HAZAROOUS AIR POLLUTANTS
A Preliminary Exposure and Preliminary Rick Appraisal
for 35 U.S. Counties
Prepared for:
U.S. Environmental Protection Agency
Office of Policy Analysis
401 M Street, SW
Washington, D.C. 20460
Prepared by:
Versar Inc.
6850 Versar Center
Springfield, Virginia 22151
and
American Management Systems
1777 N. Kent Street
Arlington, Virginia 22209
Contract # 68-01-6715
September, 1984
-------
TABLE OF CONTENTS
Page No.
1.0 INTRODUCTION 1-1
1.1 Background T-l
1.2 Overview of Methodology 1-2
1.2.1 Selection of 35 Counties 1-2
1.2.2 Selection of Pollutants 1-3
1.2.3 Source Characterization 1-3
1.2.4 Exposure and Risk assessment 1-4
1.3 Limitations of the Study 1-5
1.3.1 Oata Limitations 1-5
1.3.2 Limitations of Models 1-6
1.3.3 Limitations of Potency Factors 1-6
1.4 Outline of the Report 1-6
2.0 METHOOOLOGY 2-1
2.1 Phase I (Screening) 2-2
2.1.1 Pollutant Selection 2-2
2.1.2 Source Screening 2-5
2.1.3 Risk Screening 2-22
2.2.1 Refinement of Phase I Input Oata 2-26
2.2.2 Dispersion Modeling 2-27
3.0 CONCLUSIONS 3-1
3.1 General Findings for the 35 Counties 3-1
3.2 Cancer Risk Associated with Various Source
Categories 3-3
3.2.1 Road Vehicles 3-7
3.2.2 Gasoline Marketing (Service Stations) 3-7
3.2.3 Solvent Usage 3-7
3.2.4 Waste 011 Burning 3-7
3.2.5 Residential Wood Combustion and POTWs
(Sewage Treatment Plants) 3-12
3.3 Variation 1n the Significance of Sources and
Pollutants Across Geographic Areas 3-15
3.4 Findings on Risk to the Most Exposed Individuals 3-17
4.0 REFERENCES 3-27
-------
LIST OF TABLES
Page No.
Table 2-1 Hazardous A1r Pollutants Evaluated 1n
the 35 County Study 2-4
Table 2-2 Example of NEDS Point Source Report 2-6
Table 2-3 Example of NEDS Area Source Report 2-9
Table 2-4 Example of Special NEDS Report NE099 2-17
Table 2-5 Th1rty-f1ve Counties Selected for Phase II 2-25
Table 3-1 Summary of Estimated Lifetime Cancer
Incidence for 35 Counties 3-2
Table 3-2 Percent of Risk Associated with Point and Area
Sources 1n 35 Counties 3-5
Table 3-3 Summary of Cancer Incidence for 35 Counties 3-16
Table 3-4 Variation of Lifetime Cancer Incidence-, by Chemical,
for 6 Counties 3-24
Table 3-5 Most Exposed Individuals 3-26
-------
LIST OF FIGURES
Page No.
Figure 2-1 Modeling Receptor Pattern 2-30
Figure 3-1 Cancer Incidence By Source Category 3-4
Figure 3-2 Incidence as a Function of Pollutant 3-6
Figure 3-3 Incidence due to Road Vehicles 3-8
Figure 3-4 Incidence due to Gasoline Marketing 3-9
Figure 3-5 Incidence due to Solvent Usage 3-10
Figure 3-6 Incidence due to Waste 011 Burning 3-11
Figure 3-7 Incidence due to Woodburnlng 3-13
Figure 3-8 Incidence due to POTW Volatilization 3-14
Figure 3-9 Principal Sources of Risk for Baltimore County 3-18
Figure 3-10 Principal Sources of Risk for Cook County 3-19
Figure 3-11 Principal Sources of Risk for Harris County 3-20
Figure 3-12 Principal Sources of Risk for Los Angeles County....3-21
Figure 3-13 Principal Sources of Risk for Philadelphia County.. .3-22
Figure 3-14 Principal Sources of Risk for Sullivan County 3-23
-------
1.0 INTRODUCTION
1.1 Background
This report presents the results of a preliminary effort to
characterize the nature and extent of the hazardous air pollution problem
in 35 counties. It 1s one element of a larger EPA study, whose goal was
to define the national scope of the health Impacts of toxic air
pollutants. The technical approaches used in the effort described here
were Intended to complement those of the EPA study, and to fit Into its
overall six month schedule.
The 35 county study builds on the concepts of the Integrated
Environmental Management Division's (IEHD) geographic demonstration
projects, which emphasize analysis of geographies variations in toxic
chemical problems. The principal questions of concern here are:
• Which pollutants and sources contribute most to the hazardous air
pollutant problem?
• Do the sources and pollutants with the greatest potential impact
on health risk vary significantly from region to region?
Different areas of the country have different meteorological
conditions, topography, and source mixes. These factors are
presumed to 3ffect exposure in each area ~ the question is, how
significant are these variations? If geographic variations are
found to be significant, this could affect the design of future
air toxics control strategies.
• How do point sources (e.g.. Industrial plants) compare in
importance to areawide or dispersed sources (e.g. road vehicle
emissions, heating, and degreaslng operations)? Recent hazardous
air pollutant regulatory strategies have focused primarily on
large Industrial point sources. Experience in IEM0 geographic
studies suggests that the cumulative effect of mobile or dispersed
sources may be similar In Importance to major industrial point
sources.
Are sources not traditionally considered under Section 11? of the
Clean Air Act, such as hazardous waste facilities, or releases of
volatile organic compounds from sewage treatment plants,
significant sources of air toxics?
1-1
-------
This study could not attempt to answer these questions' definitively,
both because of the time and resource constraints of the study, and
because of the quality of the data available at this time. The purpose
was to Interpret the available data, compensating as much as possible for
Its known deficiencies, 1n order to give additional perspective on this
complex Issue to EPA policy makers. This report 1s not Intended for use
1n setting a regulatory agenda, nor does it support any rulemakings
currently under way. Its purpose 1s to characterize potential problems,
not to deal with the controllabll1ty of any problems, known or
suspected. The general goal was to support EPA's intent to Investigate
the air toxics situation 1n a broad and comprehensive way.
1.2 Overview of Methodology
. This study evaluated data from all counties nationwide, in order to
select 35 counties with high expected potential air toxics exposures. A
high!y generalized exposure and risk evaluation was performed for these
counties, based on a small subset of 21 hazardous air pollutants, and
using established approaches for evaluating health risks. Only existing
data on air toxics potencies and emissions were used; we generated no new
data.
Cancer was selected as the single health endpolnt of concern --
1n this and 1n the companion studies — because 1t 1s of wide public
concern, and because It can be an Issue for toxics 1n ambient air at the
concentrations commonly encountered 1n most urban environments. We made
no attempt to evaluate other known adverse health effects.
The major analytic components of the analysis are as follows:
1.2.1 Selection of 35 counties
The study began by gathering data on air pollutant emissions
nationwide. Although some very useful and detailed studies exist for
particular subreglons and counties, the objectives of the study required
uniform and consistent coverage. The screening analysis therefore
depended entirely on national level data.
1-2
-------
The 35 counties selected for more intensive analysis were chosen from
counties with the highest expected ambient exposures, based on high
populations and large aggregate emissions from all source categories. To
gain additional perspective, some counties were Included, even though
their populations and aggregate emissions were not the highest, because
they contain large industrial point sources of potential interest. The
possibility of selecting 3 stratified random sample was considered, but
the approach was not pursued further because the criteria for
stratification would themselves be debatable, and the time and funds
available were limited.
The 35 counties contain roughly 20% of the total U.S. population, 20%
of total releases of volatile organic compounds, and 10% of total
particulate matter loadings. They represent a variety of industrial and
population distributions, but are not a statistically representative
sample of the country. As in any case study analysis, conditions in
these counties cannot be extrapolated to the country as a whole.
1.2.2 Selection of Pollutants
The study looked at estimated releases of 21 compounds. These
chemicals were chosen from a combined list of the 37 pollutants proposed
for listing under Section 112, and 50 additional substances Identified by
EPA's Office of A1r Quality Planning and Standards (OAQPS). Selections
were made based on the following criteria:
• Sufficient evidence of carcinogenicity (based on ORG criteria)
•- Significant release rates
Readily available emissions information
1.2.3 Source characterization
The analysis focused on three categories of sources: point, area,
and "non-trad1t1onal." Point sources are generally large Industrial
facilities, such as refineries or power plants. Area sources include
many small dispersed facilities, such as dry cleaning establishments or
1-3
-------
degreasing operations, as well as road vehicles. "Non-traditional"
sources Include a group of both point and area sources that are singled
out because they have not received significant regulatory attention as
sources of air toxics 1n the past. Examples Include sewage treatment
plants waste oil combustion and residential wood burning.
Once the 35 counties were identified, the national-level emission
estimates used for screening were supplemented with more accurate
estimates to the extent possible. Emission estimates for point and area
sources were developed using several techniques. Where possible, we
relied on plant-specific data and EPA documents indicating emission rates
from specific source categories. Where this was not possible, surrogate
loadings were developed using VOC loadings shown 1n the National
Emissions Data Systems (NEOS) (the data base originally used for
screening) broken down into specific toxic compounds through the use of
apportioning factors. F.or the non-traditional souroes, we constructed
special algorithms to estimate emissions based on generic data.
1.2.4 Exposure and risk assessment
A preliminary exposure and risk assessment was completed for the 35
counties to estimate excess cancer Incidence attributable to hazardous
air pollutants based on ambient exposures. Concentrations of the 21
pollutants 1n the 35 counties were estimated using atmospheric dispersion
modeling. Aggregate excess cancer Incidence was calculated by overlaying
the modeled pollutant concentrations on the exposed population located 1n
each county. EPA's Office of Toxic Substances' model, GAMS, was used In
this effort.
Potency estimates (unit risk factors) for most of the selected
compounds came from EPA's Carcinogen Assessment Group (CAG), the other
from work completed by Clement Associates. Methods used for estimating
the cancer Incidence followed standard EPA procedures. No effort was
made to judge the appropriateness of these analytic techniques.
1-4
-------
1.3 Limitations of the Study
Although this report presents quantitative estimates of excess cancer
Incidence, the apparent precision of the numbers may be misleading. The
usefulness of the findings 1s tn .the relative significance of different
sources and pollutants 1n different regions of the country — not in the
absolute numbers. We believe that the aggregate exposure patterns and
the rankings of Individual substances are reasonable. The results shown
In the document help provide Insight Into these relationships, but 1n
absolute terms the results may not be particularly Informative.
1.3.1 Oata Limitations
As stated above, the work relied entirely on existing health,
pollutant, and emissions data. The pollutants studied do not represent a
complete set of potentially hazardous air pollutants. While the
pollutants were chosen from acknowledged priority lists, the selection
was biased toward the better documented compounds. Ethyl benzene, for
example, was excluded because there are no available health data on its
effects. 01ox1n, a known human carcinogen, was omitted because knowledge
of sources and emissions is still poor. The selected pollutants account
for an average of roughly.8% of the total particulate and volatile
organic releases 1n the 35 target counties. One can only speculate about
the potential environmental hazards posed by the remaining pollutants.
Lack of data also caused us to omit a number of sources that may, 1n
fact, contribute significantly to human exposure. For example, there was
Insufficient information on releases from non-trad1t1onal sources such as
Superfund sites, hazardous waste treatment, storage and disposal
facilities (TSDFs), and hazardous waste combustion in boilers. In
addition, for some compounds we were unable to characterize certain types
of facilities or area source categories based or the available data.
Data, limitations are particularly Important where we had to rely or
national data to supply county-specific estimates. The quality of the
NEDS data system, in particular, has acknowledged Hm1tat1ons--1t
1-5
-------
contains outdated Information, and has uneven quality control. An attempt
*as made to correct seme obvious Inaccuracies and data omission:, but
deficiencies certainly remained. None of the experts vie consulted an
this problem suggested a better alternative to NEOS.
1.3.2 Limitations of Models
A number of simplifying assumptions had to be made 1n the exposure
modeling effort. There was not attempt, for example, to model all the
stacks at an Individual facility; all stack emissions were modeled
together using one set of stack specifications typical for the plant.
