United States . Office of Air Quality EPA -450/4-8 1-035
Environmental Protection Planning and Standards December 1931
Agency Research Triangle Park NC 2771 1
Air
v>EPA Analysis of Inhalable
and Fine Particulate
Matter Measurements
-------
EPA-450/4-81-035
Analysis of Inhalable and Fine
Particulate Matter Measurements
by
John G. Watson, Judith C. Chow and Jitendra J. Shah
EPA Project Officer: Thompson G. Pace
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
December 1981
-------
DISCLAIMER
This report was prepared for the U.S. Environmental Protection
Agency by Environmental Research & Technology, Inc., Concord,
Massachusetts, under Contract No. 68-02-2542, Task 6. The contents of
this report are reproduced herein as received from the contractor.
The opinions, findings, and conclusions expressed are those of the
authors and not necessarily those of the U.S. Environmental Protection
Agency. Mention of trade names or commercial products does not
constitute endorsement or recommended use.
-------
ABSTRACT
Total, inhalable and fine particulate matter measurements
acquired by EPA's Inhalable Particulate Network in 1979 and 1980 are
summarized and analyzed in this report. The theoretical collection
efficiencies of different samplers were calculated and compared to
each other and to an acceptable performance range. The measurement
processes and several of the major urban airsheds of the IP Network
are described. The spatial, temporal and statistical distributions of
these measurements are examined. A receptor-oriented model relating
IP to TSP is derived and tested for prediction accuracy under various
situations. A mass balance receptor model is applied to IP and FP
chemical concentrations in four urban areas to estimate the
contributions of various emissions source types to ambient mass
concentrations.
111
-------
ACKNOWLEDGEMENTS
Many people had a hand in the gathering of information and the
preparation of this report. Considerable thanks is due to project
officer, Tom Pace of U.S.EPA/OAQPS, for originally formulating the
problems to be addressed and coordinating the flow of information
between disparate sections of the Environmental Protection Agency and
this contractor. Tom provided expert technical review on all sections
and prepared the structure and much of the text for Chapter 5 on urban
spatial and temporal distributions.
Bob Burton, Gardner Evans, Joe Walling and Charles Rodes of
EPA/EMSL answered scores of questions about sampler operations,
placement, precision and accuracy. Their intimate knowledge of the
measurement process and their ability to communicate it to the users
of their data added significantly to the accuracy and precision of the
conclusions drawn in this report.
Harold Barkhau of EPA/OAQPS provided consultations on the best
uses of the NEDS emission inventories and then supplied selected
contents of the data base in the most convenient manner. Neil Frank
provided expert advice on the proper methods of interpreting
statistical summaries of suspended particulate matter concentrations.
Stuart Dattner of the Texas Air Control Board supplied the inventories
for Houston and El Paso and reviewed the reduced forms presented in
this report for accuracy. Bob Scott of Philadelphia Air Management
Services reviewed the Philadelphia emissions inventories and spent
considerable time with the authors performing site surveys and
sampling local industrial sources.
Cliff Frazier of NEA Labs performed x-ray fluorescence analyses,
Kochy Fung, Betty Nuesca and Robert Swanson of ERT quantified carbon
and ion concentrations, and Russ Crutcher of Boeing Aerospace
performed optical microscopic analyses of selected IP Network filter
samples.
Tom Pace, Charles Rodes, Ralph Baumgardner, Joe Walling, Ned
Meyer, Neil Frank, John Bachman, Bob Stevens, Tom Dzubay of EPA,
Stuart Dattner of Texas Air Control Board, and Bob Scott of
Philadelphia Air Management Services provided many constructive
criticisms of the initial draft which have been incorporated into this
-------
final report. A special note of thanks is due Dr. Peter Mueller of
the Electric Power Research Institute (EPRI) for his comprehensive
critique.
Peter Mueller of the Electric Power Research Institute provided
valuable consultation, particularly as input to Chapter 6 concerning
non-urban size-selective particulate matter concentrations from the
SURE data base. Dr. Paul Lioy of the New York Institute of
Environmental Medicine also provided substantial consultation and
encouragement.
Monica Chang drew the majority of the illustrations while Audrey
Rose, Marge Hagy, Marion Keith and David Hathaway typed the text and
tables in this report.
These and many others provided the valuable contributions which
made this report possible.
J.G.W.
J.C.C.
J.J.S.
vi
-------
TABLE OF CONTENTS
Page
DISCLAIMER ii
ABSTRACT iii
ACKNOWLEDGEMENTS v
TABLE OF CONTENTS vii
LIST OF TABLES ix
LIST OF ILLUSTRATIONS xv
EXECUTIVE SUMMARY S-l
1. INTRODUCTION 1~1
2. AEROSOL SAMPLER COLLECTION CHARACTERISTICS 2-1
2.1 IP Network Sampler Configuration and
Collection Effectiveness 2-1
2.2 Collection Efficiencies of IP Network Samplers 2-7
2.3 Relative Collection Efficiencies of Different
Sampler Inlets 2-14
2.4 The SURE Sampling Inlet Characteristics 2-18
2.5 Comparison of Calculated and Measured Relative
Collection Efficiencies 2-21
3. THE INHALABLE PARTICULATE MATTER SAMPLING NETWORK
MEASUREMENT PROCESS 3-1
3.1 Sampler Locations and Site Descriptions 3-1
3.2 Filter Media 3-12
3.3 Network Operations 3-23
3.3.1 Filter Weighing 3-25
3.3.2 Field Sampling 3-26
3.3.3 Sulfate and Nitrate Analysis 3-27
3.3.4 Elemental Analysis 3-28
3.3.5 Carbon Analysis 3-29
3.3.6 Optical Microscopic Analysis 3-30
3.3.7 Accuracy and Precision of Analyses 3-32
3.4 Data Validation 3-33
3.5 Precision and Accuracy of Size-Classified
Mass Measurements 3-36
4. URBAN AREAS IN THE IP NETWORK 4-1
4.1 General Meteorological and Statistical Information 4-1
4.2 Industrial Emissions in Several Urban Areas 4-3
vii
-------
TABLE OF CONTENTS (Continued)
Page
5. GEOGRAPHICAL AND SEASONAL VARIABILITY OF TSP, IP, AND FP
IN URBAN AREAS 5-1
5.1 TSP, IP and FP Urban Concentrations 5-1
5.2 Regional Geographical Distributions 5-2
5.3 Urban Geographical Distributions 5-8
5.4 Neighborhood Geographical Distributions 5-10
5.5 Seasonal Distributions 5-13
6. GEOGRAPHICAL AND SEASONAL VARIABILITY OF TSP, IP AND
FP IN NON-URBAN AREAS 6~1
6.1 TSP, IP and FP Non-Urban Concentrations 6-1
6.2 Regional Geographical Distributions 6-7
6.3 Seasonal Distributions 6-12
6.4 Non-Urban Contributions to Urban Concentrations 6-13
7. STATISTICS OF THE IP NETWORK CONCENTRATIONS 7-1
7.1 The Log-Normal Frequency Distribution 7-2
7.2 Cumulative Frequency Distributions in the IP Network 7-6
7.3 Arithmetic and Geometric Means 7-11
7.4 Variability of Means and Maxima with Sample Size 7-17
8. PREDICTING IP CONCENTRATIONS FROM TSP CONCENTRATIONS 8-1
8.1 A Receptor Model Approach 8-2
8.2 Site Type and Concentration Stratification 8-8
8.3 Predicting IP from TSP Concentrations 8-20
8.4 Predicting Averages and Maxima of IP from
TSP Concentrations 8-33
9. SOURCE CONTRIBUTIONS TO INHALABLE PARTICULATE MATTER 9-1
9.1 Validity of IP Network Chemical Composition
Measurements 9-3
9.2 Urban-Scale Chemical Compositions and Possible
Source Type 9-20
9.3 Urban-Scale Source Contributions 9-26
9.4 Neighborhood-Scale Source Contributions 9-50
10. SUMMARY, CONCLUSIONS AND FUTURE RESEARCH 10-1
APPENDIX A SAMPLING SITE DESCRIPTIONS A-l
APPENDIX B AVERAGE METEOROLOGICAL MEASUREMENTS & URBAN
CHARACTERISTICS OF IP NETWORK CITIES B-l
APPENDIX C REFERENCES C-l
Vlll
-------
LIST OF TABLES
Table Title Page
1.1.1 Questions Concerning Inhalable Particulate 1-4
Matter
2.2.1 Geometric Mean Aerodynamic Diameters and
Geometic Standard Deviations for Selected
Atmospheric Particle Size Distributions 2-9
2.2.2 Theoretical Collection Efficiencies for Different
Aerosol Sampler Inlets and Different Collection
Effectiveness Curves for Typical Size Distributions 2-12
2.3.1 Ratios of IP Sampler Collection Efficiencies to HIVOL
Collection Efficiencies as a Function of Wind Speed
and Particle Size Distribution 2-15
2.3.2 Ratios of Dichotomous Sampler Collection Efficiencies
to HIVOL Size-Selective Inlet Collection
Efficiencies as a Function of Wind Speed and
Particle Size Distribution 2-17
2.3.3 Ratios of Hypothetical Ten Micron Collection
Efficiencies to IP Sampler Collection Efficiencies
as a Function of Wind Speed and Particle Size
Distribution 2-17
2.4.1 Theoretical Collection Efficiencies for
SURE SFS, SSI, and Hypothetical 10 urn Inlets 2-20
2.5.1 Relationships Between TOTAL and HIVOL Mass
Concentration Measurements 2-23
2.5.2 Relationships Between SSI and HIVOL Mass
Concentration Measurements 2-24
2.5.3 Relationships Between TOTAL and SSI Mass
Concentration Measurements 2-25
3.1.1 Study Scales 3-1
3.1.2 Description of Entries in Table A.I of Appendix A 3-4
3.1.3 Site Survey Summary 3-7
3.2.1 Comparisons of Arithmetic and Geometric Average
Mass Concentrations for HIVOL, SSI, TOTAL, FINE,
and COARSE taken in 1979 and 1980 3-16
IX
-------
LIST OF TABLES
Table Title Page
3.2.2 Comparison of Measured Arithmetic Average
Concentrations in 1980 on S&S Filters to Those
Predicted from 1979 Measurements on Quartz
Filters 3-17
3.4.1 Data Validation Flagging Criteria 3-34
4.2.1 Potential Sources of Inhalable and Fine
Suspended Particulate Matter in U.S. Urban
Areas 4-7
4.2.2 IP Sampling Sites in Seven Urban Areas 4-19
5.1.1 Quarterly and Annual Arithmetic Averages
of TSP, IP and FP in Urban Areas at
Thirty-Eight Network Sites 5-3
5.1.2 Quarterly and Annual Maxima of TSP, IP
and FP in Urban Areas at Thirty-Eight Network Sites 5-5
5.3.1 Ranges of TSP, IP and FP Annual
Arithmetic Average and Maximum Concentrations
Between Sites Within Urban Areas 5-9
6.1.1 Quarterly and Annual Arithmetic Averages of
TSP, IP and FP in U.S. Non-Urban Areas at
Eight IP Network Sites 6-3
6.1.2 Quarterly and Annual Maxima of TSP, IP and
FP at Eight IP Network Sites 6-4
6.1.3 Locations of Nine SURE Class I Sampling Sites 6-6
6.1.4 Seasonal and Annual Arithmetic Averages of
TSP, IP and FP at Nine SURE Sites 6-8
6.1.5 Seasonal and Annual 24-Hour Maxima of TSP,
IP and FP at Nine SURE Class I Sites 6-9
6.2.1 Ranges of Non-Urban Arithmetic Averages in
the Western and Eastern U.S. 6-10
6.2.2 Ranges of Non-Urban Maximum 24-Hour
Concentrations in the Western and Eastern
U.S. 6-10
6.4.1 Ranges and Averages of Arithmetic Average
Concentrations at Urban and Non-Urban Sites
in the Western and Eastern U.S. 6-14
-------
LIST OF TABLES
Table Title Page
6.4.2 Ratios of Non-Urban to Nearby Urban
Concentrations of TSP, IP and FP 6-15
7.3.1 Average Ratios of Geometric to Arithmetic
Averages at IP Sampling Sites 7-13
7.3.2 Average Ratios as a Function of Geometric
Standard Deviation for IP Sampling Sites 7-14
7.3.3 Variability of the Ratios of Geometric and
Arithmetic Averages with Sampling Frequency 7-16
7.4.1 Number of Measurements in Subsets of All
Measurements at Philadelphia Sites 7-19
8.2.1 Receptor Model Evaluation of SSI Concentration
as a Function of HIVOL Concentration for Data Sets
Stratified by Site Type and HIVOL Concentration 8-9
8.2.2 Receptor Model Evaluation of TOTAL Concentration
as a Function of HIVOL Concentration for Data Sets
Stratified by Site Type and HIVOL Concentration 8-10
8.2.3 Receptor Model Evaluation of FINE Concentration
as a Function of HIVOL Concentration for Data Sets
Stratified by Site Type and HIVOL Concentration 8-11
8.3.1 Prediction Error Frequency Distributions Using
SSI/HIVOL Models Derived from Single Sites for
Various Data Sets not Included in the Model 8-23
8.3.2 Prediction Error Frequency Distributions Using
TOTAL/HIVOL Models Derived from Single Sites for
Various Data Sets not Included in the Model 8-24
8.3.3 Prediction Error Frequency Distributions Using
Models Derived from St. John's to Predict SSI
and TOTAL Concentrations at Other Sites in the
Bridesburg Industrial Area 8-28
8.3.4 Prediction Error Frequency Distributions at
Philadelphia Sites Using St. John's SSI
Measurements to Predict SSI Concentrations
at Other Sites 8-30
8.3.5 Prediction Error Frequency Distribution for
Philadelphia Sites Using Models Derived from
Alternate Data Pairs 8-34
XI
-------
LIST OF TABLES
Table Title Page
9.1.1 Average and Maximum Chemical Concentrations
in Buffalo, NY (10/79 - 9/80) 9-4
9.1.2 Average and Maximum Chemical Concentrations
in Houston, TX (10/79 - 9/80) 9-7
9.1.3 Average and Maximum Chemical Concentrations
in El Paso, TX (10/79 - 9/80) 9-10
9.1.4 Average and Maximum Chemical Concentrations
in Philadelphia, PA (10/79 - 9/80) 9-13
9.1.5 Examples of Inconsistent Observations at
Angola Big Sister (NYBUB), NY (10/79 - 9/80) 9-16
9.2.1 Arithmetic Average Chemical Concentrations
Stratified by Site Type 9-24
9.2.2 Highest Average Chemical Concentrations
Stratified by Site Type 9-25
9.3.1 Source Types Used in Chemical Element Balances 9-28
9.3.2 Average Source Contributions to FINE and COARSE
Particulate Matter in Buffalo, NY 9-34
9.3.3 Average Source Contributions to FINE and COARSE
Particulate Matter in Houston, TX 9-36
9.3.4 Average Source Contributions to FINE and COARSE
Particulate Matter in El Paso, TX 9-37
9.3.5 Average Source Contributions to FINE and COARSE
Particulate Matter in Philadelphia, PA 9-38
9.3.6 Source Contributions (ug/m^) on Individual
Sampling Days in Philadelphia, PA 9-43
9.3.7 Results of Microscopic Analysis of CP on
Selected SSI Samples in Philadelphia, PA 9-46
9.4.1 Average and Maximum Chemical Concentrations in the
Bridesburg Industrial Area of Philadelphia
(10/79 - 2/80) 9-51
9.4.2 Source Contributions in the Bridesburg
Industrial Area 9-56
XII
-------
LIST OF TABLES
Table Title Page
A.I Sampling Sites in the IP Network A-l
B.I Average Meteorological Measurements for
IP Network Cities B-l
B.2 Urban Characteristics of IP Network Cities B-2
Xlll
-------
LIST OF ILLUSTRATIONS
Figure Page
2.1.1 Collection Effectiveness of Standard HIVOL
and HIVOL Size-Selective Inlets Under
Various Wind Speeds 2-3
2.1.2 Collection Effectiveness of Beckman
and Sierra 244 Dichotomous Sampler
Inlets 2-4
2.4.1 Collection Effectiveness of the SURE Sampler 2-19
2.5.1 Scatterplot of Simultaneous Measurements of
Particulate Matter with the HIVOL Size-
Selective Inlet and the Dichotomous Sampler 2-28
3.1.1 Locations of Sampling Areas in the IP Sampling
Network 3-3
3.2.1 a-b Scatterplot of Simultaneous TOTAL and SSI
Sulfate Concentrations from All IP Sampling
Sites in 1979(a) and 1980(b) 3-19
3.2.2 a-b Scatterplot of Simultaneous HIVOL and SSI
Sulfate Concentrations from All IP Sampling
Sites in 1979(a) and 1980(b) 3-21
3.2.3 Scatterplot of Simultaneous TOTAL and SSI
Nitrate Concentrations from All IP Sampling
Sites in 1980 3-22
3.2.4 Scatterplot of Simultaneous HIVOL and SSI
Nitrate Concentrations from All IP Sampling
Sites in 1980 3-22
3.3.1 Flow of Routine Tasks Performed in the IP
Network Operations 3-24
3.5.1 a-f Scatterplots of Collocated Measurements for
HIVOL in Philadelphia (a), HIVOL in Middletown
(b), COARSE in Philadephia (c), FINE in
Philadelphia (d), TOTAL in Philadelphia (e)
SSI in Middletown (f) 3-42
4 2.1 IP Sampling Sites and Industrial Point
Sources in Birmingham, Alabama 4-12
4.2.2 IP Sampling Sites and Industrial Point
Sources in Phoenix, Arizona 4-13
xv
-------
LIST OF ILLUSTRATIONS (cont'd)
Figure Page
4.2.3 IP Sampling Sites and Industrial Point
Sources in Denver, Colorado 4-14
4.2.4 IP Sampling Sites and Industrial Point
Sources in Buffalo, New York 4-15
4.2.5 IP Sampling Sites and Industrial Point 4-16
Sources in Houston, Texas
4.2.6 IP Sampling Sites and Industrial Point
Sources in El Paso, Texas 4-17
4.2.7 IP Sampling Sites and Industrial Point
Sources in Philadelphia, Pennsylvania 4-18
4.2.8 Sampling Site and Point Source Locations
in the Bridesburg Industrial Area of
Philadelphia 4-25
5.4.1 Average Concentration and Concentration
Ranges in Bridesburg Area Versus Distance
Between Sites for TSP, IP and FP 5-11
5.5.1 a-b Seasonal Variation of Quarterly TSP, IP
and FP Averages(a) and Maxima(b) in 11
Eastern and Midwestern U.S. Urban Areas 5-14
6.1.1 Locations of SURE Sampling Sites 6-5
7.1.1 Patterns 1 (a), 2 (b) and 3 (c) of TSP
Frequency Distributions 7-4
7.2.1 a-e Cumulative Frequency Distributions of
HIVOL (a), SSI (b), TOTAL (c), FINE (d), and
COARSE (e) Concentrations for All
Measurements in the IP Network 7-7
7.2.2 a-d Examples of Cumulative Frequency Distributions
from Individual IP Network Sites Representing
Log-Normal (a), Pattern 1 (b), Pattern 2 (c)
and Pattern 3 (d) Distributions 7-10
7.4.1 Arithmetic Average Concentrations for Different
Sampling Intervals at 500 S. Broad, Allegheny,
NE Airport, Presbyterian Home and St. John's
for HIVOL (a), SSI (b), FINE (c) and TOTAL (d) 7-20
7.4.2 a-d Maximum Concentrations for Different Sampling
Intervals at 500 S. Broad, Allegheny,
NE Airport, Presbyterian Home, and
St. John's for HIVOL (a), SSI (b), FINE
(c) and TOTAL (d) 7-23
xvi
-------
LIST OF ILLUSTRATIONS (cont'd)
Figure Page
8.1.1 Fractions of a Typical Particle Size Distri-
bution Collected by Sampler 1 and Sampler 2 8-4
8.2.1 Frequency Distribution of Relative Differences
Between SSI Predicted and Measured Concentra-
tions Calculated from the Linear Regression
Slope Model for All Sites 8-13
8.2.2 a-c Scatterplot of SSI/HIVOL (a), TOTAL/HIVOL (b)
and FINE/HIVOL (c) for All IP Network Sites,
0 to 300 ug/m3 HIVOL Concentration
Range 8-15
8.2.3 Average SSI/HIVOL and TOTAL/HIVOL
Ratios at IP Network Sites 8-21
8.4.1 Histogram of Percent Differences Between
Predicted and Measured 24-hr SSI Concen-
trations for Highest Values at All IP
Network Sites 8-38
8.4.2 Histogram of Percent Differences Between
Predicted and Measured 24-hr TOTAL Con-
centrations for Highest Values at All IP
Network Sites 8-39
8.4.3 Histogram of Percent Differences Between
Predicted and Measured Arithmetic Averages
of SSI at all IP Network Sites with More
Than 20 Data Points 8-41
8.4.4 Histogram of Percent Differences Between
Predicted and Measured Arithmetic Averages
of TOTAL at All IP Network Sites with More
Than 20 Data Pairs 8-42
xvi i
-------
EXECUTIVE SUMMARY
The advent of a new size-specific standard for suspended
particulate matter mass concentrations has motivated measurement
programs to quantify the concentrations of fine (that portion of
ambient aerosol consisting of particles with aerodynamic diameters
less than approximately 2.5 urn) and inhalable (the portion consisting
of particles of aerodynamic diameters less than 15 urn) particulate
matter in urban and non-urban parts of the United States. The United
States Environmental Protection Agency (EPA)'s Inhalable Particulate
Network is the most comprehensive of these monitoring programs.
Recently, it has been recommended that a size-specific standard based
on the fraction of ambient aerosol with particle sizes less than
approximately 10 urn in aerodynamic diameter be considered, but no such
measurements are yet available.
The data from the EPA and other networks are interpreted in this
report and the extent to which they might apply to a 10 um standard is
evaluated. The topics which are addressed and the major conclusions
derived from them are presented in this executive summary.
Aerosol Sampler Collection Characterisitcs
The products of typical aerosol size distributions and collection
effectiveness curves of IP Network and other samplers, resulting from
wind tunnel study tests, were integrated over particle size to
calculate absolute collection efficiencies. The efficiencies of
different samplers were compared to each other and to an acceptable
performance range. Though dichotomous sampler collection
effectiveness curves do not fall within the acceptable performance
range, the collection efficiency does meet the collection efficiency
requirements of that range under typical wind speeds and particle size
distributions. HIVOL samplers with size-selective inlets satisfy both
collection effectiveness and collection efficiency requirements. The
efficiencies of samplers which would fall within EPA's proposed
acceptable performance window range from 47 to 71% for a typical urban
size distribution. A sampler with an effectiveness curve designed to
S-l
-------
meet the lower limit would sample 66% of the mass of a sampler
designed to correspond to the upper limit. Inlets with 10 um 50%
cut-sizes can be expected to collect between 80 and 90% of the mass
collected with the present dichotomous and HIVOL size-selective
inlets. HIVOL samplers with size-selective inlets and dichotomous
samplers should sample equivalent mass concentrations within 5% of
each other under typical situations. However, under low or high wind
speeds, in particle size distributions with much coarse material, or
when interferences are present in one or both of the samplers, this
equivalency degrades. Results of the collection efficiency model
agree with average ratios determined from ambient measurements. These
average ratios show substantial variability due to insufficient
knowledge of the ambient particle size distribution and wind speed
during ambient sampling.
The Inhalable Particulate Network Mea.su rement Process
The IP Network made mass, ion and elemental concentration
measurements of total (TSP), inhalable (IP) and fine (FP) suspended
particulate matter at urban and non-urban sampling sites throughout
the United States. The interpretation of these data cannot be
separated from the measurement process.
The study scale is urban-scale with some regional-scale and
neighborhood-scale sites. Most sampling site locations seem to be
representative of the study scales for which they were selected.
A change in the fibrous filter media for high-volume sampling
after the beginning of 1980 could result in differences between
sulfate, nitrate and mass measurements that are products of the
measurement process rather than an environmental cause.
The data validation procedure to which IP Network mass data are
submitted identifies many internal inconsistencies as well as extreme
cases; the total number of these cases amounts to a large portion of
the data base. For the purposes of this report, several of these
flagged values were deleted. Data validation procedures have been
applied only to mass measurements; (no validation flags appear on
reports of chemical composition) and should be extended to ion and
element measurements.
S-2
-------
Tests on high-volume samplers show that coppper emissions from
the motor can interfere with copper measurements from nearby samplers
and that deposition on the filter during standby periods can interfere
with mass measurements.
The comparison of collocated high-volume sampler measurements in
the IP Network is similar to comparisons in other networks showing
average differences between simultaneous samples of less than 5% of
the average mass concentration. Collocated dichotomous samplers have
been shown to attain average differences of less than 10% of the
average IP mass concentration.
Urban Areas in the Inhalable Particulate Network
The urban areas sampled by the IP Network represent a broad
coverage of population, meteorological, and emissions cases. Though
not all urban areas in the United States are represented, the ones
that were chosen include major population centers with varying
population densities. Major particulate matter sources have been
identified and located on maps with respect to IP Network sampling
sites in Birmingham, AL, Phoenix, AZ, Denver, CO, Buffalo, NY,
Philadelphia, PA, Houston, TX and El Paso, TX.
Several of the sites in the IP Network were found to be in
proximity to industrial sources which emit chemical species measured
on IP Network samples. These concentrations can be used to quantify
source contributions to receptors.
Geographica1 and Seasonal Vajiability of IP and FP _Ln Urban Areas
3
Annual arithmetic averages of IP and FP can exceed 90 ug/m and
o
35 ug/m , respectively, in urban areas, though the typical average
3 3
concentrations appear to be approximately 50 ug/m and 25 ug/m ,
respectively. Primary standards for annual IP averages in the 55 to
•j
120 ug/m range (Hileman, 1981) would find most, and possibly all,
of the sites examined in compliance. This range is tentative and may
have been modified subsequent to the writing of this report.
S-3
-------
Annual 24-hr maximum concentrations of IP and FP can exceed
200 ug/nr and 100 ug/m , respectively, in urban areas. Typical
3 3
values are approximately 100 ug/m for IP and 60 ug/m for FP.
Hileman (1981) cites a 24-hr maximum IP range of 150 to 350 ug/m-'
for a primary standard and a 24-hr maximum FP range of 70 to
o
220 ug/m for a secondary standard. Once again, most, if not all IP
Network sites would be in compliance with such a standard. The values
cited by Hileman (1981) are tentative and the form and values of a
24-hr standard may have been modified subsequent to the writing of
this report.
The urban and neighborhood-scale IP and FP measurements vary
significantly from site to site within the areas studied. The
implications are that (1) local (within a few kilometers of the
sampler) sources are significant contributors to IP concentrations and
(2) present spacing between sampling sites may be inadequate to fully
represent IP concentrations in certain urban areas and neighborhoods.
Seasonal averages of IP and FP tend to peak slightly in the
summer periods, but there are many individual exceptions at IP Network
sites. This conclusion is speculative because of the inadequate
seasonal data available for this report. If the conclusion is valid,
then sampling should take place on a yearly basis for the
determination of long-term averages so that all seasons are equally
weighted. A seasonally-weighted annual average might be a more
appropriate method of calculating the annual average if the number of
samples in one or more seasons differs by a large amount from the
number of samples taken in another season.
Geographical and Seasonal Variability £f I_P and F_P in Non-IJrban Areas
Non-urban average IP and FP concentrations are nominally 30 and
10 ug/m , respectively, in the western United. States and 30 and
o
20 ug/mj, respectively, in the eastern United States. The number of
non-urban sites with sufficient data in the West is small, however,
and these observations should carry less weight than those from the
East where independent measurements corroborate each other.
S-4
-------
There is a general seasonal variability of average IP and FP
concentrations in the eastern United States in which the spring and
summer month concentrations are elevated with respect to the fall and
winter months. Independent measurements from different networks
support this conclusion. Though the measurements in the western
United States do not show significant deviations between seasons, the
number of sites examined is too small to allow a general conclusion.
Up to 60% of the IP and 70% of the FP in urban areas can
typically be accounted for by concentrations present at nearby
non-urban sites. This portion varies substantially from site to site
and is based on a limited number of sites, primarily in the eastern
United States. Much of the non-urban concentrations may have
originated from far away anthropogenic sources.
Statistics o_f_ _IP_ and F_P Concentrations
One percent of the IP Network measurements exceeded 170 ug/m
2
for the high-volume size-selective inlet samples (SSI), 150 ug/m
3
for the dichotomous IP samples (TOTAL), 80 ug/m for FP and
3
90 ug/m for coarse particulate matter (CP) samples.
The cumulative frequency distributions at individual sites show
examples of the TSP patterns advanced by deNevers et al (1979) as well
as approximations to log-normal distributions. The small number of
measurements available at most sites prevents an assessment of the
adequacy of the log-normal and other statistical distributions in
representing IP and FP data.
The geometric and arithmetic averages of IP and FP data sets
normally differ by 5 to 15% with respect to the arithmetic average,
though the difference is as large as 37% for IP Network data. The
ratios of geometric to arithmetic averages which are calculated from
different subsets of all possible measurements at a site are constant
to within _+5% until the number of measurements in the subset falls
below five.
Arithmetic averages of IP and FP tend to remain constant, within
^10%, as the number of days between samples is increased. For 15 or
fewer samples this consistency breaks down.
S-5
-------
Maximum concentrations of IP and FP are extremely sensitive to
sample size and can be reduced by 50% as the number of days separating
samples is increased. This means that any standard, such as the
present one, which specifies a maximum concentration, should also
specify a sampling frequency, or that other forms of short-term
standards should be considered.
Predicting rP Concentrat ions from TSP Concentrations
The average ratio of IP to TSP derived from a number of different
data sets was used to predict IP concentrations from TSP
concentrations under a variety of conditions. The predicted values
were compared with measurements taken simultaneously with the TSP
measurements used in the model.
No simple relationship between FP and TSP was found which would
predict FP from TSP with adequate precision. For IP Network data
nearly 70% of the FP predictions differed from the measurements by
more than +20%.
The average ratio model derived from all IP Network measurements
is the simplest relationship between IP and TSP. For size-selective
inlet and dichotomous sampler IP measurements, stratification by site
type does not increase the precision of the model's predictions. The
model is independent of TSP concentrations for size-selective inlet
predictions, but it exhibits a TSP concentration dependency for TOTAL
predictions. Seventy-six percent of the size-selective inlet
measurements and 57% of the dichotomous sampler IP measurements were
predicted to within +20% by this model in the IP Network.
Models derived from simultaneous TSP and IP sampling at a single
site in an industrial neighborhood predicted IP concentrations from
TSP concentations at nearby sites in the neighborhood with a precision
which was comparable to the difference in IP measurements obtained
from nearby sampling sites.
Arithmetic average IP concentrations at IP Network sites were
predicted from TSP concentrations to within +_20% for all
size-selective inlet IP maxima and for 89% of the dichotomous sampler
IP maxima. This degree of uncertainty may be adequate for estimating
IP averages for compliance purposes in certain situations.
S-6
-------
Maximum IP concentrations at IP Network sites were predicted from
TSP concentrations to within +207, for 77% of the size-selective inlet
IP maxima and for 60% of the dichotomous sampler IP maxima. The
uncertainty associated with predicting maximum concentrations was
comparable to that associated with predicting any concentration,
irrespective of its magnitude.
IP and FP Composition and Sources
The chemical compositions of total, fine and coarse suspended
particulate matter concentrations from sites in Buffalo, NY, Houston,
TX, El Paso, TX, Philadelphia, PA and in the Bridesburg industrial
area of Philadelphia were studied to identify and quantify likely
source contributions to the fine and coarse particulate matter
fractions. Microscopic analysis of the coarse fractions of several
samples was performed to identify source contributions.
The identification of likely contributors based on the
geographical proximity of a major source to a receptor identified from
emission inventories and site survey was often confirmed by chemical
concentration measurements and receptor model source contribution
calculations.
The chemical element balance and microscopic properties balance
receptor models exhibited major limitations. Despite these limits it
appears that major contributors to IP Network sites in the eastern
United States for the periods under study were: unaccounted-for
sulfate (possibly from the conversion of SO ) and motor-vehicle
exhaust in the average fine particle fraction and geological material
and possibly biological material in the average coarse particle
Q
fraction. Industrial point sources showed small (less than 1 ug/m )
contributions to both size fractions at urban-scale sites in most
cases. Data were insufficient to apply these observations to western
sites.
In an industrial neighborhood, where a number of sources were
located in close proximity to each other and to the IP Network
samplers, the average source contributions to the coarse fraction
varied between sites even though the average inhalable particulate
S-7
-------
matter mass concentrations were roughly the same. A receptor model
approach to quantifying source contributions in such a neighborhood
may require more than one sampling site.
In general, the IP Network has accomplished its goals of
providing a data base from which certain observations can be made and
hypotheses can be formed. The data included in this report, however,
are too limited to draw conclusions and many of the interpretation
efforts made here should be applied again when the data base is more
complete. If some of the recommendations of this report concerning IP
Network sampling, analyses and reporting are put into effect, the
value of the data base to researchers will be increased substantially.
S-8
-------
CHAPTER 1
INTRODUCTION
The primary National Ambient Air Quality Standards for suspended
particulate matter were established with the Clean Air Act of 1970.
The sampling methods and values of these standards were based on the
best technical information available at that time concerning the
quantification of total suspended particulate matter (TSP, the mass
per unit volume of particles with aerodynamic diameters less than
approximately 30utn) concentrations (Federal Register, 1971) and the
effects of those concentrations on public health (U.S. HEW, 1969).
The state of knowledge concerning the ambient aerosol, its
sampling and its adverse effects has advanced considerably in the past
decade. The Clean Air Act Amendments of 1977 recognize the value of
this increased knowledge in the modification of present standards:
"Not later than December 31, 1980, and at five-year
intervals thereafter, the Administrator shall complete a
thorough review of the (air quality standard setting)
criteria. . .and the (existing) national ambient air quality
standards...and shall make such revisions in such criteria
and standards and promulgate such new standards as may be
appropriate..."(US House of Representatives, 1977, p. 7).
The National Ambient Air Quality Standards will be set based on a
thorough review of the property damage and adverse health effects
research. The first drafts of this review have been completed. The
most important implication of the studies included in this review is
that new standards for suspended particulate matter must be related to
particle size as well as to mass concentration.
Miller et al (1979), after surveying experiments measuring the
relative amounts of particulate matter deposited in different parts of
the body as a function of particle aerodynamic diameter, concluded
that:
'...there is no standard conducting airway deposition curve,
and hence, there appears to be no clear basis for
establishing a particle size range which is exclusively
restricted to the conducting airways."
1-1
-------
They do note, however, that less than 10% of the particles with
aerodynamic diameter greater than 15 um penetrate to trachea and state
that:
"...15 um would be a reasonable particle size cut-point to
include in the design of a sampler which would differentiate
particles deposited in the upper vs. lower respiratory
tract."
Particles in the 2 to 15 um size range tend to deposit, in the
conducting airways of the respiratory system, while the majority of
particles of aerodynamic diameter less than 2 um penetrate to the gas-
exchange areas of the lungs. Though they recognize that this
penetration is variable, Miller et al (1979) propose:
"...a cut-point anywhere between 2 and 3 um would reflect
particle deposition primarily associated with the
gas-exchange areas of the lung..."
The exact size ranges to be monitored and the maximum
concentrations to be allowed by a new set of standards have not yet
been specified. The Clean Air Scientific Advisory Committee of the
Science Advisory Board of U.S.EPA recommended in its meeting of
July 29, 1981 that a 10 um upper size limit be used in establishing
standards to protect public health. While this recommendation is not
binding, it provides a reason to more thoroughly evaluate aerosol
concentrations in the 0 to 10 um size range as well as those in the 0
to 15 um range.
Until the new standards are issued, the identifier applied to the
mass concentration in the 0 to 10 or 15 um aerodynamic diameter size
range is inna1able particulate matter (IP). The name given to mass
concentrations in the 0 to 2.5 um aerodynamic size range is fine
particulate matter (FP).
When a standard is promulgated, EPA's Office of Air Quality
Planning and Standards (OAQPS) must direct the state and local
agencies responsible for compliance monitoring to collect data
appropriate for judging whether or not an area is in attainment of the
standards. The states, in turn, must propose particulate matter
1-2
-------
emissions reductions which will bring the areas into attainment. When
the standards are promulgated, sampling equipment specifications,
sampler siting, and sampling frequencies must all be tailored to the
acquisition of data compatible .with those standards. Both areas in
attainment and in violation of the present standard must gain an
understanding of their status under the new standard. Areas with
approved control strategy implementation plans must predict the
efficacy of those plans for meeting the requirements of the new
standards.
To understand these regulatory and monitoring aspects of a
size-classified suspended particulate matter standard further, the
Environmental Protection Agency's Environmental Monitoring and Support
Laboratory has deployed an Inhalable Particulate Matter Sampling
Network in key regions of the United States (Rodes, 1979). Ninety-six
sampling stations came on line in 1979, with additional sites added in
1980 and 1981. This network was designed to sample IP in the 0 to
15 urn size range. Since its deployment, size-selective sampling
inlets have been developed and are being tested which will sample IP
in the 0 to 10 urn size range. When field tests are complete and
commercial manufacturers provide a sufficient number of these inlets,
IP data in the 0 to 10 urn size range will be acquired at existing IP
Network sites. In the interim, estimates of these concentrations will
have to be made from the 0 to 15 um data.
This Inhalable Particulate Network measures particulate matter
concentrations in three size ranges (0 to 30 um, 0 to 15 um and 0 to
2.5 um). Its coverage is meant to be representative of geographical,
climatological, emissions and population areas in the United States,
but it is not as extensive in space and time as the existing HIVOL
compliance network. Sampling frequency is every 3 or 6 days.
The data from this network include mass concentrations of total,
inhalable and fine suspended particulate matter (TSP, IP and FP,
respectively). Selected samples are submitted to chemical analysis to
determine elemental and ionic concentrations for TSP, IP, and FP
fractions. Several special sampling studies have been undertaken.
1-3
-------
The purpose of this study is to begin to answer a specific set of
questions about fine and inhalable suspended particulate matter.
These questions appear in Table 1.1; they are the questions most
commonly asked by those who deal with air quality standards. They
were posed by the network designers at EPA/OAQPS to guide the
establishment and operation of the network.
A three-phase approach was used in this study:
The first phase resulted in a review of relevant literature
related to the topic being addressed. This survey of past work was
meant to suggest technical approaches to the interpretation of IP
Network data, to evaluate the validity of those approaches, and to
provide perspective to the conclusions drawn.
The second phase stated certain hypotheses related to the
questions of Table 1.1 which could be confirmed or refuted using the
IP Network data.
The third phase devised a model consistent with the IP Network
data which supported or negated the hypotheses. If the data or the
model were insufficient, then an assessment of the additional data or
model development required was made.
It was not expected that all of these questions would be answered
in their entirety in this report, or that the data collected in the IP
Network would ultimately be sufficient to deal with them. Their
statement here, however, sets several objectives to be accomplished.
Where IP Network measurements were found insufficient to answer the
questions, measurements from other studies were examined or new
measurements were suggested which might reach those objectives in the
future.
This report consists of ten chapters. The first, this
introduction, explains the purposes and objectives of the study, the
generalized technical approach, and previews what is to come in
subsequent chapters. Chapters 2 through 4 are dedicated to the tenet
that the interpretation of data should be done with an understanding
of the process of acquiring those data. Environmental measurements
1-4
-------
TABLE 1.1
QUESTIONS CONCERNING INHALABLE PARTICIPATE MATTER
SOURCES OF IP
-What are the probable sources of inhaled particulate matter
(IP) in urban areas; in rural areas?
-What are the sources and the relative impact of particulate
matter from: (a) exhaust from mobile sources (gasoline
versus diesel), (b) stationary sources, (non-ducted versus
ducted), (c) secondary particles (sulfates, nitrates,
organics), and (d) fugitive dust sources such as
re-entrained dust from paved versus unpaved roads on ambient
inhaled particle levels?
-What are regional differences in the sources of IP?
MASS, CHEMICAL, ELEMENTAL AND SIZE CHARACTERIZATION
-What are the differences and similarities in mass, size,
elemental, and chemical composition concentrations from
urban area to urban area, urban area to rural area,
industrial city to non-industrial city, eastern versus
western urban area, and heavily populated versus lightly
populated areas?
-What is the range and average of mass, size, elemental and
chemical composition in these areas? How do mass
concentrations of IP and FP relate to the TSP levels in
these areas (ratio of FP and IP to TSP on both daily and
annual basis)?
-How do data from HIVOL, dichotomous sampler and
size-selective HIVOL compare?
-What are the causes of variations in the ratio of IP to TSP?
SPATIAL PATTERNS
-For an urban area, what is the spatial distribution of
ambient IP concentrations? Are concentrations relatively
uniform across an area (indicating uniform source
contributions) or are there hot spots (indicating local
source contributions)?
-What are variations in IP concentrations among site types;
by land use?
-What is the vertical and horizontal distribution of IP near
sources?
1-5
-------
TABLE 1.1 (Continued)
-Are the gradients of IP significant with respect to
instrument siting and control strategy development?
-What is the scale of representativeness of IP monitors and
how many monitors would be needed for urban areas of various
sizes?
TEMPORAL PATTERNS
-What is the diurnal variation, weekday to weekend variation
and seasonal variation of IP concentrations within an urban
area?
-How does the chemical composition of IP change with each of
these time periods?
TRANSPORT/TRANSFORMATION/BACKGROUND
-Of the inhaled particulate matter measured in urban areas,
how much is locally generated and how much is transported
into urban areas from upwind sources?
-How much of this transported particulate matter is sulfate?
-What are the probable sources of sulfate particles in urban
areas?
-How much sulfate is emitted directly as sulfate (primary
sulfate) and what are the major sources of primary sulfate
(e.g., oil fired power plants)?
-What is the origin of the sulfate being transported into
the urban area?
-What is the scale of transport and where are the major
concentrations (clusters) of the precursors Located
geographically?
-What is the impact of the Ohio River Valley on eastern U.S.
IP levels?
-What are the particle removal processes and how do IP
concentrations decrease as distance from the urban areas
increases?
-What is the background level of IP on both a mass and
chemical composition basis?
-What is the source of the material?
-How do background levels differ in various parts of the
Nation?
1-6
-------
tend to lose their identities once they enter a large computerized
data bank. They seldom carry confidence intervals or information
pertaining to their sampling and analysis.
Chapter 2 contains a detailed description of the aerosol sampler
characteristics. Since any new particulate matter standard is likely
to be size-specific, it is imperative that the particle size
collection characteristics of aerosol samplers used to collect TSP, IP
and FP are known. Wind tunnel tests have provided estimates of these
characteristics and these tests show the variation of the particle
size collection effectiveness of dichotomous samplers with wind
speed. Integration of the product of collection effectiveness and
typical concentrations in different size fractions of ambient aerosol
allows the magnitude of this variation to be estimated. The
comparability of 0 to 15 urn and 0 to 10 urn measurements is also
evaluated by this method.
Chapter 3 presents the IP Network sampling process in detail,
geographically locating the areas sampled and, to the extent which
information is available, describing the sampling sites. Filter
media, field sampling, filter handling and processing, chemical
analysis, and data validation procedures are also summarized.
Reproducibility of measurements via collocated sampling provides the
best estimate of the precision to be attached to each value in the
data base. An estimate of this precision is important to the
comparison of one data value with another; if the confidence intervals
around two measurements overlap, then there is no basis for finding
any difference between them. This estimate is made in Chapter 3.
Most of the IP Network sampling sites are located in urban
areas. To answer the questions in Table 1.1 about the sources of
inhalable particulate matter and the effects of air movement, the
types of sources in these urban areas and their locations with respect
to the sampling sites should be specified. Eventually, all urban
areas sampled by the IP Network need to identify the particulate
matter sources and their locations if the sources of IP and FP
concentrations are to be quantified. In Chapter 4 of this report, the
character of the industrial sources and their relative location with
respect to the IP sampling sites for some of those urban areas have
1-7
-------
been compiled. Average meteorological values related to air pollution
have been tabulated. Previous studies of the source contributions to
the different urban areas have been reviewed and summarized. This
documentation will be referred to in later chapters to explain the
results of ambient sampling.
Chapter 5 examines the spatial variability of size-classified
aerosol concentrations with an eye toward relating those
concentrations to sources and also to estimate the density of sampling
sites required to represent a single land use area with one sampler.
The non-urban concentrations of size-classified suspended
particulate matter is studied in Chapter 6. Data from the Sulfate
Regional Experiment (SURE, Mueller and Hidy et al, 1981) and from
non-urban sampling sites in the IP Network are compared to the
concentrations in urban areas. The non-urban concentrations provide
upper limits of aerosol mass transported into the urban areas.
The statistical distribution from which pollutant concentrations
are drawn is an important consideration in estimating statistics from
subsets of all possible data. Log-normal distributions have been used
to describe air pollution data in the past. Several patterns of
deviations from these distributions have been observed in the analysis
of TSP concentrations from HIVOL samplers. In Chapter 7, these
patterns are sought in the size-classified data from the IP Network
and the validity of a log-normal distribution assumption to describe
the data is evaluated. Geometric and arithmetic means are compared as
reasonable statistics against which to evaluate a standard.
Until a more extensive network of size-classified monitoring
sites is deployed, it will be necessary for many communities to
estimate their IP concentrations from existing HIVOL TSP
measurements. These estimates cannot be used to determine compliance
with a standard — only actual measurements can do that. These
estimates would serve as guidance in network design and short-term
planning. Chapter 8 derives a receptor-oriented model for relating IP
to TSP concentrations and estimates the accuracy with which IP
predictions can be made from HIVOL measurements. The usefulness of
this model is evaluated against data from the Inhalable Particulate
Network.
1-8
-------
The major sources of IP and FP are not necessarily the same as
the major sources of TSP. Yet the identification of IP and FP sources
will become paramount as new control strategies are developed to meet
new standards. In Chapter 9 the results of the previous chapters are
combined with the results of chemical and microscopical analyses of
special-study samples in a receptor-oriented model to quantify within
specified confidence intervals the types of sources contributing to
ambient size-classified aerosol concentrations.
Finally, Chapter 10 summarizes the work of all chapters. The
limitations of the present data set with respect to answering the
questions in Table 1.1 are pointed out. The requirements of future
sampling and analyses necessary for obtaining more definitive answers
are stated.
It is important to reiterate that while this report addresses the
questions of Table 1.1 within the constraints of data from the
Inhalable Particulate Network, it does not answer them all. The cost
of designing experiments to address all of them was prohibitive.
Difficulties in network start-up caused many size-classified samples
to be missing or invalidated. Delays in data processing and analysis,
and the time constraints imposed on the data interpretation phase of
this project meant that much less than one year of data were reviewed
from many sites. The majority of the interpretive efforts were
dedicated to measurements acquired between May, 1979 and June, 1980.
For some applications summary statistics from October, 1979 to
September, 1980 were used. Size classifying inlets for 0 to 10 um
monitoring were not available. The chosen sampler locations, sampling
frequency and sample duration limited the information available for
answering questions about spatial and temporal distributions. Many of
the conclusions drawn from the technical approaches taken here are
limited by this lack of completeness. As new and more complete data
become available from the IP Network and other monitoring programs,
they should be used to test the conclusions drawn in this report and
to advance new ones. The hypotheses presented herein offer
opportunities for some fascinating speculations about the
distribution, transport and sources of suspended particulate matter in
various size ranges in the United States. They remain to be tested,
modified and retested to establish their veracity.
1-9
-------
CHAPTER 2
AEROSOL SAMPLER
COLLECTION CHARACTERISTICS
The purpose of the IP Sampling Network is to create a data base
of measurements as a function of particle size. The definition of the
particle sizes collected depends on the instruments used for their
collection. In this chapter the instruments used to obtain the
samples from which the data in this study were derived are described.
The particle size collection characteristics of different samplers
which are intended to sample the same and different mass fractions of
the total ambient aerosol are studied and compared. The results of
the theoretical treatment are compared with a number of actual ambient
measurements to evaluate their validity.
2.1 IP Network Sampler Configuration and Collection Effectiveness
Each sampling configuration was designed to include a standard
11 1/2" x 15" high volume (HIVOL) sampler, a HIVOL equipped with a
15 urn size-selective inlet (SSI) and flow controller, and a
Sierra 244, 244E or Beckman SAMPLAIR virtual impactor (also termed
dichotomous sampler). Wedding (1980b) provides good illustrations of
the SSI, Sierra and Beckman inlet constructions, while Dzubay and
Stevens (1975) explain the operation of virtual impactors. All sites
were equipped with HIVOL samplers. Due to technical and procurement
difficulties, some sites possessed a HIVOL(SSI) without a dichotomous
sampler, others had a dichotomous sampler without a HIVOL(SSI), while
other sites had neither. The precise configuration at a given site is
specified in Table A.I of Appendix A as explained in Chapter 3. Where
the lack of measurements from a missing sampler caused a significant
effect on the interpretation process, that effect will be noted.
The sampling configuration was intended to measure Total,
Inhalable, and Fine Suspended Particulate Matter concentrations. The
definitions of Total, Inhalable and Fine Suspended Particulate Matter
(Lioy et al, 1980) are:
2-1
-------
• Total Suspended Particulate Matter (TSP): TSP is that
portion of suspended material which is normally collected by
a high-volume filter sampler for 24-hr and has 50% cut-size
(the particle aerodynamic diameter at which one half of the
particles penetrates the inlet and one half does not)
ranging from 30 to 65 urn for wind speeds between 2 to
24 km/hr. (McFarland et al, 1979)
• USEPA "Inhalable" Particulate Matter (IP): Miller et al
(1979) proposed as "inhalable" dust that portion collected
by a sampler with a 50% cut-size of 15 urn. A. similar
definition would be the Thoracic Fraction (TF) since these
are the particles that can enter the trachea and lungs and
contribute toward the production of health effects generally
associated with ambient particulate pollution, such as
bronchial cancer, bronchitis and emphysema (Lippmann,
1980). More recent proposals impose a 50% cut-size of 10 urn.
• Fine Particulate Matter (FP): FP is that portion of an
aerosol which penetrates a particle collector with a 50%
cut-size of 2.5 urn. Depending on the penetration curve used
for comparison (American Conference of Governmental
Industrial Hygienists (ACGIH) or the British Medical
Research Commission (BMRC)), FP could be nearly equal to,
but generally smaller than the concentration known as the
"respirable" fraction, which is defined by ACGIH and BMRC as
the fraction which penetrates through the conductive airways
of the lower respiratory tract (tracheobronchial tree) of
healthy adults and is available for deposition in the
nonciliated (alveolar) zone of the lung (Miller et al, 1979).
There is no sampling instrument which can collect all particles
less than and no particles greater than a certain aerodynamic
diameter. Only recently have standardized wind-tunnel methods been
developed (Ortiz, 1978; McFarland et al, 1979b; Wedding et al, 1977)
to measure the particle size collection characteristics of different
aerosol sampling inlets under variable wind speed conditions. An
understanding of these characteristics is central to the comparison
and interpretation of inhalable particulate matter measurements.
The collection effectiveness curves of the sampling devices used
in the IP Network are presented in Figures 2.1.1 and 2.1.2. Included
in each diagram is the acceptable performance range suggested by EPA
and reported by Ranade and Kashdan (1979) for sampling inhalable
particulate matter.
Figure 2.1.1 also includes a collection effectiveness curve with
a 50% cut-size of 10 urn. This curve was obtained by shifting the
2-2
-------
x
o
c
tu
c
o
•H
+->
o
0)
o
u
0)
u
100
90'
80-
70-
60-
50-
40-
30-
20-
10-
0
A:Hypothetical 10
B:Step Function 10 ym
C:HIVOL(SSI)
D:HIVOL 24km/hr
E.HIVOL 8km/hr
F; HIVOL 2km/hr
'^Acceptable
Performance
-Range, RTI
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
ln(D)
2 3 4 5 6 7 8 9 10 15 20 30 40 50 eV 70
Aerodynamic Particle Diametei (D), ym
Figure 2.1.1 Collection Effectiveness of Standard HIVOL and HIVOL
Size-Selective Inlets Under Various Wind Speeds
(McFarland et al, 1979b). Logarithms of particle
diameters are placed on the abscissa to facilitate
reading of collection effectiveness as a function
of In D.
2-3
-------
X
o
c
o
s
UJ
o
•H
+J
U
CL>
tH
r-l
O
u
0
1—I
•J
•H
•p
oj
CX
100-
90
? 80-
70'
60-
50-
40-
30-
20'
10'
flti ill Range, RTI
2.0 2.5 3.0 3.5 4.0
A : Beckman FINE
B '• Sierra 40km/hr
C • Sierra 15km/hr
D • Beckman 2km/hr
E • Sierra 5km/hr
0
0.5
1.0
1.5
T 3 i+ 5 6 7 8 9 10 15 20 3D 40 50 60
Aerodynamic Particle Diameter (p). ym
Figure 2.1.2 Collection Effectiveness of Beckman (MacFarland et al, 1979a)
and Sierra 244 (Wedding et al, 1980a) Dichotomous Sampler Inlets,
The Beckman SAMPLAIR collection effectiveness also varies with
wind speed but only the 2km/hr curve was available.
2-4
-------
15 urn (HTVOL)SSI curve. The slope of the 10 urn curve as defined by
the parameter, s (McFarland et al, 1980b), is 1.35.
".16
where
D.16 = particle size for which collection effectiveness is 16%.
^ QA = particle size for which collection effectiveness is 84%.
For comparison purposes, a perfect 10 urn collection effectiveness
curve with a slope equal to 1 is considered in Figure 2.1.1. The
first item of note is that the collection effectiveness of the
dichotomous sampler inlet and of the HIVOL sampler inlet are dependent
on wind speed for aerodynamic particle diameters greater than 7 um.
The tests done by Wedding et al (I980a) on the inlet to the Sierra 244
at 5 km/hr wind speeds and McFarland et al (1979a) on the Beckman
SAMPLAIR inlet at 2 km/hr wind speeds are significantly different even
though the designs of these inlets are similar. Simultaneous ambient
sampling under the same wind speed conditions shows the measured mass
concentrations collected through the two inlets to be different
(Grantz, 1981). McFarland et al (I979a) measured 50% cut-sizes of
15.5, 13 and 10 um under wind speeds of 2, 8 and 24 km/hr,
respectively, for the Beckman inlet while Wedding et al (1980a)
measured 50% cut-sizes of 22, 15, and 9.5 um under wind speeds of
5 km/hr, 15 km/hr and 40 km/hr for the Sierra 244 inlet. These two
tests show significant differences between collection effectiveness of
dichotomous samplers under various wind speeds and between each other
at similar wind speeds.
Though the methodologies of the Wedding and McFarland tests are
essentially the same, they have not been demonstrated to be
equivalent. A method of standardizing these tests needs to be
developed. Wedding must test the Beckman inlet, McFarland must test
the Sierra inlet and more simultaneous samples must be taken to
resolve the discrepancy. This type of interlaboratory testing and
simultaneous sampling should be carried out on all inlets which will
2-5
-------
be used in large scale sampling programs. The other possibility is
that there are significant differences in the particle size collection
characteristics of the Beckman and Sierra inlets.
McFarland et al (1979b) tests of the HIVOL collection
characteristics show not only variability with wind speed; collection
effectiveness also varies with wind direction because of the
instrument's asymmetric design. At 20 km/hr wind speeds, the
collection effectiveness of 23.5 um particles varies from 90% when the
wind vector is perpendicular to one of the flat sides of the sampler,
to 57% when the wind vector points to the corner of the HIVOL inlet.
The wind tunnel tests of the circular HIVOL size-selective inlet
show its collection effectiveness to be virtually independent of wind
speed. McFarland et al (1979b) measured a 15 um 50% cut-size under
2 km/hr, 8 km/hr and 24 km/hr wind speeds. Wedding (1980b) found 50%
cut-sizes for the HIVOL(SSl) of 13.4, 14.4, and 12.5 um at the same
respective wind speeds which differs from the McFarland tests to a
lesser degree than that exhibited between the dichotomous sampler
tests. This agreement for the HIVOL(SSl) inlet lends credence to the
equivalency of the two wind tunnel tests and suggests that there may
be significant differences between the Beckman and Sierra particle
collection characteristics.
Because of the difference in collection effectiveness curves, IP
data reporting should also include the sampler used to collect the
measurements and typical wind speeds under sampling conditions. In
this report, the operational definitions of aerosol samples will be:
HIVOL: TSP collected by the HIVOL sampler irrespective of
wind speed. Nominally 0 to 30 um size range.
SSI: IP collected by the HIVOL equipped with a
size-selective inlet. Nominally 0 to 15 um size
range.
TOTAL: IP expressed as the sum of FINE and COARSE
concentrations collected with a Sierra 244 or
Beckman SAMPLAIR dichotomous virtual impactor
sampler, irrespective of wind speed. Nominally 0
to 15 um size range. The wind tunnel tests suggest
that Beckman and Sierra samples should receive
different designations. The EPA data base does not
provide this information.
2-6
-------
FINE: FP collected by the fine stage of the dichotomous
virtual impactor sampler and corrected for fine
particulate matter which ends up on the COARSE
filter. Nominally 0 to 2.5 um size range.
COARSE: Coarse Particulate matter (CP) collected by the
coarse stage of the dichotomous virtual impactor
sampler and corrected for fine particulate matter,
irrespective of wind speed. Nominally 2.5 to 15 um
size range.
2.2 Collection Efficiencies of IP Network Samplers
Given the results of these wind tunnel tests, there is no doubt
that different samplers under different ambient conditions measure
different fractions of the total suspended particulate matter. The
question of how much difference should be expected between
simultaneous measurements with different instruments has been examined
as follows.
The mass of particulate matter collected by a sampler in a
particle size range is equal to the mass of the particulate matter in
the air in that size range times the average collection effectiveness
of the sampler over that range. If the ambient mass concentrations
are distributed as a function F(D) of aerodynamic particle diameter,
D, and the collection effectiveness of the sampler as a function of
particle size is E(D) then the mass concentration, C, measured by the
sampler is
/* °°
C = / F(D) E(D) dD 2-1
The particle size distribution, F, is also a function of space
and time while E is also a function of wind speed. For the purposes
of this discussion, these variables will be held constant for
calculating values of C.
The effectiveness of the IP Monitoring Network inlets, E(D), are
presented in Figures 2.1.1 and 2.1.2. It remains to choose an
appropriate particle size distribution function, F(D).
2-7
-------
Atmospheric aerosols have been observed to have a bimodal mass
distribution (Junge, 1963; Whitby, 1978; Whitby and Sverdrup, 1980;
Brock, 1973; Lundgren and Paulus, 1975) with a minimum in the
neighborhood of 2 urn. Those particles less than 2 urn in aerodynamic
diameter can be broadly classifed as fine and those with diameters
greater than 2 urn are termed coarse. There is further evideirice of
these modes being log-normal and additive. Fine aerosols are
generated primarily by condensation while coarse aerosols are produced
for the most part by mechanical processes.
Whitby and Sverdrup (1980) fitted three additive log-normal
functions to atmospheric aerosol measurements made with optical
particle counters. These fits of additive log-normal distributions to
ambient measurement were found to be applicable under a variety of
circumstances as evidenced by low reduced chi-squares statistics.
Whitby and Sverdrup (1980) formulated seven different categories into
which their hundreds of fitted distributions fell. In all of these
categories, more than 90% of the aerosol volume was concentrated in
the superposition of two log-normally shaped modes. Lundgren and
Paulus (1975) also found atmospheric aerosols to have bimodal mass
distributions which could be fitted by two additive log-normal
distributions.
The fine and coarse modes have been observed to have geometric
mean diameters in the range of .2 to .7 um and 3 to 30 um,
respectively. The geometric mean diameter and geometric standard
deviation for selected categories are listed in Table 2.2.1.
The atmospheric aerosol mass distribution, assuming it to be
additive bimodal log-normal, may be represented by
D - In Df )2V
where
P = the fractional mass in the fine mode and 1-P is the
fractional mass in the coarse mode
D = particle aerodynamic diameter
2-8
-------
TABLE 2.2.1
GEOMETRIC MEAN AERODYNAMIC DIAMETERS
AND GEOMETRIC STANDARD DEVIATIONS FOR
SELECTED ATMOSPHERIC PARTICLE SIZE DISTRIBUTIONS
Fine Mode
Coarse Mode
Aerosol
Classification
Averagea
Urban
Averageb
Background
and Aged Urban
Plumea
Marinea
Geometric
Mean
Aerodynamic
Diameter
Df
(urn)
.38
.50
.47
.39
Geometric
Standard
Deviation
Geometric
Mean
Aerodynamic
Diameter
cr.
2.02
5.0
1.84
2.0
Dc
(urn)
10.20
20.00
7.27
19.30
Geometric
Standard
Deviation
cr
2.26
2.00
2.12
2.70
aWhitby and Sverdrup (1980). Optical diameters, DQ, reported in this
reference have been converted to aerodynamic diameters, Da> by D =
VP" x DO where a density, p , of 1.7 gm/cm^ has been used for the
fine mode and 2.6 gm/cm^ has been used for the coarse mode. While
this approximation may not be true in all cases, it is legitimate for
making comparisons between the collection efficiencies of different
samplers under a variety of circumstances.
bLundgren and Paulus (1975).
2-9
-------
Df & Dc = the geometric mean aerodynamic diameters
for the fine and coarse modes, respectively
°f & °c = t'ie geometric standard deviations for
the fine and coarse modes, respectively.
The parameters P, Df, D , CTf and a in Table 2.2.1 are fitted by
minimizing the sums of the squares of the differences between F(D),
integrated over the same size range as an ambient measurement, and the
fraction of the total mass measured in that size range. Equation 2-2,
with appropriate parameters is an appropriate F(D) to be used in
Equation 2-1.
The integral of Equation 2-1 was calculated by dividing the
interval from 0 to 100 um into regions over which the natural
logarithm of the particle diameter was equal to .2 . For the ith
interval
Ci = F(Di)E(Di) Aln D 2-3
= .2 F(Di)E(Di)
The particle diameter at which F(D) was evaluated via
Equation 2-2 was chosen as the center of the interval. E(D) was read
from the curves of Figures 2.1.1 and 2.1.2 at the same value of In D.
The summation of fractional mass concentrations, ZC. over all
intervals between 0 and 100 um gives the fraction of the total mass
collected by the inlet under consideration. The maximum error
introduced by numerical approximation over the intervals was estimated
to be less than 3% by dividing a few intervals with the most rapid
change of F(D)E(D) into smaller intervals and comparing the integrals
with those obtained for the larger interval.
Other sources of errors in this treatment result from the
simplification with respect to real aerosol size distributions
inherent in equation 2-1, measurement uncertainties of the collection
effectiveness curves, and the lack of consideration of mass
measurement interferences (two such interferences, artifact formation
2-10
-------
and passive deposition, are identified in Chapter 3 as agents which
could affect the mass measured by IP Network samplers). For the
present analysis these errors are noted, but not quantified.
Limitations imposed on the conclusions because of these shortcomings
will be noted where appropriate.
The collection efficiency of the sampler (i.e. mass collected by
inlet/total suspended mass) for the size distributions specified by
the parameters in Table 2.2.1 and the collection effectiveness curves
in Figures 2.1.1 and 2.1.2 were calculated and are presented in
Table 2.2.2. The fine particle fraction, P, of the total mass was set
equal to .33 and .5 for these calculations. This study and others
show this to represent a reasonable ambient range.
A close examination of the results in Table 2.2.2 provides some
important insights into the variability of the different IP Network
sampler inlets, and two hypothetical 10 urn inlets, with respect to
particle size distributions, wind speeds, and the proposed acceptable
performance range for 15 urn inlets.
The collection efficiencies of all inlets under all wind speed
conditions vary with particle size distribution. The efficiencies
increase as the total aerosol mass is shifted toward the smaller
particles because the collection effectiveness increases as particle
size decreases. The lowest efficiencies are obtained for the urban
average and marine distributions, which Table 2.2.1 shows to have the
largest geometric mean aerodynamic particle diameters (approximately
20 urn) in the coarse mode. At average wind speeds of 8 to 15 km/hr
expected at IP Network sites, TOTAL collection efficiencies range from
.52 to .87, SSI efficiencies range from .55 to .89, and HIVOL
efficiencies range from .76 to .96 for the size distributions
considered. The size distribution which is the most representative of
the majority of IP Network sites is probably the urban average
(Lundgren and Paulus, 1975) with P = .33. For this distribution, the
TOTAL, SSI and HIVOL collection efficiencies at typical wind speeds
are .52, .55 and .76, respectively.
The variation of sampler collection efficiency with wind speed is
most pronounced for the TOTAL and HIVOL samplers. For the urban
average size distribution (P = .33) the collection efficiency of the
2-11
-------
CN
CN
CN
u
pa
H
f.
PS
u
b
Q
1
CO
u
u
Z
u
M
u
M
U,
CL.
U
Z
0
H
O
s
is
-3
a
w
t-
w
a:
o
u
s
t-
z
2
E-
a:
8
CO
w
1
M
B
IK
b
Z
0
M
13
3
8
H
Z
Qtf
U
b
b
a
a
•"£
2
H
3
Z
t— i
C*
[xJ
£
i
CO
o
en
o
5
at
j_
Q,
C
£
at
Q
'
*E<
CJ
u
.u
a
A.
O
o
>
X
"3
g
5
c
(J
CO
a!
u
j
PM
X
<
CO
g
M
X
CO
C£
-J
0-
5
<;
CO
(9
Ul
CO
a
o
t—i
L >
C
QJ
a.
o
^_<
o
*-*
n
at
a
o
g CO
3 ^
o
^
in
T— (
01
a.
o
V)
CO
CO
•o
0)
at
CL u
CO -C
CN -^,
c *.
3
u
si
*" M
•o
01 U
aoo -^.
CO B
j<:
•o
c
h
CNJ
_§
£
"* 1
OJ M
0) £
0.»n ^-
01 l-H J
•c
e
3
£
10 ~B
c
** b
C9
C
O
0 B
tfl U
O -J
^ w
ca
IB
t-4
0
o\^-i ^HOO mso mo
COO*N r^r^- CTNCTN r--C30
vOf~ -a-u^ r~-oo mvc
SS SS 5iC 5°
vo OO
in vD ON CN r**« 00 ON CN
m CN CN o co o *-^ en
\DP^ <3-m r**oo m^o
m o CN en m r** ON CTN
r-- oo in NO oo oo in sC>
moo r* m CN-^ or-^
oooo \Q r** O>ON r^r-^
en o en o en o en o
c
CO
u ,a
01 t-
00 S
1- "O -O
QJ C 0V
,O > 3 ckO^Q
a> nj o < at .c
« C 60 -O 3 C
a> .0 u « P- i-
> Lj Cfl (Q
< n BS z
i
Ifl
o
>J
01
01
•o
G
01
a
01
•c
c
01
01
to
* c
o
C -H
01 4J
v< u
00 01
•O —I
C -H
3 O
J U
O t)
c
to
o:
a
3
•o
c
o
iu a.
to t^
O IB
U *
01
2-12
-------
TOTAL sampler decreases from .67 to .42 as wind speeds rise from
5 km/hr to 40 km/hr. These extremes represent variations of +29% and
-19% with respect to the collection efficiency under typical
(~15 km/hr) wind speed conditions. For the HIVOL sampling the same
size distributions, efficiencies decrease from .89 to .70 as wind
speeds increase from 2 km/hr to 24 km/hr. These extremes constitute
+17% and -8% deviations with respect to typical wind speed conditions.
The upper and lower collection efficiencies of the proposed
acceptable performance range (Ranade and Kashdan, 1979) for 15 urn
inlets offer a considerable range for possible collection
efficiencies. The acceptable performance range exhibits +29% and -15%
deviations with respect to the SSI collection effectiveness curve,
which falls in the center of the range (see Figure 2.1.1).
The dichotomous sampler (both Sierra and Beckman) and the SSI
collection efficiencies fall within the efficiency limits of the
acceptable performance range regardless of the size distribution
sampled and the wind speed with the exception of the Sierra sampler
under 40 km/hr wind speeds. This is in contrast to their collection
effectiveness curves illustrated in Figures 2.1.1 and 2.1.2 for which
only the SSI sampler meets the criteria.
HIVOL collection efficiencies are marginally within those
corresponding to the upper limit of the acceptable performance range
under 24 km/hr wind speed conditions, but they exceed this limit for
all size distributions in 2 km/hr and 8 km/hr winds.
Both of the hypothetical 10 um inlets exhibit collection
efficiencies less than that of the lower boundary of the acceptable
performance range for 15 um inlets.
The predicted collection efficiencies for the HIVOL (ranging from
70 to 96%) are generally lower than the 97% +_ 3% observed by Lundgren
and Paulus (1975). The apparent reduction in overall collection
efficiency could be due to:
• Mass measurement interferences (e.g. passive deposition and
artifact formation) not included in the model
• The error introduced by the bimodal log-normality assumption
2-13
-------
• Errors due to the extension of the HIVOL collection
effectiveness curve beyond the experimental data points
• The HIVOL collection effectiveness curve generated with the
use of uniform oil droplets (McFarland et al, 1980) not
being quite the same as that of the airborne particles
sampled by Lundgren and Paulus (1975)
No data on the absolute collection efficiencies of the
dichotomous samplers and the HIVOL size selective inlets under ambient
conditions have yet been acquired.
2.3 Relative Collection Efficiencies of Different Sampler Inlets
Each of the IP sampler collection effectiveness curves is
distinct from every other, but the collection efficiencies listed in
Table 2.2.2 are much more similar than would be initially suspected by
comparing the effectiveness curves. This is so because typically a
third to half of the mass in the aerosol size distributions examined
is in the size region where the collection effectiveness of the
samplers is close to 100%.
Three inter-sampler collection efficiency comparisons are
relevant to the interpretation of IP Network data:
1. IP inlets (both dichotomous sampler and HIVOL size-selective
inlets) to the standard HIVOL inlet
2. Dichotomous sampler inlets to the HIVOL size-selective inlet
3. The hypothetical 10 urn inlets to dichotomous sampler and
HIVOL size-selective inlets
The first comparison is important because of the plethora of
HIVOL measurements available and the desire to develop a predictive
model allowing IP concentrations to be predicted from TSP data. Such
a model is formulated in Chapter 8 of this report. The fraction of
measured TSP which would be measured as IP by the network samplers
under various wind speeds and particle size distributions can be
estimated by this comparison and is given in Table 2.3.1.
For the SSI measurements, which show no wind speed dependence,
this ratio increases with increasing wind speeds. In contrast, for
the TOTAL measurements with the Sierra dichotomous sampler, the ratios
2-14
-------
CO
CN
S
CO
e S
en _
tzJ Z
£ o
ll
O 3
I—I fa
fa •*
w «
z <
2 CO
f- u
« 3
-> u
8 S
s s
Q,
•? z
< °
w H
* §
. nJ
o 8
RATIOS
HIVOL
j
o
1-1
o
1 o
t_i 01
ET CL
« S
H-f 1*
as &
H _
co fj
S 3
S
M ^
co •£
M C
3 «
~ 1
£ £
1 s
Q C
i o
<= S
w *J
s -
CO °
Q
Z rj
i— i I"H
3 1
JJ
«
,__i
u
oi
,J
h-4
S
|
g
J
X
H
'HIVOL
M
CO
CO
Ua
O
CO
en O
01 ••*
C eg
a &.
c£
13
x:
•— ^
E
0
13
01 U-
0) 1-4
<-s a. x;
eg eo — .
^ E
lj "^ SJ
oi c
co 3 1-4
^
o
•--
c
u^
TJ
^-s 0)
c oi u
eg a. x:
E en -^.
r i *^1 *S
w Ta -X
01 C
03 "^ CN
•^ 3
^4
x;
"E
-3-
CM
T3
OJ 1-
oi x:
a.--^
en E
T3
C 00
.,_!
3
Ui
J3
•"NS^
E
>>
CN
01
8-S C
•*a
fa
X)
C
o
• ia
ia
-4 tg
O 0
co >>a
O u-i
M *i-i
01 CO
< CO
tg
0
^ sO 00 ^H 00 OO ^ (*O
^^ ^\ 00 CT^ ^^ ^"^ O^ CTN
OO OO OO OO
•s^ r~4 ^5 ^^ ^5 ^O ^"Sl f^*1
r**co ^OP^. oooo ^Or*.
^^ r^ ^5 ro ^3 ^O ^^ C^
r^oo ^Opv. oOoo r-»oo
r^j ^k oo r*> oo ^^ f*** rn
GOOO Or^ coo> r-»co
\O^ 00'—' OOOO ^^OO
v\ O^* ^ ^ VN O\ Q^ O\
ooco 00 3 -^
Ol la CO cfl
J O
0 O >
iJ J-l > 1— (
to 01 ^— i x
=8 X
U) la
o) la IH x;
• -I 0 X! ->.
U Ua -^ £
C E X
01 —4 ^
O CN 00 CN
•H •
U-l CN O O
U-l ia u
01 01
i— I T3 TJ
l-i XI Ol Ol
Ol eg )u la
r— i H tg cs
a. a. a.
E u E E
tg 01 o o
CO CO O CJ
to XI O 13
2-15
-------
decrease with increasing wind speeds. These ratios also depend on the
particle size distribution being sampled; at typical wind speeds of
8 km/hr and 15 km/hr, the SSI/HIVOL ratios range from .72 to .93 while
corresponding TOTAL/HIVOL ratios extend from .68 to .91. For the
urban average (P = .33) distribution at typical wind speeds5 the
SSI/HIVOL ratio is .72 and the TOTAL/HIVOL ratio is .68.
The second comparison demonstrates the extent to which inlets on
samplers which are intended to measure the same fraction of the
aerosol mass, the IP fraction, can be expected to do so. The ratios
of collection efficiencies of these samplers tabulated in Table 2.3.2
show the degree of equivalency which can be expected under ambient
measurement conditions. As seen in the Table 2.3.2, the ratio of
collection efficiency of dichotomous samplers to HIVOL(SSI) is close
to, but generally less than 1.0. For the typical 15 km/hr wind speed
and urban average size distribution, the samplers measure the same
mass concentrations within 5% of each other. There is a discrepancy,
as noted earlier, between the Beckman and Sierra inlets at low wind
speeds; the Sierra inlet samples up to 22% more aerosol mass than the
Beckman or SSI according to these tests. At high wind speeds, a
substantial difference, approaching 25% between TOTAL and SSI samples,
is apparent. To reiterate, this treatment involves only the
penetration properties of the inlet and does not consider the
adsorption or loss of gases by the filter media or passive deposition
on filters during sampler standby periods prior to and. after sampling.
