estimates of the source contributions, S-, which are the
unknowns to be calculated. The solution algorithm is, therefore, an iterative
one that calculates a new set of S- based on the S- estimated from the previous
J u
iteration. The first iteration is performed with the Sj arbitrarily set to
zero, and it is equivalent to the ordinary weighted least squares CMB solution
proposed by Friedlander (1973). Derivations of the effective variance solu-
tion to overdetermined sets of linear equations are presented by Britt and
Luecke (1973) and Watson et al. (1984). The algorithm was originally imple-
mented in Version 1 of the CMB software, and is carried out by the following
steps expressed in matrix notation. A superscript k is used to designate the
value of a variabl e at the kth iteration
1. Set initial estimate of the source contributions equal to zero
s/-0 = 0, j=l .... J
2. Calculate the diagonal components of the effective variance matrix,
All off-diagonal components of this matrix are equal to zero,
,,k 2 x i ck2 2
V = ar + i S. Or-
en c, j = i
3. . Calculate the k+1 value of S.
sk+1 =
4. Test the (k+1 )th iteration of the S. against the k_th iteration. If
any one differs by more than 1 percent, then perform the next itera-
tion. If all differ by less than 1 percent, then terminate the algorithm.
A-7
-------
sk+1 - s k
If -Jr ^ >0.01, go to Step 2
s5
Sk+1 - sk
If _j J-<0.01, go to Step 5.
sj
5. Assign the (k+l)th iteration to S- and o<. ^11 other calculations
J J
are performed with these final values of S,- and cc
j j ,
°s.
J
\ J = 1 J
The number of iterations required to meet the test in Step 4 normally
ranges from three to five. Version 6 allows as many as 20 iterations to take
place before returning a result. A warning message is given if a 1 percent
convergence has not been reached within 20 iterations. This lack of convergence
1s usually indicative of an unstable solution caused by collinearity of source
profiles, and other performance measures given by the Version 6 software will
normally confirm this. The source contribution estimates should not be
believed when convergence is not obtained. A different combination of species
and sources in the solution should be used in this .case.
A.6 EFFECTS OF DEVIATIONS FROM CMB MODEL ASSUMPTIONS
Assumptions 1 through 6 stated in Section A.2 are fairly restrictive,
and they will never be strictly complied with in actual practice. Fortunately,
the CMB model can tolerate reasonable deviations from these assumptions, though
these deviations increase the stated undertainties of the source contribution
estimates.
The CMB model has been subjected to a number of tests to determine its
abilities to tolerate deviations from model assumptions (Watson, 1979; Gordon
et a 1., 1981; Currie et al., 1984; DeCesar et al., 1985a,b; Watson et al., 1985;
Watson and Robinson, 1984; Henry, 1982; Dzubay et al., 1984; Javitz and Watson,
A-8
-------
1986; Watson et a 1., 1986). These studies all point to the same basic conclu-
sions regarding deviations from the above-stated assumptions.
With regard to assumption 1, source compositions are known to vary sub-
stantially among sources, and even within a single source over an extended
period of time. These variations are both systematic and random and are
caused by three phenomena: 1) transformation and deposition between the
emissions point and the receptor, 2) differences in fuel type and operating
processes between similar sources or the same source in time, and 3) uncer-
tainties or differences between the source profile measurement methods.
Evaluation studies have generally compared CMB results from several tests
using randomly perturbed input data and from substitutions of different source
profiles for the same source type.
The general conclusions drawn from these tests are:
The error in the estimated source contributions due to biases in all of
the elements of a source profile is in direct proportion to the
magnitude of the biases.
For random errors, the magnitude of the source contribution errors
decreases as the number of components increases.
With regard to the nonlinear summation of species, assumption 2, no studies
have been performed to evaluate deviations from this assumption. While these
deviations are generally assumed to be small, conversion of gases to particles
and reactions between particles are not inherently linear processes. This
assumption is especially applicable to the end products of photochemical
reactions and their apportionment to the sources of their precursors. Further
model evaluation is necessary to determine the tolerance of the CMB mocel to
deviation from this assumation.
A major challenge to the application of the CMB is the identification of
the primary contributing sources for inclusion in the model, assumption 3.
Watson (1979) systematically increased the number of sources contributing to
his simulated data from four to eight contributors while solving the CMB
equations assuming only four sources. He also examined including more sources
in the least squares solutions than those which were actually contributors,
with the following results:
A-9
-------
Underestimating the number of sources had little effect on the calcula-
ted source contributions if the prominent species contributed by the
missing source were excluded from the solution.
When the number of sources was underestimated, and when prominent species
of the omitted sources were included in the calculation of source con-
tributions, the contributions of sources with properties in common with
the omitted sources were overestimated.
When source types actually present were excluded from the solution, ratios
of calculated to measured concentrations were often outside of the 0.5
to 2.0 range, and the sum of the source contributions was much less than
the total measured mass. The low calculated/measured ratios indicated
which source compositions should be included.
When the number of sources was overestimated, the sources not actually
present yielded contributions less than their standard errors if their
source profiles were significantly distinct from those of other sources.
The over-specification of sources decreased the standard errors of the
source contribution estimates.
Recent research suggests that assumption 3 should be restated to specify
that source contributions above detection limits should be included in the
CMB. At this time, however, it is not yet possible to determine the "detec-
tion limit" of a source contribution at a receptor since this is a compli-
cated and unknown function of the other source contributions, the source pro-
file uncertainties, and the uncertainties of the receptor measurements. Addi-
tional model testing is needed to define this "detection limit."
With regard to assumption 4, it is very likely that the true number of
individual sources contributing to receptor concentrations is much larger than
the number of species that can be measured. It is therefore necessary to
group sources into source types of similar compositions so that this assumption
is met. For the most commonly measured species, meeting assumption 5
practically defines these groupings.