Certain mlcrcscale modeling factors, such as building downwash effects,
were not Included. Differences between Indoor and outdoor exposures were
not evaluated; exposures were calculated as a function of 24-hour a day
outdoor exposure. These slmpllfVzatl-MVi tto not, In our opinion,
seriously distort the relative rankings of sources and pollutants, but
they probably do have an effect on the absolute magnitude of the
calculated effects, creating overestimates of some effects and
underestimates of others.
1.3.3 Limitations of Potency Factors
The potency factors used here—developed by EPA's CAG and by Clement
Assoc1ates--are conservative upper bound estimates. The true value? are
unlikely to be higher, but could be substantially lower. As a result,
estimates of cancer Incidence presented In this document may overstate
the actual situation for specific sources and pollutants modeled.
1 • 4- Outline of the Report
This report Is divided in two sections. Section 1 provides a
detailed treatment of the methodology. This Includes a technical
discussion of the first screening phase of the study, I.e. selection of
pollutants and sources, generation of emissions estimates, and selection
of the 35 counties—and goes on to describe the detailed modeling and
exposure and risk assessment performed for the selected subset. Section
1-6
-------
2 presents the findings for all 35 counties, separately and 1n the
aggregate. Information 1s presented on pollutant and source
contributions, Importance of non-traditional sources, and the
significance of geographic variation.
References follow these two sections. Appendices documenting the
technical components and findings are also attached.
1-7
-------
2.0 METHODOLOGY
The major objectives of this study were to: (1) provide Insight on
the magnitude of the air toxics problem; and (2) explore the nature of
the situation, Including explicit consideration of nontradltlonal sources
and geographic variation. The Indicator chosen to evaluate the
potential severity and dimension of the problem was estimated cancer
Incidence associated with inhalation of ambient concentrations. The
findings from this report were also designed to form one of the
quantitative bases for a six month EPA study 1n hazardous air pollutants
entitled, "The Magnitude and Nature of the A1r Toxics Problem In the
United States." Given the ambitious scope of the project, the tight
scheduling of the EPA analysis, and the limited funds, it was decided to
focus on a subset of pollutants, sources and geographic areas
(35 countries). The first phase of the study concentrated on gathering
all uniformly available Information nationwide and ranking counties
without the use of dispersion modeling. The second phase Involved
*
refining the emission estimates for the 35 counties and performing an
exposure and risk assessment based on ambiant data to estimate cancar
incidence.
This chapter of the report summarizes the methodologies employed in
the study. The first section focuses on the technical work undertaken in
Phase I; the second section describes the Phase II activities. Detailed
discussions and documentation of the analytical components of the
analyses can be found 1n Appendices A through J.
Appendix A 1s a technical description of the data management system,
HEMIS (Hazardous Emission Model and Integrated System), used to support'
both phases of the study. This system contains all emissions data, risk
Indices and exposure coefficients for the project. Appendix J provides
the raw data printouts from HEMIS.
2-1
-------
2.1 Phase I (Screening)
The objective of Phase I was to rank all U.S. counties on the basis
of relative exposure and potential human health risk from a selected list
of pollutants, and to Identify a subset of counties for more Intensive
study In Phase II. The National Emissions Data System (NEDS) was used as
the basic data set for bath Phase I and Phase II. NEDS, maintained by
EPA's Office of Air Quality Planning and Standards, stores and reports
emissions related data for five criteria air pollutants. Including
volatile organic compounds (VOCs). NEDS was selected for this study
because 1t provides sufficient uniform Information to estimate air toxics
emissions for both point and area sources at the county level
nationwide. Despite acknowledged limitations with NEDS, Including
miscoded or outdated emissions data, the use of this data base Is
considered to be a suitable for meeting the objectives of this study.
Appendix C presents an overview of NEDS.
. Every effort was made to compile as comprehensive a file as possible
on sources of the selected pollutants for every U.S. county, but as
described 1n the following discussion, certain categories of sources had
to be omitted from the Phase I ranking effort. Significant additional
source data were later developed for the selected counties 1n Phase II,
but this effort, too, was subject to some Important limitations (see
Section 3.2).
2.1.1 Pollutant Selection
The number of noncrlterla air pollutants of- concern across the
country Is potentially very large. A small subset of pollutants were
selected that met the following criteria:
• Availability of pollutant emission factors. VQC (volatile organic
compound) apportionment factors, or point source emission rates.
For the mast part, data on emissions of hazardous air pollutants
are not available. Consequently, when possible, pollutant
loadings were estimated based upon knowledge of other pollutant
classes (such as VOC) which have been previously reported as part
of regular program operations. When data were insufficient to
2-2.
-------
estimate emissions of the most significant sources of a candidate
compound, these pollutants were excluded from the study. It
should also be noted that, for the pollutants eventually selected
for this study, 1t was not always possible to address all
sources. An example 1s dloxln, which was not examined because no
reliable emissions factors could be found for the numerous
combustion processes that release dloxln.
• Available health effects data. Primary emphasis was placed on
compounds with sufficient evidence of carcinogenicity, since this
1s the most widely studied and best quantified health effect.
This permits a direct comparison of risks of different pollutants
1n terms of estimated cancer Incidence. These factors were
provided by EPA's Cancer Assessment Group and are provided 1n
Appendix B.
• Likelihood of the substance being present 1n the atmosphere 1n
significant quantities. Many hazardous air pollutants are emitted
to the atmosphere 1n such small amounts, or are so readily
degraded 1n the air, that they would not be expected to be a
nationally widespread hazard. Only compounds suspected to present
a generalized risk to a significant population were considered as
suitable targets for Investigation 1n this study. For example,
toluene d11socyanate (TO I) which was emitted In extremely low
quantities to the air during production and processing (LaShelle
and Smith,. 1974) was omitted from this study.
The pollutants selected were drawn primarily from the 11st of 37
proposed NESHAP pollutants, and an OAQPS list of 50 additional
chemicals. The final list of twenty-one compounds Included 1n this
analysis 1s found 1n Table 2-1. Coke Oven emissions and gasoline vapors
(from gasoline marketing) were not characterized until Phase II.
The 11st of pollutants Includes both metals and organlcs. While It
1s not possible to give precise figures on what proportion of total
national hazardous air pollutant emissions these represent, 1t 1s
Important" to note that we later estimated that the list of organlcs
constitutes only about 5.7 percent of total VOCs as listed by NEDS.
While not all VOCs are expected to prove carcinogenic, 1t 1s highly
likely that other species of organlcs whose emissions are Included 1n the
NEDS' VOC category may pose significant human health risks.
2-3
-------
Table 2-1. Hazardous Air Pollutants Evaluated in the 35 County Study
Acrylonitrile
Arsenic
Ber>zo(a)Pyrene
Benzene
Beryl liun
1,3-8utadiene
Cadmium
Carbon Tetrachloride
Chloroform
Chrcmiim (total)
Coke Oven Emissions
Ethylene Dibrcmide
Ethylene Oichloride
Formaldehyde
Gasoline Vapors
Nickel (total)
Pentach1oropheno1
Perch1oroethy1ene
Styrene
Tri ch1oroethy1ene
Vinyl Chloride
2-4
-------
Several conimonly recognized hazardous air pollutants were not
selected for Inclusion. For Instance, asbestos and dloxln, although
recognized carcinogens, were not Included because emissions, from both
point and area sources, are poorly understood and Inadequately
documented; Radionuclides were not included In the study for similar
reasons. Appendix 0 presents the rationale for pollutant selection.
2.1.2 Source Screening
Industrial Point Sources
NEDS was used 1n the Phase I screening to Identify major point
sources of air toxics. An example of a NEQS point source printout Is
presented In Table 2-2. For each point source, NCOS lists the name,
address, cartographical (UTM) coordinates, and Standard Industrial
Classification (SIC) code. Data are broken down by process and emission
points within the facility. Each process Is assigned a Source
Classification Code (SCC). VOC data were obtained from NEDS for each
emission point within the facility.
Apportionment factors were then applied to the VCC amount listed in
NEDS for all facilities. Later, 1n Phase II, for modeling purposes,
point sources emitting less than 100 tons per year of VQC were treated as
area sources. Because of the wide variability In metals emissions from
source to source, allocation factors for emissions of metals from point
sources were not available and were not Included 1n Phase I. Metals
releases from particular facilities such as power plants, however, were
considered 1n Phase II. .
The apportionment factors were largely based upon work done under the
Northeast Corridor Regional Modeling Project (NECRMP). (The NECRMP was a
project undertaken by EPA's Office of Air Quality Planning and Standards,
the northeastern States, local planning agencies and Metropolitan
Planning Organizations to develop regional and urban ozone control
strategies through the use of photochemical air quality simulation
models.) Factors to apportion species were developed that break down VOC
2-5
-------
Example of NEOS Point Source Report
N f D S POINT SOURCE LISTING
NEDS-USEU FILE CREATED Oil
S1ATEI 21 >: MARYLAND
couiiiYioiaor- Baltimore
AQCR(1I5>: METROPOLITAN BALTIMORE
PLANT ID- 0021 PDIIIT ID: 02
HAHE-AOORESS: CARR-LOHREY CLASS 2201 KLOMAH ST 21230 SIC022I1: GIASS CONTAINERS
PERSONAL coin ACT; J J FELOIIAIill SCCI 3-05-01<»-021 : MINERAL PRODUCTS -GLASS MFC
HOMO AY MAT 16, 19B1 PAGE 320
CITY<0120): BALTIMORE
-CONTAINER GLASS -MELTING FURNACE
! GENERAL INFORIIA1 I Oil
UTI1 GRID COORDINATES
! tIAIID CALCULATED POINT EMISSIONS
DATE OF
LAST UPOATE: 1980
OWNERSHIP: PRIVATE
1PP PROCESS:
SOURCE: PROCESS
MQHMAL OPERATIONS
HHkHtiHHMHHHMKIIHNII
IIOUHS/DAY: 24
DAIS/WEEK: 7
WEEKS/YEAR: 51
7. ANNUAL THROUGHPUT
NMMNNriMHUNNIfMNNNIINII
OEC-FEB: 25 V.
MAR-MAY: 25 '/.
JUNE-AUG: 25 /.
SEIM-UOV: 25 7.
!
! SPACE IIEAT: 00.0 /.
\
COMPLIANCE INFO
hfcHMNMMMMMNMNKtf
VARIANCE ISSUED
SCHEDULED
J COIII'LNCE DATE: /
«
! COMPLIANCE STATUS
! UPDATE: 01/10/75
EMERGENCY COIHROL
AC I ION PLAII
NOT REQUIRED
UTrt iONE: IS
HORIZONTALS
VERTICAL:
359.<, KM !
-------
emissions for each type of source, as Identified by SCC code, Into Its
various chemical constituents. The Information was compiled for 138
Individual organic pollutants and 946 distinct SCC codes.
There are two major limitations of this approach:
1. The distribution of VOC Into speclated components from a specific
SCC code 1s highly variable from facility to facility. Although
generic data may be useful in the national perspective, process
differences between plants create substantial discrepancies
between specific releases and generic predictions. This needs to
be considered when Interpreting data on the county, and
particularly, the plant level.
2. The data In NEDS are often outdated or simply Incorrect. Coding
mistakes can result 1n orders of magnitude errors 1n the VOC value
used as the basis from which to apportion organlcs. In some
cases, facilities that are out of business still appear 1n the
data base. Stack specifications, needed for air dispersion
modeling In Phase II, often proved unreliable or nonexlstant, and
were adjusted.
In the 35 county analysis (Phase II), an effort was made to resolve
data base anomalies In NEDS for key sources. For example, VOC totals
were reviewed to delete or revise values that were obviously In error.
During Phase I screening; however, this was not feasible because of the
large number of sources Involved.
The potential for systematic errors was also present. For example,
vinyl chloride emissions were found to be higher across the board, based
on the apportioned NEDS data, than more specific inventories that were
conducted by local regulatory agencies. This could be attributed either
to a. systematic overpredlctlon of VOC, and/or to the unrealistic
assumption that 100 percent of the VOC emitted 1s vinyl chloride.