The final comparison is important to relate conclusions drawn
from a measurement system with a 15 urn cut-size to a standard which
would possibly address the mass fraction in the 0 to 10 um size
range. The ratios of the hypothetical 10 um inlet (s = 1.35)
collection efficiencies to the TOTAL and SSI efficiencies appear in
Table 2.3.3. For the urban average aerosol (P = .33) and 15 km/hr
wind speeds the 10 um inlet would collect 87% of the mass collected by
a Sierra dichotomous sampler and 82% of the mass collected by a HIVOL
with a size-selective inlet. These factors, rounded to reflect their
uncertainties, .9 for TOTAL measurements and .8 for SSI measurements,
can be used to estimate mass concentrations in the 0 to 10 um size
range from existing IP measurements. These factors must be used with
2-16
-------
TABLE 2.3.2
RATIOS OF DICHOTOMOUS SAMPLER COLLECTION EFFICIENCIES TO
HIVOL SIZE-SELECTIVE INLET COLLECTION EFFICIENCIES AS A
FUNCTION OF WIND SPEED AND PARTICLE SIZE DISTRIBUTION
TOTAL/SSI
(Sierra)
(Beckman)
Aerosol
Classification
Average (Whitby)
Urban Average
(Lundgren)
Background (Whitby)
Marine (Whitby)
J, Wind speed
Fine
33
50
33
50
33
50
33
50
5 km/hr
1.12
1.07
1.22
1.14
1.08
1.06
1.14
1.08
15 km/hr
.96
.98
.95
.95
.98
.98
.97
.97
40 km/hr
.83
.88
.76
.85
.86
.90
.84
.89
2 km/hr
.97
.98
1.00
.98
.96
.98
.98
.99
Ranges
of
Ratio
.83
.88
.76
.85
.86
.90
.84
.89
to
to
to
to
to
to
to
to
1.12
i.O/
1.22
1.14
1.08
1.06
1.14
1.08
TABLE 2.3.3
RATIOS OF HYPOTHETICAL TEN MICRON INLET COLLECTION EFFICIENCIES TO
IP SAMPLER COLLECTION EFFICIENCIES AS A
FUNCTION OF WIND SPEED AND PARTICLE SIZE DISTRIBUTION
(Slope of 10 urn collection effectiveness curve is 1.35)
Aerosol
Classification
Average (Whitby)
Urban average
(Lundgren)
Background
and Urban Plume
(Whitby)
Marine (Whitby)
%
Fine
33
50
33
50
33
50
33
50
5 km/hr
.78
.84
.67
.77
.83
.87
.76
.84
10 um/TOTAL
(Sierra)
Wind
15 km/hr
.90
.93
.87
.92
.92
.94
.90
.94
Speed
40 km/hr
1.05
1.03
1.07
1.04
1.04
1.03
1.04
1.03
10 urn/Total
(Beckman)
2 km/hr
.89
.93
.82
.88
.93
.94
.88
.93
10 um/SSI
.87
.90
.82
.88
.89
.92
.87
.92
Range of
Ratios
.78
.84
.67
.77
.83
.87
.76
.84
to
to
to
to
to
to
to
to
1
1
1
1
1
1
1
1
.05
.03
.07
.04
.04
.03
.04
.03
2-17
-------
great care, however, with full recognition of the range of values they
can take under different wind speed conditions and aerosol size
distributions. It appears that the present IP Network sampling
configuration overestimates 0 to 10 urn mass concentrations by 10% to
20%.
2.4 The SURE Sampling Inlet Characteristics
Several other samplers can be said to measure FP and IP. Not all
have undergone wind tunnel characterization and it is inappropriate to
examine them here. Lioy et al (1980) and Camp et al (1978) mention
several of them. The sequential filter sampler designed especially
for the Sulfate Regional Experiment (SURE, Mueller and Hidy et al,
1981) is important because it acquired time-resolved (3-hr samples,
8 samples/day) of IP and FP at nine sites in non-urban areas of the
eastern United States during 1977 and 1978. Some of these data will
be summarized in Chapter 6 to assess non-urban concentrations of IP
and FP, so it is important to relate the collection characteristics of
these samplers to those in the IP Network.
The collection effectiveness curves at different wind speeds for
the SURE IP and FP samplers are given in Figure 2.4.1. The IP 50%
cut-size is 9 +_ 3 urn, depending on wind speeds. Table 2.4.1 compares
the SURE IP collection efficiencies with those of the HIVOL,, SSI and
hypothetical 10 um inlets.
Most of the SURE sampler collection efficiencies at 2 km/hr and
8 km/hr wind speeds are at or slightly below the lower limit of the
acceptable performance range for 15 um inlets; the collection
efficiencies at 24 km/hr are significantly less than this lower
limit. The variability of collection efficiency with respect to
particle size distributions and wind speeds is similar to that
exhibited by the Sierra dichotomous samplers in Table 2.2.2,. For the
size distributions studied, the SURE IP sampler should yield mass
concentrations equal to 80% to 90% of those which would be measured by
a Sierra sampler. Since all SURE sites were non-urban, with average
wind speeds of 10 to 15 km/hr, the efficiencies calculated for the
background and urban plume (P = .50) size distribution at 8 km/hr for
the SURE sampler and 15 km/hr for the Sierra sampler provide the most
2-18
-------
100
90
^ 80
X
o
g 70'
o
60^
w
.2 50-
•p
o
o 30'
P.,
2o^
10-
0'
r Acceptable
Performance
Range, RTI
A : SURE FINE
B : SURE SFS 24km/hr
C : SURE SFS 8km/hr
D : SURE SFS 2km/hr
0
0.5
l.'O
2'.0 2i5 3lO
ln(D)
2 3 ii 5 6 7 § ^ l"0 15 20 30 40 50 ^0 70
Aerodynamic Particle Diameter (D), ym
Figure 2.4.1 Collection Effectiveness of the SURE Sampler, Flow Rate 130
(McFarland et al, 1980)
2-19
-------
a
v
0.
u o
m r-i
a o)
o. >
9 B
W
0)
ON r-l
00 ON
o
oo
4 Q)
8 >
01
%
0
a.
vO CM
r^ oo
1— 1
.tf
r-j
TABLE
„
M
[*
CO
S
g
ptf
o
h
CO
Id
^
0
z
id
u
M
[»
b
W
O
B
01
J
d
0
u
M
H
2
i
H
CO
H
*J
M
s
3
O
d
s
lYPOTHET]
rM
o
H
CO
CO
o
>
M
m
3
£
M
HIVOL
£
3
O.
2
W
CO
u.
«
M
8
3
CO
^
CO
CO
M
X
•* ~g
N J
T3 h
0) £
0) 00 --i
0- B
•0 .*
e
• H
»
h
X
N 1
^
vO fs| IA \£> lA ON r— ( r-<
r^ oo IA ^o oo co ^o Is*
IAC> Or^ r-in CMO>
coco r^r*- ^o^ r^ r*°
&\ CNJ \o r>j -^ vO r*« to
CO &* r^ efi ON O^ r-^ 00
iTi^O ^C^J f^-CO ONCM
lA^D OfO CSCS OO'-t
rO^J OOON OOO 3-m r>.r>- m\o
oo^c o i— ' v^cM \or**
\D T** m \o r-co iA*o
CO O
co m
o
m
O -I
-D
01
E
3
00
ffl
01
00
ffl
X
CO
H
0)
a)
,
^^
o
00
^
"^^
a
3
•o
01
>
•0
c
<0
>»
J3
4-1
• H
5
•°
,
^•^
in
r^
OS
r-4
^
cn
3
f-t
3
CO
PM
•a
c
«
c
a)
5
1
3
U
V
a
n
•o
c
• H
s
IM
O
u
c
01
•a
e
01
a.
•o
e
•^
0)
lJ
a
CO
0)
u
e
0)
• H
u
u-t
l*-l
0>
§
.p^
4J
u
01
2
o
o
•o
2-20
-------
appropriate comparison; in these cases the SURE samples would measure
aerosol mass concentrations equal to 90% of simultaneous Sierra
sampler measurements and 88% of simultaneous SSI measurements.
The SURE collection efficiencies at 8 km/hr most closely
approximate the hypothetical 10 urn inlet (s = 1.35) of all inlets
evaluated here. For the urban average size distribution (P = .33),
the SURE measurement would exceed the 10 urn measurement by 7% whereas
for the background and aged urban plume distribution (P = .50), it
would be 5% less than the 10 urn measurement.
Because the uncertainties resulting from the different sampling
locations and sampling periods of SURE and IP Network measurements are
expected to be greater than the 10% difference in mass collection
efficiencies of SURE and dichotomous samplers, the two measures will
be treated equivalently in this report.
The SURE and dichotomous FP collection effectiveness curves are
nearly identical and intercomparison studies (Camp et al, 1978)
between the SURE FP and earlier models of the dichotomous sampler show
the mass concentrations measured to be equivalent.
The SURE data base presents a large quantity of size-classified
concentration measurements with established accuracy and precision
which can be used to characterize non-urban IP and FP concentrations
in the eastern United States. The SURE IP measurements may more
closely approximate those which would be made with a 10 um cut-size
inlet than the IP Network measurements.
2.5 Comparison of Calculated and Measured Relative Collection
Efficiencies
The collection effectiveness curves and the collection
efficiencies derived from them in the previous sections are based on
laboratory measurements and certain assumptions about size
distributions which may not be valid under ambient sampling
conditions. Since many IP Network sites ran HIVOL, SSI and TOTAL
samples simultaneously, it is possible to compare the ratios of mass
concentrations collected with these samplers to those presented in
Tables 2.3.1 and 2.3.2.
2-21
-------
These average ratios for TOTAL/HIVOL, SSI/HIVOL, and TOTAL/SSI
for various site type classifications are tabulated in Tables 2.5.1,
2.5.2 and 2.5.3, respectively. Other researchers' measurements with
the same types of samplers have also been tabulated (several other
references with sampler comparison results were reviewed, e.g. Kolak
and Visalli (1981), Camp et al (1978), Dzubay and Stevens (1975),
Trijonis et al (1980), but results were not included because the
sampling inlets differed from those used in the IP Network).
Three methods have been used to establish the relationships
between simultaneous measurements of two samplers, the average ratio,
the ratio of averages, and the slope of a linear regression between
one measurement and the other. Results of each of these methods have
been placed in Tables 2.5.1, 2.5.2, and 2.5.3 because, as can be seen
for TP Network data used in this study, they are not the same. By
listing all three relationships it is possible to compare the
inter-sampler relationships of the IP Network with those of other
networks. The average ratio should be compared with the calculated
values in Tables 2.3.1 and 2.3.2. The final column in each one of
Tables 2.5.1, 2.5.2, and 2.5.3 contains the calculated ratios which
are closest to the measured ratios. The samplers, size distributions
and wind speeds corresponding to these ratios are listed in the notes
for each table.
The three TOTAL/HIVOL relationships in Table 2.5.1 show a
substantial variability for the same data. For all site types
combined, the linear regression slope is the lowest at .60 while the
average ratio is the highest at .73. This trend persists for all site
types. The linear regression slopes found in this analysis of IP
Network data are consistent with those found by Suggs et al (1981b)
(using measurements from the same network), Miller (1980), Wendt and
Torre (1981), Pashel et al (1980) and Grantz (1981). The ratios of
averages found by this study, Suggs, and Grantz are in reasonable
agreement while those reported by Miller are approximately 25% lower
than the others. The average ratios reported by Suggs are lower than
those found in this study even though the basic data set is the same.
This is due to the different methods of calculation and the different
data validation procedures used (Section 3.4 of this report describes
2-22
-------
m
CN
1
g
TION MEASUREM
CONCENTRA
en
1
J
g
32
f-
'S BETWEEN TO
LATIONSHI1
ai
Similar
CO
£
U-l
o
o
2
g
,/HIVOLf
• Regressic
er-
< « c
H CU M
pj
O
.rl tt-l
a:
o
u
S
&
TO
>
CU
j_i
•a
u
o
•r-<
4-1
S
ctJ
u
• l-t
sc Theoret
i— i
lation Samf
Ui
Ui
o
u
J_l
cu
u
cu
a
o
w
XI
(0
S
w
cu
>
to
>
4)
o
and Samplers
=
-r- 00
X-C-Cr^^^E x:.fioqx:
r-iCNi-lP^CN»-tCN OOQOOOO'-i
CN in ^O ON CN in O ON I"*" f^1*
r-» r— i CN r-t i~ ( ^-i CN r-i ON ON
rom-ctoor^cNOO \D r- * CN .— i ^H en c^c^ro oO <— < in >jD ON 00 oo r^- c^ ON ON r*» oo f^* 00 ON
m ^o NO
PO oo "tf1 c*^J O "^ IA ^ ^ * 00 -^
m NO oo *o in ^ ON **"* o ^^ ^o
O<-r'-J''^'u^fyiiin f*^ c^1 ^-rncNON
\o *o in vo in \o in oo 60 M oo oo ^o oo oo oo u™t ^o NO in u"*
iWtu*OOeuo>> ^-i -ajcutu>i -"
•OEM Ui njco-r-i coggE tc^toto
tflcOt033« 3 g -f* « B tOtOtO 33 33
*-* i-3
CU 00 O
X) •-' O w
«J i-4 pi *— -CrfOO-EO-I
r-( CO t~ 1 1— 1 ON *— '
CU< ON< W rf 1-1 T3 <
01 iJ ,-i >J pJ^^cJ
WCL| ^-'p^ y-s UlR* Cfl .*-s CU
v-'«S " E O O - £ r-l f-* S
X-S<)-< r-4-J-< 00--,OO AJ C > "- ' > C C > CU ^- C
ij 4J tO cfl »— < OJtOtOM ctft_4 cOffltOM tO cO
CO J-t Ui E rE 1-4 E rE Ur Ui X UiErC r-iUi NE
to*OCUU*^ OOCUCJT3 i—<(lJ*C 'OCUtJ'U JZ 0) C^*
•r-j x: .r-i oj c oo — -i u c »-* ••-* c c -H cu c co *H o eu
xiococQto 3cnpaco -Hcoco cucopato toco u^ea
H CO £ ES (X O
s
oo
-*
r*.
m
CN
CN
00
, Sierra
to
3
r:
to
14-4
0
o
J_>
CO
Ui
cu
x:
.u
o
AJ
Ui
CU
14-4
tu
Ui
o
iJ
tO
c
• r-4
«4-i
cu
nbers appear,
3
C
CU
CtJ
cu
c
i_l
cu
J
«
— 1
1
0
•— 1
o
•o
"
CO
cu
— 1
a.
E
CO
CO
4-4
O
n
E
3
c
CO
.Ul
cu
n
he average T
alculate t
o
•LJ
eu
CO
B
TO
O
CU
"i
3
C
0)
£
to
1
o
CO
x:
H
Av
3.4
T! 60
01
.0 CU
des
hin
to
c
.2 TO
iJ 3
TO -O
.5t
.ss
t-i c
0) -n
TO 01
ou
at
«H T3
3 §
•v
01 Ol
CO 60
3 2
2 s:
cC f
g O
Ctf CO
co tu
00
eu rt
X I-
ng so
ithme
•-i C
3 a
ogen
a o o
cu -^ -H
Ui •*-» j_i
00 CL
C E
•^ to
•—( CO
a
E tu
co en
co cu
rC
E- 3
0) B
X
" j:
XJ
C "J
0 3
60 O>
C ^
--J 0)
cu a»
Q, 4J
01 C
Q .H
o
j-
a
S
o
TO
Ol
• H
cn
o
n
Ave
Ave
rl
S3
f.
e
J£
j: X
E ~E
n TO
B rl
60 60
TO TO
Av
Av
.C X:
E "E
01 0)
> >
< <
0
7.
60
2-23
-------
CM
IA
W
J
CO
2
j[r.
a
58
r
g
<
z
u
8
CO
1
>
t-4
_
5
^
M
to
u
U3
03
w
cu
>— i
35
C/3
sc
2
5
Dd
n
O
t) 4J
l-i q)
CO Bi
• H — <;
e 00 1
t-l 01 hi
X OS 01 u
l-l hi E 01
CO B M U
CO 01
E
0>
a
o
CO
09
01
.0 oo
O B
•H U-l hi
CO >
as <
B
0 •O
••4 4J
4J CO
OB
01
00 •
B 00
>
<
U 01
4J O»
X
ON ON C*«J ON
* • • •
£ oo ••< j:
.s £ jc . . . .
CM CM CN CM CM ON P~
00 ON ON cn CM CM ON
\O CM -^
cn -"« -^ in CNI o NO
CM ON ON ** NO O CM
in ON O O -* r- >3-
1 1
sssssss
SCM 00 O*. CM 00 00
r* r* os p-* \o r**
2 2 2 2 §- S
r* r- r- NO NO NO r-
—
^ -< B
B B •*
-J ,- « H 0 E
B It -H tJ hi 0>
— .- 4j to oi -a
u u v -o 1 '»
00 CJ •U C 0 0>
B ,= § '- H U 06
0> E S 01 B E B
w 1-1 u en « B a
05 E E B hi hi hi
a a
\-O CM ^* « •* »A *3"
rn ^^ ^^ r«> en 60 «-o \o
irt CM IA IA m a\ ^3-
^ i— « -^ ** >^
oo m o
^ co o*>
^H OH
•* CO
-- CM 1*
I
CM vD 00
ec 00 eo oo ofi ao r-*r»
•rf c^ cn o P** ^^ >A
CO sO r** O^ P-» CO r*»
O 00 P*. ^ CM
•— ( O O O <-(
O Cf' Cfr CM O
(0
V
a.
«««••- « (C
flj -H .H -U 0» *^ '^
•U )- O C 4J U t-i
•<-> jj u CD •<-( u u
cnnai'Q^ co nn
3 g "- « 33
«-* *O B CB 1W -^ 'O'O
<->eoai3 — cc
< M U OC 02 < MM
a) co
,-> ON
£ C^
CO
c^ &>
^-^ U /-"v
0 P-I
-i H 00
•o --
4J C ^
(U CO
W 4J 4J
60 'O C
00 C «
3 0) U
w S u
•
(0
0
*•*
u
§
t-t
•X
M
CO
CO
B
3
•O
•J
•O
B
O
E
0
4J
a
01
T3
CO
•a
E
B
4J
m
•o
B
0>
00
hi
0>
B
U
4J
e
£
B
.
09
B
O
• H
4J
CO
hi
B
01
U
E
0
u
-3
s
s
01
00
B
U
0)
B
.^
1
hi
S
•o
E
a
O)
00
11
s
u
• H
4J
f*
4J
• H
hi
B
U-l
O
O
4J
B
OS
.C
01
00
B
0)
B
0
0 J
u >
B i-l
hi SB
01 01
£ 00
4-) B
hi
O 0>
B
01 01
OJ 4J
01
O 4J
3
^ O
CO -«
E B
•-4 O
U-l
O
^ 4J
4-1 TJ
OJ
- 01
hi 3
B
0> 01
a oi
Q-— i
B CL
S
01
.0, UJ
B 0
3
E hi
01
0) .0
£ 1
£ E
4*>
01
E £
0) 4J
*.3
C "O
0 b
01 £
0} 4J
0)
hi 0)
Q0£
0) 4J
ht
K
hi M
01 CO
E
— i 00
B
•0 hJ
E 01
B >
fl
01 0)
oi .e
B
hi 01
0) u
> a
S rl
3
O —
O U
4J O
B 4J
1^
t3
. j,
O 01
.-1 3
4J
B 01
hi 01
OJ Ct>
hi 01
01
> *w
CO 0
hi hi
o o>
u-i fi
. 1
<» e
•H
CO 01
a .c
CO
u to
n -H
•o
T3
"4-1 C
O O
u
l-i (U
a) n
| «
u
«
^
CO
CM
0)
«
60
co
u
OJ
^
^
•
CO
c
o
JJ
u
0)
a.
C 3
•^ o
1-1
•O 60
a>
^ 0)
U AJ
u
co e
"O £
4J
» •*
CA $
O 0)
CO
c
O M
w 3
CO -O
c -H
"s -S
.-^ -o
,-t C
OJ ..H
(0 01
u j;
CO 4-f
4J
OJ CO
£ §
4J ••*
O fl
£ hi
.- -D
a E
B
4J
hi 01
O E
a o
0) -H
hi U
B
09 hi
£ B
u 01
u
E E
•H O
U
01 01
01 00
3 B
hi
fl OJ
B B
•o
14-1
| 0
B 01
01 01
00
£ hi
U4 >
O
u
0! --i
e u
O 01
09 B
00 U
B -H
•H hi
01 fl
3
0)
•a u
OJ B
4J
q oo
^ E
3 •.<
u a
— i 3
B O
U hi
00
01
hi 01
01 *-•
9 **d
09
01
O £
.-i O
4-1 B
B O)
hi
E
0) •*
(ft £
0) 4J
f- '»
01
X
0)
>
B
00
01
E
hi
O
01
4J
09
O
X
m
01
09
B
00
00
E
E
fl
E
O
U
hi
3
iw
f-4
09
hi
O
B
01
M
0
iJ
E
U-l
O
E
O
4-1
hi •
O 09
01 hi
•a oi
o •-
a
- e
00 B
• H
•S.S
B£
09 U
M E
CO O
CO hi
U-l
hi
O V)
u
iJ E
O 01
1— 1 01
S hi
3
hi 01
O B
uj 0)
E
•o
01 B
01 O
3 -H
M
B CO
••4 hi
U
0) O
• * n
14^ 0]
oj i
iJ ,£
4J
C "^
O 3
60 0>
C U
>H OJ
C Ui
S."
OJ C
O •-"
U-l
.
X
•*
• CM
J-(
"^ M
B co
II
CM ft.
ey
O 60
lA CO
• U
II OJ
ft- >
0 C
• «
U U JO
1 f =
j*i ^ v-
o
00 • CM •
U 1- k<
co *B S "£ *&
Qu CM ft, CM 00
• •> M •> * ft
•c oj ro oj o cn
Q] 60 CO 60 lA CO
.--CO (0
^ V- II W II (I
•-4 OJ ft- OJ ft- ft.
OJ < < -
a o) oj u
w C C C C C
flj -H CO *^ '-•<
JJ ^ l-i
O t-i jfl l-i CO CO
Z ,=>£.=)££
60 ^ •.- -*-i .* t-H
2-24
-------
z
u
£
H
58
in
£
Z
O
CJ
x
2
£
O
z
•i
j
<
H
g
a
en
a.
a
a:
•a
u< ca
a az
u OJ
0 a
• E
&&
e
0
u
«
C/J 00 1^
—• 01 01 ^
J OS u Q
< C 01
H h i ••
p n
P 01
U3 E-
•rJ X ^ B -1
^40 r*- "~i in co in
O cy- o oo oo o
x o E O
cr> P-- oo P-N
r*. -j- m
^ i—i O
m
- CM CN
.—i ^D O " O O °^
CM oo r*« oo CM CM in
) m o
i-< PH 60
f** f** o
vD *d^00
o o o
00 60 60 60 60 60 60
O •—' O^ i
I 00 —I O
f") 60 bO 60 60
, •— 1 1 — <-H CM t— t
oo co co r- o
to
CO
i— in cy\ r*.
—i CN co o
en -O
OJ C
O 01 C C C 0)
CJ OS ffl co CO 4-i
to c c c i-i i- i-
CO CO cfl 3 3 3
^-t x x x .n x x
—i 1- I* U 3 3 3
< D 3 D CO CO CO
COO*
MOOi
CO CO
•1-3
c c
00 rH
60 *-<
3 <
3
CO
X
co
c
ra
v^
u
CO
C
O
• H
4J
CO
C
01
u
c
o
o
, •]
t— 1
QJ
60
CO
i-i
QJ
tfl
U
4-1
QJ
E
'£
CO
C
KJ
^
E-
P
01
60
ej
l-i
0)
>
U
bO
cfl
p
OJ
>
«
U-J
0
o
•rH
4-J
tfl
l-l
Ql
X
4J
O
4-J
J-l
QJ
U-I
QJ
^
O
3
4-1
i—4
Cfl
C
• rl
U-I
OJ
X
XJ
I-i
CO
QJ
CL
a.
cfl
CO
i-l
0)
1
C
QJ
QJ
l-i
X
4J
C
OJ
c
o
CO
CO
QJ
l-i
60
QJ
k
C
•H
T3
C
CO
60
CO
l-i
QJ
>
U-J
O
o
ij
^
A
o
JJ
0)
^J
01
Ui
o
U-i
to
•H
e third is the number of samples in the SSI average .
x
4J
-
00
«
QJ
>
J
<
E™1
f-
0)
u
OJ
cfl
3
U
efl
O
T3
41
CO
3
CO
OJ
r—i
o.
E
CO
o
QJ
X CM
E
3 CO
C
CM
4-1 cfl CD
Cfl
3
U
cfl
o
QJ
OJ
CO
O
cfl
^J
OJ
CO
QJ
X
H
X
01
l-l
CO
0
4-J
cfl
OJ
60
cfl
l-i
QJ
CL
3
O
60
QJ
X
C
x
•H
CO
OJ
4-1
3
•o
•H
T)
C
0)
X
JJ
cfl
to
O
4-1
Cfl
l-l
-o
c
CO
CO
c
0
•H
CO
l-i
U
C
01
o
c
o
o
OJ
60
cfl
Ut
QJ
>
tfl
O
CO
60
cfl
QJ
>
CO
u
•H
01
E
•^
j_i
cO
QJ
cfl
60
C
•r-i
a.
3
O
t-,
60
QJ
• H
(0
X
u
cfl
QJ
C
X
u
3
^,
i"-i
00
f~J
60
00
3
C
.r-l
*O
QJ
^_,
O
01
U
• H
4-1
Cfl
l-l
<
E-
r ,
t—t
CO
to
OJ
U~i
o
1*
CO
l-l
01
>
c
loss of nitrogen or sulfur containing gases may positively or nega t ively
^
o
c
o
4-1
a
o
CO
T3
„
60
C
a.
E
CO
CO
t_J
to
O
i_J
<
H
O
l-i
o
OJ
CO
3
CO
'O
QJ
E
^,
QJ
, |
U-I
OJ
J_l
c
o
60
C
•H
TJ
C
01
a
0)
Q
CO
1^
OJ
pt-
E
tO
CO
CO
u
4_»
E
O
i-i
to
C
QJ
E
u
3
CO
cfl
D
E
C
O
• H
cfl
c
01
U
c
o
u
CO
CO
cfl
E
X
4-1
3
i-i
QJ
QJ
4-1
C
Sierra ^Background, P=,50, Sierra 5 km/hr.
i-i
x
*B
QJ
c
^,
•E
C
cfl
C
g
U
OJ
CO
CO
CO
i
0»
60
CO
Q>
>
^1
X
QJ
U
C
QJ
QJ
(^
C
•l-J
QJ
O
OJ
D-
(0
4_(
2:
/hr. lUrban Average, Sierra 15 km/hr. ^Average, P=.33, Sierra 40 km/hr.
E
^
0
CO
Vj
QJ
CO
K
o
.
•.
O)
60
jj
0)
>
c
cfl
J3
•*
l-i
X
E
CM
C
O
QJ
CO
CO
i
^
OJ
60
tfl
QJ
^J
C
cfl
X-
l-l
3
^
X
"E
o
cfl
QJ
to
„
CO
II
QJ
60
Cfl
QJ
>
C
efl
w
o
X
E
o
CO
^,
OJ
• H
CO
.
o
II
*.
"O
3
O
00
u
CO
pa
2-25
-------
those procedures). This comparison of measurements from different
researchers using the same samplers highlights the difficulty of
determining an experimental relationship between the collection
efficiencies of different samplers. The data chosen and the methods
of summarizing those data must be carefully considered.
The calculated ratios summarized in the final column of
Table 2.5.1 come from size distributions and wind speed conditions
which are plausible for the site types. For example, urban average
distributions with wind speeds from 2 to 15 km/hr exhibit calculated
ratios very similar to those measured at urban industrial, commercial
and residential sites. Similarly, for rural agricultural sites the
ratio for the average size distribution at 15 km/hr is the most
similar.
Substantial variability in the individual ratios exists, as
evidenced by the large standard deviations (which range from .14 to
.28 for different site types). This is reasonable since the size
distributions and the wind speeds vary significantly for the periods
of time and geographical distances over which the samples were taken.
The close agreement between the calculated and measured ratios lends
credence to the model of Section 2.2 and to the results of Section 2.3.
The three SSI/HIVOL relationships of Table 2.5.2 do not show the
same trend as the TOTAL/HIVOL relationships. In this case, though the
linear regression slopes are usually the lowest (except in the case of
negative intercepts), the ratio of averages is greater than or equal
to the average ratio. The calculated SSI/HIVOL ratios for the urban
average size distribution with 2 to 8 km/hr wind speeds are in fair
agreement with the average ratios measured at each site type.
The TOTAL/SSI relationship is an important one, since both of
these samplers are being used to estimate inhalable particulate matter
concentrations. The calculated ratios in Table 2.3.2 show that the
calculated equivalency of the two samplers varies up to 24% with size
distribution and wind speed, though TOTAL/SSI ratios under typical
wind speed and particle size conditions are approximately .95.
2-26
-------
Table 2.5.3 compares the IP Network measured average TOTAL/SSI
ratios with the calculated ratios. Most average ratios are fairly
close to unity, though the size of the standard deviations indicates a
large amount of variability. This variability is evident from the
scatterplot of simultaneous TOTAL and SSI measurements in Figure 2.5.1.
The measured ratios at suburban sites are significantly less than
those at the urban sites and are similar to those calculated for
unlikely size distributions and wind speed conditions for these
sites. Most of the suburban measurements were made during 1980
whereas greater than half of the urban measurements were taken in
1979. Different SSI and HIVOL filter media were used during each
year, and it appears that the adsorption of sulfur and nitrogen
containing gases on the 1980 filter medium is significantly higher
than that on the 1979 filter medium. Since the TOTAL filter medium
remained the same, the TOTAL/SSI ratio would decrease in 1980 if this
adsorption contributed to SSI mass measurements. It is impossible to
quantify this effect since the filter medium used for individual
samples is not reported with the measurements. The magnitude of this
artifact is discussed in Chapter 3.
This lengthy treatment of the particle size collection
characteristics is of great importance to the formulation of
monitoring requirements for estimating compliance with a
size-classified standard as well as interpreting IP Network
measurements. The following observations are applicable to these
subjects:
Though dichotomous sampler inlets do not meet the stated
collection effectiveness requirements of Ranade and Kashdan
(1979), they do meet the collection efficiency requirements
under typical wind speeds and particle size distributions.
HIVOL(SSI) samplers attain both collection effectiveness and
collection efficiency requirements. Criteria for particle
size collection efficiency for acceptable sampling devices
should be specified in addition to or in place of criteria
for collection effectiveness. The efficiencies of samplers
which would fall within the window of Ranade and Kashdan
(1979) range from 47 to 71% for a typical urban size
distribution. A sampler with an effectiveness curve
designed to meet the lower limit would sample 66% of the
mass of a sampler designed to correspond to the upper
limit.
2-27
-------
<•>
S
3.
*
C
o
•H
+J
rt
c
c
0
u
V)
s
0)
1— 1
rt
chotomous ,
•H
Q
306
27Q
24C
ZlSi
182
156
12G
9C
66
36
'
SLOPE : .821+- .320
INTERCEPT : 5.4694- 1.314
CORRELATION : .924
NO. OF POINTS: 285
•
•
• ,
1 1
i • ' , ,»
%' *
• If'™
i ' i' *' »' ' *
• j *"»f ly
•JjiyjV'i
• , (p*11
b' L 1_ J A A 1 1 J
60 90 IZ0 ld0 180 ZT0 Z40 Z/0 30
Size-Selective Inlet Mass Concentration, yg/m'
Figure 5.1
Scatterplot of Simultaneous Measurements of Particulate
Matter with the HIVOL Size-Selective Inlet and the
Dichotomous Sampler.
2-28
-------
Present wind tunnel testing of inlets provides useful
comparisons of collection effectiveness. Tt is not certain
that all such tests are equivalent. Standardized methods of
evaluating collection effectiveness and efficiencies should
be devised and applied to all samplers used for evaluating
compliance with a standard. These methods might include
wind tunnel measurements, sampling of standard aerosol size
distributions, and simultaneous ambient sampling with
established inlets.
Inlets with 10 urn 50% cut-sizes can be expected to collect
between 80 and 90% of the mass collected with the present
dichotomous and HIVOL size-selective inlets.
HIVOL(SSI) and TOTAL inlets sample equivalent mass
concentrations within 5% of each other under typical
situations. However, under low or high wind speeds, in
particle size distributions with much coarse material, or
when interferences are present in one or both of the
samplers, this equivalency degrades.
The integration of the product of particle size distribution
and collection effectiveness provides useful information on
the average relative mass collection efficiencies of
different samplers. Results of this model agree with
average ratios determined from ambient measurements. These
average ratios show substantial variability and a knowledge
of the ambient particle size distribution and wind speed
during ambient sampling is required to truly test the model.
2-29
-------
CHAPTER 3
THE INHALABLE PARTICIPATE MATTER SAMPLING NETWORK
MEASUREMENT PROCESS
This chapter briefly summarizes the measurement of suspended
particulate matter in the Inhalable Particulate (IP) Network. In
later chapters, reference will be made to the information on sampler
siting, sampling and analysis procedures, data validation and
measurement precision presented in this chapter. Much of this
information is included as a reference for future as well as present
uses of IP Network data because it is not easily available elsewhere.
3.1 Sampler Locations and Site Descriptions
Pollutant concentrations and the causes of those concentrations
vary over many geographical scales. A monitoring network design is
expected to define the geographic scale over which pollutants are to
be measured in accordance with the purposes for which measurements are
being made and samplers must be placed in locations that are not
influenced by pollutant sources which are significant contributors on
a smaller scale. Table 3.1.1 shows the types of scales and the
nominal separation of samplers required to represent those scales.
TABLE 3.1.1
STUDY SCALES
(Federal Register, 1979)
Nominal Sampler
Study Scale Separation (km)
Global 1,000
Regional 100
Urban 10
Neighborhood 1
Middle .1
Micro .01
3-1
-------
The TF
concentrati.
of air qual
Howeve
representat
located in '
representat
source cont
industrial ,
near a heav
Approp
made by tha
between sam
terms of th
understand!
concentrati
developed t
siting on m
description
for measure
study scale
criteria de
Particulate
alleviate i
prevent mee
Figure
sampled whi
individual
Table A.I c
state, the
letter refe
numbers def
the samplir
after Trijc
classificat
within appr
twork was designed primarily to characterize urban-scale
of suspended particulate matter, since the attainment
standards is evaluated over this scale.
several of the sampling sites can be termed
of regional or neighborhood scales. Samplers; are
ely separated non-urban areas for regional
and one study was performed to assess concentration and
ution variations over the neighborhood scale in an
a. Plans are underway to complete a middle-scale study
traveled roadway.
te sampler siting is paramount when the measurements
ampler are to be used to interpolate the concentrations
rs. Therefore, descriptions of the sampling sites in
arious geographic study scales are essential to an
of the accuracy with which they represent the
within that scale. As yet, no simple model has been
lace accuracy and precision estimates due to sampler
urements taken at a particular site. The site
ust be used £ posteriori to offer possible explanations
ts which do not agree with expectations for a certain
Most sampling sites in the IP Network follow the siting
ed by Ludwig et al (1977) for siting Total Suspended
tter (TSP) monitors. Though these criteria are meant to
uences of local sources, practical considerations often
g them in their entirety.
1.1 shows the geographical distribution of the areas
T.ible A.I of Appendix A lists and describes the
pLing sites within each area. The first column of
alns a site mnemonic; the first two letters identify the
oid two letters identify the city or town, the third
t) the site within the city or town, and the final two
:he type of sampling site. These two numbers refer to
i :e type as defined in the first part of Table 3.1.2
^1980). The first number represents a regional
if the general area surrounding the sampling site
ti itely a 10 km radius, and consists of urban, suburban
3-2
-------
t/1
fn
(U
g
O
o
oj
O)
CL,
t/)
oj
0)
i
en
M-l
O
o
•H
•p
rt
o
o
&0
• H
tt.
3-3
-------
TABLE 3.1.2
DESCRIPTION OF ENTRIES IN TABLE A.I OF APPENDIX A
1. Code for Location and Land Use (Column 1)
Location Land Use
Indus_tria.l Commercial Residential Agricultural Remote
Urban
Suburban
Rural
11
21
31
12
22
32
13
23
33
14
24
34
15
25
35
A pair of numbers followed by "?" designates an uncertain
classification.
2. Code for microinventory, Local or NAMS hardcopy, NAMS soft copy.
"*" means a site survey is planned.
"0" means no site survey is planned.
"1" means a site survey has been made.
"2" means a site survey has been made but it is incomplete.
3. Site Elevation above sea level in feet.
4. Probe height above ground in meters.
5. SAROAD code: First two digits - state code
Next four digits - county, city or district code
Last three digits - Site ID
3-4
-------
and rural descriptors. The urban areas are characterized by closely
spaced and heavily traveled highways, closely spaced single and
multiple family dwellings, concentrated commercial establishments, and
r\
population densities of greater than 2,000 people/mi or more. The
suburban areas contain a small number of heavily traveled highways
(usually an Interstate beltway, spur or connector), a large number of
access roads, widely spaced, single family dwellings, widely separated
shopping centers and population densities of the order of 2,000
people/mi^. A rural area is characterized by few roads, scarcely
traveled, dwellings separated by 1 km, open or forested areas and
population densities less than 500 people/mi^.
The second digit of the site descriptor refers to the types of
land uses within approximately a 1 km radius of the sampling site. An
industrial area contains one large or several moderately sized
manufacturing facilities which commonly emit some form of air
pollutant. A commercial area contains stores, parking lots and office
buildings. A residential area consists primarily of dwellings, an
agricultural area of fields, and a remote area of mountains, forests
or meadows.
These classifications were given on the basis of site surveys if
they were available. In the absence of site surveys, the
classification assigned to a site in the Directory of Air Quality
Monitoring Sites (USEPA, 1978a) was adapted to the previously
described scheme (the classification in this directory is slightly
different from the one presented here). Three types of site survey
were used, and their existence for a particular sampling site is
indicated by a code from the second section of Table 3.1.2 in columns
3, 4 or 5 of Table A.I of Appendix A.
The most detailed site survey is the microinvetory developed by
Pace (1979). This survey includes a detailed map of the area within
1/2 km of the sampling site noting roads, traffic counts, open fields,
storage piles, and any visible emissions. Major point and area
sources are identified within a 1.5 km radius of the site and major
point sources are plotted within 8 km of the site. Estimated emission
rates are assigned to each source and summed over a set of predefined
sectors. Photographs in each of the cardinal directions are
3-5
-------
included. Sampling configuration (probe height, equipment, location,
etc.) is also described. Microinventories were supplied by the EPA
Office of Air Quality Planning and Standards (OAQPS) for the thirteen
sites indicated in Table A.I of Appendix A.
Several of the sites have been designated as National Air
Monitoring System (NAMS) sampling sites. One of the requirements of a
NAMS site is that it be subjected to a NAMS hard-copy site survey.
This survey includes the 1/2 km map, identification of point and area
sources, optional photographs, and configuration description, but it
contains no estimates of relative emission rates. These hard-copy
surveys were performed in Buffalo, NY and Philadelphia, PA as part of
this project. Others were supplied by OAQPS. Surveys for Phoenix, AZ
were taken from Richard and Tan (1977) and for El Paso, TX and
Houston, TX from Price et al (1977).
The assignment of a two digit site-classification code was made
by the authors of this report based on their review of the existing
information. Table 3.1.3 summarizes the environs in the vicinity of
each site which was identified by the survey and used to assign a
site-type classification. This table provides a more detailed
description to supplement the two digit codes.
A NAMS soft-copy survey, the existence of which is noted in
column 5, is a computer generated listing of site coordinates,
configuration and classification according to the Directory of Air
Qua!ity Monitoring Sites (USEPA, 1978a) criteria. The presence of
nearby roadways or industries is often, but not always, noted on the
survey. The site-type classifications from the soft-copy survey and
the Directory are less accurate than those derived from the more
extensive surveys because they lack detailed descriptions of the local
environments. Of the 108 sites listed in Table A.I of Appendix A, 42
presently have hard-copy surveys or microinventories and 27 are meant
to have surveys taken in the near future. These detailed surveys of
the sampling environment are necessary to understand the types of
sources which may be contributing to the suspended particulate matter
3-6
-------
f)
f-H
m
w
hJ
PQ
H
O
C
0)
P
a>
UJ
0)
oi
>>
V
P
a
en
01
j-t
en
C
O
•H
4J
a.
T-l
j-<
W
(U
01
•H
^
<
S
w
^
u
od
W
w
M
W
cJ
o
•r-
«
CJ
•H
14-1
•H
W
a
r-
CJ
at
E
ctf
23
a>
4-
C/3
Q
a>
-r4
to
4J
0,
C
M
O
u
•H
Jg
o .c
• 4-14-J
CJ >-<
1 -H W O
•H O CJ O
•H CO 14-1 CO 4J
CO Q} CO,
cn ffi IH P QJ
-i
rC fl) ,O U-l
C (-1 S
0 3 O C *0 <~s
U >- O C 6
t3 UH p CD &
a » -H
3 TJ E -^N CO
o a) j*; a) e
p > T3 ^ O
oo co -3- 3 4-1
OM -^ rH 00
§t-H O f>
C 0 >
M • C *H 4J O
<4H bOTH
C .C to m 4-1
B -H w o> ? to
*O *H CJ '•--' C(
rH .H S I-. >
iH -H 3 4J X!
-3 05 o en 4-J
*O -O C 10 CD P
0» O QJ O
CD (i 4*J C *J
U O CJ f-i ^ T3
O W QJ O O C
tJ to V- Cu C CD
iH
CD
CJ
S_i
QJ
E
O
u
1
c
CO
JH
1U
•^
»
£
CO
J=
c
-H
E
t-1
•H
PQ
W
<
pa
ij
<-
y
OJ
C
M
O
t-i
O
•H
22
OJ
CO
fx,
T3 S
QJ O
j: .w
to iw
o 4-j
U-4 O
0 C
o
•M T3
c
E re
o
4-J
73 CO
E 01
3 3
o
IH -
oo x:
B 3
0 0
IH CO
CO
E 73
re
f> O
P
73
OJ 73
4H OJ
re x
CJ IH
0 3
.J CJ
rH
re
•H
E
OJ
73
•H
CO
K
1
B
re
)j
3
X
3
CO
,j
«
o
o
Ui
cc
4H
Q
hH
pj
*c
CO
OJ
CJ
3
O
CO
4-1
E
«H
O
a
01
00
re
rH
O
Z
CJ
•H
CO
IH
4-1
01
4J
re
at
o
a>
4-1
•H
CO
f
O
a
a>
CJ
at
^
re
at
E
4-t
E
01
E
M
O
U
S
B
O
4-t
S
CN
E
O
UH
P^
p*.
fH
E
73
>
0
E
CO
re
3
^,
o
4J
OJ
u
01
rH
re
4J
CO
3
B
M
1
B
to
14
3
X
3
en
,4
K
.H
CJ
4H
B
re
re
H
w
rH
ffi
<3
•S3
3
O VO
CO •
rH
tfl ^
73
re 4J
O CO
IH re
01
73
QJ 73
Xi B
IH Q
3
"'e
73
0) CN
> —
re rH
a
B-X
03
i— 1 tfl
U CD
*O s~^
B E
3 ^
O
IH CM
00^
B
O rH
IH ^
4J 4-J
O CO
rH CO
OJ
00
E O
•H 4-1
Ui CO
a 01
a TH
IH
73 re
01 73
> B
re 3
a o
j~-i
rH
01
• at
CJ 4-t
-H CO
UH
UH 73
re c
IH re
4-1
E
re M
at
x
(V,
• re
oi 3
4J re
•H
co B
S
i -H
fc
s-^
B
(**> 4-1
) CO
-i. OJ
4J X
CO 4-J
a IH
at o
x: E
3 73
0 E
to re
4-J JZ
re 4H
CO O
at c
a
re to
B 4J
IH CJ
O
4J 14
CO (X
re
fH rH
x re
A oj
x— \ fH
B -H
y^ pj
•* t!
i re
o
4-1
a
at
CJ
OJ
IH
B
0
E
i.
O
ffi
-^
Q.
O
CJ
IH
re
1
IH
73 -H
at re
CO 73
(X E
e re
^-x.
B
• S
rH
OJ vr
OJ rH
rH I
73
B 4-1
3 co
o at
H 3
00 X
4J
at 3
!> o
o to
J3
re co
73
B re
o
in IH
rH 73
at
T3 X
OJ 1-
tO CJ
CJ E
3 §
rH
CO
3
fH
3
CJ
00
1
rH
re
3
OJ
01
IH
OJ
re
CJ
^
rl
f
OJ
•H 73
to B
re
01
X a
4-1 OJ
•H
UH 4-t
O fH
/~^ >H
E 4-1
J^ O
cQ
"^ ^j
rH 14-1
i *
U
4J t-J
tfl *H
fl( ffl
3
Xi 73
4J OJ
P 4-J
B B
•H
§ a,
E IH
3 CO
4H IH
IH OJ
O X
O.H
E 4-J
OJ E
cx re
o o
•H
73 UH
B -H
re E
an
4J *H
CO CO
3
73 O
OJ
•H •
4J re
•H OJ
00 1-
3 re
h.
at
• 4-1
73
E OJ
3 N
O -H
IH IH
eg oj
CJ O
•H re
UH IH
UH re
re x
4J
X 01
00 CO
•H at
fH 73
,
jj
O
4-1
a
OJ
CJ
at
re
OJ
B
73
OJ
•H
UH
TH
4J
E
01
73
•H
at
at
3
CO
OJ
ej
3
O
CO
B
O
cx
CJ
4-1
CO
• X
00 4J
B 14
•H O
73 E
rH
•H 73
3 E
x re
UH X
O 4-1
3
UH O
O CO
o
IH CO
73
E CO
0 0
73
B 73
3 OJ
O >
IH re
oo a
E (*>
5 x
UH re
01
E B
*sf 01
p
73 re
at
to P
cj at
3 H
re
fH
E
01
fH
to
at
tz
\
B
re
j_j
3
3
en
•H ^ fH
P 4J
OJ CO U
4J cx re
*H
W 73 IH
oj re
OJ > fH
x re 3
4-1 a. u
E fH
UH 3 X
O OJ
X >
4J 4J
CO fH E
re 3 o
U IH
re
• t* u
at x to
4J OJ 3
•H rH 73
CO rH
re at
OJ >
X 4J fH
UH 73 00
O 3
73 tin
/S re
fH >,
^~- > IH
IH re re
) IH OJ
•i, 00 E
CO
01
CJ
IH
3
O
CO
^J
E
•H
O
cx
4J
B
CO
U
•H
UH
•H
B
00
•H
CO
o
Z
E
fH
O
re
73
E
3
x
.
o
4J
CX
OJ
CJ
OJ
IH
IH
re
at
E
O
3=
OJ
. 00
00 >-
E re
73
rH O
•H Z
3
t
« rH
OJ
UH >
o re
UH 4J
o
0 4J
P X
00
O rJ
73
E •
3 73
o re
IH O
00 IH
B 73
O OJ
IH X
3
B CJ
6
in o
E
73
ai »
re oj
0 >
o re
rH
re
fH
4H
E
01
73
fH
to
OJ
1
E
re
IH
3
X
3
CO
CJ
„
re
N
3
N
^j
«£
CJ
•
IH
O
4-1
CX
OJ
o
ai
IH
j_,
to
ai
E
OJ
O
3
O
to
B
•H
O
cx
o
1
B 0)
O 4H
B fH
CO
73 UH
at o
re x~s
00 —
E rH
•H ?
73 ^
rH
fH CO
3 B
X O
•H
E 4H
O O
a)
73 IH
E fH
3 73
O
|H rH
00 rH
to
6
0 E
IH fH
to
B 73
re
v.0 O
73
OJ 73
re x
U !-<
3 5
rH
to
CJ
IH
0)
B
e
0
CJ
1
B
re
1-1
3
X
3
CO
CJ
to
01
OJ
00
E
CO
o
3
cc)
^f
CJ
CO
01
o
3
o
CO
4-t
B
•H
o
a
E
O
B
O
•H
4H
re
B
IH
O
UH
B
•H
O
Z
•
CJ
fH
<4H
«
OJ
01 rH
4H X
CO CO
OJ -H
73 re
o >
s: re
3-7
-------
•^
01
IH
4J
B
O
CJ
m
w
3
H
cy
cj
c
(U
0)
UH
(S
5s
C
r-l six O
3 CO C
JO C O
O •H
UH «H -u
O 4-» (fl
!>
to co
01 CO
CO 01 01
y > 4J
Q 01 -H
J CO 05
iH
CO
IH
E
01
•H
CO
cu
oi
(
E
CO
h-
3
"3
cn
CJ
to
E
U
•U
to
to
to
&.
CJ
3
u
vailable.
CO
CO
0)
3
0
CO
u
X
(D
>
(0
.
00
c •
-H U
TJ -H
•H UH
3 «0
.0 u.
4-1
UH
O 4J
>c
p. £•;
O *H
c
o •
r" •*
§ *
o
h e
00 .*
hi lH
UH
.*
B h
CO
r- a
•O hi
01 01
4J IH
CO iH
y CO
Q hi
•2 H
i-H
CO
«H
E
01
•a
•H
CO
CU
a:
1
E
CO
(„,
3
X
3
C/3
S
X
•H
3
§
5
CO
01
o
hi
3
0
to
4J
B
0
a
0)
00
hi
™
o
z
1*.
01
to
0
iH
1
•o
1*
3 XI
y h
E to
o oi
B E
CJ
X
o
z
I
00 •
B u
IH IH
•O UH .
IH tO l-l
3 hi X)
X) 4J to
UH X IH
O > CO
CO ^
a o< eo
o X
4J CO
0)
E • o
0 B b
^ 3
•a o
O i-H 4-»
00 E iH
•H O
1 u a
hi iH E
B to B
•00
00 CO -H
O 4J
•O hi «
CU S
-J T3 U,
CO 01 0
y > UH
O co E
iJ a -H
rH
CO
•H
B
01
•H
CO
01
cd
1
B
CO
XI
3
3
en
CJ
•
0)
0
B
01
tH
b.
CJ
CJ
X
00 01
E 4J
H IH
•O CO
•H
3 O
XI hi
UH
UH
0 B
B
E E
O -H
JS
3 a
O
00 -U
to
1 2
IH
UH «C
:!
T3 iH
01 CO
4-1 hi
CO 01
tl >
O 01
•J cn
CO
y
i^
01
B
B
O
CJ
1
E
CO
XI
hi
3
XI
3
"
g
*U
E
O
B
£
U
u
CO
CJ
E
0
B
O
•H
4-1
I
O
UH
E
•H
S
y
U4 01
hi eo
•H
£-, CO
to eo
01
jz to
01
o y
4J hi
3
01 0
4J CO
CO
h> 4-1
01 E
•^ IH
0 0
£ o-
CJ
X
o
H
00
E
•H
•D
iH
3
XI
O
o
0
B
0
•T3
3
O
00
s
Q^
0)
4>J
(0
u
,_,
CQ
•H
^j
4J
CO
3
*o
C
M
I
c
CO
IH
D
S
Ik
o
u
to
•H
y
B
CO
u.
B
CO
cn
CJ
b
tn
CJ
01
iH
XI
CO
•H
CO
01 >
4J CO
CO
hi CO
01 01
•o y
X 3
0
to
01 4-1
iJ5
» O
a
UH
0 E
O
B
.*! B
0
^i
B E
iH O
•o
CO
o •
hi y
•H
•O UH
01 UH
> co
CO hi
a. 4J
CJ
X
01
to
0
tH
y •
01
4J iH
• 0 XI
00 iH 10
E iH
-H 00 *H
•O E CO
iH -H >
3 hi
XI CO CO
Cu CU
UH y
O hi
4 3
"SIS
~~ E
B iH -H
0 0
B a
3 4J O
0 -H
hi S B
m from g
ed roads
nformatio
-a- > -H
CO
•o o. o
01 Z
4J 01
eo oi
U hi •
iH
CO
•H
y
01
Q
CJ
1
E
CO
XI
hi
3
CJ
CU
CO
0
*-)
E
CO
cn
2j
CJ
CJ
X
to
•c
§•
U-i CO
hi O "O
01 E hi
hi >49
3 M?
y > to
-£s 01
00
II*
a, co eo
01
T) E
• E
•O CO CO
B 4-1
3 /"-s cj
0 B 01
hi J«! iH
00 iH
•«• o
§1 °
A/
hi 4J
UH 4J tO
W g
iH « •
£ 01
"S tf S
4J O 10
CO B
y tl
o 4J x:
,J CO 4J
iH
CO
•H
y
hi
CU
Q
CJ
1
E
CO
XI
hi
=
8
p
O
4J
p
Q
Q
S
a
8
CO
01
y
hi
3
0
CO
.u
c
T-l
0
a.
4J
C
(0
u
f-l
UH
•H
C
•H
CO
£
UH a
CO 01
hi U
4J 01
hi
*•*
a a
cu oi
X B
^j
g"
4J
E
• CO
oo y
•H T3 UH
•0 B -H
•H CO E
*3 r^ -H*
XI B CO
UH o
o -a- z
a-i
o
4-1 4J y
CO iH
E CO UH
O 0) *+H
to
•0 - hi
E 4J 4J
3 to
0 01 .X
hi » >
00 to
to 01
§•0 x
CO
hi O
UH hi •
B -a 'B
CU .*
iH XI
-H hi SO
3 )
•O y >c,
01
4J T3 js
CO 01 4J
CJ > hi
Q CO O
,-3 OH B
iH
CO
iH
hi
4J
CO
3
•o
E
M
1
B
CO
XI
hi
3
8
;
o
CJ
hi
0>
x>
3
oe
to
01
4J
CO
o
CQ
O
8
hi
o
4J
a
cu
y
cu
hi
rces near
3
O
CO
4J
E
•H
O
a
^
4J
E
0>
E
M
O CJ
hi X
y
£
i
E E
x;
4J -a
•H J-l 0>
S e IH
«H *H
CO O CO
4J iH & 4J
O CO 01
iH O. U T3
C M O " O •
o E -H n z oi
•H *H E 'H ^ iH
4J .* 3 XI
co hi s: 'H co
4J tO CO • IH
4J 0) O to
01 -O > 4J >
u a> 4J tu a. co
•H > B cn cu
UH CO OJ y CO
a. y o) 01
UH B tO • hi y
O 3 1-1 CJ hi
UH CO UH O O
O • UH CO
O Ss « co B
0 .M E
B 0) 4-1 tM *H
o co a >, o
hi O 01 > E O.
UH iH y to CM hi 3 C
O CO ^. 'H O O
J CL ,H < CO UH
iH
eo
•H
hi
4J
to
3
B
M
1
B
to
XI
hi
o
cn
B"
o
4J
CO
tn
cu
hi
•H
b.
X
CO
UH
hi
•H
to
b.
5
§2
CJ
X
o
z
•
y
«H
UH
UH
hi
H
01 •
4J hi
CO O
oi a
•o 01
o u
£ 01
hi
• hi
T3 CO
3 B
O
hi CO
m from g
nt source
ft IH
01
4J 01
CO 00
y hi
32
i-H
CC
CJ
IH
g
CJ
1
B
CO
XI
hi
3
3
O
S
„
0)
i— I
•H
to
^
3
o
3-8
-------
TJ
QJ
3
C
4-)
O
cj
rH
cn
CQ
H
OJ
CJ
C
OJ
0)
QJ
$
QJ
t-i
CO
CU
4-1
•H
CO
C
o
ex
•H
J-i
CO
QJ
O
OJ
4J
•H
CO
C
O
•H
4-1
cd
CJ
U-I
CO
CO
TO
rH
0)
E
Z
QJ
*H
CO
D
QJ
CO
"
c
QJ
C
o
L-i
U
£
t-<
*H
U-i
O
U-i
o
o
c
o
TJ
C
3
O
^
00
E
O
i
E
O
rH
TJ
01
TO
0
rH
•H
C
QJ
"O
CO
QJ
PC
1
c
CO
X3
JH
o
c
o
"^
TO
•J
QJ
^•i
•H
CC
co
o
00
C iH
-H TO
t-l 4J
CO C
ex QJ
•o
QJ CO
> CD
a
C Ql
3 CO
C
xi a*
4-J TJ
•H
3 C
o
CO
TJ O
TO -H
O UH
t-l UH
TO
TJ J-i
QJ 4->
X3
3 >
O TO
QJ
TJ ffi
QJ
TO •
e
3
• 0
C I-
O TO
4-1 CO
4-» O
CO rH
O
TH
^
4J
o
rH
QJ
l-i
t-i
CO
3
cr
OJ
TJ
3
rH
CJ
•H
CO
CJ
1-1
3
O
en
4-J
e
o
fD
QJ
t-i
CD
l-i
O
0)
CJ
OJ
E
o
I-I
UH
E
^C
o
_)
-a-
£
^"
4-J
w
QJ
x:
4-j
0
CO
c
ex
g
o
u
u
sc
QJ
OJ
J-i
"5
CO
UH
O
"4-1
O
o
1-1
TO
C
o
c
3
O
00
E
o
IH
\O
•u
QJ
TO
O
rH
TO
p
*j
CO
TJ
C
r-t
1
C
TO
X>
t-i
Z
"
CM
=it
en
<
CQ
Z
•
4J /*N
K 5
co -a:
01
0
CO CM
CO l-l
0 I
J"l >— '
TJ x:
QJ *J
1- O
3 C
u
13
•o c
QJ TO
TO "^
P- E
• O
00—1
C ^
•S 7
rH ^
•H
3 4J
rH 5
O
o -
X! /""s
U B
hi ~v
^_J J
CO ^
01
•o
3
O
c
•H
01
CJ
l-l
3
O
CO
^J
c
0
PH
«
CO
TJ
TO
O
QJ
QJ
TO
1-*
4-1
rH
QJ
TO
QJ
O
^
*j
CO
TO
OJ
x:
3
O
CO
TJ
c
TO
4-J
CO
OJ
>>
c
TO
E
O
CJ
C
0
•H
*
OJ
jj
I-I
O
*J
ex
QJ
CJ
01
0
l-l
/- S.
E
LO
O
1
CJ
OS
^1
o
to
OJ
c
o
CO
UH
O
ex
0
4-1
C
o
13
C
3
O
00
E
o
t-l
(4-1
E
CN
T3
OJ
TO
O
TO
4j
C
QJ
TJ
CO
01
&
I
C
TO
!_,
3
Xl
3
CO
OJ
00
TO
3
01
en
i-i
0)
4-1
CO
CO
00
CQ
CC
CQ
£
E
CN
,__,
)
4-J
CO
OJ
3
4-1
CO
*o
TO
0
TJ
QJ
X>
l-i
CJ
TJ
01
TO
DM
•
00
C
rH
3
5-1
«
C
TO
PH
_J
c
Ql
TO
OJ
t_i
H
QJ
4J
E
O
U-i
E
CN
rH
)
4-J
CO
TO
OJ
.TJ
C
TO
„
E
^H
s— *"
4->
O
CO
4-1
C
CO TH
4J O
o a
jj
00 C
C TO
TH 0
t-i U-I
TO **H
a c
00
QJ CO
TO O
az
x:
4-J •
•H QJ
3 4J
CJ CO
£ UH
UH o
TO
t-i ^
4-1
•H rH
^~"'
QJ CO
•H 3
co TO
ti
O
ex
01
CJ
01
I-.
s_j
TO
QJ
C
CO
QJ
JH
O
CD
SC
t*
o
4J
CO
3
0
U-i
cO
O
UH
O
O
IH
C
O
C
3
O
00
E
O
IH
UH
E
CO
QJ
TO
O
CO
^_,
4J
CO
3
13
C
M
1
C
TO
i_i
z
*>
00
CN
*R=
CO
DM
rj
CQ
Z
I
"5
l-i CO
O ID
C TO
O
TJ t-i
C
TO T)
OJ
4-J OJ
M >
O TO
C J-i
4J
4-1
TO >s
rH
CO *H
TO TO
O OJ
tH PC
TJ
QJ •
XI QJ
3 -H
CJ CO
C
O £
C 0
TJ U^
OJ
TO E
p-l ^
• "-x^
00 i-i
c ^
rH 4-J
3 TO
CO
QJ
O
t-i
O
CO
4-)
p;
TH
o
CO
C
o
4-1
CJ
OJ
U
•H
TJ
,_!
i— 1
TO
C
E
C
S
3
O
4-J
^
rH
I
4_)
CO
0)
3
>>
C
TO
ex
g
o
o
c
o
^
c
CO
rH
Q)
4-J
CO
TJ
rH
"
«
e
CSJ
)
^i^
4J
01
3
4-1
^4
O
c
^,
CO
a.
E
o
0
c
o
•H
4J
CJ
3
O
a
01
o
CJ
'E
•^
QJ
f
LH
O
^^.
E
n
o
4J
CN
x:
IH
O
C
^
J-i
CU
c
•H
UH
QJ
TO *
QJ
»T!
.
00
C
TH
TJ
rH
•H
^
IH
O
4-J
CO
CD
c
o
TO
0
O
c
3
O
i-l
00
p
tH
UH
E
CM
01
CO
^
rH
TO
•H
4-J
CO
3
TJ
C
rH
1
c
TO
jj
3
z
c
o
4J
TO
4J
a.
E
3
PH
x:
3
E
•H
3
O
CQ
Z
CO
C TO
O
•H OJ
4J TJ
0) rH
IH CJ
•H C
rH CO
f-H QJ
TO CJ
C 3
TH O
CO
E
^ 4J
>J *H
"•^ O
c
•H •
X X
*J JD
•H 1^
3 co
o>
tn c
•o
CO CJ
O -H
hi U-i
u^
t3 CO
CU hi
J2 -U
hi
3 JS
0 CJ
D
01
CO o
ft. Z
^
o
c
i«e
4-J
CO
QJ
K
4->
CO
QJ
3
X
3
O
CO
c
TO
E
0
CJ
O
IH
TJ
C
TO
CU
CO
TJ
OJ
4-1
TO
CJ
O
rH
&s
TO
ex
E
o
CJ
QJ
0
TO
3
fa
QJ
4-1
CO
O
IH
^
E
•*^
rH
I
4-)
OJ
3
i^
4-1
£X
0>
O
0)
I-i
UH
O
E
en
o
4-1
CN
,>
^
CO
1
4J
J-i
O
*
4-J
o
c
4_j
rj
a;
c
I-H
0
I-I
CJ
•H
X
UH
o
o
o
M
C
O
TJ
C
3
O
tH
oc
E
0
i-l
U-i
E
^j.
TJ
0)
4-1
TO
CJ
O
rH
O
.J
QJ
QJ
TO
IH
3
rH
3
O
l-i
00
1
rH
TO
3
c£
O
„
C
TO
CO
CU
-H
3
TO
C/)
<
£
C6
O
QJ
•H
4-1
•H
00
3
CO
TO •
QJ CJ
x; -H
4-J UH
tn UH
o re
C IH
4-1
TO 4-J
TJ 00
re -H
O rJ
M
QJ QJ
> 4-1
TO *H
ex co
a
^ 0)
x:
4-1
J-i UH
QJ O
rH
•H /— -
TO E
4-1
QJ -^
•H ,)
4-J
c
o
a
0)
00
TO
1— 1
o
TO
QJ
l-i
TO
0)
XT
4J
QJ
N
s-t
OJ
4-J
CJ
TO
TO
x:
o
4->
'
j_,
O
a
0)
CJ
OJ
IH
TO
0)
C
CO
cu
CJ
o
CO
4-1
c
QJ
C
rH
O
l-l
CJ
£
QJ
•H
UH
UH
O
O
O
C
O
TJ
e
3
O
tH
60
E
O
E
rH
TJ
QJ
4J
TO
O
O
rH
.
TO
• TO
CO
C E
O fii
*H
CJ -^.
QJ rH
VH C
TJ O
i-H 4-J
rH O
TO 3
l-i
C 4J
•H CO
C
TJ O
TO CJ
O
1- 0)
t~t
TJ CJ
QJ TO
1 S
CJ O
QJ •
> CJ
re *H
UH
TO
O C 4J
4-* O
01 4-J >
«
rH
TO
•H
o
IH
0)
e
CJ
t
C
re
x
IH
°
0
c
o
•H
4J
TO
4-J
C/3
0)
l-i
rH
TO
4-J
C
QJ
O
CO
o
a:
o
IU VU
4J QJ
CO PS
^
Xl
re
QJ
CO
QJ
CJ
3
O
CO
^
c
•H
O
a
cu
00
TO
rH
o
Z
4-J
c
OJ
c
hH
o
CJ
•H
UH
O
0
o
c
o
TJ
g
3
O
60
E
0
IH
E
rH
TJ
QJ
+J
re
CJ
o
rH
i-1
O
4-1
ex
OJ
CJ
Q)
*
rH
re
4-J
C
0)
TH
CO
QJ
Pi
I
c
TO
S-i
3
3
CO
PH
^
O
O
CJ
CO
00
•H
X
rH
re
•H
0)
ex
i— i
<
(-H
PH
2
rH
TJ 4J
C U
re QJ
t-i
E TJ
,*/
TJ
0 C
rH «
j e
re *-*
OJ )
> *L
>H x:
TJ -U
3
OJ CO
i-> CO
3 TJ
CJ TO
O
TJ *H
QJ
^ *O
TO QJ
P- Xi
3
• CJ
O QJ
o re
co ex
4-J
C
TO
0
UH
•H
C
00
-H
CO
o
z
•
o •
TH hi
u-i o
CO CL
hi GJ
4-1 CJ
OJ
-J hi
00 hi
•H CO
|-J 01
c
• CO
o> cu
4J O
en 3
o
U4 0]
0
4-1 C
CO O
01 O.
3-9
-------
'•N
1
E
O
CJ
m
rH
W
03
H
01
O
C
OJ
kl
01
01
B.
0?
kl
3
CO
0)
4J
CO
B
0
•H
4J
O.
•H
kl
a
CO
01
a
OJ
4J
TH
CO
E
c
4-
co
c
u-
CO
CO
ft
CJ
01
B
Cl
Z
01
4J
^
Crt
C.
0
CO
4J
B
01
B
O
kl
CJ
•H
£
CO
01
u
kl
OO^ 3
E B o
•H Ji CO
rH CN 4-1
•H ->. E
3 rH TH
.0)0
^ a
CO
CO OJ
UH E 00
O O kl
•rl CO
UH 4J rH
0 CJ
o oi o
IH ki Z
TH
E -0
O •
rH CJ
TJ rH TH
C CO UH
3 UH
O B co
kl TH IH
00 4J
CO
O CO >
ki O CO
PC
S T3
CN 43 •
rH k1 OJ
3 4J
TJ CJ TH
01 CO
4J TJ
CO OJ E
CJ ^ O
O CO kl
rH
CD
•H
CJ
IH
QJ
Q
CJ
1
6
to
ki
BH
TJ
CO
O
CO
•
Q
IT)
HJ-
OH
S
kl
O
a.
01
CJ
OJ
ki
ki
RJ
01
E
4-1
E
01
E
M
O
kl
CJ
•H
£
1
kl
4-1 O
E E
OJ
E CO
01 >\
00 CO
E 13
CO CO
£ o
ki
ki
TH -0
kl CO
00 OH
1 -
IH ki
UH 01
rH
3 co
ki
S 4J
vO -
CO
•U 01
OJ O
CO >
CJ kl
O OJ
rJ CO
rH
CO
TH
CO
3
"O
E
rH
1
B
CO
O
OH
^
E
01
-C
00
01
rH
<
as
-C
OH
BH
^
^
CO
01
3= rH
• CO 00
OJ 4J E
4J • QJ o B
-a- E -a- ji
4J "-^ TH ^s.
CO rH fl •»
co ; co ) ~-
01 vi' 01 •*-* rH
-C CJ 1
4J 4J kl 4J "^
kl (0 3 CO
O 01 O OJ 4-1
B S co 3 co
~*.w,3g
E co E ki £
Ji TH O 4-1
•0 O E 3
O CO BH O
rH O TJ CO
rH • CO >*
) TJ CJ X 4-1
*±s 0) TH TH
4J CO UH CO TH
CO CL CO kt CJ
01 E ki CJ co
3 S 4J CO UH
o
J
CN
^
,-H
vi*
.E
4J
3
O
CO
o
4J
4J
CO
OJ
4J
3
O
CO
00 kl
c o
•H 4J
TJ OJ
C d
CO 01
-E ki
^j
E
01
E
0
ki
CJ
TH
£
CO
TJ
CO
O CO
kl O
• CJ
•O OJ kl
01 u 3
-0 -H 0
kl CO CO
3
O 01 4-1
1 -E B
E 4J TH
O O
E UH a
o
TJ OJ
OJ •** 00
> S k,
CO Ji CO
.Is?
TJ
3 CO •
O CO ^
ki 01 -a
00 ki •
TJ CO kl
§B 01 O
CO B 4J
UH ^ CJ 01
S TH CJ
UH kl
in CN eo
) kl O
TJ six 4J J
01 4J
co 4-1 £ ki
O 3 00 CO
O O TH 01
J CO J C
,-H
CO
•H
E
01
13
TH
CO
01
03
1
C
CO
-Q
kl
3
3
CO
4-1
B
CJ
CO
OJ
H
01
4J
CO
&
4-1
B
O
B
01 <
CO BH
U
X
BH
£
4-1
C
01
E
0
CJ
•H
£
E
4J O
E TH
TH 4J
0 0
I 13 a 3
• 01 B • B kl
.£> 4J O .—N CO 01 4J
CO •H CJ B 00 CO
rH -CO Ji 4J kl C
CO E co co O
UH O ^3- OJ rH CJ
OJ O O -~^ S
4J 00 rH -C O UH •
OI4JCO) 4-lZO-O
kl CO rH NS^ 3 O
O CO O CO TH
O -C CO 01 O 01
O4J-r)COC04JrH&
3 E 01 13 TH
EOS 10 CO 4-1 00
O to O 4J O 3 C
kl B kl S -0 TH
13 13 00 CO O rH
E E rH T3 kl - a
ScOEC-OIUHkiB
O OJ > O CO
ki -B (X 01 to s^> 4-1 CO
oo .j o oo a a c.
3 CO Ji 01 OJ
B 0 0 » C CJ -E
OCOCQOJOCNQJ4J
UH CO < CJ rH 00
13 4J TH j kl E
O • «-i 0) M
5 O M CO T3
•X3 "O ..--. *H 4J 4) CO
at a> -u » at >>
O O« <^ w W ^ 3 tO
o c j 4J oj o o a;
1— t
TO
•H
1-4
.u
CO
3
•a
c
M
1
c
CO
Iw
ex
fc
4-)
c
«
iH
IX
CD
&0
TO
2
CO
•
t
CO
o
BH
BH
4J
B
OJ
E
M
o
IH
CJ
•H
£
CO
0) CJ
kl TH
CO UH QJ •
UH j; IH
00 CO 4J O
4J E kJ 4J
E -H 4-1 UH a
01 13 O 01
§B 01 CJ
X CO 4-1 ^^ QJ
00 4-1 rH CO S kl
CO 3 IH Ji
E O ki OJ ki
CO CD 01 *O stf- to
S3 4J o ^s. OJ
13 CX. £ rH C
IH CO O *-s
•H O CJ CO
•< ki TH . je oi
rH OJ 4-1 CJ
C rH 01 4J 3 kl
O OJ X TH O 3
> CO CO O
•on en
§kl • UH 4J
CJ 01 O O 4J
O 4-1 rH E
IH TH z
TJ OJ Ji CO CO
01 CJ OJ O.
CO > — 4J js 0)
O 01 ? O TH TH
rJ CO >•> CO ft CO
rH
CO
TH
E
OJ
TJ
•H
CO
OJ
OH
1
E
CO
XI
ki
3
3
CO
BH
^J
^J
0
a
£
HgJ
•
•
Z
U
BH
OH
CJ
as
rH CJ
rH 00 T<
CO E U-l ^
TH U.I S
E Ji Cll Ji
TH kl kl
CO 4-» CN
-C TJ O.
O E 4.1 iH
ki 3 T3 -c: >
3 0 01 00 -t-
-C kl > T-l
y w ra _! x
OH E
UH CO CO
0 tJ .• a
co • oi B
a. o 01 4.1 o
O IH 4-1 T-l CJ
4J *r-t tf)
13 CO rH
B 01 u-i co
0 > B 0 4J
CO O OJ
•o a ki ^- E
E UH B
3 13 Ji! CO kl
O 01 ^ OJ O
kl •£) E *sT 'O 4J
oo ki ji ^.. 3 a
3 r-l rH 01
§CJ -3- ) CJ 0
— vi' E 01
UH CO ) 4-1
OJ ^ CO OJ UH
S rH CCI CJ O
CJ CO 0,1 kl
ul E X 3 4J
O 4-' O CO
TJ • TH k Cfi OJ
01 01 4J O ft
4J 00 CJ E 4J -E
co co oj e -J
O co TH o O O
rJ 00 TJ r-l OH E
rH
10
•H
kl
4-1
CO
3
TJ
E
rH
1
a
CO
ki
°
OH
w
j:
o
ki
3
U
CO
C
-C
o
4J
CO
rH
S3
BH
BH
CJ
X
^-^
S 4-1 «
Ji O
4-1 rH QJ
j: o 13 cu
00 rH 00 3 X
•H ~«. C rH 4.J
01 rH TH CJ
k, J Ji E C5
CO CO
4-1 4J G. CO 4J rH •
O CO OJ X * CO kl
THOj'Skit-'i'ojlS
BH x > 3 ' ji B a
4-1 CO O 4J 01
UH 1J OH CO-C:-3-TJ CJ
O O 00 — E 01
C 4J -H rH tj kl
a • E IH )
O CO OJ TH si> .« UH
4J TJ *J O s"^ "^ O
CO TH BH E: 4J E
E O CO J.! CO Ji s~.
