United States Office of Air Quality EPA-453/R-93-046
Environmental Protection Planning and Standards September 1 993
Agency Research Triangle Park, NC 2771 1
Ar
Schedule for Standards:
Methodology for the
Source Category
Ranking System
FINAL
-------
DISCLAIMER
This report has been reviewed by the Emissions Standards
Division of the Office of Air Quality Planning and Standards,
Environmental Protection Agency, and approved for publication.
Mention of trade names or commercial products is ~ot intended
to constitute endorsement or recommendation for use. Copies
of this report are available through the Library Services
Office (MD-35), U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina 27711, or from National
Technical Information Services, 5285 Port Royal Road,
Springfield, Virginia 22161.
ii
-------
CONTENTS
Section Page
Disclaimer ii
Tables iv
Figures v
1.0 INTRODUCTION 1
1.1 BACKGROUND 2
1.2 REPORT ORGANIZATION 3
2.0 SUMMARY OF METHODOLOGY 5
3.0 EMISSIONS ESTIMATES 11
3.1 EMISSION STANDARDS DIVISION (ESD) APPROACH 14
3.1.1 Production/Consumption Database 14
3.1.2. Point Sources 15
3.1.3. Modeled Area Sources 20
?/} 3.1.4 Uncertainties and Limitations 20
^ 3.2 NATIONAL EMISSIONS DATA SYSTEM (NEDS)/SPECIATION APPROACH . 21
^ 3.2.1. Point Sources 21
') 3.2.2. Modeled Area Sources 23
. _ 3.2.3. Limitations and Uncertainties 23
V 3.3 TOXIC RELEASE INVENTORY SYSTEM (TRIS) APPROACH 23
^ 4.0 HEALTH EFFECTS : 24
rf> 5.0 EXPOSURE SCORES 32
VA£ 5.1 METHODOLOGY 33
5.2 MODELED AREA SOURCES 35
5.3 FUGITIVE SOURCES 36
5.4 POINT SOURCES 39
6.0 SOURCE CATEGORY SCORING AND RANKING 44
7.0 REFERENCES 47
m
-------
TABLES
1 SCRS Emissions Data Requirements 12
2 Overview of the Three General Approaches Used to Estimate
Emissions 13
3 Summary of Average ESD Emission Factors, Emission Rates, and
Discharge Parameters , 17
4 Facility List for Monomer "A" Production 19
5 Consumption Source Categories Emitting Monomer "A" 19
6 U. S. Monomer "A" , 20
7 Dose Conversion Methodology and Associated Species Weight and
Respiration Data 28
8 SCRS Emissions Data 32
9 SCRS Population Data . . 33
10 Source Category Exposure Data 43
-------
FIGURES
1 Schematic Diagram of the SCRS Methodology 6
2 Methodology for Producing Source Category Risk Scores 10
3 SCRS Scoring and Ranking Scheme 45
-------
1.0 INTRODUCTION
This report explains the methodology used by the Source Category
Ranking System (SCRS). The SCRS is a tool devised by the U.S.
Environmental Protection Agency's Office of Air Quality Planning and
Standards to help prioritize source categories for regulations development
under the Clean Air Act, as amended in 1990 (CAA). The SCRS addresses the
first 2 criteria stated in Section 112(e)(3) of the CAA. These criteria
require that the EPA consider, when prioritizing source categories for the
development of the schedule for standards, "adverse effects of" hazardous
air pollutants (HAP's) and "the quantity and location of emissions." The
system generates a score for each listed source category based on emissions
estimates, estimates of toxicity of the HAP's and, to a lesser degree, the
location (when known) of the emitting facilities within a listed source
category. Details of the scoring methods and data input are discussed in
subsequent chapters of this document.
The SCRS does not generate population exposure assessments per ie,
and does not generate risk estimates. The results of the SCRS are, at
best, a relative ranking of source categories based on readily available
information on emissions and toxicity. The data incorporated into the SCRS
are of greatly varying quality and depth, and the algorithms contain many
assumptions. Therefore, the results of the SCRS should not be
misinterpreted. The SCRS was designed as a tool to aid in prioritizing the
source categories listed under Section 112(c) of the CAA.