Area Sources
Although the Impact of large point sources can be considerable, air
emissions from area sources can also pose significant environmental
risk.. Relatively small stationary and mobile sources may cumulatively
account for the- majority of air emissions 1n a geographical area.
1-7
-------
In order to gain additional 1ns'1ght Into the overall contribution to
estimated cancer Incidence due to these area sources, county-wide
emissions of the study pollutants were estimated for the following
sources:
1. Road vehicles
• light duty gasoline
• heavy duty gasol1ne
• heavy duty dlesel
2. Gasoline marketing (service stations)
3. Solvent usage
• degreaslng
• surface coating
• printing and publishing
• miscellaneous Industrial uses
4. Heating
residential - wood, oil, coal, gas;
o Industrial - oil, coal, gas;
• commercial - oil, coal, gas.
These sources were selected because they are known emitters of one
or more of the study pollutants, and because emission factors were
readily available to estimate loadings to the air. Ourlng Phase II (see
Section 2.2), the area source Inventory was expanded to Include
emissions of pentachlorophenol from preserved wood (used 1n fenceposts,
landscape timbers, wooden patio decks, etc.) and cooling towers.
All of these area source emission estimates were based on data
provided 1n NEDS. Table 2-3 provides an example of the NEDS Area Source
Report, which' was utilized 1n many of the emission estimates. For the
Phase I scan, 1f NEDS reported a zero value or no data, for a category
(e-.g-. solvent consumed 1n drycleanlng, residential coal consumption, or
vehicle- miles traveled), estimated emissions from that source were
calculated to be zero. Because of the size of the Phase I data base It-
was not feasible to Identify these data gaps for each area source
category In each county. During Phase II, data gaDS for the 35 county,
were filled, when practical, using alternative methods (see Section 2.2).
2-8
-------
Tabie c-3
Example of NEDS Area Source Report
)0-26-62 NATIONAL EMISSIONS DATA S U T f II PAGE 0009-2
AREA SOURCE EMISSIONS REPORT fOK *£AR 1900 DATA FILE LAST UPDATED OM JULY 21. 1962
STATEI11): HEM JERSEY COUHTYI0660 I: BURLINGTON CO AMCRI045): METROPOLITAN PHILADELPHIA
l
f u f (. CONS
U M P T I 0 N
E M I S S
I 0
N S
IN TONS
P E
R YEAR
PART
s
0 X
N 0 X
V
O C
C O
fiESlDEIfl JAt
155
516
670
52
911
ANTHRACITE COAL
3,000
tcns
l'»
41
5
15
115
BITUMINOUS COAL
0
TC!IS
0
0
0
0
0
MOOD
6.900
TOIIS
66
0
1
6
621
DISTILLATE Oil
14,240
THOUSAND GALLONS
41
491
106
12
66
RESIDUAL OIL
0
THOUSAND G^LIOliU
0
0
0
0
0
NATURAL CAS
7.070
MILLION CU'JIC FEE 1
11
2
154
19
71
COmEHCIAL AID INSTITUTIONAL
155
790
557
11
74
ANTHRACITE COAL
1.020
TONS
5
14
6
0
0
BI1UM1NOUS COAL
0
TOIIS
0
0
0
0
0
MOOD
0
TONS
0
0
0
0
0
DISTILLATE OIL
11,600
THOUSAND GALLONS
14
199
lie
2
15
RESIDUAL OIL
14,510
THOUSAND CALLOUS
136
577
199
6
16
NATURAL GAS
260
MILLIOi! CUBIC FEET
0
0
14
1
1
IIOUSIRIAL
4
42
57
3
12
ANTHRACITE COAL
0
TONS
0
0
0
0
0
BIIU1INOUS COAL
0
TONS
0
0
0
0
0
COKE
0
TOIIS
0
0
0
0
0
WQUO
0
TOIIS
0
0
0
0
0
DISIILLATE OIL
2,920
THOUSAND GALLONS
4
42
47
1
11
RESIDUAL OIL
0
THOUSAND GALLONS
0
0
0
0
0
NATURAL GAS
70
MILLION CUBIC FEET
0
0
10
2
1
PROCESS GAS
0
MILLION CUOIC FEET
0
0
0
0
0
TOTAL FUEL COIISUMPTION
114
1
,166
1,264
66
999
AIUHRACIIE COAL
4,020
TONS
20
55
U
15
115
BIlUtllNOUS COAL
0
TOIIS
0
0
0
0
0
COKE
0
TONS
0
0
0
0
0
MOOD
6.900
TOIIS
66
0
1
6
621
DISTILLATE OIL
50,960
THOUSAND GALLONS
61
714
491
15
112
RESIDUAL OIL
14.510
THOUSAND GALLOPS
116
577
199
6
16
HAIURAL GAS
7,420
MILLION CUHIC FEET
11
2
176
22
75
PROCESS GAS
0
MILLION CUOIC 1 LET
0
0
0
0
0
FUEL CHARACTERISTICS:
ANIHRAC1TE COAL 0.7 7. SULFUR, II. I Z ASH OlST (LLATE OIL 0.2 /. SULFUR
BITUMINOUS COAL 7.0 /. SULFUR, 11.0 I: ASH RESIDUAL OIL 0.5 X SULfUH
-------
Table 2-3 (continued)
}o-26-ez national emissions data s u t t n
AREA SOURCE ^MISSIONS REPORT fOR YEAR 1980 DATA HIE LAST UPDAIED OH JULY 23, 1982
STATE I }1 \ • IIEM JERSEY COUNTY! 0660 I: BURLINGTON CO AQCHI045>: METROPOLITAN PHILADELPHIA
PACE 0009-J
3OI.|0 HASTE OISPOSAL
PfSlDEMflAL
Otl SITE lt 224
COtntHClAL AND INSTITUTIONAL 11.800
OH SITE INCINERATION 11,800
OPEN BURNING 0
45
45
0
1*
14
0
21
0
26
0
61
61
0
INDUSTRIAL
ro
I
ON SITE INCINERATION
OPEN BUHHIHG
TOTAL SOLID UASTF OISPOSAL
on site incineration
OPEN BURNING
300
300
0
39,320
12,720
26,600
343
57
286
25
14
11
94
21
73
442
56
386
1.372
148
1.224
-------
Table 2-3 (continued)
}0-26-« national emissions
AREA SOURCE EMISSIONS REPORT FOR YEAR 1900
STATE!311: "EH JERSEY COUHiri 0660 I: BURLINGTON CO
DATA SYSTEM
DATA FILE LAST UPDATEO Oil JULY 23, 1902
AQCR(045>: METROPOLITAN PHILADELPHIA
PAGE 0009-'.
tiPTPP VEHICLES
EMISSIONS IN TOMS PER YEAR
ro
i
6AS0LIIIE POWERED VEHICLES
highway Vehicles
LIGHT DUTY VEHICLES
LIHITED ACCESS
RURAL
SUBURBAN
URBAN
LIGHT DUTY TRUCKS
LIMITED ACCESS
RURAL
SUBURBAN
URBAN
HEAVY DUTY VEHICLES
LIMITED ACCESS
RURAL
SUBURBAN
URBAN
off-highway vehicles
GALLONS FUEL
I THOUSAIIOS >
139.074
137.791
112.*25
17,847
7,519
CALCULATED
VEHICLE MlttS
I THOUSANDS I
1,992,496
1,725,720
0
431,410
0
1,294,290
223,091
0
55,773
0
167,310
43,665
0
10,921
0
32.764
PART
fi,922
a.916
7,663
0
637
0
7,046
996
0
03
0
913
237
0
27
0
210
SOX
311
30a
247
0
62
0
155
44
0
11
0
33
17
0
4
0
13
1,263
N O X
7,041
6,971
5,520
0
1.405
0
4,123
090
0
229
0
661
553
0
161
0
392
70
V O C
10,129
9.933
7.949
0
1.179
0
6,770
1,197
0
170
0
1,027
707
0
107
0
600
196
C O
107,230
104,966
02,005
0
a,093
0
' 73.192
11.007
0
1.130
0
9.949
11.794
0
1.505
0
10,209
2,264
DIESEL POUtREO VEHICLES
HIGHWAY VEHICLES - HEAVY DUTY
LIMITED ACCESS
RURAL
SUUURBAN
URBAN
OFF-HIGHWAY VEHICLES
RAILWAY LOCOMOTIVES
14,651
11,061
2,200
590
63.219
0
15,605
0
47.414
539
409
0
75
0
414
43
7
246
195
0
49
0
146
34
17
2.490
1.967
Q
4 75
0
1.492
414
109
302
293
0
40
0
253
62
27
1.025
063
0
119
0
744
124
30
TOTAL - ALL MOTOR VEHICLES
153.725
9.461
557
9,531
10.511
100,255
-------
Table 2-3 (continued)
19-26-82 UMIOHU EMISSIONS DATA SYSTEM
AREA SOURCE EMISSIONS REPORT FOR YEAR 1930 DATA FILE LAST UPDATED ON JUL* 21, 1982
STATE 111)'• HEM JERSEY COUNTY!0660>: BURLINGTON CO AQCR1045I: METROPOLITAN PHILADELPHIA
PAGE 0009-5
T (I » II 3 p P » H U o !'
EMISSIONS IN TONS PER YEAR
K>
I
ro
LAND VENICI.ES
HIGHWAY VEHICLES
LIMITED ACCESS
RURAL
SUUURBAN
URBAN
Off-HIGHWAY VEHICLES
RAILHAY LOCOMOTIVES
TOTAL - ALL GASOLINE VEHICLES
TOTAL - ALL OIESEL VEHICLES
GALLONS FUEL
(THOUSANDS)
151,725
149,652
CALCULATED
VEHICLE MILES
(THOUSANDS)
2,055,715
0
511,929
0
1,541 >71)6
i.48i
590
119,074
14,651
AIRCRAFT
hi I 11ARY
CIVIL
COtUIERCIAl
47,210 LAIIDIHG-TAKEOFF CYCLES
19,600 | AIIOING-TAKEOFF CYCLES
7,610 LAllOUIG-TAKEOfF CYCIES
0 LAIIUING-TAKEOFF CYCLES
PART
9.461
9,405
0
622
0
6.501
49
7
6.922
519
120
120
0
0
SOX
557
501
0
126
0
177
17
17
111
246
21
21
0
0
N 0 X
9,511
6.91&
0
2,270
0
6,668
464
109
7,041
2,490
166
165
1
0
V 0 C
10,511
10.226
0
1,496
0
6,710
256
27
10,129
162
186
164
4
0
C 0
106.255
105,629
0
11.715
0
94.094
2,166
16
107,210
1.025
666
765
61
0
VESSELS
COAL
OIESEL OIL
RES10UAL OIL
GASOLINE
0 TONS
1.690 THOUSAIU GALLONS
0 1IIOUSAI10 GALLONS
481 THOUSAND GALLONS
44
0
44
0
0
57
0
55
0
2
419
0
411
0
6
278
0
65
0
211
870
0
145
0
725
TOTAL TRANSPORTATION
9,625
635
10,116
11.177
109,991
-------
Table 2-3 (continued)
ro
»
19-26-02
NATIONAL EMISSIONS
DATA
S
Y S T E N
PAGE 0009-6
AREA SOURCE EMISSIONS REPORT fOH YEAR 1900
OATA FILE
LAST UPDATED ON
JULY 21. 1962
ST*TM?H: l(EH JERSEY COU
-------
A brief discussion of each of the four major Phase I area source
categories 1s given below.
(1) Road Vehicles
The road vehicle category Includes all gasoline and dlesel powered
vehicles that are operated on streets, roads, and highways. The Phase I
hazardous air pollutants emitted by road vehicles Include formaldehyde,
benzene, ethylene dlbromlde, 1,3-butad1ene (a constituent of motor oil
and rubber 1n tires), cadmium, and benzo(a) pyrene.
Emission estimates for road vehicles were based the annual vehicle
miles traveled (VMT) as listed 1n NEDS. NEDS lists, for each county 1n
the country, the VMT for each of three major categories of road
vehicles: light duty gas, heavy duty gas, and dlesel. Emission
estimates were made for each category based upon the assumption that
emissions are proportional on an area-wide basis to VMT's. The total
emissions from the three categories were then summed for each pollutant.