The linear independence of source compositions required by assumption 5
has become a subject of considerable interest since the publication of Henry's
(1982) singular value decomposition (SVD) analysis. This analysis provides
quantitative measures of col linearity and the sensitivity of CMB results to
specific receptor concentrations, which can be calculated analytically in
each application. Henry (1982) also proposed an optimal linear combination
A-10
-------
of source contributions that have been determined to be col linear. These
"regression diagnostics" have been summarized by Belsley et al. (1980) and
have been applied to the CMB by DeCesar et al. (1985 a,b). Version 6 of the
CMB has incorporated these regression diagnostics to test their ability to
identify potential deviations from model assumptions. Section A.7 describes
them as they have been implemented in Version 6. At this point, the exact
values of these diagnostics that correspond to intolerable degrees of multi-
coll inearity are unknown, though this is an area of active research. These
diagnostics are available to advanced users of Version 6, and for research
purposes only, via a set of undocumented commands which may be obtained from
the author.
Tests performed on simulated data with obviously collinear source pro-
files typically result in positive and negative values for the collinear
source types. Unless the source profiles are nearly identical, the sum of
these large positive and negative values very closely approximates the sum of
the true contributions.
With most cormionly measured species (e.g., ions, elements, and carbon)
and source types (e.g., motor vehicle, geological, residual oil sea salt,
steel production, wood burning, and various industrial processes), from five
to seven sources are linearly independent of each other, in most cases.
Gordon et al. (1981) found instabilities in the ordinary weighted least
square solutions to the CMB equations when species presumed to be "unique" to
a certain source type were removed from the solution. Using simulated data
with known perturbations ranging from 0 to 20 percent, Watson (1979) found:
"In the presence of likely uncertainties, sources such as urban dust and con-
tinental background dust cannot be adequately resolved by least square fitting,
even though their compositions are not identical. Several nearly uniaue ratios
must exist for good separation." It is widely recognized at this time that a
substantial amount of research is needed to minimize the effects of source
profile collinearity in the CMB. The Version 6 software has been structured
to facilitate this research as well as to alert regulatory users to the
potential of excessive collinearity in routine applications.
With respect to assumption 6, the randomness, normality, and uncorrected
nature of measurement uncertainties, there are no results available from
A-11
-------
verification or evaluation studies. Every least squares solution to the CMB
equations requires this assumption, as demonstrated by the derivation of
Watson et al. (1984). In reality, very little is known about the distribution
of errors for the source compositions and the ambient concentrations. If
anything, the distribution probably follows a log-normal rather than a normal
distribution, since none of these values can ever be negative, while a normal
distribution allows for a substantial proportion of negative values. For
small errors (e.g., less than 20%) the actual distribution may not be
important; but for large errors it probably is important. A symmetric dis-
tribution becomes less probable as the coefficient of variation of the mea-
surement increases. This is one of the most important assumptions of the
solution method that requires testing.
A.7 CMB MODEL PERFORMANCE MEASURES
Though the CMB model evaluation studies cited in Section A.6 provide
some general conclusions regarding the effects of deviations from model
assumptions, these studies cannot assess the validity of a single CMB model
application. Several analytical performance measures have been incorporated
into Version 6 software to allow the confidence in the model solution to be
evaluated. The values taken by these performance measures will suggest
actions which might be taken to correct or add model input data, change fitting
sources or species, or to combine source profiles.
A.7.1 Reduced Chi Square
The reduced chi square and its associated degrees of freedom indicate
the "goodness-of-fit" of a solution to the CMB equations.
2 1 *
Reduced chi square = y T , = jr Z
i=i
Degrees of freedom = DF = I - J
U
C-- F. -S-
1 J=1 ^ J
v2..
ei i
A-12
-------
For a perfect least squares fit with normally distributed random deviations
in the measurements of C- and F.., the reduced chi square will equal 1
i i J
(Bevington, 1969).
Values of reduced chi square which exceed 10 indicate that one or more
ambient species concentrations calculated from the source contributions differs
from its measured value by more than the measurement and source contribution
uncertainties. The calculated-to-measured ratios and residual to uncertainty
measures indicate which species are most responsible for the large reduced
chi square.
A.7.2 Percent Mass
Percent mass is the fraction of the measured mass concentration which
is accounted for by the source contribution estimates,
J
Percent mass = ( z S-/C.) x 100
j = l J
A percent mass less than 80 percent usually means that not all source types
have been included in the CMB.
A.7.3 R-Square
R-Square is a measure of correlation between the calculated and measured
species concentrations. The definition of R-Square in the CMB assumes a
least squares solution with a zero intercept.
R-Square = 1 - p
z /Ci
1=1 I A.
\ ei i
A-13
-------
A.7.4 Measures of Residuals
The calculated to measured concentration ratio (RATIO C/M-). indicates
how closely the species concentrations calculated from the source contribu-
tion estimates compare with each other. Only the RATIO^ is displayed
J
RATIO C/M. = i F.-S-/C.
0 = 1 J -
J
RATIO R/U. = ( Z F..S. - C,)/aD
1 j=l 1J J 1 Ki
A.7.5 Singular Value Decomposition (Uncertainty/Similarity Clusters)
The Singular Value Decomposition of the weighted F matrix is given by
where U and V are Ixl and JxJ orthogonal matrices, respectively, and D is a
diagonal matrix with J nonzero and positive elements called the singular val
of the decomposition. They are also given as the square roots of the eigen-
value of the symmetric matrix F^ V~* F with the columns of V the associated
J -e ^ %
eigenvectors.
It is these eigenvectors which give rise to the uncertainty/similarity
clusters since rewriting (I) gives
Thus, if a singular value in £ is "small" one has a "small" or nearly
degenerate linear combination of the columns of the weighted F matrix with
coefficients given by the elements of the column of V corresponding to the
"small" singular value. In the CMB model the sources appearing in the un-
certainty/similarity clusters arise from these eigenvectors and are
characterized by:
A-14
-------
1. Eigenvector elements which have absolute value greater than or
equal to 0.25
2. At least two sources satisfy (1)
3. At least two of the sources satisfying (1) have a TSTAT value less
than or equal to 2.0.
A. 8 SOLUTION METHODS
Version 6 of the CMB model is the first that contains no user options
for using the least squares solution procedure (or fitting method). The
current version employs the effective variance weighted least squares method
exclusively. In previous versions, the following solution methods have been
used:
o Ordinary weighted least squares -
choice in Versions 1 through 4. Versions 3 and 4 provided for
an implicit intercept
o Effective variance weighted least squares -
choice in Versions 1 through 4. Versions 3 and 4 provided for
an implicit intercept
o Britt - Luecke exact solution choice in Versions 1 and 4
o Ridge regression -
choice in Versions 3 and 4.