O ki co B
UH • O OJ CN Ji
.TJ O rH --,
TJ 01 01 j TJ rH -3-
3 kl B TH CO *•-> rH
O 3 Ji CO 4.1 >* )
IH CJ CO 4J -±s
00 s3- UH QJ CM CO
- — O 5t CO CO 4-1
§TJ rH j: kl OJ CO
01 I •-*• 4.J CJ -C OJ
UH CO Ji O ki -E
OH 4J E kl O W
01 — .ED. O
O -C ( «H O OJ
01 IH ( O E
cooE-JcJoioia
O 4H 3 3 4J Ji B
OOOOki -HOO
rH
CO
TH
kl
4J
CO
3
E
M
1
B
CO
kl
a
4J «£
-E OH
00
TH *
01 •
kl O
rV. CJ
4J kl
0 0
rH 4-1
•H 0
OH £
r^
X
BH
BH
3-10
-------
oi
^
•In
d
o
u
m
rH
rJ
S
H
(U
O
d
QJ
JH
QJ
0>
QJ
IH
3
cn
4-i
*H
d
0
•H
4-1
CX
•H
0
CO
o
QJ
4-J
cn
d
o
*H
4-1
re
o
•H
UH
•H
CO
CO
re
i-H
CJ>
QJ
B
cu
4-J
•H
cn
Q
cu
*J
•H
cn
EC
•
/—v
• e
CU rC
4-1 4J
B TH -3-
01 CO — -
B *-H
4-1 0) )
to £ ^
tU 4-1
VH to
4J UH B
0 0
UH *H
0 n 4-J
O. A! HI
O rl
4J 0") TH
C i-t
0 ) rH
vi- rH
E 4J
3 CO C
O tO TH
VH CU
00 ID
§tO to
O O
V4 rl L4
UH
B >H tu
•rl J2
in Q y
•o o
cu •
4J 4J TJ
tO C QJ
O iH tO
iJ aOn
to
TH
rl
4J
cn
3
•a
c
M
1
C
to
jO
rl
OH
CJ
PH
3
u
Z
1
PH
OH
^
s
;*
ro
4-t
CO
cu
3
J3
4J
O
d
m
I
M
O
•H
UH
JH
4-1
CU
4->
re
o>
0
S
CO
• t— i
CO • CO
re oo 4-t
QJ d QJ
(H -H E
CO i-H
cx*o
QJ re re
CO CO
O 00
ex Q) d
•X rH «H
01 *H 4J
pC i-H
B S QJ
O B
IH 4-1 CO
re in
CO CX CX
3 CX
til OJ O
00 O
QJ cfl
> 3 o»
•H 01 TJ
4-1 CO 3
*H rH
60 UH O
3 0 C
[>H -H
O to
tU 4J O
4-1 O (H
•H 3 3
to IH O
CU CO
J£ C 4J
UH CU O
o « OH
01
4-J
O
^-x
*s
^
T— 1
Xi-
co
cu
3
£
4-J
3
O
CO
•0
B
to
£
3
o
CO •
S»i O
E 4-1
CO £X
a cu
B CJ
O 01
CJ rl
B
cu
•O 4J
cu
> UH
to o
PH
S
ooo
14 -^
tO rH
oo I
UH
O 4J
to
a to
O 01
4_J ^
C 3
O O
cn
TJ
C CO
3 TJ
o to
IH O
00 IH
§C
CO
V4 CU
UH T— 1
B
tu
CU VH
4J 3
CO U
J I
iH
re
•H
ti
CO
3
•a
d
M
i
d
CO
in
<;
O-i
4-1
01
IX
H
EC
(X
ex.
4_|
CO
OJ 0)
*"O * 3
3 s~*- X
O cAi JH
d o
•H i-H d
CO ^ 1-0
QJ
O X X
3 3 IH
OOO
CO CO C
-U 60 CO
d d cu
•H -H -H
O rH p
d cx
re s
• _c o
o o
•H rH
UH « iH
UH IH CO
re cu o
In d *H
4-1 -H E
B 0>
0) JZ
4J » O
re so
0) -H
•U TJ CO
o re i-i
X o re
i-H 4-1
0)
• d B
0) -H
co oo re
•
IH
O
4-)
OJ
o
cu
IH
01
.c
o
rC
60
r*"i d
• 0> 4-J *H
IH D .C «H W
0) 4J 4-J i-H rH
CO CO UH U B cj
d o ca co in
re 0) UH i4
01 ^ d CX OJ IH
Ci3 3 UH *H CXrH O
tS rC O CN T3 O CU J-1
4J "^. CO O CX
UH iH '""•« rH O d QJ
d ^ N!- oj
O CO >J X O 13 d
re rH IH IH rH JS
d o ) o H* o '"*> w
O IH ^ d d B
•H _^ UH
*O *O W "O • O
O VH 3 re UH O rH S
60 O 4-1 O CO 3 ^
iU rH V-l O " I
« 4J VH Cfl iH X
^^-vT3 CUO-into
o B QJ *D o re
rH • ^ > O d QJ
oo co SS •
QJ M \ d 4J d d
4-ITJrH3 «iHfljre
re <-H cu co cx cx
(jjo^ocoreoo
tH
CO
VH
4-1
CO
3
T)
d
FH
1
d
re
in
CU
jj"
OJ
UH
CO
d
re
IH
H
2
i
(X
Pi
EC
• CO
-C • 0) QJ
60 -C 0) 4J IH
IH 01 4J «H re
r-i |H CO T3 _d 4-J IH
CO 3 0) 0) 4J CO O
•H O 4J CX CO X 01
IH d UH X 4-i O
* O d O JH
t* S ^ IH S !>-i .C
a re ^ UH o d 4J
OP- CN o re
•^ rH CO 01 B O
O 00? ^"OO O'^
C ^ OB
C <~H C tO TH CJ fl
3 TH O 0) 4-J CU TH
^4»434-J4-J 003QJ4-J
00 O tO 3 iH *B
IH 0) fc CJ O CN
§CU IH ^ C ?
4-t TH C TH (0 ^
UH cu o cj cn *o 4-J
S rH 00 UH O to CU
C CO UH IH ^
in O >i co 3 ^ X
cu to a. o
4H cu cn 4J 4J H IHOrH'CU 'tO
t-H
CO
•H
J-j
4J
CO
3
TJ
C
M
1
C
to
.c
IH
CJ
J
CU
00
IH
3
CO
01
•H
M *.
rH TJ >
TH OJ CO
3 W OJ
* to B
o
rH O
O ^
f to B
U 4*1 Jti
CO O
rH 00
UH )
O tt> -is
C
a TH .c
O J<1 4J
-JUS
CO O
o
TJ T3
- 0) C
TJ > tO
C to
O B B
IH 3 J
00
e C iH
o to ;
UH CO
B to 4J
0 IH
VD IH O
C
0) tU >,
4-J > rH
CO tO 4-1
OOO)
f-J -3 IH
to
VH
4-J
to
3
TJ
C
I
B
to
°
X
H
rH
C/J
o
PC
o
X
H
B s~^
0 B
•H .y
O 00
3 /
T3 ^
JH 4-1
a co
re
QJ OJ
B _n
•H 4-J
rH 3
0
Q) CO
TJ
3 CO
I-H d
O QJ
d >
•H O
CO QJ
oi _*«;
o o
JH O
3
O -
CO '"N
B
4-1 _^
^H CX)
O /
DJ vi/
4-)
• CO
o re
•H 0)
UH ,C
UH 4J
jo 5
JH O
4J CO
Q)
4_)
O
s
rH
si^
4-J
«
QJ
4-1
3
0
CO
60
•H
i-H
d
jr
c •
•H IH
cfl O
JH 4-)
60 CX
0)
•°3 O
£ 0)
Cfl IH
EC
QJ
* 4-1 -H
4-J CO 4J
a re -H
O 3
•a EK
.c d
rH •
re *^N o
QJ E -H
ffi _« UH
UH
UH CM CO
O *-^. IH
rH 4J
CX )
O ^ QJ
4-) 4-1
.C CO
O JH QJ
0 T3
-do
T) S
d cn
3 TD
O CO -
IH O 0)
60 in 4-t
§T3 Cfl
CU
JH > QJ
UH re .c
£
UH
r- - o
60
*o d '"^
QJ -H B
4-1 *\3 ^
CO i-H
O 3 )
r-3 ,C \i/
rH
re
•H
o
IH
QJ
B
B
O
C_J
1
d
re
^
X
H
o
CO
re
Cu
i-H
u
3
W
X
H
QJ
3
tH
O
d
•H 4J O •
Cfl *H LJ
co re 4J o
QJ rH >-, 4-J
O XJ rH CX
3 T3 4-1 O
o re re QJ
CO QJ O IH
4-1 'O 0)
d - -H x
O B I-H
Cu j^ UH B
O
• ^ d UH
re ^— ' re
QJ ^v
in U x-v £
re cn E ,y
CU ^
QJ 3 00
43 ^ in >
-J ^^
CO O *J
OJ d 4J CO
N co re
•H oo
-------
concentrations at each site. Surveys of the remaining 39 sites should
be planned before the termination of the IP monitoring program to
complete the record.
Table A.I of Appendix A contains other information about the
sites. The elevation above sea level, the EPA SAROAD identifier, the
UTM coordinates, and the sampler height above ground level are in
columns 6 to 10. Columns 11 and 12 present the inclusive sampling
dates from which measurements were drawn for the majority of the work
in this report. Because this amount of data was insufficient for some
purposes, data which became available later in the study were added to
this set. This will be noted when the data are presented. The* final
three columns denote the presence (1) or absence (0) of HIVOL,
HIVOL(SSI), and dichotomous samples, respectively, for the designated
period.
3.2 Filter Media
In recent years certain filter media have been found to bias the
measurement of suspended particulate matter. This section examines
the extent of those biases in IP filter media. Special attention is
given to the change of SSI and HIVOL media between 1979 and 1980 and
the effect this change could have on ion mass concentration
mea sur ement s.
To adequately quantify ambient mass concentrations of suspended
particulate matter, the filter media selected for sampling should have
collection efficiencies greater than 95 to 99% for all particle sizes,
should have tensile and mechanical strength to withstand sampling,
analysis and transport, should present a low enough flow resistance to
its sampler that an adequate aerosol deposit can be obtained, and
should be free of interferences and variable blank concentrations that
might affect measurements to be made on the aerosol deposit.
The fiber filters used for HIVOL and SSI sampling and the Teflon
filters used for dichotomous sampling generally meet all of the
criteria except for that of blank levels and interferences.
Stevens et al (1978) have evaluated the Ghia Teflon filters for
blanks and interferences and found them generally acceptable for the
IP Network measurements except for the possible loss of nitrate. The
3-12
-------
glass fiber filters are another story. Two types of fiber filter were
used in the IP Network; HIVOL and SSI samples in 1979 were taken on
Microquartz fiber filters (Walling, 1981) while those after 1980 were
taken on Schleicher and Schuell (S&S) HV-1 EPA grade glass fiber
filters with an organic binder (Rodes, 1981).
Sixteen samples of blank quartz filters and two samples of blank
S&S filters of the same type which were used in the IP Network were
submitted to automated colorimetric and carbon analysis as described
in Section 3.3. The blank values for sulfate and nitrate
concentrations for Microquartz filters used during 1979 were .08 and
2 33
.03 ug/cm (equivalent to .02 and .01 ug/m if 1600 m are
2
sampled) and .34 and .05 ug/cm for S&S filters (Clark, 1981,
Walling, 1981). The total carbon and total organic carbon blanks for
quartz filters were measured by the authors of this report to be
2 3
7.5 + 3.5 and 5.2 + 2.2 ug/cm (1.9 + .9 and 1.3 +_ .6 ug/m ). The
total carbon and organic carbon measured on two S&S filters were
186 + 23 and 118 + 10 ug/cm2 (47 + 6 and 30 + 2.5 ug/m3),
respectively.
While the sulfate and nitrate blanks of both filters are lower
than most ambient levels, the high levels and variability of the
carbon blanks make the S&S filter unacceptable for carbon analysis.
The quartz filter carbon blank is variable, but it is lower than most
ambient carbon levels. The quartz blank could probably be reduced by
pre-firing the filter before initial weighing. Because the organic
binder is an integral part of the the S&S filter, pre-firing might
damage it.
Also of great concern is the possibility of artifact formation of
sulfate and nitrate; not only would these artifacts bias sulfate and
nitrate concentration measurements, they would also bias the mass
measurements. There is substantial evidence for sulfate and nitrate
artifacts on commonly used HIVOL filter media. It has been shown that
filters can adsorb SC>2 which is oxidized to SO,. Thus sulfate
measurements on the filter represent both ambient sulfate and artifact
sulfate. The factors affecting artifact sulfate formation have been
reported to be filter alkalinity, SO concentration, humidity,
3-13
-------
temperature, and volume of air pulled through the filter (Coutant,
1977, Pierson et al, 1980). Coutant estimated the range of normal
sampling error due to SC>2 adsorption on a basic filter (like glass
2
fiber) to be between .3 to 3 ug/m . Shaw et al (1981) found that
mass and sulfate collected by the SSI were higher than simultaneous
TOTAL mass and sulfate measurements. They hypothesized that the extra
mass was due primarily to sulfate artifact and, to a lesser extent, to
the combination of the positive and negative nitrate artifacts.
Nitrate sampling errors are primarily due to gaseous nitric acid
in the atmosphere and have been shown to be substantial for glass
fiber filters (Spicer and Schumacher, 1977; Appel et al, 1981b;
Meserole et al, 1979). The use of an inert filter, such as Teflon, to
minimize the positive nitrate interference is prone to a negative
interference resulting from the volatilization of HNO- (Appel et al,
1981b, Pierson et al, 1980).
Filter media may not be the only factor influencing the reported
nitrate concentrations measured with IP Network samplers, however. In
a field test of four samplers in California, Wendt and Torre (1981)
found that the particulate nitrate collected by the SSI averaged
3
3.8 ug/m less than that of the corresponding HIVOL samples,. It was
hypothesized that the observed difference was due to loss of gaseous
nitric acid on the aluminum surface of the size-selective inlet.
Analysis of washings of the aluminum surface accounted for 92% of the
missing nitrate (or HNO vapor). They found no difference in
sulfate concentrations between the SSI and HIVOL samples. Appel
(I981a) is currently investigating the potential of artifact: formation
for different filter media used by EPA.
The IP Network measurements offer the possibility of evaluating
the difference in artifact formation properties of IP Network
filters. Because the data are sparse and preliminary, this evaluation
must be considered illustrative rather than definitive. It does
provide a motivation for further study of IP Network filter mdeia. It
also raises some cautions about the interpretation of sulfate, nitrate
and even mass measurements made in the IP Network. For this report,
the IP mass concentration data were separated into two subsets, 1979
(quartz) and 1980 (S&S), to evaluate the changes which might: have
3-14
-------
occurred because of the switch in the filter medium. The geometric
and arithmetic averages, standard deviations, and number of samples
for the ambient aerosol concentrations as measured by HIVOL, SSI, and
dichotomous samplers are given in Table 3.2.1. The cumulative
frequency distributions for all measurements at all sites were also
plotted to see if any changes were evident before and after 1980;
these plots showed the 1980 distribution shifted several ug/tn3
higher than the 1979 distribution for HIVOL and SSI, but not for 1979
and 1980 TOTAL mass concentrations.
As seen in the Table 3.2.1, the COARSE average as sampled by the
dichotomous sampler remained the same from the 1979 to the 1980
subset, while the FINE average decreased. This is reflected in the
TOTAL arithmetic average which decreased from 47 to 43 ug/m3. The
HIVOL and SSI arithmetic averages increased from 70 to 71 ug/m3 and
54 to 58 ug/m3, respectively. Though these increases were not
large, the trend was opposite that of the dichotomous sampler.
A rough estimate of the additional average mass collected by the
S&S filters over the quartz filters is tabulated in Table 3.2.2. The
assumption is that if the quartz filter had been used in 1980, the
ratio of average SSI (1980) to SSI (1979) and average HIVOL (1980) to
HIVOL (1979) would be the same as average TOTAL (1980) to TOTAL (1979)
ratio. As seen in the Table 3.2.2, the average mass concentration
measured on S&S filters seems to be higher than that measured on
quartz filters by about 7 ug/m3 for both SSI and HIVOL. This
amounts to about 11% additional mass measured for HIVOL and 17%
additional mass for SSI with the use of S&S filters. The slightly
larger additional mass for SSI, if it could be attributed to artifact,
might be due to the lower flow rate for the SSI (Coutant, 1977) as
compared to the HIVOL (the SSI flow rate is 40 cfm while the HIVOL
flow rate is 50 cfm) which would make the relative increase of the SSI
average mass higher than that of the HIVOL mass if the S02
adsorption reached saturation during the sampling period. Thus, mass
measurement comparisons show that, on an average, 7 ug/m3 of
additional mass was collected in 1980 over that collected in 1979.
3-15
-------
TABLE 3.2.1
COMPARISONS OF ARITHMETIC AND GEOMETRIC AVERAGE MASS CONCENTRATIONS
FOR HIVOL, SSI, TOTAL, FINE AND COARSE
TAKEN IN 1979 AND 1980
Geometric Average (ug/tn3) and Arithmetic Average (ug/m3)
Standard Deviation and Standard Deviation
Category
a
HI VOL
SSI'
TOTAL
FINE
COARSE
Avg
58
(
47
(
40
(
22
(
16
(
1979
Std Dev
2
l,041)c
2
316)
2
714)
2
714)
2
714)
1980
Avg Std Dev
60 2
(1,046)
51 2
( 455)
37 2
( 795)
18 2
( 795)
16 2
( 795)
19
Ayg
70
(1
54
(
47
(
26
(
21
(
79
Std Dev
44
,041)
30
316)
31
714)
18
714)
18
714)
Avg
71
(1
58
(
43
(
22
(
21
(
1980
Std Dev
42
,046)
31
455)
22
795)
12
795)
16
795)
aSampling done using quartz filters for 1979 subset while S&S
filters were used in 1980.
"Dichotomous samplers employed Teflon filters throughout.
cNumber of measurements in average.
3-16
-------
TABLE 3.2.2
COMPARISON OF MEASURED ARITHMETIC AVERAGE CONCENTRATIONS IN
1980 ON S&S FILTERS TO THOSE PREDICTED
FROM 1979 MEASUREMENTS ON QUARTZ FILTERS
Arithmetic Average
Concentrations in uj*/_m3
HIVOL SSI
1980 Arithmetic average 71 58
from S&S filter
measurements
1980 Arithmetic average 64 49
predicted^ from 1979 quartz
filter measurements
Difference 7 9
aObtained by multiplying the 1979 averages for HIVOL and SSI by .915,
the ratio of arithmetic average TOTAL in 1980 to TOTAL in 1979.
TABLE 3.2.3
COMPARISON OF 1979 and 1980 ARITHMETIC
AVERAGES OF MASS CONCENTRATION RATIOS
SSI/HIVOL
TOTAL/HIVOL
TOTAL/SSI
Ayg Std Avg Std Ayg Std
1979 (Quartz) .734 .124
(293)a
.783 .158
(578)
1.078 .226
( 86)
1980 (S&S)
.715 .150
(390)
.668 .178
(557)
.902 .206
(179)
Number of measurements in average.
3-17
-------
These averages do not result from simultaneous measurements, and
they could be different just because the data sets from which they
were derived were different. However, one possibility is that
additional mass collected by the 1980 SSI and HIVOL samples is due to
a sulfate and, to a lesser extent, a nitrate artifact which occur on
the S&S filters but not on the quartz filters.
The same conclusion is drawn from the comparison of 1979 and 1980
arithmetic average ratio measurements given in Table 3.2.3. These
ratios are calculated from measurements made at the same site at the
same time. The average ratios of TOTAL/SSI and TOTAL/HIVOL decrease
by about 15% from the 1979 to the 1980 subset. This corresponds to
about 7 ug/m3 of additional mass collected, on an average, by HIVOL
and SSI samplers in 1980 over that collected in 1979. The ten percent
difference between the TOTAL/SSI ratios is similar to that noted in
Section 2.5.
The portion of the additional 1980 mass attributable to sulfate
and nitrate artifacts can be estimated by comparing the sulfate and
nitrate concentrations measured on SSI, HIVOL, and TOTAL samples taken
during 1979 and 1980. Once again, since the number of data points is
small, this comparison cannot be considered conclusive. The 1979
scatterplots for nitrate are not included because nitrate measured on
SSI was close to the lower quantifiable limit.
Figures 3.2.1 a and b are scatterplots and linear regression
parameters of the TOTAL/SSI sulfate measurement from samples taken
during 1979 and 1980, respectively. In 1979, the simultaneous
measurements were nearly equal for all samples considered. The linear
regression slope of 1.06 shows overall equivalency within experimental
precisions. The data points approach zero for TOTAL and SSI
measurements. In 1980, the TOTAL sulfate was considerably less than
the concurrent SSI sulfate in almost all cases. The linear regression
slope is .68, a considerable reduction from that calculated for 1979
measurements and well outside the bounds imposed by normal measurement
precisions. The 1980 plot shows minimum sulfate concentrations of
approximately 3 ug/m3 for SSI measurements even though the TOTAL
sulfate concentrations approach zero. The maximum absolute amount by
which an individual SSI sulfate measurement exceeded a TOTAL
measurement in 1980 was 8 ug/m3 while it was only 3 ug/m for the
1979 data.
-------
o
00
Oi
3"
1
£9
U3
<5
'
I
CM
(0
CO
_
UJ
0-
O
W
s
o
1
j
1
1 r«
: a?
Ol
r-
C71
-H
"a
%u
•
I
•
^^
1
co
'
H *
2 • .
fe
§ . -
-
i i i < ' i i i '
co
.r~-
CO
_•*•
M
"c\i
-00
u>
CM
O3
(0
CO
s
s
•8
-t^-
CM
-T}-
CM
lf^
"CM"
-«)
»••
-U>
-CJ
'CO
•o
-C\l«
rt
t*H
"3
CO
CO
CO
o
E
*^.
00
3.
n
o
• H
rt
(-.
+->
C
(U
o
§ M
u CO
o CO
i2 T3
M-( *^
__. (-*
S $
z 3
g
t"1
w
o
0
^H
&
rt
^
+-i
C
(D
O
C
O
CD
•M
ro
<4-l
t— i
3
CO
/ — \
ja
V /
o
CXD
CTl
i — l
Td
'c?
v — /
CTl
r~-
C7i
i— 1
C
•H
W
0)
4->
•H
co
3-19
-------
Similar observations do not apply to the comparison of
simultaneous HIVOL and SSI sulfate measurements for 1979 and 1980
presented in Figures 3.2.2 a and b. The two measurements during both
years are nearly equal, with linear regression slopes which do not
significantly differ from unity. This is to be expected when filter
media for simultaneous SSI and HIVOL samples are the same, either
quartz or S&S; any artifact is the same on both samples. These
figures also show that most sulfate is contained in particles with
diameters less than 15 urn.
The TOTAL/SSI and HIVOL/SSI nitrate comparisons are plotted in
Figures 3.2.3 and 3.2.4, respectively, for 1980 data; SSI nitrate
measurements from 1979 were inadequate to form such a comparison for
that year. There is little agreement between TOTAL and SSI
measurements, though SSI is almost always in excess of TOTAL. The
combination of a variable loss of nitrate from the TOTAL and a gain of
nitrate on the SSI samples could account for the discrepancies. In
the extreme case, one SSI nitrate concentration exceeded the
corresponding TOTAL nitrate concentration by 5 ug/m3. SSI nitrate
concentrations approach a minimum on the order of 1 ug/m3. The
SSI/HIVOL nitrate comparison for 1980 exhibits the same properties as
those shown by the corresponding sulfate plots; the measurements are
essentially equivalent.
If the sulfate and nitrate on the Teflon filters are used to
estimate the actual ambient concentrations, the linear regression
lines corresponding to Figures 3.2.1 and 3.2.3 can be used to estimate
the additional mass collected on S&S filters due to artifact
formation. Average TOTAL sulfate and nitrate concentrations from
various sampling sites are tabulated in Chapter 9 of this report and
are of the order of 10 ug/m3 for sulfate and 2 ug/m3 for nitrate.
The linear relationships of Figures 3.2.1 and 3.2.3 yield expected SSI
concentrations of 14 ug/m3 sulfate and 4.3 ug/m3 nitrate. This
o
amounts to approximately a 6.3 ug/m increase in average mass
concentrations which could be attributed to artifact formation, an
amount very close to the average mass concentration differences of
Table 3.2.2.
3-20
-------
o
oo
CTl
•°
•"
§<£
0)
.
1 1
03 (A IA *
i 8 a 2
"si! v- '
r "^ *
ui S d !*>
Q P S
3> ?5 8 S * *-
' • ' • 1 1 1 1 1
1 1 1 1 1 1 1 1 1
55
"cv
"w
'w
.00
•j£
'-
•a
"<0
•9)
- o>
C\J (S
s •»
_ -
10 4 N
LO OJ 0) 09
0} — O) —
£ "
z z
OH
H H O
& 5 ^ -
CJ J U.
UJ & UJ O
8s tf & .
_1 Z O O
W H 0 Z _
1 I 1 I I 1 I 1 1
CO
"N
.^
C\J
--
-CO
""
-10
-Csl
-0)
-
M
4J
C
V
0
c
o
CJ
<2
^
W
CO
6
o
6 -H
•H CO
00
0 -H
r~^
0 §"
rH CJ
(_,
<1> CL,
•P
rt (~H
O ^H
CO <
1
fN
•
rsi
.
tn
h
00
•H
fui/8n UOIJEJ1U3DUOO
TOAIH
3-21
-------
6
io 0
0
00
o
*
»
G> U)
8 %
. ,
CO U) CO
<9 Ot —
r_l •
„
T^
£ S £ ".
UJ Q! SJ 0
§> £ S .
-J Z Q O • •
R a z .
• • i • i • t i i
— j ^
O 'w
HH [/) .
- at 3= C 0
» 0 00
I/) -H C7^
3 +J ^H
O nJ
Sis .5
c^ C
"e I-H o m
- r, -^ 3 C +->
•^ ^ E 0 -H
•H U CO
CO
s CO I-H
" t rt ^
LJ Cd *^5 i™H
w 0 C -1
•£ CO rt nj
- « ~
w "^
CN
- N ^
0)
fH
— &0
•H
i i i i 1 — i 1 1 1 r IB
ftu/8rt 'suoilBJjuaouoo aie-iliM 10AIH
§
—
"
(O
? OD
«0 —
oo u) co
v o> 03 a
CO C 00 —
• • •
V)
OH
{r H p
U S d I
O P S *
^ R s § •. : to
<0
E
o
« £^
O tn •
E- C 0
O 00
m (/I «H O^
3 4-> ^H
O rt
0) ^ C
m_ C -M -H
• r- < rt c
M 4-1 1) I/I
=• i-l O 0)
3 C •*->
^2 E o -H
• c
•MB ° S ^
« MI—I
O 4-* ^ O^
i s s s
• v « &. co
rH '""1
M z c3 T3 i— i
_ o c <-i
10 co rt rt
VI
- CM M
tN
•
• -
D
(H
3
' » -H1
eai/8rt
3-22
-------
It should be cautioned once again that the above observations are
based on scatterplots made using 18 to 19 data points. However, all
evidence points in the same direction: filter artifacts due to the use
of the S&S filter for IP sites in 1980 could account for a significant
part of the difference between SSI and TOTAL average mass measurements
and can bias the reported sulfate and nitrate concentrations.
For a variety of reasons, EPA has changed the type and
manufacturer of the filter media which are used in its ambient
monitoring networks. This procedure poses serious problems to the
interpretation of data. As pointed out in this section, the switch
during 1980 to S&S filters resulted in possible biases of sulfate,
nitrate and mass measurements. Also, S&S filters have very high and
variable total carbon blank values which makes carbonaceous aerosol
analysis by combustion methods impossible. Switching of filter types
also limits the use of data for long term trend analysis. It is not
too unusual to analyze for the first time or to reanalyze samples
collected years ago. This limits the potential use of such samples.
3.3 Network Operations
Mass, elemental and ionic concentration measurements in the
Total, Inhalable, and Fine Suspended Particulate Matter are produced
by the IP Network Monitoring System (USEPA, 1980c). In addition to
these routine analyses, several special filters were selected for
optical microscopic and carbon analysis in addition to non-routine
analysis for ions and elements as part of this study and separate from
normal IP Network procedures. This section briefly describes the
analysis procedures followed for all of these measurements and lists
some of the limitations of the methods used.
The overall flow of routine tasks performed on the samples
collected is summarized in Figure 3.3.1. All the samplers in the
network were supplied by Environmental Protection Agency's
Environmental Monitoring Systems Laboratory (EPA/EMSL) and operated by
state or local agency personnel. The filters were weighed and sent to
sites by EPA/EMSL in Research Triangle Park, NC. The sampling was
carried out for 24 hours, midnight to midnight, every sixth day (or
3-23
-------
re in
4-1 V
re u wr
Co
o
o>
PL
O
O
z
(X
13
3
O
«:
«4H
o
o
r-(
H*
CD
3
00
u.
3-24
-------
third day in some cases'). The samples were returned to EPA/EMSL for
weighing and possible further analysis. Mass concentrations were
calculated for all samples. Sections of a selected number of filters
were submitted for chemical analysis to determine sulfate, nitrate,
and a variety of elemental concentrations. Sulfate and nitrate were
measured on all filters using automated colorimetry. Elemental
concentrations on HIVOL and SSI filters were measured by optical
emission spectroscopy and on Teflon filters by x-ray fluorescence.
Ion and elemental analyses were performed on approximately 25% of all
samples.
3.3.1 Filter Weighing
The HIVOL and SSI fibrous filters (8" x 10") were allowed to
equilibrate for at least 24 hours in a control box where relative
humidity was below 50% and temperature was between 15 to 35°C. The
weighing was done using a Mettler balance with a sensitivity of
.1 mg. Internal quality control was carried out by checking the zero
of the balance after every fifth weighing and by requiring immediate
investigation of any blank filter that weighed above or below the
range of 3.5 to 5.0 g. External quality control was carried out by
the reweighing of 4 filters out of each set of 100 by another
technician. If all of the reweights were within 2.8 mg of the
original weight for blanks and 5 mg for exposed filters, all weights
were accepted. External audits by supervisory personnel were
conducted at unannounced times.
The Teflon membrane filters (37 mm) used with dichotomous
samplers were also equilibrated as desribed earlier and weighed using
a Mettler microbalance precise to _+ 1 ug. The exposed filters were
reweighed on the balance on which their tared weights were obtained.
The internal quality control was carried out by weighing a "standard"
filter, arbitrarily selected for that purpose, at the beginning of
each day. If the weight of the standard filter was not within 20 ug
of the established value, a full-scale balance checkout was undertaken
before regular filter weighing commenced. About 6 filters were
3-25
-------
reweighed by another technician for each balance per day. If the
reweights were not within 20 ug of the original value, the difficulty
was found and corrected and all filters were reweighed. The zero and
calibration of the balance was checked after every fifth weighing.
Any blank filter weight outside of the normal range of 90 to 110 mg
resulted in immediate investigation.
A unique sample identification number was assigned to each
filter. Fiber filters were placed in filter folders and Teflon
filters were placed in petri dishes which contained the sample ID for
shipment to and from the field.
3.3.2 Field Sampling
In the field, each filter was loaded into the sampler at the time
the previous sample was removed. Sample start and stop times were
controlled by mechanical or electronic timers and sample durations
were quantified by elapsed time meters. All Teflon filters were
handled while wearing disposable gloves. Fiber filters were handled
on the edges with bare hands.
Average flow rates before and after filter exposure were measured
for HIVOL and SSI samplers with a continuously recording Dickson
Mini-corder. This device provided a constant trace over the sampling
period so that the occurrence and duration of power or motor failures
could be evaluated. The Mini-corder was calibrated at 5 flow rates
every three months, when a motor or brushes were changed, or when
successive calibration checks disagreed by 10% or more. The
calibration standard was a variable resistance orifice calibrator
which was in turn calibrated once per year with a Roots meter.
Calibration was checked with the calibrated orifice by using a clean
filter to provide flow resistance; this check was performed every
other sample (approximately once each 12 days). Nominal flow rates
o
were 1.4 and 1.1 m /min for HIVOL and SSI, respectively.
The Beckman SAMPLAIR dichotomous sampler was operated in a manual
mode controlled by a master timer independent of the sampler's
microprocessor unit. Initial flow rates through the FINE and COARSE
filters were 15.0 and 1.67 1/min, respectively, and were measured with
3-26
-------
separate rotameters. Flow rates were also measured following
sampling. Rotameters were calibrated at the factory and the total
flow rate (sum of flows through FINE and COARSE samples) was subject
to a one point check every other sample (approximately 12-day
intervals) using a calibrated orifice meter. Sierra 244 and 244E
dichotomous sampler flow rates were measured the same way.
The 1.67 1/min flow rate through the COARSE filter meant that
approximately 10% of the FINE particulate matter was deposited on the
COARSE filter. Corrections to the measured FINE and COARSE masses
were made as part of the data processing step.
Sample volumes were calculated from the average of initial and
final flow rates multiplied by the sample duration. Samples were
collected for about 24 hours every sixth day. The nominal volumes of
air pulled through HIVOL, SSI, and dichotomous samplers were 2000,
1600, and 22 m-^, respectively.
3.3.3 Sulfate and Nitrate Analysis
Automated colorimetry was used for the analysis of ions from
HIVOL, SSI, FINE, and COARSE filters. All the analyses were done by
EPA/EMSL. Their procedures are summarized briefly.
Filters were extracted (l/12th of an exposed filter for SSI and
HIVOL, all of FINE and COARSE) in 40 ml of distilled deionized water
by sonication for 30 minutes followed by centrifuging at 2000 rpm for
20 minutes. For HIVOL samples, one ml of each extract was poured into
a specimen vial and the sample turntable rate was 40 samples per hour
with 2:1 sample to wash-time ratio. The sample tray run at the
beginning of each day was loaded with duplicate calibration
standards. Quality control standards were placed in every tenth
position.
Sulfate was determined using the methylthymol blue (MTB) method
(Technicon, 1972). The extract was first passed through an ion
exchange resin to remove the interferring cations. Then it was mixed
with a solution of MTB and barium chloride at pH 3 to 4. Sulfate ions
in the extract reacted with the barium to form barium sulfate, thereby
3-27
-------
lowering the ratio of barium to MTB. Sodium hydroxide was then added,
raising the pH to 11 or 12. At high pH, barium and MTB form a
blue-colored chelate. The excess uncomplexed MTB is gray. The amount
of uncomplexced MTB, monitored colorimetrically at 460 nm, is
proportional to the sulfate concentration.
Nitrate was determined using the cadmium reduction method in
which the sample was mixed with ammonium chloride, then passed through
a copperized cadmium column which reduced the nitrate to nitrite
(Technicon, 1976). A mixture of N-(l-naphthyl)ethylenediamine
dihydrochloride (NEDA), sulfanilamide and phosphoric acid was added to
the sample. The nitrite and NEDA reacted to form a pink dye with a
peak absorbance at 520 nm which was proportional to the nitrate
concentration.
Interlaboratory and intermethod comparison results of this method
(Mueller and Hidy et al, 1981) show it to be free of interferences and
with excellent reproducibility.
3.3.4 Elemental Analysis
The elemental analysis of HIVOL and SSI filters was accomplished
using inductively coupled argon plasma emissions spectroscopy (Lynch
et al, 1980) according to USEPA (1978b). Elemental concentrations on
HIVOL and SSI filters were reported in EPA data summaries for 11
different elements as shown in Figure 3.3.1 (USEPA, 1981a). To
extract the elements, 1" x 8" of each filter was placed in a
polypropylene centrifuge tube to which 12 ml of extracting acid
(2.23 M HC1, 1.03 M HNO-j) was added. The mixture was ultrasonicated
for 50 minutes and 28 ml of distilled deionized water was added to the
tube. The mixture was centrifuged for 20 minutes at 2500 rp»m. The
clear solution was then transferred to a 30 ml acid-clean
polypropylene bottle, taking care not to disturb the solids in the
bottom of the tube. The analysis was done on a Jarrell-Ash
Inductively Coupled Plasma Emission Spectrometer at EPA/EMSL. Two
quality control standards were prepared from the stock calibration
standards and run with each set of samples.
3-28
-------
The FINE and COARSE membrane filters were analyzed by x-ray
fluorescence analysis following the procedure described by Dzubay and
Rickel (1978) at Lawrence Berkeley Laboratory. Several additional
filter samples were analyzed by NEA Laboratories for this study for
Al, Si, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Br and Pb.
X-ray fluorescence analysis is a non-destructive technique.
Samples go through this analysis prior to the colorimetric analysis
where they are destroyed. Atoms in the sample are excited from their
ground states to higher energy levels by x~radiation from x-ray
tubes. The energy emitted by atoms as they return to their normal
ground level is characteristic of the emitting element and is used to
quantitatively identify the element. Details on the replicate
analysis and calibration procedures are not available in USEPA
(1980). Elemental concentrations on FINE and COARSE filters are
reported in EPA data summaries for 15 different elements as shown in
Figure 3.3.1 (USEPA 1981a).
3.3.5 Carbon Analysis
Total and organic carbon were determined for a selected subset of
SSI and HIVOL filters at Environmental Research and Technology, Inc.
by the method of Mueller et al (1980). These measurements are not
part of the routine IP Network analyses undertaken by EPA/EMSL.
A Dohrmann DC-50 carbon analyzer was used for the determination.
A sample to be analyzed was loaded onto a platinum boat containing
manganese dioxide. The boat was then advanced to a pyrolysis zone at
850°C where all carbonaceous materials were oxidized by the manganese
dioxide. Organic carbon was determined by lowering the temperature to
550°C. The carbon dioxide thus generated was then hydrogenated
catalytically to methane and measured using a flame ionization
detector. Calibration of the instrument involved nulling with
acidified distilled deionized water and span checking with 180 ppm and
360 ppm carbon standards consisting of potassium hydrogen phthalate
solutions. Detector output was set to correspond directly to the
carbon concentration of the standards. A separate punch from every
tenth filter was reanalyzed to ensure precision of the analysis was
maintained.
3-29
-------
3.3.6 Optical Microscopic Analysis
Optical microscopic analysis of portions of the deposits on
nineteen SSI filters and some source samples was performed by Boeing
Aerospace Corporation's Fine Particle Laboratory. The details of the
optical microscopic analysis applied to these selected SSI filter
samples are described by Crutcher and Nishimura (1981) with their
results of the analysis.
The particles from a 12 cm^ section of each 8"xlO" filter were
removed by sonication in toluene and the extract was centrifuged.
Portions of the particulate matter residue were mounted on slides for
viewing under a polarizing microscope.
First, a qualitative identification of particle types common to
the sample was made by moving the slide on the microscope stage and
classifying the observed particle types by morphology, refractive
index, color, birefringence, and association with other particle types,
The particle types identified were divided into five major
categories with a number of subcategories:
Minerals, including quartz, calcite, gypsum, clay and other
minerals.
Biological materials, including plant parts, pollen and
spores, starch and other biological material.
Industrial and transportation particles, including weathered
industrial, calcium hydroxide, hematite (rust), hematite
(industrial), glass, slag, paint spheres, tire wear, golden
needle, lead oxide needle, hydrated cement, clinker, YC2S,
€38, and glass cement.
Combustion products, including fused flyash, unfused flyash,
charred wood, soot, oil soot, black carbonaceous spheres,
black carbonaceous spheres, cenospheres, bagasse, black
irregular, charred paper, magnetite spheres and coal.
Particles not contained in other categories, including
molybdenum disulfide and any other unidentified substances.
3-30
-------
The particles were identified as members of specific categories
based upon recognition of the features determined in the qualitative
analysis; this recognition was made by the microscopist. The diameter
of each particle was estimated by identifying which circle of a Porton
graticule best approximated the projected area of the particle. Each
successive circle of the graticule increases its diameter by 1.41
times that of the previous circle; nine circles were used for
comparison ranging from 1 um to 16 um in diameter. The majority of
the particles observed had optical diameters corresponding to the
circles less than the 8 um diameter circle.
Each sample was moved across the center of the field of view and
particles were identified and counted as they reached the eyepiece
crosshairs until approximately 1,000 particles were included. The
numbers of particles within each particle type and size range were
totaled, multiplied by a density typical of the particle type
(densities of minerals were obtained from Deer et al, 1966, while
densities of other materials were derived from density gradient
measurements of other researchers) and the volume of a sphere having
the diameter of the Porton graticule circle corresponding to that size
range. The masses achieved by this process were summed and the
percent of this total comprised by each category was calculated by
dividing the mass of particles so classified by this sum.
The following limitations of this method are realized:
The process of removing particulate matter from the filter
may result in some loss, dissolution, or preferential
extraction of different particle types.
One thousand particles is small compared to the total number
of particles collected on the filter. If one assumes
1,000 particles of 8 um diameter and 2 gm/cm3 density are
counted, the total mass observed amounts to approximately
.5 ug. This represents .0003% of the mass deposited on a
typical HIVOL filter.
Few particles are spherical and the adjacent size ranges
differ by a factor of 2.8 in particle mass. Particles
falling between ranges can have mass estimates differ from
their true values by up to 50%.
Only coarse particles (greater than 1 to 2 um optical
diameter) are observed and counted. Therefore, the
proportions of particles resulting from each category
correspond only to the COARSE size fraction.
3-31
-------
The uncertainties in particle source-type contribution estimates
resulting from these limitations of optical microscopic analysis are
difficult to quantify; no precision model incorporating all of them
has yet been proposed. The portion of this uncertainty due to
statistical counting error in the sample examined was calculated for
each of these samples using the method of Crutcher and Nishimura
(1978) for each particle type. This uncertainty could exceed a factor
of 2 in a given particle category.
3.3.7 Accuracy and Precision of Analysis
Each of the procedures followed in the IP Network involves
replicate analysis and audit components, the data from which can be
used to estimate accuracy and precision over specified time periods.
Though the collocated mass measurements presented later in this
chapter give an idea of the mass concentration measurement precision
which is attainable by following IP Network standard operating
procedures, they do not tell what the relative precisions of
individual measurements are. This must be done through the use of
replicate analyses of measurements for specific sites and times.
Many comparisons of measurements are made in this report and
conclusions are drawn when measures differ from each other. If the
uncertainty intervals of two measurements do not overlap, then they
significantly differ from each other and a cause of that difference
can be sought. If the intervals do overlap, however, then no
significant difference exists between them. Without knowing these
intervals for the measurements being compared, it is impossible to
determine the significance of differences or similarities among these
measurements.
At this preliminary stage, very little Quality Control (QC) and
Quality Assurance (QA) data from the IP Network have been compiled and
summarized for those using the data for interpretive purposes. Such a
summary is required if the interpretation of measurements by those who
did not participate in their acquisition is to be meaningful.
Baumgardner (1981) reports that such a summary will be issued in EPA's
1981 data summary.
3-32
-------
3.4 Data Validation
The HIVOL, SSI, TOTAL, FINE and COARSE mass measurements used for
the analyses in this report went through various validation steps
which verified their acquisition according to prescribed procedures,
their transfer through the data management system and their internal
consistency with respect to each other. Some of these procedures were
applied by EPA/EMSL to eliminate and flag suspect data. Additional
criteria were applied to the resulting data set by the authors of this
report. Both sets of procedures are outlined in this section. The
EPA/EMSL mass concentration validation procedures are presented in
USEPA (1980c). Validation takes place in the field, in the weighing
laboratory, and at data processing.
Samples were invalidated by the field technician if:
• sampling did not start at 0000 + 0030 local time,
• the sample duration was outside of the interval 24 +_ 1 hr,
• post-sampling flow rates were 10% less than pre-sampling
flow rates,
• flow rate calibration checks (scheduled for every other
sample change) made before and after a sample was taken
differed by more than _+ 10%
• leaks around filter were indicated by a discolored edge, or
• the filter was damaged.
In addition to invalidating the sample under the above conditions, the
field technician noted on the data card any emissions or
meteorological conditions which might bias the measurement.
In the weighing laboratory, glass fiber filters were weighed in
sets of 100. Five of these were reweighed by a different technician
and if the reweights did not agree within prescribed tolerances, all
weights for that set were invalidated until the filters were reweighed
and the reweight test was passed.
For dichotomous samples, 5 to 7 filters were reweighed every day
for each balance in operation. If the prescribed tolerances were
3-33
-------
exceeded, then all of the weights measured on that balance on that day
were invalidated and the procedures were repeated for those filters.
At the data processing step, all records which were keypunched
from the data forms were checked for faithful transcription. After
association of mass and flow rate measurements and calculation of
ambient mass concentration by a computer, the measurements were
screened by:
• identifying negative or less than detection limit values,
• statistical outlier tests, and
• comparison of measurements from simultaneous samples in
different size ranges.
Input data records associated with values flagged by this screening
were examined and corrections to the data base were made when
discrepancies were found.
Values which did not pass through this screen after the
correction process was complete were:
• invalidated and removed if they were negative,
• flagged with as S if they were statistical outliers, and
• flagged with an R according to the criteria of Table 3.4.1.
Table 3.4.1
DATA VALIDATION FLAGGING CRITERIA
1.
If
-HI VOL -
SSI
TOTAL
FINE
-COARSE-
value is <
-20-
15
15
10
- 5^
ug/m3, or >
-120-
100
100
40
- 60-1
ug/m3, flag
value.
2. If SSI/HIVOL ratio is > 1.1 or < .4, flag both values.
3. If TOTAL/HIVOL ratio is > 1.1 or < .4, flag both values.
4. If TOTAL/SSI ratio is > 1.2 or < .8, flag both values.
5. If COARSE/FINE ratio is > 1.3 or < .3, flag both values.
3-34
-------
The EPA/EMSL defined limits in Table 3.4.1 are empirical and are
used only as a guide to more extensive validation; it is possible,
though unlikely, that these limits are exceeded under ambient
conditions.
For the purposes of this study, additional criteria were applied
to remove some, but not all, flagged values from the data set used for
the majority of the interpretive efforts. To remove all flagged
values would have made the data too sparse to be of any use, but some
of the discrepancies were so obvious that any interpretation of the
data would have been severely compromised.
For the purposes of this report the measurements of HIVOL, SSI,
TOTAL, FINE and COARSE for a site and day were removed by the authors
if any of the following inequalities was found to be true:
/ COARSE \ 3-1
\HIVOL - FINE
i ^^ ^^ „ „„ I "" .3
HIVOL - SSI
HIVOL
HIVOL - TOTAL\
HIVOL )
-.15 3-4
Inequalities 3-1 and 3-2 state that the COARSE material, that
within the 2.5 to 15 urn range, must equal at least 30% of the material
in the 2.5 to 30 urn range for the set of samples to be accepted. The
30% value is arbitrary and was chosen because the IP data suggest it as
a natural dividing point. Many samples, particularly those from the
Phoenix and Birmingham areas, showed very low COARSE sample masses. The
ratios of COARSE/(HIVOL - FINE) for these cities were in the
neighborhood of .1 to .2, a very unlikely situation given the typical
particle size distributions discussed in Section 2.2.
Inequalities 3-3 and 3-4 eliminate measurements for which the mass
of particulate matter in the 0 to 15 urn range exceeds the mass in the 0
to 30 um range by more than 15%, a reasonable tolerance for measurement
error.
3-35
-------
The resulting subset of the inclusive sampling dates listed in
Appendix A.I contains 2,087 HIVOL, 771 SSI, and 1,509 dichotomous
(TOTAL, FINE and COARSE) values out of a possible 2,675 HIVOL, 1,045 SSI
and 1,960 dichotomous samples (Suggs et al, 1981b).
The analyses presented in Chapters 7 and 8 this report used the
data set filtered by inequalities 3-1 to 3-4. Statistical data
summaries were received from EPA for the treatments in Chapters 5, 6 and
9; these were based on the EPA validated data which were not subject to
the screening of inequalities 3-1 to 3-4. The appropriate notation will
be made when screened and unscreened data are used.
The extensive data validation procedures outlined here have
resulted in individual measurements which are internally consistent.
The criteria have taken their toll, however, in leaving large gaps where
samples have been invalidated. The validation and flagging process is
necessary, but the lack of measurements resulting from it in many
situations introduces a sampling error due to the reduced number of
values, particularly for simultaneous measurements. As the network
monitoring progresses and the standard operating procedures become
ingrained, fewer samples will be invalidated and the size of the
validated data base will grow.
3.5 Precision and Accuracy of Size-Classified Mass Measurements
The purpose of this section is to identify, and quantify where
possible, the biases to mass concentrations in the IP Network and to
estimate the precision attainable with these measurements from
collocated sampling.
Though the HIVOL sampler has been the mainstay of suspended
particulate matter sampling for nearly three decades, recent studies
show that it is subject to certain biases other than those already noted
which, if not considered, will appear to be properties of the ambient
aerosol rather than consequences of the sampling method. An examination
of these studies is useful in estimating the bias introduced into the IP
Network's HIVOL measurements and in indicating studies which need to be
carried out with regard to the other IP samplers. These studies include
the effects of:
3-36
-------
• Carbon and copper emissions and recirculation into the
sampled airstream of the HIVOL or other concurrently
running samplers.
• Deposition and removal of particulate matter during passive
sample exposure periods.
The motor of a HIVOL sampler contains a copper commutator which
turns between carbon brushes. The abrasion of these surfaces causes
copper and carbon particulate matter to be entrained in the air
exiting the HIVOL. It is possible that some of this particulate
matter finds its way onto the sampling substrates, thus biasing the
ambient air measurements. In the IP Network, carbon was measured as
part of this study on selected glass fiber filter samples. Copper is
routinely quantified both on selected glass fiber and Teflon filter
samples. Contributions from both of these substances can add to the
mass concentration measured on a filter. Bias due to HIVOL emissions
is a possibility and the maximum likely contribution must be
quantified.
Fortunately, studies of these emissions have been performed by
3
other researchers. Countess (1974) measured 150 ug/m carbon in the
output of a typical HIVOL sampler. He injected known concentrations
of dioctylphthalate (DOP) into the HIVOL discharge and measured the
amount collected on the filter under various ambient airflow
conditions. He found the greatest recirculation occurred with the
HIVOL blower removed from its housing in a horizontal position; .1% of
the HIVOL DOP emissions were collected. The maximum carbon
3
concentration bias is therefore .15 ug/m , a quantity below the
lower quantifiable limits of carbon measurements.
The worst case recirculation efficiency of .1% is valid for
copper emissions as well, and, since the commutator wears at a slower
rate than the brushes, the contribution to the sample mass is
negligible. King and Toma (1975) demonstrated that under normal
operating conditions, a HIVOL can collect as much as ten times the
ambient copper levels as a result of blower emissions. This
recirculation was reduced substantially when a tight seal between the
HIVOL blower and the housing was formed so that no recirculation could
travel through the housing. Because the dichotomous sampler inlets
3-37
-------
are normally at least 1 m above and 2 m to the side of the HIVOL
exhaust, the copper contribution is probably negligible to FINE and
COARSE samples under most conditions. Significant contributions to
the HIVOL and SSI samples is a possibility, however. High copper
values in the .1 to .6 ug/nr' range should be suspected of being
contaminated (King and Toma, 1975), except in situations where a known
industrial source of copper, such as a copper smelter, might be the
cause of elevated copper concentrations. The IP Network chemical
analysis data show many copper (Cu) concentrations on HIVOL and SSI
samples in this range, while FINE and COARSE Cu measurements taken
simultaneously are substantially lower. Cu contamination from the
HIVOL motors is a viable possibility.
In the IP Network HIVOL sampling configuration, a HIVOL filter
could remain in an idle sampler for up to six days prior to and after
sampling. During this idle period, wind blows particle laden air
through the HIVOL housing with some of those particles depositing on
the filter. This is termed passive mode deposition. McFarland et al
(1979b) have quantified this deposition as a function of particle
size, wind speed and ambient concentration. They found that the
deposition rate increases with increasing values for any one of these
variables. For example, in an 8 km/hr wind, McFarland et al (1979b)
found 8 ug of the 5 urn particles they generated deposited on the HIVOL
filter for each 1 ug/m^ concentration of these particles in the
sampled air over a 24-hour period. This increased to 40 ug/24-hr for
1 ug/nr* of 15 um particles and to 100 ug/24-hr deposited for
o
1 ug/nr concentration of 30 um particles. The fraction of total
suspended particulate matter sampled by the HIVOL greater than 15 um
is in the vicinity of .33 (Pace, 1979b). If the ambient air contains
100 ug/m^ total suspended particulate matter, then the deposition
bias on the filter during the 6-day standby period would be about
6 ug/m^ for 15 um particles and 13 ug/m^ for 30 um particles
according to McFarland's data. The actual bias in the mass
concentrations could be calculated by integrating McFarland"s
deposition function times the ambient particle size distribution by
the method of Section 2.2. Ambient experiments have been performed,
however, which provide more useful estimates of passive deposition.
3-38
-------
Bruckman and Rubino (1976) found positive biases ranging from 5
to 28%, with an average of 15% when comparing collocated HIVOL samples
from 6-day and 0-day standby periods. Blanchard and Romano (1978)
found a range of 2 to 48% with an average of 13% (by inverting the
sampler they seem to have eliminated the deposition bias). Chahal and
Romano (1976) measured a range of -5 to 29% with an average of 7%.
Swinford (1980) found a -5 to 25% range with an average of 10%. These
experimental results agree with the wind tunnel measurements of
McFarland et al (1979).
Sweitzer (1980) examined several passively loaded HIVOL filters
from Peoria and Decatur, Illinois with an optical microscope and found
90% of the particles to have optical diameters greater than 10 um with
some diameters as large as 60 um. This qualitative particle
evaluation identified fly ash, sulfate particles and coal fragments on
the Peoria samples and corn starch balls, bundles and grains on the
Decatur samples. These were similar to presumed emissions from nearby
industrial sources.
Thanukos et al (1977) have raised the concern that some of the
particulate matter collected by the HIVOL is removed by wind blowing
through the sampler during the standby period following sampling.
They compared TSP measured by a standard HIVOL with that obtained from
a HIVOL preceded by an Andersen Cascade impactor; presumably the
impactor did not allow wind to remove particles. For TSP
concentrations less than 150 ug/m^ they found the geometric mean of
unprotected HIVOL measurements to be 5% less than the geometric mean
of measurements by the wind-protected HIVOL with a 1-day standby
period following sampling. Swinford's (1980) tests confirm this.
This bias is not significant for IP Network samples.
Deposition during passive periods can cause samples taken with
the HIVOL to be significantly biased. Normally, approximately 10% of
the average TSP can be attributed to this bias, but this can be
significantly higher for individual measurements made in atmospheres
with a great quantity of large particles.
Passive deposition experiments have not yet been performed on
HIVOL(SSI) and dichotomous samplers. Presumably, the size-selective
inlets will prevent the larger particles responsible for this
3-39
-------
deposition from being blown into the sampler, but the bias is
impossible to estimate without wind tunnel and ambient tests similar
to those performed for the HIVOL. Such tests need to be planned in
the near future.
The best estimation of the precision attainable from the samplers
used in the IP Network is from the results of collocated sampling of
the same ambient concentrations by similar instruments.
McKee et al (1972) conducted a HIVOL sampling collaborative
testing program with twelve laboratory groups participating. The
2
samplers were equally spaced on a flat, 465 m rooftop at a height
of 15 m above ground level. Each group weighed its own filters,
calibrated its own samplers, and measured its own flow rates. The
sampling time was measured by the referee of the study. The mean
coefficient of variation for the results from 11 of the 12 groups (the
results from one group were deemed outliers) was 3.8% for particulate
3
matter concentrations in the 70 to 140 ug/m range. The mean
coefficients of variation of mass measurements made with two
collocated standard HIVOL samplers in another study (Camp et al, 1978)
3
was 4.1% in the 50 to 180 ug/m range.
Several collocated sampling studies have compared the dichotomous
sampler models used in the IP Sampling Network. Shaw et al (1981)
compared Beckman samplers and concluded:
"...over a broad mass range, fine fraction aerosol mass can be
collected using the standard commercial inlets with a
reproducibility of 5% or better. Coarse fraction aerosol mass
can be collected with a reproducibility of 10% or better except
for very light mass loadings (less than 200 ug). Fine and coarse
masses can be collected with a reproducibility of 5% or better."
They found that the mass measurement precision dominates the
3
precision attainable below approximately 200 ug (or 8 ug/m )
loadings. This uncertainty rises rapidly below this point for the
beta gauge mass measurements Shaw et al (1978) made and passes +50% at
3
4 ug/m concentrations.
Grantz (1981) compared collocated measurements from the Sierra
244 and Beckman samplers by linear regression and found slopes
(intercepts) of .73 (12.9), .79 (6.13) and .53 (7.11) for TOTAL, FINE
3-40
-------
and COARSE, respectively, when the Beckman measurement was the
independent variable. The correlation coefficients were low, .71, .77
and .53, respectively. Grantz (1981) does not present the data used
to calculate these parameters, but the agreement between most
individual values was probably not within ^10% with such high
intercepts, slopes different from one, and low correlation
coefficients.
Pashel et al (1980) ran collocated Sierra dichotomous samplers in
the vicinity of two steel mills and found percent differences
(calculated as the ratio of the differences to the average of the two
measurements) ranging from -40% to +40% for TOTAL, -59% to +65% for
FINE and -72% to +67% for COARSE at one site. At another site the
ranges were -49% to +90% for TOTAL, -10% to +115% for FINE and -200%
to +15% for COARSE. Pashel et al (1980) noted several sampling
difficulties, such as dirty inlet tubes and a buildup of particulate
matter around the collection nozzle.
Collocated sampling studies with the HIVOL(SSl) have not been
reported.
Three collocated sampling sites have been established in the IP
Network to assess network precision, one at 500 S. Broad St. in
Philadelphia (PAPHA), one at ARMCO Coke Plant Gate in Middletown, Ohio
(OHMII) and one at ARMCO Research (OHMIB) in Middletown. Data are
available for HIVOL, TOTAL, FINE and COARSE measurements at
Philadelphia, PA and for HIVOL and SSI measurements at Middletown,
OH. Figures 3.5.1a-d present scatterplots and linear regression
parameters for each of these comparisons. Though the correlation of
HIVOL measurements is quite good at both PAPHA and OHMII, a bias of 9%
seems to exist between samplers at both sites as evidenced by the
linear regression slopes. The average of ratios of paired samples for
HIVOL is .97 and 1.02 for PAPHA and OHMII, respectively, so the
difference in slopes is probably an artifact of the regression
process. The precision is still acceptable for most HIVOL
measurements, however. The average of absolute percent differences is
6.5% at PAPHA and 4.7% at OHMII. Eighty percent of all paired
measurements at PAPHA are within _+10% of each other and all are within
+_20%. The nominal reproducibility of HIVOL sampling using IP Network
3-41
-------
_
I
n
-j
o
i
.e
\*
*
•
* •
§8 *
N
H.
So? 09 CD „
a
.j
|
£
'a
CO CM
s s
i * *,
§CS| w * *
CO O) (0 ~w" •
CO
••.•••• "
H
0 H
»- jj ° :-
fo 5
D 1 Uv ™
UJ m Q O
fe P S • "
-J Z Q O
W H 0 Z
t, ,, L 1 I I 1 1 ]_ 1
• -.HKBPBB'B"'
IW
g
0
3
M
s
s
s.
8
£
T
D
n
£
b>
K)
B
"M
a
C
0
4-»
S
u
g
o
u
h
U
c/)
V
c
o
Of
to
E
M
^
O
'^
rt
^t
u
o
u
^
o
"S,
ra
„
c
•H
4-)
3
^
O
M-(
tf)
4^
0)
lU
^
t/)
rt
0)
S
13
0)
rt
o
o
1— 1
r— 1
O
U
4-i
0
to
4->
o
i— i
p {
^
0)
4->
rt
o
co
4-1
1
rt
i— i
L/)
bO
(U
3
•H
•V
/• — ^
-13 /— v
2"^^i
' — < rt
•H C
§ PH O
O i-H 4->
+-* Oi O
0) T3 i-H
i— i rt TJ
T3 -H ;H
IS ex *"
cc.5
•H -H
H- 1
J UJ CO
O Z CO
>> I-H
HH u, -a
X C
•» rt
•* /~^
/— s O r~^
rt <-' a>
rt
rt -H rt
•H J3 -H
PH--H P-i
I-H a) I-H
(U T3 (D
'^ rt ^3
rt I-H rt
r j ,pj I_J
•H x; -H
X ex x
ex ex
C
c -H c
• H -H
m
J CO J
O Oi <£
> < H
HH O O
33 Cj H
i
ui/3rt '
3-42
-------
5
c-
-10
r-. U)
rt
LU
W
JX
u
U
-
g«0 -" -
00
iL
S«r * w
r- o) —
£ "
s@i ' " •
LU ^ •
C J U.
uj pi cn o
_J Z Q O
W H O Z
II ] 1 1 1 1 1 1
atntotoijtofa^feini
O'TVCOCOCMN — —
U>
^
f
Kl
6
Ig "Si
co ^
2fe 2fe 3fe
ampler Concentration,
6 ib ifc
Routine S
&
0)
C
•H
O
u
LO
txO
ui/8rt '
3-43
-------
a
8
8 2
ii-
CD
O>
r*
W 0» —
w
UJ
fe
W
fe
O
-fe—fe-
-k—fo-
-ir
E
M
V
I-H
O.
U)
C
• H
4-1
3
W 0)
r- ca
eg N
«'
i i -
§0) U>
co o)
rs
§
H
UJ DC (D
o »- 2
CO H 8
a a
w/Zfi -uotaejauaouoo aatdlutiS p3aBO°U°3
t ' CS
-
•
m
.
10 "
-
co
§
£
u.
o
i
&' 1 I I 1 1 1
a
CP
•Q
3
s
9
O
8
s
^
a
to
s
a
tshtsishkifsdf
to
B
D4
A
c
o
•H
?
2
u
J
,
o
1-
03
M
0)
C
•t-t
3
0
a;
3
C
O
u
ir;
at
^
W)
•H
tt,
m/Srt 'uotjuj;uo3uo^
3-44
-------
standard operating procedures is in the neighborhood of +5%, which is
comparable to the precisions previously cited.
The slope of the collocated PAPHA COARSE measurements' regression
line does not show the bias exhibited by the HIVOL comparison, though
the correlation is not nearly as good. The average of absolute
percent differences is 11.5%. Fifty-three percent of the paired
measurements are within +10% of each other, 87% are within +^20% and
all are within +30%. The COARSE measurement uncertainty using IP
Network standard operating procedures is in the neighborhood of +^10%,
agreeing with Shaw et al (1981) measurements.
The FINE reproducibility is illustrated in Figure 3.5.Id. The
average absolute percent difference between values is 7%; 80% of the
values are within +107, of their average, 93% are within ^20% and all
are within +307,. The average ratio is 1.04.
The TOTAL scatterplot shows an average ratio of 1.05, an average
absolute percent difference of 9%, and 60% of the values within +10%,
93% within ^20% and all within ^30% of their average.
The SSI average ratio is .96, absolute percent difference is 4.5%
and all values are within 11% of their average.
More collocated sampling needs to be done in different parts of
the network to assure that the precision estimates made here have
wider application. These estimates show that precisions within +10%
are attainable, and in the SSI and HIVOL cases within +5%, for I?
Network measurements. This is a snapshot of the entire network,
however, and individual measurements may have greater or lesser
uncertainties associated with them.
This chapter is an archive of information about the measurement
process which should be used as an aid in interpreting data from the
IP Network data base. It brings to light several features of the
existing data base and suggests some additions to the measurement and
reporting process that could be made to enhance the usefulness of IP
Network data.
3-45
-------
• Most sampling site locations seem to be representative of
the study scales for which they were selected. Site surveys
should be standardized, completed, compiled and made
available for all IP Network sites. The site-type
classification scheme is still subjective and it is likely
that some sites are misclassified. The types of samplers
(e.g. Beckman or Sierra) and their installation and removal
dates should also accompany this survey.
• The change in SSI and HIVOL filter media after the beginning
of 1980 could result in differences between sulfate, nitrate
and mass measurements which are products of the measurement
process rather than an environmental cause. The exact dates
on which the change took place at each site should be
tabulated, and some consideration should be given to a well
planned, executed and documented switch back to the
Microquartz filter which seems to minimize sampling
artifacts.
• The routine IP Network measurements comprise mass, elements
and ions on HIVOL, SSI and TOTAL filters which are acquired
via a comprehensive set of procedures and analytical
methods. What is lacking is documentation concerning the
quality control and auditing of these procedures. Blank,
standard and replicate checks should be used to estimate the
measurement precision of individual data items and these
should be recorded in the data base and included in data
reports. Network audit results should be summarized to
provide a statistical measure of the measurement accuracy
for each observable.
• The data validation procedure to which IP Network mass data
are submitted identifies many internal inconsistencies as
well as extreme cases; the total number of these cases
amounts to a large portion of the data base. For the
purposes of this report, several of these flagged values
were deleted. The number and types of flags should be
summarized with the data capture statistics for each site.
Data validation procedures have been applied only to mass
measurements; (no validation flags appear on reports of
chemical composition) and should be extended to ion and
element measurements.
• Tests on HIVOL samplers show that copper emissions from the
motor can interfere with Cu measurements from nearby
samplers and that deposition on the filter during standby
periods can interfere with mass measurements. The emissions
should be eliminated by filtering or ducting sampler
exhaust. Passive deposition should be minimized by
shortening exposure times during the sampler standby mode.
Passive deposition measurements on HIVOL(SSI) and
dichotomous samplers should be carried out.
3-46
-------
The comparison of collocated HIVOL measurements in the IP
Network is similar to comparisons in other networks showing
average differences between simultaneous samples of less
than 5% of the average mass concentration. Similar results
are obtained for HIVOL(SSI) samplers. Collocated
dichotomous samplers have been shown to attain average
differences of less than 10% of the average IP mass
concentration. Collocated measurements between Beckman and
Sierra samplers need to be made; since these two samplers
are assumed to be equivalent for IP Network measurements, a
test is required to support that assumption.
3-47
-------
CHAPTER 4
URBAN AREAS IN THE IP NETWORK
At the time of this report, the Inhalable Particulate Matter
Sampling Network measured concentrations in 26 heavily populated areas
of the United States.
This chapter presents general meteorological and statistical
information for those areas that are related to the particulate matter
concentrations. In seven cities, Birmingham, AL; Phoenix, AZ; Denver,
CO; Buffalo, NY; Philadelphia, PA; Houston, TX; and El Paso, TX; the
industrial source types and their geographical locations with respect
to receptors have been identified on maps. This descriptive
information will prove useful when the relationship between chemical
composition of the aerosol and its sources is examined in Chapter 9 of
this report.
4.1 General Meteorological and Statistical Information
Table B.I of Appendix B lists typical (the average of
approximately 30 years of data) annual average meteorological
measurements in the urban areas of the IP Network. The first three
columns show variations in average, maximum and minimum temperatures.
The averages range from 44.1°F (Minneapolis) to 70.3°F (Phoenix). The
maxima vary from 53.8°F (Minneapolis) to 85.1°F (Phoenix), and the
minima fall between 34.3°F (Minneapolis) and 58.0°F (Houston).
Precipitation is lowest (less than 10 inches) in Phoenix and El
Paso, and highest (greater than 40 inches) in Birmingham, Hartford,
Atlanta, Boston, Baltimore, Raleigh, Houston, and Richmond. The
sunniest areas (more than 70% of the time) are Phoenix, Los Angeles,
Denver, El Paso, and Salt Lake City, while Seattle, Portland, and
Pittsburgh enjoy the available sunshine only 50% of the time. Thirty
year average wind speed for most urban areas is in the range of 14 to
18 km/hr with a low speed of 10 km/hr in Phoenix and a high of
20 km/hr in Boston. Most daytime relative humidities are between 50
to 60%.
4-1
-------
Seasonal degree days follow the temperature averages with low
values (less than 2000 DD) in Birmingham, Phoenix, Los Angeles,
Honolulu, and Houston and high values (greater than 6000 DD) in
Denver, Hartford, Des Moines, Chicago, Detroit, Minneapolis, Buffalo,
and Cleveland. Snowfall is non-existent in the Arizona and California
urban areas, minimal in Texas, Oregon and North Carolina, and heavy in
Denver, Hartford, Boston, Minneappolis, Cleveland, Pittsburgh, and
Salt Lake City. Snowfall levels in Buffalo are extremely high.
Because of the long averaging times involved, the data in this
table obscure relationships between meteorological factors and
individual particulate matter concentrations. The data in this table
are intended for use as a screening device to decide whether or not a
more detailed examination of specific meteorological or emissions
measurements available at the time the samples were taken is warranted.
For example, degree days are related to wintertime heating
requirements and the concomitant combustion of wood, oil, coal or
natural gas. Differences in wintertime fine particle carbon
concentrations, should they be measured, between areas with vastly
different degree day values could be due to this source type.
Similarly, elevated wintertime concentrations of sodium and chloride
in Buffalo, where average snowfall is 93 inches, might be caused by
the salting of roads whereas this would be an unlikely contributor to
elevated concentrations of these elements in Houston.
Table B.2 of Appendix B presents general population, area and
industrial activities information about each of the urban areas. Once
again, this information is for screening purposes rather than to
provide a direct relationship between anthropogenic activities and
suspended particulate matter concentrations. Populations of the
metropolitan areas and the space which they encompass vary
substantially. Population density is probably a better indicator of
potential emissions which might affect the measured concentrations.
Highest population densities (greater than 10,000 people/mi^) are in
the eastern and midwestern cities of Washington, D.C., Chicago,
Boston, Baltimore, Detroit, St. Louis, Buffalo, and Phiadelphia. One
western city, San Francisco, falls into this category. Lowest
population densities (less than 3,000 people/mi^) are in Phoenix,
4-2
-------
Kansas City, Raleigh, El Paso, and Houston. The general summary of
major industries gives some idea of what the possible industrial
suspended particulate matter emissions might be.
Many other summaries of descriptive information would be helpful
as pointers to more detailed information when the manmade progenitors
of measured pollutant concentrations are sought. Average mixing
heights, geography, climatological cycles, number of vehicle miles
traveled, gasoline and fuel oil consumption for the periods over which
samples were taken would be helpful. Much of this information is
available and should be summarized as part of the IP Network's data
record.
One of the most important issues deriving from a new standard for
inhalable particulate matter will be the extent to which industrial
sources are responsible for any violations of that standard which are
observed at particular monitoring sites. Some initial estimates of
industrial contributions in typical urban areas can be obtained from
the chemical concentrations acquired through the IP Network. For
these data to be meaningful, however, it is necessary to gain a more
detailed understanding of the industrial source character of
individual urban areas.
4.2 Industrial Emissions in Several Urban Areas
Though it is possible to hypothesize the potential contributors
to measured suspended particulate matter chemical concentrations,
there is no substitute for a knowledge of the types of sources within
an urban area and the locations of those sources with respect to the
receptors. All urban areas will contain emissions from area sources
such as motor vehicle exhaust, residential combustion of gas, oil, or
wood, and resuspended geological material from roadways and vacant
lots.
The specific area sources most likely to affect the measurements
at a specific sampling site are those in immediate proximity to it.
These roadways, homes and open areas are best identified from the
individual site surveys described in Chapter 3.
4-3
-------
Industrial point sources, however, are not homogeneously
distributed throughout an urban area. Emissions are often from stacks
which provide more opportunity for pollutants to be transported
further than those from ground-level sources. The contributions from
point sources are highly dependent on source-receptor geometry with
the receptor being downwind of the source during the sampling period
(Gordon, 1980). When conclusions are drawn concerning the likely
sources of chemical concentrations measured by the IP Network, as they
will be in Chapter 9 of this report, it is important to know the types
of point sources contributing to an urban area and their locations
with respect to the samplers.
In several cases, this has been done in great detail. Portland,
OR and St. Louis, MO have detailed maps showing their major industrial
point sources. IP Network sites ORPOA and ORPOB identified in
Table A.I of Appendix A, correspond to sites 1 (exactly) and 3
(nearby) in Watson (1979) and IP sites MOSLA and MOSLB correspond to
sites 119 (nearby) and 111 (exactly) in Dzubay (1980). On a
neighborhood scale, Grantz (1981) presents a map for the IP monitoring
sites in the vicinity of ARMCO Steel in Middleton, OH. IP Network
sites OHMIA, OHMIB, OHMIC, OHMID, and OHMII correspond to Grantz1
sites 8, 6, 5, 4 and 3 respectively. Many other such maps may exist
in more obscure references which remain to be identified.
Seven such maps showing major industrial sources, their types of
emissions, and relationship to IP sampling sites have been prepared
for this study and are presented here. The cities were chosen because
they are representative of the western, eastern, southwestern, and
southern parts of the United States and contain varying degrees of
industrial activity. The basis for identifying and locating the
sources was the condensed point source listing of the National
Emissions Data System (USEPA, 1980a).
The information in this listing is supplied to the EPA by state
and local air pollution agencies. The listing contains the industry
name and address, its location in UTM coordinates, estimated annual
emission rates for all criteria pollutants, including TSP, and a
source classification code (SCCXUSEPA, 1972) which identifies the
4-4
-------
process or function which causes the emissions. Several of these
processes may be present in a given industry.
In certain cases, the SCC is not adequate to identify the
process. A Standard Industrial Classification (SIC) (U.S. Office of
Management and Budget, 1972), which identifies the primary product or
activity of an industry, is used in these cases. Only one SIC is
assigned to a company even though that company may have several types
of processes, and therefore several SCC. The SCC is preferable as an
identifier of particulate matter source types.
As part of this study, NEDS inventories for Philadelphia,
Birmingham, Phoenix, Buffalo, and Los Angeles metropolitan areas were
obtained from EPA/OAQPS. NEDS inventories from Denver, Houston, and
El Paso were obtained from the Colorado and Texas state agencies as
part of other projects (Heisler et al, 1980a; Heisler et al, 1980b;
Heisler et al, 1981) and were adapted to the format of this study.
Presumably the information in the state-generated inventories and
federal inventories are the same, though feedback from officials in
Colorado, Texas and Pennsylvania indicate that the state inventories
are more accurate because they are more frequently updated.
Because so many point sources (literally thousands, most of them
being small boilers in apartment houses, schools and hospitals which
should be treated as area sources) are listed in the inventories, a
selection criteria was devised to narrow those chosen to the most
likely contributors. For Birmingham, Phoenix, Buffalo, and
Philadelphia, only sources with TSP emission rate estimates exceeding
5 tons/yr were drawn from the inventories for location on maps. In
Houston, El Paso, and Denver, many of the point sources which met this
criterion were more similar to area sources than to point sources.
For these areas, a 50 tons/yr cutoff was used. The Los Angeles
inventory is nearly as large as those of all other cities combined and
it was only half completed by the time resources available in this
project were expended on it, so results are not presented here.