A variety of EPA reports and papers concerning emissions from road
vehicles were reviewed and emission factors compiled based upon the
available data. These factors are listed 1n Appendix E, along with the
appropriate references. One major assumption made 1n developing these
factors was that half of the gasoline consumed 1n the U.S. 1s unleaded.
Emission factors for gasoline powered vehicles, were apportioned to
reflect this 50-50 mix, as necessary. One major exception was emissions
of 1,3-8utad1ene, which are Independent of fuel type.
(2) Gasoline Marketing
The gasoline marketing area source category 1s composed entirely of
emissions from- service stations; 1t' Includes emissions both from vehicle
refueling, and from the delivery of gasoline to the service station.
Emissions from' bulk terminals and other parts of the wholesale gasoline
distribution network are considered by NEDS to be point sources and
consequently, were treated as point sources In this study. High
exposures near-the pump from self service vehicle refueling were also
excluded from- this scoping analysis.
2-14-
-------
For the Initial scan, emissions from gasoline marketing were
estimated for three pollutants: benzene, ethylene dichlorlde (EDC), and
ethylene dlbromlde (EDB). A fourth pollutant, gasoline vapors, was later
added for the Phase II analysis.
The quantity of VOCs emitted from all service stations In each
county Is given 1n NEDS. Apportionment factors were used to estimate the
emissions of EDC, EDB and benzene each Phase I contaminant In the study.
In Phase II, gasoline vapors were considered equivalent to the VOC
amounts found 1n NEDS. It should be noted that since different cancers
are caused by exposure to gasoline vapors than to EDC, EDB or benzene,
the Incidences are assumed to be additive. Appendix E also provides
these apportionment factors.
(3) Industrial, Commercial and Institutional, and Residential Heating
(Including wood combustion)
For this study, all large boilers listed in NEDS were reviewed as
point sources. Emissions from the combustion of fossil fuels and wood by
small sources were considered to be area sources. Pollutants emitted
during combustion Include benzo(a) pyrene, formaldehyde, beryllium,
nickel, arsenic, and cadmium. Residential wood combustion 1n fireplaces
ancf wood stoves 1s a major source of polynuclear aromatic compounds.
However, for this study, the only polynuclear aromatic quantified was
benzo(a)pyrene.
NEDS contains data on the county level for Industrial, commercial
and Institutional, and residential consumptions of residual oil,
distillate oil, coal, natural gas, and wood. These values were used 1n
conjunction with emission factors to estimate annual emissions of each
pollutant. These emission factors were derived.from a variety of
sources, are listed. 1n Appendix E'.
Emissions from wood stoves and fireplaces were differentiated while
NEDS provides total wood consumption, 1t does not distinguish between
stoves and fireplaces. DeAngelis etal. (1980) provided a state-by-state
2-15
-------
breakdown of wood burned 1n both fireplaces and stoves. For lack of
better data, 1t was assumed that these apportionment factors were
consistent for all counties within each state. This breakdown can also
be found 1n Appendix E.
(4) Solvent Usage
Solvents are liquid organic compounds that are used for (1) cleaning
or (2) product application, (e.g. surface coatings or aerosol
propellants). Because solvents are used throughout the economy In both
the private and Industrial sectors, compiling Inventories of emissions
can be quite complicated. While large sources must be considered as
point sources, for dispersion modeling purposes, smaller sources were
treated as area sources. In many cases, however, these small sources can
collectively compose the bulk of the solvent emissions In a typical urban
area.
Two solvents, perchloroethylene and trlchloroethylene, were among the
11st of study pollutants. Perchloroethylene 1s emitted -by degreasers and
drycl-eaners, with a small amount emitted In a miscellaneous Industrial
category (not Including degreasers, drycleaners, graphic arts, or the
rubber and plastics Industry). Trlchloroethylene 1s used almost solely
as a degreaslng agent.
A NEDS "special Interim report", NE099, was used as the basis for
estimating the solvent emissions. This report provides county-wide
breakdowns on solvent usage by category (I.e., solvent use, degreaslng,
drycleaning, rubber and plastics, graphic arts, miscellaneous,
Industrial, and non-1ndustr1al; Table 2-4 1s an example of this report).
It 1s standard practice to> conservatively assume that all solvent that
was consumed was emitted to the atmosphere. Factors provided by QAQPS
were- then applied to apportion solvent usage (emissions) by pollutant.
These- factors are also listed In Appendix £.
2-16
-------
Table 2-4
Example of Special NEDS Report NE099
AREA SOURCE FUEL CONSUMPTION ALLOCATION RESULTS BY COUXTY
STATE fSAROAD COOEI = 39
COU^Y (SAROAQ COOfl = 1200
STATE NAME = PENNSYLVANIA
COUNTY NAME = DUCKS
YEAR OF RECORD * 01
AQCR = 045
SULPHUR COIfTEIfT < X p -
AI(TH.
fPAV
.70
BITU1.
fOAL
DIST.
PH
.20
RESIO.
OIL
.as
ASH CONTENT IZI
AIITII. BITUM.
COAL COAL
11.06 9
RESIOEIITIAL FUEL
ANIH. COAL
(10»«l TOtISI
124
BITUH. COAL
<10««1 TOIISI
24
DIST. OIL
I10"«4 GAL.I
7,064
NAT. GAS
(10»»7 CU.fT.I
455
COTTIERCIAL AID INSTITUTIONAL FUEL
ANTH. COAL BITUH. COAL
flO»«| TPNSJ
P1ST. OIL RESIO. OIL MAT. GAS
I10""4 GAL. I 110*«7 CU.FT. I
BITUH. COAL
(10««1 TOIIS I
502
169
434
INDUSTRIAL FUEL
DIST. OIL RESID. OIL NAT. GAS
(10""4 GAL. I (10a«7 CU.FT. I
1.111
647
GASOLINE FUEL
LIGHT VEHICLE HEAVY VEHICLE OFF HIHAY
(lp«!«jpAL.| 110»*3 GAL. )
HEAVY VEHICLE
I10«»J GAL.I
DIESEL FUEL
OFF II1HAY
RAIL LOCOMOTIVE
(10««4 GAL.I
f50,51S
11,464
3.563
20,457
446
502
AIRCRAFT
VESSELS
EVAPORATION
MILITARY
l|0 CfC I10M2I
CIVIL
lto circ \ iommi i
13,050
C0I01ERCIAL
LTO CYC (10«»1I
GASOLINE
(10«»J GAL.I
710
SOLVENT PURCHASED
(TOIIS YEAR I
GASOLINE MARKETED
110««5 GAL. )
TRADE PAINT SIC 7535 SIC 371 SIC 25
$7? 04 0 227
MISC. pFG. SIC 373 --TOTAL SURFACE COATING--
522 0 2,052
9.549 1.042
SOLVENTS BY USER CATEGORY I TONS I
SIC 34 SIC35136 SIC 26 SIC 2431244 TRANSPORTATION SIC 36
73 24 1.193 40 0 10
OEGREASIIIG DRY CLEANING PRINTING RUDDER OIIIER USES
1,130 463 120 1.065 3,003
-------
Nontrad1t1onal Sources
A number of sources were addressed 1n this analysis that
historically have not been considered major sources of air pollution, or
whose emissions have not been quantified. These "nontraditional11 sources
may, at times, be dominant local contributors of air toxics. An attempt
was therefore made to estimate emissions and levels of risk for waste oil
combustion, volatilization from municipal wastewater treatment plants,
combustion of hazardous wastes In boilers, emissions from hazardous waste
treatment storage and disposal facilities, and superfund sites.
Unfortunately, data limitations prevented the successful quantification
of emissions from the latter three groups. The methods used to estimate
emissions from each of the other sources are as follows:
(1) Waste 011
Every year over 500 million gallons of waste o11 are burned 1n
boilers, kilns, diesel engines, and waste oil heaters (PEQCo 1983).
These oils can contain high levels of metals and organlcs, and, when
burned, are a likely source of airborne metals and, to a lesser extent,
organlcs.
Lead 1s usually recognized as the waste oil metal that poses the
greatest concern. However, it was not Included 1n this study, since it
1s not a recognized carcinogen. Of the metals Included 1n this analysis,
chromium which has an extremely high potency value, was found to be the
most significant metal emitted from waste oils.
Emission levels from waste oil are difficult to estimate because of:
(1) the wide fluctuation 1n oil contaminant' concentrations; (2)
uncertainties in the ratio of airborne emissions to bottom ash; and (3)
uncertainties 1n the- volume of oil burned in a locality. An approach
using data provided by the Office of Solid Waste concerning waste oil
composition, and estimates of combustion efficiency were used to estimate
emissions. This resulted 1n- conservatively high estimates of emission
levels. Since there- 1s little information identifying specific sources
that burn the. fuel, waste oil combustion was considered to be an area
source rather than a point source, for modelIng purposes in Phase II.
2.-18
-------
The volume of waste oil burned on a county-wide basis was assumed to
be proportional to the county-wide quantity of residual and/or distillate
oil used, as reported 1n the NEDS area source reports. The proportioning
factor between residual and distillate oil consumed 1s unique for each
state as supplied by the Office of Solid Waste, was presumed to be
constant for each county 1n a state. It was assumed that 75 percent of
all metals and 0.1 percent of all organic contaminants that are in an
"average" waste oil, as defined by PEDCo (1983), were emitted to the
air. Appendix F-l provides more detailed documentation of the waste oil
emission algorithm.
(2) Publicly Owned Treatment Works (POTWs)/Sewage Treatment Plants
Sewage treatment plants, also known as POTWs, are an Interesting
component of this study because their emissions are a result of
Intermedia shifts of pollutants from wastewater to the air via
volatilization. An objective of the study was to compare emissions from
POTWs with emissions from more traditional point and area sources.
Previous IEMD studies of Philadelphia, for example, suggest that a major
wastewater treatment plant may be a major source of toxic air pollutants.
For modeling purposes in Phase II each POTW was considered to be a
point source. The Industrial Facilities Qischarge (IFD) file, plus the
use of prototype facilities (model plants), was used to estimate
volatilization from all POTWs in the U.S. that accept Industrial waste
a total of approximately 1600 facilities. While IFO does not contain
data on volatilization, 1t does contain Information on the name of the
plant,, flow, latitude, longitude, SIC codes discharging to plant, and
control technology level. Criteria used in selecting the appropriate
model plant to serve as the basis for emission estimation, included flow
rate, the constituency of Industrial wastes recovered by the POTW, and
the' type- of wastewater treatment used by the plant. For each facility,
emissions were estimated for the following pollutants: benzene, ethylene
dlchlorlde, carbon- tetrachloride, chloroform, vinyl chloride,
2-19
-------
trlchloroethylene, perchloroethylene, and acrylonltrlle. These
pollutants were selected because of a high Henry's law constant - (I.e.
high volatility, low solubility). Appendix F-2 provides more complete
documentation of the procedures used to estimate emissions from this
non-trad1t1onal source category.
(3) Combustion of Hazardous Waste 1n Boilers
This source category was considered for Inclusion because standards
are currently being developed by the EPA and preliminary data Indicate
that significant quantities of hazardous materials are being burned 1n
boilers. However, based on limitations of available data (on such
factors as the volume types and compositions of wastes being burned,
destruction efficiencies, sources, and associated risks),1t was felt that
we could not adequately address this source category.
(4) Treatment, Storage, and Qlsposal Facilities (TSDFs)
Like the combustion of hazardous waste in boilers, this source
category was found to be too poorly defined by readily available
Information to address 1n this study. Existing data bases do not provide
the necessary data to predict emission levels. No two TSDFs are alike;
they accept different waste streams often with uncertain compositions,
use different engineering processes, are subject to different, often
undocumented, environmental conditions (such as soil type and cover
depth). All of these parameters affect resulting emission estimates.
RCRA Part A data, as documented In the Hazardous Waste Oata Management
System (HWOMS), was. considered to be unsuitable for Input to emissions
models because of Inflated capacity data and the often unspeclflc nature
of the data entries.
Current models for estimating emissions from TSDFs are considered to
be-Inadequate-. Although models of releases from various TSDF sources
have- been developed (see Appendix F-3), they still require additional
refinement and validation.