All solution methods use the same multiple linear regression approach
and matrix solution of the simultaneous equations. The differences lie in
handling of the error term(s). In ordinary weighted least squares, each of
the i equations is modified by dividing both sides of the equation by the
square of the standard error of the species concentration (cf ) to weight the
i
more accurate concentration data more heavily. However, the errors in the
source profile fractions (aP ), which are usually larger than cr , are
i j i
ignored in the ordinary weighted least squares method.
The error term V in the effective variance weighted least squares
i i
method includes both a^ and ap components, as explained in Section A.5.
i i j
Effective variance is an estimate of total uncertainty, which takes into
A-15
-------
account errors in all input data. The same error term V is also used in
e..
the Britt-Luecke solution, which is intended to provide an exact solution of
the CMB equations. The Britt-Luecke procedure is not employed in Version 6
because it has not been shown to significantly improve the CMB, despite its
greater complexity and computational requirements.
Both the effective variance least squares and Britt-Luecke solutions
require the iterative process for estimating S- values because the error
J
term, V . , is a function of S,.. The ordinary weighted least squares method
" ' 1 J
has no cv term, and therefore does not require a convergent solution,
i n
The ridge regression fitting method is described in the User's Manual
for Version 3 of the CMB Model (EPA-450/4-83-014, July 1983).
A-16
-------
APPENDIX B
DIAGNOSIS OF ERROR MESSAGES
B-l
-------
t
Error message
File system error in file CMBOUT
Error Code 1034, Status 0004
Error opening (XXXXXX.XXX)
What is the name of your __
AKT*JEFFIN*AK MATRIX IN SUB-
ROUTINE CEB2 NEEDS IMPROVEMENT
00
1
^ Some receptor concentration
standard errors are less than or
equal to zero. Weighted regression
cannot be done in this case
File not found in file Error
Code 1032, Status 000A
RECORD NUMBER XX OF SOURCE NAME FILE
XXXXXX.XXX DID NOT MATCH THE REQUIRED
FORMAT (12, 2X, A8, 2X, Al) AND WAS
IGNORED
(continued)
r
Probable cause
Corrections
CONFIG.SYS file not on DOS used to boot
system or not in root directory
1) File name does not exist or is in-
correct
2) File name extender not used
3) CONFIG.SYS has insufficient number
of files
1) Too many fitting sources for number
of fitting species (No. species
minus No. sources must be >_ 1)
2) A single constituent source (e.g.,
secondary sulfate or nitrate) is in-
cluded, but that constituent is not a
fitting species
<
A receptor data set has not been selected
for modeling and each ambient concentra-
tion is set at a default value of 0.0
File name extension not included
Increase number of files to
14 in CONFIG.SYS file
Leave interactive session
using CTRL-ALT-DEL and
correct file name or
CONFIG.SYS
Reduce sources or add
species
Add omitted species if
ambient data are available
otherwise delete the
single species source
In response to ENTER
COMMAND, type SELECT and
then specify site, date,
and size range
Specify full file name as
given in directory, e.g.,
XXXXXXX.DAT
If input data are improperly formatted,
these data will not be read into memory
and may result in erroneous data
Change the record (line) in
the XXXXX.XXX file to match
the required format
-------
Error message
Probable cause
Corrections
RECORD NUMBER XX OF POLLUTANT NAME Input data are improperly formatted
FILE XXXXXX.XXX DID NOT MATCH THE
REQUIRED FORMAT (12, 2X, A8, 2X, Al)
AND WAS IGNORED
RECORD NUMBER XXX OF SOURCE COMPOSI-
TION FILE XXXXXX.XXX DID NOT MATCH
THE REQUIRED FORMAT AND WAS IGNORED
AN UNIDENTIFIED POLLUTANT CODE it XX
WAS FOUND IN FILE XXXXXX.XXX SOURCE
- S04
FITTING ELEMENT § XX HAS NON-
POSITIVE UNCERTAINTY - .00000 PLEASE
REPLACE WITH POSITIVE DETECTION
LIMIT PROGRAM TERMINATED
, view the
results and use the CMB
Application and Validation
Protocol to help locate the
problem
(continued)
-------
Error message
Probable cause
Corrections
SSCONT output wraps around so each
species takes two lines instead of
1. The top of the table scrolls
off of the screen
More than nine sources are included in
the source list
Species number has zero ambient data The referenced ambient species has an
uncertainty uncertainty < or = to zero
Reenter the SSCONT command
and press
simultaneously to stop the
scroll. Press any key to
continue the scroll
Make uncertainty equal to
detection limit
CP
I
i
-------
APPENDIX C
PRINTOUT OF TEST DATA SET FOR PACS1
(«**** USER RESPONSE)
C-l
-------
A>CMB6
MAKE SURE THAT YOUR CAPS LOCK IS ON ! !!
<<**~~
DISK FILE FOR INITIAL INPUT?
IF NOT ENTER CARRIAGE RETURN
IF SO ENTER NAME OF DISK FILE
INPORT.DAT <<****
DO YOU WISH TO RENAME CMBOUT?
IF NOT ENTER A CARRIAGE RETURN.
IF SO ENTER THE FILE NAME.
CMBEXAM1 <<****
**************** kk ************** * * ********** **********************************
* *
* U. S. EPA CHEMICAL MASS BALANCE RECEPTOR MODEL *
* *** IBM-PC VERSION 6.0 (REVISED 7/20/87) *** *
* *
* EPA PROJECT MGft: THOMPSON G. PACE III, PE *
* U.S. ENVIRONMENTAL PROTECTION AGENCY *
* OFFICE OF AIR QUALITY PLANNING AND STANDARDS *
* RESEARCH TRIANGLE PARK, NC *
* (919)5415585 *
* *
* *
* PRINCIPAL AUTHOR: DR. JOHN G. WATSON *
* DESERT RESEARCH INSTITUTE *
* UNIVERSITY OF NEVADA SYSTEM *
* (702)9721676 *
* CONTRIBUTING AUTHORS: *
* DR. J.C. CHOW MR. P.L. HANRAHAN DR. N.F. ROBINSON *
* MR. J.E. CORE DR. R.C. HENRY DR. H.J. WILLIAMSON *
* MR. D.A. DUBOSE MR. T.G. PACE DR. L. WIJNBERG *
* *
************************************ ****************************** ********** **
C-2
-------
THE "HELP" COMMAND LISTS COMMANDS.