While the SCRS assisted in developing the schedule, it was not the
only factor considered. Some of the other.considerations that may
influence the regulatory schedule include, but may not be limited to, the
efficiency of grouping categories according to pollutants emitted or
processes or technologies used, the degree of current emission control
under the CAA or other statutes, the efficiency of continuing regulatory
-------
efforts currently underway, the amount of information available on
demonstrated emission control technology, and the nature of the adverse
health effects associated with a source category.
For purposes of the SCRS, source category designations generally are
similar to those used by previous regulatory programs under the CAA, for
example, national emission standards for hazardous air pollutants (NESHAP)
under Section 112 of the Act and new source performance standards (NSPS)
under Section 111 of the Act. There were, however, no explicit or
predefined source category definitions developed for the SCRS. Because of
the large number of pollutants and sources involved, source category
designations were influenced heavily by the data sources that were
available to EPA. These data sources represented varying degrees of
quality and precision.
For these reasons the designation of source categories in the SCRS
was not necessarily the same that EPA uses for specific regulatory
purposes. As more information becomes available during the regulatory
process, the Agency may redefine, combine, or subcategorize source
categories in order to most effectively accomplish the objectives of the
CAA. From this view point, the SCRS provides a starting point for
initiating the regulatory process. The system is highly flexible and
allows incorporation of additional source categories as new information
becomes available.
As a general rule, a source category is a group of processes or
equipment that are functionally and economically related for the purpose of
producing a product or an intermediate. Examples include petroleum
refining, benzene production, synthetic rubber production, metal coil
surface coating, steel foundries, and stone quarries. Source categories
may also include operations which are common to a number of industrial
processes and are best handled separately, such as solvent degreasing
operations or external combustion space heating.
1.1 BACKGROUND
In the past, NESHAP have been promulgated on a pollutant-by-pollutant
basis. The program relied primarily on assessments of the public health
risks from exposure to single pollutants. This narrow definition of
exposure and risk can underestimate the total risk posed by the mix of
-------
pollutants commonly emitted by many industries. As the Agency developed a
better understanding of the nature and extent of the air toxics problem,
the inefficiency and ineffectiveness of this approach became apparent. For
example, under this approach, the source category of acrylonitrile-
butadiene styrene (ABS) resin production was first studied because of the
possible public health risks from acrylonitrile emissions. Later, after a
decision had been made not to develop a NESHAP for acrylonitrile, ABS resin
production was assessed again for emissions of 1,3 butadiene, and still
later for styrene emissions. None of the studies considered the public
health impacts of all three compounds collectively. Repetitive assessment
of source categories is an unnecessary burden for EPA, the industry, and
the public.
Another disadvantage of the pollutant-by-pollutant approach is that
it forced the Agency to make regulatory decisions for all source categories
that emit a particular hazardous air pollutant. When assessing a single
hazardous pollutant, a given set of source categories may account for all
the significant public health risks. Other source categories may be
identified that release a small amount of hazardous emissions but pose an
inconsequential impact on risks. Often, the regulatory process has been
slowed and scarce resources spent to complete risk assessments and make
regulatory decisions for these low risk source categories.
To enhance the effectiveness of the program, EPA began developing a
new approach for air toxics regulations. The new approach involved
prioritizing and assessing source categories, considering multi-pollutant
impacts on public health. The major advantages over the pollutant-by-
pollutant approach are that it speeds the regulatory process, avoids
repetitive consideration of source categories, and provides program focus
on the highest priority regulatory candidates from the viewpoint of public
health risks. The Agency developed the SCRS to help implement this new
approach. The SCRS helps prioritize source categories emitting hazardous
air pollutants for the purpose of developing national emission standards
under the CAA.
1.2 REPORT ORGANIZATION
Section 2.0 of this report summarizes the overall methodology
employed by the SCRS. In Section 3.0, the background for developing the
-------
emissions component of the SCRS is described. Information on the
development of health effects and exposure scores is provided in
Section 4.0 and 5.0, respectively. Section 6.0 describes how the health
effects and exposure scores are integrated to produce various rankings.