Records in the NEDS inventories are dated 1975 to 1977 and it is
recognized that TSP emission rates calculated at that time would be
substantially different today. For example, a source in Philadelphia,
4-5
-------
the NE Incinerator, is no longer in operation even though NEDS shows
it emitting 356 tons/year. However, the NEDS emission rates are of
value in that they reflect the potential contributions of different
source types even though the recorded emission rates may not be
current in all cases.
Staff members of the Texas Air Control Board and Philadelphia Air
Management Services reviewed the emission inventories in El Paso,
Houston, and Philadelphia and found the point source types and
locations specified by the NEDS to be correct, for the most part,
though some errors in location and present level of activity were
found. This local agency review is still needed for the Birmingham,
Phoenix, Denver and Buffalo inventories.
Since measurements at IP sampling sites include elemental and
ionic concentrations, and in a few special studies might involve
elemental and organic carbon measurements as well as particle size
segregation, knowledge of the major chemical species and the dominant
size range (FINE or COARSE) constituting the emissions from these
potential urban sources is required to relate them to the ambient
concentrations. After the source types in the seven urban areas were
identified, the aerosol source characterization literature was
reviewed to determine the major chemical component and dominant:
particle size of each source.
Table 4.2.1 presents the potential sources of inhalable and fine
suspended particulate matter, the dominant particle size and chemical
composition of emissions from those sources according to the given
references, and a number corresponding to the point source (•) on one
of the maps in Figures 4.2.1 to 4.2.7.
Each number represents a facility which could (but not
necessarily does) contribute to the given particle size and chemical
concentrations measured at receptors. The IP sampling sites((x))are
also located on the maps with numbers which correspond to the site
codes in Table 4.2.2. Reference to the site survey summary, Table
3.1.3, gives an idea of the likely areas sources affecting each site.
With the information in Table 4.2.1 and the geographical
relationships of sources to receptors specified in Figures 4.2..1 to
4.2.7, it is possible to state some hypotheses concerning the
4-6
-------
i-H
«N
-*
a
3
H
to
<
u
5
B>
a:
a
en
D
Z
M
1
H
£
B
5
B
M
H
1
cu
Q
Id
O
Z
u
&
en
3
en
u
z
M
h
e
5
3
3
z
M
O
en
u
B:
o
en
^
H
Z
H
s
o
J5
c
CO
c
•H
e
o
a
rH
CO
U
•»•«
e
5
01
u
fe
0
to
CO
a.
iH
U
c
0
CO
3
O
EC
•H
rH
•H
£
CH
o
iH
CO
U-l
3
CO
(U
Q
X
•H
c
O
£
S
CO
00
c
E
fl
«H
cc
OJ
t-l
cy
t*H
CD
4i
CJ
a
0)
•H
S
X
H
^
£
£
0
"
[O
•^
j
<
s
H
0)
O
t-l
3
0
c/>
CT*
r-*
ON
t-H
.^
e
o
CO
*J
*
*
•H -H
CO H
* A
rH 0)
< U.
w
en
I
CM
0
i-H
0
CN
CN
f>
O
00
o
-00
m o
o
* I*--
-* 0
o
.sD
< 0
00
CN
<•
CN
*
^
CT\
t-H
> —
tH
O
^1
H
•,
CJ
o
*.
u
* TH
* H
(0
4-1 i-H
CU 00
o\
C rH
O ^-(
cn
IH X
(U CQ
•H £
CO iH
U 0
11 <•
UQfHO h CJ O CJU
ocnuo PQOO oo
*.*•>*, *.*,*. * *
CJffltOCJ ^ CJ CJ CJCJ
W bU
CO CO
p£ p£ M U f*1 f*1 I"*1
< < 2 2 Z 2 Z
O O *-l I-* M hH M
CJ CJ f* fa fe Ln Cb
O
w
r*.
O
rH
o <•<<<<
< < < < <
O^ rH
^ ^ ^ ^ -^
< < < < <
< < < < <
CN
rH < < < < <
00
c
•H
J= Tt
U QJ
*J T3 J^ C W
nj Ct3 co CJ-HO.*
•^ Oil) CU 303-HW
eo ^ T3 rH tjj-pCcy
cu at 4-1 co c *j 4-1 a; w
W > CO CU O x-' -H
QJ -H*JrH C i-t fH AJW
i-i cc LJ cu ty • cu
o reoo o wcoej-j^
C l-i O.4JW4JCO CU OJ « ^ CO
cj < eH
CU
rJ ^H
«
C
o u
> 01
„
O -H
Z
- 1
w>
Id
2
i— (
b
^D
O
fc
*tf
O
rH
0
rH
O
S3
So
CM 00
0 0
rH U~i
O O
O
rH
CNJ
O
r-t
O
0 0
CN "X>
O 0
rH
•r*
0 §
iH rH *H
CU CO W
3 3 CO
E^ TI 3
•H ^
rH M g
CO PS CJ
CO
o •
E*-i CC
m
5
rH
V^ '
c
o
CO
4J
a)
3
•*
O
3 rH Si^ £ 3 S
sssssss
csi-d-^rHmvocomco
^H^^eNCNCNCNfncn
rHc^imocN**r^--aFvc
rH rH rH CN CN CN CN J*1 (*1
ON •*• vTi ON r^
O rH rH rH CN
oo m in r^ CN
O rH rH i-H CN
4-7
-------
o
CJ
10 4>
rJ O •*
O i-l O
T-> B 0
ffl V O.
CJ
o
- IW
O O
u en
•H tocj
CO S! u
•H W CJ
<: &. o
Cj
u
u
O
CJ
u
CJ
O
CJ
Ed
CJ
U
CJ
o
CJ
u
CJ
u
B rH
CO U
B ^ en
o a
a a,
g
Ex
u
en
8
u
2
3
en
o
CO
O
B
O
n X
3 H
S
CM in *
o o i
*O ff> ON iH \O
O CM -»• U"i in
O
-»
s
to
S
r*. M
-------
41 ON
00 r-{
41
§8
» «
u u
CO 4>
3
i
10 IU
o -rt
-H O
e
CJ
o
u
JJ
o
u
u
£
CJ
o
cj"
u
b
^
CO
rH
-i to
o a
a PM
I
u
CO
8
z
rH
b
•O
41
3
C
3 H
0
-*r en
o CM
co in
o N
£M O NO CO
O IH en ^r
t-H -a- 00
O CO CO
in co o
O iH ^
H
0)
O
r-l
3
0
vO
(N
O
CN
C
•H
to
c6 a1
O t£ *H
•H i— (
CC 0) C
.u 4) nj
01 En BC
OO
V •
•
01 00
c c
U CO
TO
u >
3 M
*O 4J ^
C C 4)
M i-l -U
MH (0
>-, (U 0)
> Pd ffi
to
4) •
X 10
•
00 ON
O IN
i — r^
0 CN
rH in
O CM
B
41
o
4)
0
CJ
,
-Q
rH
CM
•^ rH
0 CO
CO 00
O CM
00
rH C
tO -rl
i* rH
41 T3
C C
•H tO
£ X
,
0
r Smelting 14
41
a
a.
o
CJ
.
T)
Smelting 14
duction
0
•O ti
a o-
41
)J bO
.
4)
•^
rH
num Smelting 14
Galvanizing 11,
•H
§U
B
< N
, ,
^ exj
vD
o
CNJ
0
rH
0
r-(
D
0)
4J
K)
C
0
M
,
_c
^ in ON
O r~< rH
oo -J- r*.
O «— < » e^4
CM rn
m o
CN CO
m o>
CN CN
60
C
U
fH
3 4>
C CJ
rH 3
•rf b
,
•H
4-9
-------
00
r^.
HIXT-
.0
B-H
V Q
01
b 4-J
u oi
B iH
NCJ
•H U
e 01
o> a
CJ
u
CJ r-l 3
O < CJ
CJ.O -O
HI CM CJ
u
z
bl
M
U.
g
[X
u
2
O O O O »H
r-t CO %O 00 O
o o o o IH
r-t ^H en IT,
§
U
•»
a
b
01
C8
o> <
o
s
fH f^ 00 f^ 9*
O O iH (M tS
OS
O
;00
(N
U
U
U
«
e
u
^
CO
rbon
rnace
CJ [i
j^
01
a
01
X
K
a
u
fH
B
01
Q
B3
•H
CJ
e
B
•H
to j*
3 B
0 Z
a
B *->
n 01
o> to
U CJ
(0
•H •
X to
•o
B
O
•H
rH U
a a
CJ b
IS.
01 01
.e b
CJ IX
.0
3 -H h
•O b o
O 3 w
p fH 4J (0
&. a u M
•H B 01
-J 1J iu B
E II 3 -H
•H J-> C CJ
a a a B
OH £ Z M
00
X B
o a
4J
« O
u (C
£9
6 E
4-10
-------
i E 01
it o
j= to
rg .3-
O O
C
O
u
•«
s
?
3
o
C/J
c
o
u
3
^j
O.
p_l
p
CO
CD
CO
OJ
TJ
3
C
ion, petroleum processing, gypsum, asphalt rotary dryer,
4J
CO
O
^4
*0
U-i
(3
0)
E
to
4*
T3
3
C
ferromonganese alloy processing.
f.
bO
C
r^
c
JC
Of
o
0»
tfl
0)
c
m
in blast furnace.
•M
K
0>
JJ
to
c
o
• ^
CO
Ol
3
C
cturing, molybdenum-sul fate roas t ing, PVC product ion , agr
«
3
C
E
U
(0
CO
a.
to
0)
T!
3
C
, detergent .
00
c
l-l
3
4-1
U
CO
U-J
3
C
CO
a>
TJ
u
• .-4
at
ent , motor vehicle parts and accessories , scrap and waste
E
a.
•1-4
3
cr
00
c
a
E
3
a.
CO
"O
3
C
^
V
c
l-l
3
JD
£
U
c
0)
^
4_l
l-l
o
u
cfl
Vj
01
c
u
c
(0
Jj
i_t
(0
3
TJ
C
T3
C
(9
O
tfl
0)
c
u
c
• r-i
.— (
to
a.
••^
u
c
3
E
to
a>
T)
3
C
OC is Organic Carbon.
f,
c
o
j_i
o
p—l
to
iJ
c
0>
E
0)
w
(0
0
Iz3
o>
u
l-l
3
0
(0
cB
-------
(s-^/)sampling Site
• Point Source
BIRMINGHAM AL
Figure 4.2.1 IP Sampling Sites and Industrial Point Sources in
Birmingham, Alabama
4-12
-------
*/Sampling Site
Point Source
Figure 4^2.2 ip Sampling Sites and Industrial Point Sources in
Phoenix, Arizona
4-13
-------
Site
• Point Source
Figure 4.2. 3 IP Sampling Sites and Industrial Point Sources in
Denver, Colorado
4-14
-------
:££Lake Erie ;:;x;:;:;:;:;x;:::;:;:;:;:;:::o:;!;:;x;:;:|x:S::
US 5
Sampling Site
Point Source
Figure 4.2.4 IP Sampling Sites and Industrial Point Sources in
Buffalo, New York
4-15
-------
xJSampling Site
Point Source
Figure 4.2.5 IP Sampling Sites and Industrial Point Sources in Houston, Texas
4-16
-------
t/i
03
X
oj
D-
c/1
0)
o
fn
O
c
•H
O
1—I
rt
•H
^
+-)
in
3
oj
4-1
•H
c/:
•H
I-H
I
ex
\o
CM
3
bfl
4-17
-------
56 •
0 1 2 Miles
PHILADELPHIA, PA
(/Sampling Site
Point Source
Figure 4.2.7 IP Sampling Sites and Industrial Point Sources in Philadelphia
Pennsylvania.
4-18
-------
TABLE 4.2.2
IP SAMPLING SITES IN SEVEN URBAN AREAS
City, State
Map ID
Site Code
Birmingham, AL
**
••
»
»
Phoenix , AZ
••
ii
Denver, CO
••
Buffalo, NY
••
••
ii
Houston, TX
••
>t
El Paso, TX
Clint, TX
Philadelphia, PA
"
"
••
••
••
••
••
••
••
"
"
••
••
1
2
3
4
5
1
2
3
1
2
1
2
3
4
1
2
3
1
2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
ALB I A
ALBIB
ALB 1C
ALBID
ALB IE
AZPHA
AZPHB
AZPHC
CODEA
CODEB
NYBUA
NYBUB
NYBUC
NYBUD
TXHOA
TXHOB
TXHOC
TXELA
TXELB
PAPHB
PAPHI
PAPHJ
PAPHK
PAPHL
PAPHM
PAPHN
PAPHD
PAPHC
PAPHE
PAPHG
PAPHH
PAPHF
PAPHA
Site Name
South Birmingham
Inglenook
North Birmingham
Mountain Brook
Tarrant City
Carefree
Maricopa County Health Dept.
North Phoenix
Downtown
Gates Rubber Company
P. School No. 26
Big Sister Sewage Treatment Plant
P. School No. 28
Wilmuth Pump Station
Aldine Mall Road, CAMS 8
CAMS 1
Seabrook
Tillman
High School
Allegheny
St. John's Catholic Church
Pilot Freight Motor Company
N.E. Wastewater Treatment Plant
T&A Pet Shop
N.E. Transfer Station
Bridesburg Recreation Center
S.E. Water Treatment Plant
Belmont Filter Plant
North East Airport
Presbyterian Home
Temple University
H. Gratz College
500 S. Broad
4-19
-------
difference in chemical concentrations that should be observed between
sampling sites and urban areas if certain industrial point sources are
making significant contributions to fine and inhalable suspended
particulate matter loadings. If these differences are found in the
receptor measurements, then there is a good possibility that the
sources may be contributors and further work should be specified if it
is deemed desirable to confirm and to quantify those contributions.
If significant differences are not found between sampling locations
with different source characteristics, then the point sources with
unique chemical characters are probably not major contributors
compared to the more homogeneously distributed and common area sources.
One difficulty with this approach is that many of the emissions
listed in Table 4.2.1 are the products of combustion and have no
distinct chemical character other than their elemental and organic
carbon content. This does not allow them to be distinguished at the
receptors, and since routine carbon measurements are not made (and
cannot be made on the sampling substrates presently used in the IP
Network), an upper limit estimate of their contributions is not yet
possible. In spite of this limitation, Table 4.2.1, Table 4.2.2 and
Figures 4.2.1 to 4.2.7 contain a substantial amount of information
about the possible and impossible source-receptor relationships.
Table 4.2.1 shows that the seven cities under consideration
contain a large number of point source types as well as the usual area
source types. Reading across the rows horizontally gives an idea of
the number of source types in all of the urban areas. Rock crushing
is prevalent in the western cities of Phoenix and El Paso, but not so
common in the eastern cities. Major (recall that smaller units are
not listed) residual oil combustion sources are concentrated in the
eastern and southern cities. Feed and grain handling are minimal in
Denver and El Paso while fertilizer production occurs only in
Houston. Mineral handling and iron and steel foundries are sources
common to several cities. Coal combustion is most prevalent in
Philadelphia and Buffalo while most of the forest products operations
are in Houston. Coke ovens are found in Philadelphia, Houston, and
Buffalo. Philadelphia has the largest concentration of smelting
operations.
4-20
-------
Table 4.2.1 provides an idea of the particle sizes to be expected
from different primary sources. The COARSE particle samples should
receive their major contributions from geological material, wood and
paper products, feed and grain processing, pollen, and mineral
handling. Though other sources contribute some COARSE particulate
matter, their contributions to the FINE samples should be greater than
their contributions to the COARSE samples. Most of FINE contributors
are point sources, though the residential heating and motor vehicle
area sources also play their role.
Concerning chemical composition, the COARSE particulate matter
concentrations should be rich in the geologically-related species of
Al, Si, Fe, and Ti, with additional concentrations of carbon from feed
and grain handling and pollen. In certain cases, mineral handling
operations near a receptor site may raise the concentration of a
particular element in which the particular mineral or ore being
handled is enriched.
The FINE chemical concentrations will not present such a rich
variety of chemical concentrations. The primary constituents of most
fine particle sources are organic and elemental carbon. In certain
cases, smelters or steel operations will contribute Al, Pb, Cu, Zn,
Fe, and Mn to this fraction. Pb and Br can be expected from
automobile exhaust and V and Ni will often be contributed by residual
oil combustion. Naturally, some geological elements will be collected
in the FINE particles. FINE particulate matter concentrations will
also receive significant contributions which do not result from the
primary emissions summarized in Table 4.2.1. Sulfate, nitrate, and a
certain amount of the organic carbon found (Stevens et al, 1978) in
ambient samples has been shown to result from the conversion of vapors
to particles in the atmosphere. Thus, even though a large variety of
sources emitting sulfate and nitrate do not appear in Table 4.2.1,
concentrations of these species should be expected on ambient samples.
Reading down the city columns of Table 4.2.2 and relating the
source locations to the receptor locations for individual urban areas
by using the appropriate maps of Figures 4.2.1 to 4.2.7, intimates the
types of chemical species which would be observed if the nearby
sources were making contributions to each receptor.
4-21
-------
In Birmingham, AL, all sites except site 4 are surrounded by a
number of industrial point sources and might be expected to receive
some contributions from them. Because there are a large number of
iron and steel foundries and blast furnaces, chemical concentrations
for Fe, Mn, Cr, and Zn are expected to be higher at sites 1, 2, 3, and
5 than at site 4. Residual oil burning is associated with the
industrial activities which would elevate V, Ni and carbon
concentrations. The zinc smelter (11) would contribute Zn to sites 3,
2, and 5 more than to sites 1 and 4, particularly with wind directions
from the north or northwest. The transportation activities associated
with the industry would contribute to organic carbon, elemental
carbon, Pb and Br concentrations. Site 4 is a suburban residential
site which is the best reference with which to compare concentrations
at the other sites.
Compared to Birmingham, the population (referring to Table B.2 in
Appendix B) in the Phoenix metropolitan area is twice as high but
there is much less industrial activity. The major industries in
Phoenix are feed and grain handling, mineral handling, rock crushing
and aluminum smelting. Thus, point sources could be expected to make
their major contributions to the COARSE particles. Site 2 should
receive the greatest contributions from most sources including the
rock crushers, mineral handling, feed and grain facilities and
automotive emissions from major highways. Site 2 should see higher
concentrations of Al, Si, Fe, Ti, Mn, and organic carbon in the COARSE
and Pb and Br in the FINE than should be seen at the other sites. The
aluminum smelter (3) may contribute Al to the FINE under westerly wind
conditions. Residual oil combustion is low, Ni and V concentrations
should not be as high as in other cities with more of this activity.
Site 1 is a rural remote site which can be used as a background
reference for the other two sites.
Denver is another city with few major industrial point sources.
Mineral handling and the lead smelter (5) constitute the major
operations. The population of Denver is high and the population
density as shown in Table B.2 in Appendix B is twice as high as in
Phoenix. Also, the location of the city of Denver is unique; it is
elevated and surrounded by mountains. The amount of COARSE
4-22
-------
particulate matter and chemical species like organic carbon, elemental
carbon, Mn, Al, Si, V, Ni, and Fe might be elevated at sites 1 and 2
due to the mineral handling facilities. The proximity of Pb smelters
to site 1 could affect the Pb measured. Because of the high
population density downtown, the area sources, like transportation and
street dust contributions, might be elevated. Denver has exhibited a
significant photochemical contribution and possible contributions from
uninventoried wood combustion (Courtney et al, 1980, Heisler et al,
1980a, 1980b). FINE concentrations of elemental and organic carbon
could be due to these sources. Neither site 1 nor 2 is an adequate
reference site so it will be difficult to determine increments due to
urban-scale sources in Denver.
The population of Buffalo is lower than that of Phoenix and
Denver but the population density is very high; the weather is also
very cold. This might cause FINE organic carbon, elemental carbon
concentrations due to residential heating and Pb and Br concentrations
due to motor vehicle exhaust to be elevated. Buffalo contains the
largest number of coal combustion facilities, and FINE As
concentrations might be elevated if they were contributors. Residual
oil combustion units are widespread, and FINE V and Ni concentrations
at sites 1 and 3 might be higher, because of their proximity to these
sources, than the corresponding concentrations at site 4. Steel mills
and blast furnaces make Buffalo similar to Birmingham and Mn, Fe, and
Cr might be enriched at all sites in the area. Buffalo receives
contributions from air masses passing over the Ohio River Valley where
substantial quantities of SO- are injected into the atmosphere by
power plants (Mueller and Hidy et al, 1981) and can convert to
measurable sulfate by the time they reach Buffalo (Kolak et al,
1979). The background site for Buffalo is NYBUB, site 2, which is
30 km south of Buffalo's city center.
The Houston area contains a wide variety of source types.
Natural gas combustion is a major source because of its association
with petroleum refining, and whereas contributions from this
clean-burning fuel will be low in other areas where its use is
primarily for domestic heating and cooking, elemental and organic
4-23
-------
carbon contributions to FINE samples from this source might be
significant in Houston. Sampling site 2 is closest to the highly
industrialized ship channel and might expect elevated levels* of Fe,
Mn, and Cr due to the steel foundry. The background reference site 3
is classified as rural residential and though nearby natural gas
combustion sources might affect its carbon concentrations, it: should
serve as a good reference with which to assess the industrial
influences on the other sites.
The effect of emissions on the air quality in El Paso cannot be
considered independently of the city of Juarez which lies just across
the Rio Grande River in Mexico. Emissions in Juarez are not. well
identified nor quantified so little can be said about them here.
Site 1 might be influenced by the nearby smelter (1), though Hubert et
al (I981a, 1981b) do not find Pb and As measurements at this site to
be enriched (they did not measure Cu or Zn). The steel foundry (3) is
too far away to have much influence as are the fluid crackers (5 and
6), concrete batching (7) and rock crushing (2 and 8). The background
reference site is in Clint, TX, 45 km southeast of the El Paso city
center.
Among all the sites under consideration, Philadelphia has
(referring to Table B.2 of Appendix B) the highest population per
square mile. Industrial activities are numerous and include mineral
handling, residual oil combustion, coal combustion, incinerators, and
various metal smelters. Philadelphia is also an important area
because of the high concentration of IP sampling sites, special
studies and long-term monitoring.
The long-term sampling site 14 at 500 S. Broad is removed from
most point sources. It might show enriched FINE concentrations of Pb
and Br from local traffic, but few enrichments from industrial
sources. Site 11 in a residential area should exhibit elemental
concentrations similar to those measured at the background reference
site 10, though the mineral handling facilities close to site 11 may
cause COARSE Al, Si, Fe, Mn, and Ti to be enriched over the levels at
site 10 and the FINE fraction at both sites may be enriched in V and
Ni due to the nearby residual oil boilers.
4-24
-------
A special neighborhood-scale study was carried out in the
Bridesburg industrial area which is shown enlarged in Figure 4.2.8.
The Bridesburg area of Philadelphia is a combination of heavy
industries, residential and commercial land uses. It covers
approximately a 2 km x 4 km area 10 km northeast of the city center.
As a special study in EPA's Inhalable Particulate Network, a
seven-station, four-month sampling program was carried out in this
area. The most interesting effects of sources on receptors are
probably to be observed in the Bridesburg area. Grain loading (33)
might contribute to COARSE carbon concentrations at site 1 while
manganese ore handling (59) and the zinc galvanizing (31, 32) might
contribute to sites 1 and 5. The copper and secondary metals smelter
(34) may elevate Zn, Pb and Cu concentrations at sites 4, 5 and 6.
The coke oven (40) with its associated coal storage piles would most
likely elevate FINE and COARSE elemental carbon concentrations at
sites 2, 3 and 7. The lead smelter (38) would be expected to raise Pb
concentrations at 2, 3 and 7 more than at the other sites. This
neighborhood-scale study offers a good opportunity to determine the
spatial variability of source contributions to ambient concentrations.
Once again, this examination of individual urban areas is a
screening rather than a definitive statement of which sources affect
which receptors. Little is known about the operating cycles, control
equipment or meteorological conditions prevailing during the times
when the airshed was sampled by the IP Network. The value of this
screening is in determining which source types need not be further
investigated within an urban area. For example, Phoenix, with few
industrial sources, presents a much simpler situation than
Philadelphia with many and varied industrial sources.
There is value, too, in comparing the source characters and
ambient concentrations of different cities with each other. If
chemical concentration measurements are similarly enriched between
cities with many different source types but with a few in common, then
the common types must be causing the enrichments. Similarly,
differences ia chemical enrichments which correspond to those expected
from the different source types will support the hypothesis that those
4-25
-------
O in
co f->
3
M O
C CO
C. C
03 O
CO Q.
3 -H ™
-° *4
to -d ^
ID CD
13 -H 0)
•H <4-i i-i
!H -H ^2
CQ 4-> oj
ID a)
45 T) C
•4-1 I—I -1-1
C 0) TJ
•H fH 0)
rt -H
10 <4H
C W -H
O (U 4-1
•H 4-> C
+J .H V
Cd CO T3
O i-i
o
>-J • CD
rt M
flj -H nj
O JZ
O -
C/3 T3
fi -H
•H X
O O-i
o.
CD
4H O
T3 O CO
ni cd *o
0) C
a> M ca
•H tN
c M TT
• H 4J
rH (/) 0
^ nl C
~ CO i-<
6/1
•H
u.
4-26
-------
additional source types do contribute to ambient concentrations of
inhalable and fine suspended particulate matter.
The discussion in this chapter results in several observations:
• The urban areas sampled by the IP Network represent a broad
coverage of population, meteorological, and emissions
cases. Though not all urban areas in the U.S. are
represented, the ones that were chosen include major
population centers with varying population densities. Major
particulate matter sources have been identified and located
on maps with respect to IP Network sampling sites in ten of
the network urban areas, and seven of these maps were
presented in this chapter. Similar maps for other urban
areas in the network should be compiled.
• Several of the sites in the IP Network were found to be in
proximity to industrial sources which emit chemical species
measured on IP Network samples. If these sources are
contributors to suspended particulate matter concentrations,
the concentrations of chemical species which they emit
should be elevated above those at more distant sampling
sites. This analysis is done in Chapter 9 of this report.
• Some of the sources located in this document may not have
been in operation during IP Network sampling. The National
Emissions Data System is a useful adjunct to the IP
Network. Many of the entries are dated, however, and the
report formats are not convenient for the receptor-oriented
assessments of particulate matter concentrations, so
considerable effort is required to extract information from
it.
4-27
-------
CHAPTER 5
GEOGRAPHICAL AND SEASONAL VARIABILITY OF
TSP, IP, AND FP IN URBAN AREAS
The spatial and temporal distributions of the size-classified
samples acquired in the IP Network can provide information on a number
of issues related to TSP, IP and FP concentrations:
• Standard Attainment: The IP Network is research oriented
rather than compliance oriented, and the measurements
obtained from it are not intended for testing an area's
attainment of a standard. However, the distribution of
average and maximum concentrations provides an estimate of
which areas are most likely to be in or out of compliance
with a standard and can serve as a preview of that which
routine compliance monitoring might eventually find.
• Causes of Elevated and Depressed Concentrations: The
variation of concentrations with meteorological subregions,
urban emissions and time of year provides a basis for
speculation on causes of observed particulate matter
measurements which can be supported or refuted by further
study.
• Sampler., Spacing: The spatial variability of ambient
concentrations within an area allows an estimate to be made
of the density of samplers required to truly represent the
pollutant concentrations in that area.
• Sampling Schedule: Different seasons of the year are
influenced by different meteorological and emissions
conditions. An annual average will be affected by the
frequency of samples taken within each season if the
seasonal averages differ substantially.
The seasonal and annual averages drawn from urban sampling sites are
presented and examined in this chapter to address these issues.
5.1 TSP, IP and FP Urban Concentrations
Data collected by the IP Network present several opportunities
for the examination of spatial distributions of TSP, IP and FP on
regional, urban and neighborhood scales. The network's widespread
geographical location of samplers allows concentrations from various
urban areas in the United States to be compared and regional patterns
to be discerned. Availability of a complete year of data from two or
5-1
-------
more sites in the environs of Birmingham, AL, Los Angeles, CA, San
Francisco, CA, Buffalo, NY and Philadelphia, PA provides insight into
urban scale spatial distributions of particulate matter. The special
study of the Bridesburg Industrial area of Philadelphia provides
similar insight into spatial distributions on the neighborhood scale.
Table 5.1.1 contains the arithmetic average concentrations of
TSP, IP and FP at sites in the IP Network for a full year of
monitoring beginning on October 1, 1979 and ending on September 30,
1980. The IP and FP concentrations correspond to the TOTAL and FINE
measurements at the sites. At the time of this writing, SSI
measurements of IP were much less complete than the TOTAL measurements
of IP. Each quarter's values have been averaged separately to reveal
the seasonal distribution. Only averages containing more than five
values have been included. The annual arithmetic average was
calculated as the average of the four quarterly averages rather than
from the entire set of data for that year. Early in the network
operations, many samples were lost due to start-up problems, resulting
in a nonuniform distribution of samples between quarters. This
averaging procedure gives each season equal weight in determining the
annual mean, regardless of the number of samples acquired.
Table 5.1.2 contains the maximum concentrations of TSP, IP and FP
found during each season. Only maxima from quarters with more than
five TSP, IP and FP values are listed. The annual maxima are the
highest values which occurred in any one of these quarters. The data
in this table were taken from summary reports produced by EPA/EMSL and
were not screened by the procedure of inequalities 3-1 to 3-4.
5.2 Regional Geographical Distributions
The urban areas have been grouped by five geographic regions of
the country. At those sites for which annual arithmetic averages were
calculated, the highest average IP concentrations were found in Los
Angeles (92 ug/m3), El Paso (68 ug/m3), Buffalo (63 ug/m3) and
Birmingham (58 ug/m3). There is no clear regional pattern but a
wide variability in concentrations between urban areas is seen.
The average of FP at urban sites ranges from 13 ug/m3 in San
Francisco (Richmond) to 37 ug/m3 in Los Angeles. The regional
5-2
-------
cn
H
cn to
4J~
U 01
O B
e z
10
£
S
o.
H <0
H o
0
cn 4-t
o- u
n
z 01
M b
CO
IX
rb Pu
[K
c
o
flj
4)
Jj.
CU
rH
. *-*.
CM vO 1
vD v£>
CO
CM CO r-
CM CM CO
rH O CO
^ CO vD
CO OS 1
m m
J
O Cu
H C
O * O «T)
•O C -U JO
* J-( O 60 00 O
•0 O W C K »
jj )-< MH 60 -H 3 i— 1
10 O -U C JT -X.1 4)
(Q IH I- -H K CO N
x tJ ffi »s £j ac
M EC IS Pu
o
2s
CM
CM
CM
O
iH
C
0)
00
0)
1— 1
<
CO
CM
"••*.
ro
OS
•*
CO
CM
--*.
m
<»•
-^
m
vO
Os
1-1
in
CO
--^
CO
m
.H
CM
CO
CO
in
CO
-*
CM
*3"
co
•^.
CM
>3-
-U
0
D_
h
*H
^J
W
rZ
CM
CO
^,
CO
X> W3
Os sD f1^ O Os
CO CM CM CM rH
iH CM 00 in CO
in *^ "^ ft co
"*-. *^ ^s. -^ *V.
CO IT| | CM ^.
rH 0
m vj-
CM CO
rH
-3"
CM
O
m
CO
r-.
CO J
CM 4J
CO rH 4-1
DH 3 a- os
CM »* m
~^. — *-*.
m <± os
m os r^.
-a- m os
rH r-t rH
•^ -v. -^
r** m o
CO ^ ^
^* -^ -^
CM iH O
r-x os os
CO CO rH
rH CM CM
-^. "^ "^
O CO iH
CO ^ CO
-j- m co
>3- vO «^>
^- \o
iH rH
*^, ^^
m r-.
CM CO
-^ *-^
CO CO
co m
g<«
m co o
K ^_( *H y*
« rH rH
•H 0 O
rH O. CX. W
0 (TJ « «H C
d. 0> 4) 30
ft) C C O U
-------
41
cn _ __
I > 31
•a
CM
•^
O
sO OS Csl
CM *~< CO
in O oo
in ro m
00
so
•a
CM
O
oo
§
ON
j co co CM co
co o o •» oo
oo m CM so «a
co
•a
SO
•3-
^
00
O
-31
ON
oo
so
IN
ON
00 ^H CO
fM to4>
C "<
41 °)I-H O < 3 to "J j3 b>UCuC
ft 0* —J O M 41 « 3 .ri .ri . «
X ^H « •< ft fe « CiJoicnw
3 ui » 2 cn
O 4>
cn ft
id
4)
cn
41
J-l
cn
a.
b.
D-
B
o
a
o
60
5-4
-------
CM
rH
m
S
00
^
H
to.
O
£
rH
X
_i
§
>4
S
s
W
H
06
&
*
CO
u
H
M
cn
X.
at
9
g
g
O
P_|
M
|
S
as
t-t
X
H
S
i
§
OQ
a:
a
z
M
d,
Cb
Q
i
CU
rH
&T
cn
t*
*+,
CO
^J
c
B
eg
u
s
CO
(Q
X
1.4
41
fH
O.
e
n
cn
tn
:hotoraou
w
^H
a
g
5
u
U-l
£
CO
ft*
fb
•D
C
CO
IX
M
•^f*
C
O 01 ^
OB oocn
<0 W S
01 t* ""o*
cn ft) oo
1 > 3
r^
CN /-v
fH fl
B
» ""^
i-H 00
*H 3
CO N-'
U4
C
CO 0)
OH .0 *J
M t- -H
D cn
CN
Icn i-t
cn rH
^^
-a- NO
cn -3-
-H fH
"*x •"•*,
f- rH
cn so
i-H rH
cn co
*.
-•*. -a1 m
co cn m I rH o I -^ -a
rH CO CO fH r-^ sO O>
f— i o 1 vO MS r*. cn co i
m en cn CM r-% r-. cn
rH CM
CM rH
cn cn CM o rH o
CO P- \O CO Cn iH rH
-3- -^ -^. -^. -V. "-». "V.
•«^ ^ r*. -a •& -H u >, a E - o)
T3 C *J J= O J= 03 C lu O C *M
-i- o oo oo o a a>i-t£j= o>o
•OO WC 1-3 rH*X< O 1- E
4j t- «- oot-t 3i-H tucnoo *•-? o -H
CQ O -J CJC .O.OI T3 QJ.W B W
q u~ii* -^t w XN «
< 0 Z
Z w ^
C ffl -CM
« o o o
C W 00 <-H •
o co tQ cn
•U O • UH *
to CQ U U-i DL.
O 3
co pa
CO
m
^»
fH
rH
rH
i— |
O\
rH
GO
m
"•"K.
m
cn
*^»
cn
NO
rH
r-.
•*
•^i
cn
*H
cn
rH
•* m
m cn
**-^ -x.
^H CM
FH rH
CM CM
in CM
rH CM
CO
cn
0
^-.
-a-
-a-
rH
CO J=
CM 4J
3
* B
Cn rH W
• *H CO
(X 3 0)
3
•O
^H
Z
r*.
«a- -a-
^* -^-
*^fc m
rH O
SO *H
sO *H
CM tN
FH CM
FH
rH Cn
CM -*^
**^ in
a* o
-a- ^H
**»* ^**.
tn rH
CO CM
CM
rH CM
CM CM
-^ -^
*H CM
sO CO
so cn
CM CO
FH «-H
ss- r-
<• ^a-
m cn
tn NO
co cn
cn cn
r- cn
CM CN
-^ -^
CM m
•^ in
•^ "^-^
CM OO
so cn
2 < pa
CQ <0
» «H -H
CO rH fH
•H O O
FH & O.
O CO CO
O. 01 0)
CQ C C
01 C C
C —• -H
c z z
•T-i
£
CO
«tf
in
00
CO
rH
i-H
o
*3-
•••^
cn
NO
r«*
cn
FH
cn
cn
sO
m
-*
cn
i
CO
3 C
0 0
J W
y-i
• <
4J
cn
m
tn
"^.
m
o
FH
CM
rH
m
m
^•^
rH
CO
-a1
rH
FH
cn
-a-
•-^
r»»
r**
-s^
r*.
rH
rH
CO
*a-
m
cn
rH
CM
rH
r-.
-*
*»x.
-a-
•*^
o
o
rH
J=
4J
CQ
0)
rc
X
o x
o>
* i-H
C 1-
o o
h £
^
<
cn
r*»
"^.
cn
CO
rH
CN
FH
m
r*.
****
cn
CO
i— i
CM
FH
r««.
m
•^
•*
sO
*»^
•^>
FH
rH
tn
-a-
<•
sO
CO
sO
PN.
^a1
cn
^^
CO
CO
z
o
* «J
•H ce
w c
IB C
C TH
C U
•H C
O fH
C 0
•H
o
5-5
-------
o a)
co acco
to * S
O M
CO 01
i >
-a- <
o
00
OS
o
oo
OS
a
en
O
CM
CM
r*
—
OS
OS
SO
^
rH
OS
O
rH
o
o
CM
IN
OS
CO
O -<
SO l~.
r-~ rH
00 00
m
CO
01 M
B
•H
y
rH DO
*-! 3
o
CO
sC S
^ in
PH m
o m oo
en CM en
so Os o
r-* m os
CM
U"l O ^ *» CO
m \r\ -^. CM ^>
•^.^^ M ^-^
sC ON O iA W
to r- r-( \o r*
-^ ""s. -^ -^. --^
m CNJ m i in
Cs O •
CM vO <
3
er
u
-a- 01
en i-t
~~~ a
in
c
a
j:
Ch
SO
lO
iH
00
rH
CM
r*-.
*•%.
OA
vO
rH
VO
r-.
*-,.
CO
rH
OA
rH
o
rH
fH
fs.
O
CN
•tf
CO
-.,.
fH
-^
t
O\
t*\
IT)
"v.
ki
0
e
f
»
a
03 QJ
CU »Q 4J
rH rl *H
P W
S •
O £
CJ
•> w
rH B
rH O
•H 4J
> M
C C
01 .£
£ 0)
S «J
CO
^
U *H
U to 3
to 0
to n to
to E »
C S X £
CQ 4J
t^ 3
0
en
B B
>j n «
•< J= XI
00 M
- C C
eio-H -H
B ea pa
•H
E • •
I- en 2
•H
pa
4J
^
rH
ij CO
to oo to
o c to
0 < 3
to <
to 1 1
0)
2
Sx
33
u o
•O T5
to to *o
0) CO 3
a a, £
5
o
u
U 01 4J
•H kl TJ (0 0»
0 0 B 10 «
B B O W 0
CO IH B -1
IH C £ •
t, > U CJ- C
fH fH • tO
c pJ K en «
to
en
4J
no
•H
^ fj
^g
CO
" rH
01 4J
rH 4J
U Cfl
W t)
co en
01
en
u OH
to ~~.
a.
I-l rH
0 -~-
UH PM
•^ en
H
to
oi n
4J fH
TH
en u
to
rH E
rH I-l
< O
li-
ra
5-6
-------
distribution of these annual averages is fairly homogeneous in the
3 .
eastern United States, ranging from 22 ug/m in Birmingham to
3
32 ug/m in Philadelphia. Average FP concentrations west of the
Mississippi River are generally lower, with the exception of El Paso
and Los Angeles.
Maximum 24-hr concentrations of IP vary substantially and rarely
occur on the same day at nearby sites. They appear to be more
affected by local than by regional phenomena. The spatial variability
of these maxima in the West seems comparable to the spatial
3
variability in the East. These maxima range from 58 ug/m in
3
Seattle (Seattle Light) to 200 ug/m in Los Angeles (Rubidoux).
These measured maxima might be larger if sampling were undertaken
daily rather than every sixth day. This subject is examined in
Chapter 7.
3
FP annual maxima range from 44 ug/m at Minneapolis (site A) to
3
112 ug/m at Presbyterian Home in Philadelphia. The variability
within regions of the U.S. is comparable to the variability between
regions.
Thus, with respect to a standard which might be formulated in
terms of an annual average and a 24-hr maximum, all sites would be in
3
compliance if an IP standard specified a 92 ug/m annual average and
a 213 ug/m maximum and if an FP standard required concentrations
3 3
less than a 37 ug/m annual average and a 112 ug/m maximum. For
lower values, the number of potential non-compliance sites could be
evaluated by selecting those with averages and maxima greater than the
standard from Table 5.1.1.
The urban areas with the highest IP average concentrations are
3 3
Los Angeles (92 ug/m ), Birmingham (58 and 55 ug/m ), El Paso
(Texas has a rule for invalidating TSP samples collected during
duststorms which has not been applied to IP Network data. Thus, the
IP average of 68 ug/m and maximum of 213 ug/m may be biased if
3
duststorms occurred) and Buffalo (63 and 52 ug/m ). Of all the
sites evaluated, these same urban areas exhibit the highest 24-hr IP
3
concentrations (maxima are 134 and 111 ug/m in Buffalo, 140 and
3 3
110 ug/m in Birmingham, 99 and 200 ug/m in Los Angeles).
3 3
San Jose (113 ug/m ) and Philadelphia (134 and 146 ug/m ) also
5-7
-------
show high 24-hr maximum concentrations. For the most part, the
remaining urban areas exhibit average IP concentrations of less than
50 ug/m-* and maximum concentrations of less than 100 ug/m-*.
The IP Network was never intended to measure compliance. A
network oriented toward monitoring compliance with a standard might
find higher or lower annual average and maximum concentrations than
those cited here depending on sampler locations and sampling
frequencies.
5.3 Urban Geographical Distributions
The ranges of TSP, IP and FP annual average concentrations and
maximum concentrations within urban areas having two or more sampling
sites are presented in Table 5.3.1. The range of values for all sites
appears in the last row of the table. The variability of both
averages and maxima between different sites within each area is
evident from these ranges. In no case can data from a single sampler
be said to spatially represent the urban area with reasonable
precision. Even for FP, which shows the smallest range of values, the
minimum ratio of the high average to the low average occurs for the
Buffalo area at 1.22. This implies that if only the P.S. 28 sampling
site were used to represent the entire Buffalo area, then average FP
concentrations at P.S. 26 would be underestimated by 22%. The
relative differences between spatially separated averages within the
same geographical area are higher for other urban areas and size
fractions. For example, if the West LA IP average were used to
represent the average IP concentration at Rubidoux, the estimate would
be incorrect by 100%.
Similarly, maximum concentrations vary from site to site within
an urban area. The minimum ratio of FP maxima is in Philadelphia,
where the ratio is 1.13, but in other areas this ratio is
substantially higher. Once again, in Los Angeles the IP maximum in
Rubidoux is twice that of West LA.
The variability among sites within a single urban area does
appear to be less than the variability of all the urban areas sampled,
though it may be unfair to infer this because the 19 sites of the
5-8
-------
rH
en
in
U
1
^^
P4
C*
jK
<«
PH
H
PH
CO
H
ft.
O
W
o
z
^c
PH
MIRATIONS
w
£
o
CJ
M
>$
;§
Q
§
W
1
w
«jj
o
M
H
H
ȣ
H
HH
5j
,
5
3
%
«
pd
13
55
^
B
M
CO
W
H
}_^
CO
SB
w
w
B
w
pa
|
•rt
00
. 0)
P-l OC
tn cO
Ml
vw
|
-_J
•M
en X
8 tj
00
PM 01
M 00
CO
fll
W
•£
g
J3
^
CO JO
6 5s
^•^
00
9
•»
PH 0)
CO 00
H CO
I-i
01
00
c
CO
o
(X iH
01
-o
PH
O
a
0)
c
CO
01
4->
•H
CO
5-9
-------
combined data represent a considerably larger population to draw from
than the 2 or 3 sites in a single urban area; if each urban area had
averages and maxima from a larger number of sites, the comparison
would be fairer. Still, the range over all sites is large. For
averages, the ratios of maximum to minimum are 4.1 for TSP, 3.8 for IP
and 2.8 for FP. For maxima, these ratios are 4.4 for TSP, 3.4 for IP
and 2.5 for FP. The ratios of these overall range limits are
significantly higher than the corresponding ratios within an
individual urban area.
It is evident that the monitoring of urban concentrations of TSP,
IP and FP will require at least an urban-scale (see Table 3.1.1)
monitoring network to adequately represent the spatial distribution of
these concentrations. In certain situations, a smaller scale, a
neighborhood scale, of sampler spacing may be required to adequately
represent an area.
5.4 Neighborhood Geographical Distributions
To understand more fully the spatial patterns within arid around a
heavily industrialized area, the Bridesburg industrial area of
Philadelphia was the subject of an intensive sampling effort: from
October 3, 1979 to February 15, 1980. Seven sampling sites were
located within a 2 km x 4 km area. Three sites were within the
Bridesburg industrial area and four other sites were on its
perimeter. These are shown in relation to each other and nearby
industrial sources as sites 1 to 7 in Figure 4.2.8. Sites 1, 3 and 4
are the core sites and sites 2, 5, 6 and 7 are the perimeter sites.
An urban background site, NE Airport, was located approximately 10 km
away to the northeast at a small airport in a less densely populated
area in a generally downwind direction. This is site 10 in
Figure 4.2.7.
Figure 5.4.1 compares the average and range of concentrations of
TSP, IP and FP within the core, perimeter and NE Airport subgroupings
as a function of distance from the industrial area. The data used for
this figure did not go through the screening process of inequalities
3-1 to 3-4. The ranges show substantial variability within each
5-10
-------
14
100 -i
I
6 8 10
DISTANCE FROM BRIDESBURG CENTER , km
Figure 5.4.1
Average Concentrations and Concentration Ranges in
Bridesburg Area Versus Distance Between Sites for
TSP, IP and FP
5-11
-------
classification, but there are major differences between the core,
perimeter and airport groupings.
TSP and IP are 45%, and 31% higher at core sites when compared to
perimeter sites. Absolute concentration differences are 30 ug/nr*
for TSP, and 16 ug/nr* for IP. These differences occur over an
average distance between sites of 1 km. It is apparent that TSP and
IP concentrations are not homogeneously distributed on this
neighborhood scale. From the study area to the NE Airport, a distance
of 10 km, an additional average concentration decrease of 25 ug/nH
for TSP and 17 ug/nr* for IP was seen.
A similar decrease was not observed for FP between the industrial
and perimeter sites. In fact, with one exception, the FP
concentrations in the Bridesburg study area all exhibited a rather
flat profile. There was a substantial decrease in FP levels between
the industrial area and the NE Airport (10) site, indicating that an
average of 8 to 10 ug/tn of fine particles was associated with
sources in or near the industrial study area.
This poses some important problems for the siting of monitors to
represent a neighborhood. These samplers were spaced within 1 km of
each other, a spacing which is much closer than that which would be
found in a routine compliance monitoring network. Any one of them
could have been chosen as the compliance monitoring site for the
neighborhood. Yet, given appropriate values for a standard, this area
could be in attainment or non-attainment depending on whether the
monitor selected was in the perimeter or in the industrial core area.
This case is an extreme one; most neighborhoods are not so highly
industrialized and would probably not show the extreme IP
concentration gradients found in Bridesburg. Yet it is just this type
of neighborhood which is most likely to be in non-attainment.
Since both perimeter and industrial core sites show IP and TSP
concentrations substantially higher than concentrations at the
background site, the incremental amounts must be due to sources of
particulate matter originating in the neighborhood. Speculations on
what those sources might be were advanced in Chapter 4. In Chapter 9
a receptor model approach will be applied to further identify likely
contributors to the observed concentrations.
5-12
-------
The problem of siting a single sampler to represent an entire
area still remains, however, and development of a method to perform
such siting for compliance monitoring program will be just as
important as the formulation of a new standard when it comes to
determining attainment status.
5.5 Seasonal Distributions
The IP Network offers the opportunity to evaluate seasonal
patterns in various urban areas and regions of the country. All
samples are of 24-hr duration, which precludes examination of the
hourly variability, and the sixth day sampling schedule (third day at
some sites) does not provide a strong data base for investigating
daily distributions or weekday/weekend patterns. The seasonal
averages presented in Table 5.1.1 do allow variability between seasons
to be assessed.
Seasonal patterns of average and maximum IP and FP concentrations
are examined by reading across a row of Table 5.1.1 or 5.1.2 for a
specific sampling site. Trijonis et al (1980), studied the St. Louis
regional monitoring data and concluded that IP and FP concentrations
peak in the summertime. This hypotheses can now be evaluated in other
parts of the country.
Figure 5.5.la shows the seasonal variation in average quarterly
concentrations for IP and FP in 11 eastern and midwestern cities.
This geographical subset of sites was selected because of the regional
patterns shown above for FP concentrations. There is a slight
increase in FP in the summer, which causes IP to be higher, although
the increase is only 7 ug/m^ over the fall-winter-spring averages.
The ranges appear to be generally similar among the quarters, except
that FP seems less variable in both spring and fall. Figure 5.5.1b
shows the seasonal variation of quarterly maxima. FP is seen to be
lower in the spring. The CP fraction can be observed in these figures
as the difference in FP and IP levels. It is interesting that in the
spring the FP/CP ratio is lower than in the other seasons, suggesting
a different mix of sources of IP in the spring.
5-13
-------
n
I
z'
o
K
<
CZ
K
Z
LU
U
Z
o
u
140 —
120 —
100 —
80-
60 —
40-
20 —
160-
140'
120 —
100 —
80-
60-
40-
20 —
A. AVERAGE
LEGEND
TSP
IP
FP
t
I
&**
B. MAXIMA
223 T 226 T191
1
T221
FALL WINTER SPRING SUMMER
1979 1980
Figure 5.5.1a-b Seasonal Variation of Quarterly TSP, IP
and FP Averages(a) and Maxima(b) in 11
Eastern and Midwestern U.S. Urban Areas
5-14
-------
The data suggest that FP might be slightly lower in the spring
season and higher in the summer, causing IP to be slightly higher in
the summer. However, there is not sufficient evidence to suggest that
there is a general seasonal pattern in IP and FP similar to that shown
in St. Louis.
It is evident that though there may be specific areas in which
seasonal differences are unimportant, enough variability exists
between seasons to require sampling on an annual basis, with
approximately equal numbers of samples taken in each season.
The temporal and spatial coverage of the data available for this
report severly limits the certainty with which conclusions can be
drawn about the attainment of standards, causes of elevated and
depressed concentrations, sampler spacing and sampling schedules. The
following observations about these issues based on the preceding
discussion must be considered tentative; their applicability will be
greater when a similar analysis is applied to the more complete IP
Network measurements in the future.
Annual arithmetic averages of IP and FP can exceed
90 ug/m-5 and 35 ug/m , respectively, in urban areas,
though the typical average concentrations appear to be
approximately 50 ug/m3 and 25 ug/m3, respectively.
Primary standards for annual IP averages in the 55 to
120 ug/m3 range (Hileman, 1981) would find most, and
possibly all, of the sites examined here in compliance.
This range is tentative and may have been modified
subsequent to the writing of this report.
Annual 24-hr maximum concentrations of IP and FP can exceed
200 ug/m3 and 100 ug/m3, respectively, in urban areas.
Typical values are approximately 100 ug/m3 for IP and
60 ug/m3 for FP. Hileman (1981) cites a 24-hr maximum IP
range of 150 to 350 ug/m3 for a primary standard and a
24-hr maximum FP range of 70 to 220 ug/m3 for a secondary
standard. Once again, most, if not all IP Network sites
would be in compliance with such a standard. The values
cited by Hileman (1981) are tentative and the form and
values of a 24-hr standard may have been modified subsequent
to the writing of this report.
The urban and neighborhood scale IP and FP measurements vary
significantly from site to site within the areas studied.
The implications are that 1) local (within a few kilometers
5-15
-------
of the sampler) sources are significant contributors to IP
concentrations and 2) present spacing between sampling sites
may be inadequate to fully represent IP concentrations in
certain urban areas and neighborhoods. Methods of
determining the sampling uncertainty due to sampler
placement should be developed and applied.
Seasonal averages of IP and FP tend to peak slightly in the
summer periods, but there are many individual exceptions at
IP Network sites. This conclusion is speculative because of
the inadequate seasonal data available for this report. If
the conclusion is valid then sampling should take place on a
yearly basis for the determination of long-term averages so
that all seasons are equally weighted. A seasonally
weighted annual average might be a more appropriate method
of calculating the annual average if the number of samples
in one or more seasons differs by a large amount from the
number of samples taken in another season.
5-16
-------
CHAPTER 6
GEOGRAPHICAL AND SEASONAL VARIABILITY OF TSP, IP AND FP
IN NON-URBAN AREAS
While the geographical and seasonal distributions of suspended
particulate matter at the urban sites showed the potential of certain
urban areas to exceed a size-classified standard, an examination of
the geographical and seasonal variability of non-urban concentrations
assists in estimating that portion of an urban average or maximum
concentration which is not subject to control of emissions within the
urban area.
These non-urban measurements are particularly useful when they
are from the same air masses that affect nearby urban areas but which
are not significantly influenced by the emissions from those areas.
The roles of synoptic weather conditions covering broad geographic
scales and of long-range transport from one urban or industrial area
to another are better defined when the variability of ambient
concentrations due to local emissions is removed.
No site within the United States, and possibly even the world,
will be completely free of anthropogenic contributions to its ambient
aerosol. Yet, from the practical standpoint of a localized urban
area, it is sufficient to separate that which is transported into the
area from that which originates in the urban area, regardless of the
origins of the transported material. Regional-scale studies must then
be undertaken to separate the natural from anthropogenic contributions
at non-urban sites. In this chapter, regional-scale measurements from
the IP Network and from the SURE Network are tabulated, examined, and
compared data from nearby urban sites.
6.1 TSP, IP and FP Non-Urban Concentrations
Several sites in the IP Network are removed enough from populated
areas that they can be considered non-urban, even though in some cases
they may be influenced by nearby urban areas. Eight of these sites
had acquired a large enough number of measurements by the time of this
writing to determine seasonal averages and maximum concentrations for
TSP, IP and FP. As in Chapter 5, only quarterly periods with more
6-1
-------
than 5 measurements were examined and the values were taken from IP
Network data summaries produced by EPA/EMSL that did not go through
the data filtering of inequalities 3-1 to 3-4. IP and FP correspond
to the TOTAL and FINE measurements of the dichotomous sampler. The
averages at these sites appear in Table 6.1.1 and the maximum
concentrations are presented in Table 6.1.2. The annual averages in
Table 6.1.1 were calculated as the average of the four seasonal
averages to remove any bias caused by an unequal number of samples per
season. As for the urban concentrations in Tables 5.1.1 and 5.1.2,
the non-urban sites have been grouped by regions of the United
States. At the bottom of each seasonal column, averages of the
averages from each site have been recorded as indicators of the
overall non-urban concentration average. A similar set of numbers
appears in Tables 5.1.1 and 5.1.2 for urban measurements.
Figure 3.1.1 shows where these non-urban sites are located in relation
to the urban areas sampled by the IP Network.
Measurements in the eastern United States can be supplemented by
TSP, IP and FP measurements from the SURE (Perhac, 1978, Mueller
et al, 1980, Miller, 1981, Perhac et al, 1981) which were taken during
6 seasonally representative month-long periods from August, 1977 to
October, 1978. Figure 6.1.1 shows the locations of the 9 SURE Class I
sampling sites which are identified in Table 6.1.3. Detailed
descriptions of these sites, the measurements made and the
regional-scale interpretations drawn from them are to be presented in
Mueller and Hidy et al (1981). The comparison of the SURE and IP
Network sampling inlets in Chapter 2 showed that the IP fraction of
the total aerosol mass collected by the SURE sampler is within 10% to
20% of the IP fraction collected by the IP Network samplers. The FP
fractions collected by both samplers were shown to be identical.
SURE aerosol samples were taken on Teflon coated glass fiber
filters which were specially chosen to be free of sulfate artifacts,
and though a nitric acid artifact was found, it was much less than
that of the standard glass fiber filters. Arithmetic averages of TSP
at each site for each of the seasonal months were calculated from all
the valid, 24-hr samples available for that month; the arithmetic
averages reported here differ from the geometric averages calculated
6-2
-------
s
m
cj
M
H
W
H
M
ss
en
<
w
CO
e>
z
M
CLl
fc.
O
CO
w
tH
^
oi
I
B
z
rJ
OS
w
c
O 01 ^M
to ooco
to m a
01 )H -~-
CO 01 00
> 3
•a- < -~-
o
00
e^
CO
. 6
rl -»
0) Of
a 3
CO
o
oo
c*
CO
- 8
00 ""^
C 00
•H 3
Q.
CO
o
00
i—4
rJ
CO
4-) 3-
ON
r-t
•^v
1
-v.
fV
OJ
vD
1-t
.—1
CO
-*
CN
K
o
A
T3
C
CO
i— i
CO
M
0)
•H
>
3
to
Cfl
en
rH
^*.
ON
m
00
CO
CO
rH
^.
-a-
•^i
i— i
r-
m
i— i
-a-
r*v
\£>
O
1-1
O
rH
i— I
m
•^
t£>
r*»
CN
rH
CO
>£>
O
O
rH
X
H
A
•U
C
*H
i— (
O
I— i
rH
*-*.
sj-
ro
•^^
sy-
in
<— i
»H
•v^
o->
OJ
-*^.
f*^
v£)
*
ro
--*
ON
sf
\D
rH
"^,
ro oi
m oj
rH OJ
O vo
ro 04
-•>. -s.
vD O
ro ro
i-J
M
•a
o
o o
3 'z.
T)
CO CU
^ H
cc &
ro
OJ
m
•
rH
04
**
•^
0\
Xi
CO
a>
w
vD
rH
*-x.
rH
OJ
00
ro
1
•^.
1
in
m
CO
rH
^
OI
•*
00
ro
3
CD
W
Vj
0)
4-1
to
•H
W
bC
•H
en
ON
rH
ON
01
vD
sr
ro
Ol
•^v.
sD
ro
O
vO
r*.
rH
ON
0]
r-v
in
oo
rH
-^,
r-
OJ
---,
OJ
ro
oo
rH
*""*.
O
ro
m
ro
CO
0)
4J
U *H
CO CO
OJ p
OO t-<
CO (U
V- 4J
CJ CO
> CO
< w
m
rH
OJ
CO
*-^
o
m
i —
rH
--v.
ro
CO
•-*.
CO
vD
in
rH
-^>
ro
i —
in
m
rH
"^
O
ro
r-
ro
m
rH
-*
ro
01
rH
< CO
OH
fc
0-1
L<
O
fa-
ro
6-3
-------
o oi -H
« aon
a a e
-» < >--l
IA <*•>-*
'z
a
o
o
3
8
01
c
«H
3
§
_
•u
c
Q
(0
H-f
01
«rt
>
3
n
»
X
H
4J
C
•H
»H
U
to
at
4J
M-t CO
O
C
g£
•H U
X ID
a ai u
r 3 to
a
u
r4
M X
H
•o •
0 *
QUO
3 Z 0
•a b
•H -A
a DU «
l< H 01
03 ai co
01
M
S
0)
CO
V
4J
u
•H
U
00
•H
pa
U-i
0
§
•H
X
£
0)
01
4J
to
e
i*
01
4J
(0
a
u
U-l «
O 01
4J
§1-1
CO
•H
X rH
rt (H
S <
Cu
tu
^-.
Pu
M
^-
a.
12
6-4
-------
c/l
-------
CO
4-1
C
ao a
c a
•H V
4-1 14
CO 3
•H tO
x a
ta v
Q
3 C
o
B
&4 CO
CO i-H
H.
. 01
CNi-4
O
CO PI
L4
HI
3
O
H
0>
E
o
2
01
B
O
z
cu
i
w
CO
e
m
o >,
14 o
cu
Z
»in
w m o
m • >-,
•o
c
o in
3 •
-
ui
a
CO
B
o
z
JO B
10 O
14 4J
a B
—i to
CJ CJ
3
B .o
O B
o>
H CO
i-l 01
•H B
> B
CO 01
S CJ
n B
> -H
u >
O ID
O CJ
0>
Oil
•H 3
CJ
c •
3 4J
X b,
o
X .a
60 CO
•H B
01 01
1-4 01
a H
c
•c o
i-4 4J
01 01 «
•H Ji OJ
-UH 3:
Mrt
Vi 3 -
3 CO CO
£ DO
" C
rt E
O3
O EC
u
CO
CO
3
O
01 -H 4J i-l
-H & T4 14
cy i a
|J 5 CO
in
CN
01
B I
O Z
"ill
01
3
00
B
E
to
14
u
CO
•a
E
B
CO
O
E
3
Q
o
CJ
-E 01
CJ iH
14 00
eg B
01 to
CO TH
01 L4
00
o
00
I.
3
XI
CO
01
£
4-)
O
B
O
h
U-l
c
0
•H
4J
U
01
k4
«H
•o
01
JZ
u
01
t-
10
CO
E
o
•H
4J
y
01
h
•H
O
CO
14
01
CO
a
1
U
•H
X
a
to
b
00
o
a
o
4J
O
o
o
o
m
rg
r^
U)
O
w
S
XI
N-X O
c
• 0)
«-H 3
« iH
MH
U C
0) -H
>. 0)
T3 O
•H LJ
X S
o
•o tn
c
id 1-1
(Q
^ o
0) o
i-H i— 1
1— (
O QJ
3 i-H
S X»
•H
O CO
4J CO
o
M a.
OJ
M-( -a
it e
Ctf (Q
U
6-6
-------
by Mueller and Hidy et al (1981). Eight IP and FP samples of 3-hr
duration were taken each day. SURE data summaries report monthly
averages for each 3-hr period; the average of these averages was used
to represent the monthly average. This would be equivalent to the
monthly average of all 3-hr samples if no samples were missing; since
few months experienced 100% data captures, there may be slight
differences between the averages reported here and those calculated by
other methods. Table 6.1.4 contains the results of these
calculations. The fifteen-month averages for each site are the
averages of the averages from the six seasonally representative months.
The maximum 24-hr TSP, IP and FP concentrations during each month
of sampling at each site were selected from the SURE data summaries
and are tabulated in Table 6.1.5. For the 3-hr duration IP and FP
samples, only days with 6 or more valid samples were considered. The
averages of the 3-hr samples for the two or three days on which the
highest 3-hr concentrations occurred were calculated and the highest
of these averages was taken to be the maximum for that month.
With the data in Tables 6.1.1, 6.1.2, 6.1.4 and 6.1.5, it is
possible to observe the regional and seasonal distributions of TSP, IP
and FP concentrations for non-urban areas, analogous to the
examination of urban concentrations presented in Chapter 5, and to
compare urban to non-urban concentrations.
6.2 Regional Geographical Distributions
The ranges of TSP, IP and FP average concentrations in the
western and eastern parts of the United States are presented in
Table 6.2.1. The SURE ranges (except for the low end of TSP) are in
agreement with those found by the IP Network, lending credence to the
regional representativeness of the sites in both networks since the
measurements were derived from different sites at different times.
The first feature of this regional comparison is that the ranges
in the West for TSP and IP are larger than the ranges in the East and
that western FP average concentrations are lower than eastern averages
in that size range. Part of this is due to the selection of sites.
Hawaii has little industry and little pollution transported to it,
aside from marine aerosol, from outside. It must represent one of the
6-7
-------
en
u
1-
u
2
DM
CO
-a- H
a
x>
^rr s^ f** CO ON O fn ON o O ^
U 0*<^ i-t iH" rH CM CM i-f (si (Sj *H
c oo e -^-^^-^.-^^^.^•^
O flJ *••- moNONi-ir-.ONooooeM
•CJaOO CMCMCMCnmCMeNCMCM
—• •^»>^'^.i^.*^i^.^^-^>x
f^.r^oo r^so^rooo
inmmin-3-enmcn
O
CO
1 CMCMCMCMCMCMCMCM
« •xfc*^'««^-^'xfc'^.'«%.*xl
\ocnincM sos&oo«?O
^ *^ cn «^ CO O vO 00 ONinrH^iH
l-i 00 iHCMCMCMcnCMCnCMCM
cos •^•^•^.•^•*^-^.<»^*^.'^.
3s^- moNrHincnoN^HvooN
C rHCMCn-^-lTlCMinCMCM
r*.
r^
ON
^ m so r»». ON so
-^ -•-. *^. -^ **^ *^.
S oo>-'-a>si'-3'.tf-*m
3
CO ON
CM CM
o
-a-
o
•a
in ON
CM CM
X
o
2
H
U oi
K tJ
S -H
cn cn
| «
e re
re TM
lJ -O
o c
DM
H
>
00
V.
3
01
00
2
01
ai -H s
oo « 5
a E
U pH *r^
01 i-l X
>«, J
01
00
a
u
as 01
IH CM cn <•
o o o o
o
d.
0)
01
00
«
kl
01
>
<
6-8
-------
CO
w
H
<
X
<
.J
u
w
PS
3
w
f -o
.u B
C 3<
o 6 a
00 t—< en r*- 43 t—i »-H O CO
>>o co '•o -4- co m WP m m
-•x. --^ --v -x. *-^ --^ *-^ *"* *-*
CN CM O voOm*j^r^
CT\ O ON r-« CN ^ 00 ^O "^
•"-*. -^\ -^•*-,-*^*^--^.-1^.
CN r-^rg enr^-^-rH^^o
OO ^Ha^ vDONinOCT\«£>
CM i-H CM «-*
^H m oo CM *J e»j vo o fJ^
-3-enrM fMenencJincM
*^. "^ •->. •^-^.-^•^.•^.-^
iH iH i—(
oornen r^ VD •-( r-t -* m
^£j co ^o •j>mm'»£>-3'm
-^.•^"^ ^^-^•^v-^.-^.*^,
CM t-^O vOCMvC-^OOf11-
a\o^ r^rvinoom-^-
rH
•*^*^*^- *->. --S. *^ -X. -N. -^
rMm>* rx-a-o»Hm^
00 rHOO OOfHOOO^m
CM i—t i—(
coo-tn r^a>-^^^.
«^ooo r-1 CM ^o o> i-* m
en in w*i *^ vo in en in en
\*-^\ ^-x."^*^,-^,-^,'^
oo >H m r^o-0
CM
o
a
B
a
z
H
• TH TO
C
3
Q
g
O
0
X rH
« «
[K
*^
o<
oo
C
P2 0)
O <-J
tn -H
co ON
00
6-9
-------
Table 6.2.1
RANGES OF NON-URBAN ARITHMETIC AVERAGES
IN THE WESTERN AND EASTERN U.S.
INCLUSIVE SAMPLING PERIODS OF TABLES 6.1.1 and 6.1.4
West (IP Network)
East (IP Network)
East (SURE Class I)
TSP Range
ug/m3
33 to 88
38 to 55
22 to 55
IP Range
ug/m3
16 to 59
21 to 34
22 to 37
FP Range
ug/m3
6 to 15
15 to 23
16 to 23
Table 6.2.2
RANGES OF NON-URBAN MAXIMUM 24-HOUR CONCENTRATIONS IN THE
WESTERN AND EASTERN U.S.
INCLUSIVE SAMPLING PERIODS OF TABLES 6.1.2 and 6.1.5
West (IP Network)
East (IP Network)
East (SURE Class I)
TSP Range
ug/m3
57 to 240
85 to 161
66 to 254
IP Range
ug/m3
56 to 171
51 to 78
47 to 120
FP Range
ug/m3
11 to 55
31 to 57
47 to 81
6-10
-------
most pristine sites in the IP Network. Clint, Texas is subject to
duststorms on occasion which will elevate the larger particle
concentrations. The remaining western sites in Winemucca and Sauvie
Island exhibit TSP and IP within the range of the eastern sites.
The generally lower average concentrations of FP in the West
compared to the East are associated with lower densities of industrial
emissions in the West and possibly with differences in synoptic-scale
meteorology. A number of studies (compare Pierson, 1980; Heisler
et al 1980, for example) have shown eastern aerosol to contain a
greater proportion of sulfate than western aerosol while the western
aerosol, as in the cases of Winemucca and Clint which are in arid,
desert-like areas, tends to be enriched in geological material over
the eastern aerosol. The sulfate is primarily in the FP fraction
while the majority of the geological material is concentrated in the
coarse fraction of IP and TSP.
The ranges of maximum concentrations are presented in
Table 6.2.2. It is more difficult to compare maximum values than it
is to compare averages because of the random nature of individual
measurements taken at different times. The maximum TSP and IP
concentrations in the West correspond to the Clint site and are
consistent with the duststorm activity already mentioned while the
Hawaii site shows the lowest values for TSP, IP and FP.
The differences from East to West are not so marked for the
maxima as they are for the averages. None of the maximum TSP values
at these non-urban sites exceeds the primary standard of 260 ug/np.
The maximum concentrations vary substantially from site to site
for all particle size ranges. Given the extreme differences between
the air masses affecting the sites in the West, this is
understandable. The air masses affecting sites in the eastern United
States are similar, however, and the range of maximum concentrations
exhibited by the eastern sampling sites might be, in part, due to some
local emissions or meteorological events which affected the particular
site on the date the sample was taken.
It is better to examine the averages to get an idea of regional
variability in the East. The picture of IP and FP average
concentrations is fairly well represented by the average of all the
6-11
-------
eastern annual averages for both IP Network and SURE sites. The
3
nominal IP eastern regional average is 29 ug/m for the IP Network
data and 28 ug/m for the SURE data while the FP regional average is
3
19 ug/m for both data sets. This consistency between both
averages, taken at different times and different sites supports the
generalization that average non-urban eastern concentrations are
3 3
30 ug/m and 20 ug/m for IP and FP, respectively.
6.3 Seasonal Distributions
There seems to be no general seasonal trend in average or maximum
concentrations at western sites. A few anomalies show themselves in
the seasonal data; at Winemucca and Sauvie Island the Fall, 1979,
maximum TSP values are less than the corresponding IP maxima and the
TSP average is less than the FP average, but this is probably due to
non-simultaneous sampling for TSP and IP.
In the eastern United States there is a tendency toward lower
concentrations in all size ranges in the fall and winter months than
in the spring and summer months. SURE Montague and Scranton sites are
the most illustrative of this phenomenon, but it is evident in the
averages of most eastern sites. IP and FP from the eastern IP Network
were 36 and 23 ug/m respectively during Summer, 1980 but were 27
3
and 18 ug/m during Winter, 1980. In the SURE Network, all site
3
averages of IP and FP were 33 and 22 ug/m in August, 1977 and 38
3
and 25 ug/m during July, 1978. However, these values were only 26
and 20 ug/m , respectively, during January, 1978. Snowcover in many
areas can suppress the coarse fraction of IP during winter while the
high humidity and temperatures of summer will increase the likelihood
of S02 conversion to sulfate in the eastern United States. These
hypotheses need to be tested for IP and TSP concentrations by
examining the chemical composition of selected samples.
Maximum concentrations in the East do not seem to follow a
general seasonal pattern. Though the maxima for all particle sizes at
Montague (SURE site) occurred during the summer months, the maximum
6-12
-------
concentrations for Rockport (SURE site) occurred in January, 1978.
Once again, maximum concentrations seem to be randomly occurring with
many causative factors.
6.4 Non-Urban Contributions to Urban Concentrations
The geographical distributions of non-urban TSP, IP and FP
concentrations presented in this chapter can be compared with the
urban concentrations of Chapter 5 to come up with a rough estimate of
that fraction of the urban concentration which does not originate
within the urban area. Ranges of annual averages for non-urban and
urban sites in the East and West are grouped in Table 6.4.1.
In all cases the upper end of the urban range of averages is
higher than the corresponding upper end of the non-urban range. In
the West, the highest urban averages are approximately double the
non-urban averages for TSP, IP and FP. In the East, they are nearly
double for TSP and IP, but only half again as large for FP. A fair
amount of the FP in eastern cities seems to be of a regional-scale
rather than urban-scale nature.
The averages of concentration averages at urban and non-urban
sites over the entire United States gives a rough cut of the nominal
fraction of the urban aerosol which could be the result of transport
into the urban area. This fraction must be considered an upper limit
since even well-located regional-scale sampling sites will be affected
by localized sources to some degree. Up to 50 to 60% of the average
TSP, 70% of the average IP, and 60 to 80% of the average FP found in
urban areas could be accounted for by transported material.
A comparison of urban and non-urban concentrations is more
accurate when measurements at a non-urban sampling site close to an
urban sampling site are compared. Such a comparison of average
concentrations is made in Table 6.4.2 where an IP Network urban site
has been paired with an IP or SURE Network non-urban site which should
sample the same types of air masses as the urban sites, but without
the local sources.
The ratios of non-urban/urban average concentrations have been
tabulated to gain an idea of the possible distributions of the
6-13
-------
w
J
5
NCENTRATIONS
RAGE
E-t
W
ES
RAG
CO
u
O
5
OJ
CO
w
CO
55
pa co
&
r
Ji 4-1
M OJ
0K
4-1 PM
01 l-l
w co
01
CO A
M
C &
n) i
-0 C
^ o
4-l|
col
cal
w|
m ON ON
CM i— 1 i— 1
CO CO CO
CO CM CM
O O 0
4J 4J 4-1
o
00 S r-l
^i 4-1 U
I-i 0)
0 S5 W
S w
4-1 P-l C3
Ol M CO
S5 1^'^--
PL, CO CO
M 01 01
s_^ 4-1 4J
• H -i-l
05 CO CO
01
i-i C C
•H ca CB
CO XI Xl
i-i M
C P !=>
ca i i
Xi C C
MOO
D S3 2
CO
CM
r^-
co
O
4J
CO
r— 1
vO
rH
O
m
oo
oo
0
4-1
CO
CO
^"\
X
l-l
o
3
4-1
01
E5
PL,
i-l
\^r
U)
Ol
4J
•r-l
co
C
ca
Xi
M
D
1
C
o
S5
i-l
i— 1
<
id
O
cj
CO
01
U
C
w
I
4J
3
O
CO
3
O
CO
U
Si
T3
U)
I
(0
tS
0)
T3
3
i— I
U
6-14
-------
TABLE 6.4.2
RATIOS OF NON-URBAN TO NEARBY URBAN
CONCENTRATIONS OF TSP, IP AND FP
Northeast
MA
Montague, MA/E. Boston
/E. Boston
NY
Big Sister, NY/P.S. No. 26
/P.S. No. 28
/Wilmuth
Midwest
OH
Duncan Falls, OH/Akron
/Cincinatti
/Steubenville
South
AL
Giles County, TN/S. Birmingham
/N. Birmingham
/Inglenook
/Mtn. Brook
/Tarrant City
Southwest
TX
Clint, TX/E1 Paso
West
OR
Sauvie Island, OR/Portland
All Non-Urban/Urban Sites
Ratio of Arithmetic Mean
Non-Urban/Urban
TSP
.35
.46
.35
.33
.25
.59
.76
.42
.73
.52
.36
1.08
.31
.79
.49
.56
IP
.56
.68
.33
.40
.64
.57
.74
.49
.62
.55
.48
.83
.33
.76
.43
.65
FP
.77
.89
.52
.64
.72
.66
.77
.53
.83
.86
.73
1.12
.44
.42
.52
.71
6-15
-------
portions of the urban concentrations which could be attributable to
non-urban aerosol. These ratios are crude estimates, since the urban
and non-urban sites are not always close to each other (Figures 6.1.1
and 3.1.1 will locate the relative positions of sampling sites) and
the measurements do not encompass the same periods of time. The
averages of IP Network data were calculated from the data acquired
over the periods listed in Table A.I which went through the filtering
of inequalities 3-1 to 3-4. SURE averages were taken from Table 6.1.4.