2-20
-------
The feasibility of estimating emissions from TSDFs was considered
again for the subset of counties 1n Phase II, but data were
Insufficient. Survey data for 3,000 TSDFs compiled by the Office of
Solid Waste survey were evaluated as a source of Information, but only
limited data were found for the 35 counties. The following specific
limitations were Identified:
• A number of study pollutants are reported as components of generic
(or combined) waste streams and process waste streams. Data were
not available to determine the actual percentage of the waste
stream that corresponds with the target chemical (e.g.,
trlchloroethylene Is reported as a component of spent halogenated
solvent waste streams such as Nos. F001 and F002).
• The data base does not contain detailed Information describing
specific design and operating parameters for many of the reported
waste management processes.
• .The study pollutants did not appear often 1n the survey data.
Only 10 of the pollutants were even reported as being managed by
any of the 244 TSDFs surveyed 1n the 35 counties. Of these ten
pollutants, five are reportedly handled by only one facility.
Four of the remaining pollutants were reported to be treated at
fewer than six facilities.
A qualitative report was prepared' that summarizes the state of the art 1n
this area 1n Heu of attempting to quantify emissions; this 1s presented
in Appendix F-3. It was concluded that a large proportion of liquid
wastes handled at TSDFs are managed 1n open systems, such as impoundments
and landfills. The potential for emission of toxic pollutants 1s
considerable, since even compounds with extremely low vapor pressure and
low solubility 1n water can be subject to significant volatilization
release' when placed In surface Impoundments, landfills or landfarms.
Ambient, monitoring at selected TSDFs may be. the only meaningful way
of characterizing- Impacts from this type of source. Appendix F-3 also
documents; the analysis of available monitoring data that was performed.
However, considerable' developmental work remains to be conducted 1n order
to design- sampling methods adequate to address the range- of situations
2.-21
-------
that exist at the TSDFs. Some of this work 1s currently being conducted
for OAQPS. These efforts are designed to test field sampling methods.
(5) Uncontrolled Hazardous Waste Facilities (Superfund Sites)
Contaminant release to air must be observed before a site can be
added to the National Priorities List (NPL) because of air contamination;
16.5 percent of all the 409 currently listed NPL sites were listed
because of air contamination problems. (The NPL 1s CPA's ranking scheme
to set priorities among candidate superfund sites, and 1s based on the
potential magnitude of human exposures and risks due to contact with
chemicals at, or migrating from the site). A comparison of data
quantifying concentrations of specific air toxics at superfund sites with
similar data for TSDFs determined that the Superfund sites concentrations
were in most cases lower than those found at TSOFs (see Appendix F-3).
This 1s not surprising as highly volatile materials will have had ample
time to escape from abandoned sites, while their release from active
sites would be ongoing. However, since the toxics concentrations
observed at Superfund sites may, 1n certain Instances, compare with
levels that have been reported 1n the vicinity of a number of TSDFs, the
air toxics loading contribution of this source category should not be
overlooked, especially 1n local (as opposed to regional or national)
evaluations.
Unfortunately, for many of the same reasons that were listed above
for TSDFs, 1t was also Impossible, at least at this time, to adequately
predict emissions from this potentially significant source of toxic air
pollution.
2.1.3 Risk Screening
Ranking, methodologies were developed to provide perspective and to
form a. basis from which to select counties for further study. The three
principal ones are as follows:
2-22
-------
m n
"Risk Index 1" = l l (kkg/yr) (potency)(county population)
3-1 1=1 J1 1
m n
"Risk Index 2" = I I (kkg/yr) (potency)(county population density)
j=l 1=1 J1 1
m n
"Risk Index 3" = t I (kkg/yr) (potency)
j=l 1=1 Ji 1
where: kkg/yr = annual emission rate 1n metric tons per year
1 = summation across "n" pollutants emitted from a source
J = summation across "m" sources within a county
The proceeding risk Indices were used to rank counties rather than
exposure modeling because of the massive number of sources involved;
dispersion modeling was neither practical nor necessary to meet the
objective of providing perspective on the counties selected for more
detailed review.
The use of different risk indices provided several approaches to
characterize counties. While all Indices consider the mass and potency
of sources within each county, the population term 1s different in each.
The reasons are as follows:
Risk Index 1 - This approach 1s heavily weighted, by the total
population 1n a county. For cumulative exposure, total population is an
Important factor to exposure. However for geographically large counties
with widely dispersed sources, this approach may unreal 1stlcally express
the county's risk.
Risk Index 2 - Although much like the preceedlng Index, by using
population- density (population t miles'") rather than total
population-, the previously stated bias associated with large, spread-out
counties 1s reduced.
2-23
-------
Risk Index 3 - This Index was used to help identify counties with
relatively high emissions, but low populations.
Thirty-five counties selected for Phase II analysis are presented In
Table 2-5. Twenty-two of these counties were near the top of all three
lists. An additional seven counties that were typically high in at least
two of the Indices were also selected. The remaining six counties were
selected to consider areas with very large Industrial emissions of one or
more of the selected pollutants, regardless of population. These
counties, which were chosen to explore aggregate versus maximum
Individual risk, are listed below:
• Brazoria, Texas
• Kanawha, West Virginia
• West Baton Rouge, Louisiana
• Calcasieu, Louisiana
• East Baton Rouge, Louisiana
« Sedgewlck, Kansas
Approximately twenty percent of all VOC and 10 percent of Total
Suspended Particulates (TSPs) emitted 1n the nation are released to the
air 1n these 35 counties; they also contain 20 percent of the U.S.
population. These counties also account for a significant percentage of
"cumulative risk", as determined by the three surrogate risk Indices.
While the Phase I analysis provided perspective on the counties
ultimately selected for further review, greater detail was needed to
adequately consider the magnitude and nature of the air toxics Issue for
the- selected pollutants. In Phase II, the emissions data were improved
for the 35 counties selected, dispersion modeling was performed and
population data assessed to estimate exposure, and some preliminary
analyses were performed to consider most exposed Individuals.
2-24.
-------
Table 2-5
Thirty-five Counties Selected for Phase II
Los Angeles, CA
Baltimore City, MO
Bergen, MO
Cook, IL
Queens, NY
Oakland, MI
Harris, TX
Middlesex, NJ
Orange, CA
Philadelphia, PA
Essex, NJ
Wayne, MI
8urlington, NJ
Metro Boston, MA
Santa Clara, CA
Dallas, TX
Marion, IN
Cuyahoga, OH
Jefferson, KY
Nassau, NY
Sullivan, TN
Allegheny, PA
Summit, OH
Alameda, CA
Tarrant, TX
San Oiego, CA
Jefferson, TX
Maricopa, AZ
West Baton Rouge, LA
East Baton Rouge, LA
Sedgewick, KS
Kanawha, WV
Calcasieu, LA
Galveston, TX
Brazoria, TX
2-25
-------
2.2.1 Refinement of Phase I Input Data
As mentioned 1n the Phase I discussion, there were several
limitations to the emissions data used 1n Phase I. To Improve the
emissions data, the following four major data sets were added to HEMIS
during Phase II: (1) Versar generated plant-specific data, (2) SAI
plant-specific data, (3) estimated trace metal emissions, and (4) area
sources. Each of these 1s discussed below:
(1) Versar data - This Information consisted of plant-specific
emissions estimates for large major Industrial facilities
located 1n the 35 counties. Emissions factors and production or
capacity data were used to estimate emissions. If these data
were not readily available for pollutants or facilities, we did
not estimate emissions. Listing of point source data generated
by Versar 1s presented 1n Appendix G. Since these data were
based on some plant specific Information, they were always given
preference over NEDS when data from both sources existed.
(2) SAI data - These emissions data were estimated based on review
of permit data, site Inspections, and engineering analysis. A
listing of these data and more Information on their derivation
are presented 1n Appendix G. Since these SAI data were based on
the most s1te-spec1f1c Information, they .were given preference
over NEDS or the Versar estimates when data from all three
sources existed for a particular facility.
(3) Hetal emissions - Trace metal emissions from utilities were
estimated based on readily available emissions factors and
production data from NEDS. A complete discussion of the
calculation procedures and .assumptions used Is presented In
Appendix G. In addition to emissions from utilities, some
additional estimates were made for trace metal releases from
processing; additional Information 1s given 1n Appendix G.
(4) Area Source Emissions - Area source emission estimates were also
expanded or upgraded 1n several ways. We estimated
pentachlorophenol (PCP) emissions from cooling towers and
preserved wood based upon national emission estimates of PCP.
It was assumed that PCP emissions from commercial and
Institutional sources, and residential cooling towers, and large
air conditioning systems are distributed proportionately to
population density. Data were available listing preserved wood
usage' 1n each of five regions of the county. 8y assuming that
emissions were proportional to usage, 1t was possible to
estimate emissions on a regional level. It was then presumed
2-26
-------
that emissions were distributed by square mllege within counties
throughout each geographic region. Appendix E presents a
further description of. these methods. Gasoline vapors, emitted
from service stations (gasoline marketing), were considered to
be equivalent to the VOC emissions listed In NEDS.
Because of concern about the direct use of NEDs data for this
application, additional quality control was performed on the emissions
and stack specifications data that were used 1n modeling. All of the VOC
data were reviewed at the stack level to delete records that appeared to
be anomalously high. In the same manner, stack specifications data such
as height of release, exit velocity, exit temperature, and the diameter
of stacks were reviewed for anomalous or Inconsistent stack data, (e.g.,
unreal 1stlcally high stacks, exit velocities, stack height/stack diameter
ratios, etc.). Plants with unreal1st1cally high stack heights, exit
velocities or flow rates were defaulted to a conservative set of stack
specifications. Appendix H documents the procedures used to screen and
select stack specifications for the 35 county study.
Finally, although coke oven emissions were not estimated in this
analysis, exposure and cancer incidence calculations were provided by
OAQPS (see Appendix I). These numbers were developed to support the
listing of coke oven emissions as a hazardous air pollutant under Section
112 of the Clean A1r Act.
2.2.2 Dispersion Modeling
Dispersion modeling is used for a wide range of applications. While
detailed modeling. 1s often essential for applications associated with
regulatory actions or mlcroscale analysis, 1t was feasible for this
scoping study, where emissions from thousands of sources had to be
modeled-. Inputs, to the model were simplified by assuming: one set of
stack release specifications, such as height, temperature, and flow for
each facility; uniform- distribution of area source emissions, regardless
of- land' use- patterns or population density; and the use of meteorological
data, representative on the county level. One reasonably representative
National Weather Service (NWS) station was selected in each county to
2-27
-------
model all sources. Stability Array (STAR) data were used to represent
wind speed, wind direction, and atmospheric stability. Wind flow and
stability can be markedly different for various subreglons within a
county. While a study to evaluate emissions from one source 1t may, In
some cases, be advlseable to obtain s1te-spec1f1c data for modeling
purposes, for a study with as large a scope as the 35 county study,
simplifying assumptions were needed. The use of one NWS station to
represent sources within a county 1s consistent with the screening-level
modeling that was performed.
The GAMS model was selected as the modeling system for this study
because 1t contained a suitable point source dispersion model
(ATM-SECPOP), area source box model (BOXMOD) and the national 1980 census
data base (GSC 1984). This system was developed and 1s maintained by
EPA's Office of Toxic Substances.
The dispersion models 1n GAMS are compatible with the screening
objectives of this program. ATM-SECPOP contains the basic core of most
EPA models, I.e., PasqiM 11-Guifford dispersion coefficients as Inputs to
Gaussian dispersion modeling. The additional features of ATM-SECPOP such
as dry deposition, wet deposition, and decay, were conservatively set
equal to zero. 80XMO0 1s a box model that greatly simplifies the
dispersion of area source emissions by assuming that these releases are
unlformaly distributed across an area. The main reasons that GAMS was
selected Include:
•• point source algorithm Is generally consistent with EPA models
•i contains 1980 census data, disaggregated at the block
group/enumeration district level.
* box model provides the simplification needed to reasonably manage
the' volume: of area sources to be modeled.
For screening- level purposes, 1t 1s likely that other modeling- systems
could.' also have satisfied the objectives of the study.