USE COMMANDS AE,DE,AS,DS FOR CHANGES.
ENTER COMMAND
HELP «**++
HELP-LISTS THESE COMMANDS
DATA ACCESS AND SEQUENCING
SELECT-SELECT DATA SET FOR CMB
SIZE-CHANGE SIZE FRACTION
EXIT-CLOSE FILES AND LEAVE
CMB OPERATIONS
AE-ADD A POLLUTANT TO THE FIT
DE-DELETE A POLLUTANT FROM THE FIT
AS-ADD A SOURCE TO THE FIT
DS-DELETE A SOURCE FROM THE FIT
CMB-PERFORM CMB
-= SCREEN DISPLAY
PINFO - PRINT CURRENT STATUS ON SCREEN
PMATRIX - PRINT SOURCE SIGNATURE OR RECEPTOR CONCENTRATIONS
SSCONT - PRINT SOURCE CONTRIBUTIONS TO ELEMENTAL CONCS. TO SCREEN
PCOMP - PRINT COMPUTED AVERAGES OF CMB SERIES
DATA STORAGE
WRITE - WRITE CMB RESULTS TO DISK
SELECT SHOULD NORMALLY BE THE FIRST COMMAND
ENTER COMMAND
SELECT <<***~
ENTER DESIRED CMB SITE CODE: XXXXXXXXXXXX
PACSl
ENTER YEAR: YY
77 <<**~~
ENTER DATE: MMDD
0813 <<***-*
INPUT DESIRED SIZE FRACTION:(FINE OR COARSE)
F <<**~~
¦DATA SEARCH BEGUN FOR
SITE: PACSl YEAR: 77 DATE: 0813
SITE: PACSl YEAR: 77 DATE: 0813 FC
C-3
-------
ENTER COMMAND
CMB
SOURCE CONTRIBUTION ESTIMATES - SITE: PACS1
SAMPLE DURATION 24 START HOUR
R SQUARE .98 PERCENT MASS
CHI SQUARE 1.12 DF
DATE: 0813 77 VERSION:
0 SIZE: FINE
98.7
13
6.0
SOURCE
~
TYPE
SCE(UG/M3)
STD ERR
TSTAT
1
MARIN
12.3882
2.2452
5.5177
3
UDUST
9.5918
1.3876
6.9127
4
AUTPB
10.0834
1.4941
6.7486
5
RDOIL
11.0601
1.9238
5.7490
8
KRAFT
4.6928
5.0424
.9307
11
ALPRO
10.6022
3.5896
2.9536
12
STEEL
8.6728
1.3771
6.2980
13
FERMN
11.8754
1.8321
6.4820
MEASURED CONCENTRATION FINE/COARSE/TOTAL:
80.0+- 8.0/ 80.0+- 8.0/ 160.0+- 11.3
UNCERTAINTY/SIMILARITY CLUSTERS VERSION: 6.0 SUM OF CLUSTER SOURCES
1 8 17.081+- 4.238
1 5 8 28.141+- 3.829
C-4
-------
SPECIES CONCENTRATIONS - SITE: PACS1 DATE: 0813 77 VERSION: 6.0
SAMPLE DURATION 24 START HOUR 0 SIZE: FINE
R SQUARE .98 PERCENT MASS 98.7
CHI
: SQUARE 1.
12
DF
13
SPECIES-
I-M-
MEAS-
CALC-
RATIO
C/M--
RATIO
R/U
1
TOT
80.00000+-
8.00000
78.96679+-
4.82217
.99+-
.12
TOT
-.1
9
F
.88300+-
.08800
.67644+-
.24792
.77+-
.29
F
-.3
11
NA
~
6.93000+-
.69300
6.97037+-
.56443
1.01+-
.13
NA
.0
12
M6
*
.43000+-
.04300
1.60949+-
.62626
3.74+-
1.50
MG
1.9
13
AL
~
4.66000+-
.46600
4.02416+-
.88919
.86+-
.21
AL
-.6
14
SI
~
3.02000+-
.30200
2.92212+-
.13329
.97+-
.11
SI
-.3
16
S
2.95000+-
.29500
3.02499+-
.31806
1.03+-
.15
s
.2
17
CL
~
5.95000+-
.59500
5.69358+-
1.24829
.96+-
.23
CL
-.2
19
K
~
1.64000+-
.16400
1.73088+-
.46411
1.06+-
.30
K
.2
20
CA
~
1.78000+-
.17800
1.43535+-
.11366
.81+-
.10
CA
-1.6
22
TI
~
.08300+-
.00800
.10088+-
.01630
1.22+-
.23
TI
1.0
23
V
~
.37200+-
.03700
.39756+-
.08308
1.07+-
.25
V
.3
24
CR
~
.31500+-
.03200
.20977+-
.12151
.67+-
.39
CR
-.8
25
MN
*
2.99000+-
.29900
2.82843+-
.14115
.95+-
.11
MN
-.5
26
FE
~
4.53000+-
.45300
4.24446+-
.33269
.94+-
' .12
FE
-.5
28
NI
~
.76500+-
.07700
.68248+-
.13427
.89+-
.20
NI
-.5
29
CU
~
.04400+-
.00400
.05274+-
.00510
1.20+-
.16
CU
1.3
30
ZN
~
.22500+-
.02300
.26786+-
.03966
1.19+-
.21
ZN
.9
35
BR
~
.41900+-
.04200
.56133+-
.17386
1.34+-
.44
BR
.8
82
PB
~
2.53000+-
.25300
2.13748+-
.30300
.84+-
.15
PB
-1.0
91
OC
~
7.54000+-
.75400
8.50981+-
1.35631
1.13+-
.21
OC
.6
92
EC
*
1.42000+-
.14200
1.33579+-
.34011
.94+-
.26
EC
-.2
93
S04
*
10.30000+-
1.03000
9.78932+-
1.47512
.95+-
.17
S04
-.3
94
N03
~
.63800+-
.06400
.88401+-
.35938
1.39+-
.58
N03
.7
C-5
-------
ENTER COMMAND
SSCONT
<<**~*
CALC SPECIES(PER SOURCE)
INDIVIDUAL RATIO =
MEAS SPECIES(ALL SOURCES)
SPEC\SOURCE 1 3 4 5 8 11 12 13
1 TOTAL .155 .120 .126 .138 .059 .133 .108 .148
9 F .000 .000 .000 .007 .000 .720 .000 .039