The range of ratios for TSP is the largest, as is to be expected
since the largest particles, (greater than 15 urn diameter), are rarely
transported far and are likely to be of local origin. Up to 31% to
100% of the urban TSP concentrations could come from outside the urban
area, though the mode seems to be in the 50 to 60% range.
For IP, 33 to 83% of the urban average concentrations can be
found at nearby non-urban sites, though the majority of values are in
the 55 to 75% range.
For FP, the non-urban/urban ratio of averages ranges from 42 to
100% with a large number in the 65 to 80% range.
The extreme cases can be explained in several ways, mostly by
looking at the site surveys of Table 3.1.3. Mountain Brook, AL, for
instance, is practically a non-urban site, being located on the
outskirts of town, so it is expected to receive little local
influence. Clint, Texas experiences the same duststorm activity as El
Paso, but not the winter burning for heating that occurs in El Paso
and Juarez, so the TSP and IP non-urban/urban ratios are high whereas
the FP ratio is low.
These specific site comparisons, even in their crude form,
support the generalization made from the grand mean ratios of
non-urban/urban average concentrations: upper limits of 50% of the
TSP, 60% of the IP and 70% of the FP are not subject to the control of
emissions within the specific urban area being sampled, particularly
in the eastern United States.
If an urban area is found in non-attainment of a new standard,
practical control strategies may require regional in addition to local
planning. The regional-scale monitoring data, aside from the now
6-16
-------
non-existent SURE Network and the few sites available in the IP
Network, is presently inadequate to provide a technical basis for such
planning.
Several generalizations can be made from the non-urban
measurements and their comparison with urban measurements:
Non-urban average TSP, IP and FP concentrations are
nominally 50, 30 and 10 ug/m , respectively, in the
western United States and 50, 30 and 20 ug/m3,
respectively, in the eastern United States. The number of
non-urban sites with sufficient data in the West is small,
however, and these observations should carry less weight
than those from the East where independent measurements
corroborate each other.
There is a general seasonal variability of average TSP, IP
and FP concentrations in the eastern United States in which
the spring and summer month concentrations are elevated with
respect to the fall and winter months. Independent
measurements from different networks support this
conclusion. Though the measurements in the western United
States do not show a significant deviations between seasons,
the number of sites examined is too samll to allow a general
conclusion.
Up to 50% of the TSP, 60% of the IP, and 70% of the FP in
urban areas can typically be accounted for by concentrations
present at nearby non-urban sites. This portion varies
substantially from site to site and is based on a limited
number of sites, primarily in the eastern United States.
Much of the non-urban concentrations may have originated
from far away anthropogenic sources.
6-17
-------
CHAPTER 7
STATISTICS OF THE IP NETWORK CONCENTRATIONS
The primary air quality standards for suspended particulate
matter are intended to protect public health with an adequate margin
of safety. They are designed to prevent problems caused by long-term
and short-term exposures to suspended particulate matter. Thus, the
present particulate matter standard consists of two components; an
annual component (75 ug/nH annual geometric mean) against which to
o
judge long-term exposure and a 24-hr component (260 ug/m not to be
exceeded more than once per year) with respect to short-term exposure.
An area's attainment of these standards is determined by ambient
sampling at representative sites according to a prescribed procedure
and schedule. With regard to the present standard, 24-hr HIVOL
sampling is required every sixth day at a location sufficiently
removed from significant local sources of particulate matter (within
approximately 200 feet).
Thus, two things determine whether or not an area is in
attainment or non-attainment of the standard: the form of the standard
with the values attached to it, and the measurement process used to
obtain ambient concentrations to compare with those values. This
section deals with the frequency of occurrence of different
particulate matter concentrations and the statistics, particularly the
mean and maximum value, derived from a finite set of samples.
First, the log-normal probability distribution is examined as a
reasonable model for representing the population of particulate matter
concentrations and for summarizing the large number of measurements
taken in the IP Network. Next, the distributions of each size
fraction at individual sites and over the whole network are examined.
Finally, geometric and arithmetic means are compared as possible
statistics for representing the true distributions and their
variability as a function of sample size is studied. Because of the
limited quantity of data available for examination, the cases studied
are to be considered illustrative rather than conclusive. All data
used in this chapter went through the screening procedure of
inequalities 3-1 to 3-4.
7-1
-------
7.1 The Log-Normal Frequency Distribution
Two issues must be addressed in the setting of an ambient air
quality standard:
1. The determination of the actual long-term and short-term
concentration levels.
2. The statistical form by which the exceedance of those levels
is determined.
The second issue is important from a standard-setting (as opposed to a
standard monitoring) point of view because any set of measurements
taken to assess a sampling site's attainment of the standard is merely
a statistical sample of all possible measurements. Even if 365 24-hr
particulate matter samples per year are taken, this stiJ.JL represents a
single random sample of size 365 from an infinite population of
possible levels (Neustadter and Sidik, 1974). In most cases, the
sample size will be less than 365 cases/year and estima te s of the true
mean and maximum concentrations must be made from this sample. These
estimates are statistics which assume the cases are drawn from the
same population. The precision with which these statistics are
estimated decreases as the number of cases used to calculate them
decreases. The estimate of the mean is not affected by the choice of
an assumed statistical distribution of the population. The selection
of a distribution can, however, change the estimate of a maximum value
for a given number of days of sampling.
The log-normal distribution has long been used to represent the
statistical population from which pollutant concentrations are drawn
(Larsen, 1974, 1975, 1976). It offers some practical advantages over
the more commonly known normal distribution in that:
No negative values are allowed, which is consistent with
ambient measurements.
Histograms of the frequency of occurrence of various
concentrations are skewed toward the lower concentrations.
High concentrations receive lower weighting in the
estimation of a geometric mean of a log-normal distribution
than they do in the estimation of the arithmetic mean of a
normal distribution.
7-2
-------
No physical justification for a log-normal distribution of
ambient pollutant concentrations has yet been found. Benarie (1971)
first suggested that they are log-normally distributed because wind
speeds are log-normally distributed. Making concentration predictions
from a variety of dispersion models with log-normally distributed wind
speeds as input, Bencala and Seinfeld (1976) concluded that "... if
wind speeds are nearly log-normally distributed then resulting
concentrations will be nearly log-normally distributed." They are
careful to add the caveat that "this result does not establish that
wind speeds are the primary influence on concentration distributions
in the atmosphere, since other effects are most certainly influential."
Kahn (1973) showed that a log-normal distribution follows from
the assumption that the change in pollutant concentration between
successive measurements is proportional to the first measurement. He,
too, admits that "the complexity of the emission-transport-receptor
process makes it difficult, if not impossible, to prove that air
pollutants are log-normal."
Studies of total suspended particulate matter (TSP) concentration
distributions at a number of sites yield mixed results. Hovey
et al (1976) and Phinney and Newman (1972) applied Kolmogorov-Smirnov
and Chi-square goodness of fit tests to TSP data in New York and
Indiana, respectively, and did not reject the hypothesis of log-normal
distributions. Kalpasanov and Kurchatova (1976), however, did reject
the log-normal hypothesis for TSP concentrations in Sofia, Bulgaria.
The most comprehensive studies of TSP concentration distributions
were carried out by deNevers et al (1977, 1979). After examining
4,112 annual cumulative frequency distributions from HIVOL sampling
sites in California, Georgia, Illinois, Missouri, New Hampshire, New
York and Pennsylvania, they (deNevers et al, 1979) observed four
patterns which occur in the mid-range of the data. Patterns 1, 2 and
3 are illustrated in Figures 7.1.1, a, b and c. Pattern 4 is the
inverse of Pattern 3 with a horizontal section in the middle.
deNevers et al (1979) drew random samples of 25 and 75 values
from a large set of random numbers generated from a log-normal
7-3
-------
a) Pattern 1
_ 100
"E
b) Pattern 2
50 «0 95 99 999
PERCENTILE
50 80 11 '(9 99 »
PtRCENTILE
E 1
»
a.
c) Pattern 3
20 5C 80 95 99 999
PERCENTILE
Figure 7.1.1 Patterns l(a), 2(bJ, and 3(c) of TSP Frequency Distributions
Found oy deNevers et al (1979).
7-4
-------
distribution and found excellent examples of Patterns 1, 2 and 3
(Pattern 4 did not frequently occur in the ambient or simulated data).
In agreement with other researchers, deNevers et al (1979) warn
that each of the patterns in Figure 7.1.1 "....could have arisen by
simple random sampling from truly log-normal populations of data. It
is not convincing evidence that it did."
In search of this convincing evidence deNevers et al (1979)
applied the Kolmogorov-Smirnov goodness of fit test to 3,992 of the
annual distributions (patterns with extremely high or low end values
were deleted from the original set as providing unrepresentative
statistics and were eliminated from consideration). These tests
yielded the result that most patterns appeared with frequencies
similar to those to be expected if samples were drawn from random,
log-normally distributed populations. The exceptions consisted of an
inordinate number of Pattern 1. deNevers et al (1979) postulate that
these cases might arise from samples being drawn from two independent
log-normal distributions, each representing a distinctive
meteorological regime.
Other distributions for suspended particulate matter
concentrations have been proposed. Bencala and Seinfeld (1976) tested
the three-parameter log-normal, the Weibull, and the Gamma
distributions. Mage and Ott (1978) advocate a censored
three-parameter log-normal model as a possible alternative to the
traditional two-parameter log-normal distribution. deNevers et al
(1979) propose the use of more than one log-normal distribution to
represent one year of measurements (Mage, 1979).
It seems that the assumption of a log-normal distribution for
suspended particulate matter concentrations is not completely
supported by a physical rationale nor always by statistical
measurements of validity. It is "....defensible on the grounds that
it works" (Lodge and West, 1969), and that is considered sufficient
justification for choosing it over others in order to present and
evaluate size-classified suspended particulate matter concentrations
collected in the IP Network.
The statistics of the extreme values of a distribution are not
necessarily the same as the central part of the distributions, and
nearly all log-normal plots of suspended particulate matter
7-5
-------
concentrations show deviations at the upper and lower concentration
extremes (deNevers et al, 1977). Roberts (I979a) shows that when
dealing with a "not to be exceeded..." type of standard applied to
individual measurements, "the probability of exceedances.... is
independent of the distribution of the random variable..." Roberts
(I979b) shows several examples of this applied to gas concentration
measurements.
The frequency distributions of TOTAL and FINE particulate matter
measurements need to be studied to determine whether or not log-normal
distributions and their associated statistics adequately describe the
actual measurements for practical purposes even though fits to such
distributions may not always be statistically significant.
7.2 Cumulative Frequency Distributions in the IP Network
Cumulative frequency distributions for all validated data
specified in Table A.I of Appendix A for all sites in the IP Network
appear in Figure 7.2.1 a-e and are not necessarily representative of
distributions at a single site. Each plot contains the calculated
geometric average concentration and geometric standard deviation.. In
all cases these calculated values at the average, the average minus
one standard deviation, and the average plus one standard deviation
match fairly closely with the values read from each of the frequency
distributions at the 50th, 16th and 84th percentiles, respectively.
The central portion of each of these curves tends to approximate a
log-normal distributions fairly accurately. It is obvious, however,
that at the extremes there are substantial deviations from this
distribution, particularly in the case of the TOTAL, FINE and COARSE
suspended particulate matter concentrations. Approximately 3% of the
FINE samples have concentrations less than 4 ug/m^. Nearly 10% of
the COARSE measurements are less than 5 ug/m , while 2% are less
than 1 ug/m3. These extreme values propagate to the TOTAL
distribution.
The central and upper extremes of the FINE and COARSE
distributions are similar to each other, which is reasonable when one
compares individual FINE/COARSE pairs and the average of FINE/COARSE
ratios of 1.21. At the lower extreme, however, there exists a
7-6
-------
-a
nj
o
4-1 W
o oo
C O
O U
c
•H
\O
CTl
O
•H
Q
U PL,
c
0 -N
3 r-,
c* o
0 ^—'
fH H-3
PL, <
H
£g
•H
CO
u co
I
cd
CM
W)
•H
u,
o
<4-l
0
fH
03
to
4->
c
nJ
0
S
0
0
+J
10
CL, Ctf
7-7
-------
disproportionate share of low COARSE concentrations, notwithstanding
the elimination of many of these records according to the validation
criteria of Section 3.4. As deNevers et al (1977) observed for HIVOL
measurements, "...some of these low values, particularly the extremely
low ones, are an artifact of the data gathering and recording system
and most likely are not true representations of the atmospheric
conditions." As noted earlier, some of the COARSE material may not
adhere to the filter thereby causing many of those samples to exhibit
lower than expected concentrations.
If the SSI and the TOTAL are sampling the same aerosol, then
their cumulative frequency distributions should be similar.
Comparison of Figures 7.2.1b and 7.2.1c show a marked difference
between the distributions, both in terms of the geometric averages and
the maximum values. The SSI distribution shows a lower proportion of
values less than 10 ug/m3 than does the TOTAL distribution. The
proportion of high concentrations is greater for SSI than it is for
TOTAL; 7% of the SSI measurement exceed 100 ug/m3 whereas only 3% of
the TOTAL values exceed 100 ug/m3. Only 285 of the 771 SSI and
1,509 TOTAL were taken at the same site at the same time, so
Figures 7.2.1b and 7.2.1c may not represent the same population. In
fact, the proportion of SSI samples obtained in industrialized areas
was considerably higher than the proportion of dichotomous samples
taken in such areas. However, the preponderance of low concentrations
shown by the FINE and COARSE distributions casts some doubt on the
representativeness of some of these measurements.
As stated in Section 7.1, there is no reason to expect the
distribution of ambient concentrations at a particular site to follow
a log-normal curve, let alone measurements from various sites
scattered all over the country. The value of the plots in
Figures 7.2.1a-e is in conveniently summarizing all the data in the
network in a concise manner. It is remarkable that the HIVOL and SSI
concentrations appear to be so log-normally distributed, though given
the large number of degrees of freedom, a statistical test might
reject the hypothesis that they were drawn from a log-normal
distribution.
The 99th percentile concentration as read from each graph gives a
good indication of what the highest 1% .of the concentrations in the
7-8
-------
network are, approximately 210 ug/m3 for HIVOL, 170 ug/ra3 for SSI,
150 ug/m3 for TOTAL, 80 ug/m3 for FINE and 90 ug/m3 for COARSE.
The IP Network data presented in Figures 7.2.1a-e are highly dominated
by measurements taken in the Philadelphia area (39% of the HIVOL, 59%
of the SSI, and 29% of the TOTAL, FINE and COARSE values in the plots
are from Philadelphia sites). Cumulative frequency distributions for
all sites outside of Philadelphia were generated which yielded very
nearly the same results as those in Figures 7.2.1a-e. SSI, TOTAL and
FINE plots for data grouped by different site type classifications
were similar, though they were often shifted upward or downward
because the average concentrations for different site types were
different. When the number of points dropped to less than 30, the
linear part of the log-normal distribution disappeared in these plots.
For all sites where SSI or TOTAL measurements were 20 or greater,
the SSI, TOTAL and FINE cumulative frequency distributions were
plotted in an effort to see whether or not deNevers et al (1979) TSP
patterns occurred for IP and FP measurements. Most sample sets
contained fewer than 30 values, and the plots, in general, did not
appear to follow any pattern. The best examples of deNevers et al
(1979) from these plots are presented in Figures 7.2.2a-d. These may
be compared with Figures 7.1.1a-c. It seems that the same types of
patterns observed for TSP concentrations may be present in IP
concentrations, but given the small number of measurements available
from most IP Network sites at this time, it is too soon to tell. When
more data become available, it would be useful to determine whether or
not deNevers et al (1979) patterns exist and to test their hypothesis
that they result from two different log-normal distributions related
to different meteorological regimes. If TSP, IP and FP distributions
from the same site and simultaneous samples exhibit the same patterns,
it would reinforce the hypothesis that some environmental (as opposed
to measurement) factor is causing the variability of the
concentrations.
7-9
-------
») PAPHA SSI, Los Norul
b) PAPHB TOTAL Pattern 1
Geometric Mean .41.1 Vg/m
Geometric Std. *• 1.6
•No. of Pts. • 111
— . —
/"
y*
»
it 1 1 1 1 1 1 1 1 t { 1 1 1 1 1 1 . 1 1 1 1
200
iaa
E
Concentration, y
8
I?1
Geometric Mean > 33.1 Ve/m
Geometric Std. > 1.6
No. of Pts. * 91
~ —
»-***
X*
X
— i — ' 1 — i — *- ' • ' * ' • ' 1 1 1 1 1 1 i i i i ,
.zaasoB .Bn .100.100 ,aca.:
Percentile
.zaasaa .can .7aa.ua .aaa.
Percent!le
c) PAPHB SSI Pattern 2
iaa
IB
i [ i [ i l ' 1 ' I ' I ' I ' 1 ' I
Geometric Mean » 64.2
Geometric Std. •> 1.5
No. of Pts. • 85
i
d) PAPHG TOTAL Pattern 3
iaa
2 sa
.200.383 .EH .7Bo.ua .aso.sfa
Percentile
.sazue .u
ia
— f-T - , — i ' M i " j
Geometric Mean « 39.6 Hg/IH
Geonetric Std. « 1.6
No. of Pts. - 53
' i ' i ' i '"
t i
i t i
, i . i ,
i ,
.OEa.ioa .zao.5ao .czar .7oataa .aaa.aca
Perccntile
Figure 7.2.2 a-d Examples of Cumulative Frequency Distributions from
Individual IP Network Sites Representing a Log-Normal
(a), deNevers et al (1979) Pattern 1 (b), Pattern 2 (c)
and Pattern 3 (d)
7-10
-------
7.3 Arithmetic and Geometric Averages
Though a geometric average and standard deviation can be
calculated from a set of ambient concentrations at a site and these
parameters can describe a log-normal distribution which is close to,
though not exactly equal to, the central part of the ambient
measurement distribution, there are good reasons to formulate a
long-term standard in terms of an arithmetic rather than a geometric
mean (Mage, 1980):
• The arithmetic mean has physical significance for human
health whereas the geometric mean does not. The integral of
an ambient concentration times a breathing rate over time is
proportional to an arithmetic rather than a geometric mean.
• The use of the arithmetic mean for the particulate matter
standard will allow comparison with other average air
quality data (such as 802) and provide more meaningful
confidence intervals (an arithmetic mean has a symmetric
confidence interval while that of a geometric mean is
asymmetric).
In practice, the difference between the geometric average and the
arithmetic average of a set of ambient suspended particulate matter
measurements may be small enough to allow one to be estimated from the
other. Mage (1980) calculated the arithmetic averages of the data
used to support the present National Ambient Air Quality Standard for
TSP and compared them with the geometric averages of the same data
sets. The average ratio of the geometric to arithmetic average was
.87 with a standard deviation of .02 for 21 pairs of averages. The
ratios ranged from .82 to .92. Frank (1980) calculated and compared
the geometric and arithmetic averages of TSP for 11,451 site-years
included in the National Aerometric Data Base (NADB). He stratified
these averages into four geometric average concentration ranges to
identify variations with concentration. In the range which included
3
the level of the TSP standard, the 61 to 90 ug/m range, he found
that out of 3,756 site-years, over 50% of the geometric/arithmetic
average pairs differed by less than 11% and over 90% of the pairs
differed by less than 20% from each other (it should be noted,
however, that 1% of the pairs differed from each other by more than
7-11
-------
44%). If the difference between averages for IP and FP is just as
constant, then, depending on the precision required, little is lost
in transferring from a geometric mean, which might be more descriptive
of the distribution, to an arithmetic mean, which may be more closely
related to health effects (Mage, 1980).
Arithmetic and geometric averages were calculated for all IP
Network sites with available HIVOL, SSI, TOTAL and FINE measurements
using data validated by inequalities 3-1 to 3-4 and the ratios of
geometric to arithmetic averages were produced. The averages, ranges
and standard deviations of these ratios for each sample type are
presented in Table 7.3.1. The average for each size fraction is in
the vicinity of .9 with standard deviations of approximately .05.
This ratio varies with the scatter in the data. Sample sets having
larger standard deviations (geometric or arithmetic) have lower and
more variable ratios. Table 7.3.2 presents the distribution of ratios
as a function of geometric standard deviation. As the geometric
standard deviation increases, the ratio decreases. The range for a
given standard deviation depends more on the number of samples than on
the standard deviation. Most geometric standard deviations are in the
neighborhood of 1.5 to 1.7 for all size fractions. In this
neighborhood, the ratio varies from .85 to .95 for all samples.
As long as both averages can be calculated from a set of data, it
makes no difference which average is chosen. The calculated average
is only an estimate of the true mean, however, and its precision
depends on the sample size. The calculation formulas for the
geometric and arithmetic averages are derived from maximum likelihood
theory (Meyer, 1978) to provide the most probable estimate of the mean
based on a log-normal or normal distribution, respectively. Different
sample sizes will give different estimates of the true mean, and if
the assumed distribution (log-normal or normal) is closer to the real
distribution of particulate matter concentrations, the calculated
average (geometric or arithmetic) from a given number of samples will
be closer to the true mean.
This is studied from a practical standpoint in Table 7.3.3 where
the ratios of geometric to arithmetic averages are calculated for data
sets which assume 3, 6, 12, 24 and 48 days between samples. The data
7-12
-------
TABLE 7.3.1
AVERAGE RATIOS OF GEOMETRIC TO ARITHMETIC
AVERAGES AT IP SAMPLING SITES
Sample
Type
Number
of
Averages
Average
Ratio
Geometric
Average/
Arithmetic Average
Standard
Deviation
of
Ratio
Range
of
Ratio
HIVOL
SSI
TOTAL
FINE
82
37a
70
70
.91
.91
.90
-88
.05
.07
.05
.06
.70 to .98
.63 to .98
.71 to 1.00
.69 to .98
Site OHMEA eliminated as outlier.
7-13
-------
TABLE 7.3.2
AVERAGE RATIOS AS A FUNCTION OF GEOMETRIC STANDARD
DEVIATION FOR IP SAMPLING SITES
Geometric
Standard
Deviation
Sample ug/m^
HIVOL 1 .
1.
1.
1.
1.
1.
1.
1.
2.
2.
SSI 1.
1.
1.
1.
1.
1.
1.
2.
3.
2
3
4
5
6
7
8
9
1
6
2
3
4
5
6
7
8
0
6
Range of Ratio of
Geometric Average to
Arithmetic Average
.98
.96
.94
.91
.88
.87
.87
.81
.78
.70
.98
.97
.94
.91
.89
.87
.85
.79
.63
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
.99
.98
.95
.95
.92
.94
.87
.85
.78
.70
.99
.98
.97
.95
.94
.90
.86
.81
.63
No. of
Ratios
4
7
9
15
19
19
2
3
1
1
3
3
7
6
9
4
2
2
1
Range of Number
of Data Pairs in
Each Ratio
3
3
14
5
6
2
35
6
27
12
2
2
7
2
2
7
9
10
6
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
9
16
38
48
107
99
41
26
27
12
5
12
25
85
111
35
33
14
6
7-14
-------
TABLE 7.3.2 (Continued)
Geometric
Standard
Deviation
Sample ug/m3
TOTAL 1
1
1
1
1
1
1
1
1
2
2
2
.1
.2
.3
.4
.5
.6
.7
.8
.9
.0
.2
.7
FINE 1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
2.1
2.3
3.0
Range of Ratio of
Geometric Average to No. of
Arithmetic Average Ratios
1.0 to 1.0
.98
.96
.94
.91
.88
.86
.85
.82
.81
.77
.71
.97
.94
.90
.88
.85
.83
.82
.77
.78
.69
.69
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
.98
.97
.98
.94
.92
.89
.88
.88
.85
.78
.71
.98
.96
.94
.94
.90
.86
.87
.85
.85
.72
.69
1
1
3
6
19
19
4
7
4
3
2
1
2
4
7
20
15
4
8
4
3
2
1
Range of Number
of Data Pairs in
Each Ratio
2
3
9
2
4
7
13
3
5
10
9
9
2
3
9
2
13
28
9
5
3
26
19
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
2
3
22
35
49
91
25
33
29
35
19
9
9
35
43
51
91
49
25
29
14
35
19
7-15
-------
TABLE 7.3.3
VARIABILITY OF THE RATIOS OF GEOMETRIC AND ARITHMETIC
AVERAGES WITH SAMPLING FREQUENCY
Sampling Site
500 S. Broad (PAPHA)
HIVOL
SSI
TOTAL
FINE
Allegheny (PAPHB)
HIVOL
SSI
TOTAL
FINE
NE Airport (PAPHE)
HIVOL
SSI
TOTAL
FINE
Presbyterian Home (PAPHG)
HIVOL
SSI
TOTAL
FINE
St. John's (PAPHI)
HIVOL
SSI
TOTAL
FINE
Sampling Intervals (days)
1 A ii 24 48
Ratio of Geometric to Arithmetic Average
.92
.90
.93
.92
.87
.91
.92
.90
.91
.94
.91
.89
HG)
.91
.95
.90
.88
.91
.89
.88
.88
.94
.91
.94
.92
.88
.92
.95
.96
.90
1.00
.90
.87
.94
1.00
.94
.93
.90
.90
.88
.86
.95
.92
.99
.96
.89
.93
.96
.97
.92
-
.93
.91
.95
-
.93
.92
.88
.88
.85
.84
.95
.91
.99
.98
.91
.94
.97
.97
.96
-
.95
.92
.94
-
.96
.95
.92
.92
.99
1.00
.94
.88
.99
.99
.98
.97
.96
.96
.95
-
.95
.88
.96
-
.94
.91
.91
.89
-
_
Range of
Sample Size
113
111
43
43
99
85
20
20
82
2
91
91
98
2
53
53
L07
96
18
18
to 7
to 8
to 4
to 4
to 7
to 5
to 2
to 2
to 7
to 1
to 6
to 6
to 7
to 1
to 4
to 4
to 8
to 6
to 3
to 3
7-16
-------
set is essentially halved from one interval to the next and the number
of samples used to calculate the averages for the 3-day and 48-day
intervals is listed in the last column.
The ratios of the geometric average to the arithmetic average
rarely vary by more than .03, even when the initial sample size is
reduced by a factor of 16. When significant changes dp occur, the
total number of samples is normally less than ten.
These case-studies indicate that both the geometric and
arithmetic averages are of equal value in representing the
distribution of ambient concentrations for all particle size ranges
and reasonable numbers of measurements. If an arithmetic mean is
deemed more representative of exposure levels, then a standard which
is specified in terms of such a mean seems reasonable. These studies
are not as extensive as those of Frank (1980), however, who did find
large differences between geometric and arithmetic averages for data
from a small number of site-years. More comparisons of these averages
of SSI, TOTAL and FINE measurements should be made.
7.4 Variability of Means and Maxima with Sample Size
Every statistic is an estimate calculated from a finite sample
drawn from a larger (often infinite) population. The frequency of
sampling and the total number of samples available will affect the
estimates of the mean and maximum values for particulate matter
concentrations in all particle size ranges. Akland (1972) describes a
practical method of assessing the adequacy of a sample size. He used
6 years of TSP measurements which resulted from daily measurements in
Buffalo, NY. He calculated the annual arithmetic average TSP
concentrations for each of these years using all samples, the samples
on every third day, and the samples on every 13th day. The maximum
deviation from the population mean was +.9% for the third day subset
and -8.0% for the 13th day subset. The standard deviations of the
subsets and the population were nearly the same.
Five sites in the IP Network collected enough data (approximately
every third day for one year) to make such a comparison possible for
the size-classified concentrations. While five sites are not
7-17
-------
sufficient to form sweeping generalizations, they are sufficient to
illustrate the types of effects different sampling frequencies can
have on the statistics calculated from their measurements. The data
records for the 500 S. Broad, Allegheny, NE Airport, Presbyterian Home
and St. John's sampling sites in Philadelphia were successively halved
by removing every other data record. This is equivalent to, though
not exactly the same as, sampling every 3rd, 6th, 12th, 24th and 48th
day. It is not exactly the same because sometimes samples were taken
more often than every third day, and in some cases no sample was taken
on a particular day. This procedure is legitimate because it
simulates what would happen under normal sampling conditions. In a
routine sampling program, some extra samples are taken and others are
lost. Table 7.4.1 shows that this procedure generally halved the
number of samples from one sampling interval to the next except when
the number of records became very small, i.e. less than five. The
arithmetic averages of each of these subsets were calculated for each
size-classified sample and the results are plotted in Figures 7.4.1a-d.
The important feature to note in these figures is the variability
of the averages as the interval between samples increases and as the
number of samples used to calculate the average decreases. The cases
considered in Figures 7.4.1a-d do not show general upward or downward
trends in average concentrations with sample size; these trends cannot
be expected because of the random way in which samples used to
calculate each average were selected. The average calculated from the
rejected data set after each split would provide an equally valid
estimate of the true mean. If an average shown in one of the plots of
Figures 7.4.1a-d for a given sampling interval is a certain amount
less than the previous average resulting from more frequent sampling,
the average calculated from the rejected data set is greater than the
previous average by the same amount. For clarity, only the averages
from the data sets corresponding to the pre-ordained sampling schedule
are shown.
Some of the arithmetic averages vary substantially as a function
of sample size. SSI at Presbyterian Home and NE Airport can be
dismissed as having too small a sample size. St. John's FINE and
TOTAL averages show extreme variations by the time the original
7-18
-------
TABLE 7.4.1
NUMBER OF MEASUREMENTS IN SUBSETS OF ALL
MEASUREMENTS AT PHILADELPHIA SITES
Sampling Site
& Size Fraction
500 S. Broad (PAPHA)
HIVOL
SSI
TOTAL
FINE
Allegheny (PAPHB)
HIVOL
SSI
TOTAL
FINE
NE Airport (PAPHE)
HIVOL
SSI
TOTAL
FINE
Presbyterian Home (PAPHG)
HIVOL
SSI
TOTAL
FINE
St. John's (PAPHI)
HIVOL
SSI
TOTAL
FINE
Days Between Samples
12
No. of Measurements
113
111
43
43
99
85
20
20
82
2
91
91
:G)
98
2
53
53
107
96
18
18
50
57
21
21
No. of
49
44
10
10
No. of
41
1
48
48
No. of
48
1
25
25
No. of
54
50
10
10
29
28
11
11
15
15
5
5
7
8
4
4
Measurements
27
23
5
5
14
11
3
3
7
5
2
2
Measurements
21
-
24
24
12
-
12
12
7
-
6
6
Measurements
23
-
14
14
13
-
8
8
7
-
4
4
Measurements
28
24
5
5
14
12
3
3
8
6
-
—
7-19
-------
CO
e
M
SL
03
O
110-
loo-
80-
60-
a) IIIVOL
0 3 . 6 12 2i* *»8
Sampling Interval (days)
m
e
-^
ec
A
V
80-H
60-
20
b) SSI
0 3 6 12 21 *»8
Sampling Interval (day)
C
"ea
rt
-------
18 values captured with the 3-day interval (values weren't available
every third day, but the situation is analogous to what would happen
if many randomly occurring samples were invalidated, a very real field
problem) are reduced to 5 values with the 12-day intervals; TOTAL
averages at Allegheny begin with 20 samples for the 3-day interval,
end with 2 samples for the 48-day interval and show the same
deviations. FINE averages at Allegheny, however, do not vary by more
than 5% until the number of samples is reduced to three for the 24-day
sampling interval.
In general, when the number of samples used to calculate the
average falls below 15, there is often, though not always, an abrupt
change in the average concentration estimate. In most cases, this
corresponds to the 24 or 48-day sampling interval.
Except for the case of Allegheny, the variability of the averages
with sample size (or sampling interval) is essentially equivalent for
all site types (as Table A.I of Appendix A shows, these five sites
represent a number of different site classifications) and for all
particle sizes sampled in the IP Network until the number of samples
in the average is approximately 15 or less. The averages of the less
frequent sampling schedules (sampling interval of 12-day instead of
3-day) are within + 10% of the averages calculated from the 3-day
schedule for most cases. The same conclusion would also be drawn if
all the subsets were individually examined.
The case of Allegheny is interesting in that both HIVOL and SSI
averages contain a large number of samples (113 and 99, respectively,
for the 3-day intervals) but the HIVOL averages show considerable
variability as the sampling interval is increased whereas the SSI
averages remain nearly constant. Allegheny (see Table 3.1.3) is a
site which is influenced by a number of nearby non-ducted area
sources; many of the individual HIVOL measurements are high and vary
from day to day. The SSI concentrations are not as variable. It
appears that for this type of site, an average concentration estimate
for a size-classified sample requires fewer individual measurements
than does an average for TSP. In certain situations, the temporal
concentration homogeneity introduced by the removal of highly
7-21
-------
variable, local source contributions of particles with diameters
greater than 15 urn might offer the advantage of reducing the number of
samples required to estimate the true mean.
The maximum concentrations from each one of these data sets are
shown for the five sites and particle sizes in Figures 7.4.2a-d.
Estimates of the true maximum concentration are severly dependent on
the sampling interval for all particle sizes. An important point in
all the plots is that the maximum concentration always goes down as
the sampling interval increases. The decrease in the maximum
concentration is not a function of sampling interval. Sometimes a
maximum value will persist through all cuts of the data set, as for
the HIVOL maximum at Presbyterian Home. At St. John's the maximum
value persists for three cuts for all sizes. The maximum drops sharply
and remains the same for the remaining cuts for 500 S. Broad SSI and
Allegheny FINE and COARSE. As the plots show, the decrease in maximum
concentration from one set to the next can be quite large for all site
classifications and particle sizes. The greatest decrease is for
Allegheny HIVOL which falls by over a factor of two between the 6-day
and 12-day sampling intervals. Even for the change between the 3-day
and 6-day periods, the maximum decreases by over 35% for all particle
size fractions at 500 S. Broad and for FINE and TOTAL at Allegheny and
Presbyterian Home. In contrast to the average concentrations, which
generally do not change substantially when sampling intervals are
lengthened, the maximum concentration is very sensitive to the number
and frequency of samples.
If a new standard is couched in the same terms as the present
standard, an annual average and a 24-hr maximum (or 2nd maximum), then
sampling intervals of up to 12 days seem adequate to assess the
average, but not the maximum. Under such a standard, the best
strategy of a region desiring compliance with a 24-hr standard would
be to perform the minimum amount of sampling allowed. As
Figure 7.4.2a shows, the samplers at 500 S. Broad, Allegheny and NE
Airport may be penalizing themselves by sampling every third day
instead of every sixth day as required by the reference method for TSP
compliance monitoring. The same holds true for IP (both TOTAL and SSI
measurements) and FP concentrations. This means that the monitoring
7-22
-------
200-
3. 150-
6
S
• H
X
cfl
12 100-
50
43P
a) HIVOL
0 3 6 12 2% »»8
Sampling Interval (days)
200-
e
e
•H
X
efl
ISO-
KJO-
25
b) SSI
0---— 0
0 3 6 12 2"t
-Ao- o
0 3 6 12 2<» «»8
Sampling Interval (days)
Figure 7.4.2 a-d Maximum Concentrations for Different Sampling Intervals
at 500 S. Broad(•), Allegheny(x), NE Airport(A),
Presbyterian Home(o), and St. John's(V) for HIVOL(a),
SSI(b), FINE(c), andTOTAL(d).
7-23
-------
agency which gains a more accurate estimate of the true mean by
sampling more often than required runs a greater risk of noncompliance
with the 24-hr standard than it would if it sampled less often.
Some other form of a 24-hr standard than the present one is
required to eliminate or reduce the penalty of more frequent
monitoring. Frank (1981) suggests that each maximum concentration
measured should additionally count as having occurred on as many days
as separate the samples, and that the standard should be based on the
expected number of exceedances above a specified concentration level.
The EPA (1981b) Office of Air Quality Planning and Standards Staff
paper recommends that either "...the allowable number of exceedances
would be expressed as an average or expected number per year...,"1 or
that "...a given percent of the daily values would be expected to be
less than or equal to the standard level." Roberts (1979a, 1979b)
suggests some other approaches as do Curran and Frank (1975). Several
of these forms need to be formulated by qualified statisticians and
tested by changing the sampling intervals of large data sets, similar
to what has been done in this section, to evaluate their practicality.
It should be emphasized once again that the paucity of data
available at the time of this report limits the generality of the
observations made in this chapter. The illustrative cases indicate:
One percent of the IP Network measurements exceeded
210 ug/m3 for HIVOL, 170 ug/m3 for SSI, 150 ug/m3 for
TOTAL, 80 ug/m3 for FINE and 90 ug/m3 for COARSE.
The log-normal cumulative frequency distributions at
individual sites show examples of the TSP patterns advanced
by deNevers et al (1979) as well as approximations to
log-normal distribution. The small number of measurements
available at most sites prevents an assessment of the
adequacy of the log-normal and other statistical
distributions in representing IP and FP data.
The geometric averages of IP and FP data sets normally
differ by 5 to 15% from the corresponding arithmetic
averages, though the difference is as large as 37% for IP
7-24
-------
Network data. The ratios of geometric to arithmetic
averages calculated from subsets of all possible
measurements containing decreasing numbers of measurements
at a site are constant to within +_5% until the number of
measurements in the subset falls below five.
Arithmetic averages of IP and FP tend to remain constant,
within +10%, as the interval between samples is doubled and
the number of samples is decreased. For 15 or fewer samples
this consistency breaks down.
Maximum concentrations of IP and FP are extremely sensitive
to sample size and can be reduced by 50% as the interval
between samples is increased. This means that any standard,
such as the present one, which specifies a maximum
concentration should also specify a sampling frequency or
that other forms of short-term standards should be
considered.
7-25
-------
CHAPTER 8
PREDICTING IP CONCENTRATIONS FROM
TSP CONCENTRATIONS
Given numerical values associated with a 24-hr maximum and yearly
mean for a size-classified standard, communities must make some
assessment of the extent to which they attain those values.
Ultimately, this task can only be accomplished via actual ambient
monitoring using samplers and procedures which meet the particle size
collection specifications of the standard. The deployment of a new
sampling network will not take place immediately, however, and in the
interim, estimates of size-classified aerosol concentrations must be
made for planning and network design purposes.
Many HIVOL TSP measurements exist at a wide variety of sites
throughout the United States. In Chapter 2 it was demonstrated that,
for a given size distribution and wind speed, relationships between IP
and TSP measurements can be established; these derived relationships
agreed remarkably well with the averages of many measurements. It is
tempting to use these relationships to estimate IP and FP
concentrations from TSP concentrations. A substantial uncertainty is
associated with this estimate, however, and it is important to
evaluate the magnitude of that uncertainty.
In this chapter, the validity of a receptor-oriented model for
estimating IP from TSP concentrations is formulated and evaluated.
This process involves:
• Deriving average slope and linear regression relationships
between IP, FP and TSP measurements using IP Network
measurements.
• Calculating the parameters of these relationship for data
sets stratified by site type and HIVOL concentration range
to evaluate whether or not such stratification improves the
predictive ability of the model.
• Evaluating the accuracy of a model derived at one site over
a period of time to predict IP concentrations at another
site for the same time period.
8-1
-------
• Evaluating the accuracy of a model derived at a site over a
given period of time to predict IP concentrations at the
same site over a different time period.
Once again, the caveat concerning the illustrative rather than
conclusive nature of this treatment needs to be emphasized; IP Network
data from which these models are derived are still limited. The
methodology followed here may be useful for specific applications, but
the actual models used in a given situation should be tested with more
exhaustive sets of data to verify their predictive abilities. All
measurements used in this chapter were screened by inequalities 3-1 to
3-4.
8.1 A Receptor Model Approach
Two approaches to estimating IP and FP concentrations are
available, one source-oriented and the other receptor-oriented.
In the source-oriented approach an inventory of particulate
matter emission rates in the appropriate size range is formed using
emission factors (Shannon et al, 1980) and size-classified source
sampling (Harris and Smith, 1980). With appropriate meteorological
input, the transport of these emissions between source and receptor is
simulated mathematically (Cramer and Bowers, 1980) to provide
estimates of expected contributions.
This source-oriented method has the following limitations:
• Emission factors are inaccurate and variable.
• Many sources of particulate matter emissions are
unidentified.
• Source models cannot adequately represent the deposition of
particulate matter from the atmosphere.
• Source models cannot adequately represent the chemical and
physical transformations of particles and gases.
The receptor model approach to estimating size-classified
suspended particulate matter concentrations involves simultaneous
size-classified and HIVOL sampling at representative receptors, the
8-2
-------
establishment of a relationship between size-classified concentrations
and HIVOL concentrations (this is the receptor model), and an
evaluation of the precision of this model in predicting
size-classified concentrations at receptors similar to those from
which the model was derived. Once formulated and validated, this
model could be used to estimate size-classified concentrations with a
definable uncertainty at receptors with HIVOL measurements.
The advantages of this receptor model approach are that it
overcomes the limitations of the source-model method: no emission
rate estimates, source identification, deposition rates, or chemical
reaction mechanisms are required. The disadvantage is that the
receptor model formulated for one receptor may not be applicable to
another receptor. The relevant receptor characteristics which affect
the transferability of the model must be identified in its formulation.
Any receptor model must be derivable from a source model to be
physically significant. The starting point for the receptor model
relating size-classified particulate matter concentrations to TSP
concentrations is the typical urban mass distribution as a function of
particle size (Willeke and Whitby, 1975) which is reproduced in
Figure 8.1.1. If an aerosol sampler has an upper particle size
cutpoint of D , the area under the distribution between 0 and D
c c
will be proportional to the ambient concentration measured by that
sampler.
If two samplers with different cut-sizes, D and D , as
Cl C2
illustrated in Figure 8.1.1, sample from the same size distribution,
i, simultaneously, then the ratio of the concentrations, C,. and
^2^i they measure will be proportional to the ratio of the areas
indicated in Figure 8.1.1.
R - °2i
Ri ~
ll ll
This assumes all other sampling biases (e.g. artifact formation) are
negligible. If the particle size distributions for all samples,
i=l ..... n, retain the same shape, R. will be the same for all i. In
8-3
-------
Csl
O
o
e
3.
CD
N
00
CD
i—H
O
rt
DH
6
C flj
O VH
•H M1*-!
+J nj O
3 -H
tfl
• H
O
CD
N
• H
co
CD
r—t
O
fn
0,
Onr-l
<4H
O
to o
g^
.§
(N
to
J-l 'H
CD
r-i T5 CD
PH C r-H
6 rt
in
H
V] (/)(/)
CD !/)
T3 W O
C O PH
«g,
:^§
^ -H
CD CD 4->
PH-'H £>
+-> -H
rt ^
+•> t/i
u
rt
JH
U-
T3
CD
• H N
O ^ tfl
CD O
^ (4-1 X
—i C
O
-------
reality, the shapes of the size distributions will vary from sample to
sample, and so will R-. Sampler-specific biases, some of which were
identified and estimated in Chapter 2, will also cause R. to vary.
The receptor model must include easily identifiable conditions which
will minimize the variability of the size distribution and biases
between samples and thereby reduce the variability of R..
Since the R- are variable, the best estimate of one R for a set
of samples must be a statistic which is a function of the R^. Two
simple statistics are available, the slope, R,, of a linear
regression between GI and €2 and an average of the individual
ratios, RA-
The linear regression slope formulation modifies Equation 8-1 by
adding a constant, b. Thus, an estimated concentration in size range
2, C?., as a function of the concentration in size range 1, C .,
will be:
C = R C, . + b
2i L li 8-2
The values of RL and b and the correlation coefficient, r, are
(Bevington, 1969):
C C - C C
RL = — i £ 8-3
7 - C2
Cl Cl
where
n
C2 ' Cl ' n . CliC2i
— 1 n 2
9 = — V C
C2 n . Ci 8-6
8-5
-------
C2 - R^ 8-7
C2 '
Cl
- Cl •
C2
r = — 8-8
(c2 - c2)1/2 (c2 - c2)172
V^-i 1' • \ V->^ 9'
Since IP and FP concentrations must equal zero when TSP goes to
zero, b should be equal to zero. Due to measurement uncertainties, b
can differ from zero, but it should be within a few times the lower
quantifiable limit of zero.
The correlation coefficient (r) is indicative of the strength of
the relationship between C, and C_. For the purposes of this
application, a correlation coefficient in excess of .9 means that the
linear approximation is a good one; if r is less than .9, then the
correspondence between the two variables is considered to be low and
the independent variable is not deemed to be a good predictor of the
dependent variable — the receptor model is therefore an inaccurate
representation of reality.
The other statistic is the arithmetic average of the ratios
- I R. 8-9
n .=1 x
The standard deviation of the set of n R. values,
2 8-10
is the best measure of the precision associated with estimates made
from the relationship:
8-6
-------
A
The lower 0R, the more precise will be the estimate C^--
Trijonis et al (1980) applied the linear regression model of
Equation 8-2 and the average ratio model of Equation 8-11 to HIVOL and
TOTAL data acquired from simultaneous sampling with HIVOL and
dichotomous samplers in St. Louis, Missouri during the 1976 Regional
Air Monitoring Study (RAMS). The dichotomous samplers used in this
program were of an earlier make (Loo and Jaklevic, 1974) and the upper
size cutoff has been estimated at a larger size than that of the
dichotomous samplers used in the IP Network. At the ten sites he
considered, Trijonis found RT varying between .20 and .74 for the
TOTAL/HIVOL relationship and between .11 and .26 for the FINE/HIVOL
relationship. The correlation coefficients ranged from .42 to .88 and
.29 to .59 for TOTAL/HIVOL and FINE/HIVOL, respectively. Trijonis
conclusion was that "...TSP measurements alone cannot adequately
explain the day-to-day fluctuations in TOTAL... and FINE."
Trijonis et al (1980) also calculated average ratios, R,,» and
standard deviations, aR, for TOTAL/HIVOL at each of the RAMS
sites. These average ratios at each site contained from 53 to 84 data
pairs and ranged from .49 to .71 with standard deviations between .17
and .31. The overall ratio for all sites was .61 with a standard
deviation of .22. Trijonis conclusion was that in the case of St.
Louis, "...(HIVOL) is a poor predictor (of TOTAL) on a daily basis."
Trijonis et al stratified their sample with respect to position, time
and meteorology and found that for TOTAL/HIVOL the estimates of the
average ratio were not substantially refined by this stratification.
These conclusions about the IP prediction receptor models applied
to the RAMS data base are discouraging. They do not show a definitive
relationship between IP or FP and TSP measurements. The data set used
was limited, however, and it may be premature to dismiss the utility
of the receptor models before exploring their applicability to a more
complete set of measurements.
3-7
-------
8.2 Site Type and Concentration Stratification
The IP Network provides the best set of simultaneous
size-classified and TSP measurements for formulating and evaluating
such a model. The sampling sites encompass a wide variety of land
uses, which were identified in Chapter 3. The measurements are
numerous enough that a wide range of concentrations for TSP, IP and FP
concentrations are available. Each of these easily identifiable
characteristics of samples could have an effect on the homogeneity of
the particle size distribution within a sample set.
To form the receptor model, FINE, TOTAL, SSI, and HIVOL mass
concentrations were used in Equations 8-3 and 8-7 to calculate the
slope, RL, and intercept, b, of the linear regression formulation
and in Equations 8-9 and 8-10 to calculate the average ratio, R,.,
and the standard deviation, aR, of the data set used to achieve it.
This was done for various groupings of data which were expected
to have sampled size distributions which were more similar to each
other than to the distributions sampled by the data base as a whole.
The following groupings were chosen:
• all data,
• site type classification, and
3
• HIVOL concentrations above and below 100 ug/m .
Both the linear regression slope and the average ratio models
were applied to each group of measurements. These models were used to
relate FINE, TOTAL and SSI concentrations to HIVOL measurements. The
results of this sample stratification appear in Tables 8.2.1, 8.2.2,
and 8.2.3 with stratifications by site type and concentration.
In deciding between which of two equally valid theories applies
to a complex situation, the simpler one is that which should be
chosen. Given the choice between the linear regression slope model
and the average ratio model with both showing equal predictive
ability, the average ratio model would be chosen because it is simpler
to calculate and does not contain the constant, b, of the linear
regression slope model.
8-8
-------
CM CM O
-* o m
00 -* rH
CM VO -*
ON -
- o i
eo CM eo O
-31 CM O
o en o
NO •«• oo
rH rH 00 CM in
o> -* eo m r-*
u a
0 P-
m o so
SO in 00 rH rH rH
rH *d" ON O\ O eo
o o o o o o
cn
z
o
rH O.
H O
< tt.
H W
Z Z
u o
CJ M
Z H
8 |
M Z
cn u
00 U
• [b Q
CM O O
00 Z -J
o o
SM >
H rH
a 3 :
< o
u z
rJ S
i y
s iS
RECEPTOR
AS A
u
a.
£
B 1
M M
cn H
m H
0 S
u u
M Z
rV O
S °
2 o
cn M
i
u z
cn <
H
Q
C
•H
U
A
•H
4J
CO
•H
a
u
i
cr
0)
o
(d
O
•H
U
•H
a
OH
>resslon
Linear Rej
B
O
CO
CO
00
tH
CO
01
B
•H
rJ
O
S
01
00
t-t
01
u
b°
+ 1
rJ
co -4- m
O 0 O
O O O
o o o
CM oo en
•* O en
O o O
CM en eo
O O O
in *3* "fl*
O 0 0
CM co en
rH rH rH
-a- r-« -^
co CM en
-a- CM co
Os O eo
O eo o\
rH m
+ 1+1+1
o\ in CM
rH o m
rH
CM •* rH
O O O
+ 1+1+1
o o -a*
r^ so NO
O O o
O O O
O iH O
rnSrH
CM ON CM
000
CM rH in
0 O O
-a- CM eo
O rH O
eO rH NO
rH CM rH
£ en en
rH r*. ON
r-os rn
SO 00 •»
rH SO rH
+ 1+1+1
0 rH OS
so CM OS
CM
CO m CM
O O O
+ 1+1+1
CO 0 00
O O 0
rH O CM
O 0 O
in o m
o o o
rH in iH
O CM -^
CO CN CO
000
rH 0 rH
O O O
in o u">
O O O
rH m o
rH i-H rH
en cn co
sO OO sO
ON CO -a-
00 OO ON
!*•* -a- m
rH CM rH
rH
+ 1 + 1+1
i
CO 00 CM
O O O
+ 1+1+1
rH 0 CN
r- oo r^
in oo so
O O 0
O 0 O
O O O
§88
ON 00 CM
rH O iH
cn o en
en m oo
o o o
§88
0 O O
o o o
•tf 00 CM
rH O rH
00 00 CO
•a- O cn
eo in oo
CM m m
ON 00 ON
o -a- -a-
rH
+ 1+1+1
CM ON ST
CM sO
rH
O ^H O
+ 1+1+1
00 so so
O O O
o o o
§0 •»
0 0
o o <•
o o o
rH 0 O
rH O O
CM CO O
CN CO CO
r^ rH CO
8§S
o o o
o o o
rH O -JT
rH O O
888
rH CO CM
rH CM CM
co r^ o
ON in CM
ON f-* ON
sO NO CO
in so so
rH
+ 1+1+1
00 O SO
CM scr **
rH
0 rH 0
+ 1+1+1
rH SO CM
oo in so
Os
0
in
o
o
0
§
so
CO
0
in
o
in
O
in
O
in
O
CM
in
00
00
rH
SO
+ 1
Os
SO
1
Os
o
+ 1
o
00
OS
o
0
§
0
o
in
CO
CM
in
O
0
-a-
o
-a-
0
o
CO
CM
O
er.
in
in
+ 1
O
1
00
o
+ 1
§
CM r^ eo
O o O
CM O CM
o o o
o o o
rH O «-H
m en CM
CM CO CM
CM en eo
m in m
S5S
CM r*. CM
o o o
ON O 00
000
o o o
CM CM CM
rH O CM
•^ rH NO
ON Oi ON
O r*^ ON
CM sO rH
rH
+ 1+1+1
rH O
• • CN
CO -3-
1 1 1
CO CM CM
0 rH 0
+ 1+1+1
o CM -a
oo o oo
rH
Arlth.
^
O
>
rH
X
b
V
00 O +|
CO f4
b U
0> CO
> c2 <
^ IK
E
0
4J CO
CO E
U 00 00
C B S
01 CO ^
Cl U
E
o
u
gj 01
•H >•,
cn H
sr co co
+ 1+1+1
-a oo CM
r-* so r>-
O O O
O o O
rH CO CO
000
000
0
rH
CO
0)
4-1
w
<— 1
.-H
<3
en -a- eo
+ 1+1+1
s;r sO eM
P-. so r«*
O 0 0
o o o
rH eo ro
o o o
O O O
rH
rH
CO
•H
(_,
4>l S~\
C t» rH
Cd 3 rH
,O "C **-*
^ C
Z3 rH
CM rH CM
+ 1+1+1
*3" o cn
r-. r^. r-*
§0 0
O O
rH CO CO
o o o
000
o
rH
rH
CO
•H
CJ
C OJ CM
CO S rH
& e ^
rJ O
n cj
CO CM CO
+ 1+1+1
0 O O
o o o
rH CO CO
o o o
o o o
o
rH
rH
ca
•H
4-1
E
01 ^-.
C T3 CO
CO *H rH
,0 co ^-*
IJ 111
3 OJ
K. 00 Os
000
+ 1+1+1
0 0 O
0 0 O
o o o
o o o
o
rH
rH
«
B i-t
CO Ij
J2 4-1 ^
kJ CO rH
3 3 CM
M -a *~-
3 E
CD rH
CM
rH
+
OO
SO
O
0
O
0
c
a
.0
i~i
3
XI
3
cn
CM
rH
+ 1
CO 00
r-* so
§0
0
0 0
8°
rH
rH
TO
•H
O
l-j .-v
(U CM
E CM
Q
O
rH CO O
CM rH CM
+ 1+1+1
O O O
O 0 O
o o o
o o o
o
rH
rH
CO
•H
C u
« c
,O QJ ^— *
l-i TJ CO
D 2 CM
& CO *> —
•=> 0)
CO Q-<
O 0 O
o o o
o o o
o o o
o
rH
rH
RJ
^
-J
4J
rH
rH O **
tB -H en
^ r-l -W
3 OC
od <
O
O
n
0
rH
ft)
|^
S
az
O 0
0 0
0 0
O 0
o
rH
4) .»~s
4j m
o co
E ^^
0*
CO
c
0)
e
m
rJ
3
to
(0
^
T3
OJ
1-
•H
(TJ
a
M-i
0
n
(U
&o
CO
O
^
CO
-------
41
00
3
M
<
f
4J
•H
M
<
E
•H
4J
&
*H
•W
CO
Q
• Frequencj
o
b
ta
o
•rl
U
•H
01
^i
cw
gresslot
Linear Re
Arlth.
Average
E
O
4J
HIVOI
Concenti
rH
cn
CO
£
§
£
(0
4J
CO 09
0 H
•H
SR (Q
£
B
0
M
to
r-l
00
PS
CD
01
•rl
J
O
•H
CD
«
O
00
hi
>
IH
b
+ 1
b*
+ 1
j
S + l
Jj
3 <
en
00 00
B 9
$ -
« H
CM rH O
o> r- in
m CM m
»3- CM O
m -a- r^
iH
i-H
r»- CM SO
000
-3- -* m
0 O 0
00 O"\ 00
o o o
O m rH
CM CM CM
&\ ON O
CM CM m
CN rH rH
en en en
oo CM r-.
0 O O
Sen m
O 0
CO O 00
O rH O
CM SD CM
CM CM CM
co r-* m
CM CM CM
CM CM CM
en en en
en eft O
oo r*. o
m
+1+1+1
in in en
1
rH ST rH
O O 0
+ 1+1+1
CM 00 O
NO NO NO
oo sr oo
+ 1+1+1
in en cn
Px NO Px
000
O 0 O
o o o
o
rH
tfl
1 5
1 :
H
<
m o r^.
*>* CM CM
O 1*- (N
o en o
sO -tf 00
•rH
so en o>
o> so m
rH CM
en O en
O O 0
m so en
000
m rH r^
0 rH 0
en •* -*
rH rH rH
en in CM
en en cn
en en o
-3- en -3-
en CM cn
0 0 o
cn en *3-
O 0 0
m rH so
O rH O
in so r*.
rH iH rH
en CM oo
en en CM
rH r^ CM
-3- cn -3-
SO 00 rH
00 1 — CN
oo en in
i-H Ov rH
+ 1+1 + 1
<• in oo
•* -* so
en r-* CM
O 0 O
+ 1+1 + 1
oo *n en
so so so
^ en «a-
+ 1+1+1
NO oo sr
P* NO Px
o o o
0 O 0
0 O 0
Urban 0 t
Industrial 100 1
(11) 0 t
in CM ON
O o sr
Px rH rH
sr ON -a
m en NO
rH
in sr ON
rH en sr
CM CM
00 O NO
o o o
in o sr
O 0 0
oo in o
O rH rH
O rH rH
CM CM CM
00 NO 00
CM CN CM
rH 00 rH
cn cn en
§O ON
O 0
o o o
oo in o
O rH rH
CM CM CM
co m CT»
CM cn CM
sO sO CM
CM CM CM
-31 m r**
P*» 00 00
o o» m
rH O\ rH
+ 1 + 1+1
so so *3-
O> fv, 00
en r-. CM
O o O
+ 1 + 1+1
CM CM ^»
m so to
oo CN oo
+1+1+1
CM 00 O
p* m p*
8§8
O 0 0
Urban 0 1
Commercial 100 t
(12) 0 t
CM cn ON
OO O rH
oo CM in
rH
O P^ P-
00 00
m o NO
0 0 O
sr sr ON
O rH O
ON en en
o sr o
NO sr oo
rH rH rH
ON O NO
CM O CM
00 ON p-
en e\i en
NO O ON
0 O O
sr sr CM
O rH O
o sr o
rH rH rH
ON o sr
CM O CM
O ON O
sr CM sr
sr cn oo
00 Px 00
oo oo en
CM CM CM
en
+ 1+1+1
in m CM
CM rH NO
m en sr
O CM O
+ 1+1+1
CM m sr
Px NO NO
P-. rH 00
+ 1+1+1
oo NO r-
p» NO r~
§8§
o o o
Urban 0 t
Residential 100 t
(13) 0 t
r>. so »^-
so m -a-
-* O C?\
sO -3" O^
r— 1
m cn oo
i-H rH CM
§o o
o o
§O r^.
O O
l^ O -*
O 0 O
O 00 r-
CM 0 0
en oo so
en en en
-* m -^
8§§
§00 sr
o o
p~ O O
o o o
r» O oo
CM O rH
CM sr cn
O NO ON
sr sr en
NO ON O
00 CM ON
H NO NO
NO r~ sr
rH
+ 1+1+1
cn in ON
ON cn sr
0 rH O
+ 1+1+1
sr sr r»
m IH sr
O ON ON
+ 1+1+1
c^ m r-*
m m m
8§8
o o o
Suburban 0 t
Industrial 100 t
(21) 0 1
O P^ rH
00 00 O
m sr CM
in rH NO
rH
in NO rH
NO O P»
ON O 00
O O O
S8o
sr r~ 00
i-H rH rH
O en m
CM en rH
oo o O
CM o cn
O O en
CN in CM
ON O 00
O O O
O 0 O
m p^ sr
rH rH rH
CM p~ sr
CM rH CM
NO p* m
CM rH CM
O O rH
CM in CM
ON rH 00
NO in px
H NO O
sr CM en
sr
+ 1+1+1
NO ON O
NO NO m
p- oo m
o cn o
+ 1+1+1
CM in in
in in in
oo m oo
rH rH rH
+ 1+1+1
sr IH sr
NO NO NO
§8§
0 0 O
Suburban 0 t
Commercial 100 t
(22) 0 1
NO oo cn
r- rH en
NO in oo
in sr NO
rH
p*. oo in
CN cn
O 0 O
CM sr sr
o o o
in oo pv
o o o
CM ON rH
CM rH CM
cn Px O
cn CM en
in oo CM
en cn cn
NO CM NO
0 O 0
CM NO in
o o o
in oo NO
o o o
CM rH CM
CM CM CM
en ON p-
en CM CM
CM en sr
en en en
00 en rH
00 00 ON
CM 00 CM
rH ON rH
+ 1+1+1
ON ON m
rH in sr
CM 00 CM
O 0 O
+ 1+1+1
o in cn
P- 00 NO
m cn NO
rH rH rH
+ 1+1+1
r*-. so r-
8§8
o o o
Suburban 0 t
Residential 100 t
(23) 0 t
rH r^ m
rH rH sO
i-H rH rH
m -3- m
1-^ rH 00
T-. rH 00
m o m
rH O rH
•st o en
0 0 O
m o co
O O O
CM O rH
r-. in en
CM •$• CM
O SO CM
cn en en
O\ OS O>
O 0 O
co o r»-
O O O
'« O en
O O rH
00 00 r*.
O -H O
'
O O r^
IN) in rH
rH
+ 1+1+1
oo rv in
'O 00 ON
SrH cn
rH O
+ 1+1+1
in en in
^ in in
00 CM OO
CM rH CN
+ 1+1+1
*3" O rH
00 SO 00
O O 0
o o o
0 O O
Rural 0 t
Agricultural 100 t
(34) 0 l
m m
CM CM
r*- r^
r* r^
CM CM
m O m
0 0 O
CM O CM
§O O
O O
8§8
0 O O
CM O CM
0 0 O
CM O CM
0 O 0
sr o sr
o o o
CM O CN
o o o
0 O O
888
o o o
CM O CM
0 0 O
CM O CM
O O O
sr o sr
NO NO
p~ px
en en
00 00
+ 1 +1
00 00
sr sr
cn en
+ 1 +1
px p^
00 00
CM CM
+ 1 +1
rH rH
8§8
000
o o o
o
rH
01
co o m
b i cn
£ " ^
n
8-10
-------
m oo \o
CM CM CM
00
3
I
•J
g
en CM en
•a- CM o
O ro o
vo -4- co
co in r-* in O
m ** \o vo »tf
in -a- CM NOinoo i-trn
LI
<
CM m m
m oo co
NO in ON
ON rv m
i-H CM
m -a- on
rH CO -tf
CM CM
O
00
00
in en oo
i-H rH CM
m vo *
vo o
r*v oo in
rx -a- CM
CM CO
. r-t oo
. FH oo
3
J3
U
I
ey
V
Ll
U
•o
cj
LI
OH
r-
CM
O
CO
CO
1-f
rH
m
rH
vD O
co eo
0 O
iH iH
in co
vo co
CM ro
rH rH
O vD
00 «* ON
o o o
vo co r-.
ooo
O -I rH
ss^
r-s ON vO
CM CO CM
eo co O
CM CM CM
CM ft vO
CO ** CO
.H CM O
rH i-t t-H
rH O «-l
r-» CM vo
eo in CM
rH rH «H
CO ON CM
CM m CM
O rH O
ON O vO
O O rH
383
CO O\ CM
CM CM CM
on O co
O m
CM rH
O O
0 O
r^ co
O eo
O m
O m
CM CM
eo co
00
o
rH
rH
rH
rH
•»
ON
CM
00
rv o rv
en in en
in o en
CM O CM
ON 0 0
O O rH
rH rv O
CO rv rH
O i-H rH
rH rv O
O en NO
CM CM CM
00 rH rH
rH CM rH
CM O CM
rH rH rH
O en ON
oo NO rv
•H 0 rH
rH FV NO
-a- en
rv oo
rH rH
ON O
0 0
CM ON
rH O
O CM
-S- ON
o
in
NO
rH
en
O
O
rH
O
SC
O
CM
g
g
g
O
NO
0
o
CM
O
O
§
o
0
o"
NO
o
o
•H
s
11
06
1-.
5
i
00
CM
r^.
o
en
rH
00
t-H
NO
rH
00
rH
NO ON
en CM
CM 00
rH O
NO •»
rH rH
o- rv
rH rH
CM m
•H rH
rH rv
rH rH
ON en en
rH O CM
00 rH rv
O 0 O
ON CM sr
0 rH rH
i-H rH NO
CM CM rH
ON NO ON
I-H en I-H
in rv CM
CM CM CM
en oo NO
en in en
00 CM 00
O rH O
ON CM O
rH rH CM
rv in rH
r-l rH rH
en CM vT
rH rH rH
rH CM O
rH rH rH
*3- en m
CM <»• CM
ON ON 00
O CM O
ON O rH
O O rH
NO O NO
i-l O iH
ON -3- 00
CM rH rH
^
CM rH rH
§in -a-
rH rH
O 00 00
CM O rH
rv o ON
O O CM
o m rH
-3- rH rH
en NO -a-
rH -4* rH
CM rv in
-a NO rH
in o m
i-H O i-H
m o .3-
rH 5 I-H
ON o 00
O O 0
S8S
ON en o
O en rH
O ON in
CM CM CM
ON in oo
O rH O
O O NO
i-H rH rH
en en vo
i-H rH rH
rv. rH in
in en o
CM rH CM
oo m i-t
en -a- -a-
NO o en
O rH rH
O ON ON
rH O O
ON ON 00
rH O rH
rv ON o
ON rv ON
O CM O
0
o
o
CM
O
O
o
CN
o
vt
o
CM
O
o
o
CM
O
CM
0
CM
O
-»
O
CM
CM NO CO
NO -a- NO
o NO o
m o m
oo NO rv
en o -a-
NO m rv
NO m NO
rv CM ON
en .3- o
HI
Ll
oo
£
+ 1
oocosD iHi
>n
+ 1 + 1+1 +1+1+ 1 +1+1+ 1 +1+1+ 1 +1+1 + 1 +1+1+ 1 +1+1+ 1 +1+1+ 1 +1 +1
en o en co O oo
O en
CM
o\ m ro
O CO
<
oo o
It -H
Ll U
01 CO
+ 1
It*
CO
|J Ll Q)
O W 00
> C C
rH 0) tO
sue:
§
cj
menm ooenco ONI-HO oooo oo co co
en oo rv o co
+ 1+1+1 +1+1+1 +1+1+1 +1+1+1 +1+1+1 +1+1+1 +1+1+1 +| +|
CMOOO mencM oo-a-NO rv
OI-HNO CMIHCM mrvcn
on
S
•& CM -a1
§o o
o o
ooo
ooo
0
• -*•
CO 3 rH
,O tJ rH
Li C v-'
01 ,
. HI '
a.
.O TJ
3 C •
cn M
eo
C -rt
« cj
.0 ki
3 E
C u
CO C
XI 01
U T3 ^
3 -H en
ja m CM
3 a) v^
en PC:
o m
S en
QJ ^x
I
3-11
-------
Similarly, given equal predictive power and the choice between a
model requiring stratified data sets and one without stratification,
the one without stratification would be simpler and therefore more
desirable.
The key question to be answered is that of equal predictive
power. Two criteria have already been advanced and one more is
proposed. One model will be a better predictor than another model if:
• the standard deviation of the average ratio is significantly
reduced by stratification,
• the correlation coefficient of the linear regression slope
model is significantly increased, and
• the distribution of percent differences between predicted
and measured values is shifted to the lower percent
differences.
For each pair of data points represented in Tables 8.2.1 to
8.2.3, the absolute value of the relative difference, d., between
the measured and predicted values has been calculated from:
8-12
and the fraction of the d. in the intervals 0 to .1, .1 to .2, .2 to
.3, .3 to .4, .4 to .5, and greater than .5 have been totaled. These
distributions were formed for each stratified set of data for both the
linear regression slope, and the average ratio receptor models.
Figure 8.2.1 illustrates the distribution drawn from the comparison of
SSI concentrations calculated from the linear regression model with
3
the measurements for all sites in the 0 to 300 ug/m HIVOL
concentration range. This histogram reflects the set of six numbers
(43, 33, 13, 4, 3, 5) taken from the third row and eighth column of
Table 8.2.1. Every set of six numbers in columns 7 and 8 of
Tables 8.2.1, 8.2.2 and 8.2.3 can be interpreted in the same way. For
example, of the 683 predicted SSI concentrations from the linear
regression receptor model, the set of numbers in column 8 states that
di =
C2i -
A
C2i
C2i
8-12
-------
50-
50
>50
Absolute Value of Difference Between
Predicted and Measured Values, Percent
Figure 8.2.1 Frequency Distribution of relative differences between
SSI predicted and measured concentrations calculated
from the Linear Regression Slope Model for all sites.
3-13
-------
43%, or 294 of those predictions were within +_10% of the measured
value, 33% or 225 of the predictions were within +10% to +20% of the
measured value, 13% or 89 were within +20% to +30% of the measured
value, etc.
If the receptor model derived from a stratified subset of all
data shows a shift in the error distribution toward lower errors when
compared with the overall data set, then the receptor model derived
from that data set is a better predictor than the model derived from
the entire data set.
With the three evaluation criteria in mind, it is possible to
examine the receptor model relationships between SSI and HIVOL, TOTAL
and HIVOL, and FINE and HIVOL to define the simplest model and to
estimate the accuracy with which IP and FP concentrations can be
predicted from routinely acquired TSP concentrations.
Table 8.2.1 shows the model calculations and evaluation
parameters for the SSI/HIVOL relationship. The scatterplot of all
validated data points available from the IP Sampling Network appears
in Figure 8.2.2a. This figure shows a strong linear relationship
between SSI and HIVOL measurements. This is verified by the results
in Table 8.2.1. The correlation coefficient of the linear regression
model is .93 (recall that collocated HIVOLs showed a correlation of
.98 in Section 3.5). The relative standard deviation of the average
ratio is +18%. Most important of all, 90% of all the SSI
concentrations predicted by this model are within jH30% of the measured
values for both the average ratio and linear regression slope models.
The simplest model derived from the entire data set is
SSI = .72 x HIVOL 8-13
with a precision of
°SSI = -13 x HIVOL 8-14
Within the same data set, stratification by concentration does not offer
any advantage. The standard deviation of the average ratio is not reduced
by the stratification, and the average ratio itself changes by at most .04,
8-14
-------
a) SSI/HIVOL
b) TOIAL'IIIVOI
sag
Z7t
Z-4!
lit
I8«
c 158
1
i IZ1
^
>
n.
Ui
EC
39
(
SLOPE. • .644*- .010
INTERCEPT i S.ZZS+- .985
CORRELATION : .SZ7
NO. OF POINTS i 683
,
i i i »
" ' t i *» '
: .^
sb eh gb isa (tin isa zla ^4u z;a 3
330
Z76
Z4H
ZIC
"4 18C
c
1EC
IZt
6fl
38
i (
SLOPE • .588-1- .888
INTERCEPT : 6.335-t- .674 t
CORRELATION : .898
NO. OF POINTS 1135
,,., X ;",'; '•' '
'^if^'*
HIVOL Concentration,
HIVOL Concentration,
c) FINE/HIVOL
30,
SLOPE , .227+- .008
INTERCEPT : 8.87&1- .7Z8
CORRELATION : 589
Z4 J. NO OF POINTS: 1 135
zic-
g IZI-
nha ria—rta—zfa
HIVOL Concentration,ug/n
Figure 8.2.2a-c - Scatterplots of SSI/HIVOL(a), TOTAL/HIVOL(b),
FINE/HIVOL(c) for all IP Network Sites,
0 to 300 vg/ra HIVOL Concentration Range.
8-15
-------
or by less than 6%. The prediction error frequency distributions are not
shifted to the lower percentages by the HIVOL concentration
stratification. There is little difference between distributions resulting
from the average ratio and the linear regression slope model. Therefore,
the simplest model against which to evaluate other SSI/HIVOL relationships
is that of Equations 8-13 and 8-14.
The remainder of Table 8.2.1 contains the model parameters for
different site-type stratifications of the SSI/HIVOL data sets.
Stratification by HIVOL concentration range does not substantially
improve the predictive ability of the model within any of these subsets.
At most, there is an 8% difference between RA calculated for the 100 to
3 — 3
300 ug/m range and the RA calculated for the 0 to 300 ug/m range
derived from the urban industrial stratification. All of the other
differences are substantially less than this.
The a_ do not decrease for the more specific concentration
K
ranges. The correlation coefficients are actually higher for the 0 to
3 3
300 ug/m ranges than for either the 0 to 100 or the 100 to 300 ug/m
ranges. (This is more a function of sample size and range of data values
than of an improvement in the model. As the number of points in a sample
increases, so does the minimum significant correlation coefficient.)
Most important, the prediction error frequency distributions between
concentration range subsets are nearly the same. Only the urban
3
residential 100 to 300 ug/m range shows a substantial shift from the 10
to 20% and 20 to 30% intervals to the 0 to 10% interval. Because the
number of samples in this subset is only 12, this only represents shifts of
a few samples which is not significant. The conclusion to be drawn with
respect to stratification by TSP concentrations for predicting SSI
concentrations is that it offers no improvement over the model formed from
the unstratified data.
With regard to the choice between an average ratio and a linear
regression slope model, the site-type stratifications confirm the
conclusion drawn from the unstratified data set, that one model is as good
a predictor as the other. There is very little difference between the
prediction error frequency distributions of either model applied to the
same data set. Thus, within each stratified data set, the average ratio
without TSP concentration stratification is the simplest receptor model.
8-16
-------
Comparing these average ratios between site-type stratifications shows
little difference between the average ratio calculated from all sites and
the average ratios determined from sample stratification. The average
ratios range from .66 to .77, a maximum deviation of 7% from the value of
.72 for all sites. Only in the suburban industrial case is the standard
deviation of the average appreciably smaller than that for all sites, but
since so few samples are in this subset, this reduction is probably
insignificant.
The conclusion of this analysis is that the best receptor model for
the prediction of SSI concentrations from HIVOL measurements is that of
Equations 8-13 and 8-14. A caveat is important here. Four hundred, or
59%, of the data points in the "all sites" sample come from Philadelphia,
and most of the sites in this area are of the urban industrial type. The
model is heavily influenced by them. Even though a variety of site
classifications is represented, the particle size distributions which were
sampled in Philadelphia may be more similar to each other than they are to
particle size distributions in different cities.
The relationship between TOTAL and HIVOL may be a more representative
receptor model to predict IP from TSP because the number of simultaneous
TOTAL/HIVOL measurements is nearly twice as great and the measurements are
more uniformly distributed throughout the network than are the SSI/HIVOL
pairs. It's predictive ability may suffer, however, because of the change
in filter media from 1979 to 1980 which was shown in Chapter 3 to induce a
potential mass bias under certain conditions. This relationship is
examined in Table 8.2.2.