Section 7.0 contains references cited in this report.
-------
2.0 SUMMARY OF METHODOLOGY
The Source Category Ranking System was developed by EPA to identify
and prioritize source categories that emit potentially hazardous air
pollutants. The SCRS is a tool that combines emissions estimates, health
effects data, and limited population data to rank source categories. A
major strength of the SCRS is its flexibility, which allows the emissions
and health effects data input to the System to be drawn from multiple
information sources and varied according to the user's needs.
Figure 1 is a schematic diagram of the System. The following
paragraphs describe in general terms the information sources used at each
step in the diagram and the analytical methods used to integrate the
information to produce the ranking of source categories. These steps are
described in greater detail in the remaining sections of this report.
Specify Pollutant List
The initial step in the SCRS is user specification of a list of
pollutants for which the source category list will be generated. The list
may contain both organic and inorganic pollutants. Once the list is
generated, it may be modified relatively easily to add or delete
pollutants. For the purposes of Title III, the SCRS has been implemented
using the list of 189 toxic pollutants contained in the CAA of 1990.
Development of Source Categories
The source category list includes known source categories that emit
one or more of the specified pollutants. Identification of source
categories is dependent on the availability of resources and references
that relate emissions information to specific industrial sources. To be as
-------
Estimate Emissions by
Source Category
Develop Hea]
Effect Scoi
Concentration
Estimates
U.S. Census Data
Develop Exposure
Scores
Develop Combined
Scores
Develop Source
Category Scores
Rank Source
Categories
Figure 1. Schematic Diagram of the Source Category Ranking
System Methodology
-------
complete as possible in identifying source categories, the System
interfaces with several automated databases and allows the user to input
manually from other available references or studies.
The SCRS may access EPA's National Emissions Data System (NEDS) and
the Toxic Release Inventory System (TRIS) for source category data. In
addition, the System's data files contain information developed by the EPA
during the course of recent NSPS and NESHAP regulatory activities. In
particular, the Emission Standards Division (ESD) of the Office of Air
Quality Planning and Standards has developed a data file of source
categories for the synthetic organic chemical manufacturing industry
(SOCMI). The ESD approach is described in Section 3.1 of this report. The
ESD data have been incorporated into the SCRS data files and are available
for use. For any given application of the System, the user may choose to
use other available information such as combinations of available automated
databases or the ESD data files for generating a source category list.
Emissions Estimates
For each source category, the emission estimation technique depends
on whether the source category is comprised of point sources or "modeled
area sources". For point sources, emissions are estimated on a plant-by-
plant basis. For each plant, emissions estimates are developed for both
stack and fugitive emissions. For modeled area sources emissions are
estimated for the source category as a whole. The term "modeled area
source" in the SCRS does not have the same meaning as "area source" as
defined in Section 112 of the CAA. In the SCRS a modeled area source is a
source category for which information on individual facilities is not
available, and therefore, emissions are reported in the aggregate. These
source categories may contain CAA-defined major and/or area sources, but
are modeled as area sources in the SCRS. Many of the information sources
used to develop the source category list also are used in making emissions
estimates. For example, NEDS and TRIS contain emissions estimates for
individual plants. In addition, NEDS contains area source emission
estimates. The ESD database contains emission factors from which emissions
for the SOCMI industry may be estimated based on plant production and
consumption data.
-------
The SCRS has been designed to accept data from information sources
other than the ones listed above. As described in Section 3.0, emissions
estimates may be developed using one approach or a combination of
approaches that include data from multiple references.
Health Effects
In order to rank source categories, the SCRS requires that a health
effects database be assembled for each of the 189 chemicals and group of
chemicals listed in Title III of the CAA. Four general types of health
effects are evaluated for each hazardous air pollutant in each source
category: carcinogenicity, reproductive toxicity, acute lethality, and
other toxicity, each of which is explained in Section 4.0 of this report.