11 NA .715 .017 .000 .056 .086 .063 .016 .053
12 MG 1.383 .290 .000 .000 .069 .690 1.311 .000
13 AL .000 .182 .024 .013 .003 .614 .012 .016
14 SI .000 .708 .027 .035 .002 .012 .144 .039
16 S .139 .012 .014 .499 .186 .050 .058 .068
17 CL .833 .000 .051 .000 .014 .024 .027 .008
19 K .106 .060 .004 .019 .043 .014 .049 .760
20 CA .097 .131 .071 .098 .000 .020 .302 .087
22 TI .000 .740 .000 .147 .003 .051 .209 .066
23 V .000 .006 .000 1.023 .000 .018 .014 .008
24 CR .000 .014 .000 .017 .042 .000- .578 .016
25 MN .000 .004 .000 .002 .000 .000 .252 .687
26 FE .000 .127 .047 .073 .012 .011 .613 .055
28 NI .000 .001 .002 .775 .008 .026 .079 .000
29 CU .000 .065 .167 .189 .022 .106 .552 .097
30 ZN .000 .047 .157 .197 .014 .007 .463 .306
35 BR .059 .005 1.203 .003 .015 .009 .000 .045
82 PB .000 .014 .797 .005 .000 .001 .026 .002
91 OC .000 .150 .669 .103 .011 .055 .000 .142
92 EC .000 .125 .270 .241 .007 .172 .000 .125
93 S04 .120 .004 .013 .516 .182 .045 .021 .048
94 N03 .000 .000 .144 .113 .000 .068 .000 1.061
ENTER COMMAND
WRITE <<¦*~**
WRITTEN
ENTER COMMAND
SIZE <<*~~*
SIZE IS COARSE
C-6
-------
ENTER COMMAND
CMB
<<~~~~
SOURCE CONTRIBUTION ESTIMATES - SITE
!: PACS1
SAMPLE DURATION
24
START
HOUR
0
R SQUARE
.97
PERCENT
MASS
100.6
CHI SQUARE
1.44
DF
13
SOURCE
* TYPE SCE(UG/M3)
STD ERR
TSTAT
1 MARIN
10.6029
1.8239
5.8132
3 UDUST
9.5985
1.2616
7.6081
4 AUTPB
9.0906
1.3961
6.5113
5 RDOIL
9.7127
1.6107
6.0302
8 KRAFT
12.3264
7.9091
1.5585
11 ALPRO
11.0997
2.2441
4.9462
12 STEEL
8.1587
1.5239
5.3540
13 FERMN
9.8720
1.6165
6.1072
OATE: 0813 77 VERSION: 6.0
SIZE: COARSE
MEASURED CONCENTRATION FINE/COARSE/TOTAL:
80.0+- 8.0/ 80.0+- 8.0/ 160.0+- 11.3
UNCERTAINTY/SIMILARITY CLUSTERS VERSION: 6.0 SUM OF CLUSTER SOURCES
C-7
-------
SPECIES CONCENTRATIONS - SITE: PACS1
SAMPLE DURATION 24 START HOUR
R SQUARE .97 PERCENT MASS
CHI SQUARE 1.44 DF
SPECIES-
-I-
¦M MEAS-
CALC-
1
TOT
80.00000+-
8.00000
30.46153+-
9
F
.73400+-
.07300
.50073+-
11
NA
*
6.33000+-
.63300
6.07759+-
12
MG
~
1.48000+-
.14800
1.48676+-
13
AL
*
4.84000+-
.48400
4.40992+-
14
SI
*
3.27000+-
.32700
3.38783+-
16
S
2.50000+-
.25000
2.41255+-
17
CL
~
4.67000+-
.46700
5.19693+-
19
K
*
1.12000+-
.11200
1.44197+-
20
CA
~
1.52000+-
.15200
1.47195+-
22
TI
~
.14000+-
.01400
.13692+-
23
V
~
.277Q0+-
.02800
.34841+-
24
CR
~
.00800+-
.00100
.24531+-
25
MN
~
2.47000+-
.24700
2.43813+-
26
FE
~
5.41000+-
.54100
4.11644+-
28
NI
~
.77900+-
.07800
.63016+-
29
CU
~
.05000+-
.00500
.06613+-
30
ZN
~
.21400+-
.02100
.23739+-
35
BR
~
.52000+-
.05200
.51379+-
82
PB
~
1.78000+-
.17800
1.93077+-
91
OC
~
10.10000+-
1.01000
8.38185+-
92
EC
~
1.68000+-
.16800
1.34287+-
93
S04
~
8.10000+-
.81000
8.11918+-
94
N03
~
1.13000+-
.11300
.71048+-
DATE: 0813 77 VERSION: 6.0
0
SIZE:
COARSE
100.6
13 "
RATIO C/M-
RATIO
R/U
6.36333
1.01+-
.13
TOT
.0
.22362
.68+-
.31
F
-1.0
.47089
.96+-
.12
NA
-.3
.52389
1.00+-
.37
MG
.0
.55081
.91+-
.15
AL
-.6
.15779
1.04+-
.11
SI
.3
.28216
.97+-
.15
S
-.2
1.07658
1.11+-
.26
CL
.4
.38620
1.29+-
.37
K
.8
.10333
.97+-
.12
CA
-.3
.03204
.98+-
.25
TI
-.1
.07289
1.26+-
.29
V
.9
.11640
30.66+-15
.05
CR
2.0
.12229
.99+-
.11
MN
-.1
.30571
.76+-
.09
FE
-2.1
.11802
. 81+-
.17
NI
-1.1
.00683
1.32+-
.19
CU
1.9
.03413
1.11+-
.19
ZN
.6
.15795
.99+-
.32
BR
.0
.27347
1.08+-
.19
PB
.5
1.29686
.83+-
.15
OC
-1.0
.39741
.80+-
.25
EC
-.8
1.25659
1.00+-
.18
S04
.0
.41362
. 63-1
.37
NO 3
-1.0
C-8
-------
ENTER COMMAND
SSCONT
<<~~~~
CAIC SPECIES(PER SOURCE)
INDIVIDUAL RATIO =
MEAS SPECIES(ALL SOURCES)