2.-28
-------
Normalized exposure coefficients were computed for both point and
area sources based on an emission rate of 1 kkg/yr. These values were
subsequently linked with the emissions Inventory to estimate cumulative
exposure. Cumulative exposure out to 50 km from each point source was
computed, as well as cumulative exposure for all populations within a
county for area sources. Area sources were generally assumed to be
distributed equally throughout each county. However, for those counties
that had a well-defined urban area and rural area, two overlapping area
sources were used for modeling purposes. The unit emissions were
apportioned between the two area sources on the basis of population.
All facilities with VOC emissions greater than 100 tons/yr were
modeled as point sources, as well as all utilities, POTWs and facilities
with emissions data more specific than the apportioned NEDS values.
Facilities not modeled as point sources were modeled as area sources.
Sixteen wind directions and 3 ring distances (15 - 50 km) were
evaluated around each source, for a total of 128 sector segments.
Figure 2-1 summarizes the sector segments that were modeled around each
point source. Average concentrations for each sector segment were
computed along the centerllne of the segment based on receptors at three
equally spaced distances. The average concentration for the sector
segment was assumed to apply to the entire population within each block
group/enumeration district (BG/ED) that was 1n the sector segment. We
also computed exposures on the 8G/ED level, and then summed across all
sector segments. Population data are compiled on the county level,
census tract level and 8G/ED; the BG/ED level 1s the finest level for
which population data were available.
For concentrations estimated by BOXMOD, the same value applied to all
receptors within the area. Exposure was summed across all BG/EDs within
an urban area, source or county area source, as applicable.
2-29
-------
A««rau* concentration for
each tagmant it batad on
pradictad concantrationi
from Ihrw awnly ipacarf
rscapton.
Figure 2-1 Modeling Receptor Pattern
2-30
-------
Exposures are computed as follows:
128
Ej = I (*1) (Pi) (365 days.) (24 nj3)
1*1 yr day
where Ej = total cumulative exposure (ug-persons/yr) from source "j"
x^ = annual average concentration for sector segment "1",
(ug/m3)
Pj = total population 1n sector segment, "I"
3
The breathing rate 1s assumed to be 24 m /day for all Individuals.
Complete retention of inhaled pollutants 1s also assumed.
The following additional factors should be considered when
Interpreting the results of the modeling:
• If a point source was located within 50 km of 3 county line,
exposure outside of the county (out to 50 km) was considered.
However, only sources within a county were used 1n making model
runs for each county.
• The effective height of release of pollutants 1s the actual
release height plus the additional rise of the plume due to
buoyancy and momentum of the effluents, I.e., stack height plus
plume rise. The plume rise term in ATM-SECPOP may overestimate
effective release heights for relatively hot effluents, which in
turn could lead to underestimates of concentration. Comparisons
of the plume rise term In ATH-SECPOP to basic references, and
• GSC (1983) indicate that ATM-SECPOP appears to compute
relatively high plume rise values for hot effluents. For some
sources, this term could make as much as an order of magnitude
difference 1n total exposure.
•' Pollutants emitted from low. level stacks or vents on or near
buildings are not necessarily dispersed the same as those from
an isolated stack. Instead" of dispersing in a Gaussian form in
the- vicinity of the release, the pollutants are more likely well
mixed 1n the- turbulent zone downwind of nearby or contiguous
structures. All point sources, including low-level sources,
were modeled as point sources without accounting for the
influence of building wake effects. For low-level releases of
hot stack or vent gases, this factor could lead to substantial
underestimates of concentrations because the model may compute
an- unreal 1 stically high' effective- release, height.
2.-31
-------
• Diurnal and seasonal variation 1n emissions was assumed to be
zero. In reality, many pollutant releases are probably
correlated with these factors, especially diurnal conditions.
For example, annual emissions were assumed to be released
continuously over a period of one year. If a facility were to
operate on a one shift schedule, emissions could be zero during
the night when extremely stable and low mixing heights (poor
dispersion) 1s occurring. This factor would tend to result In
an overpredlctlon of impacts.
• Area source emissions on the county level typically would be
expected to show substantial variations between urban and rural
sections of a county. Although this Issue was somewhat
mitigated by selecting overlapping urban and rural box model
components, this factor still Is greatly simplified by assuming
uniform distribution of emissions. In total, 1t would be
expected that the use of the box model would provide generally
comparable results compared to treating each of those sources as
point sources. For sources modeled In this manner that are far
from population centers cumulative exposures may be somewhat
overestimated; while
sources in the urban center could have cumulative exposure
somewhat underestimated. In balance, the use of the box model
is expected to reasonably represent exposures from the numerous
small sources.
•• We estimated concentrations out to 50 Km. A different cutoff
could have been selected , e.g. 20 km. 3y that approach,
however, exposures 20-50 km from the source would have been
Ignored. Since many Industrial centers In urban areas are
located within a range of 20-50 km of high population centers,
1t was considered essential to provide screening estimates 1n
this range, rather than to systematically reduce estimated
exposures by cutting back the range. In keeping with the
objectives of the study, these exposure values are screening
estimates; the modeling components of this study were not
designed for refined analyses.
The- preceding approach was applied to all sources listed 1n HEMIS.
Total cumulative exposures were computed for each source. When linked
with emissions' and potency data, these- values provide Indicators of
estimated cancer Incidence.
While; aggregate exposures provide insight Into the overall
populations 1n contact with air toxics, Individuals located near the
fencellne of major emitters of toxic, pollutants can be exposed to
2-32
-------
Individual risk orders of magnitude higher than the general population.
Even though the cumulative risk may appear relatively low, a small
percentage of the population may be exposed to risks that are
unacceptable 1n the ambient air. Consequently, an attempt was made to
evaluate lifetime risks to the most exposed Individuals (MEI) for a
limited number of major facilities that were selected as examples. This
measure Indicates the probability of an Individual contracting cancer as
a result of exposure to an ambient pollutant concentration for 70 years.
Estimates of lifetime cancer Incidence 1n each county were obtained
by multiplying total cumulative exposure (normalized to 1 kkg/yr) times
the emission rate times the pollutant potency. This was calculated for
all pollutants and sources. This approach provides an estimate of excess
cancer deaths that may occur over a 70 year period. Annual cancer
Incidence 1s estimated by dividing the lifetime estimates by 70.
The total concentration at any receptor 1n a county Is a function of
contributions from numerous sources. The gradients In concentration
across an urban area are often cfu1te complex. Maximum .Impacts C3n occur
at locations where source Impacts overlap, rather than 1n the 'vicinity of
a major source. However, for the example facilities, the releases were
typically extremely large 1n comparison to other point sources, and were
released near ground-level from vents. For these reasons, 1t 1s expected
that the MEI calculations made for receptors located near the fencellne
that are Indicative of worst case Impacts for these pollutants, based on
available data. It should be noted, however, that the analyses were
screening level, and not the level of mlcroscale analysis, needed to
definitively estimate concentrations.
The-ATM-SECPOP results were evaluated to assess toxics exposure at
close-In receptors, 1n the range of 0-500 m from each selected point
source1. The- results of the above evaluations are documented In Appendix
J. and are discussed. 1n Section 3.0.
2-33
-------
3.0 CONCLUSIONS
The findings of this study are summarized In the tables and figures
presented in this section. The detailed figures from which these
highlights were drawn are found 1n Appendix J.
3.1 General Findings for the 35 counties.
Table 3-1 shows the estimated lifetime cancer incidence, In number of
cases, for the 21 pollutants studied. It should be noted that although
lifetime numbers are presented, annual Incidence of cancer can be
obtained by simply dividing the lifetime cancer estimates by 70 (the
assumed duration of exposure). The numbers reflect the general
dimensions of the potential hazardous air pollution problem for the 35
counties (the first significant decimal of the calculated values 1s
provided only to show relative values near and below zero). The total
estimated annual Incidence of cancer for these areas is 83.6. The
greatest proportion, of risk 1s attributable to benzene, which accounts
for about 23 percent of the total. Total chromium is shown to present
the second highest degree of risk -- approximately 16 percent of the
total.
It should be noted, however, only the hexavalent form of chromium
has been proven to be carcinogenic; there 1s still substantial debate
regarding the potential carcinogenicity of the trlvalent form. Since
there was no Information on the ratio of trlvalent to hexavalent, either
1n the- available amblant air data or 1n emissions, this study
conservatively assumed that total chromium releases were carcinogenic and
that trlvalent 1s as potent as hexavalent. A similar situation also
arises* with nickel. Only two rare- species are considered carcinogenic
(nickel' carbonyl and nickel subsulflde), but this study assumed that all
nlckeT forms, were equally potent.
Next 1n order of Importance are formaldehyde (12%), vinyl chloride (10%),
gasollne-'vapors (8*/.), and the common solvents (perchloroethylene -- 8% and
trlchloroethylene — 8%) all of which appear to pose comparable levels of
3-1
-------
Table 3-1
Suimary of Estimated Lifetime Cancer Incidence for 35 Counties
Listed by Pollutant Studied
Pollutant
(70-yr)
(70-yr)
(70-yr)
(70-yr)
Total
Total
Industrial
POTW
Total
Area
Estimated
Estimated
Point
Sources
Point
Sources
Lifetime
Annual
Source
Sources
Cancer
Cancer
Incidence
Incidence
Benzene
332. 5
4.3
336.8
961.1
1297.9
18.5
Chromium (Total)
512.9
512.9
423.2
936.1
13.4
Formaldehyde
430.5
430.5
269.6
700.0
10.0
Vinyl Chloride
540.6
30.3
570.9
570.9
8.2
Gas Vapors
0.0
474.8
474.8
6.8
Trich1oroethy1ene
139.5
15.4
154.9
319.6
474.5
6.8
PerchIoroethy1ene
186.2
6.7
192.9
274.6
467.5
6.7
Acrylonitrile
287.0
5.5
292.5
292.5
4.2
Coke Oven Emissions
170.1
170.1
170.1
2.4
Ethylene Chloride
20.0
34.4
104.3
0.1
104.4
1.5
Cadniun
26.8
26.8
50.4
77.1
1.1
Benzo(a)Pyrene
0.0
74.9
74.9
1.1
Arsenic
23.4
23.4
50.5
73.9
1.1
Ethylene Brenride
0.0
70.7
70.7
1.0
Nickel
46.4
46.4
5.0
51.4
0.7
Carbon Tetrachloride
2.5
8.0
10.5
10.5
0.2
Chloroform
0.7
4.7
5.4
5.4
0.1
Styrene
1.3
1.3
1.3
0.0
1,3-8utadiene
0.5
0.5
0.5
1.0
0.0
Serylliun
0.3
0.3
0.5
0.7
0.0
Pentachlorophenol
0.0
0.0
0.3
0.3
0.0
TOTAL
2721.T
159.3
2880.4
2975.5
5855.9
83.6
Because of the-uncertainties in the incidence-
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
3-2.
-------
risk. Risk levels appear to taper off significantly for the remaining
pollutants, with styrene, 1,3-butad1ene, and pentachlorophenol showing
virtually no significant risk throughout the 35 counties.
Figure 3-1 shows the general contribution to total estimated cancer
Incidence from point and area sources. (Volatilization of pollutants
from publicly owned sewage treatment plants 1s 1n the point source
category but 1s distinguished here from the major permitted Industrial
sources because 1t has not traditionally been considered as a point
source of air emissions). As Indicated, the division between point and
area sources 1s essentially equal — 49% (point) vs. 51% (area).
3.2 Cancer Risk Associated with Various Source Categories
When all sources are viewed together (see Table 3-2), 1t 1s
Interesting to note that no one source category seems to dominate. Road
vehicles appear to contribute the single largest amount (23%) of total
estimated cancer Incidence for the 35 counties (see Section 2.2.1 for a
discussion of the derivation of these risk estimates). Roughly equal
degrees of cancer Incidence are associated with solvent usage (10%),
gaso.Hne marketing (9%), synthetic organic chemical manufacturing and
usage (10%), waste oil burning (8%), and metal manufacturing (8%), but
each of these categories appears to be less than half as Important as
road vehicles 1n the total risk picture. With the exception of metal and
chemical manufacturing, these categories fall Into the "area source"
category, since the emitting points are either mobile or small and
dispersed.