Figure 8.2.2b shows the scatterplot of TOTAL/HIVOL pairs. Though a
linear relationship is discernible, it is not as strong as the one in
Figure 8.2.2a. The correlation coefficient of .90 is still fairly high,
however, indicating that a linear model is a reasonable approximation.
Both the linear regression and the average ratio models predict TOTAL
concentrations within ^30% of the measured values for 80% of the data
pairs. This pattern holds true through all of the stratified data sets.
There is, therefore, no reason to choose the more complicated linear
regression slope model over the average ratio model. This model shows less
prediction accuracy than the corresponding SSI/HIVOL model.
8-17
-------
Stratification by HIVOL concentration has a more significant effect on
the TOTAL/HIVOL relationship than it does on the SSI/HIVOL relationship.
This is consistent with the cumulative frequency distributions of the TOTAL
and SSI concentrations examined in Chapter 7. Figure 7.2.1c shows a
horizontal tendency for the high concentrations for the TOTAL samples while
the SSI (Figure 7.2.1b) distribution continues to follow a straight line.
This is possibly due to the loss of material from the COARSE filters. This
3 3
difference between the 0 to 100 ug/m and 100 to 300 ug/m HIVOL
concentration ranges appears to be significant in most of the stratified
data sets. The suburban industrial and suburban commercial subsets show
only small differences between average ratios for the concentration
3
stratification, but since the number of samples in each 100 to 300 ug/m
category is small (less than 13), great significance should not be attached
to this observation.
3 3
The average ratios in both the 0 to 100 ug/m and 100 to 300 ug/m
ranges for the urban industrial, urban commercial, urban residential, and
suburban residential site types differ by no more than 8% from those
calculated from all sites. This is not surprising since 81% of the data
pairs in all sites come from these four site types. The simplest model for
these sites should be:
for HIVOL < 100 ug/m3
TOTAL = .75 x HIVOL 8-15
aTOTAL = -18 x HIVOL 8~16
for HIVOL > 100 ug/m3
TOTAL = .63 x HIVOL 8-17
0TOTAL= ' 16 x HIVOL 8-18
The suburban commercial category shows a very low correlation,
.79, between TOTAL and HIVOL and it yields an average ratio which is
15% lower than the average ratio for all sites. The average ratio of
.68 for the SSI/HIVOL relationship in Table 8.2.1 is also lower than
that for all sites, but not to such a great extent. The poor
8-18
-------
correlation, the limited number of data pairs and the inconsistency
with the SSI/HIVOL relationship for the same site-type stratification
cast doubt on the use of a separate receptor model for the suburban
commercial stratification.
The rural agricultural classification also shows a substantial
increase in the average ratio with respect to the ratio for all sites,
but the standard deviation of this ratio is large, indicating a great
variability of this ratio. The prediction error frequency
distribution for this classification shows the average ratio model to
be a much better estimator than the linear regression model, though
3
for the 0 to 300 ug/m range, the correlation coefficient of .92 is
higher than that for all sites. No SSI/HIVOL data is available for
comparison. This site-type classification could be a legitimate
alternative to Equations 8-15 and 8-16 (notice that for concentrations
in the 100 to 300 ug/m range, the average ratios and standard
deviations are nearly identical for rural agricultural and all sites)
but more confirmation is needed.
It appears, then, that the receptor model embodied in
Equations 8-13 to 8-16 is the most reasonable one for predicting TOTAL
concentrations from HIVOL concentrations. For HIVOL concentrations
less than 100 ug/m , the TOTAL/HIVOL model is almost identical to
the SSI/HIVOL model, with the average ratios differing by only 4%.
3
For the HIVOL concentrations greater than 100 ug/m , however, there
is a substantial difference.
Table 8.2.3 is the counterpart to Tables 8.2.1 and 8.2.2 for the
relationship between FINE and HIVOL concentrations. The scatterplot
of this relationship for all sites appears in Figure 8.2.2c. A strong
linear relationship between these two measurements is not visually
evident, and the low correlation coefficient of .59 confirms this lack
of association.
As Table 8.2.3 shows, neither stratification by HIVOL
concentration nor by site type increases the correlation between FINE
and HIVOL to a significant level. The highest correlation
(discounting the rural remote classification for which only five data
points are available) is only .69, far below the .90 which was
specified in the model evaluation criteria.
8-19
-------
The prediction error frequency distributions consistently yield
more than half of the predictions differing from measured values by
more than ^30%, showing that none of the models can be relied upon to
accurately estimate FINE concentrations from HIVOL concentrations.
It must be concluded from the results in Table 8.2.3 that the
proposed receptor model is inadequate for predicting FINE
concentrations from HIVOL measurements.
Average ratios, R., for SSI/HIVOL and TOTAL/HIVOL from sites
having at least 19 data pairs are plotted in Figure 8.2.3 where they
have been stratified by site classification code. No HIVOL
concentration stratification was used in calculating the ratios for
•^
TOTAL/HIVOL because the numbers of points in the 100 to 300 ug/nf
range of individual sites were too low to make a calculated ratio
significant. The average ratios, RA, and standard deviations,
aR, drawn from Tables 8.2.1 and 8.2.2 have been placed on the
graphs for comparison.
As indicated by the stratifications in Tables 8.2.1 and 8.2.2, a
site-type classification is irrelevant to the relationship between IP
and TSP. Apparently the particle size distributions are just as
variable at rural and suburban sites as they are at urban sites. This
does not imply that the distributions vary in the same way, but that
the effect of those variations on the IP/TSP ratios is the same.
8.3 Predicting IP from TSP Concentrations
The average ratio model provides a reasonable prediction of SSI
and TOTAL concentrations from HIVOL concentrations when the average
ratio and standard deviation are calculated from the same data set. for
which the predictions are being made. In practice, if IP measurements
were available, then there would be no need for a predictive model.
To make this receptor model a practical predictor, it is necessary to
demonstrate that the model derived from one set of SSI/HIVOL or
TOTAL/HIVOL pairs can predict, with a given degree of accuracy, the
SSI or TOTAL values associated with a set of HIVOL concentrations
which had nothing to do with the formation of the receptor model. For
example, given one year of SSI/HIVOL or TOTAL/HIVOL data at a site,
8-20
-------
1.0-1
o .9-
•H
rt
as
j .8-1
.7-
w
(U
oo
rt
o>
M .6H
.5H
.4
b1
+ 1
.8-
H .7H
o
00 . i
a .6-
—r~
12
b
+ 1
35
11 12 13 21 22 23 34 35
Site Classification Code
Figure 8.2.3 Average SSI/HIVOL and TOTAL/HIVOL Ratios at
IP Network Sites
8-21
-------
how well could one predict SSI or TOTAL at another site in the same
vicinity or at a site in a completely different area? Given one year
of simultaneous IP and HIVOL measurements at a site, how well could
the IP values from the previous or later time period be predicted?
These questions can be answered by using measurements from the
Philadelphia sampling sites, the ones with the most complete data sets
in the network. The receptor models can be formed from the data pairs
acquired at one site with simultaneous SSI/HIVOL or TOTAL/HIVOL
measurements and applied at another site with similar measurements.
The percent error frequency distribution can then be used to evaluate
the effectiveness of the model as a predictor of SSI or TOTAL
concentrations. Comparing this frequency distribution with the
distribution yielded by the model derived from the site's own data set
will demonstrate the extent to which a receptor model obtained from
one site can be used to predict concentrations at another 'site.
Tables 8.3.1 and 8.3.2 contain the results of applying the
average ratio and the linear regression slope models derived from the
year-long data sets acquired at 500 S. Broad Street (PAPHA), Allegheny
(PAPHB), NE Airport (PAPHE), Presbyterian Home (PAPHG) and St. John's
Church (PAPHI) to the SSI/HIVOL data pairs of all sites in the IP
Network outside of Philadelphia, to all sites outside of Philadelphia
of one site-type classification (urban commercial), and to itself.
The prediction error frequency distributions obtained by using each
model are compared in Table 8.3.1 for SSI and in Table 8.3.2 for TOTAL.
For example, in column 4 of Table 8.3.1, the average ratio of .75
derived at 500 S. Broad (PAPHA) was used to predict the SSI
concentrations at all sites outside of Philadelphia. The distribution
of the absolute values of the percent differences between calculated
and measured values is (43, 31, 12, 8, 1, 5). This distribution can
be compared with the distribution of the model derived from the
predicted data set to evaluate its effectiveness. For this example,
the average ratio yielded by the predicted data set is .72 and the
predicted error frequency distribution is (43, 33, 14, 3, 2, 4).
8-22
-------
U as
Z o
s :fl
I li
^ U *T*
& 3
CO O *"*
3 M S=
H 5 M
•:! S |
2g S i
all 2
2 a S g
H u. o z
S 3 "
1 1 :
i i s
S S g
S 5 s
w w S
OS CO S
Cu CO g
oo -a-
r* oo in
. . * rH
to cn ON
0 X II II II II
<2 B •-<
.H0~ 0
rl S O — < *
W • ft.
<
cn
Jj
•^J
cn
3
1
CO
3
1
cn
•<
.^
cn
3
g
f** »* co o IA m
CO CM —1 rH
O GO -^ r»- ^ ^
iri oo
JJ T-l . 0 • CS
•H JC
cn PH
II II II II
,— 1 U
rH 3 <" i-J^J C
•< Ji IO2 lOC.
r«* oo r*" CN oo oO
CM ^H CN rH
r««. in CN oo oo O
CS( CN CN •-*
m co oo tn oo o
CO CN rH rH
in o HI r»* 01 m
CN n CM O *A 00
CH CN CN rH
co m in CN CN GO
CN Cn rH rH
mo IA oo o CM
>* ro
«tf
J=
o.
I—I
rH 0»
qf -0
.H rt
U rH
(_, .H
g Qy
O a-»
U O.
1 0)
c u ^5 [ibl • • * ^
M m
,P CD
U II II II II
rH 4J
rH .rJ ,< J.T1 C
< cn la: to
CM *A 00 n rH i— 1
in en
m 1-1 oo CN m rH
m f)
CN »H <7\ *d- i—l un
CO CO rH
CM CN oo O m -d-
rH \O CT» rH rH CN
m m
in CM rH
U 60
co c
at >H
0.-0
a «
(0 (U
x
u
M 01
a> B
u co
a> c S
E o
cfl a> i-i
^ AJ
4) (T) -H u
•H t-l
W rH W .H
ft) QJ '*•'
§TD T3
o c c
o n 3 "-*
o
o
o
.0
r-t
,c
o
X
(J
cO
0)
o
U-r
•O
U
0)
(4-1
tw
•o
jr
3
U •
3 >>
*3 ft)
• H
C-H ^
M U
U
•o a
at u
> tu
1-1 r-f
O
W "
& M
°S
Is
£
•o "
" t1,
tj f
• H U
4; 4>
CL 00
ft) 13
£ g
O »^S
0
C u"1
0
.H O
4-1 -U
o
rH (0 O
4) r-l
rH CO 01
4i co O
•O C Q* f*1
O O r-
CO 01 B*S
O CO r- O
.H QJ m
4J W tO
co C-0 U O
OS OJ ft* JJ
a; go
M) V4 3 tN
(0 co C
V* ft) •>
> .S -no
02 J= o
< J t- ^
Ct) ^N U
8-23
-------
CM
,
00
3
9
H
5
S
1
03
M
e
CO
W
D
.E QUENCY
E
ERROR
1
^-
*"*
g
£
g
CO
Id
H
CO
g
CO
E?
lERIVED I
a
MODELS
j
g
M
X
i-3
g
j
Cd
I
i
z
^—
o
u
T INCLUD
s
85
&
u
i
CO
1
I
CO O O
10 CO -4 CO
C .—• i-3
O SB H II II II
"" 5; < jo c
^» ft* |OB IOC g£
CO 4
to
e
o
4-1 O
3 01 ^i *-4 i-l to
js e oo oo • ct
.ri O <-N • • esl 00 J
ki a: o -a-
« • o. n n n n
.-. w <
Q U O. < -U C g
>> 0- ^
u
g
3
cr
01 in r» O
kl 4J 00 IX .tf CO
[14 kl ... Qj
M §.•£ "* S
ki •* £ n n u n
kl < «£
u o. < -u E pi
U _. |o: la; >
•OK <
01
u
0) 1-4 SO O
kl 00 ** IX CO
C ^"* lx -d1 »J
01 09 -" rt
0) •<
-J A- < -1C B g
< •*
•o CTS en o
flj lx sO 00
ki x-> oo in ag
• a- ii n ii ii
o ^-^ lotf totf !>
O ^
^
01 «
CO » u
kl 19 C
19 i) -0 •-!
C9 01 kl W C
D E 01 00 B J2
19 -e B 3 n
T! 1- C •-> — I OS
oi a 3 -e o -3
-JO. « 0
U ki 01 tg
'•o "ol 01 r » g
is S" « it-
's u
u tn
•u a
kl 19
0, 0
*o c
en CM CM
IX r- CM -IMS IX
CM CM CM *-l
to
0)
.,4
CO
ffl
1
at
•o
to rt O oo O O
a> r-i r-. in CM oo
u .MI . * , r*
•H js m
M Cu
II 11 1) II
^ 3 < J^l C
f-» 00 so "-i O f-
-^- C?N r-. ON ON <">
CM •-« <-4 CM
CM -^ in co oo oo
CM i-4 r-* F* CM
m vo ro i— ' •—* NO
CM fH »H .-1 ^^ CM
CM (-4 ^4 t-4 CM
en \o m o —« *o
CM PH fH i— *«N CM
^^ CO O ^* m O
CM CM CM i-H CM
m oo co oo m NO
CM CM «-4 i-l
m CM ^ »-< oO i-»
CM CM i-t i-l CM
•^" O \O ON pH ^
CM »-< P-4 »H CM
0 r^ ^ ^m*
f^> CM CM i-l
in ON r*. oo m so
CM CM «-* r-l
t
IB "0
.- B
O i-< to
B S "
B co
u a
B 0 CM
JO U . . • OS
ki m ^H
3 u
01 II II II II
^ •" < -]J3 B
< CO IOC |Q£
sO 00 sO O O O
cr> m so o o o
men
00 00 * ~*
0)
ki M
IB B
01 --J
a-e
a 19
(0 V
j=
ki
0) 4J
01 to ft
B O
19 « ki
kl U
oi 19 a u
u aT3 to
"4 kl
CO r-4 kl -H
Ot U 1W
J-o -o
O B C
S3 —
•J
•o
Jg
01
o
-I • CO
•C E
£ O
to
O to
— 01
U kl
<9 60
cd oi
01 ' *
60 ki
19 19
01 C
•S J
1 1
g to
IB JO
8-24
-------
For the SSI/HIVOL relationship, the receptor models derived from
each of the three Philadelphia sites predict the SSI concentrations at
all sites in the IP Network outside of Philadelphia with nearly the
same accuracy as the model derived from the data taken at those
sites. Ninety percent of the SSI concentrations are predicted within
+30% of the true value by the average ratio model derived from the
predicted data set while 86%, 92% and 82% of the SSI concentrations at
sites outside of Philadelphia are predicted within +307° by the
receptor models derived from 500 S. Broad, Allegheny and St. John's,
respectively.
Table 8.3.1 confirms the conclusions drawn from Table 8.2.1 that:
• The average ratio and linear regression slope models predict
with approximately equivalent accuracy; hence, the simpler
average ratio model is to be preferred.
• Stratification by site type does not enhance the predictive
ability of the model.
Concerning site-type classification, the one urban commercial,
and two urban industrial site types represented by 500 S. Broad,
Allegheny, and St. John's, respectively, predict with approximately
equal accuracy. In fact, the average ratio model (.67 average ratio)
derived from the urban industrial site at Allegheny is closer to the
model (.66 average ratio) derived from the urban commercial data set
than is the model (.75 average ratio) yielded by data from the urban
commercial site at 500 S. Broad Street. It is interesting to compare
the average SSI/HIVOL derived from 60 measurements at urban commercial
sites outside of Philadelphia, .66, to the average ratio derived from
194 measurements from all urban commercial sites, .73 (from
Table 8.2.1). Apparently the large portion of Philadelphia
measurements greatly affects the model parameters.
The final prediction error frequency distributions for average
ratio and linear regression slope models show that the model derived
from each individual site is a reasonably good predictor for that
particular site. Over 96% of the predictions are within j+30% of the
measurements.
8-25
-------
Table 8.3.2 presents the same analysis for the TOTAL/HIVOL
relationship. Here, the number of data pairs available for
calculating the ratios is low for all sites except NE Airport (PAPHE)
which had 74 simultaneous measurements. None of these models is as
good a predictor of TOTAL concentrations at other sites as the
SSI/HIVOL models of Table 8.3.1 are, even though they are extremely
effective in predicting the TOTAL concentrations at the sites from
which they were derived. Approximately 20% of the predicted values in
each case differ from the measured values by more than 50%. This is
in contrast to only 7% with errors greater than 50% for the
TOTAL/HIVOL model derived from all data in Table 8.2.2.
With the exception of the Allegheny model, none of these TOTAL
predictors derived from individual Philadelphia sites is as good as
the predictor derived from all data, in sharp contrast to the models
derived from the SSI/HIVOL relationship in Table 8.3.1 which are just
as valid as the model of Equations 8-13 and 8-14.
The SSI/HIVOL average ratios obtained from 500 S. Broad,
Allegheny and St. John's differ from the average ratio of
Equation 8-13 by +4%, -7% and +8%, respectively while the TOTAL/HIVOL
average ratios at 500 S. Broad, Allegheny, NE Airport, Presbyterian
Home and St. John's differ from the unstratified ratio of .73
calculated from all data by +8%, -8%, +11%, +11% and +7%,
respectively. The higher percent differences are probably due to the
lack of HIVOL concentration stratification for the TOTAL/HIVOL model
which was found in Table 8.2.2. Because each of the Philadelphia data
sets examined in Table 8.3.2 lacked sufficient HIVOL values greater
than 100 ug/m^ which were associated with TOTAL measurements, it was
not possible to create such a stratified model from measurements at
these sites. The importance of this TOTAL/HIVOL concentration
stratification will be addressed shortly. The other reason for this
lack of agreement may be the change in HIVOL filter media between 1979
and 1980. The majority of the Philadelphia measurements was taken in
1979 whereas the number of measurements in other areas is greater in
1980 than in 1979. The Philadelphia models may not apply to
TOTAL/HIVOL data for this reason. A 1979 and 1980 data stratification
might be appropriate to test this.
8-26
-------
The relationship between SSI and HIVOL concentrations derived
from single sites in a single geographical location is able to produce
a model similar to the one derived from many sites and many locations
and to predict SSI from HIVOL at geographical locations other than
those at which it was derived. Due to the inadequate amount of data
available and the possibility of interference due to filter media
changes, this conclusion cannot yet be drawn concerning the
TOTAL/HIVOL relationship.
Another strategy for forming a receptor model relationship is to
locate one set of HIVOL, SSI and/or TOTAL samplers in an area while
keeping the existing HIVOL network intact, then using the set of
simultaneous measurements from the single site to derive a prediction
model for SSI and TOTAL concentrations at the other sites. This
possibility is examined for the Philadelphia Bridesburg industrial
area (See Figure 4.2.8) using the long-term monitoring site at
St. John's to derive the model. The percent error frequency
distributions at each site for SSI and TOTAL appear in Table 8.3.3.
The distances of these sites from St. John's, as shown in Table 8.3.4,
range from .70 (Pilot Freight) to 1.1 km (NE Transfer) with the
closest long-term site being Allegheny at a distance of 3.2 km.
The average ratio at St. John's is .78 compared to .67, .61, and
.71 at Allegheny (3.2 km distance), Pilot Freight (.70 km distant) and
the T & A Pet Shop (2.8 km distance), respectively. SSI data at other
sites in the Bridesburg area were insufficient for comparison. Even
though these sites are very close together, the average ratio of .72
derived from all sites will be closer to the true model at each site
than the one derived from the St. John's data. This is borne out by
the SSI/HIVOL prediction error frequency distributions at Allegheny
and Pilot Freight where 40% and 39%, respectively, of the predicted
values differ from the measured values by more than +30%. The
St. John's model is a good predictor for NE Wastewater, nearly as good
as the model derived from that site.
The TOTAL/HIVOL model comparison is more informative since all
sites in the Bridesburg area have some TOTAL data with the exception
of Allegheny and Pilot Freight. Average ratios at Bridesburg sites
range from .75 to .79 and are well represented by the .78 ratio found
8-27
-------
,J
d s
M W
O M
S Otf M
g ft- W
C <
ET O 06 M
S 33 "2
g "* s 3
H B fH M
" O < OS
0 *> H
•x CO CO
en f; • Z D
.SHOP
en g co M g
•1 In i
cp w w
< gg O O CO
I H 8 S
a s j g
I ° s M
g a e g
f 1 | g
^* 2 CO
M CO
CO
o
rH
CO
CO
- ^co o >r\ ai
C M i-. i-v . _l
X S3 • • en p-
O PM rH
0. « H « »
w < ,40 e
co |os|« g
a
01 u
u 01
U CO rH
•H 01
•O W "0
01 U O
rl «fl X Pd
ft. Q >
01 01
4J CO
U
•H n
is*
".
.• s s s s
C "^ ' ' en rH
OB CK
< II « II II
. S p!
U ^s < rU C >
co jttio: <
«
01 4J
U 0) rH
U CO O
•H TJ
•o n o
0) 4J Z
P. Q >
TJ
W 01
U CO
•H
•a n
o> u
b a
CM Q
en rH rH CM
C/* rH r^ ^ O *O
CM CM en
CM en CM
*^ vo r^h ON • • •
0) pa rx <•
JC a i-H rH
01 < » « U II
rH 6,
rH >_ < JJ2 B
CM rH rH rH rH rH
ON ON CM 00 00 *7
CM rH rH rH rH
en en rH
CM CM oo O en *a*
en en rH rH
r~ o o
(s vo in rH
§^> . , .
pa vo r~
£ S rH t-
ooo.
r> r~ in
01 X-N . . .
4J X CM
a CL,
U «— < rJ.0 C
BE |Ptf |PB
in rH oo NO NO en
•* CM rH
CM IH in NO en en
m CM rH
oo NO CM O p en
•» en rH
m O CM O en O
in en rH
0)
4J
(0 rH 00 O
JJ r-* r^ ^
01 x-v . • •
w M •» en
n S en
(B PH
3 <• II II II
g N^ < JJ3 C
nn m
i^n en
O rH NO en o O
o oo ON en o o
in en
00 rv CM
f. f- ON
4J t-9 CM
01 S en
IX 0,
< II II H II
5 N^ < ,JjO C
H I.P4 103
n
p
4J
e
0)
•H
>H
M-l
3
to
C
M
(S DM
.sS
8-28
-------
m n ooo o
C "-N
O SB
1 GO O
•o w
01 01
I (Oi-<
I OJ
I WO
•O 4J O
2<§z
«* iH O OOO SO
m ro
rH O CO O OO
c
o
U
00
U
n
H
•o u
01 01
4J CO
U
•H
•O U
*<
II N
< J.
IP^ltt!
01
M
4-> < II
CO &4
O OM
a.
V 01
4-1 CO rH
111
(t "O
ij O
CU B J5
2
*i
CO 4J
U OJ
CJ CO
01
00
01
o ^--
U Z
•U <
o
^£
.2 "
jj s ^
co ••••
rg in oo m I-H t-t
in M oo CM m i-t
£ M
O SB
: n n
\u tei
8-29
-------
CO
oo
s
H
CO
z
s
g
M
M
H
co
M
Q
&H
W
&
8
PH
o$
Si
z
8
PREDICT
CO
g
O
•
H
C/3
0
z
H
CO
CO
M
CO
M
3
1
3
rH
a
PL)
co
a
H
i_^
co
od
g
EH
o
co
1
H
S
[ CONCENT]
co
co
H
O
o
%
PM
8
SMENTS
2
1
s
HH
CO
CO
•o
8
r<
pa
0
O Jl
*3
o
co
g
•a
0)
r-l
r-l
O
CO CO CM r-l r-l
CM
co m «tf^ocor-i^oo
(O r-l i-H CM <•
I-l
oo
r-i m cooo
CO CM CO CM
-*
•
vfr m oo »* o CM CM
CO r-l r-l CM CM r-l r-l
3 g
CO -i-l
- 4J
(3 0 •
J3 -H Cd
>-j o) cd
• CO {X
J-1 1-1 4-1 CO
CO -H (30-
Cg QJ O C
B CX OMjS6^5v5S~SB~SS-S
0 r<>4-iOOOOOO
i-i cd a) "~) i— i CM co «* m
q-l 4J pu (U
cd O'OOOOO^g
0) 13 CUC4JW4J4J4J4-IO
u -u 4) co m
C M-I 3 (-1 s-s s-e s^s B-e
cd O r-l 0) 60O OOOO
4J O'Wfi r-l
-------
at St. John's. Eighty to ninety percent of the individual TOTAL
concentrations at these sites are predicted to within +20% by the
St. John's model.
In a normal sampling network, a single site, St. John's in the
Bridesburg case, would be chosen to represent the entire land use
area. Another sampling site meeting the siting criteria which was
located several kilometers away in the same area would be just as
acceptable. The differences between measurements at these nearby
sites reflect the accuracy of using a single sampling site to
represent the suspended particulate matter concentrations of an entire
area. The differences between mass concentrations at nearby sampling
sites in the Bridesburg area was noted in Chapter 5. These
differences can be expressed in the same form as the prediction error
frequency distributions to compare the measurement accuracy due to
sampler siting with the accuracy of the IP prediction models.
In Table 8.3.4, the SSI measurements at St. John's were used to
predict the SSI concentrations on the same days at other Philadelphia
sites. Unfortunately, only two nearby sites in the Bridesburg area
had enough SSI measurements on days simultaneous with those at
St. John's to be included. No sites had an adequate number of
simultaneous TOTAL measurements to perform a similar analysis for
TOTAL concentrations.
The distributions in Table 8.3.4 show that only 57% of the SSI
measurements at Pilot Freight, which is less than one kilometer from
St. John's, were within _+30% of the St. John's measurements. From
Table 8.3.3, the average ratio model predicts 62% of the Pilot Freight
values to within ^30%. In this case, the accuracy of the average
ratio receptor model is comparable to the accuracy of representing the
Pilot Freight SSI concentrations with a nearby, and equally valid,
sampling site.
The St. John's average ratio applied to NE Wastewater in
Table 8.3.3 predicts more than half of the SSI concentrations at NE
Wastewater within _+10% and nearly 90% of the SSI concentrations to
within +30%. This is a substantial improvement over representing the
SSI concentrations at NE Wastewater by the simultaneous SSI
measurements at nearby St. John's. From Table 8.3.4, only 24% of the
NE Wastewater values are estimated to within _+10% by the simultaneous
St. John's measurements.
8-31
-------
The percent error frequency distributions for Allegheny and
500 S. Broad are included in Table 8.3.4 for comparison purposes even
though they are not part of the land use area intended to be sampled
by St. John's. The St. John's average ratio model is a better
predictor of SSI at Allegheny than are the simultaneous St. John's SSI
measurements, though neither one is too accurate; St. John's SSI
differs from Allegheny SSI by more than +30% for more than
three-quarters of the simultaneous measurements. The average ratio
model predictions differ from measurements by more than +30% for 43%
of the Allegheny measurements.
These data are limited and similar comparisons need to be made in
other areas; however, if these relationships between SSI
concentrations at nearby sampling sites are representative of the
situation in other land use areas and geographical locations, then two
conclusions can be drawn:
• The accuracy of the average ratio model for representing IP
concentrations is comparable to or better than the accuracy
of sampler siting in measuring the overall concentrations in
a land use area.
• There is no advantage in deriving an area specific average
ratio SSI/HIVOL model over using the average ratio derived
from the entire IP Network data set.
Unfortunately, since the TOTAL measurements are so few at closely
located sites, similar conclusions cannot be drawn about the
TOTAL/HIVOL relationship.
Another approach to defining a model at a specific site which
might be more accurate than the average ratio derived from all data is
to take simultaneous IP and TSP samples over a period of time,
establish the average ratio, and collect only TSP samples thereafter,
using the average ratio to estimate IP concentrations. The
simultaneous sampling should cover an entire year to contain all
probable meteorological and emissions events. None of the sites in
the IP Network had collected more than one year of data at the time of
this analysis. The ability of one year to represent another year is
not yet subject to evaluation. It i._s possible to divide the SSI/HIVOL
and TOTAL/HIVOL data sets from selected sites into two data sets,, each
8-32
-------
containing alternate pairs of values, to calculate the receptor model
from one subset, and to test its effectiveness in predicting the SSI
or TOTAL measurements from the HIVOL measurements in the other subset.
The prediction error frequency distributions for this procedure
applied to SSI/HIVOL and TOTAL/HIVOL measurements at 500 S. Broad,
Allegheny, NE Airport, Presbyterian Home, and St. John's sites are
presented in Table 8.3.5. The alternate measurement subsets are
designated by "odd" and "even", though these designations have nothing
to do with the dates on which sampling took place. The model derived
from the "odd" set is used to predict the data in the "even" set.
Table 8.3.5 also contains the average ratios, slopes and intercepts
derived from the "even" sets and the prediction error frequency
distributions of the models for each subset predicting their own
values, for comparison.
Each of the "odd" models predicts "even" SSI and TOTAL
concentrations as well as the models derived from their same data
sets. The prediction accuracy is very good for all models with over
80% of the "even" SSI concentrations, with the exception of Allegheny,
being estimated to within _+20% by the "odd" models. The TOTAL
predictions are nearly as accurate with 100% of the predicted values
within j+30% of the measured values.
Thus, it seems that the variability of the size distributions at
a specific site is less than the variability of the distributions
between sites. If simultaneous IP and TSP data are collected at a
site over a period of time, it appears possible to use the model
derived from that data to extend the IP data base over adjacent or
included periods for which TSP measurements have been taken. This
possibility must be evaluated using additional data when they become
available.
8.4 Predicting Averages and Maxima of IP from TSP Concentrations
The simplest models derivable from the current IP Network data
set for estimating IP concentrations and their precisions from TSP
measurements are those of Equations 8-13 and 8-14 for IP measured with
the HIVOL size-selective inlet (SSI), and Equations 8-15 to 8-18 for
8-33
-------
<
o
s £
g i
D °
s «? ,„
g 23 £
J2 § <
S SB ft.
-J 1 S 3
" W g °
00 §• P
a |El
(Q W ofi
s g w B
1 | *
" £ r
M1
H 3
i £
S-,
g
6u
M
C/9
CO
a CQ
- x-, oo so m a
S t-i *—v r*. •a • ^
£ S T> • • ix CM
O ft. -O IH i-l
"-> < O
O. ^- 1 « 1 1
• w
s Nv° c g
<
in
3
01 w
4J 0)
o cn -H
•rt 01
•O B tj
0) jj o
£<§* |
V 4J
01 01
4-1 CQ
CJ
•* B
T> 4J
01 B
n O
a.
CA
co moo 3
r- oo m
C **• • • •
C M x^ CM cn
o SB *o m
^ ft. Ti
< O « • « i
• ft. *-* 96
u >— < -10 B >
co ips««r <
2
5 w
4J 01 F-4
U CQ 01
^ ^3
•o n o
01 u X
b B B6
B.O |
•O
0) 4J
u oi
U CQ
•H
•0 B
0) 4J
U CO
B. O
co rx r«. oo O O
m t-. r^
CM m cn O O O
-» CM cn
O m r* o o O
in cn IH
cn cn r*. co o O
-a- cn •-.
s~*
•c I
O < iH CM O
h CM CO IX CO
oa —• • • •
cn CM
• ^ ^H
CQ B
o ,, , , , ,
o oi < >J.e B
m s_>|Bj|ee
rx
m «n
CO CO OS CM O CM
in CM
5 CM SO O ^ SO
cn
O SO OX CM O CM
SO CM
<
•O B
B 0.
O < in -a- O
u ft. rx eo rx
ea N^ • • •
«cf cn
. <-N m
cn E
o . , , , ,
O 01 < JX B
m ^-IK \et
m cn cn o O O
cn cn cn
cn o O o f~ O
cn m n
3
o
£
•o <
o a. co so o
Bag^s:-!:
s^ rx CM
• iH !-(
CQ /-s
0 ? " " * "
O "o < JJ3 B
m ^- IBS IDS
m cn
m oo os CM o CM
m CM
M
%
•si
o a. m o o
L. < ?- oo m
co a. . . .
s—» CM cn
• m
cn *-.
•c B • « n
O "O
O O < -JJ= B
m >-ioe|oc
so O O
>s *^ ^^so m -^
B CA *C3 • • •
S S ? sr
.c a, o P-I rx
60 ^ *~*
01 RH 1 1 I 1
FH ^ < JO C
< IB; IB;
•o
01 u
u 01
o cn —I
*H 01
•o « -o
01 *J O
£2*
SO Os O
>, so -a- IH
B ^ ...
u ea ^-> in co
X X -o FH cn
COB. -a
41 < o 1 i i i
.-1 a. <~*
s^ .-V406
•o
V m O O O O
tH «H m *H
^s
i
< CM O O
XB. SD •» O
B <_>•••
01 iH SO
£ *-v CM
CO B
OP S 1 1 1 1
-H >
r4 01 < -JXl B
< ^|K|OS
CM cn CM
CM so oo co cn cn
cn CM *H *H
CM so •* >H cn m
cn CM CM rt
cji Os eo IH fH cn
CM CM iH pH *4
ea
s
DM
< co o o
x a, so m •«
c ^ • • •
01 IX 00
f ^x. IH cn
o> oi ii I « n
iH >
— Ill < «JJ9 B
< X^IOSIB;
as as os o cn o
CM CM CM IH
*a- cs a^ -a- ^y O
rt CM CM rt rt
ea
a
ft. so O O
>,< so msr
c ft. . • .
01 v^ «» IX
£ >H
00/~-
01 -o i H n »
i-l O < JJ= B
< xx leaps
00 Os 00 O O V*
m CM o
os CM oo m o m
CM -ff — I
^\
ea
x
a. so o> o
X< SO •» iH
B £ • • •
0> w in oo
£ FH r*
00 ^x
ar-o ii ii n n
^ T(
-H O < -I.O B
< x^iea IK
8-34
-------
»J
I
?
3
B
-H
4J
C
O
u
in
m
00
S
H
H(
W
i— < ON 00
oo r- oo
i* m cn
o cc
a/^ H n H n hJ
Iw U /-v
•H S3 t> < lOC C
< Pu •olaiictf
•a
01 4J
U 01
u cn
•H
•o a
0) 4J
Pu Q
m oo ^* *^ o o
•3- sf
r^ 00 CT\ O O O
n ro ro O O O
•a- -a- iH
n oo -a- -» o o
•a- •*
Ed
£
4-1 •< e*J CN o
M PM 00 ON m
O ^— ' • • •
Q. -* CO
u* /-\ esj
*H C
< Q) II U II H
>
u o) •'I »J»o c
a
a
o
ti
a
•H
u
•f-t
14^
•s
u
B
O
1- U
•H PC
< S
§ S^
i-t O\ O O O O
m oo o o o o
^^
u
4-1 Cu r-l ON 00
tj < 00 1^ 00
0 Cu • • •
jj PO
< -C « II II II
•o
U 0 < JX> B
n oo *a -a o o
O
01 33 t-l oo O
B Pu oo ^ ^
a PU ' 4 CM
B/-. n u n n moowooo
PU o IBS IP<
r-i rn m O O O
•» •» tH
•o
01 4J
4J OJ fH
o cn 01
tH 73
•O « O
h <8 -a- -»
Pu Q
C?
Pu
01 < CM CM O
O *-^ • • •
^-\ CM
• C
n oi II II « N
01 >
^
01 X
0 <
pe PU
a ^
0) -o
tl *O
Pu 0
*~s (fl
to
o
4J
c
CU
*O 4-» O
d) (1> iH *H
4J C/3 01 (4-1
0*0 y-i
•H co O 3
T3 AJ IE tD
OJ TO C
M Q M
d-
»H ON O O O O
r*. esj
CO 00 O O O O
vO CO
^^
o
X
Q> OM iH 00 O
E < oo r-
-------
H
B
^
3
B
•H
e
0
o
in
,
CO
3
9
H
M
CO
CO
to M
- EC
c £
X < CO
O CM r»
. ^ II
T>
4J 41 M
U CO 01
•rt -O
•0 C8 0
01 *J £
£2
M
to I O
* OH 00
e < •
j: PM
o *-*
JJ T3 <
Stl-H
O CO 01
•H -O
•O CO O
oi w SB
US
oi
o
rt O
^t ^ ON
c^c |
CO
3
1
•0
01 *-"
u 01
O CO
•H
T3 CO
0) 4J
h CO
OH Q
CO
* s
u
JJ a>
o en
•H
•o w
u 01 ^ • • •
o ^ -»
-: B
01 H n H n
*J 01 ^I *JJ3 E
\D -a- o o o o
in "*
•a- •» o o o o
^
M
0) DM 00 -«
8-36
-------
IP measured by the dichotomous sampler (TOTAL). In the final
analysis, the effectiveness of these models can only be judged by
their ability to estimate the high 24-hr and annual mean
concentrations of IP for comparison against the values of a standard.
To test the effectiveness of the models in making these
predictions, sets of measurements from each site were chosen which met
one or more of the following requirements:
• The two highest HIVOL concentrations with simultaneous SSI
or TOTAL measurements.
• The two highest SSI measurements with simultaneous HIVOL
measurements.
• The two highest TOTAL measurements with simultaneous HIVOL
measurements.
Some sets of values satisfied more than one of these requirements
and in many cases data were insufficient to satisfy all of them.
Because of the limited number of simultaneous measurements available,
the highest concentrations meeting these criteria at several sites
were not "high" by common standards. HIVOL concentrations ranged from
31 to 439 ug/m3 with the majority in the 100 to 200 ug/m3 range.
Seventy-two pairs of SSI/HIVOL and 144 pairs of TOTAL/HIVOL values met
the criteria.
Equations 8-13 and 8-14 were used to predict SSI concentrations
and Equations 8-15 to 8-18 were used to predict TOTAL concentrations
from HIVOL measurements. The percent error between the predicted and
measured concentration was calculated for each SSI/HIVOL and
TOTAL/HIVOL pair. The distributions of these errors for SSI and TO^AL
appear in Figures 8.4.1 and 8.4.2, respectively. Superimposed on
these figures is the Gaussian uncertainty distribution predicted by
Equation 8-14 for SSI predictions and Equation 8-18 for the TOTAL
predictions.
The first feature to note is that the errors for these high
concentrations are not distributed according to a typical normal
distribution. In both cases, the distribution seems uniform between
-30% and +30%. According to the normal distribution, 68% of the
predictions should fall within plus-or-minus one standard deviation
8-37
-------
•* CM
(N t--
CD •
f-i in
3 CD
tfl 4->
OJ -H
CD w
C O
« S
•P .
T) d> C
CD 2 §
4-) CD
CD 3
CO !-H
CD
o 4->
c
CD CD
-a
c
(-1
03
Q fn
O
4-J U-l
C
CD W
O C
>H O
CD -H
D. 4->
aj
"+-I *H
O 4->
C
e CD
03 O
^ C
O
a3
CD
3
O
4->
-------
o
ro
O
CM
O
o
I
in
T
O
O
-d
0)
C
(U
0)
5
•M
0
CO
H)
(J
c
(U
+J
C
CD
O
(X
"0
CO
S .3
V T3
•P 4-> ^
U rt S
00"'
fn 3 -X
a. I-H f-i
ctf aj
C > E
0
0 P I/)
2 w -P
4-> 0 O
0 x; a
02 bo
« X •
0 T3
O f-i 0
C O T3
0M-I3
f-i
0 (/) O
4-i C C
4-1 O
•H -H
Q P 0
n} fn
P ^ 03
C
0 C w
O 0 0
M O 3
0 C
a,
O a3
>
to
I
V
S H
03 O
h E--
M
O f-i
•H ^J-
S CN
00
f-i
CxO
I/)
0
-t-J
•H
3-39
-------
and 95% of the predictions should be within two standard deviations
(assuming the precision of the measured values is much smaller than
this standard deviation). Figure 8.4.1 shows approximately 62% of the
predictions within one sigma and approximately 86% within two sigma of
measured values for SSI.
Figure 8.4.2 shows approximately 60% of the predictions within
plus-or-minus one standard deviation and approximately 93% within
+2 sigma of the TOTAL measurements. Even though the distribution of
uncertainties does not appear to be normal, the confidence interval
defined by the standard deviation of a normal distribution by
Equation 8-10 adequately represents the error to be associated with
the high concentrations of suspended particulate matter.
The prediction error frequency distributions for these high
concentrations are (44, 31, 12, 3, 3, 3) and (35, 25, 25, 8, 6, 4) for
SSI and TOTAL, respectively. These compare to (43, 34, 13, 4, 3, 4)
and (32, 25, 22, 8, 5, 7) taken from Tables 8.2.1 and 8.2.2 for the
data at all sites in the network. The receptor model predicts
individual high concentrations with the same degree of accuracy with
which it predicts lower concentrations.
Figures 8.4.3 and 8.4.4 show the percent error distributions of
the arithmetic average predictions for SSI and TOTAL, respectively.
Only averages containing 20 or more SSI/HIVOL or TOTAL/HIVOL pairs
were used. The receptor model is a better predictor of SSI averages,
with 100% of the averages being predicted within +20% and 66% being
predicted within +_10%. For TOTAL concentrations, 89% of the averages
were predicted within +20% while 61% were predicted to within +_10% of
the measured values. The average ratio receptor model predicts
averages with greater accuracy than it predicts individual values.
8.5 Using the IP/TSP Average Ratio to Evaluate Compliance with
Standards
If the data used to form and test these receptor models are
representative of the measurements which would be made in any part of
the United States, then the receptor model of Equation 8-13 is the
simplest and most accurate one for predicting IP concentrations
8-40
-------
3 5 10 15 20
ference
Between Predicted and
51 at all IP Network
s. 12 Values are included
-20 -15 -10 -5 (
Percent Dif
Histogram of Percent Differences
Measured Arithmetic Averages of S
Sites with more than 20 data pair
CD
fn
3
bC
'siU3ui9jnsB9H jo
8-41
-------
o
CO
A
O
CO
in
CM
o
CM
in
in
in
i
o
I
in
i
CM
I
CD
o
CD
fc
0)
K-l
<*•<
• H
Q
f-H
CD
fn fn
3 O
CD
tt) -H
4-> 00
O
rt
4n O >
'•H H
•H 00
Q <4-l (Nl
O
CD
CD
C3
CO
CD V)
bo i-i
rt -rt
f-i rt
CD c_
'-H < rt
O -P
O rt
e -H Q
oj •(->
Vi CD C
to 6 CM
O f,
•P -M C
23 < •*->
oo
CD
•H
U-
jo
8-42
-------
measured with a size-selective inlet from HIVOL TSP measurements
provided the uncertainty expressed by Equation 8-14 is also used to
define the confidence interval to be placed around it. Similarly,
Equations 8-15 to 8-18 prove to be the simplest and most accurate
prediction mechanisms for IP measured by the dichotomous sampler.
Compliance with a 24-hr or arithmetic average determined from
these predictions can be looked at in two ways.
First, if the confidence interval around the predicted IP value
is comparable to the difference between nearby sampling sites
assessing ambient concentrations in the same portion of a
neighborhood, as the example of Section 8.3 showed, the argument can
be made that the uncertainty of Equation 8-14 is comparable to the
sampling precision so that there is no significant difference between
the TSP derived IP concentration and one which is actually measured.
Therefore, the predicted concentration is as good as a measured one
and should be compared against the values of a standard in the same
way an ambient measurement is.
A more restrictive, yet safer approach is to add some number of
standard deviations to the predicted IP concentration and to compare
this value to the standard. If twice the standard deviation derived
from Equation 8-14 is added to the prediction of Equation 8-13, and
the resulting concentration does not exceed the standard value, then
one can be 95% certain that a measurement of IP at that site would not
show a violation.
Of course, these estimates should only be used where monitoring
data do not exist and decisions based on them should be tempered by
the comparisons between predicted and measured IP concentrations made
in this chapter. A small but significant portion of the predictions
for IP Network data examined here differed by more than 50% from the
actual measurements. Without the simultaneous IP and TSP measurements
it is imposible to tell which IP prediction will fall into this
category.
8-43
-------
This chapter has explored the possibility of estimating IP and FP
concentrations from TSP concentrations using average ratio and linear
regression slope receptor models. The frequency of occurrence of
percent differences between predictions based on these models and
ambient measurements were used to evaluate the extent to which these
models represent reality. Based on the limited data available from
the IP Network, the following observations resulted from this analysis:
No simple relationship between FP and TSP was found which
would predict FP from TSP with adequate precision. For IP
Network data nearly 70% of the FP predictions differed from
the measurments by more than +20%.
The average ratio model derived from all IP Network
measurements is the simplest relationship between IP and
TSP. For SSI and TOTAL IP measurements, stratification by
site type does not increase the precision of the model's
predictions. The model is independent of TSP concentrations
for SSI predictions, but it exhibits a TSP concentration
dependency for TOTAL predictions. Seventy-six percent of
the SSI measurements and 57% of the TOTAL measurements were
predicted to within +20% by this model in the IP Network.
Models derived from simultaneous TSP and IP sampling at a
single site in an industrial neighborhood predict IP
concentrations from TSP concentrations at nearby sites in
the neighborhood with a precision which is comparable to the
difference in IP measurements obtained from nearby sampling
sites.
Arithmetic average IP concentrations at IP Network sites
were predicted from TSP concentrations to within +20% for
all SSI maxima and for 89% of the TOTAL maxima. This degree
of uncertainty may be adequate for estimating IP averages
for compliance purposes in certain situations.
Maximum IP concentrations at IP Network sites were predicted
from TSP concentrations to within +_20% for 77% of the SSI
maxima and for 60% of the TOTAL maxima. The uncertainty
associated with predicting maximum concentrations is
comparable to that associated with predicting any
concentration, irrespective of its magnitude.
8-44
-------
CHAPTER 9
IP AND FP COMPOSITION AND SOURCES
The chemical composition of suspended particulate matter has been
the subject of a great deal of study in recent years, primarily for
two reasons:
• The effects of different chemical species are not equal;
some are more harmful than others.
• The composition of suspended particulate matter is
indicative of the sources from which it originated.
It is well-known that certain substances, for example lead,
cadmium and arsenic, are poisonous at much lower mass concentrations
than the tolerable concentrations of calcium or aluminum. The recent
National Ambient Air Quality Standard for lead, a 1.5 ug/nH
quarterly average, was promulgated in recognition of its toxic
properties. It is conceivable that other chemically specific
standards might be formed in the future. While the toxic nature of
certain pollutants measured in the IP Network may be of importance, it
is not appropriate to examine this subject here.
Receptor models, which were introduced in Chapter 8, have been
devised as a means of using aerosol composition information to
estimate the quantities contributed by various emissions sources
(Watson et al, 1980). These models have suffered in the past from a
lack of sufficient information about the composition of emissions and
ambient samples. These needs are slowly being met, and the chemical
analyses performed in the IP Network offer a possibility for testing
the usefulness of receptor models as a tool for developing national
particulate matter emissions control strategies.
In Chapter 4 the potential sources of IP and FP in selected urban
areas were identified, their major chemical components were tabulated
and their likely particle sizes were noted. Major point source
emitters of TSP which had been identified by the National Emissions
Data System were located on maps with respect to the sampling sites
and speculations were advanced about the relative concentrations of
9-1
-------
different chemical species in the FINE and COARSE fractions of the
ambient samples which would be observed if certain sources were making
significant contributions to certain receptors.
This chapter examines the chemical concentration measurements
acquired in four of the urban areas described in Chapter 4 to compare
the chemical compositions within them and between them and to estimate
which sources are the major contributors to FINE and COARSE chemical
concentrations. The broader purpose of this examination is not to
characterize the source contributions in these urban areas so much as
it is to determine the value of routine measurements acquired by the
IP Network in attributing ambient concentrations to sources. Chapter
4 took the source-oriented point of view, identifying and locating
sources and predicting which aerosol components should be observed at
nearby sampling sites if a source were a significant contributor.
This chapter looks at the same problem from a receptor-oriented point
of view, looking for elevated chemical concentrations that are
associated with the emissions of identifiable source types. The
approaches are both complementary and reinforcing. Where a
receptor-oriented interpretation indicates a source type which is not
included in the source-oriented interpretation, such a source must be
sought. When the source and receptor-oriented approaches agree, there
is strong evidence that the indicated source is affecting the receptor
in question.
To evaluate the IP Network data for these purposes, the
discussion is broken up into four sections. The first section
presents the average and maximum chemical concentrations obtained from
EPA validated data summaries for one year at each site in Buffalo,
Houston, El Paso, and Philadelphia and examines their external
consistency. As noted in Chapter 3, the chemical composition data did
not appear to pass through EPA's data validation criteria summarized
in Section 3.4, nor were uncertainty estimates available to accompany
the data. The data show some inconsistencies among themselves and
these are pointed out in Section 9.1.
The second section examines the differences in particle
composition as a function of size and location. The central city
compositions are compared to the urban scale background site
compositions, and similar site types from different urban areas are
compared.
9-2
-------
In the third section, the chemical element balance receptor model
is applied to the average concentrations presented in Section 9.1 to
evaluate its usefulness with respect to IP Network data and to
estimate source contributions to TOTAL, FINE and COARSE
concentrations. The model is also applied to more extensive chemical
measurements made on individual samples taken in Philadelphia to
determine whether or not additional aerosol measurements would aid
source apportionment. The results of optical microscopic analysis of
coarse particulate matter from several sites in Philadelphia are also
examined in this section.
The final section of this chapter applies the receptor model to
average concentrations on a neighborhood scale to study the
uncertainties of source attribution which could be due to sampler
siting.
9.1 Validity of IP Network Chemical Composition Measurements
Annual average and maximum 24-hr chemical concentrations from the
HIVOL, FINE and COARSE samples taken at IP Network sites in Buffalo,
Houston, El Paso and Philadelphia are tabulated in Tables 9.1.1,
9.1.2, 9.1.3, and 9.1.4, respectively. Birmingham, Phoenix and Denver
are not examined here because these sites lacked sufficient continuous
and consistent data at the time of this report. The values in these
tables were taken from EPA validated data summaries (EPA, 1981a) and
have not been passed through the filter of inequalities 3-1 to 3-4.
Before attempting to draw any conclusions from the data, it is helpful
to examine their consistency with each other to separate real
variability from that variability introduced by the measurement
process.
The data validation process through which individual data records
pass was insufficient for producing valid summary statistics. The
validation procedure did flag inconsistancy of mass measurements;
however, flagged data were included in the calculations of averages.
The average and maximum concentrations for Big Sister offer a good
illustration. Table 9.1.5 presents a series of values which were
included in the routine statistical summary supplied by EPA/EMSL.
9-3
-------
1— 1
•
r-l
•
ON
W
nJ
«
<:
H
AT IONS
aS
H
Z
W
o
z
8
HEMICAL
0
a
X
3
•<
w
CJ
/-s
o
oo
• — .
ON
1
ON
f-
O
i— 1
s^
>•
Z
A
s
0
3
S^
CO
Z
o
CENTRATI
Z
8
,j
<
CJ
M
S
w
g
J
p
>
0>
4-1
• -^
co
c
• I-l
1— 1
o
i
CO
•
ctf
4J
0. X
fi'-N Ctf
-1 /"N V
(Xi D
^^
4J Z
i^
,5 M
r-l >
9 <
^
X
ff\ fji
CM £
«
*£
• D
CO PQ
. »vj K*
Cu Z >
<^^ N— ' •
v^ <:
vO X
«M ct
• /™\
0 <
85 g
. &
CO Z
• ~— '
PLI W
u
Chemical
Species /
o
m o
oo o NO
ON
ON
CM
P~
ro
,-( CO
in o
vO o *D
O
ON
i-l
i— I
i— I i— I
ON o CM
CM
O
i-l
ON
00
O NO
m o CM
1-1
00
V0
o m
vO o "-•
ON
00
CO
M
cfl I-i 3
£ u U
m CM
m vj- vf CM co
OONOoOOi-lO I-ION
m m m oo
•-I CO i-l
i-t O
O i-1 CM i-l CO i-* i-4
r-lONCOOOOOOi-l
P~ O NO
CM
i-l vO
ON CM CM sf 1-1
COOvOOOOOON
00 l-H r-l P~
CO •-(
vO
•-I CO vO O CM ON
vD CO i-H O O <(• 00
CM m m
i— i
* O i-l CM r-l px
COvti-IOOO COOO
CM in co
I-l
H-* n
01 ,0. C «rl -H O O
fao,s;zf->toz
en co CM
OO i-l 1-1
m i-i i-i
r-»
CO
CM
in
w
c
o
•r*4
4-1
t
01
CO
43
O
CA
M-l CO
o ca
£
•
O
Z
rH ON
CM rH
09
4J
6
v
e co
0) C
r-l O
W M
o>
r~^
.0
ca
4J
U
0)
4-1
01
•o
4J
o
z
«
9-4
-------
_. x~>
V "e
01 <
1 5
o g
U 0
""" w m
Hcu
_, , 4~l
z 5 -H
p CO
3 B J?
**"(
1 § 1
£3 B
PQ •, <8
£ w
rH f5
M
rH 5
» s
s 5
Ja W
5 2
3 E
•
n)
4J
CO
X
o, ct)
§^N S
O
Cn O
«
.C P
£ z
3 ""*™^
B 60
rH >
i ^f*
•H •<;
»
x-\
«*
•*~s
to x
CM ig
£
• X"N
O U
Z D
tf)
• *
CO Z
• >-/
rtj >\p
M-t OU
X-v >
co **
*~s
/"•*.
v^ X
>- «
d) *5*
w ^M
4J /-N
CO PQ
•H »
CO PQ
>>
ooz
^J .. , JkM
•rt >— ' O4J
PO >
x^ •<
CM
X
CM £
*^* A*
• x-s
o <:
ZC-N
t— J
.C
CO Z
• ^^ be
fr, >
x-s <
iH
^X
i-H
(0 CO
U 01
•H -H
€ u
(U CU
J2 Ct
CJ CO
X— \
o
x^ r-t
00 •
. in
^/ N^ V-' S_/ V-'
COOOO"*CMm (OCMvOr^COCnrHiH iH
encM«JtinOocM ooeococorHinen- CM \O i-t
CM
cMOOiHCMi-HcocMCMr^enr-icnenin . ^o r^ r-t
•sT -H
menmr-icoi^ rtcMcMcor-fOOsj--* co
rHCMOCMOOO OOrHrHCMOr-tenm
r>. CM m
iH
srvOi-tcMOmrocoeniHvoOooo rH
OOmCMr-lr-lrHCMOCMCMCOenCnvOCO rHCM
CnrHr-lr-l tH rH OOSTCn
vO CM
ooOencocMOOr-icoencocM-a-rHcn en
mcococooocMOOOvO'>*-*t— ICM vor~
CO CO en rH
co
CO
CO IKT I CO
IDrH AC3i-H-H-HPCUX)-HC OO
S^^osuuzE-ioafaftHeoNcocoz
-» 1 CM
rH
oo en o
m rH
CM •* en
/•*
•U
l-i C8
CU (0 4-1
CO 4J U
.0 C 01
O 01 4-1
co B co oi
M-l CO O> C *Q
O (0 rH O
S W M 4J
O
0 Z
Z n)
9-5
-------
-o
0)
9
c
o
o
JH
S3
03
rH
•
i-H
ON
a
co
00
3
CO
23
o
w
H
w
o
55
8
53
o
H
CO
!
fl)
4-1
CO
|?
00
CM
o
CO >->
. 60
ro
PQ
4-1 PQ
CO p
60
CM
60
p p
en
(0 CO
O d>
•H T^
a o
e *
CO
o
CO
o
00
in
I 1 1 1 1 1 1 1 1 1 1 1 1 1
I CM
o
r-.
co
co
in
CO
' V»/ \^ '
tH 00 *H -* O O 00
in CM
ON
iHrHrHOOrsCNOO
in
H GO
m
^^ PO ^^ c^
OO iH •* 00 rH
r>. \o
m
oo
in
CM ON ON
in IH IH
CO
o
0) CO
CO 4J
§ g
M-l CO
-------
TABLE 9.1.2
AVERAGE AND MAXIMUM CHEMICAL CONCENTRATIONS
IN HOUSTON, TX (10/79 - 9/80)
(Data taken from USEPA, 1981a)
HIVOL CHEMICAL CONCENTRATIONS (ug/m3)
Sampling Site
(l)CAMS 8
Chemical (TXHOA)
Species Avg Max
Mass
Cr
Cu
Fe
Pb
Mn
Ni
Ti
V
so £
NOjj
No. of
Mass
49.2
.0077
.28
.78
.30
a
.011
.01
a
10.5
1.34
Observations:
Elements
Ions
(113.5 )
( .0169)
( .45 )
( 1.67 )
( .49 )
( .06 )
( .031 )
( .01 )
( .014 )
( 18.5 )
( 2.78 )
18
5
5
(2)CAMS 1
(TXHOB)
Avg Max
86.0
.0112
.10
2.19
.35
.09
.024
.01
a
8.3
1.72
(157
(
(
( 7
(
(
(
(
(
( 16
( 5
23
10
9
.9 )
.0029)
.16 )
.90 )
.69 )
.25 )
.109 )
.02 )
.030 )
.4 )
.23 )
(3)Seabrook(B)
(TXHOC)
Avg Max
43.7
.0054
.27
.50
.11
a
a
0.01
a
6.7
1.35
(87.4 )
( .0089)
( .43 )
( 1.28 )
( .16 )
( .04 )
( .006 )
( .02 )
( .017 )
(10.9 )
( 2.71 )
22
6
6
aNot detectable
9-7
-------
TABLE 9.1.2 HOUSTON, TX (Continued)
FINE CHEMICAL CONCENTRATIONS (ug/m3)
Sampling S ite
Chemical
Species
Mass
Al
K
Ca
Mn
Cu
Cl
Ni
Ti
Br
Fe
Pb
Si
Zn
S
SO 4°
N03
(DCAMS 8
(TXHOA)
Avg Max
17.4
.43
.12
.32
.02
.03
.16
.01
.02
.05
.20
.32
.27
.07
3.18
9.2
.41
(40.5 )
( .7 )
( .17)
( .45)
( .02)
( .05)
( .32)
( .02)
( .03)
( .08)
( .39)
( .50)
( .36)
( .12)
( 5.28)
(15.7 )
( .55)
(2)CAMS 1 (3)Seabrook(B)
(TXHOB) (TXHOC)
Avg Max Avg Max
23.7 (48.4) 14.3
.24
.14
.34
.01
.01
.14
a
a
.01
.08
.10
.27
.02
2.15
6.0
.68
(43.1 )
( .35)
( .33)
( .72)
( .01)
( .02)
( .54)
a
( .02)
( .02)
( .15)
( .14)
( .37)
( .03)
( 4.37)
(12.7 )
( 1.67)
No. of Observations:
Mass 8
Elements 4
Ions 4
aNot detectable
38
0
0
3
6
6
9-8
-------
TABLE 9.1.2 HOUSTON, TX (Continued)
COARSE CHEMICAL CONCENTRATIONS (ug/m3)
Sampling Site
dl)CAMS 8 (2)CAMS 1
Chemical (TXHOA) (TXHOB)
Species Avg_ Max Avg Max
Mass
Al
K
Ca
Mn
Cu
Cl
Ni
Ti
Br
Fe
Pb
Si
Zn
S
S0£
N03
No. of
Mass
17.0 (28.2 ) 8.5 (12.0 )
3.15 (
.17 (
1.65 (
.02 (
.02 (
1.11 (
.01 (
.02 (
.02 (
.61 (
.08 (
1.79 (
.04 (
1.01 (
2.1 (
1.15 (
Observations :
8
7.69)
.22)
2.57)
.02)
.03)
2.95)
.02)
.04)
.03)
.96)
.12)
3.37)
.07)
2.45)
4.6 )
1.68)
38
(3)Seabrook(B)
(TXHOC)
Avg^ Max
18.3
1.20
.20
2.01
.01
.10
.70
a
.03
a
.32
.03
2.30
.02
.49
1.1
1.33
Elements 4
Ions
4
-
(37.2 )
( 2.14)
( .29)
( 3.25)
( .02)
( .51)
( 1.99)
( a)
( .05)
( a)
( .64)
( .04)
( 4.02)
( .07)
( 1.65)
( 2.7 )
( 2.00)
3
6
6
aNot detectable
y-9
-------
TABLE 9.1.3
AVERAGE AND MAXIMUM CHEMICAL CONCENTRATIONS
IN EL PASO, TX (10/79 - 9/80)
(Data taken from USEPA, 1981a)
HIVOL CHEMICAL CONCENTRATIONS (ug/m3)
Sampling Site
Chemical
Species
Mass
Cr
Cu
Fe
Pb
Mn
Ni
Ti
V
805
NOo
(1)E1 Paso
(TXELA)
Avg Max
115.2
.0047
.37
1.60
1.16
.07
a
.02
a
5.7
1.69
(319.8 )
( .0089)
.80 )
( 2.44 )
( 2.27 )
( .15 )
( .006 )
( .03 )
( .015 )
( 8.4 )
( 3.04 )
(2)Clint(B)
(TXELB)
Avg Max
86.7
.0042
.46
1.05
.18
a
a
.02
a
4.3
1.13
(240.2 )
( .0089)
( 1.37 )
( 2.67 )
( .59 )
( .09 )
( .0006)
( .06 )
( a. )
( 8.5 )
( 2.59 )
No. of Observations:
Mass 43
Elements 9
Ions 9
40
40
16
aNot detectable
y-io
-------
TABLE 9.1.3 EL PASO, TX (Continued)
FINE CHEMICAL CONCENTRATIONS (ug/m3)
Sampling Site
Chemical
Species
Mass
Al
K
Ca
Mn
Cu
Cl
Ni
Ti
Br
Fe
Pb
Si
Zn
S
SO £
NO-j
(1)E1 Paso
( TXELA)
Avg_ Max
23.0 (
.41 (
.23 (
.83 (
.01 (
.05 (
.43 (
a (
.02 (
.16 (
.17 (
.74 (
.52 (
.14 (
.94 (
2.3 (
.47 (
137.3 )
.60)
.50)
1.21)
.02)
.10)
1.63)
a)
.02)
.43)
.33)
1.59)
.73)
.35)
1.65)
4.1 )
2.18)
(2)Clint(B)
(TXELB)
Avg Max
12.3 (28.
.32 ( .
.16 ( .
.48 ( .
.01 ( .
.02 ( .
.08 ( .
a (
.02 ( .
.03 ( .
.12 ( .
.15 ( .
.50 ( 1.
.03 ( .
.96 ( 3.
2.6 ( 9.
.54 ( 1.
9 )
53)
45)
76)
01)
04)
20)
a)
02)
11)
28)
45)
50)
12)
43)
2 )
55)
No. of Observations:
Mass
Elements
Ions
27
6
7
38
15
16
aNot detectable
9-11
-------
TABLE 9.1.3 EL PASO, TX (Continued)
COARSE CHEMICAL CONCENTRATIONS (ug/m3)
Samp1Ing Site
Chemical
Species
Mass
Al
K
Ca
Mn
Cu
Cl
Ni
Ti
Br
Fe
Pb
Si
Zn
S
SO 4=
N03
(l)El Paso
(TXELA)
Avg Max
46.6 (
1.87 (
.50 (
6.05 (
.04 (
.07 (
.24 (
a (
.07 (
.04 (
.92 (
.21 (
5.24 (
.07 (
.33 (
.7 (
.47 (
151.7 )
3.82)
.92)
14.12)
.08)
.12)
.54)
a)
.14)
0.12)
1.74)
.54)
9.79)
.18)
.52)
1.4 )
1.17)
(2)Clint(B)
(TXELB)
Avg Max
45.8 (
2.01 (
.53 (
2.86 (
.02 (
.03 (
.20 (
a (
.05 (
a (
.67 (
.05 (
4.93 (
.02 (
.28 (
.8 (
.45 (
141.8 )
3.69)
1.07)
5.59)
.03)
.11)
.38)
a)
.13)
.02)
1.27)
.15)
10.33)
.04)
.72)
1.5 )
.87)
No. of Observations:
Mass
Elements
Ions
aNot detectable
27
6
7
38
15
16
9-12
-------
S5 rH W
8 ' . §
•S S
*T* rrt W
i £ " g
SM" g
s s
ij a
v.^ X
O [z]
O.X
lj Bu
r «
c^ sc
Ij X
o> cu
J_f
OS CO C*4
vD CO CO
rH OS OS
•* ^o
o o o oo ^
CM CO
o c
0>
MH w g M
o w
OO
O
z
«
9-13
-------
^
T3
V m
3 °1
C <•
2 M
= 3
0 ^
u w
•* °
£ B «
<; ±j
a 1 s
S i »
W Z -H
9 8 a
3 ^ 1
£ 5 w
* i
-J 5
* w
S g
g
CO
^
u £
M i-
&s
si i
I. -5
o
1
a u
^ a.
0) <
><
m >
01 «S
a.
^
K X
£ S
o a.
^ £
4J >
tn <
N
X
01 es
eo PM
01 •*
s.^-11
T-H
"S X
O TO
U ^S3
60 S
• Oj
0 ^ M
^.^
-»
rH
rH
TO CO
CJ QJ
6 U
01 01
u tn
m
o\
0
i-H
CM
m
CM
rH
O>
O
m
\o
o
o
\-^
rH
0
m
^^
rH
CM
CO
rn
CM
/-\
O
vO
m
en
CM
(0
to
as
** f^ ^" (M^jorrir^mvDfHro
o
-------
<->
"X3 ^"*
01 ro
3 6
C ^
•H M
4J 3
c ^^
O
u tn
•-- 2
O
< M
a, H
< 01
cd u
< H -H
M Z C/)
£ 8 M
J Z C
W O -H
O U rH
< cx
3 J E
rH ^1 Cl
a cj «
CM M
X
-* u
• K
,-1 U
a* u
w
u o:
>J <
S 0
s °
/^N
CQ
"^^
*J
^ *^x j
o u
CL £
v-< CM
3S
u
2
o"
rH
01
o
n:
C /™v J
n u
•H sc
t-< cu
01 <:
4J ^
>< ^-
•O 1
to >
01 <
L-i
CM
i— I
iH
CO X
•• /™N n
££s
o £
•-) <
£
• v^ 0
•U >
W . R
iJ ca
x: K
&OCM
^x <;
i— i
3
0 X
Is*
^j ^
X
• ft.
03 £
O v- ' GO
O >
^ <3
rH
"^
n |
to co
U 0)
VH T-l
E U
01 01
^ a
u w
eMOr^rrJfsi*jpcNCMQO^oooomi-fOomo'-tosr ,£
>-i 03
QJ ^j
CO CO (J
O C 4J
oj q>
y-i to E to *o
o to o> c
* « rH O 4-t
w lh*|rn • S w M o
^^^cS^acsss^ssj^^gg i ?
9-15
-------
in
•
rH
•
o>
a
5
H
/->
O
H 00
< W
CO
K 1
o
M ON
H r*-
<• -~^ /-N
> O «
C^ rH rH
w *~? oo
CO ON
PQ >H rH
0 B5
•t
EH « 3
M
P3
(u
^J
OOOO OOOOOOOO
r-»oooooooo oooooooooooooooo
**^ *^^ ^1*^ *^^ *^fc^ "^^fcl ^**. "^H. ^^'N* *^^l *^^» "^^l ^^*
OOONinrHr- COONinrHVOCMOOO
rHOrHCNO rHrHCMCOOrHrHCO
QCMcMtNOO OOOOOOOOO\ONONO\
rHOOOO OOOOOOOO
^N
CM
CM
•
-*
****
ON
m
•
r*"-
rH
s
•
o^
CO
rH
^^
00
•
rH in
CM m
**~\
vO
«
•vf
-S rJ
01
• (0
60 .0
> O
*
<-^
in
rH
•
CO
•^s
O
r»
•
•*
rH
s^.
!_)
•
r~
rH
^^
ON
in
•
m cs
<}•
^^
in
cS
^— "
oo
oo r-
co CO
CO
C
0
•rH
4-1
n)
60 >
> M
CO
•O B rO
0) 3 O
•u 3
o -i-i m
OJ X 0
r4 ClJ
vj s; •
O — ' O
U S3
9-16
-------
These thirteen sampling days were singled out because they are
internally inconsistent; the TOTAL concentrations are greater than the
HIVOL concentrations, in many cases (8/7/80, 8/13/80, 8/19/80,
8/25/80, 8/31/80, 9/6/80, 9/12/80, 9/30/80) by over a factor of two.
The aluminum concentrations measured on 8/7/80 and 8/31/80 are
unreasonable for a rural site such as Big Sister (they are
unreasonable at any site). The effect of removing these 13 records on
the average and maximum concentrations is seen in the final rows of
Table 9.1.5. Though the HIVOL average and maximum do not change by
much, the COARSE mass and aluminum average and maximum drop
precipitously and the FINE mass average and maximum are considerably
different. The averages and maxima for Big Sister which are reported
in Table 9.1.1 are the averages of the data set without the 13 records
of Table 9.1.5.
Other internally inconsistent records were observed in the data
sets from which the statistics in Tables 9.1.1, 9.1.2, 9.1.3 and 9.1.4
were derived, but they were not as extreme as those evidenced in
Table 9.1.5 and corrections were not made. Significant biases in
these summaries could exist, however, and some additional data
validation criteria to remove grossly inconsistant values need to be
applied to mass and chemical data acquired by the IP Network before
statistical summaries are produced. Summaries should not be produced
or used without a prior examination of the individually flagged data
records.
Most of the individual chemical concentration measurements from
which the averages were derived were made on the same days, so there
should be consistency between HIVOL and TOTAL (sum of FINE and COARSE)
measurements. HIVOL measurements, eg. Mass, Fe, Pb, Mn, Ni, SO
and NO , should be greater than or equal to TOTAL measurements of
the same species. For those species which are known to be
concentrated in the fine particles, Pb, Ni, SO, and NO.,,
HIVOL concentrations should not exceed TOTAL concentrations by a large
amount.
For Mn, Ti and Ni, many measurements are at or near their
detection limits, so the sum of FINE and COARSE concentrations of
these elements and the HIVOL concentrations will experience large
uncertainties. The HIVOL and TOTAL concentrations are normally within
factors of 2 or 3 of each other, which is to be expected.
9-17
-------
The possibility of a bias to Cu concentrations due to the. HIVOL
exhaust was raised in Chapter 3. This would cause HIVOL and SSI Cu
concentrations to be higher than TOTAL concentrations. This is borne
out by the Cu measurements. Average HIVOL Cu concentrations are
factors of 2 to 5 times higher than corresponding average TOTAL Cu
concentrations at most sites. At Clint, Texas, this approaches a
factor of 9, but because the TOTAL Cu concentration is so close to the
lower quantifiable limit there is a large uncertainty associated with
its measurement. These discrepancies indicate that some of the HIVOL
exhaust may be recirculated through the sampler.
Pb and Fe average concentrations on HIVOL and TOTAL samples
generally agree with each other. In a few cases, notably Big Sister
in Buffalo and 500 S. Broad in Philadelphia, the HIVOL Fe and Pb
average concentrations equal approximately 60% of the corresponding
TOTAL concentrations. These averages come from different sample sets,
however, (only 5 samples are contained in the TOTAL averages while the
HIVOL averages used 98 samples at 500 S. Broad) so the difference is
probably not due to biases in the measurement process.
The sulfate and nitrate concentrations on HIVOL and SSI samples
were shown to possess a possible bias due to adsorption of sulfur
dioxide and nitrogen dioxide gases by the filter medium. The
comparison of HIVOL and TOTAL average sulfate and nitrate
concentrations in the four cities bears this out. Presumably most of
the variability between HIVOL and TOTAL sulfate and nitrate
concentrations is due to samples taken in 1980. The average HIVOL
sulfate is normally 1.5 to 2 times higher than the TOTAL average and
the average HIVOL nitrate is 2 to 3 times higher than corresponding
TOTAL averages. These results are consistent with the artifact
hypothesis.
On FINE and COARSE samples, sulfur was measured by x-ray
fluorescence and sulfate was measured by automated colorimetry. If
most of the sulfur is in the form of sulfate, then the sulfate
concentrations should equal approximately three times the sulfur
concentrations. For all FINE samples the SO^/S concentration
ratio varies from 2.5 to 3.2. Since the average S and SOT
concentrations were not necessarily derived from analyses on the same
9-18
-------
set of samples, this agreement is reasonable and lends credibility to
both the XRF and colorimetric measurement methods. Several of the
COARSE SO^/S ratios are closer to 2, notably those at NYBUA,
NYBUB, TXHOA, TXHOC, and TXELA. This could be due to sulfur in the
COARSE particles which is not in the form of sulfate, and given the
good agreement between analytical methods for the FINE sulfate and
sulfur averages, there is no reason to suspect a measurement
difficulty.
The Cr and V concentrations measured on the HIVOL samples place
an upper limit on what should be expected on the FINE and COARSE
samples. X-ray fluorescence is perfectly capable of measuring Cr and
V on the FINE and COARSE samples at no additional analysis cost, but
these concentrations are not included in the IP Network data
summaries. They should be reported for intercomparison with the
corresponding HIVOL concentrations and because, as Table 4.2.1 shows,
Cr and V are important constituents of steel making and residual oil
combustion emissions, respectively.
This check of external consistency is by no means a substitute
for sample by sample validation and flagging of individual
measurements. These can only be done as part of the normal validation
procedure, however, and cannot be attempted here. In general, the
average chemical concentrations of HIVOL, COARSE and FINE samples from
the urban areas under study are within the ranges to be expected from
such samples. They should be used with the following caveats:
• Ti, Mn, Ni and Cu concentrations are often at or near their
lower quantifiable limits (approximately .01 ug/m^) where
their uncertainties will be large.
• Cu concentrations on HIVOL samples may be positively biased
due to exhaust recirculation.
• HIVOL sulfate and nitrate values may be positively biased
due to the adsorption of sulfur and nitrogen containing
gases by the filter medium. This is probably more prevalent
for samples taken in 1980.
With these observations about the validity of IP Network chemical
composition data in mind, it is possible to study the relationships
between sources and receptors in some representative urban areas.