SPEC\SOURCE 1 3 4 5 8 11 12 13
1 TOTAL .133 .120 .114 .121 .154 .139 .102 .123
9 F .000 .001 .000 .007 .000 .635 .000 .039
11 NA .670 .027 .000 .054 .103 .042 .016 .048
12 M6 .344 .100 .000 .000 .000 .202 .358 .000
13 AL .000 .131 .021 .011 .007 .718 .011 .013
14 SI .000 .822 .023 .029 .005 .003 .125 .030
16 S .140 .000 .015 .517 .163 .000 .064 .067
17 CL .908 .000 .058 .000 .077 .029 .032 .009
19 K .133 .088 .006 .024 .044 .000 .067 .925
20 CA .098 .189 .075 .101 .029 .059 .333 .084
22 TI .000 .692 .000 .076 .000 .060 .117 .032
23 V .000 . 009 . 000 1.206 . 000 . 016 . 018 . 009
24 CR .000 .540 .000 .571 7.396 .222 21.417 .518
25 MN .000 . 004 . 000 . 002 . 003 . 000 . 287 . 691
26 FE .000 .102 .035 .053 .042 .008 .483 .038
28 NI .000 .000 .002 .668 .035 .030 .073 .000
29 CU .000 .058 .133 .146 .148 .311 .457 .071
30 ZN .000 .049 .149 .182 .000 .005 .457 .268
35 BR .041 .001 .874 .002 .013 .026 .000 .030
82 PB .000 .020 1.021 .006 .000 .000 .035 .002
91 OC .000 .032 .450 .067 .193 .000 .000 .088
92 EC .000 .089 .206 .179 .132 .106 .000 .088
93 S04 .131 .001 .015 .577 .180 .023 .025 .051
94 N03 .000 .002 .073 .056 .000 .000 .000 .498
ENTER COMMAND
WRITE «~~**
WRITTEN
WRITTEN
ENTER COMMAND
EXIT «*~**
Stop - Program terminated.
C-9
-------
~~~ CUB SOURCE CONTRIBUTION SUMARY
RESULTS FOR CUB SITE: PflCSI YEAR: 77 DATE: 0S13 VERSION: 6.0
SAMPLING DURATION: 24 HRS. KITH START HOUR: 0
SOURCE FINE COARSE TOTAL
U6/JIJ I U6/N3 I U6/H3 I
MARIN
12.388 ~-
2.245
15.485 *-
3.205
10.403 ~-
1.824
13.254 ~-
2.437
22.991 +-
2.393
14.349 ~- 2.074
UDUST
9.592 f
1.388
11.990 +-
2.109
9.599 *-
1.242
11.998 *-
1.982
19.190 *-
1.875
11.994 ~- 1.447
AUTP3
10.083 ~-
1.494
12.404 f-
2.253
9.091 ~-
1.394
11.343 ~-
2.082
19.174 -
2.045
11.984 ~- 1.533
RSfllL
11.060 +-
1.924
13.825 ~-
2.774
9.713 ~-
1.411
12.141 ~-
2.351
20.773 +-
2.509
12.983 ~- 1.817
KRAFT
4.493 ~-
5.042
5.844 +-
4.330
12.324 ~-
7.909
15.408 ~-
10.004
17.019
9.380
10.437 +- 5.910
ALPRO
10.402 *-
3.590
13.253 ~-
4.479
11.100 ~-
2.244
13.875 ~-
3.129
21.702 *-
4.233
13.544 ~- 2.814
STEEL
8.473 ~-
1.377
10.841 4-
2.034
8.159 +-
1.524
10.198 ~-
2.141
14.832 ~-
2.054
10.520 ~- 1.484
FERKN
11.875 ~-
1.832
14.844 ~-
2.729
9.872 ~-
1.414
12.340 +-
2.348
21.747 *-
2.443
13.592 «'- r.804
CALC.NASS
70.947 *-
4.022
98.708 +-
11.564
80.442 ~-
4.343 100.577 ~-
12.823
159.428 ~-
7.984
99.443 ~- 8.434
HEAS.MASS
80.000 ~-
3.000
80.000 ~-
8.000
140.000 *-
11.314
C-10
-------
RESULTS FOR CUB SITE: PACS1 YEAS: 77 DATE: 0813 VERSION: 4.0
FINE PARTICULATE FRACTION
SAMPLING DURATION: 24 WS. WITH START HOUR: 0
S-SOIKWE: .98
CHI SOUARE: 1.12
DF: 13
I TYPE
1
NARIN
12.388+-
3
UOUST
9.592+-
4
AUTPB
10.083+-
5
RDOIL
11.060+-
8
KRAFT
4.693+-
11
ALPRO
10.602+-
12
STEEL
8.673+-
13
FERHN
11.875+-
TOTAL: 78.967+-
IJ I
2.245 15.485*- 3.205
1.388 11.990*- 2.109
1.494 12.404+- 2.253
1.924 13.825+- 2.774
5.042 5.864+- 6.330
3.590 13.253+- 4.679
1.377 10.84I+- 2.034
1.832 14.844+- 2.729
4.822 98.708+-11.566
'JKCERTAINTY/SIHILARITY CLUSTERS: SUM OF CLUSTER SOURCES
1 8 17.061+- 4.236
15 8. 28.141+- 3.829
HISS FINE SUSPENDED PARTICULATE
SPECIES INCL FLE REAS. US/13 PERCENT CALC. UE/H3 RATIO R/U
1
TOTAL
30.00000+-
8.00000
100.00000+-
14.14214
78.96679+-
4.82217
-.1
TOTAL
9
F
.88300+-
.08600
1.10375+-
.15583
.67644+-
.24792
-.8
f
11
KA l
6.93000+-
.69300
8.66250+-
1.22506
6.97037+-
.56443
.0
NA
12
N6 *
.43000+-
.04300
.5375d+-
.07601
1.4094?+-
.62626
1.9
fl£
13
AL
4.64000+-
.46600
5.82500+-
.32378
4.02416+-
.8891?