Figure- 3-2 shows each of the 20 pollutants studied 1n terms of total
risk,, point source. (Industrial) risk, area source risk, and point source
risk, from POTWs. For six of the- leading pollutants — benzene, total
chromium, formaldehyde, gasoline vapors, trlchloroethylene, and
perchloroethylene, — 1t Is Interesting, to note the significant, fraction
of the- total risk presented by area sources. Only vinyl chloride and
acr1lon1tr1le appear to be-exclusively point source related. Coke oven
emissions, not. shown 1n this figure, are- attributable entirely to point
sources..
3-3
-------
Sources or Cancer Incidence Tor 35 Counties Nationwide
Lifetime
Cancer
Incidence
(Cases)
3000 -
2976
2719
tvC * ^ N»V
k *w\,\x\V«. \\\v^.v. Vn\-
2000
-SS\V\>s ¦¦:
t 'M'x r. ¦"•¦ ¦ •¦ ¦•¦¦¦¦¦ ^ •
V. ..'•. '-''O.;:- 'V |
•uuv
:v-r>y>
y ->VX
>\\v ;wvo y*\ v.; \y
kv>\-v.y'.V ¦•¦
r N
V ,. . .1 - J
I- I
j1 ;"v": ;.-.;'v v * \V ¦•]
'• -,^- ^ Y^" V '. ; V- j
\ "v'm
n,V,N%\X».».'C,w>-'-\"* j
\ ' .'• ,V. I
k\\'L's ¦ I
r~
Point Sources
Area Sources
Sourer of risk
PQT\v Volatilisation
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Cancer Incidence by Source Category
Figure 3-1
3-4
-------
Table 3-2 Percent of Risk Associated with
Point and Area Sources in 35 Counties
Point Sources \ Total Incidence
Chemicals Production/Usage 10
Hetal Manufacturing 8
Petroleum Refining 5
Rubber Production 5
Utilities 3
POTWs 2
All Others ]6_
TOTAL POINT SOURCE 49
Area Sources
Road Vehicles 22
Solvent Usage 10
Gasoline Marketing 9
Waste Oil Burning 8
Heating
Woodsmoke (stoves/fireplaces) 0.5
A11 other 1 -5
TOTAL AREA SOURCE 51
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
3-5
-------
Benzene
Total Chromium 22
¦W- vO- VS O \
Formaldehyde
Viny! Chloride T
Gas Vapors
Tnchloroethylene
Ethylene Dicnloriae _u
CaGmium £¦ I:
B?nzo(a!Pyr»T» S3
Arsenic £33
Ethylene Oibromide
Nickel IH
Carbon Tetrachloride
Chloroform
3tyrene
i ,3-Butadiene
Beryllium
PenLacfiloropnenol j
1—
0
T
Percnloroethylene £;.-:-s.vc-x-•• i
- Acrylomtrile ¦_ i
| a PO?w Vootiii*3tion
!
j L! Industrial Point Sources
j 3 Area Sources
200 4C0 500 200 '000
Lifetime Cancer Incidence
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the nurrbers should.be regarded as rough in-
dicators only.
Incidence As a Function of Pollutant
Figure 3-2
3-6
-------
Some of the major source categories are discussed 1n more detail
below.
3.2.1 Road Vehicles
Figures 3-3 shows the relative pollutant contributions to total
estimated cancer Incidence from road vehicles. The largest component Is
benzene, which accounts for 73% of the calculated values. The remaining
pollutants of concern are: formaldehyde-(16%); ethylene dlbromide (5%);
B(a)P (4%); and cadmium (2%).
3.2.2 Gasoline Marketing (Service Stations)
The pollutant of most Importance 1s gasoline vapors, accounting for
roughly 95% of the total estimated lifetime cancer Incidence (see Figure
3-4). Benzene contributes a little under 5% to the total Impact, and
ethylene dlbromide and ethylene dlchlorlde are relatively Insignificant.
3.2.3 Solvent Usage
Trlchloroethylene and perchloroethylene, which are two solvents
widely used for Industrial degreaslng and drycleanlng, apear to be among
the more major sources of estimated lifetime cancer Incidence accounting
for approximately 600 lifetime cases (8.6 cases annually) 1n the 35
county study area. Contributions to total estimated cancer Incidence
appear to be about evenly split between the two substances (Figure 3-5).
3.2.4 Waste 011 Burning
Figure 3-6 presents the estimated lifetime Incidence from the burning
of waste, oil for heat recovery. Chromium 1s by far the major source of
cancer Incidence, accounting for nearly ninety percent of the risk.
It should- be noted that lead, typically the major contaminant 1n
waste oil, was one of many pollutants not Included In this analysis since
1t 1s not a carcinogen.
-------
Road Vehicles
Ethylene Oibromide
g(a)p Cadmium
4% 2"
Formaldehyde
163
.... .
>v^- c«I%prijirfi
Risk Contributed by
Various ?oi!u:3n:5
for 35 counties
Benzene
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Incidence Due to Road Vehicles
Figure 3-3
3-8
-------
6asoline Marketing
500 -
475
Lifetime ;
Risk Contributed oy
Various Pollutants
for 35 counties
1
1
Cancer
400 ¦
Incidence
300
(Lifetime
Cases)
200
100
24
1
! • :
U
Gas Vaoors
Benzene
Pollutant
Ethlylene
Oibromide
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Incidence Due to Gasoline Marketing
Figure 3-4-
3-9
-------
Figure 9
Solvent Usage
400 r
Cancer 300 f !
Incidence
(Lifetime
Cnn)
200
I
•¦00 f
320
r •
I-: .
I L.
i r-
Lifetime
Risk Contributed by
Various Pollutants
for 35 counties
275
r-
Trichloroelhylene Perchlcroethylene
PolluUnt
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the- numbers should be regarded as rough in-
dicators only.
Incidence Due to Solvent Usage
Figure 3-5
3-10
-------
Waste Oil Burning
Cancer
Incidence
(Lifetime
Cases)
500
400
300
200
100
0
423
I f
I E
f
Total
Chromium
49
I
Lifetime
Risn Contributed by
Various Pollutants
for 35 counf.es
Arsenic Cadmium Total Mick si
Pollutant
3!.u;P
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the mmbers should be regarded as rough in-
dicators only.
Incidence Due to Waste Oil 3urning
Figure 3-6
3-11
-------
3.2.5 Residential Wood Combustion and POTWs (Sewage Treatment Plants)
Although these particular sources did not appear to contribute
significantly to the total estimated cancer Incidence for the 35
countries, (POTW: 3%, residential wood combustion 0.5%), they deserve
special mention because of our Interest 1n these nontradltlonal sources.
Wood combustion has recently become a source of Interest as a
possible contributor of hazardous air pollutants, 1n particular
polycycllc organic matter (POM). In this analysis, woodburning 1s shown
to be a rather moderate source of air toxics risk; however, of the many
POMs that may be released from residential wood combustion, only B(a)p
was characterized. Also, the 35 counties are not representative of the
areas where woodsmoke may be a problem. Figure 3-7 shows the breakdown
of component risk factors for two categories of wood burning —
residential fireplaces and residential woodstoves. Woodstoves appear to
«
present considerably higher risks from benzo(a)pyrene than fireplaces,
but the total estimated cancer incidence for each category 1s still small
compared, to the other sources studied.
The estimated cancer Incidence attributed to POTWs comes, predictably,
from a wider range of substances than for many other sources. Figure 3-8
shows the relative contribution of eight pollutants to the estimated risk
from POTWs for the 35 counties analysis. The three leading substances are
ethylene' dlchlorlde (53%), vinyl chloride (19%), and trlchloroethylene
(9%), followed by carbon tetrachloride (5%), perchloroethylene (4%),
chloroform (3%), acrylonltrlle (3%), and benzene (3%).
The contribution to aggregate risk 1s more significant than for
woodstoves but 1s. 1s nonetheless fairly small. As shown 1n the- next
section on geographic variation, sewage1 treatment plants, like may not be
of national' concern, but they may be- relatively Important sources of air
toxics for some Individual counties.. These sources may also be. of
concern from- the standpoint of risk to the most' exposed Individual.
3-12
-------
Woodburning
Cancer
Incidence
(Lifetime
Cases)
r
t
2 f
Formaldehyde ' S(aP
Lifetime
Pisk Contributed by
Various Pollutants
:or
; counties
3 Residential Fireplaces ;
G Residential Woodstoves ;
JLiZZL
Arsenic Total MicVe!
Pollutant
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the nurbers should be regarded as rough in-
dicators only.
Incidence Due to Woodburning
Figure. 3-7
3-13
-------
POTW Volatilization
Cancer
Incidence
(Lifetime
Cases)
90
50
70
60
SO
40
30
20
10
b
t 84
30
I-
i
Lifetime
Risx Contributed bv
Various Pollutants
for 35 counties
Ethyl,
dichlor.
Vinyl
Chlor.
In.
cnioro
5 th.
Cjrbon Perch Chloro Acr ?er;
tei eth. form nitr.ie snt
Pollutant
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Incidence Due to POTW Volatilization
Figure- 3-3
3-14
-------
3.3 Variation in the Significance of Sources and Pollutant: Across
Geographic Areas
One of the major objectives of the 35 county study was to explore the
extent to which the air toxics problem varies geographically. This
section first discusses some of the major source variation evidenced for
the 35 counties, and then takes a more 1n-depth look at six illustrative
counties.
Table 3-3 shows total estimated-1ifetlme cancer incidence for the
three general source categories for the 35 counties studied (complete
figures are shown in Appendix 3). Although Incidence numbers are
presented for each county, 1t 1s Important to recall that these numbers
should be construed only as rough Indicators. This table shows wide
geographic variation 1n the balance between point sources and area
sources of risk. Area sources dominate such locations as Los Angeles,
CA; Cook County, II; Queens County, NY ; and Oakland County, MI. Point
sources dominate 1n Baltimore County, MO; Bergen County, NJ; and
Middlesex County, NJ. Note also that about 75 percent of the total risk
calculated for the 35 county area 1s attributed to the top ten counties.
The succeeding figures look at six representative counties 1n more
detail: Los Angeles, Baltimore, Cook, Harris, Philadelphia, and Sullivan.
For Philadelphia 1t appears, that POTVJ-related risks may outweigh
Industrial point sources 1n Importance (72%) (POTW-18% vs. Indust.-10%).
Overall, area sources predominate In the total risk analysis. POTW
emissions also appear significant 1n Cook (9%); however, they are less
evident In- Harris (2%) and Los Angeles (1%). It becomes apparent through
such- Illustrations that sources not significant at the national level may
be- relatively Important for particular areas of the country.