Exposure Scores
The exposure scores do not accurately represent actual population
exposure since the many assumptions, constants, limited population data and
estimated emissions incorporated into the score may not accurately
represent the actual situation for any given source category. For example,
if facility location data was not available for a source category, then a
constant based on average national population density was used to calculate
the long-term aggregate exposure score. Exposure scores are
mathematically-derived by processing emissions data through simplified
atmospheric dispersion algorithms to develop estimates of ambient pollutant
concentrations at each facility. These ambient concentrations are then
multiplied by limited population data to obtain the exposure score.
Four types of exposure scores are derived in order to match short-
and long-term health effects: long-term maximum exposure, long-term
aggregate exposure, short-term maximum exposure and short-term aggregate
exposures. Details about how the exposure scores are calculated are
provided in Section 5.0 of this report.
Develop Source Category Scores
The source category scores are derived by combining the health
effects scores with the exposure scores for each pollutant and summing the
scores for all pollutants emitted by the source category. The data files
(developed in previous steps) contain four different exposure parameters
8
-------
and a variety of health effects end points. Figure 2 provides a general
overview of how the health and exposure scores are combined. The process
is explained briefly below and in more detail in Section 6.0.
The System first condenses the health effects data for each pollutant
into a single long-term score and a single short-term score. To do this,
the user must input a set of health effects multipliers. These multipliers
specify the types of health effects to consider and the relative weight to
assign to each. In a parallel step, the exposure scores for each source
category are condensed into a single long-term exposure score and a single
short-term exposure score for each pollutant emitted.
The System then combines this information to produce the overall
scores for the source category. The exposure scores and health scores are
multiplied to produce a short-term and long-term pollutant score for each
pollutant emitted. Then, the two scores are combined to form a single
score by pollutant, using weighing factors specified by the user. This
weighing factor allows the user to vary the relative importance placed on
long-term and short-term health effects for the purposes of producing any
single SCRS ranking.
This process is repeated for each pollutant that is emitted from a
source category, and the individual pollutant scores are summed to produce
the final scoring for the source category. Source categories then are
ranked by a sorting based on the final score.
-------
Health Effects
Data File
Input
health effect
multiplier
Compute a
short- and long-
term health
effects score
Source Category
Exposure Scores File
Pollutant A
Compute a
short- and long-
term exposure
score
Pollutant
Compute a
short- and long-
term pollutant
score
Input
short- and long-
term weighting
factors
Pollutant
Compute net
pollutant
score
Accumulate pollutant
scores to,produce
source category
score
I
Rank source
categories
by score
Figure 2. Methodology for Producing Source Category Scores
10
-------
3.0 EMISSIONS ESTIMATES
This section explains the procedures for estimating emissions in the
SCRS. Both point and modeled area sources (see Chapter 2.0 for
clarification of "modeled area sources") are included in the System. Point
source emissions are generated by estimating emissions from individual
plants, assigning plants to source categories, and summing emissions for
each source category. For modeled area sources, emissions are estimated in
the aggregate for the entire source category. Several optional methods are
available for estimating emissions for both modeled area and point sources.
These methods are described in detail in this section.
The System requires development of an emissions database for each
source category. Table 1 lists the emissions parameters and other
information required. This information is used to estimate emissions and
to generate exposure scores (Section 5.0) associated with each source
category. Many of these parameters in Table 1, such as pollutant release
height, temperature, release flow rate, and vent diameter, were not
incorporated for each source category individually in the calculation of
exposure scores; instead, representative values were incorporated into
constants in the simplified algorithms described in Chapter 4.0 "Exposure
Score".
11
-------
TABLE 1. SCRS EMISSIONS DATA REQUIREMENTS
Emissions rate (kg/yr)
Emissions type (point, area, fugitive)
Pollutant release height (m)
Pollutant release temperature I'C)
Pollutant release flow rate (m/s )
Pollutant release vent diameter (m)
Pollutant identification
Plant identification
Source category
State/county code
Some reference documents and automated databases contain
information for all of the parameters shown above. Other references
contain only chemical production/consumption data or emission factors.