-.6
AL
14
SI «
3.02000+-
.30200
3.77500+-
.53387
2.92212+-
.13329
-.2
SI
16
S
2.95000+-
.29500
3.68750+-
.52149
3.02499+-
.31806
.2
3
17
CL ~
5.95000+-
.59500
7.43750+-
1.05182
5.69358+-
1.24829
-.2
CI
19
K »
t.64000+-
.14400
2.05000+-
.28991
1.73088*-
.464!!
\
20
CA »
1.78000+-
.17800
2.22500+-
.31466
1.43535+-
.'.'.366
CA
22
TI ~
.08300+-
.00800
.10375+-
.01441
.10089+-
.01630
1.0
n
23
V »
.37200+-
.03700
.46500+-
.04555
.39756+-
.08308
. 3
V
24
CR ~
.31500+-
.03200
.39375+-
.05613
.20977+-
.12151
-.5
CR
25
JIN »
2.99000+-
.29900
3.73750+-
.52856
2.82843+-
.14115
_ »
KN
26
Ft i
4.53000+-
.45300
5.46250+-
.80080
4.24446-r-
.33249
-. j
cr
23
NI
.76500+-
.07700
.95625+-
.13568
.68248+-
.13427
N!
29
CU «
.04400+-
.00400
.05500+-
.90743
.05274+-
.00510
1.3
CI!
30
ZN »
.22500+-
.02300
.28125+-
.04022
.26786+-
.03966
4
-------
RESULTS FOR CUB SITE: PAC51 YEAR: 77 DATE: 0613 VERSION: 6.0
COARSE PARTICULATE FRACTION
SAMPLING DURATION: 24 HRS. KITH START HOUR: 0
S-SOUAKE: .97
CHI SQUARE: 1.44
OF: 13
I TYPE L'S/N3
1
NARIN
10.603+-
1.824
13.234+- 2.637
3
UDUST
9.599*-
1.262
11.993+- 1.982
4
AUTPB
9.091*-
1.396
11.363+- 2.082
5
RDOIL
9.713+-
1.611
12.141+- 2.35!
8
KRAFT
12.326*-
7.90V
15.408+-10.006
tl
ALFRO
11.100*-
2.244
13.875+- 3.129
!2
STEEL
8.159+-
1.524
10.198+- 2.161
13
FtRNN
9.872*-
1.616
12.340+- 2.368
TOTAL: 30.462*- 6.361 100.577+-12.823
UNCERTAlNTY/SINiLARITY O.USTERS: SUK OF CLUSTER SOURCES
KISS COARSE SUSPENDED PARTICULATE
STECIES INCL FL6 NEAS. U6/W PERCENT CALC. US/IB WIO R/U
1
TOTAL
80.00000+-
8.00000
100.00000*-
14.14214
80.46133*-
6.36333
.0
TOTAL
9
F
.73400*-
.07300
.91750*-
.12940
.50073*-
.22362
-1.0
F
11
KA l
6.J300C+-
.63300
7.91230+-
1.11900
6.07759*-
.47089
-.3
KA
12
KG <
1.4B000+-
.14800
1,85000*-
.26163
1.46676*-
.52389
.0
N6
13
AL »
4.84000+
.48400
. 6.05000+-
.85560
4.40992+-
.55081
-.6
AL
14
SI f
3.27000+-
.32700
4.08750*-
.57806
3.38783+-
.15779
.3
SI
!6
S
2.50000*-
-.25000
3.12500+-
.44194
2.41255*-
.28216
-t2
3
17
CL ~
4.67000*-
.46700
5.83750*-
.82555
5.19693+-
1.07658
.4
CL
19
K »
1.12000+-
.11200
1.40009*-
.19799
1.44197+-
.3B620
.8
K
20
CA ~
1.52000+-
.15200
1.90000*-
.26870
1.47195+-
.10333
-.3
CA
22
n «
.14000+-
.01400
.17500+-
.02475
.13692+-
.03204
TI
23
v
.27700+-
.02E00
.0492!
.34641+-
.07289
.9
V
24
CH »
.00800+-
.00100
.01000+-
.00160
.24531+-
.11440
2.0
CK
NN »
2.47000*-
.24700
3.08730*-
.43664
2.43B13+-
.12229
i
NN
26
FE i
5.41000+-
.54100
6.76230+-
.95636
4.11644*-
.30571
-2.1
¦E
28
NI <
. 77900*
.07800
.97375*-
.I37SC
.630*6+-
.11802
-1.1
M!
29
CU «
.05000+-
.00500
,0s250+-
.0085*
.06613+-
,00683
!.?
a
30
ZK ~
.21400+-
.02100
.26750+-
.03743
.2373?+-
.03413
.6
ZN
35
8K I
.52000+-
.05200
.65000+-
.091?Z
.51379+-
.15795
.0
3R
92
PB ~
1.78000*-
.17800
2.22500+-
,3146c
1.93077*-
.27347
.5
PB
?!