The smallest and least populated of these selected counties, Sullivan
shows; the greatest majority of risk contributed by Industrial point
sources (97%), 1n particular, manufacture of fibers, Inorganic chemicals,
plastics and organic chemicals. The primary pollutants of concern are
3-15
-------
Table 2-3
Sumtary of Cancer Incidence for 35 Counties
listed by Category of Source, in Order of Total Incidence
County CTO—yr) (70-yr) (70-yr) Total Total
Point Area POTW Estimated Estimated
Sources Sources Sources Lifetime Annual
Cancer Cancer
Incidence Incidence
Los Angeles, CA
541.4
862.0
a.6
1412.1
20.2
Baltimore City, MO
438.8
112.7
1.3
552.7
7.9
Bergen, MO
328.5
96.9
2.1
444.0
6.3
Cook, II
124.5
281.0
38.5
426.5
6.1
Queens, MY
93.1
217.1
26.7
336.9
4.8
Oakland, NI
43.9
248.3
0.0
292.1
4.2
Harris, TX
122.5
128.6
4.2
255.3
3.6
Middlesex, NJ
210.7
15.0
17.1
242.9
3.5
Orange, CA
61.0
147.8
1.1
209.9
3.0
Philadelphia, PA
18.7
129.2
32.7
180.6
2.6
Wayne, J»I
81.2
91.2
1.9
174.3
2.5
Essex, M
81.7
65.5
11.0
158.2
2.3
Burlington, NJ
142.4
5.5
0.3
148.1
2.1
Netro Boston, (w
46.5
97.9
1.5
145.9
2.1
Santa Clara, CA
13.1
104.4
0.7
118.2
1.7
Cuyahoga, OH
63 ,7
49.5
0.1
113.3
1.6
Allegneny, PA
52.3
40.2
1.8
94.2
1.3
Marion, IN
63.2
24.0
0.2
87.4
1.2
Oallas, TX
9.2
64.6
1.1
74.9
1.1
Jefferson, KY
21.7
38.1
0.5
60.3
0.9
Nassau, NY
5.7
47.3
0.6
53.5
0.8
Sullivan, TN
50.1
1.5
0.1
51.7
0.7
Sumrit, OH
27.0
10.1
0.4
37.5
0.5
Almeda, CA
12.7
20.1
2.6
35.4
0.5
Tarrant, IX
11.7
13.3
1.4
32.9
0.5
San Diego, CA
7.3
21.8
0.0
29.1
0.4
Jefferson, TX
12.7
3.2
0.7
16.5
0.2
Maricopa, AZ
2.7
10.8
0.7
14.3
0.2
West Baton Ro, LA
9.9
0.5
*
10.4
0.1
East Baton Ro, LA
1.2
9.2
*
10.4-
0.1
Sedgeaick, KS
0.3
9.0
0.8
10.0
0.1
Kanawha, WW
3.6
5.7
0.4
9.6
0.1
Calcasieu, LA
5.9
z.\
*
8.0
0.1
Galveston, TX
3.2
l.o.
0.3
4.2
0.1
Brazoria, TX
3.2
0.2
0.0
3.4
0.0
TOTAL
2721.1
2975.5
159.3
5855.9
83.6
*Oata.not available. „ ,
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only. 3-16
-------
vinyl chloride and acrylon1tr1le. Baltimore also shows a significant
point source contribution (76%), attributed primarily to chromium
releases from metal manufacturing. The relative importance of
Baltimore's Industrial sources, however, may be partly an artifact of the
high risk numbers used for total chromium — 1f this were lowered,
Baltimore's cummulatlve risk picture, like the four other heavily
urbanized counties shewn, might tend to be similarly dominated by area
source contributions.
Focusing on the contribution of particular source categories to total
estimated cancer Incidence, there 1s also significant variation among the six
Illustrative counties. For example, whereas metal manufacturing Is Important
for Baltimore (Figure 3-9), this category does not appear significant for Cook
(Figure 3-10), where the leading two sources are both area-related; the same
applies for any of the other five counties. Harris (Figure 3-11) 1s dominated
by road vehicles, refining, and chemicals; solvent usage — so Important in
Cook — ranks fourth 1n Importance, and 1s only half as large in absolute
terms as It 1s 1'n Cook. Los Angeles (Figure 3-12) 1s Interesting-in that out
of the five leading apparent categories of cancer risk, four are related to
area sources. Philadelphia, too (Figure 3-13), shows four out of five leading
sources to be area-wide; the large contributions of POTWs appear as second
ranking 1n importance overall. The exception of this group 1s the less
populated Sullivan (Figure 3-14), where, as Indicated above, almost all the
hazardous, air pollutant risk appears to come from Industrial point sources.
Table 3-4 summarizes the variation In cancer Incidence across these six
counties for all pollutants studied; substantial variations are shown.
3.* Findings on Risk to the most Exposed Individuals
The- preceed1ng: sections addressed, cumulative Impacts and the
significance.-of sources and pollutants. However, the significance of
sources: depends on one's, perspective. What. 1s a relatively small impact
on the national level, may be quite significant for some counties.
Similarly,, arr evaluation- of cumulative exposure within a county may not
3-17'
-------
1%
Baltimore. HD
:a
i
i rn
1 irfW
>cjc Venic'es
!~
Solvent L'sao?
• ~
i
Waste Oil Burning '
Gas
i ™
.-ieaf.r.g
i i_i
Ai; otheri
8ecause of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Principal Sources of Risk for Baltimore County
Figure 3-9
3-18
-------
253
Cook, IL
53
xt
S ,. fevr
/•; v^, i7*
¦j -- ¦¦- ¦
¦ r-,'aS^
¦
a
Road Ventci*?
Solvent Usage
~
Waste Oil 3i;rn
~
6as Market
~
POTW Vol
d
. fc=3
~
Uti! 'ties
E!ec Eauio MF
!~
AH other
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the nurbers should be regarded as rough in-
dicators only.
Principal Sources of Risk for Cook County
Figure 3-10
3-19
-------
. V-.\V\
^\\\V
las
Harris. TX
!¦
j
i
P.oad V»nic!?a
j E!
Petro Pefine j
! ^
Org Chem j
! r-»
11—j
i
\
Solvent Usaqe i
;~
'3as Market
i
i
Was'i 0i: 2:jrr. |
!~
Heating
t
iu
Aii others
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Principal Sources of Risk for Harris County
Figure 3-11
3-20
-------
jjj"
9R
Los Anqeies, CA
¦
Motor Vshic!?s
i 2
i
Qsf'-P'fiy
i n
i «—*
Gas :iarkat ina S
i L-i
Solvent osao* >
I .
i U
1
Waste Oil Bummg ¦
I s
. —'
r-3ns Eauio
13
Cherr, "rec
1 i—i
Aii others
8ecause of the uncertainties in the incidence
calculations used to derive these estimates,
the nuntoers should be regarded as rough in-
dicators only.
Principal Sources of Risk for Los Angeles County
Figure 12
3-21
-------
jt i jIuit;jr <
m i I * 1111 i t M1 *\
s ?: i M I {(; j j ji 114; j j |;», •.
Philadelphia. PA
!a
i
1
Road Vehicles ¦
Its?
i"'
POT// Vol ;
!~
Wasts Oil 3ur~. 1
! r"1
l
, Lj
i
Solvent Usage ¦
jQ
Gas M'ir^.st
jsa
1H
Crjcas ¦:< Dyes
i
i i_i
All others
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicators only.
Principal Sources of Risk for Philadelphia County
Figure 3-13
3-22
-------
\or.
/ -nikim
/ • ... Ni'iiH;
M^^SggSgar
»:
Sullivan. TN
;a
,-iber M.f
~
incrg Cher"-
i ^
Org Chen
i ~
Exoio 3*ves
¦ ^
'
«;3,; Mf
i m
> t_J
Kcad Vehicles i
: ¦—11
Aii others !
Because of the uncertainties in the incidence
calculations used to derive these estimates,
the numbers should be regarded as rough in-
dicator; only.
Principal Sources of Risk for Sullivan County
Figure 3-14
3-23
-------
Table 3-4
Variation of Lifetime Cancer Incidence, by Chemical, for 6 Counties*
Philadelphia
Sullivan
Harris
Baltimore
Ccok
Los Angeles
Total
Tota
PA
TN
TX
MO
IL '
CA
Lifetime
Annua 1
lenzene
44
6
58
54
103
411
676
9. -
otal Chromium
26
0
18
423
58
124
649
9. •
Formaldehyde
16
1
69
17
29
258
390
5.6
Tasoline Vapors
18
0
21
11
43
159
252
3.6
Trichlorethulene
9
0
10
7
63
148
237
3.4
.'erchlorethylene
16
0
18
12
43
145
234
3.2
Acrylonitrile
4
18
20
2
17
44
105
1.5
•thylene Oichloride
31
0
22
1
26
0
80
1.1
/inyl Chloride
0
26
6
0
0
22
64
0.9
B(a)P
4
0
3
7
6
21
41
0.6
Ethylene Dibrcmide
3
0
3
3
6
24
39
0.O
:adnium
4
0
c
¦J
1
6
20
36
0.5
Arsenic
3
0
2
1
17
13
36
0.5
Total Nickel
1
0
1
0
2
12
16
0.2
Carton Tetrachloride
0
0
0
0
5
0
5
0. ;
Chloroform
0
0
0
0
O
C
0
2
0.0
Beryllium
0
0
0
0
0
0
0
0.0
Styrene
0
0
0
0
0
0
0
0.0
1,3-Butadiene
0
0
0
0
0
0
0
0.0
Pen tach1oropheno1
0
0
0
0
0
0
0
o.c
Coke Oven Emissions
-
-
-
13
18
-
-
-
Total
179
51
256
552
442
1411
2891
41.3
Signifies that no data were available.
3-24-
-------
Indicate relatively high Impacts; however, localized ambient
concentrations 1n the Immediate vicinity of large Industrial facilities
may be substantial. Since residences often adjoin Industrial property,
some Individuals may be exposed to concentrations well above the
population-weighted average for a county.
This study was designed to evaluate cumulative impacts.
Nevertheless, a screening-level analysis was performed to address the
risk to the most exposed Individuals. The Intent was to highlight this
Issue and to provide perspective. A mlcroscale analysis would need to
have been performed to definitively address this problem.
The Atmospheric Transport Model (ATM) was used to model the
emissions from major point sources. Output was computed for 16 wind
directions and sector segments of 0-1, 1-3, 3-5, 5-10, 15-20, 20-25,
25-30, 30-40, and 40-50 km. Based on this output, maximum concentrations
were extracted for six facilities associated with substantial emissions
of toxic compounds. These can serve as examples; the 11st could be
expanded. Clearly, for many of these facilities, higher Impacts could
occur at the fencellne. In addition, 1t should be noted that the
emissions were assumed to be released from one stack at each facility,
building nonreactlve. Table 3-5 summarizes the results.
3-25
-------
Table 3-5
Host Exposed Individuals
fruity
County
State
Po|lutant
Concentrations lug/n?)***
Normalized Actual
to 1 kkg/yr
Emissions
(kkg/yr)
Potency
(ug/m?)-'
Haximum
per capita
risk*
Hooker
Burlington
WJ
Vinyl chloride
6.4xl02
56.7
886
2.61x10-®
I.5X104
piamoncf S(iamrpc|c
mrrls
TX
perchloroethylene
1.20xHT2
6.0
496
1.7X10"6
l.OxlO-5
Goodyear**
Harris
T*
1,3-Butadiene
9.71xl0-2
15.0
152
4.6x10"'
6.9x10"®
fMC
Kanawha
wv
Carbon tetrachloride
3.97xl0"4
0.4
991
1.SxlO-5
6.0x10"®
PPG Ind.
Calcasieu
IA
Trichloroethylene
3.36xl0-2
5.2
IS4
4.1x10"®
2.1xl0-5
Allied
Baltimore Cjty
Total chromium
8. IOxKT2
0.3
3.5
1.2xl0"2
3.6xl0-3
ffonsaptg
prazoria
TX
Acrylonitrile
1.2xl0-3
0.2
160
6.8x10"**
l.4xI0-5
* These are strictly screening estimates and should not be interpreted as a definitive risk assessment.
** These estimates include 0.3 ug/m^ Incremental localized iopact frun Petro Tex. Other source areas were not found to have significant impacts
from other sources reviewed.
t** For these calculations it was assumed that background concentrations are zero.
Because of the uncertainties in the incidence
calculations used to derive tfcese estimates,
the punbers should be regarded as rough in-
dicators only.
-------
4.0 REFERENCES
DeAngells 0, Ruffln 0, Reznlk R.. 1980. Source assessment: Residential
combustion of wood. Research Triangle Park, NC: USEPA. Office of
Research and Development. EPA-600/7-80-040.
General Software Corporation. 1983. Comparison of the atmospheric
transport model with the Industrial source complex long term model and
human exposure model. Washington, D.C.: USEPA. Office of Toxic
Substances.
General Software Corrportlon. 1984. User's guide for prototype GAMS.
Washington, D.C.: USEPA. Office of Toxic Subtances.
LaShelle, Smith. 1974. Characterization of atmospheric emissions from
polyurethane resin manufacture. Research Triangle Park, NC: USEPA
PB 237420.
PEDCO. 1983. A risk assessment of waste oil burning 1n boilers and
heaters. Final draft report. Washington, D.C.: USEPA. Office of Solid
Waste.
USEPA. 1984. The magnitude and nature of air toxics problems In the
United States. Washington, O.C.: USEPA. Office of A1r and Radiation and
Office of Policy Analysis.
3-27
------- |