Information from these types of references must be combined with SCRS
default values to obtain parameters such as release height and release
temperature. Thus, emissions estimates may be developed using one approach
or a combination of approaches using data from multiple references.
Four general approaches were used to develop emission estimates for
the current SCRS: the Emission Standards Division (ESD) Approach, the
National Emissions Data System (NEDS) Approach, literature searches of
existing EPA studies, and the Toxic Release Inventory System (TRIS)
Approach. A hierarchy was established in order to apply the most
app opriate technique for each source category. The ESD approach was used
for certain source categories of organic chemical manufacturing processes.
Next, the NEDS approach was applied. Then, a review of existing EPA
studies was done, including, but not limited to: source assessments of
pollutants, locating and estimating documents, and data developed in
support of previous Section 111 and Section 112 regulatory decisions.
Finally, the TRIS approach was used for source categories not covered by
the other approaches. The ESD, NEDS, and TRIS approaches are explained
briefly in the fol wing paragraphs, and a summary table containing the
strengths and linr :ions of each, is shown in Table 2.
One feature the SCRS is that it allows the user the flexibility to
use these three databases in a variety of ways to generate source category
rankings. For example, a ranking can be produced using either NEDS or TRIS
12
-------
trt
O
U)
I
kU
ESTINAIi
O
o
UJ
V)
t/1
£
a.
*
«
3
UJ
vu
Ul
0
3
LU
oc
nj
u
I
Limitations and Uncertainties
.c
^
u»
-g
«tf
Approach
-Ihe ESO approach can only be
used for organic chemical
manufacturing processes for uhich
published production/consumption
data are available.
41
s si
5 "O
J*3
s|]
^ u
0
« u u
» « s
«. *< a
4- (A
SO +- V
O »>
« *rf Ul
< C VI 41
2 2 S.01
i i 5 2
UJ C ** &
1
.-
3
*=8
ill
i (4
II-?
>*«***
* §J
o S 8
a |.s
^ &
»>
M!
I5J
|
Ul
-Ihe ESO approach uses values that
are averaged across large industry
cross -sect ions. It does not take
into account at 1 processes or
chemical-specific information.
w V
«* S
c « s
v e u
t» .2 M
i .5 Z
O f u
u C «
« 3
tA C M
U (1
3 ^ C
g " 1
O
< O U
11 = 1
'
-Ihe year-to-year variability of
production/consumption data add
uncertainty to this approach.
-NEDS only contains data for point
sources emi t ting more than 100 tons
per year of a criteria pollutant.
,
|.
i_
u
8 c
si
r
|l
a
3
VI O
0 J= **
ill
U " V
x 15
O
* * -
- 2!
c " t
S*< u
i*
!s?
§.222
n|!
Hit
,
1
(A
13
-------
only. Alternatively, a ranking could be generated using a hierarchy of
databases that reflects the user's perception of the quality and
reliability of the data. Thus, a ranking could use TRIS as the primary
data source, with NEDS used as a supplement for source categories riot found
in TRIS. Another example would be to use the ESD approach for the SOCMI
source categories with NEDS used for all other source categories arid TRIS
used as a supplement for source categories not found in NEDS. The System
requires that the user specify the data sources for each ranking.
3.1 EMISSION STANDARDS DIVISION (ESD) APPROACH
The ESD Approach relies on published production/consumption data for
organic chemicals in the specified pollutant list, and emission factors
developed by ESD. It has been used by EPA to complete preliminary source
assessments for several organic chemicals. As noted in Table 2, use of
this approach is limited to organic chemical manufacturing processes,
defined here as processes that produce organic chemicals or consume organic
chemicals as feedstock to produce other chemicals.
3.1.1 Production/Consumption Database
Production and consumption data were obtained for each chemical from
readily available literature. The majority of production data were
available from SRI's Directory of Chemical Producers, while most
consumption data are taken from Chemical Marketing Reporter and
Mannsville . Chemical Products Synopsis.
Source categories were identified for the production and use of the
chemicals and defined as all emission points, including raw material
storage, process emission points (e.g. process vents, leaks from valves,
flanges, and drains, and wastewater treatment), and product storage or
loading. Where applicable, source category definitions developed from
previous regulatory assessments were maintained.