9-19
-------
9.2 Urban-Scale Chemical Compositions and Possible Source Types
Three comparisons of the average chemical concentrations
presented in Tables 6.1.1 to 6.1.4 are important to gain an
understanding of the chemical character and possible sources of the IP
concentrations in urban areas:
• Comparison of the chemical compositions of the FP and CP size
fractions.
• Comparison of the FP and CP chemical concentrations at. urban sites
with those at the urban-scale background site.
• Comparison of typical urban concentrations in one city to those in
another city.
In the process of making these comparisons, reference will be
made to the maps locating sampling sites with respect to sources in
Chapter 4, to the speculations advanced there about which chemical
concentrations might be elevated by contributions from nearby sources,
and to Table 4.2.1 which relates the particle size and chemical
composition of the aerosol to sources types in each urban area. In
this way differences between chemical concentrations for different
size ranges, sampling locations and cities can be related to
differences in the size, location and city of source emissions.
With regard to particle size, all sites in all of the cities
studied show the Al, Si, Fe and Ca concentrations to be significantly
s
greater in the CP fraction. Similarly, the Pb, Br, S, SO,,
concentrations are highest in the FP fraction. Referring to the
dominant particle size and chemical concentration columns of
Table 4.2.1 shows that the CP species are common to geological
material while the FP Pb and Br are associated with leaded auto
exhaust. Chlorine concentrations are highest in the CP fractions in
Buffalo, Houston and Philadelphia. The exception is in El Paso,, TX
where the average FP Cl is nearly twice the CP Cl. CP Cl could result
from sea salt or the salting of roads in the winter. The FP Cl, as
Table 4.2.1 shows, might be associated with vegetative burning.
Average snowfall in Houston (Table B.I of Appendix B) is only
.4 inches/year, so road salting is improbable there, but it is close
9-20
-------
enough to the Gulf of Mexico to receive a marine influence. Buffalo
is too distant from the ocean for sea salt to be a large contributor
to CP Cl, but its snowfall is sufficient to require extensive salting
of roads. Philadelphia's CP Cl could receive influences from both
source types. In El Paso, neither the marine nor salting (annual
snowfall is only 4.7 inches) would be major contributions to CP Cl,
but the large quantity of wood, trash and brush burning in neighboring
Juarez could contribute Cl to the FP fraction.
The FP and CP fractions of Cu, Mn, Ni, Ti, and K concentrations
tend to be equal at most sites. As noted in the previous section, Cu,
Mn, Ni and Ti concentrations are often at or near the detection limit,
which is consistent with this equality. Nitrate and zinc
concentrations tend to be nearly equally distributed between the FP
and CP fractions or in the FP fraction, depending on the site.
With respect to comparisons between sampling sites within each
urban area, chemical concentrations in both FP and CP fractions at
urban sites often show enrichments over corresponding concentrations
at the urban-scale background sites. Thus, local sources are likely
contributors to most of the chemical species measured.
In Buffalo, NY, the two urban sites show average FP chemical
concentrations which are nearly equal to each other. The Al, Si, K,
Fe and Cl concentrations at P.S. 26 are 1.5 to 2 times higher than
they are at P.S. 28. Reference to the site survey summaries in
Table 3.1.3 shows that moderately traveled roads are closer to P.S. 26
then they are to P.S. 28. Higher FP and CP Pb and Br levels at
P.S. 26 might be due to more auto exhaust from cars on these roads.
The elevated CP Cl suggests resuspended road dust with salt in it.
Both FP and CP Fe concentrations at these sites are 5 to 8 times
higher than background, consistent with the proximity to steel mills
noted in Chapter A. FP and CP sulfate and nitrate at P.S. 26 are also
higher than those at P.S. 28 with no obvious explanation. Because
vanadium data are unavailable and Ni concentrations are so close to
detection limits, the effects of residual oil combustion proposed in
Chapter 4 cannot be discerned from these data.
In the Houston urban area there is some enrichment with respect
to the urban-scale background site for most FP chemical concentrations
at CAMS8; Al, Fe, Zn, Pb, Br and sulfate show elevated levels. In the
9-21
-------
CP fraction Al, Cl, Fe, Pb and sulfate are higher at the site closer
to the city. Oddly, several concentrations, Si, Ca and Cu, are higher
at the background site. The number of samples from which these
averages were calculated is small for both sites, so they may not be
representative. At CAMS8, the CP Al concentration is nearly double
the Si concentration, which is unusual for the geologically dominated
CP fraction. The maximum Al concentration, 7.7 ug/m^, is higher
than the maximum at any other site, and because only four samples were
included in the average, it raises the average substantially. This is
another example of the need for validation of individual data items
prior to the calculation of averages. The elevated Fe levels may be
due, in part, to the nearby steel foundry, while the Pb and Br are
most likely contributed by automobile traffic along the ship channel.
In El Paso, Pb and Br are elevated at the urban site, most likely
due to auto exhaust, though a portion of the Pb may be due to the
smelter. Ca, Cu, Cl, and Zn are also elevated in the FP fraction.
The Cl, as noted earlier, may result from vegetative burning while the
Zn and Cu may come from the smelter. In the CP fraction the Ca
average is very high, but the small number of samples gives the high
maximum concentration, 14.1 ug/m , more weight than it deserves;
this appears to be another example of the need for data validation.
Pb and Zn are higher, possibly due to the smelter, but most other
urban CP concentrations are similar to those at the background site.
Of the four cities studied here, El Paso is the least industrialized
and the comparison between urban and background chemical
concentrations reflects that.
In Philadelphia, both the NE Airport and Presbyterian Home sites
are in residential areas; they are not true background sites. FP Pb
and Si are high at 500 S. Broad with respect to the other sites. Pb,
and the associated Br, once again are probably due to auto exhaust.
The FP Si maximum of 2.4 ug/m? is much higher than the maxima at
other sites and in other cities; this biases the average. In the CP
fraction Al, Si, and Ca are elevated at 500 S. Broad with respect to
the other sites. Resuspended dust from traffic is consistent with the
site survey summary in Table 3.1.3 and with the elevated Pb and Br in
9-22
-------
the FP fraction. Reference to Figure 4.2.7 shows no major industrial
sources close to any of the sites and this is consistent with the
uniform average concentrations among the three sites.
With regard to comparisons between cities, several differences
can be noted. It is helpful to perform this comparison with respect
to a fixed reference as well as among the cities included in this
study. Tables 6.2.1 and 6.2.2 have been compiled for this purpose.
Table 6.2.1 lists averages of the average chemical concentrations at
IP Network sites reporting over 5 measurements. These averages are
classified by site type and size fraction. Table 6.2.2 presents the
highest average found at any one of the sites listed in the last row
of Table 6.2.1. Observations similar to those which have been made
for the individual cities apply to the figures in these tables.
Automotive-related elements and sulfate are in the FP fractions while
geologically related elements, Al, Si, Ca, Ti, and Fe tend to be
higher in the CP fraction. With the exception of the rural
classification, where all concentrations are lower, the CP and FP
average concentrations are relatively constant for all site types.
This is to be expected since the site-type classifications do not
consider specific sources affecting individual sites; these sources
can vary considerably for different sites even though the site types
may be the same.
The FP chemical concentrations at the urban sites in the four
cities are generally similar to those of urban and suburban sites in
Table 9.2.1 and generally less than the extreme values of Table 9.2.2,
except where individual high values at a site bias the average, as has
been previously noted. FP Fe is higher than the other cities in
Buffalo, which is consistent with the large steel industry there, and
FP Pb in El Paso is elevated with respect to Table 9.2.1 and other
cities, which might be due to the smelter. FP sulfate concentrations
in El Paso are significantly less than those in Table 9.2.1 and the
other cities, which is consistent with the regional-scale nature of
this aerosol component. El Paso receives virtually no transported
pollutants whereas the other cities are subject to regional-scale
sulfate.
9-23
-------
^
CN
ON
a
9
p
S
en
>4
CO
o
u
b
M
H
en
en
z
2
fM
z
1
8
j
5
M
£
u
e
g
<•
u
M
H
I
H
M
s_
BO
• <^
BO
T3
U
to
U
• p4
•o
c
• f^
-C
JJ
u
C0
n
60
«
£
JJ
IW
O
Q
60
03
M
0»
>
«
0)
Ui
Q
to
e
o
• *4
JJ
(0
c
u
u
c
o
y
11
09
U
.C
4J
m
^B
^
c
to
C
O
.-J
u
— *
^
i CM
CO 0
3
T)
C
1
§ CM
1 b
|
C
4
.O
S 0-
f\ tf\
•§ H
CO
^H PM
« U
1-1
M
U)
1 B-
M U.
1
C
10
1 ft.
•§ i£
en
2 B-
U U
4J
CO
3
0
C
*4
I CU
c £
o
i
5
j3 CM
U CO
D P
PL,
1-4 O
<0
^J
u
•S a.
£ U.
I-M
1
C
a
ji
U CM
D tn
c
0
• ^
« 4J
g. S^
>^ 14 to 10
E- b. u v
•1-4 •*<
o> 41 e u
U N 4) 41
.rf "J J= (
en w O en
O^O\O-l « ^H o\
"" O * en o
• 1 1 | 1 1 1 | « . d . to . 1 I 1 .
s °
fnmoNo^^-min CMO CM^CM^O
*no-*mr-.cMCMo O\o o^oo^
ON —t CM* ^
<-HONONesjCM^\oeM «-*m cnsO^^eM
incn^^-cMr-jmo O^-* oO'**®-*
CM
m »-H en r*. ^4 ^ psj
O* o O CM O en m
•tf -4
rv.
OmcMvncM.— lONin CMCM ON^m**.-!
r-HincMfnr-4fMmo Or* o O o O ,_(
ix »-v m m
CM
in
Of^r^v"imf*coeM en^- \r\ r* r* 9* a\
m-*sOf*.'-'r-*sOo Om o^1™'®^
d PM '
en
m -H oo ^-i m •*
^o o o oo -* -*
• 1 1 | 1 | 1 | OJ . . . U£buKUNefiCA&4
o NO
co m
O «N
so CM
en »-t
s3- O
r^. r-
m •-'
en m
DO O
<-4
CM r*
ON CM
^- F-4
in m
oo f*-
r- -*
O oo
r* so
* .*
O «M
CM in
00
in en
os r*.
d ^
1-1
in -tf
^* r-
i~4
i—t f— (
sO 1**
in
ON ON
r-* o
ON en
in m
O ON
r~4
tf\ 00
t^ -4
r> •*
o •*
•-i O
< < pa
£33
hi •< U
-< u H
o z H
« •> « M
•« "
^Eo33§ £55
~WOt-iiJp B^r^iJ
tn ta X en CM en u
-< •< Xo ef «ss x
OuzzaeouoH
S4<*
acts en
C Ql OJ
.5 •§ M
— 3 <«
a-j n
•< g o eu
S c >
en 1-1 <
CM
u
14}
b PM
U
^i
i2 *"
CM
« CK
to ^
•" ,2
U-4
i— 1
(8 (0
U U
g -3
cy c
•-H O
CU «H
u eu
S CO
CO CO
J= £
u u
0 0
B C
O O
•S -o
U 10 M
-j 41 11
10 — i — '
U 10 CO
u
U M M
UBS
Q> •-! •-<
•O »-( rM
a. a
u e e
O CO to
z co en
CO JO 0
9-24
-------
OJ
l-t
-D
a
H
C
U
Ph 0)
>< X
H ^
CO O>
e
Q C
U O
H 01
en oo
co
co h
z oi
o >
H OJ
W Z X
J Id 00
cq o .H
01
X
o
u ••*
O 4J
> d
< 0)
Id
X
O
C
o
u
3
e
i
O OLI
Z b
I
rt
-H CM
« O
I
C
o
z
1
e o<
« en
A H
3
•o a.
e t
^
h a.
a —i
u o CD
fa o
a) e _
N 01 01
•rJ X »•
en u en
CN r^ O CM
II • • CO • . «
OCNOOOOO^H O
OOO OOOOCM
oo^r^-3-cvli^iri mo^r^r^omcM &
^j-u~is^-r«s^-cMO OcMOO>—'cMOoo "^ 00
,-4 roesoncsim ON CN
• o o *•"' r^- o r~ oo *^ ^
& | I I 1 I I I I f I •
-3FmcN CN
^- vD r*. •—it—i *D O Oen OCN^-IO*^ ^ in
'-iOr--
^- so en oo r- CN T—<
-tflllllll-^^^-^lll
^-(inincNcNcNm ^>oo
'^""^^^.•1 I ••
in r- 1 CN i— 1 r
oo i I i I I I I
9-25
-------
With respect to the CP fraction, the geological chemical species
dominate it at all site types in all cities. As would be expected,
because of its larger particle size, the source contributions to
chemical species in this size fraction should be more localized in
nature, and consequently, more variable between sites.
Many other individual comparisons can be made, but a sufficient
number have been made here to have addressed the source-oriented
hypotheses advanced in Chapter 4. In nearly all cases, the average
concentrations measured at the urban and background sampling sites
support those hypotheses. At least some of the potential sources of
FP and CP in the selected urban areas have been identified. It
remains to quantify their contributions to ambient concentrations and
to determine whether or not other sources must be sought to account
for the mass concentrations measured at each site.
9.3 Urban-Scale Source Contributions
The chemical element balance receptor model was used to estimate
the contributions of various sources in Buffalo, Houston, El Paso and
Philadelphia to the FP and CP average mass concentrations presented in
Tables 9.1.1 to 9.1.4. The model has been thoroughly explained
elsewhere (Watson, 1979) and will not be elaborated on here.
The chemical element balance model, like any other model, is only
a simplified representation of reality. The results it produces,
therefore, must be considered in conjunction with its limitations.
Its greatest limitation is the inadequacy of the input data on which
it functions. These data include:
• The types of sources which could contribute to a receptor.
• The fractional chemical compositions of identified source
types in the desired size fractions and valid estimates of
their uncertainties. This fractional composition must be
that which is perceived by the receptor, not necessarily
that which is measured at the source.
9-26
-------
• The chemical concentrations at the receptor in the desired
size fractions and valid estimates of their uncertainties.
These chemical species must be the same as those of the
source emissions and the number of such species must exceed
the number of source types.
The potential source types in each urban area were identified in
Table 4.2.1. One difficulty which was noted in the discussion of this
table was that many of these source types lack chemical species or
other properties of their particulate matter emissions that allow them
to be differentiated from each other. This effectively groups sources
with similar compositions into the same source types. Some of the
source types in Table 4.2.1 contain organic and elemental carbon as
major constituents, and these species are not quantified on IP Network
samples. Similarly, vanadium and chromium, which are species common
to residual oil combustion and steel production, are not included in
the IP Network data summaries. Sodium and ammonium are two other
measurements which would aid in the attribution of receptor
concentrations to marine or road salt and secondary aerosols,
respectively; they are not routinely measured in the IP Network.
The source compositions of source types identified in Table 4.2.1
which were used to perform the chemical element balances are
identified in Table 9.3.1. Most were taken from Watson (1979) because
this is one of the few resources which reports most of the important
chemical species in the FINE and COARSE size ranges with estimates of
their uncertainties. Crustal shale, crustal limestone and steel
industry compositions were taken from Dzubay (1980). Dzubay reports
no uncertainties so arbitrary precisions of ^10% were assigned to each
of his values. He does not segregate source compositions by particle
size so the same values were used for chemical element balances on
both the FINE and COARSE fractions. Because smelting was identified
as a potential contributor in El Paso and Philadelphia, chemical
analyses of copper smelting emissions from Zoller et al (1980) were
adopted for the FP and CP fractions; the values used and the way they
were calculated are included in Table 9.3.1. Unfortunately, no source
compositions for lead smelting operations were available.
9-27
-------
TABLE 9.3.1
SOURCE TYPES USED IN CHEMICAL ELEMENT BALANCES
Mnemonic
Source Type
Reference
CONTDUST
DZCRUSTA
DZLIMEST
PAPHSOIL
URBNDUST
Continental Dust
Crustal Material
Limestone
Soil from Philadelphia
Urban Dust
Watson, 1979
Dzubay, 1980
Dzubay, 1980
*
Watson, 1979
ALUMPROC
DZSTEELI
MARINE
PAPHMN
PBAUTOEX
RESIDOIL
FERROMAN
VEGBURN1
VEGBURN2
ZOLLERCU
Aluminum Processing
Steel Industry
Marine
Mineral Industry
Leaded Auto Exhaust
Residual Oil Combustion
Ferromanganese
Vegetative Burn 1
Vegetative Burn 2
Copper Smelting
Watson, 1979
Dzubay, 1980
Watson, 1979
*
Watson, 1979
Watson, 1979
Watson, 1979
Watson, 1979
Watson, 1979x
Zoller et al, 1978*
ELEMENTC
ORGANICC
NITRATE
SULFATE
*See next page.
Elemental Carbon from
unidentified sources
Organic Carbon from
unidentified sources
Nitrate from unidentified
sources
Sulfate from unidentified
sources
Single Constituent Source
Single Constituent Source
Single Constituent Source
Single Constituent Source
9-28
-------
TABLE 9.3.1 (Continued)
*SOURCE COMPOSITIONS UNIQUE TO THIS STUDY
PERCENT COMPOSITION
Chemical
Species
Mass
OC
EC
Al
Si
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Ni
Cu
Zn
Br
Pb
FINE
PAPHSOILb
100+10
3.1+2.4
9.1+4.3
.207.15
,53+_1.3
10+3.4
227_8. 5
.77+. 053
.327.44
1.4+. 71
3.0+3.2
. 33+^. 19
.014+". 014
.031+.033
.99^1.1
7.273.2
0082^.0054
.040+". 027
.187.19
00827.0052
.207.13
PAPHMNC
100+10
.61+6.4
4.6^1.3
.12+". 17
.257.35
3.7+3.8
5 . 2+2 . 1
5.2+". 12
.327.13
.737.081
.967. 021
.26+_.071
.14+". 07 7
.19+". 026
40_+13
6.3+_2.5
.010^.0063
.0307.031
.0467.042
.0025'+. 0064
.157.029
ZOLLERCUd
100+_10
a
a
a
60+_6.0
.86+_.78
a
20+^2.0
a
a
a
a
.0054+_.0044
a
.0014+_.0012
a
a
.61+5.0
6.6T9.5
.00627.0038
. 94~1 . 2
COARSE
PAPHSOIL
100+10
1.47.58
1 . 57. 69
.0377.015
.3U.19
3.7+1.5
12+_4.5
. 257. 14
.027+. 029
. 607. 20
2.27l.3
. 35;+. 10
00857.01
.027+. 038
.3S+.62
5 . 37l . 9
0015^.0013
.0597.084
.093+. 094
0015+.0013
.0447.050
PAPHMN
100+10
1.17.26
2.47.37
.067.021
.20+_.082
2.2+1.6
3.8^1.0
.40+. 052
.20~.012
.407.087
.60+_.068
.30+_.0072
.10+. 017
.10+. 001
23+4.3
4.77.002
.OU.01
.03+. 014
.30+. 006
.00137.006
.107.0064
ZOLLERCU
100+10
a
a
a
60+_6.0
.86^.78
a
20^2.0
a
a
a
a
.0054+_.0044
a
.0014+_.0012
a
a
6.1+5.0
6.6+9.5
.00627.0038
. 947l . 2
aNot detectable.
bAverage of resuspended road dust samples taken from NW of NE transfer site, NW of
NE waste site and SW of Allegheny Site.
C0ne soil sample taken around the manganese ore facility and 3 baghouse samples
from the manganese ore facility in Philadelphia.
dAverage of 5 samples from Zoller et al (1978) with sulfur concentrations
arbitrarily set to 20% for each sample.
9-29
-------
Source samples of a manganese ore handling operation and street
dust were taken in Philadelphia as part of this study. These samples
were dried, seived through a Tyler 400-mesh screen which eliminates
particles with nominal diameters greater than 38 urn, and resuspended
for sampling with the SURE cyclone pre-separator, which was shown to
yield a sample equivalent to the FP fraction, and an open-faced 47 mm
filter. These samples were taken on Fluoropore Teflon filters for XRF
and ion analyses and on quartz filters for carbon analysis. Four
samples from the manganese ore facility and four soil samples were so
analyzed and their fractional chemical compositions were calculated.
The average and standard deviation of the four measurements for each
chemical species were used for the source composition value and its
uncertainty within each size fraction. Because the mass concentration
of FINE particles was approximately one-tenth of that of the TOTAL
particles in these samples, the TOTAL compositions were used for the
COARSE compositions without correction for the FINE fraction. The
resulting source compositions of PAPHMN and PAPHSOIL are presented in
Table 9.3.1.
This selection of source compositions is admittedly inadequate
for an application of the chemical element balance receptor model in
four different cities. There is no reason to suppose that source
compositions measured several years ago in Portland, Oregon, (Watson,
1979), soil compositions from a geochemistry handbook (cited by
Dzubay, 1980), samples from five copper smelters in the southwestern
United States (Zoller et al, 1978), a 1971 estimate of steel industry
compositions (cited by Dzubay, 1980) and eight samples from two source
types taken one year after the ambient sampling used in this study
should closely approximate the source compositions perceived at the
receptors in Buffalo, Houston, El Paso and Philadelphia between 1979
and 1980. The uncertainties associated with the source compositions
specified in Table 9.3.1 are large, however, and it is hoped that they
are sufficient to account for their lack of specificity. There may be
significant discrepancies between the model results and the true
source contributions.
Since the chemical element balance is based on a linear
combination of source contributions it can be applied to average
9-30
-------
receptor concentrations to calculate average source contributions.
The model equation is
C., = Z a..S.. 9-1
lk XJ Jk
where the C., is the concentration of chemical species i in the kth
IK
receptor sample, S. is the contribution of the jth source to the
JK
kth receptor sample, and a. . is the fractiona
in source j. Averaging over n samples yields
kth receptor sample, and a. . is the fractional quantity of species i
_ 1 n , n P P i n p_
C. = - 1C., = - I Z a. .S.. = Z a. . - I S. = Z a-.S. 9-2
i n k=1 lk n k=1 j=1 U Jk j=1 U n k=1 Jk j=1 V J
where C. and S. are the average concentrations and source
contributions, respectively.
The precision of an average is equal to the standard deviation of
the data set used to calculate it divided by the square root of the
number of samples in it. The average concentrations presented in
Tables 9.1.1 to 9.1.4 with their associated precisions were used as
input to the effective variance least squares fit of source
contributions to the chemical element balance equations. This fitting
method propagates the uncertainties of the source compositions and the
receptor concentrations through the solution process to provide
estimates of the uncertainties of the source contributions. Because
of the small number of values in some of the average concentrations
and because of extreme values for some of the measurements used to
obtain them, the width of the precision intervals associated with them
might be underestimated. Thus, due to recognized limitations
associated with the source and chemical composition data available to
the chemical element balance receptor model, the source contributions
calculated here are meant to be illustrative rather than conclusive.
Each chemical element balance was performed using the
interactive, iterative process described by Watson (1979) using his
criteria for the best least squares fit of source contributions to the
9-31
-------
chemical element balance equations. The combination of sources
finally selected after various trials was that for which:
• The ratios of chemical concentrations calculated from
Equation 9-1 to the measured concentrations most closely
approached unity.
• The sum of the source contributions most closely approached
the measured mass concentration.
Source types in a particular fit were selected from the set in
Table 9.3.1 regardless of whether or not such a source type was listed
in Table 4.2.1 as existing in the urban area under study. This was
done for two reasons:
• Source types similar in composition to those included in the
fit but not included in Table 4.2.1 may exist. Such sources
should be sought, their true composition should be
determined, Table 4.2.1 should be corrected and the balances
should be re-run with the new source composition.
• If such source types do not exist in the area under study,
the fact that they appear to be contributors raises
legitimate doubts regarding the validity of the receptor
model input data.
Any source contributions which were less than their associated
uncertainties were removed from the fit. Normally all chemical
concentrations were included in the fit except sulfur, because it
duplicated the role of sulfate, and those below detection limits as
indicated in Tables 9.1.1 to 9.1.4. The single constituent source
types (Watson, 1979) in Table 9.3.1 are meant to show that portion of
the chemical species they represent which is not accounted for by
other source types. In the case of sulfate and nitrate this is an
upper limit on the aerosol which forms in the atmosphere from
emissions of sulfur and nitrogen-containing gases. For organic and
elemental carbon this represents the upper limit of contributions from
many sources, such as biological material and natural gas combustion,
for which carbon is the major, and sometimes the only, constituent.
Between five and ten iterations of different source type and chemical
concentration combinations were required before the best fit was
9-32
-------
achieved. These iterations were performed by a technically qualified
person whose only training in receptor models was that received at one
of EPA's receptor model workshops; the purpose was to simulate the
type of application which would normally take place for compliance
modeling purposes. The specific details of all chemical element
balances presented here can be obtained from the authors on request.
The results of the chemical element balances applied to the
average chemical concentrations at sites in each of the four cities
appear in Tables 9.3.2 to 9.3.5. The sum of all the average source
contributions and the fraction of the measured mass that they account
for are included in the final rows of the tables. For site types with
a full complement of IP Network chemical concentrations, 60 to 70% of
the average FP mass and 80 to 130% of the CP mass is accounted for.
The exception for the FP mass occurs at NE Airport in Philadelphia and
for the CP at Clint near El Paso.
The source contributions in these tables need to be examined for
their consistency with the known source types in each urban area
recorded in Table 4.2.1 and with the observations of Section 9.2.
The geological source types are most dominant in the CP fraction
at all sites in all areas. In general, between 50 and 95% of the sum
of the source contributions can be attributed to geological material.
None of the four geological source compositions seems to be preferred
by the fitting procedure at any particular site. Philadelphia's soil
composition fits the COARSE chemical concentrations in Buffalo and
El Paso better than it does those in Philadelphia.
Geological CP fraction source contributions at the urban sites
are invariably elevated over similar contributions at corresponding
urban-scale background sites, showing that some portion of the
difference must be coming from local sources. It appears that even
though size-classified sampling may eliminate some of the fugitive
dust contributions which have so often lead to violations of standards
with HIVOL samplers, geological material from a myriad of sources will
still be a major contributor to inhalable particulate matter
concentrations.
Unaccounted for sulfate is by far the largest contributor to the
FP fraction though its presence in the CP fraction is small in all
9-33
-------
CM
W
H
1 s "5
" bfi
WO 3
a .j c,
H^c ^^
^*4
fe & to
8 § 1
Z] 0)
co a C 4J
1 « S «
g E B .!
a a s i
aw « §
Q H a co
3 D
SB §
04 ^ w
O Prf fi
co o r
PLI
W M
r_b M *^
^ H
0 Q Q PL, C CO a
r-l 0
0 0
0) a)
c5 co
•j- oo r-i m r>- ^s
i1-0^,;^?, y, 5,
i— I • ON ,0 + | ,0 ,Q .0 +| +|
CO CO rH
vO • vO
• o
I-l
vo ^d* oo O m
CO O • CM -* • 6«S
• • i— 1 • • CM OO
+ I + I + I-H±I-Q +1 +1
C^ 00 ON «^ C3 CO i— 1
• i— i • vo in • ^o
CM • CM • • O
-------
^-v
CO
e
"^-^^
^ 60
•0 3
0) >— '
3
C co
•H Z
4-> O
C w
O H
0 £3
>_, pq co
M 4J
>H Orf -rH
Z H CO
z
3 8 ?
d >H
< W r-l
PL, o a
CK p«: e
ED E> CO
pa o co
co
CM
w
co co
P4
cy\ <
o
W 0
i 8
H <
K
is
5
0-
e
3
PL, cd
43 C5
J_l ^^
+J ^J
§e
r-< Z
•rl >— '
D5
/"-s
^
Q)
O Q)
b W
2 >^
0 H
CO
• • r-l
r-l r-l •
+ |43 43 +| +|43
CO vO CJN
• • m
f-- -^ •
r~ r^
CO CO O O
• • • •
+ |+|43 43 +| + |
CO CM
• r-i m r-i
+ 143+143 43434343 +| +|
vO -* O CM
-S- • -CM
CM ^O r-l
CO
CO
CO
£
T3
0)
r4
3
CO
cd
CO
X^J Mr-jCMDCJ &
W r-l J Z Z O O
OOWwMW^Kl ^
HOZWD&WPij 0
tai-ir-i HPqpQrJS
^jcooicoOOjS 4-1
pqw-gJcMWwOrJ •• C
PU«SQ>>N-< 1-1 cu
(0 rj
JJ £
O Q)
H CM
•
4J
•i-l
M-l
cd co
ti -^
Cd 4J
T3
C
JJ -H
suff icien
t needed
C 0
M Z
Ctf J3
9-35
-------
TABLE 9.3.3
AVERAGE SOURCE CONTRIBUTIONS TO
FINE AND COARSE PARTICUALTE MATTER IN HOUSTON, TX
AVERAGE FINE SOURCE CONTRIBUTIONS (ug/m3)
Sampling Site
Source
Type
Geological
DZLIMEST
PAPHSOIL
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
RESIDOIL
VEGBURN2
ZOLLERCU
Total:
Percent of Measured Mass:
(1) CAMS 8
(TXHOA)
.75+. 29
2.0+.6
8.7+1.5
.37+. 08
1.3+.4
.18+. 10
1.2+.7
.60+.45
15.1+2.7
87+26%
(2) CAMS 1 (3) Seabrook (B)
(TXHOB)a (TXHOC)
.91+. 33
1.5+.. 4
5.8+1.8
.64+. 2
b
b
1.5+.8
.18+. 14
10.8+2.1
76+24%
AVERAGE COARSE SOURCE CONTRIBUTIONS (ug/m )
Sampling Site
Source
Type
Geological
DZLIMEST
CONTDUST
PAPHSOIL
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
RESIDOIL
MARINE
Total:
Percent of Measured Mass:
(1) CAMS 8
(TXHOA)
4.8+1.7
b
8.2+2.2
1.6+.10
1.1+.4
.34+. 10
.18+. 10
3.3+1.6
19.7+3.4
116+32%
(2) CAMS 1 (3) Seabrook (B)
(TXHOB)a (TXHOC)
6.1+1.6
6.0+.9
b
.88+. 36
1.3+.3
.14+. 03
b
2.0+.9
16.5+2.0
90+19%
alnsufficient data.
bNot needed in fit.
9-36
-------
TABLE 9.3.4
AVERAGE SOURCE CONTRIBUTIONS TO
FINE AND COARSE PARTICULATE MATTER IN EL PASO, TX
AVERAGE FINE SOURCE CONTRIBUTIONS (ug/m )
Sampling Site
Source
Type
Geological
DZLIMEST
PAPHSOIL
URBNDUST
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
VEGBURN2
ZOLLERCU
PAPHMN
Total:
Percent of Measured Mass:
AVERAGE COARSE
Source
Type
Geological
DZLIMEST
CONTDUST
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
VEGBURN2
ZOLLERCU
ALUMPROC
MARINE
Total:
Percent of Measured Mass:
(1) El Paso
(TXELA)
2.2+. 6
2.5+. 6
b
1.4+.17
.36+.32
3.3+. 9
3.7+2.0
1.0+.6
b
14.6+2.6
63+18%
SOURCE CONTRIBUTIONS
Sampling Site
(1) El Paso
(TXELA)
21.8+5.7
13.7+2.4
b
.46+. 16
.89+. 30
b
.94+. 21
b
.53+. 22
38.3+6.1
82+17%
(2) Clint (B)
(TXELB)
1.3+.2
b
2.6+. 3
2.3+. 6
.51+. 13
.65+. 15
1 . 0+ . 5
.33+. 24
.012+. 004
8.7+. 9
71+9%
(ug/m )
(2) Clint (B)
(TXELB)
8.8+1.4
b
.41+. 21
.42+. 12
.21+. 07
2.1+1.0
.31+. 25
3.5+1.2
b
27.0+_2.5
59+8%
alnsufficient data.
bNot needed in fit.
9-37
-------
m
•
en
Cft
W
1
pS
„
M
a
PM
W
•<
t-H
H- 1
sa
PM
Z
M
)V>
w
C_(
H rn"^
^ 4
60
H v5
tJ CO
§ §
M
» fa
?• CP
^
pq
4J
M
0 ^
p| M
•H pu
z ^
O
^^
-) PB
• 51
£ £
^••^
CM
V cQ
•a s
i) S
3 3
^^
•-I
T3
cd
Q
I-l
«
CO 33
o 51
0 &j
/^N
I— 1
^"^
cu
0 0)
M CX
§ ^
CO
•^ in vo ^* e?^ CM
cM«d" ejN 1-1 sf o en en »B>8
+*|^5 +| +*|+"| +'|+*|X> +I+*M3 +| +|
^^ ^^ f4 ^D OQ 00 ^^" 00 ^M *^^
CO • *.
r-l
vo o en •*
oo r^ CM oo • x
ja +|.a o o +|+|A +|A jo +| + 1
oo O en CT* • J £> PM •• (j
O C l-i i-l a)
•-I o 4-l
o
o
cd
^
o
Cd 4J
4J •!-( 0)
cd y-i 3
•d -d
C!
4J >i-l -d
(3 v
a) -d -d
•H (U 3
o "d 1-1
•t-l CD o
M-l CD £3
«w S .5
tD W 4-1
COO
cd ft o
9-38
-------
s~\
•n
w
1
H
s-\
CO
e
60
-j
N«X
CO
z
o
s
H
g
M CU
Oi 4J
H -rl
Z co
8 ,
a •*
CJ r-l
P4 a
8 §
co co
W
53
1
O
W
r h
1
5
«
co PH
0) <
H PH
n J '^dr'
tM >~^
-r-N
rH
x^
CO
—
C cd
J3 x-^
O r-l
•-} 33
• %
4J PH
CO v-'
x-\
CM
v-X
>-,
c
0) c«
ob to
a? S
_«| Qj
3 3
vo m o m
COin i-lr-l O 1-ICOO-* CTl &•?
• • •• • •§•• • QQ
+ 1 + 1 +1+1 +I-Q+I+I + I + I +1 +1
>d-co r-~r^- m r-i-*r~co r- o\
• • m*o vj- cMr^r-i« • r>-
r-im •• • • • o >~i o
• 1-1
r^
r-IP^. vOvO CTir-100 CM
incO CM r-l r-IO r-IOvO • S-S
+ 1 + 1 +1 + 1 +1 + 1-2 +1 + 1 + 1 +1 +1
^J- r-l VO l • * * * * CO r^
+ 1 + 1 o o +1 + 1 + 1+1 + 1,0 +| +|
o O m in ON m ^- r~ r-i
• • cyi«-iincy> ?J
4-1-1-10)
cd >4-i 3
•0 T)
C
4J -rl T3
C a)
0) -a T3
•rl CO 3
0 T3 r-l
•rl CO O
14H CD C3
>W t3 -rl
3
W 4J 4J
COO
H Z Z
* rQ U
9-39
-------
urban areas except El Paso. This is consistent with other
observations in the eastern United States (Mueller and Hidy et al,
1981; Dzubay, 1980). The FP sulfate is usually considered to be
regional in nature, though the contributions at urban sites in
Buffalo, Houston and Philadelphia are significantly higher than those
at the corresponding background sites, usually by a factor of 1.5 to
2. This could be due to chemical concentration averages calculated
from non-simultaneous samples at the different sites, some of which
might include more sulfate-rich days than others; it could also be due
to other sources of sulfate within the urban area which need to be
identified. El Paso experiences low unaccounted for sulfate
contributions even though the smelter is a major source of SO . No
regional-scale pollution episodes are known to affect El Paso as have
been documented for Buffalo and Philadelphia (Mueller and Hidy et al,
1981).
Unaccounted for nitrate is generally evenly distributed between
FP and CP fractions and is not substantially elevated over background
levels except at P.S. 26 in Buffalo where the FP nitrate is nearly
twice the CP nitrate and three times the background value. The
nitrate source contribution appears to be of a regional scale and of a
3
small (less than 1 ug/m ) magnitude.
The other source types appear to follow the expectations of
Table 4.2.1 with respect to dominant particle size and play
subordinate roles to geological material in the CP fraction and, with
the exception of El Paso, to unaccounted for sulfate in the FP
fraction.
Leaded auto exhaust is a familiar contributor at most urban sites
with contributions on the order of 2 to 3 ug/m . The PBAUTOEX
contribution in El Paso may be elevated over its true value by
accounting for some of the lead from the lead smelter; it will be
recalled that no source composition for lead smelters was used in the
chemical element balances.
The residual oil combustion contributions are generally very
o
small (less than .2 ug/m ) if present at all, but without a vanadium
concentration this source type is most sensitive to nickel
concentrations, and these are often at or below detection limits.
9-40
-------
The MARINE contributions in Buffalo, Houston and Philadelphia are
in the CP fraction, are elevated with respect to background sites and
3
are 1 to 3 ug/m of the COARSE mass. Because they are so much
higher at inland sites, the source is probably not so much marine as
it is some sort of salting activities. A source composition for this
source type should be obtained. There may be other urban sources of
chlorine in these cities which could also be contributors; it is
improbable that much road salt would be applied in Houston. Chemical
element balances for summer and winter averages might help to resolve
the issue. The small CP MARINE contribution in El Paso might be
mistaken for vegetative burning which is also sensitive to chloride
concentrations.
The VEGBURN1 and VEGBURN2 contributions are among the most
tenuous of all the sources. Watson (1979) expressed grave doubts
about the representativeness of the measurements made to obtain his
source compositions, and the only reason he included them in chemical
element balances was because they accounted for much of the organic
carbon measured on Portland's aerosol samples. The vegetative burning
source contributions are sensitive to chlorine and potassium
concentrations. At P.S. 28 in Buffalo, for example, a CP MARINE
contribution would probably be more consistent with other sites, and
3
reality, than the 12 ug/m vegetative burning contribution without
significantly deteriorating the quality of the fit. In general,
however, the vegetative burning contributions are concentrated in the
Fp fraction, where they belong, are of a reasonable magnitude, 1 to
3
4 ug/m , and are highest in Buffalo and El Paso, where heating by
wood burning is a likely possibility. The contributions from this
source type must be considered as speculative.
The industrial process contributions must also be considered to
be speculative. For the most part the steel, copper smelter, aluminum
processing and manganese ore handling source types are small
3
contributors (less than .5 ug/m apiece to either size fraction)
and, because their source compositions are usually dominated by one
chemical species, they are probably just filling in small differences
between measured and calculated concentrations for those species,
similar to the role played by the single constituent source types.
9-41
-------
The exceptions are in Buffalo, where steel industry contributions
3 . .
approach 1 ug/m in the FP fraction and where this industry is
expected to affect the urban sampling sites. Even if this
contribution is real, however, it is not a major contributor to either
size fraction and, consequently, to average IP concentrations.
Several individual samples from Philadelphia were chosen from the
IP Network archives for non-routine analyses. The purpose of these
analyses were twofold. They were intended to see whether or not any
advantage could be gained from analytical methods not: part of the IP
Network analyses, in this case carbon and optical microscopic
analyses, and to further test the validity of the chemical element
balance receptor model by applying it to individual samples taken
simultaneously at different sites. In order to perform carbon
analysis on HIVOL and SSI filters it was necessary to choose samples
from 1979, prior to the use of the S&S filter described in Chapter 3.
Unfortunately, sampling problems severly limited the number of
simultaneous samples available so that, out of the ten sampling sites
operating in Philadelphia during the latter part of 1979, only two at
a time yielded simultaneous dichotomous samples. Up to five sites at
a time yielded SSI samples. Sulfate, nitrate and carbon measurements
were made on HIVOL or SSI quartz fiber filters while elements were
measured on the FINE and COARSE Teflon filters. The sulfate, nitrate
and carbon concentrations were placed in the FINE size category
because their concentrations are usually highest in that fraction; in
reality these measurements are a combination of concentrations in both
the FP and CP fractions.
October 18, 1979 and November 20, 1979 were days with very high
IP concentrations in Philadelphia. The measured mass for each sample
in the final row of Table 9.3.6 shows that most of the IP mass was in
the FP fraction on these days. November 17, 1979 showed IP
concentrations which were within the typical urban ranges observed in
Chapter 5.
On October 18, 1979, the major contributors to the FINE fraction
were auto exhaust and unaccounted for sulfate. The auto exhaust
contribution was several times larger than that found for the average
source contributions in Table 9.3.5. The FINE geological, unaccounted
9-42
-------
OH
I s
£ Z
~ M
Z
o
m
B
no
3
Q
H
CQ
M
Qj
P
8
B
o
M
ON
ON
r-l
0
CM
£
O
Z
ON
ON
„
^-
•
|
ON
r-.
ON
<— t
00
.— i
.
U
o
O
U3
W
8
E
H
fa
A
8
u
H
to
bl
«
8
z;
b.
01
o
>
e-
(.
_•
o
to
U b (i)
z o s
£ Q.£
-H .J 0-
^ <
u
oi u
K 01 33
ft* 8 PUI
— , O i— ^
CM r-4
m • •
4- \-& + I ^ ^ + 1 w O U
O CM cn
^ Ov M3
CM
00 •
+ | + |.a JD ^ o o o o
-* o
CM
** S * J J
* •<—*,.
CM
m CM in in
CM m . o • •
• • "tf • i— 1 rH
• rM 0 . oo -»
fv.
e^-tf o
+ ( + |JD jo +| u u u u
000 CM
CM in
r»-
r*. \o o «A
+ 1 + (.c -O +( +| u u o
in ^H *»
p-t
m
I-H in o
JD A +|jD .A +1 +1 o O
*-i CM ^i
.
CM »n
CM - 0 00 ^
jQjD+|,OJ3 +| +|+| + |
CM CM in M
CM <J O U
COWU)W uj W O E^
f— i w & D o u Pf^^z
y i— t p 53 «c n ^ fa o£ ^ £
^OoBBfti-E Ctnzow
O 0
0) 01
U cn
m ,* t*
m *9 • • o f**
4 • rH CM —
~4 CM »-4 Csl
•*
CM O ON
CM ^ O
in
^ONSf cn
m • • o • M
. <~( r-l • IA 00 CM
O m ^5 »-< Q CM CM
CM O *M • *-* *-*
r*4 QQ
ON
ON 0 »4
O (^ *-* O m oo CM M?
+ 1 + 1 +• i + IJD +1 + 1 +1 r~>
. • . O • • ON
* r^ P-t
t— *
00 (N ^-i
o CM CM o • *e
• * . • t-i ON in
+ | + | + | + tja JD +| +| •
. • . O « f"-. CM
CM
ON CM m
o * oo O r- • M
i|Al±!ili|'° i1 i1 ^
.."to. • ^ CM
r-i r* w
,-. «A «n o • CM cn
^W1-*-0 i1 ^ s
CM
O ""* < -4 OJ 3 3
Q) (0 U CO 10
j: 4J u< nj «
J-> O CU (U (U
0 f- cu Z Z
,
C0
a
•0
o
u
a
1— 1
o
a u
jj .FH" 01
tj -o
C
c a>
01 TJ T3
•t-i ty 3
U T3 t-4
•r-l (U O
*W 0) C
3
09 U AJ
coo
n) jo w"
9-43
-------
for organic carbon, unaccounted for elemental carbon, residual oil
combustion, auto exhaust and copper source contributions were
uniformly distributed (i.e. the uncertainty intervals of each
contribution overlap) among the two sites. The manganese ore handling
contribution at NE Transfer was consistent with its location relative
to both sites in Figure 4.2.7. The FINE unaccounted for sulfate
contribution near Presbyterian Home was significantly greater than
that at NE Transfer on October 18, 1979; this source contributed more
to the FINE mass concentrations at Presbyterian Home than at NE
Airport on the other days examined and on the average, as shown in
Table 9.3.5. It is not obvious where this additional sulfate came
from. Since sulfate was measured on glass fiber filter media for the
individual days an artifact problem is possible, though samples taken
in 1979 should not be susceptible to this.
The large differences between unaccounted for nitrate and
vegetative burning contributions to the two sites is not consistent
with the area-wide emissions nature of these sources.
In the COARSE sample, contributons of geological material
dominated at both sites, though the contribution at NE Transfer was
much higher than the CP mass measurement there; since all chemical
species in the chemical element balance fit for this sample showed
relative concentrations which were consistent with each other, it is
suspected that the filter which was analyzed and its mass
concentration do not correspond to each other.
On November 17, 1979, the FINE sample at 500 S. Broad showed
higher contributions from geological material similar to that found in
Table 9.3.5. The unaccounted for carbon contributions were also much
higher. Residual oil combustion, copper smelting and manganese ore
handling contributions were low and did not differ significantly
between sites, which is consistent with previous observations that
they are part of the noise in the measurements. The auto exhaust
contribution at 50Q S. Broad should have been greater than that at NE
Airport but the FINE chemical element balances did not show this.
Once again, geological material dominated the COARSE fraction.
9-44
-------
On November 20, 1979, auto exhaust and unaccounted for sulfate
were major contributors to the FINE mass. The residual oil combustion
contributions were higher than any others observed at both sites; if
November 20 were a very cold day, this would make sense because of the
proximity of residual oil burning sources to these sites shown in
Figure 4.2.7. The COARSE fraction, once again, was dominated by
geological material.
The comparison of source contributions to simultaneous samples at
different sampling sites coupled with a knowledge of the geographical
relationships between source and receptor locations aids the
acceptance or rejection judgement to be made for each chemical element
balance. The examples in Table 9.3.6, limited though they are, show
that conclusions about measurement validity, unidentified sources, and
the real (as opposed to the estimated) lower quantifiable limits for
source contributions can be drawn from these comparisons.
The carbon measurements did not add much value to the balances;
the fractions of the FINE mass concentrations accounted for did not
increase significantly over those which were found without carbon
measurements. The use of carbon concentrations in this application is
crude, however, due to the lack of a filter medium for FINE and COARSE
samples which is amenable to carbon analysis.
The source contributions to the COARSE fractions of IP have not
exhibited great variety; geological material has been the major
contributor. A more detailed examination of contributions to this
fraction in Philadelphia using optical microscopy is presented in
Table 9.3.7. Unfortunately, in only one case, 500 S. Broad on
November 17, 1979, were SSI, FINE and COARSE samples available
simultaneously.
Comparing the COARSE source contributions to these samples from
Tables 9.3.6 and 9.3.7 shows that while the chemical element balances
calculate a 57% contribution from geological sources, the microscopic
analysis shows only 22.5% due to minerals. The chemical element
balance biological and combustion sources are negligible whereas these
sources constitute 50% and 26%, respectively, of the CP fraction
according to microscopic analysis results. The more specific
groupings of the microscopic analysis (Crutcher and Nishimura, 1981)
9-45
-------
1^.
•
CO
ON
rJ
H
<
PM
W
fa
hJ
W
ffi
fa
W
Q
w
H
o
3
U
CO
S3
O
P-i
CJ
fa
o
CO
M
CO
PLl
8
CO
O
erf
U
CO
a
H
2
0
^
,
o
J_l
srcenl
fa
o
*j~j
j^*
CO
0
a
e
o
u
CO
I i pj
£^60
^
01
0
c
o
•H
4-J
c
•H
0
0
C
i^J o
-H
rH 4-J
cfl cfl
•H 4-1
4J 0
co PL.
3 co
73 C
C CO
M p
H
Cfl
rH
CO
O
•H
60
O
rH
O
•H
PQ
CO
i-_4
Cfl
01
d
*rH
s
60
C
•H Q)
i-H 4->
O. Cfl
e o
cfl
CO
d
o
•H
4J
CO
O
O
ON
3
0
m
CM
CO
vO
m
t-^
3
CM
O
CM
ON
f-^
ON
rH
*
00
rH
•
4-1
CJ
0
73
CO
0
PQ
•
CO
o
o
m
in o r-.
ON ON CO
i-H
rH
•
0 O
!*-*• vo in
O •* CM
CO CM CO
vO vo CO
CM VO O
rH rH rH
rH O m
r-- m o
CM rH CO
m oo r^
ON CO vO
CM -tf CM
4-1
43
60
W -H
S*>- o)
13 C M
0) 43 fa
43 0
60 "-3 4J
01 O
rH • i-H
rH 4-1 -H
•< C/3 fa
1
o
ON
CM
rH
i-H
CM
rH
O
rH
sfr
ON
CO
CO
0)
4->
CO
Cfl
^
S3
sf CO vO
-* oo h~
o o o
CM O 00
vo in i —
CO CM CO
vo m vo
in in ,-
PQ C C
0) 43
. 43 0
C/3 60>-5
01
O rH .
O rH 4J
m •< w
CM CM
O vo
i-H
o o
00 rH
m co
CM CO
CM CM
ON in
ON rH
rH f~~
rH CO
rH VO
CO * m CM m m
CM rH rH i-H rH
ON t^ in co mo
VO r-. CM O rH rH
CO CM CM CO in SJ-
o^ vo r*** ^* co ^^
vO 00 00 ON m CM
CM rH CM CM rH CM
CO
4J 4-1
43 43
60 60
CO iH -H
t^ OJ 0) CO
c d ^ >-> a) 01
01 43 fa fa 4J 4-J
43 O CO CO
60 ••} 4J 4J CO cO
0) O O S S
rH « i-H rH
1 — I 4J «H »H M M
N
PQ C
0)
CO £,0
01
0 rH
0 rH
m <•
in rH
rH i-H
00
O 00
00 N
43
cu
73
•rl
rH
CO
CU
e
CO
01
43
4-J
14-1
o
CO
•H
CO
rH
cfl
rj
cfl
0)
CO
CJ
•H
rH
a,
cu
(2
cfl
9-46
-------
showed most of the biological material was present as plant parts (17%
of total mass) and other biologicals (20%) and most of the combustion
products as charred wood (19%).
These major differences in source contributions to the same
sampling site at the same time cast doubt not only on the source
contribution estimates but on the methods of achieving them.
One source of this difference may rest with the limitations of
these methods. The limitations of the input data to the chemical
element balance were discussed in this section and those of the
optical microscopic analysis were presented in Section 3.3. The
replicate microscopic analyses of samples offers an estimate of the
uncertainty associated with each of these measurements. These
replicate analyses were performed by different microscopists on the
same slides prepared from SSI filters taken at 500 S. Broad, Pilot
Freight and NE Waste on November 17, 1979. The results of the
contributions to the five general categories are included in
Table 9.3.7. The most notable feature of these replicates is that
when the contribution of one category drops below 15 to 20%, the
replicates exhibit extreme differences; industrial and transportation
at 500 S. Broad and NE Waste, and the "other" category at Pilot
Freight are the most striking examples. Significant deviations at
higher contribution levels are also present, notably in the biological
and combustion contributions to NE Waste.
These differences become larger when the specific categories of
Crutcher and Nishimura (1981) are compared. At 500 S. Broad,
Nishimura found 27% of the CP fraction composed of plant parts and 20%
other biologicals while Crutcher, examining the same sample, found 23%
due to plant parts and only 3% contributed by other biologicals. In
combustion products, Nishimura estimated that 19% of the CP mass was
due to charred wood with only 3% from fused flyash while Crutcher
found that 18% was contributed by fused flyash with only 2% coming
from charred wood. Similar differences showed themselves at NE Waste
and Pilot Freight. These detailed summaries are available upon request
to the authors of this report.
9-47
-------
Replicate analyses of chemical compositions of selected samples
always showed agreement within +10% for concentrations greater than
ten times the lower quantifiable limit. To estimate the true
uncertainty associated with chemical element balance source
contributions, however, the entire modeling process should be
duplicated by different researchers. Each should choose the source
types and their compositons independently of the other and each should
perform the chemical element balance fitting procedures using his
choice of chemical species and sources. Only after this procedure is
followed and the source contributions calculated from the same
chemical measurements compared will the process be similar to that
performed by two separate microscopists looking at the same sample.
Major differences between the two methods will probably persist,
however, for the CP fraction simply because the geological source
compositions do not vary by great amounts; it is likely that greater
than 50% of the CP fracton will be attributed to geological material
whereas the results in Table 9.3.7 show most of this material
contributing in the 20 to 30% range. The biological, industrial,
transportation and combustion source types appear to be major
contributors by microscopic analysis, but their contributions via
chemical element balances are much smaller.
Certainly more work needs to be done to perfect both methods and
to evaluate their applicability to estimating source contributions
from IP Network samples.
The treatment here, which is a small part of a moderate-sized
study, cannot be expected to resolve all of the problems and
inconsistencies which have been noted; however, these efforts do
suggest an iterative scenario which could be followed to provide a
cost-effective yet defensible estimate of source contributions. The
steps would include:
1. Identify and locate likely source types contributing to an
area's air using existing emissions inventories.
2. Use whatever source compositions are available, even if
their source types do not exist in the area under study, to
perform chemical element balances on average and selected
individual chemical concentrations.
9-48
-------
3. Compare simultaneous source contributions from different
sites with each other and with the proximity of sources to
receptors documented in Step 1. Eliminate those localized
source types which make small but uniform contributions over
a large area.
4. Obtain source samples from source types in the area under
study which have been identified as possible contributors.
Submit these to chemical and microscopic analysis.
5. Reanalyze those receptor and source samples for which
chemical compositions seem inconsistent.
6. Run chemical element balances with new sources. Perform
microscopic analyses of selected samples using a
standardized methodology and use chemical element balance
fitting procedures with microscopic concentrations and
source compositions to determine source contributions.
Though this procedure should help to refine the source
contribution estimates, the analysis in this section strongly suggests
three things:
• The greatest contributors to average and high FP mass
concentrations in the eastern United States are unaccounted
for sulfate (possibly due to the conversion of 802) and,
secondarily, automobile exhaust.
• The greatest contributions to CP concentrations are from
geological material.
• Industrial point sources generally make small contributions
to average and elevated IP mass concentrations.
These observations are based on the very tenuous assumptions
already stated about the receptor models used and need to be verified
using the steps which have been proposed in this section.
These observations apply to urban-scale measurements, but in many
cases source emissions may be clustered in such close proximity to
each other that the source contributions estimated at one site may not
be completely representative of the neighborhood under study. A
neighborhood-scale sampling network is necessary to ascertain the
variability across an industrial neighborhood.
9-49
-------
9.4 Neighborhood-Scale Source Contributions
In an urban-scale compliance monitoring program, one sampling
site is used to represent an entire neighborhood. Though the TSP, IP
and FP concentrations measured at this station may be representative
of the area as a whole, the source contributions to the concentrations
measured at that station may not be the same as the source
contributions to an equal mass concentration measured at: another
station in the neighborhood.
To study this issue, the samples collected in the Bridesburg
industrial neighborhood of Philadelphia, depicted in Figure 4.2.8 are
used. Each of the seven Bridesburg sampling sites was intended to
take 24-hr samples every third day. Due to technical difficulties,
this intended sampling program was not completely carried out.
Unfortunately, sites having SSI measurements had incomplete
dichotomous measurements and sites with a whole set of dichotomous
measurements lacked complete SSI data. Only three Bridesburg sites,
T&A Pet Shop (5), NE Transfer (6) and Bridesburg Recreation Center (7)
yielded enough dichotomous samples for x-ray fluorescence analysis.
These three sites are on the perimeter rather than in the core of the
Bridesburg neighborhood. NE Airport (10), the urban-scale background
site, also provided a number of dichotomous samples on which
measurements could be made. These samples were taken between 10/3/79
and 2/15/80. The same type of analysis applied to urban-scale
chemical concentrations in the first three sections of this chapter
can be applied to the neighborhood-scale concentrations in Bridesburg.
Table 9.4.1 presents the average and maximum mass and elemental
concentrations of FP and CP measured at the three sampling sites.
There are no significant differences among these sites In the FP
averages for mass, Al, Si, S, K, Ca, Ti, Ni, Br, and Pb. Similarly,
in the CP averages, the concentrations of mass, S, Cl, K, Br and Pb
are homogeneously distributed over the area encompassed by the
Bridesburg perimeter samplers. Comparing these elements to those
associated with the known sources in the Bridesburg area in
Table 4.2.1 shows that the FP fraction might be equally affected by
urban dust and motor vehicle exhaust at all sites. The CP fraction
shows equal effects of motor vehicle exhaust at all sites.
9-50
-------
TABLE 9.4.1
AVERAGE AND MAXIMUM CHEMICAL CONCENTRATIONS
IN THE BRIDESBURG INDUSTRIAL AREA OF PHILADELPHIA (10/79 - 2/80)
(Data from Chow et al, 1981)
FINE CHEMICAL CONCENTRATIONS (ug/m3)
Sampling Site
(5)T&A Pet
Chemical (PAPHL)
Species Avg Max
Mass
Al
Si
S
Cl
K
Ca
Ti
Mn
Fe
Ni
Cu
Zn
Br
Pb
29.70
.36
.35
2.65
.30
.20
.30
.03
.10
.34
.02
.02
.15
.19
.69
(108.80)
( 1.36)
( 2.82)
( 12.59)
( 1.50)
( .62)
( 1.48)
( .27)
( .93)
( 2.48)
( -ID
( -10)
( .98)
( .59)
( 1.82)
(6)NE Transfer
(PAPHM)
Avg Max
32.50
.34
.29
2.70
.67
.18
.28
.03
.02
.25
.02
.14
.55
.16
.66
(106.60)
( .83)
( .98)
( 7.94)
( 4.05)
( -36)
( .56)
( .16)
( .06)
( .78)
( .09)
( 1.83)
( 7.17)
( -47)
( 2.02)
(7)BRC (10)NE Airport (B)
(PAPHN) (PAPHE)
Avg Max Avg Max
29.10
.29
.31
2.58
.52
.18
.25
.03
.02
.20
.02
.05
.19
.15
.56
(77.80)
( .40)
( 1.19)
( 5.28)
( 3.24)
( .39)
( -39)
( .15)
( -16)
( .47)
( .04)
( -61)
( .93)
( .43)
( 1.81)
21.00
.26
.30
2.52
.12
.13
.17
.02
.01
.14
.01
.02
.09
•12,
.53
( 51.60)
( .40)
( 1.13)
( 5.65)
( .50)
( .23)
( .26)
( .03)
( .02)
( .26)
( .03)
( .04)
( .25)
( -33)
( 1.31)
NO
No. of Observations:
Mass 33
Elements 40
Ions
37
40
28
40
34
17
9-51
-------
TABLE 9.4.1 BRIDESBURG, PA (Continued)
COARSE CHEMICAL CONCENTRATIONS (ug/m3)
Samp ling S i te
(5)T&A Pet
Chemical (PAPHL)
Species Avg Max
Mass
Al
Si
S
Cl
K
Ca
Ti
Mn
Fe
Ni
Cu
Zn
Br
Pb
S°l
N°3
20.80
.95
1.89
.60
.42
.21
.88
.07
.38
.96
a
.02
.12
.05
.24
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
75.20)
2.90)
6.39)
7.01)
2.86)
.65)
2.59)
.26)
6.85)
4.38)
.05)
.14)
1.35)
.57)
2.03)
-
™
(6)NE Transfer
(PAPHM)
. Avg. . _Max _
23.00
1.55
2.79
.56
.40
.28
1.30
.11
.05
1.17
a
.11
.52
.08
.20
( 56.
( 6.
( 10.
( 2.
( 1.
( 1.
( 12.
\ •
\ •
( 3.
V. •
( 1.
( 6.
^ •
\ •
-
™**
90)
64)
47)
87)
87)
09)
07)
54)
32)
58)
07)
17)
99)
12)
84)
(7)BRC (10)NE Airport(B)
(PAPHN) (PAPHE)
Av& Max Avg Max
21.50
1.10
2.10
.51
.36
.21
.83
.07
.03
.86
a
.03
.14
.03
.15
(77
( 3
( 5
( 2
( 1
(
( 2
(
(
( 2
(
(
(
(
(
.30)
.03)
.66)
.16)
.98)
.58)
.14)
.36)
.12)
.16)
.02)
.18)
.79)
.11)
.36)
-
*~
10.06
.82
1.46
.31
.27
.15
.55
.04
.02
.56
a
.01
.05
.03
.13
(22.70)
( 1.33)
( 3.07)
( 1.64)
( 1.90)
( .28)
( 1.09)
( .12)
( .04)
( 1.49)
( .02)
( .03)
( .15)
( .08)
( .28)
-
~
No. of Observations:
Mass
Elements
Ions
33
40
37
40
28
40
34
17
aNot detectable.
y-52
-------
The most striking information conveyed by Table 9.4.1 is in those
average concentrations which show differences from site to site. FP
Cu and Zn concentrations are 3 and 4 times higher at NE Transfer than
they are at the other sites. This site is also the closest to the
copper smelting (34) operations pictured in Figure 4.2.8. The highest
Cu and Zn concentrations at this site are 6 and 8 times the average of
the highest values at the other two sites. The second highest Cu and
Zn concentrations at site 6 are 3 are 8 times the average of the
highest values at sites 5 and 7 for FP. These few extremly high
values shown in Table 9.4.1 at NE Transfer cause the average at that
site to be higher than the averages at the other two sites. These
highest values show that Cu and Zn concentrations are periodically
higher at site 6 which is consistent with the concept of a point
source which is sometimes upwind of a nearby receptor. Average FP Mn
and Fe concentrations are 5 and 1.5 times higher, respectively, at
site 5 than they are at the other two sites, and the highest values
show that this is not a general phenomenon, but one that occurs on
just a few days. Since Al and Si are not simultaneously elevated, it
is unlikely that geological material is the contributor here. The
close proximity of the manganese ore unloading facility (59) could
explain this. The average FP Cl concentration at site 5 is only half
as high as that at the other two sites; no single source of this
element is identified by Table 4.2.1.
The lack of spatial homogeneity of several of the CP elemental
concentrations is consistent with the contention that geological
contributions vary from site to site. CP concentrations of Al, Si,
Ca, Fe and Ti are approximately 1.5 times higher at site 6 than they
are at the other two sites. This is consistent with the larger
quantity of truck traffic at this site described in the site survey
summary of Table 3.1.3. The elevated CP Cu and Zn levels at site 6
are consistent with the source/receptor relationship established for
the corresponding FP concentrations. The high Mn concentration at
site 5 is also consistent with that observed for the FP Mn. A
corresponding elevation of the Fe concentration would be masked by the
urban dust contributions.
9-53
-------
It appears that local metallurgical sources and nearby
resuspension of urban dust affect site 6 and that mineral handling
affects site 5. It is interesting to note that the coke oven (40),
with some of the highest emissions in the Philadelphia airshed
according to the NEDS emissions inventory, does not show itself to be
a contributor to sites 5 and 7 which are closest to it. A glance at
Table 4.2.1 shows that carbon, which has not yet been measured, is the
major chemical species to be expected from this source.
3
The average Bridesburg FP mass loading is 7.6 ug/m higher than
the urban-scale baseline FP mass. This difference accounts for 25% of
the Bridesburg FP concentrations. Comparing elemental concentrations
indicates local contributors to this increment. Most geologically
related elemental concentrations (Al, Si, K, Ti) are not appreciably
higher in Bridesburg than they are at NE Airport. Ca and Fe are
slightly enriched (less than a factor of 2), but this is consistent
with measurements of urban dust which show elements of concrete (Ca)
and metal (Fe) abrasion (Hopke et al, 1980). Cu and Zn concentrations
are higher at Bridesburg sites than at the urban-scale background
site, supporting the contention that local sources are responsible.
The doubling of Pb and Br concentrations from the urban-scale
background to the industrial setting shows the importance of
automobile exhaust as a local Bridesburg source of these elements.
Sulfur (S), presumably in the form of sulfate, is homogenously
distributed across the industrial and urban-scale background sites,
confirming what is known of its regional nature. Though some of the
Bridesburg Cl could be due to automobile exhaust, it is a factor of 4
or 5 times higher at two Bridesburg sites than it is in the
urban-scale background average. There is an unidentified local source
of this element in the area, possibly the salting of roads during the
winter months of this sampling program.
3
For the CP, the average Bridesburg mass loading is 8.2 ug/m
higher than the average concentration; thus, local activities account
for approximately 38% of the CP. The geologically related elements
are the most highly enriched and urban dust is the most likely
contributor to these increments. However, Table 4.2.1 shows a number
of local emission sources in the CP mode which do not contain the
9-54
-------
elements for which measurements were made. Feed and grain handling,
and coal and coke storage piles contain significant quantities of
carbon in CP particles which may be part of the local contribution.
Automobile exhaust, urban dust, copper and possibly zinc
operations appear to be among the local sources contributing to
ambient concentrations, though there may be others whose chemical
structure would show no affect on the chemical concentrations which
have been measured to date.
Table 9.4.2 presents the chemical element balance source
contributions calculated from the average concentrations in
Table 9.4.1. These source contributions, for the most part, support
the observations made from the chemical concentrations. FP source
contributions in the Bridesburg area are fairly evenly distributed if
one discounts the vegetative burning contributions which are probably
due to the Cl concentrations; these might be better assigned to the
marine aerosol, or more probably, road salt. The CP fraction shows
higher geological contributions at NE Transfer, and a higher manganese
ore handling influence at T&A Pet, both of which are consistent with
the relative locations of sources and receptors.
The conclusions drawn from this neighborhood-scale study of the
sources affecting FP and CP particulate matter concentrations in the
Bridesburg industrial area are:
• Approximately 25% of the FP, 38% of the CP, and therefore,
30% of the IP average mass concentrations are contributed by
local Bridesburg sources. This means that up to 70% of the
average IP mass loadings are contributed by the urban-scale
aerosol and are not subject to local control in Bridesburg.
• Both average FP and CP mass concentrations, and therefore IP
concentrations, are homogeneously distributed over the three
Bridesburg sampling sites.
• Average FP mass concentrations are almost equally affected
by locally generated sources, though certain small mass
contributors may have higher influence on certain chemical
species.
• Average CP mass concentrations are uniformly distributed,
but samplers in close proximity to local geological and ore
handling sources show greater contributions from these
source types.
9-55
-------
TABLE 9.4.2
SOURCE CONTRIBUTIONS IN THE BRIDESBURG INDUSTRIAL AREA
AVERAGE FINE SOURCE CONTRIBUTIONS (ug/m3)
Sampling Site
Source
Type
Geological
DZLIMEST
PAPHSOIL
CONTDUST
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
RESIDOIL
ALUMPROC
ZOLLERCU
VEGBURN2
VEGBURN1
PAPHMN
Total:
Percent of Measured
PAPHL
b
3.9+. 9
b
7.4+1.2
c
3.4+. 5
.36+. 10
b
.52+. 39
b
4.1+1.3
.15+. 14
19.9+2.1
Mass: 67+11%
PAPHM
b
2.4+. 8
b
c
c
3.0+.5
.36+. 10
b
3.0+2.3
b
18.9+3.3
b
27.7+4.1
85+15%
AVERAGE COARSE SOURCE CONTRIBUTIONS
Source
Type
Geological
DZLIMEST
DZCRUSTA
Secondary
SULFATE
NITRATE
Other
PBAUTOEX
MARINE
ZOLLERCU
ALUMPROC
PAPHMN
Total:
Percent of Measured
Sampling
PAPHL
2.1+.4
6.9+. 8
c
c
1.1+.2
.83+. 31
.46+. 35
2.4+. 6
5.9+1.6
19.7+2
Mass: 95+17%
Site
PAPHM
3.1+1.0
12.8+_1.0
c
c
.77+. 18
.75+. 25
2.5+. 5
4.6+. 9
.16^.05
24.8+1.8
108+11%
PAPHN
.38+. 26
1.85+.6
b
b
c
2.8+. 4
.36+. 09
b
1.2+.9
4.5+2.0
b
b
11.1+2.3
38+9%
(ug/m )
PAPHN
1.9+.3
9.1+.8
c
c
.61+. 14
.72+. 23
2.2+. 4
3.2+. 6
.09+. 03
17.9+1.1
83+14%
PAPHE
.36+. 10
b
1.5^.2
7.2+1.1
c
2.5+. 5
.18+. 06
b
.52+. 38
1.1+.5
c
c
13.4+1.4
64+8%
PAPHE
1.3+.2
5.9+. 5
c
c
.59+. 13
.51+. 30
1.0+.4
2.4+. 5
.06j+.02
11.8+.9
111+12%
needed to fit.
cNot included due to lack of data.
9-56
-------
These conclusions justify the use of one sampler on the perimeter
of the Bridesburg area to represent average mass concentrations of FP
and IP, but they do not justify the use of a single sampler to
identify and quantify the sources of CP, and therefore of IP.
Apparently, relative source contributions to IP vary from sampling
site to sampling site even though the sum of those contributions may
be the same. Thus, detailed receptor-oriented source apportionment on
the industrial neighborhood-scale may require a number of sampling
sites. This condition may be less stringent on the urban-scale or on
the non-industrial neighborhood-scale. These observations are, of
course, to be considered with their limitations outlined in
Section 9.3.
This chapter has attempted to identify and quantify the sources
of FP, CP and IP using receptor-oriented models. The limitations of
these models applied to the present situation are evident and have
been pointed out. Nevertheless, these efforts have resulted in some
observations about the chemical concentrations from receptor samples
in the IP Network data base and the possible contributors to those
concentrat ions.
• Data validation procedures should be applied to IP Network
chemical concentrations to identify and remove suspicious
values which can bias the results of receptor model source
contribution estimates.
• The identification of likely contributors due to the close
geographical proximity of a major source to a receptor is
often confirmed by chemical concentration measurements and
receptor model source contribution estimates.
• The chemical element balance and microscopic properties
balance receptor models exhibit major limitations. Despite
these limitations, it appears that major contributors to IP
Network sites in the eastern United States are unaccounted
for sulfate (possibly from the conversion of 802) and
motor vehicle exhaust in the average FP fraction and
geological material and possibly biological material in the
average CP fraction. Industrial point sources show small
(less than 1 ug/m3) contributions to both size fractions
at urban-scale sites in most cases. Data are insufficient
to apply these observations to western sites.
9-57
-------
In an industrial neighborhood, one sampling site may be able
to measure average IP mass concentrations equivalent to
those at a nearby site, but the sources contributing to
those concentrations may be different. A receptor model
approach to quantifying source contributions in such
neighborhoods may require more than one sampling site.
9-58
-------
CHAPTER 10
SUMMARY, CONCLUSIONS AND FUTURE RESEARCH
This study of measurements acquired during the first year of
monitoring of the EPA's IP Network has answered some questions and
raised others. It is appropriate here to summarize the contents of
this report, to determine the extent to which questions have been
answered, to outline further work which will be required to confirm
those answers and to address those issues that remain to be explored.
The underlying message of this work is that the interpretation
of environmental measurements for descriptive, cause-effect, and
.policy formulation considerations cannot be separated from the
measurement process. Thus, this report has combined at every step
assessments of the effects of measurements on conclusions. This
summary must include the conclusions and recommendations about the
measurement process as well as conclusions about total, inhalable and
fine suspended particulate matter concentrations in the United States.
Chapter 1 set the stage by proposing a list of questions
concerning inhalable particulate matter which might or might not be
answered through the analysis of IP Network measurements and
information obtained in other studies. A technical approach involving
literature review, hypothesis forming and hypothesis testing using IP
Network measurements was outlined and was followed throughout the
study.
Chapter 2 dealt with the sampling devices used to acquire TSP, IP
and FP samples. These included high-volume samplers, high-volume
samplers with size-selective inlets, and dichotomous samplers. The
collection effectiveness curves, which show the percent penetration of
particles through a sampling inlet as a function of particle
aerodynamic diameter, were drawn from a number of wind tunnel tests of
samplers used in the IP Network. These curves were compared with an
acceptable performance range which was proposed for 15 urn cut-size
inlets.
The collection efficiency, the fraction of the total aerosol mass
collected by each sampler, was calculated by integrating the product
of collection effectiveness and fractional aerosol mass concentration
10-1
-------
over all particle sizes. These efficiencies for different sampling
devices under different wind speeds and particle size distributions
were compared with each other. This treatment showed that, under wind
speed and size distribution conditions typical of urban areas, the IP
concentration should equal about 70% of the TSP concentration, TOTAL
and SSI samples should not differ by more than 5%, and samples taken
with a hypothetical 10 urn cut-size inlet should sample 10 to 20% less
mass than a 15 um cut-size inlet.
The Sulfate Regional Experiment .(SURE) IP and FP collection
characteristics were also examined. The IP fraction collected by this
sampler was found to be approximately 10% smaller than that collected
by the dichotomous sampler inlet while the FP collection efficiency
was indistinguishable from the IP Network's FINE sample. This sampler
was included so that SURE IP and FP data could be compared with IP
Network data later in the report.
Only the SSI collection effectiveness fell within the proposed
acceptable performance range though both the SSI and dichotomous
samplers fell within the range of collection efficiencies
corresponding the upper and lower limits of the acceptable performance
range for typical wind speeds and size distributions.
A comparison of the ratios of collection efficiencies derived
from the integration process with average ratios from ambient
measurements showed a reasonable agreement, though the range of
measured ratios was much larger than that predicted by the integration
process. There was also some evidence that SSI mass measurements on
certain samples consistently disagreed with simultaneous TOTAL
measurements; an interference due to a change in filter media rather
than changes in the inlet characteristics was suggested as a cause
which merits further investigation.
Chapter 3 described the network, locating the areas sampled
throughout the United States. When site surveys existed, the sources
within the immediate environment of the sampler were noted. Each site
was classified as being representative of an urban, suburban or rural
area. The site survey information was found to be incomplete, and
those sampling sites requring greater documentation were identified.
10-2
-------
The significance of sulfate and nitrate artifacts was examined
and these interferences were found to cause possible biases in mass,
sulfate and nitrate measurements for a portion of the HIVOL and SSI
samples. The fibrous filter medium used in HIVOL and HIVOL(SSl)
samplers in the IP Network was changed from a quartz fiber filter used
in 1979 to a glass fiber filter with organic binder used in 1980.
Comparisons of average mass ratios of TOTAL/SSI showed 1.08 _+ .23 in
1979 and .90 +_ .21 in 1980. Similarly, the regression line slope of
TOTAL vs. SSI sulfate measurements was 1.07 ± .07 for 1979 samples and
.68 _+ .05 for 1980 samples. A significant fraction of the difference
between SSI and TOTAL mass measurements could be accounted for by
artifact formation on samples taken after the beginning of 1980.
The measurement methods followed to obtain the data in the IP
data base were briefly described. Standard operating procedures
summaries for filter weighing, ambient sampling, and chemical analyses
for ions, elements, carbon and microscopic analyses were presented.
The descriptions were brief and were intended to be pointers to
the more detailed procedures. It was observed that complete
characterization of the methods and procedures was not available to
this study due to the youth of the IP Network, but that such a
characterization would be necessary by the time IP Network sampling is
completed.
The data validation process through which all IP Network
measurements passed is probably the most comprehensive of any of EPA's
large-scale monitoring networks. It involved timing, duration, flow
rate, calibration, and visual inspection checks in the field.