OC «
<0.10000+-
1.01000
12.62509+-
1.785*4
9.38185*-
1.29686
-1.0
ac
92
EC ~
1.68000+-
.16800
2.10000+-
.29698
1.342S7+-
.39741
-.8
EC
93
504 »
8.10000*-
.81000
10.12500+-
1.43199
9.11918*-
1.25659
.0
S04
?4
N03 ~
1.13000+-
.11300
1.4'250+-
.19976
.71048+-
.41362
-1.0
N03
MEASURED AK8IENT HASS (U6/N3): FINE: 80.0+- 0.0 COARSE: 50.0+- 8.0 TOTAL: i60.0*-11.3
C-12
-------
RESULTS FOR CHS SITE: PACS1 YEAS: 77 DATE: 0813 VERSION: 6.9
TOTAL PARTICULATE FRACTION
SAMPLING DURATION: 24 HftS. KITH START HOUR: 0
RESULTS DERIVED FROM FINE AND COARSE SAMPLES
t TYPE US/N3 I
1
MARIN
22.991+-
2.893
14.369+- 2.074
J
UDUST
19.190+-
1.875
11.994+- 1.447
4
AUTPB
19.174+-
2.045
11.984+- 1.533
5
RDOIL
20.773+-
1.509
12.983+- 1.817
8
KRAFT
17.019+-
9.380
10.637+- 5.910
11
ALPRO
21.702+-
4.233
13.564+- 2.814
12
STEEL
16.832+-
2.034
10.520+- 1.484
13
FERNN
21.747+-
2.443
13.592+- 1.804
TOTAL: 159.428*- 7.984 99.643+- 8.634
HISS TOTAL SUSPENDED PARTICULATE
SPECIES IffCL RE HEAS. U6/N3 PERCENT ' CAiC. US/113
1
TOTAL
I60.0000Q+-
11.31371
100.09000+- 10.00000
159.42830+-
7.98406
TOTAL
9
F
1.61700+-
.11434
1.01063+-
.10106
1.17717+-
.33387
p
11
NA
13.26000+-
.93858
8.28730+-
.82917
13.04796+-
.73307
NA
12
IK *
1.91000+-
.15412
1.19373+-
.12806
3.09625+-
.81630
KG
13
AL ~
9.50000+-
.67187
5.93730+-
.59380
8.43408+-
1.04397
AL
14
SI *
6.29000+-
.44512
3.93125+-
.39328
6.30995+-
.20656
SI
16
S
3.45000+-
.38668
3.40625+-
.34121
5.4J754+-
.42518
S
17
CL ~
10.62000+-
.75638
6.4I750+-
.66616
10.89051+-
1.64841
CL
19
K t
2.7M00+-
.19860
1.72500+
.17402
3.17285+-
.60378
K
20
CA ~
3.30000+-
.23407
2.06250+-
.20657
2.90730+-
.15361
M
22
TI »
.22300+-
.01612
.13938+-
.01410
.23781+-
.03594
TI
23
V *
.64900+-
.04640
.40563+-
.04079
.74597+-
.11052
V
24
CR »
.32300+-
.03202
.201B7+-
.02458
.4550B+-
.16B26
CR
25
(IN ~
5.46000+-
.38783
3.41250+
.34202
5.26657+-
.18676
NN
26
FE »
9.94000+-
.70541
6.21250+-
.62247
6.36090+-
.45181
FE
28
NI «
1.5*400+-
.10960
.96500+-
.09669
1.3126*+-
.17B77
NI
29
CU »
.09400+-
.00440
.03875
.00577
¦ H08B+-
.00832
CU
30
ZN <
.43900+-
.03114
.27436+-
.02748
.50525-
.05232
:h
35
BR <
.93900+-
.06684
.58688+-
'.05860
1.07511+-
.23469
BR
82
PB «
4.31000+-
.30934
2.69375-
.27141
4.06825+-
.40316
PB
91
DC »
17.64000+-
1.26040
11.02500+-
1.10829
16.69165+-
1.87655
oc
92
EC »
3.^00OO+-
.21997
i.937SC+-
. 19409
2.47565+-
.52308
CP
93
S04 ~
!8.4QOOO+-
1.31034
I1.50000+-
1.13410
i 7.90850+-
1.93778
S04
94
N03 <
^ .7&800+-
.12787
1.10500+-
.1126b
1.59449+-
.54794
N03
MEASURED AtlBIENT MASS (U6/H3): FINE: B0.0+- 8.0 COARSE: 30.O*- 8.0 TOTAL: 160.0+-U.3
C-13
-------
TECHNICAL REPORT l#ATA
(Please ruJ Instructions on the re-ersc bejort- LOriph tin ^
|i report no.epa/SU/DK-89/033 a 2
1 EPA-450/4--83-014R
3. RECIPIENT S ACCESSION NO.
PRft9-181101
4'Rece6ptorSKodfetETechnical Series Volume III (Revised):
CMB User's Manual (Version 6.0)
6. REPORT DATE
May 1987
6. PERFORMING ORGANIZATION COC'E
7. AUTHORIS)
Kenneth Axetell, John G. Watson, and Thompson G. Pace
S. PERFORMING ORGA Ni " AT ION REPCflT NC.
I
9. PERFORMING organization name and aooress
PEI Associates, Durham, NC
Desert Research Institute, Reno, NV
U.S. EPA, OAQPS, RTP, NC
10. program element no. \
11. CONTRACT'GRANT NO
68-02-3890, WA No. 38 and
CX-813087-01-1
12. SPONSORING AGENCY NAME AND AOORESS
U.S. EPA
OAQPS, MDAD, Mail Drop 14
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVEPE j
14 SPONSORING AGENC CODE
15. supplementary notes
EPA Project Officer: Thompson G. Pace For Diskette, see: pb89-181093
'^fieStfienncal Mass Balance (CMB) Model uses the chemical composition of an ambient
particulate sample to estimate the relative contributions of different source cate-
gories to the measured particulate concentration. The chemical composition of each
source category's emissions (source profile) must also be known 1n order to run the
model.
Thlsmanual describes Version 6.0 of the CMB Receptor Model. It is designed to allow
persons to use the CMB Model constructively with only a few hours' learning time.
Because of this emphasis on rapid command of modeling procedures, the manual covers
primarily the mechanical aspects of running the model. Users seeking more Information
on the statistical or theoretical bases of chemical mass balance receptor modeling are
referred to "Appendix A: Theory Of Chemical Mass Balance Model."
The manual is intended for wide use by State and local air pollution control agency
personnel in developing State Implementation Plans (SIPs) for PMiq. The U.S.
Environmental Protection Agency has published a companion document to this manual
that should be consulted for this application. The Protocol For Applyinq And
Validating The CMB Model,' EPA-450/4-87-010, provides guidance on applicability,
assumptions, and interpretation of results. The protocol provides a practical
strategy for obtaining valid results.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSati I ield/Croup
Receptor Models
Chemical Mass Balance
Source Apportionment
Least Squares
ftjltlple Linear Regression
Microcomputer Software User's Manual
1
1». DISTRIBUTION STATEMENT
19. SECURITY Class (ThuRtport,
Unl1m1ted
2i. no. of pages
82
20 SECiipiTY CLASS (This page/
unlimited
22. PRICE
iPA Farm 2220-1 (R«». 4-77) mivioui coition is oilOLCTt
------- |