Production of one chemical or a group of similar chemicals in the
same process was considered as one source category. For example,
production of 1,3-butadiene is one source category, and production of
chlorinated hydrocarbons is another. The latter includes production of
methylene chloride, chloroform, trichloroethylene, carbon tetrachloride and
perchloroethylene.
14
-------
In some cases, the end use of one of the chemicals was considered a
separate source category, while in other cases source categories
represented the use of several chemicals in one process. For example, the
category of degreasing/metal cleaning was identified as an end use of
toluene, trichloroethylene, and several other pollutants. Since the
consumption of those chemicals was distributed among several end uses,
consumption values were allocated to source categories using percentages
found in the literature. For example, 90 percent of all trichloroethylene
was used for degreasing and metal cleaning. Thus, 90 percent of
trichloroethylene consumption was assigned to the degreasing/metal cleaning
source category.
After identification of the source categories, the Directory of
Chemical Producers was used to assign specific facilities to appropriate
source categories. Source categories with available facility lists were
identified as point source categories. If a list of facilities was not
available, the point source category was treated as an area source.
3.1.2. Point Sources
Emission estimates were developed on a source-specific basis by using
emission factors presented in Table 3 (U.S. EPA, 1987a). Production and
use of organic chemicals were divided into seven groups-- production
processes, polymerization, solvent use, reactant use, blowing agents,
pesticide production, and Pharmaceuticals productionaccording to
similarities in processes. Within each facility production or consumption
of a pollutant was associated with one of the seven process groups.
Within each of the process groups are emission factors for:
process vents
storage
equipment openings
handling
secondary waste streams
equipment leaks
liquid spills
accidental gas releases
relief discharges
15
-------
These emission factors were based on pollutant-specific data collected by
ESD through Section 114 of the Clean Air Act.
The ESD emission factors (U.S. EPA, 1987a) was based on the
assumption that emissions from process vents, storage, handling, secondary
waste streams, and equipment openings are a function of consumption or
production of the pollutant. Table 3 presents the emission factors for
these operations. Equipment leaks, relief valve discharges, accidental
gaseous releases, and liquid spills were assumed to be independent of
consumption or production rate. For these operations, Table 3 presents
average annual emission rates. The table also shows the release
parameters and a designation for each operation as either "stack" (S) or
"fugitive" (F).
Facility-specific production data were obtained from the Directory of
Chemical Producers when available. When specific production data were not
available, total production for the source category was divided equally
among the facilities in the source category. Consumption of a chemical in
a process was assumed to be proportional to the production capacity of the
chemical being produced. In this case, total consumption for a source
category was divided among the facilities in proportion to the relative
production capacity of each plant. (If production capacities were not
available, then total consumption was divided equally among the plants.)
Emission estimates were developed on a point-source by point-source
basis by multiplying the appr ^-iate emission factor by the
production/consumption rate, cr by assigning the appropriate emission rate.
"he average release parameters shown in Table 3 were then assigned to each
.lission source to complete the necessary emission parameter inputs
required in Section 5.0 for estimating population exposure.
Example of ESD Approach
To illustrate the methodology, this section uses the ESD Approach to
estimate the emissions associated with the production and use of a
hypothetical organic chemical ("monomer A").
The Directory of Chemical Producers lists the facilities that produce
monomer "A" by company name, location and capacity (Table 4). Information
in Chemical Products Synopsis and Chemical Marketing Reporter indicates
that monomer "A" and the primary feedstock used in its production
16
-------
5 J
(A
UJ
O ^
*" u
«1 i
UJ
!i
i/» «
«J
1
s
(A
Ul
's
?
!
o
X
a
V
a
u
V
u
sasss"aa
r- r> o o o o o
j^ooooogg
GO rvj rg ^
tf* « ^ i/^ O O
oooo«« O (v* ra
z z z z
rocess Vents !
torage !
quipaent Openings i
landl ing 1
iecondary Uaste Streaas
quipaent Leaks
iquid Spills
Accidental Gas Releases
elief Discharges !
on Processes
i
a.