Re-weights were performed in the laboratory. Transcription, outlier
and internal consistency checks were made at data processing. Still,
inconsistencies in individual data values were common, and though
these had been flagged in most cases by the routine validation
procedures, an additional screening based on reasonable limits
suggested by typical ambient particle size distributions was followed
to eliminate the most blatant inconsistencies. Later in the study a
similar, but more subjective, screening was applied to chemical
concentration measurements. In the cases of the seasonal averages at
urban and background sites, this screening procedure was not followed
10-3
-------
and the summary statistics were used exactly as they came from EPA.
Where differences are detected between the results of analyses
presented here and similar analyses by other researchers using the
same data, they are most likely due to this additional validation
process.
The final section of Chapter 3 evaluated the accuracy and
precision of the mass measurements obtained by IP Network sampling
equipment. The effects of recirculation of HIVOL copper and carbon
emissions were examined, and studies were cited which showed that
these contributions to mass and carbon concentrations were negligible,
but that HIVOL and SSI copper measurements could be biased.
Comparisons of simultaneous TOTAL and SSI copper measurements lent
credence to this proposition.
The effects of passive deposition on HIVOL filters over the 6-day
period between samples was examined via a review of several studies
which consistently showed an average bias of +10 to +15% with up to
+25% increases in TSP for individual samples under certain
conditions. Removal of particulate matter from filters during passive
sampling was found to be negligible. No reports of such passive
sampling tests on HIVOL(SSI) and dichotomous samplers were found.
Collocated sampling results indicated precisions of +5% for HIVOL
and SSI mass measurements, +7% for FINE measurements, and +10% for
~~ 3
TOTAL and COARSE measurements greater than 10 ug/m . These
precisions must be considered those attainable by IP Network sampling
and not necessarily the precisions to be associated with any
particular set of values.
In Chapter 4, the characteristics of the urban areas containing
IP sampling sites were summarized. Long-term average meteorological
information and current demographic and geographic descriptions were
given.
Seven urban areas were selected for more detailed examination:
Birmingham, AL, Buffalo, NY, Denver, CO, Phoenix, AZ, Houston, TX,
El Paso, TX, and Philadelphia, PA. Point sources of TSP emissions
were identified and placed on maps in relation to the receptor sites.
The source types identified were tabulated and assigned dominant
chemical components and particle sizes on the basis of literature
10-4
-------
review. Speculations were made concerning the expected ambient
concentrations, speculations which were tested in Chapter 9. In
particular, a lead smelter was expected to have a major influence on
concentrations in El Paso, TX, the iron and steel industry was
expected to influence samples in Buffalo, NY, and a number of closely
spaced industries was expected to affect neighborhood-scale sampling
sites in an industrial area of Philadelphia.
Chapter 5 undertook the task of presenting the spatial and
seasonal variation of TSP, IP, and FP concentrations in urban areas
with an eye to determining which areas of the United States would be
most likely to exceed a standard, what the causes of elevated and
depressed concentrations might be, the necessary sampler spacing
required to represent different neighborhoods and the sampling
schedule required to estimate an annual mean. Los Angeles,
Birmingham, and El Paso showed the highest IP concentrations based on
IP Network data. Urban FP concentrations in the West were generally
lower than those in the East, though the paucity of western sites with
sufficient data limits the validity of this conclusion.
Several sampling sites seem to be required to characterize IP and
FP concentrations in a single urban area. In a highly industrialized
neighborhood, major differences in TSP and IP concentrations were
found between sites spaced only one or two kilometers from each other
even though these mass concentrations were similar to those at other
nearby samplers. Significant seasonal variability of averages and
maxima did not manifest itself at all sites, but it was observed at
enough sites to conclude that year-long sampling is required to
achieve an unbiased average.
The spatial and seasonal variability of non-urban TSP, IP and FP
concentrations was dealt with in Chapter 6. Non-urban IP average
concentrations were found to be more or less comparable in the eastern
and western United States though the distribution between fine and
coarse particles was different. It was speculated that the higher FP
component in the East was due to a greater preponderance of sulfate
and that the larger CP component in the West was due to greater
quantities of suspended geological material. This observation was
tempered by the small number of western sites reporting data. No
10-5
-------
marked seasonal variations were noted in western IP and FP
concentrations, but these concentrations did increase during the
summer months in the eastern United States.
When comparing non-urban to urban concentrations it was concluded
that about 50% of the TSP, 60% of the IP and 70% of the FP found in
urban areas could be due to origins outside of those areas. These
percentages are upper limits due to the likelihood that a portion of
the non-urban concentrations is derived from local, non-urban sources.
The frequency distributions of mass concentrations were studied
in Chapter 7; the log-normal model was chosen as the most appropriate
one for representing ambient suspended particulate matter
measurements, and HIVOL and SSI measurements seemed to fit this model
fairly well. TOTAL, FINE, and COARSE concentrations for the whole
network showed a significant deviation from the distribution. The
99th percentile concentrations of all measurements acquired in the
network were 210, 170, 150, 80 and 90 ug/m3 for HIVOL, SSI, TOTAL,
FINE, and COARSE mass concentrations, respectively.
The use of an arithmetic over a geometric average was addressed.
From a practical standpoint it was found that the two averages vary by
only about 10% and that the variability is fairly independent of
sample size. Thus, it seems that either one could be used for
long-term standards.
The variability of the arithmetic average and 24-hr maximum
concentrations as a function of sampling frequency was studied by
re-estimating them using data sets from which every other record had
been removed. Sampling intervals of up to 24-day were possible before
major differences in the calculated averages appeared. Maximum
concentrations were found to be highly dependent on sampling intervals.
The average ratio of IP/TSP was examined in Chapter 8 as a
function of site classification and TSP concentration. Different
relationships were found for SSI/HIVOL, . 72+_. 13 irrespective of HIVOL
concentration, and for TOTAL/HIVOL, .75.+ . 18 for TSP less than
100 ug/m3, and . 63+_.16 for TSP greater than 100 ug/m3.
The utility of using these ratios to predict IP concentrations
from TSP measurements was examined and for all IP Network data they
predicted SSI concentrations within +30% in 89 out of 100 cases and
10-6
-------
TOTAL concentrations within j*30% in 80 out of 100 cases. The
prediction error was found comparable to the error of estimating the
concentration at a location from an IP measurement at a nearby site*
No consistent relationship was found between FP and TSP concentrations.
In Chapter 9, the spatial distributions of chemical species
concentrations in the Buffalo, Houston, El Paso and Philadelphia areas
were examined to address the speculations of Chapter 4. In general,
the chemical species measurements confirmed those speculations. Urban
sites showed enrichments over background sites, and receptors closer
to industrial sources with unique chemical emissions exhibited higher
concentrations of these species in certain cases.
The chemical element balance receptor model using two original
source samples from Philadelphia and others from previous studies was
applied to average chemical concentrations at urban-scale and
neighborhood-scale sites. This model and the microscopic properites
balance model were also applied to a selected number of individual
samples. The conclusions concerning results derived from these models
are especially guarded since the major efforts required to overcome
their limitations were not undertaken as part of this study.
Nevertheless, it appears that in urban areas of the eastern United
States, the major contributor to the FP fraction is sulfate which
cannot be accounted for by other sources and is possibly of secondary
origin. The next largest general contributor is automobile exhaust.
From 25% to 50% of the FP mass is often unaccounted for, however,
which may be due to chemical species, such as ammonium and carbon,
which are not measured in the IP Network. On CP samples, geological
material seems to dominate and 100% of the mass is normally accounted
for, though there is a discrepancy between the chemical and
microscopic evaluations of the sources contributing to this size
fraction.
A set of questions was proposed in Chapter 1 which the IP Network
was intended to answer. After this review of the data acquired in
that network, it is possible to evaluate the extent to which the
network has answered these questions.
With regard to the sources of IP, the question of probable
sources has been answered by the compilation of emissions inventories
10-7
-------
and the examination of the chemical concentrations. Several important
chemical components are missing, however, notably ammonium, sodium,
elemental carbon, organic carbon and vanadium, which would aid in the
quantification of sources. Aerosol formed from the conversion of
gases, fugitive dust, and motor vehicle exhaust all seem to be
important contributors in the eastern United States while ducted,
industrial emissions appear to be minor contributors. The data from
the western United States were insufficient to answer the questions.
In relation to mass, chemical and size characterization, it
appears that differences between these characteristics are not
generally related to site-type classification with the exception of
urban vs. non-urban categories. This may be due to the subjective
manner in which classifications are assigned. The ratio of IP to TSP
is variable and does not appear to depend on the site type. Though
theoretical studies show that IP measurements made with HIVOL
size-selective inlets and dichotomous samplers should be equivalent,
simultaneous sampling shows that they differ; this may be due to the
adsorption of gases on the SSI filters. The ratio of IP to TSP varies
because of sampling as well as ambient changes which include wind
speed, size distribution and filter media. The ranges and averages of
mass concentrations in different areas are obtainable from IP Network
data, but they depend upon the number of samples obtained and the
frequency of sampling.
The spatial distributions of IP and FP concentrations show that
FP mass concentrations are fairly uniformly distributed across urban
areas, with some notable exceptions, while JP mass concentrations are
not. Gradients are significant with respect to instrument siting and
control strategy development; even though the average mass
concentrations at neaby sites may be nearly the same, the sources
contributing to those concentrations may not be the same. The scale
of representativeness of IP monitors is very specific to the area
under study and only warnings, not generalizations, can be drawn from
IP data. IP Network data is insufficient to determine the vertical
distribution of concentrations near sources.
The IP Network sampling schedule does not permit diurnal or
weekday/weekend variability to be evaluated. Changes in seasonal
10-8
-------
averages of mass and chemical concentrations can be discerned, but the
data available for this report was too limited to form
generalizations. When more than two years of data become available
these observations can be made.
The number of regional-scale sampling sites in the IP Network
with sufficient data was too small to form major conclusions
concerning the transport, transformation and background
concentrations. Associated meteorological measurements and
short-term, multi-day samples required to characterize the events
associated with long-range transport are also lacking. Other networks
clo provide IP and FP data, e.g., the SURE/ERAQS network in the eastern
United States (Mueller and Hidy et al, 1981) and the WRAQS (Hilst,
1981) and VIEW (Cahill et al, 1981) networks in the western United
States which meet these requirements.
Some of the SURE data was used in this report, but not to the
extent necessary to answer the questions in Table 1.1. The
unification of the IP Network and other, related data sets could
provide a cost-effective extension of IP Network measurements.
The text of this report is laced with recommendations about the
measurement process and the need for further data interpretation.
There is no doubt that the wealth of information in the IP Network
data base has barely been tapped. With some reasonable changes in
operating and reporting procedures this wealth could be enhanced.
Since a "tentative" label has been applied to most of the
observations in this report due to the limited quantity of
measurements available to it, the major recommendation is that the
treatments of Chapters 5 to 9 should be applied to a more extensive
data set, presumably the one which will be available in the IP Network
after this report is published. Since attention is shifting from
inhalable particulate matter in the 0 to 10 um range in place of the 0
to 15 um range, data from networks acquiring measurements in this new
size range should be submitted to the interpretative efforts of this
report and others as soon as they are available. These further data
analyses could benefit from the following suggestions:
10-9
-------
• The equivalence of wind tunnel tests should be established,
and the Beckman and Sierra inlets should be tested under
similar conditions to evaluate their equivalence.
• A complete evaluation of the glass fiber filter media needs
to be made and, if possible, one should be used which will
not exhibit artifact formation.
• Site surveys should be completed for all sites and a
meaningful site classification scheme should be devised and
applied.
• Data validation procedures should be devised and used to
reject certain values when calculating summary statistics.
These need to be applied to chemical components as well as
to mass measurements. Precision and accuracy results should
accompany measurements. These should be specific to each
measurement and not applied to the IP Network as a whole.
• Vanadium and chromium concentrations obtained from x-ray
fluorescence should be included in data reports.
Consideration should be given to non-routine but regular
sodium, ammonium, carbon and microscopic analyses of samples.
• Emissions inventories in each urban area sampled by the IP
Network should be reduced to locations with respect to
sampling sites, and source compositions for similar sources
should be assembled. Initial receptor modeling should be
performed and appropriate samples should be taken of the
source types they identify for more refined modeling.
• Spatial and seasonal averages of IP and FP for multiple
years should be examined to confirm whether or not they
repeat themselves.
• The measurement processes of data from other networks which
might supplement IP Network data should be evaluated so that
interpretation of that data can be properly carried out.
• Patterns in log-normal frequency distributions of IP and CP
similar to those found for TSP should be sought at
individual sites. Forms of a standard other than annual
average and maximum concentration should be formed and the
effects of frequency, location and method of sampling on
their attainment should be evaluated.
• Additional models for predicting IP and FP concentrations
from TSP should be devised and tested.
• The present and additional receptor models should be tested
in greater detail according to a predefined protocol to
determine their applicability to IP Network data.
10-10
-------
In general, the IP Network has accomplished its goals of
providing a data base from which certain observations can be made and
hypotheses can be formed. The data included in this report, however,
are too limited to draw definitive conclusions. Many of the
interpretative efforts made here should be applied again when the data
base is more complete.
10-11
-------
APPENDIX A
SAMPLING SITE DESCRIPTIONS
-------
9
X
41
in v
X
in
tn
41
e
X
X
in
X
O
IM
x
•o
or
o
K Ul
« »-
K <
•90
o *in
o *-t-
CE X Ul
0. Z
9
»4
X
o
x
h»
0
IM
X
O
X
»••.
0
e
x
e
x
*-
0
x
0
x
r-
41
ru
x
e
x
^.
c»
e
x
o
x
e
0
4>
0
x
e
x
*w
m
x
o
X
r»
it
x
o
x
r»
0
it
•^
o
•^
r-
(•<
x
o
X
l»
0
e
x
e
x
h»
o
IM
X
O
X
r*
in
ru
x
o
X
r*.
9
IM
X
e
x
r>
9
IM
X
0
X
4»
X
X
*•
O
tt
X
o
X
l»
e
0
x
e
x
9)
Ul
X
«
09
o
or
Z *»
O •
u
a
09
09
at
o
o
«•
9
41
0
!>•
O
»-
e
Irt
4>
0
in
it
o
o
e
0
o
e
l
i
e
o
i
1
4>
M
4>
It
t*
r^
IM
4)
O
IM
in
4>
IM
O
0
0
1^
o
o
1
1
o
e
1
1
o
IM
9
IV
»••
e
o
ft
r»
«•
tn
it
IM
e
o
0
•A
^1
e
e
•
o
0
«
K»
O
e
o
e
it
o
h-
o
o
«<
9
IM
m
^^
0
e
e
9
in
IM
e
i
i
e
e
sin
o
e
0
»
in
r»
0
o
in
_^
IM
in
•• c
0 •
o -
e e
o •
Al 1*
ft C
O C
1 1
t 1
e c
e e
1
4>
9
IM
9
^>
9
41
0>
41
v«
9
> 41
• e
• 0
> 0
> .9
1 9
1 ^
. 0
1 t^
> e
i
i
> o
» ~«
w«
0
0
M
ru
o
>»
in
0
V*
m
o
9
IM
0
O
O
o
4)
0
f^
e
•
in
•t
«4
«
1
1
O
9
e>
IM
•4
h.
it
41
IM
.J
e
9
9
o
e
o
0
4)
O
It
0
1
1
0
1
1
c*-
»
9
in
IM
0
in
in
0
c»
in
P*
IM
O
0
o
IV
0
IM
m
o
i
0
o
in
o
h.
it
e^
*•
*•
ru
e»
0
. 9
«•
9
IM
e
o
o
0
1A
M*
O
tfA
o
•
in
41
IM
4>
0
It
e
o
0
9
m
o
i
i
i
— 0
4)
O
O
IM
•«
»
F»
^
O
IM
9
4>
O-
It
9
O
0
O
4)
^^
m
Ift
e
•
0
m
0
IM
!••
O
9
41
h.
in
>«
41
IM
4)
9
^,
e
e
in
it
in
in
e
•
o
IM
0
IM
0
.O
IM
f»
«*
It
It
m
0
e
41
IM
O
0
o
IM
O
in
e
l
i
i
—
Cf
o
IM
h>
in
9
C*
0
IM
41
4>
in
in
it
0
e
e
0
it
in
0
•
IM
ru
0-
0
»
e
IM
41
^
O-
It
9
—
in
in
it
o
e
e
4>
^9
m
e
i
i
*•
9
e
e
e
o
it
it
«*
0
e
9
0
0-
in
9
0
e
o
0
0^
m
o
•
0
0
IM
9
»
It
0
0
O-
It
^t
c»
41
O
O
in
—
o
0
o
0
in
o
4)
e
•
41
in
4>
in
IM
z
<
z
a
z
<
f.
Z
a
x.
49
I
ae
z
•x.
K
»4
OB
a
u
tn
ui ua
H- o
•_ c
05 U
< o
t-
z in
3 —
o c
z «
(9
Z H-
** «9
Z. U
z e
w e I
co u.
h- < O
05 Z UJ
>- C I
H- Z •- 09
«« O »- CB
U 09 •< Ul
Ul
Z •• IM
h- x «
ce M
« Z •.«
a. ui x »
Ul C M Z
ox z a:
i a. ui o
: x o u.
« 0
U Ul «
> O « O _! O U
• U Ul
O) o «.
t- » z 0
Z 09 •« Ul
M < _l
O -I 09 Ul
Q. Ul O 19
(9 -I Z
: Ul Z •«
ru
u
ru
a
*-
m
z & ui z
< X
a « u «5
or — « o
« «t a. u.
H- — «t z
P"
•• e
(\J c
ui <
oJ H-
tr ce
U. 3
Ui CL
ce ce
« XX
U «
fw
in
x >- i-
_i x 09 a.
«UJ "— O
o » z x
U Ul Ul t- >•
o> o -o H-
«t O X •-•
CU O &. • O9
c a z or
U X Ul
o: 3 a: to i-i
3
U. 09 >
O -I -I
• -I OP
z •
O » Z I-
M ui e 09
> V)
05 Z Ul Z
Ul CD -I
• cc c <
a, o ce z
x _i • «
Ul < • Z •
• 0> Ul Ul
• < z • o
_| UK 4>
rt it 09 o-
ouo
*^ fe
• 09
« U
_l
• Ul
_IU
z «
M
« en
Z
«
09
o •
a. o
c >
ce, _>
i- cc
ui
z r
o
>^ «•*
CO Ul .
x en ce
o i-.O
o y y
o ce
i-" C UJ
D co ;>
O -
u c
•9 U
II 09
U M
Z U
« z
ce «
u. cc
l- u.
Z 0
« •< z
89 Ul •
09
a
z
«
09
O
U
o
ce
CE O
z
— X
09
cn •
• o t-
t- 0 • 09
09 « H-
l-i 0 5- X
SO H- >-
t- Z O •- 9 Z X
It « CC 0 S h-
— CC K> Oc. *C 9
u. ru ui z K «
s > Z
9 Z O PI O f9
— < o z ru o —
-• w o> s -« a 9
x
a.
«• ru KI
n o <
x x a.
a a. u.
M IM M
ru ru ru
< cc
-------
Ul
> t-
J <
Ul O
ac ui
(0 >UI
Ot-l-
ac z ui
o. z
z
3
h- .o r-
•XI — IM
XXX
in •< in
-• o
f» O
— e
o o
o ixt —
I- IT
IM •«
-" O
o w> -»
o o •«
X X X
o ixi ixi
o o «•
X X X
o o o>
o
X
X
e
e
V.
o
0
0
X
0
o
x
in
0
X.
«s
o
m
X
o
1
o
9
C»
X
o
o
X
e
0
0
X
0
1
X
0
0
X
o
e
X
o
*
X
•o
o
X
o
«a
X
e
1
w*
«•
••
X
fa
X
e
it
X
-
i
«
«
o
X
e
e
X
0
e
X
o
I
V
r»
o
X
0*
0
X
0
in
X
o
1
o>
•0
0-
X
•e
o
X
e
ifl
e
X
o
e
e
9-
in
X
9
O
X
O
^t
X
1
in
o-
X
in
o
X
o
«•
X
«rt
1
«J
*4
ni
X
in
0
X
e
in
0
X
i
o
1*1
ni
X
o
o
X
e
o
o
e
X
o
e
i
o
M
•a
X
«•»
o
X
0
co
m
X
o
*•
e
0-
0
h-
X
0
o
X
o>
^
Kl
X
0
r-
IV
9
1".
1X1 0
o m
9 in
O 1X1
r- 0
— m
e o-
in •a
o m
e> o»
9 »•
— o
o in
in e
9 ni
in (xi
m
-
ft
K,
1*
.O
in
9
O
C»
•
•O
in
(XI
•o
CO
*•
•e
••
-•
"
(XI
»
« 9
O O
3 5
e> e
•• e
9 in
9 in
o- in
f» M
•a M
9 «
»xi e
fl »O
e
in
in
•XI
(XI
fl
o»
to
O 0
1X1 <0
•6 1^
0 O
in 9
0- -C
(XI <0
»• 1X1
fl ft
e» in
•O 9-
in «
o
K
«
CO
« Z '
>0
I w
(XI
in
(XI (XI
I <0
I -O
nt
(VI
in
• «XI — -« — KlftlWO' —
w(xio-«4>
•«M«iM«oeoeo
999r*0>(XI(M — — —
— — -r-. — fW(X)(VI«V(M
-• I
•« I
i in i in i
i rvi i e i
i i — i
0>
0>
Ul
IT
O
ac.
u
0}
U1UI
>-o
co u
o
u
OL _
I > K
O « O
u y > u.
a OEH-
« « «
« — a.
— (XI
** ^*
e «
U ^L
o z
O I-
u u
o
ac
o
ac
•<
z
z •
o or
c z*
_l •«
uj c, ,
N^
u. co
X i-t
u or
K- (E
»• c
o
K
U
«. o
< z
o
• Z CE «
•-> I- « iu u
z a. _»>-
«o z uit- z _i
< o o < ui ac
X >-> €J «
«- Z •• Ul
«« i- ui _i a.
ft- _j co «
Z CO < U •
Ul ••* ^
• UI I »- C <0
K U CO Ul
CO •• « ¥ -X
-I Ul O (E >-
•« z
•> u
OC Q.
>_> ui o
z^zonioavL.»_ico
• Ul
I- I -I
a. e o> H- •-
> Ul O «
I- I- S X Ul •
— z o >- a o
U UJ CO 9 K C
u o
a: ee - z z -• < «->
Z CO O Z <
ui 9 1-1 o ui a
U (XI U. (XI CO 00
X 0»
U Z Ul
O K <« < K
C3 «_!«-. :»• _l UJ I Z «. O
< O »- _l H- Ul Z
u »c « M oz«. * ac M
« O Z O U CLUIZ Z O t-
zcauu u o o z -j
u«co^CDKCO « •
»XU>-« >U K a CO l-i Ul i-t
< Z CO CE CO
19 O *5
«••« x a.
ZXI—K3UIXC4COO: ^COZCO£2:«
u 9cnacac-IJOOLO
Or K O COi-'U. CO
o *u •u.jx at
ocozui «u.««_>«Z2 >ouic •ecor
_i (T rf u. I I- ut UJ O U! O. O.COUJUI
KCXOUI atn«« cj»- • e • .j >-
o:oino>i'-uioZ9.jz
«KI«ITN«<<««4HM«iCCC«" ar** *f1CUI
Ute fl ZtOXOUkfOUJC^^CQ^^CQIPCO^COIiJI
a)
UL
Z
(XI (XI (XI K>
^ -. _ (M
U O *- C
o o •« z
U Ul Ul *-<
O O U X
a>
<
x
IE
CD
00 O
X X
u u
— — -, — rvj ru ru
< •< •« a> « 03 •«
« — o c « « ::
X. X » » 35 31 O
en < « < o c c
3£ -I Z Z Z Z X
A-2
-------
o
o
o
z
x
o
o
o
o
o
o
o
o
o
O V* V4 •-« •*
o o o »* o
CO O 9
•» ru •<
XXX.
•O -A -
I- <
10 O
o
X
o
X
0
0
0
X
o
X
o
o
o
X
o
X
o
o
ru
X
o
X
o»
^.
ru
x
o
X
o-
F^
0
X
„
X
O-
|-fc
O
X
o
X
o
OD
ru
X
o
x
0
OD
O
X
O
X
o
o
o
X
o
X
o
o
in KI -o
— O ru
XXX
9 in »-
o o o
o
X
00
o
o-
o
o
CO
o
OD
o
o
o
o
o
o
o
o
o
Ui
z
o
u
U I
1C •
O t-l
a: x i
a. :
x
Z
1 9
1 0
3 O-
D KI
T in
D in
r- oo
D •a
r 9
9 ru
D —
1 KI
n -a
n ru
n KI
r-
o
o
o
00
CO
*0
9
O
in
ru
KI
KI
in
o
ru
in
•o
o
9
0
in
00
r-
9
g
0
O
O
OO
o-
9
O
in
f~
00
r*.
9
I
1
in
OD
ru
Kl
9
O>
•>.
in
KI
•D
Kl
KI
O
o
0
-o
-o
ru
9
O
O
in
_J
KI
I*.
O
in
o-
•A
ru
9
Kl
O-
9
OD
Kl
r^
Kl -O
— O
9 —
O- IT
0 CO
0 0-
in KI
co o
i» in
oo —
-• o-
0 CO
9 •£>
in
ru
o
in
03
o-
Kl
(^
«•-
•c
o-
00
•o
o
o
•o
ru
n*
in
9
O
O
CC
03
in
in
o
o
in
o-
Kl
a
9
o
o
o
Gr-
in
in
o in
o •*•
0 9
•o o-
Kl 0
in in
9 9
O Sf
o a
O CO
CO CO
Kl CO
9 in
O
a;
• o
i <-> i
11- i
UJ I
-* in o
o -i tv
o o o
o o o
M
ru
(M
ru
9
ru
in
in
o
>o
ru
ru
9
ru
ru -« t- -«
o
o
o
Kt O
ru o
•a -o
ru ru
O -"
o o
OD -«
o» Ki
9 o
•« r>
ru ru
o o
«-t o
in o
9
-• »< -« «n
o o o o
-< o o o
o o o o
•4) ^> OD O
Kl O
O
OD
in
& -4 ~ ~
O «-t -^ ~* •*• Kt O
9 9 9 9 •« — O
Kl Kl Kl Kl Kl Kt ru
o
in
O
o
09
•o
9
Kl
Kl
o
KI
U
cc
ru
«
ru
«
o
o
ru
«
o
o
o
o
o
o
o
o
co
co
u
a:
a
o
co
00
Q,
•-4
or
a:
o
ffi
a
(TKI
« KI
a. ru
o 1-1
»-
« o
z z
«t
< _l
1-1 CC
o <
« u.
u u
i-t O Z
o z z
or «.
M »uj co eo co
z o o lu uj >-i
x cc or _i
«. .o o o
i— _i uj u. u. a.
»-• o
o cc
or i- «t < _iz
I- Z Z Z Z O Z
U1O O O O O »
_IO<_>tO«-ii-lIZ
o or I- i- o
O >.Z UK < CO >
X t- O U- Z Z UJ
u or 1-4 u. x >
ooo»-ujzzi9«r
U. ^> -> 1-1 1-1 1-1
I J « «t X Q
U «_! >Z Z CC
H 2 O Ul
x a.
o z o o «
o
•> o z
00 Z
_i • m
O >- 1-1
. o. »- =>
< •-> o
Ul U -I
z
Z 00 •
Z CO CO
o -> o z o
z
t-i in
19 in
UJ ru
or 9
m co
>- z a z
UJ O I- => O
_l i^ CO CD »-i
_l»- O O t-
z o «t u z i O
CO (J
in
KI
u
<
3:
z o or « cc or
.^) 1-1 Kl < •<
eo >o < ^ Z Z
»» ru o KI ru ru
»» — o ~ -« •«
< CD < < OD «
Z Z
2 2
z uj r- M ru uj
z o oc. — < -o or
o o 1-1 in or o 1-1
o KI u. *^ f- in u.
O Z Z
_j o t-»- or
11 Z CO uJ
• ^ ^ UJ CD
i- it « o or
co or »- z o z
« co < u. «u
• a. -i z
x: or X -i
U _t UJ CO 4 &
or « o • z uj h- or
< Z U Z 3
z ec o •-« < x
o uj x cc *— >—
o to • co o in :
or D >- 1-1 or KI .
CD o u. o o 9 i
a.
i—
or
o
t—
co
z •-•
X
or
O CO
1-4 Z
CD 19
Z
o «
UJ Z
ui >
z _»•-<
Z < -I
< I-
h-l •
oo. a
z a: co or
U I
O X OO
X. -I • -J
»- •-! or oo _J
ru
CO CO
O O
Z Z
CO
<
O
u
z z
X X
x a.
.13 ZJ
> O O
I"-
O
c
CD
O
(_>
CO CC X
or < t-
t- ui z a
1-1 o or o
Z Z < X
13 < CD to
< X
UJ UJ *O
m -i
co o
ru
KI
ro
UJ O- ••
-t Z
1-4 Z O _t
> or it
uj »t- o
co •a Z C
1-1 U O CC
X U O CD
z -s. z. •
« UJ O UJ
> Z K. >
Z I- «
«. 1-1 =3
OD Z J »-
ru _J z
in «o 1-1
t t- a. o
r^ co O-
ru « t-
»- in o i z z
to o >- «-i u
CC >• 3 9 O Ul
zi z z o. or
X X UJ • (9
UJ Z Z Z
5C U. Z UJ —
u. 1-1 in UJ c
o z ru a: KI
CD
J-
2
A-3
-------
OOOOOOOOO--O-OB- — —-• — •« — «>
oruruiMruoruruo-«-«oiv — — rurururuo
w xxxxxxxxxxxxxxxxxxxx
OH- o^^^^o^^ruitKi^in^^inininin^
2« oooooooooooooooooooo
UJ O XXXXXXXXXXXXXXXXXXXX
oooooooooooooooooooo
O tfl Ift 1^ O O 0 ^* «O K) ^K AJ f^ f^ f^ M PO P^ •* ^T
orururu — oooruooru — •* — -• — — ruru
>» xxxxxxxxxxxxxxxxxxxx
K UI OBBJC^O-OBt^ruiM — KimKlKlffldimO-
« h- OOOOOOOOOOOOOOOOOOOO
*- « xxxxxxxxxxxxxxxxxxxx
mo o*<**o»o**s>ooooooooo»^
oi^f»r*i»>oi^r»r»a>«e«Ba>oeieh»^
UI K
d9*u I « ru e ru 9
o »-•- i o o o o ru
e
in
-•
in
IM
9
«
9
•e
r»
o-
in
e
V
V
V
e
o
•a
r>
^
•a
e
iu
M
9
e
o
o
e
r>
•o
e
o-
V
e
e
«
r>
r»
•a
e
«
e
0
•o
•a
r»
9
0
O
9
^
h-
O
9
M
9
e
e
B
*»
9
9
e
•a
•o
9
e
e
•
o-
t*t
o
ru
»
9
O
O
«
9
9
9
9
*•
4J
9
ru
•e
*•
9
ru
9
in
Kl
ru
9
in
o
ru
9
ru
^.
in
*>
ru
9
in
o
ru
9
ru
p»
in
10
ru
9
in
o
ru
9
ru
^
in
ft
ru
f
in
o
ru
9
ru
p>
in
M
ru
9
in
o
ru
9
ru
^-
in
w
IM
9
in
o
IM
9
IM
l«»
O
in
t-
9
e
e
0
ru
1*1
in
X
X
« oooooooooooooooooco— •
O oooooooooooooooooooo
(e •>4O4ru*k4niooo9999997rue
iu o • loiiimiiiru
_l h- U. I IllxlllOllllt
UJ I ^ X
O
U O-«-«OO— "OOOfUOOOOO
o « x o c o o o o o
O Z
•c oooooooooo»o_joooooooo-«
a uj
•> i- z »
01 _j a • « u x
u< a. >• >- -i z o _i _i x o z a:
V ZZOi-iXXIU_lUO CO
o >-1- oo of. >o »~
c zz * » x >• j x *o ui o • » *•> ui o) »
UI O»-O-t t-MX*
t- >. _j << u u. xzz«c«c»ccc < > 2: _i
« z u. a « = z o •- >-1- a: zouixmt-ieux t- >-« uiua
o u. H- a> o -i x t- _> ^ m ir» u K ae «> _J « < _i a c
•« ir x • «-«<«< o*-ujeoxo« •. c o o _j r uj a
a> aouio •»- u, z x uj z z _i »-xa._i>-o uj a _i uj 3:
z — is a: uj« u. o z x • •-< o o x =i «^-x i- H- >— o c >-«"••
o »• • > f- o o < >• uu csinxzzKu « z z —' o »3 13 z
w —uiruK-iujru«"X»<(-<-«ffl cuo a o_ir_ilu«o>-"
a. •« • m o • ^ a • • _j o >- o o « •? o. tr H- ~ 2 a uj c a.oa.oa)H-zcn
ac zazoaooz* — >«>cca Kouioa2H>u nzuiz^zzzuizruirz _i •-
iu ccooacm^O'XX a • x>—_iaxo.o< oj z i— a: >—>•>" H- LJ t— <_> i— o < o
x x x < « j x o 3 _i ru u;
— — ru — — «•« — rvi — — -.ruiMrvjruruoc — KI
«>— — — i— c
t- O 33SlCOICX«UUUUXXX£XX£0}a.
•-O >->->-V>->-XXXXIIXXXXXXXQ:
SO O ZZZZZZQ9OOQOO3Q3O = OO
A-4
-------
9 in
x e
9- -O
o o
•x. x
e> o
v
s<9
o
v
o
o> -• in »•»
•a
o
x
xx
intn
x
in
x
in
x
in
in
»
x
o
X
a .»-
O t- IU
K X X
O
IU
r
o
o
X
X
-o
IM
in
(M
in
M (VI (VJ «. IM
in >»»
ivj
IM 0
e ni
M
w o- o
0«0
co IM in
e o
o o
t- in
in
•o
IM
•e
IM
o
(M
IM
in
in
e
•o
•o
in
o
OE
9- e
— (M
O
9
IM
O
•o r-
O — «M
o e o
I I I
II III
i -i
z s
uj in o
o in o
ro 9
" Kl
X •«
O C
a. o
a: <
o a.
o
at
o
tr
< O
T a
99 O
or •
« n
— c
in in
a. — — x «. x a. _ia.i—
x»- x a. •-• a. x>-xo ui •
< o a. uj x a. > < o o *xuu
CE> « Ct O >- >- H- »
IU^- K« Z • >3XZ09««
te z >- z x o >• tr 11- •< ui
« « z uj tr t- o e « uj -X O X H- t— UJ
x t- jr o x tr z z >-o u o« if
3 *C U CL O) UJ ^3 O ^J ^^ C9 ** liW ^J ID QC U
o • >• u. uj
>- «c .
•05 uj u
at
• ui
rf U H- UI -IK- •
• u a o co K
iu^o«eoiu4>-*->- i— z uj u. u < co
xxzai »~cn>-coiucoxc<9 xie
cacao >- x ^- <; e _j o cr >- »-
Z»-zo:xeoxQ.x-^ujCo «z
oor«ciH-ujt-3:>— •^-_j9iua
u-Zox—o-irt— — er>»«a.»\jza.
cv. r- ru — — ^ — — ^
OUOIUU.CAXIX1V
xxxxxxxxxx
0.0.0.0.0.0.0.0.0.0.
uj uj z c
_i_)_iz
_i _J uj c
C O IU C
Z >-• > 1-1
IU U
K CO CC K
O « CJ
X K X UJ
vu « tr
• u. _i
T O) UJ tf
Z O K
•« < =>
tr »o
a >- 2 en
LU ai
u •-• e
tr iu o M
u «e> a:
X z
co _i x
o u
09 O 09
3 ~X
1-1 UJ <_> X
1C »- CO U
u «- —
3 « X X
CD U
O O "-> >
«• < IU X Z
O I- UI
C tt < _i X
Z ID (J « a
C «C 1-1 UJ
X CO -I X _!
X -J UJ _1
u e o a. <
« o u x
(M -• Kl •«
X
a.
x
a.
z
a.
— IM
a «
a. a.
-------
o
o
UJ
o »-
z <
UJ O
in rr 9 9 9 in
o o •» «< -« ••«
X X X X X X
•* 9 * NO .o 3
x O C O O O
o
CD
o
03
O
00
o
CD
in «-> o
M ru o
XXX
in 3 o
o o o o
X X X X
o o o o
CO OD O O
X
o
o
o
o
o
o
o
o
X
o
o
o
o
o
o
o
o
rvi
X
•o
o
X
o
t-
«
in
a
IT
Kl
1
1
Kl
•O
O
^
O
Kl
Kl
00
^
Kl
in
K
ru
ru
o
o
00
Kl
in
o-
ru
Kl
•ft
9
in
CO
ru
I
*
Kl
fw
OD
ru
?•»
01
Kl
Kl
3
•O
3
O
Kl
1
f
Kl
3
•0
0
in
3
3
3
in
^
o
3
1
*
O
o
00
,,
v«
in
9
O
O
Kl
in
ru
9
1
t
O
O
OD
9
•£>
ru
9
O
O
in
r>
Kl
ru
^H,
^*
O
9
^
•ft
ru
in
o
c-
r>
9
in
3
0
0
O
o
Kl
OD
ru
in
0
*•«
K>
0.
9
in
O
o
m
•
r> 9 r» *
Kl
~ — ru IM
3 * „ « .. h- Kl
KI o o o o in •»•
o o o o o o o
o in o o o o o
•o •» ru ru CD 9 9
in »•• in o OB oo CD
ru 9 o o o -» •«
in in oo r> r>
9993939
0190
in • — —
Ml Kl
9
ru o •» -«
CO
CO
U
IT
oo
•e
o,
•*4
K,
CO
UJ
o
» o
CO U
or
• >- a «« co
a. n. uj co
o o _j co
<_> o o. <^ z>
Z «i O
O*~ « CO X
ex u. co co o
X 03 ^
>- O O X
ct to 03 > :» o
o z: x. i-> 1-1 >-•
»-< «i r I o
z z z
UJ u. u. u.
> cr sc o o o
zoo
M at a a
o —J -j uj uj ui
U 0 (J y X I
1-1 O O 3 3 -3
X. -i-J* Z Z
c o o c c c
u u u u u u
A-6
-------
APPENDIX B
AVERAGE METEOROLOGICAL MEASUREMENTS &
URBAN CHARACTERISTICS OF IP NETWORK CITIES
-------
O 00
o >
M <
S 3
o c
C C
VI <
CO 60Q CO
C C CO
O -H 01
oi 60 in
Q •^
I OOOOOrotn I
a
a
B 0) 3 C
• B! « C
n •<
60
•a >
01 <
01 <~-
d, rH .£
co co a.
•O 3 S
c c —
•H C
0) 01 <
i-l C
.O ^H «H
vn J= 0)
01 M 3
as C C
O 3 C
0* CO •<
c
o •
•H 60
-rf a. s
« B <
a 01 •
H iH b
rH OO
ca • 3 ^
B c e
t4 -H C
o S <
^H a >
n S <
• 60
a. >
6 <
n c so
B ffl C •
ki 01 C
o 2: <
C^O»-HsDOC>ir^O^OC^OOrHLnOmu"ivD*£>v£>irimiri
vD I— *£> m u^ -^- u"l
X
H
Birmingham, AL
Phoenix, AZ
Los Angeles , CA
San Francisco, CA
Denver, CO
Hartford, CN
Washington, D.C.
Atlanta, GA
Honolulu, HA
Des Moine , IA
Chicago, IL
Boston, MA
Baltimore, MD
Detroit, Ml
Minneapolis, MN.
Kansas City, MO
St. Louis MO
Raleigh-Durham, NC
Buffalo, NY
Cleveland/Akron, OH
Cincinnatti, OH
Portland, OR
Philadelphia, PA
Pittsburgh, PA
Dallas - Ft. Worth,
XI X
H H
0 C
CO O
rt *H
O< CO
3
rH O
U 33
Salt Lake City, UT
Richmond, VA
Seattle, WA
B-l
-------
a
4)
Ha jo
Industr
>^ a •« o "w
- t ' 5s
£ *« 34) 1-
4j jj O C "O 41
3 g ** ^ S «
1-1 * - ^ &.
w c *J •" ** *fl
*j c a. i- w -H co
o co a s ^i ^-' w
«M W U *M CO ^* W
3 C • CO >* *J U
C4!>*'HT3 -^^-i 3
(08 U.3C 4» ^ • t:
e a. -J c « w s • u o
-7 CO (5 «J4> 41 U
D3 3 B 00 ^H« 4J a
k< to a.j= Uk> eoucacgca *
4IU30 •O'OXI 41 0 « Li 3 QJ 4)
cnwB<«'«i«a'Oe: o < « H M «
U ^-»
•o
41 -
lufacturer
9. tourism ant
reduction, foe
tool ma
Buslnes
Steel p
industry
r
frozen
O k- "V
0 3 C
w 4-1 eg
u
TD 03 4-t
" SS
•S 25-
L products
1 trucks , mete
ilc computer o
ary, farm equ]
jducts
s ca u w 13
U O U W IM
ca
*J CO
C 41
CO -H
>
X iH
1 truck asseml
>nt related at
1 steel compar
o cu c
-i«S£
products
cosmetics
s .
0 CR
CO
•^ u
CO S
4)
(J
"B en
41 »-f
£ O
cj O
TJ
c
"8
0
-
CU
B
u <-<
41
irel manufacti
ilc Industrie;
aducte
y -o
c c
CO CO
B 41
0 0
ki CO
CU 3
CO
S
apparel, foe
. manufacture
:oal „ alum i nun
•" a *
•Q
O
C
c ^
o *
J-i CQ
V
4) U
power plants
Ld machinery,
»rgent produce
:e
_l 3
CJ ~4 -W CO
?
3
C
B
CO
01
C
41
U<
CO
CJ
Electro
c "o
CO C
O. CO
1 «
3
at ^
3 CO
3 k< C
CJ 3 CO
•^ 4-1 B
emical and ag
Winery manufac
Tics, apparel
industries •
«o o e *j e
a. u. u cu
B 4J -Q cu •41
"O CJ
N CJ
U -D
O O
k. 0
U w O C
a. cu CL o
•*i to -H
•U 4) *J
c «-i en «j
CJ 01 (X (A
co E a c
.A E •«•» «
U
if
1
! 2
1 «
E £
(- O
k,
•
CO
n
5
8^
t- O
> w
O S
= 3
o
oo n"
C w
— C
ca w
» <
to
• a
3 •*
** C
3 0
•3*
0 41
X O
ri
O
00
CO
CJ
o
r
- o
C E
o -5
o CO
a r «
ki
a Q
•H t
3 JZ
41
w ca
en (£
0 .
Z * *•!
- C cq
O ca C
•H ^H C
CO 4* -rt
w- > 0
CQ U U
°
C
CQ
0
CO
CL
>H
41
CO
a-
00
ki
3
cn
di
u
O
u*
in
co
Q
B-2
-------
APPENDIX C
REFERENCES
-------
REFERENCES
Akland, G. G. (1972). "Design of Sampling Schedules", Journal of Air
Pollution Control Association, 22, 264.
Appel, B. R., Wall, S. M., Tokiwa, Y, and Haik, M. (1979).
"Interference Effects in Sampling Participate Nitrate in Ambient
Air", Atmospheric Environment, 13, 319.
Appel, B. R. (1981a). Personal Communication, California. Department
of Health, Air and Industrial Hygiene Laboratory, Berkeley, CA.
Appel, B. R., Tokiwa, Y., and Haik, M. (I981b). "Sampling of Nitrates
in Ambient Air", Atmospheric Environment, 15, 283.
Barrett, R. E., Margard, W. L., Purdy, J. B. and Strup, P. E.
(1977). "Sampling and Analysis of Coke-oven Door Emissions",
EPA-600/2-77-213, Research Triangle Park, North Carolina.
Baumgardner, R. (1981). USEPA/EMSL. Personal Communication.
Benarie, M. M. (1971). "About the Validity of the Log-normal
Distribution of Pollutant Concentrations", Proceedings of the
Second International Clean Air Congress, Academic Press, New
York, p. 68.
Bencala, K. E., Seinfeld, J. H. (1976). "On Frequency Distribution
of Air Pollutant Concentrations", Atmospheric Environment,
H), 941.
Bevington, P. R. (1969). Data Reduction and Error Analysis for the
Physical Sciences. McGraw Hill, New York.
Blanchard, G. E., Romano, D. J. (1978). "High Volume Sampling:
Evaluation of an Inverted Sampler for Ambient TSP Measurements",
Journal of Air Polluation Control Association, 28, 1142.
Brock, J. R. (1973) "Process, Sources, and Particle Size
Distributions", In Fogs and Smokes, The Faraday Division, The
Chemical Society, London.
Bruckman, L., Rubino, R. A. (1976). "High Volume Sampling: Errors
Incurred During Passive Sample Exposure Periods11, Journal of Air
Pollution Control Association, 26, 881.
Cahill, T.A., Ashbaugh, R.L., Eldred, P.A., Feeney, P.J., Kusho, B.H. ,
and Flocchini, R.G., (1981). "Comparisons Between Size
Segregated Resuspended Soil Samples and Ambient Aerosols in the
Western United States". In: Atmospheric Aerosol: Source/Air
Quality Relationships, (Macias, E.S. and Hopke, P.K. editors),
ACS Symposium Series 167, American Chemical Society, Washington,
D.C.
Camp, D. C., Cooper, J. A., Rhodes, J. R. (1974). "X-Ray
Fluorescence Analysis. Results of a First Round
Intercomparison. X-Ray Spectrometry, J3, 47.
C-l
-------
Camp, D. C., VanLehn, A. L., Rhodes, J. R. and Pradzynski, A. H.
(1975). "intercomparison of Trace Element Determinations in
Simulated and Real Air Particulate Samples", X-Ray Spectrometry,
4, 123.
Camp, D. C., VanLehn, A. L. and Loo, B. W. (1978). Intercomparison of
Samplers Used in the Determination of Aerosol Composition,
EPA-600/7-78-118, Research Triangle Park.
Capar, S. G., Tanner, J. T., Friedman, M. H. and Boyer, K. W. (1978).
"Multielement Analysis of Animal Feed, Animal Wastes and Sewage
Sludge", Environmental Science & Technology, 12, 785.
Chahal, H. S., Romano, D. J. (1976). "High Volume Sampling: "Effect
of Windborne Particulate Matter Deposited During Idle Periods",
Journal of Air Pollution Control Association, 26, 895.
Chow, J. C., Shortell, V., Collins, J., Watson, J. G., Pace, T. G.
and Burton, B. (1981). "A Neighborhood Scale Study of Inhalable
and Fine Suspended Particulate Matter Source Contributions to an
Industrial Area in Philadelphia", APCA Meeting, Philadelphia, PA,
1981.
Clark, J. B. (1981). Personal Communication, EPA, North Carolina.
Coles, D. G., Raggini, R. C., Ondov, J. M., Fisher, G. L. ,
Silberman, D. and Prentice, B. A. (1979). "Chemical Studies of
Stack Fly Ash from a Coal Fired Power Plant", Environmental
Science & Technology, 13, 455.
Countess, R. J. (1974). "Production of Aerosol by High Volume
Samplers", Journal of Air Polluation Association, 24, 605.
Courtney, W. J., Tesch, J. W., Russwurm, G. M., Stevens, R. K.
and Dzubay, T. G. (1980). "Characterization of the Denver
Aerosol Between December 1978 and December 1979", APCA Meeting,
Montreal, Quebec.
Coutant, R. W. (1977). "Effect of Environmental Variables on
Collection of Atmospheric Sulfate", Environmental Science &
Technology, U, 875.
Cramer, H. E. and Bowers, J. F. (1980). "An Overview of
Atmospheric Dispersion Modeling of Particulates"„ Proceedings:
The Technical Basis for a Size Specific Particulate Standard
Parts I & II Specialty Conference, p. 266, Air Pollution Control
Association, Pittsburgh, PA.
Crutcher, E. R. (1978). "inter- and Intra-Laboratory Comparison
of Optical Microscopical Analysis of HIVOL Filters", Boeing
Particle Identification Laboratory Report to EPA Region X, Ref.
AHM-78, Seattle, WA.
Crutcher, E. R. (1979). "A Standarized Approach to Airborne
Particulate Analysis Using Quantitative Light Microscopy",
Proceedings: Institute of Environmental Science,.
C-2
-------
Crutcher, E. R. and Nishimura, L. S. (1978). "Estimation
of Error in Quanitative Microscopic Analysis of Compounds",
Boeing Aerospace Company, Report, Seattle, Washington.
Crutcher, E. R. and Nishimura, L. S. (1981). "Optical
Microscopic Analysis of 19 Airborne Particulate Samples", Boeing
Technology Services Report BTS 987-1, Seattle, Washington, Sept.
28, 1981.
Curran, T. C. and Frank, N. H. (1975). "Assessing the Validity
of the Lognormal Model When Predicting Maximum Air Pollution
Concentrations", APCA Meeting, Boston, MA.
Deer, W. A., Howe, R. A., and Zussman, J. (1966). Rock Forming
Minerals, John Wiley and Sons, New York, New York.
deNevers, N., Lee, K. W. and Frank, N. H. (1977). "Extreme
Values in TSP Distribution Functions", Journal of Air Pollution
Control Association, 27, 995.
deNevers, N., Lee, K. W. and Frank, N. H. (1979). "Patterns
in TSP Distribution Functions", Journal of Air Pollution Control
Association, 29, 32.
Dzubay, T. G. and Stevens, R. K. (1975). "Ambient Air Analysis
with Dichotomous Sampler and X-Ray Fluorescence Spectrometer",
Environmental Science & Technology, _9_» 663.
Dzubay, T. G. and Rickel, D. G- (1978). "X-Ray Fluorescence
Analysis of Filter-Collected Aerosol Particles",
EPA-600/J-78-120, Research Triangle Park, NC.
Dzubay, T. G. (1980). "Chemical Element Balance Method Applied to
Dichotomous Sampler Data", Annals of the New York Academy of
Sciences, 338.
Farewell, S. 0. Gage, D. R., Jernegan, M. F. and Felkey, J.R.
(1981). "Inhalable Particulate Levels in Northern Idaho
from the Initial Eruptions of Mt. St. Helens", Journal of
the Air Pollution Control Association 31, 71.
Federal Register (1971). "National Primary and Secondary Ambient
Air Standards, Appendix B - Reference Method for the
Determination of Suspended Particulates in the Atmosphere",
Federal Register, ^6(84): part II, April 30, 1971.
Federal Register (1979). "EPA Rules and Regulations: SLAMS
Monitoring Objectives and Spatial Scales", Federal Register,
jW(92), 27586, May 10, 1979.
Frank, N. H. (1980). "Difference Between Arithmetic Mean and
Geometric Mean TSP", EPA/OAQPS memo to John Bachman and Henry
Thomas, Ambient Standards Branch, June 2, 1980, Research Triangle
Park, N.C.
C-3
-------
Frank, N. H. (1981b). Personal Communication, US EPA/OAQPS,
Research Triangle Park, NC.
Gordon, G. E. (1980). "Receptor Models", Environmental
Science & Technology, 14, 795.
Grantz, J. A. (1981). "Inhalable Particulate Matter in the
Vicinity of an Integrated Iron and Steelmaking Complex", APCA
Meeting, Philadelphia, PA.
Greenberg, R. R., Gordon, G. E., Zoller, W. H., Jacko, R. B.,
Neuendorf, D. W. and Yost, K. J. (1978). "Composition of
Particles Emitted from the Nicosia Muncipal Incinerator",
Environmental Science & Technology, 12, 1329.
Harris, D. B. and Smith, W. B. (1980). "Sampling and Data
Handling Procedures for Inhalable Particulate Emissions from
Stationary Sources", Proceedings: The Technical Basis for a Size
Specific Particulate Standard, Parts I & II Specialty Conference,
p. 107, Air Pollution Control Association, Pittsburgh, PA.
Heisler, S. L., Henry, R. C. and Watson, J. G. (1980a). "The
Source of the Denver Haze in November and December of 1978",
presented at APCA Meeting, Montreal.
Heisler, S. L., Watson, J. G., Shah, J. J., Chow, J. C.,
Collins, J. C., and Whitney, J. (1981). "Recommendations for the
Design of Aerosol Characterization Studies in Houston and El
Paso, Texas", prepared for the Texas Air Control Board.
Environmental Research & Technology Document P-A716-110, Concord,
MA.
Hileman, B. (1981). "Particulate Matter: The Inhalable Variety".
Environmental Science & Technology 15, 983.
Hovey, H. H., Jones, H. C. and Stasiuk, W. N. (1976). "Short
Term Air Quality Standards for Suspended Particulate Matter in
New York State", Journal of the Air Pollution Control
Association, 26, 129.
Hubert, J. S., Candelaria, R. M., Rosenblum, B. and Appelgate, H. G.
(1981a). "A Survey of Ambient Air Levels of Lead in El Paso,
Texas from 1972-1979". Journal of the Air Pollution Control
Association, 31, 259.
Junge C. (1963). Air Chemistry and Radioactivity, Academic
Press, New York
Kahn, H. D. (1973). "Note on the Distribution of Air Pollutants",
Journal of Air Pollution Control Association, 23, 973.
Kalpasanov, Y. and Kurchatova, G. (1976). "A Study of the
Statistical Distribution of Chemical Pollutants in Air", Journal
of Air Pollution Control Association. 26, 981.
C-4
-------
King, R. B., Toma, J. (1975). "Copper Emissions From A High
Volume Air Sampler", NASA Technical Memorandum NASA-TM X-71693.
Kolak, N. P., Hyde, J. and Forrester, R. (1979). "Particulate
Source Contributions in the Niagra Frontier", EPA 902/4-79-006,
Region II, New York, NY.
Kolak, N. P. and Visalli, J. R. (1981). "Comparison of Three
Methods for Measuring Suspended Particulate Concentrations",
Environmental Science and Technology, 15, 219.
Lane, H. U. (1981) Editor. The World Almanac and Book of
Facts 1981, Newspaper Enterprise Association, Inc., New York, NY.
Larsen, R. I. (1973). "An Air Quality Data Analysis System for
Interrelating Effects, Standards, and Needed Source Reductions",
Journal of Air Pollution Control Association, 23, 933.
Larsen, R. I. (1974). "An Air Quality Data Analysis System for
Interrelating Effects, Standards and Needed Source Reductions -
Part 2", Journal of Air Pollution Control Association, 24, 551.
Larsen, R. I., Heck, W. W. (1976). ""An Air Quality Data
Analysis System for Interrelating Effects, Standards and Needed
Source Reductions - Part 3, Vegetation Injury", Journal of Air
Pollution Control Association, 26, 325.
Larsen, R. I., Heck, W. W. (1977). "An Air Quality Data
Analysis System for Interrelating Effects, Standards and Needed
Source Reductions - Part 4, A Three Parameter Averaging Time
Model", Journal of Air Pollution Control Association, 27, 454.
Lioy, P. J., Watson, J. G. and Spengler, J. D. (1980). "APCA
Specialty Conference Workshop on Baseline Data for Inhalable
Particulate Matter", Journal of Air Pollution Control
Association, 30, 1126.
Lippman, M. (1980). "Position Statements on Particulate
Matter", ISO/TC 146 AD HOC Working Group, Gainsville, FL,
March 25, 1980.
Long, S. J., Suggs, J. C. and Walling, J. F. (1979). "Lead
Analysis of Ambient Air Particulates: Interlaboratory Evaluation
of EPA Lead Reference Method", Journal of the Air Pollution
Control Association, 29, 28.
Loo, B. W. and Jaklevic, J. M. (1974). "An Evaluation of the ERG
Virtual Impactor", LBL-2468, Report of Lawrence Berkeley Labs,
Berkeley, CA.
Ludwig, F. L., Kealoha, J. H. S. and Shelvar, E. (1977).
Selecting Sites for Monitoring Total Suspended Particulates,
EPA-450/3-77-018, Research Triangle Park, NC.
C-5
-------
Lundgren, D. A. and Paulus, H. J. (1975). "The Mass Distribution
or Large Atmospheric Particles", Journal of Air Pollution Control
Association, 25, 1227.
Lynch, A. J., McQuaker, N. R. and Brown, D. F. (1980).
"ICP/AES Analysis and the Composesition of Airborne and Soil
Materials in the Vicinity of a Lead/Zinc Smelter Complex", Air
Pollution Control Association Journal, 30, 257.
Mage, D. T. and Ott, W. R. (1978). "Refinements of the Lognormal
Probability Model for Analysis of Aerometric Data", Journal of
Air Pollution Control Association, 28, 796.
Mage, D. T. (1980). "The Statistical Form for the Ambient
Particulate Standard Annual Arithmetic Mean vs. Annual Geometric
Mean", Journal of Air Pollution Control Association, 30, 796.
Mamuro, T., Mizohata, A. and Kubota, T. (1979). "Elemental
Compositions of Suspended Particles Released from Various
Boilers", Annual Report of the Radiation Center of Osaka
Prefecture, 20, 9.
McFarland, A. R. and Rost, G. A. (1979a) "Design and Evaluation
of an Automated Aerosol Sampler", Texas A&M Air Quality
Laboratory Report 3397/01/79/ARM, College Station,Texas.
McFarland, A. R., Ortiz, C. A. and Rodes, C. E. (1979b).
Characteristics of Aerosol Samplers Used in Ambient Air
Monitoring", 86th National Meeting of the American Institute of
Chemical Engineers, Houston, TX, April 1-5.
McFarland, A. R., Ortiz, C. A. and Rodes, C. E. (1980a).
"Characterization of Sampling Systems", in Proceedings the
Technical Basis for a Size Specific Parituclate Standard Parts I
& II Specialty Conference, Air Pollution Control Association,
Pittsburgh, PA.
McFarland, A. R., Ortiz, C. A. and Meemong Kolkiat, B. (1980b).
"Characteristics of the ERT Aerosol Sampler in the SURE-TSP
Mode", Texas A&M University Air Quality Laboratory Report
4295/10/01/80/ARM, College Station, TX.
McKee, H. C., Childers, R. E., Saenz, 0., Stanley, T. W. and
Margeson, J. (1972). "Collaborative Testing of Methods to
Measure Air Pollutants - The High-Volume Method for Suspended
Particulate Matter", Journal of Air Pollution Control
Association, 22, 342.
Meserole, F. B., Schwitzgebel, K., Jones, B. F., Thompson, C. M.,
and Mesich, F. G. (1976). "Sulfur Dioxide Interferences in the
Measurement of Ambient Particulate Sulfates", EPRI Research
Project 262, Palo Alto, California.
C-6
-------
Meserole, F. B., Jones, B. F., Rohlack, L. A., Hawn, W. C.,
Williams, K. R., and Parsons, T. P. (1979). Nitrogen Oxide
Interferences in the Measurement of Atmosphere Particulate
Nitrates, EPRI Report EA-1031, Palo Alto, California.
Meyer, S. L. (1975). Data Analysis for Scientists and Engineers
John Wiley & Sons, New York.
Miller, F. J., Gardner, D. E., Graham, J. A., Lee, R. E.,
Wilson, W. E. and Bachman, J. D. (1979). "Size Considerations
for Establishing a Standard for Inhalable Particles", Journal of
Pollution Control Association, 29, 610.
Miller, J. D. (1980). "Middle Scale Distribution of Inhalable
Participates and TSP within a Metropolitan Area", Journal of Air
Pollution Control Association, 30, 1320.
Miller, S. (1981). "SURE Air Quality Data", Environmental Science &
Technology 15, 989.
Mueller, P. K., Hidy, G. M., Warren, K., Lavery, T. F., and
Baskett, R. L. (1979a). "The Occurrence of Atmospheric Aerosols
in the Northeastern United States". Conference on Aerosols:
Anthropogenic and Natural Sources and Transport, New York Academy
of Sciences.
Mueller, P. K., Fung, K. K., Heisler, S. L., Grosjean, D.
and Hidy, G. M. (1980). "Atmospheric Particulate Carbon
Observations in Urban and Rural Areas of the United States",
presented at the International Symposium on Particulate Carbon:
Atmospheric Life Cycle, General Motors Research Laboratory,
Warren, MI.
Mueller, P. K., Hidy, G. M., Baskett, R. L., Fung, K. K., Henry,
R. C., Lavery, T. F., Lordi, N. J., Lloyd, A. C., Thrasher, U.
W., Warren, K. K. and Watson, J. G. (1981). Sulfate Regional
Experiment (SURE); Report of Findings, Section 3,
Report EA-1901, Electric Power Research Institute, Palo Alto, CA.
Neustadter, H. E. and Sidik, S. M. (1974). "On Evaluating
Compliance with Air Pollution Levels 'Not to be Exceeded More
Than Once a Year1 ", Journal of Air Pollution Control
Association, 24, 559.
Nitrate by Cadmium Reduction Method, (1976). In Standard
Methods for the Examination of Water and Waste Water edited by A.
E. Greenburg and M. J. Taras, Washington, D.C., American Public
Health Association, p. 620.
Ortiz, C. A. (1978). "Aerosol Collection Characteristics of
Ambient Aerosol Samplers", Air Quality Laboratory Publication
3565/10/78/CAO, Texas A&M University, College Station, TX.
C-7
-------
Pace, T. G. (1979a). "An Empirical Approach for Relating Annual
TSP Concentrations to Particulate Microinventory Emissions Data
dn Monitor Siting Characteristics", EPA-45/4-79-012, Re£>earch
Triangle Park, NC.
Pace, T. G. (1979b). "Preliminary Characterization of 1P/TSP
Ratio", APCA Meeting, Cincinatti.
Pace, T. G. (1980). "Ambient Particulate Baseline Conditions -
Sources and Concentrations", Proceedings: The Technical Basis for
a Size Specific Particulate Standard, Part I & II Specialty
Conference, Air Pollution Control Association, Pittsburgh, PA.
Pace, T. J., Watson, J. G. and Rodes, C. E. (1981). "Preliminary
Interpretation of Inhalable Particulate Network Data", APCA
Meeting, Philadelphia, PA.
Pashel, G. E., Egner, D. R., Winkley, J. C. and Sistek, R. E.,
(1980). "Preliminary Results of AISI Particulate Matter
Monitoring Study", Proceedings: The Technical Basis for a Size
Specific Particulate Standard Parts I & II Specialty Conference,
p. 252, Air Pollution Control Association, Pittsburgh, PA.
Perlac, R. M. (1978). "Sulfate Regional Experiment in Northeastern
United States: The "SURE" Program". Atmospheric Environment,
.12, 641.
Perlac, R. M., Hilst, G. R., Mueller, P. K., Hidy. G. M.,
Lavery, T. F. and Watson, J. G. (1981). EPRI Sulfate Regional
Experiment: Results and Implications. EPRI Research Project
862-2, Report EA-2077-SY-6D, Palo Alto, California.
Phinney, D. E., Newman, J. E. (1972). "The Precision Associated
with the Sampling Frequencies of Total Particulate at
Indianapolis, Indiana", Journal of Air Pollution Control
Association, 22, 692.
Pierson, W. R. and Brachaczek, W. W. (1976). "Particulate
Matter Associated with Vehicles on the Road", SAE Automotive
Engineering Congress and Exposition, Detroit, Michigan,
February 23-27, 1976.
Pierson, W. R., Brachaczek, W. W. , Korniski, T. J., Truex, T.. J.
and Butler, J. W. (1980). "Artifact Formation of sulfate,
Nitrate and Hydrogen Ion on Backup Filters: Allegheny Mountain
Experiment", Journal of Air Pollution Control Association, 30, 30.
Price, J. H., Gise, J. P., Sievers, H. E., Ehlers, S. E. and
Knape, B. K. (1977). "Attainment Analysis: Volume I, Causes of
Nonattainment", Texas Air Control Board, Austin, TX.
Rahn, K. A. (1976). "Silicon and Aluminum in Atmospheric
Aerosols: Crust-Air Fractionation?", Atmospheric Environment, 10,
597. ~~
C-8
-------
Ranade, M. B. and Kashdan, E. R. (1979). "Critical Parameters
for the Federal Reference Method for the Inhalable Particulate
Standard", Final Report Technical Directive 222, EPA Contract
68-02-2720, Research Triangle Institute, Research Triangle Park,
NC.
Richard, G. and Tan, R. (1977). "An Implementation Plan for
Suspended Particulate Matter in the Phoenix Area Volume I, Air
Quality Analysis", EPA-450/3-77-021a, Research Triangle Park, NC.
Roberts, E. M. (1979). "Review of Statistics of Extreme Values
with Applications to Air Quality Data: Part I", Journal of Air
Pollution Control Association, 29, 632.
Roberts, E. M. (1979). "Review of Statistics of Extreme Values
with Applications to Air Quality Data: Part II, Applications",
Journal of Air Pollution Control Association, 29, 733.
Rodes, C. E. (1979). "Protocol for Establishment of a Nation-
wide Inhalable Particulate Network", USEPA/EMSL Report, Research
Triangle Park, NC.
Rodes, C. E. (1981). Personal Communication, EPA, North Carolina.
Schwitzgebel, K., Coleman, R. T., Collins, R. V., Mann, R. M.
and Thompson, C. M. (1978). "Trace Element Study at a Primary
Copper Smelter", EPA-600/2-78-065a, Research Triangle Park, NC.
Shah, J. J. (1981). "Measurements of Carbonaceous Aerosol
Across the U.S.: Sources and Role in Visibility Degradation",
Ph.D dissertation, Oregon Graduate Center, Beaverton, OR.
Shannon, L. J., Reider, J. P. and Cowherd, C. (1980). "Emission
Factors for Inhalable and Fine Particulates", Proceedings: The
Technical Basis for a Size Specific Particulate Standard, Parts I
& II Specialty Conference, p. 281, Air Pollution Control
Associaton, Pittsburg, PA.
Shaw, R. W., Stevens, R, K., Lewis, C. W. and Chance, J. H.
(1981). "Comparison of Aerosol Sampler Inlets", APCA Meeting,
Philadelphia, PA.
Spicer, C. W. and Schumacher (1977). "Interference in Sampling
Atmospheric Particulate Nitrate", Atmospheric Environment 11, 873.
Stevens, R. K., Dzubay, T. G., Russwurm, G. and Rickel, D. (1978).
"Sampling and Analysis of Atmospheric Sulfate and Related
Species", Atmospheric Environment, 12, 55.
Suggs, J. C., Rodes, C. E., Evans, G. E. and Baumgardner, R. E.
(198lb). "Inhalable Particulate Network Annual Report:
Operations and Data Summary (Mass Concentrations Only) April 1979
- June 1980", USEPA Environmental Monitoring Systems Laboratory
Report, Research Triangle Park, NC.
C-9
-------
Suggs, J. C., Burton, R. M., Pace, T. G., Himmelstein, L. and
Hauptman, F. (1981c). "Philadelphia Inhalable Particulate
Intensive Studies: Spatial Correlation in a Metropolian Area",
APCA Meeting, Philadelphia, PA.
Sweitzer, T. A. (1980). "Characterization of Passively Loaded
Particles on HIVOL Samples", Journal of Air Pollution Control
Association, 30, 1324.
Swinford, R. (1980). "The Assessment of Passive Loading Effects
on TSP Measurements in Attainment Areas", Journal of Air
Pollution Control Association, 30, 1322.
Taylor, S. R. (1964). "Abundance of Chemical Elements in the
Continental Crust: A New Table", Geochimica et Cosmochimica Acta,
28_, 1273.
Technicon Industrial Systems (1972). Tarrytown, NY, Technicon
Industrial Method No. 118-71.
Thanukos, L. C., Taylor, J. A. and Kary, R. E. (1977). "High
Volume Sampling Particulate Removal from Filter Surface by High
Winds", Journal of Air Pollution Control Association, 27,, 1013.
Trijonis, J., Eldon, J., Gins, J. and Berglund, G. (1980).
"Analysis of the St. Louis RAMS Ambient Particulate Data", Report
to EPA under contract 68-02-2931 Task 6, Research Triangle Park,
NC.
U.S. Department of Commerce (1979). Statistical Abstract of
the United States; 1979 (100th Edition), Bureau of the Census,
Washington, D.C.
U.S. Department of HEW (1969). Air Quality Criteria for
Particulate Matter, National Air Pollution Control Administration
Publication AP-49, Washington, D.C.
U.S. EPA (1972). "NEDS Source Classification Codes and
Emission Factor Listing", Environmental Protection Agency,
National Air Data Branch, Research Triangle Park, NC.
U.S. EPA (1978a). Directory of Air Quality Monitoring Sites
Active in 1977, EPA-450/2-78-048, Research Triangle Park, NC.
U.S. EPA (1978b). "OES Procedures for Determining Trace Elements
in 1978 NASN Samples", Document PSD/Be/01, Research Triangle
Park, NC.
U.S. EPA (1980a). "National Emission Data System Condensed Point
Source Listing for Particulates", EPA/OAQPS, Research Triangle
Park, NC.
C-10
-------
U.S. EPA (1980b). "National Emission Data System Source
Classification Codes and Emission Factor Listing - Appendix A
(2/80)", EPA/OAQPS, Research Trinagle Park, NC.
U.S. EPA (1980c). Inhalable Particulate Network Operations and
Quality Assurance Manual U.S. Environmental Monitoring Systems
Laboratory, Research Triangle Park, NC
U.S. EPA (1980d). "Inhalable Particulate Matter Network Valid
Data Averages Summary", report generated by EPA/EMSL July 25,
1980.
U.S. EPA (1981a). "Inhalable Particulate Matter Network Valid
Data Averages Summary from 10/79 to 9/80", report generated by
EPA/EMSL on August 14, 1981.
U.S. EPA (1981b). "Review of the National Ambient Air Quality
Standards for Particulate Matter: Draft Staff Paper", prepared
for U.S. EPA Clean Air Science Advisory Committee Meeting of
November 16-18, 1982, Research Triange Park, N.C.
U.S. House of Representatives (1977). "Clean Air Amendments of
1977", Conference Report, #95-564, August 3, 1977, U.S.
Government Printing Office, Washington, D.C.
U.S. Office of Management and Budget (1972). Standard
Industrial Classification Manual, U.S. Department of Commerce,
Office of Federal Statistical Policy and Standards, Washington,
D.C.
Von Lehmden, D. J., Jungers, R. H. and Lee, R. H. (1974).
"Determination of Trace Elements in Coal, Fly Ash, Fuel Oil and
Gasoline — A Preliminary Comparison of Selected Analytical
Techniques", Analytical Chemistry, 46, 239.
Walling, J. (1981). USEPA/EMSL, Personal Communication,
Research Triangle Park, NC.
Watson, J. G. (1979). Chemical Element Balance Receptor Model
Methodology for Assessing the Sources of Fine and Total Suspended
Paritculate Matter in Portland, Oregon, Ph.D. dissertation,
Oregon Graduate Center, Beaverton, OR.
Watson, J. G., Mueller, P. K., April, S., Chow, J., Fung, K.,
Gendron, L., James, P., Jones, R., McFarlane, I., Moore, E.,
Nuesca, B., O'Toole, D., Putney, K. and Warren, K. (1981).
Eastern Regional Air Quality Study; An Extension of the SURE
Measurements, Report EA-1914, Electric Power Research Institute,
Palo Alto, CA.
Wedding, J. B., McFarland, A. R. and Cermak, J. E. (1977).
"Large Particle Collection Characteristics of Ambient Aerosol
Samplers", Environmental Seience & Technology, 11, 387.
C-ll
-------
Wedding, J. B., Weigand, M., John, W., and Wall, S. (1980a).
"Sampling Effectiveness of the Inlet to Dichotomous Sampler",
Environmental Science and Technology, 14, 1367.
Wedding, J. B. (1980b). "Ambient Aerosol Sampling: History,
Present Thinking, and a Proposed Inlet for Inhalable Particulate
Matter", in Proceedings, the Technical Basis for a Size Specific
Particulate Standard, Parts I & II Specialty Conference, Air
Pollution Control Association, Pittsburgh, PA.
Weisz, P. B. (1972). "Finagle's Laws: The Present Official
Formulation", Journal of College Science Teaching, 2^, 16.
Wendt, J. G. and Torre, K. J. (1981). "Field Test of Four Size-
Segereated Samplers", APCA Meeting, Philadelphia, PA.
Whitby, K. T. (1978). "The Physical Characteristics of Sulfur
Aerosols", Atmospheric Environment, 12, 135.
Whitby K. T. and Sverdrup G. M. (1980). "California aerosols:
Their Physical and Chemical Characteristics", in The
Character and Origins of Smog Aerosols edited by Hidy G. M.
Mueller P. R. Grosjean D., Appel B. R. and Wesolowski J. J.
published by John Wiley & Sons.
Willeke, K. and Whitby, K. T. (1975). "Atmospheric Aerosols:
Size Distribution Interpretation", Journal of the Air Pollution
Control Association, 24, 529.
Zoller, W. H., Small, M., Germani, M. and Moyers, J. L. (1978).
"Atmospheric Trace Elements Emissions from Copper Smelters",
presented at Division of Environmental Chemistry, American
Chemical Society, Miami, Florida.
C-12
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO.
EPA 450/4-81-035
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Analysis of Inhalable and Fine
Particulate Matter Measurements
5. REPORT DATt
December 1981
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
John G. Watson, Judith C. Chow and
Jitendra J. Shah
8. PERFORMING ORGANIZATION REPORT NO.
ERT Document No. A394-140
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Research £ Technology, Inc.
696 Virginia Road
Concord, MA 01742
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-2542 Task 6
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. EPA
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Total, inhalable and fine particulate matter measurements acquired by EPA's
Inhalable Particulate Network in 1979 and 1980 are summarized and analyzed in this
report. The theoretical collection efficiencies of different samplers were calcu-
lated and compared to each other and to an acceptable performance range. The
measurement processes and several of the major urban airsheds of the IP Network are
described. The spatial, temporal and statistical distributions of these measure-
ments are examined. A receptor-oriented model relating IP to TSP is derived and
tested for prediction accuracy under various situations. A mass balance receptor
model is applied to IP and FP chemical concentrations in four urban areas to
estimate the contributions of various emissions source types to ambient mass
concentrations.
17.
a.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Particulate Matter
Inhalable Particulate
b.IDENTIFIERS/OPEN ENDED TERMS C. COSATI l;ield/Group
Fine
Inhalable
18. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21 NO OF PAGES
334
20 SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
-------
c:
8
Do ~3
3 Jn
^ CD ;-*
-.tJ cr
:r° -D m -n TO
2° 2 ? S I
£2 83"S
A-
01
: £ D Q]
J to ~
•3 o
-Q.3
S
SL
------- |