-
(M O «- O> O> 00
FO tf\ K1 CNJ CM IM
\f\ ij\ O O ^
i> O NI o O rvi
co rg
(\J K. 1*1 K)
-roo rg
o a o < < o
o - o> a o o
^
Cr»
3 m z z z
'rocess Vents '
itorage '
'quipaent Openings
landling
iecondary Uaste Streaas
quipaent Leaks
Accidental Gas Releases
Ielief Discharges !
i
1
Kt
saaaaaaa
O> rg o
«- -o «
* K1 Kl *- O t- «-
O »- rg ^ rg o O
OOOOO«OO
r-^^>o>rooo
s
»- o
K> »n «* O
'rocess Vents '
itorage !
Equipaent Openings !
landling !
Secondary Uaste Streaas
Equipaent Leaks
Accidental Gas Releases 1
(elief Discharges !
»
I
*
17
-------
w
<«
"
**
*^
U
X
O)
<4f
w
^
u
CL
a
z
>^
0>
V
>
)
V
u
1
uojssiur
1
£
X
£
5
u
Source
1
§!*> N^ m
> > >
Ft IM fNJ IM
-j; o o o
Kl O O O
O in
p- o
0 0 < <
z z
o> %»
o- -o o o
O* >f K1 O
< < O !"-
Z Z CO IM
O ^
"I
S "" z z
>
(A V% Ifc tfc
U)
^
m v a
g : =
* ISf
3 S!T,
o o a--
^ *a*
U M 8"
<
fctf
i
c
?
1
ca
^n
f^« fe^ B^ f^ ff^ fj^ ff^ ff^
saaaaaaa
o o o o o o
J-OOOOOO1M
flO flO ^ ^
^ ° *" .
» = =;<<<<«=;
!C ~° o S
IM "* "** W 2
<«<««QOO
""ggs
kT>
f» 1O
§"liS0SzZZ
«,««.-.»«.«.-.
i
tl
I/I W
? " S
fw j< c.
S 8^5
^ '^iJ
«i«a.Cga-'v
I
X
g
a.
3
u
w
o
~~
«3
S2
o a
0 e
< ^
S N
«,«,
l«
I
>
28
?S
S
1
h.
&
«l
"
U
«^
z
i
i
^
asaaa
0 -0 S 0
*/> o ^ o o
s s
0<0<<
e a o o o
-
« < < o a
-"Is
K1
rS 3 S z
"»--.".
i
V
< M
? v»
-------
("feedstock") are the only pollutants of interest that are emitted during
monomer "A" production.
Chemical Products Synopsis and Chemical Marketing Reporter also
indicate two uses of monomer "A" and the distribution of the relative
consumption between the two. These uses are classified as source
categories. Table 5 lists the categories and the percent of consumption
allocated to each. Table 6 summaries the estimated monomer "A" emissions
from production and consumption using the ESD approach.
TABLE 4. FACILITY LIST FOR MONOMER "A" PRODUCTION
Plant Name
Company # 1
Company # 2
Company # 3
Company # 4
Company # 5
Company # 6
Company # 7
Company # 8
Company # 9
State Name
LA
TX
NY
CA
IL
OH
NJ
OK
PA
Capacity
(Mq/yr)
750
1000
800
400
560
1000
500
840
700
TABLE 5. CONSUMPTION SOURCE CATEGORIES EMITTING MONOMER "A"
Category Name
Pollutant Name
CAS Number Percent
Allocated
Production of polymer "A"
and copolymers
Misc. Monomer "A" uses
Monomer "A"
Monomer "A"
XXOOO
XXOOO
3
97
19
-------
TABLE 6. U. S. MONOMER "A" EMISSIONS
Category Name Total Emissions
(Mq/yr)
Monomer "A" 10,900
Misc. Monomer "A" uses 99,320
Production of polymer "A" and copolymers 46,660
3.1.3. Modeled Area Sources
When a facility list was not available for a source category, the
source category was treated as an area source (i.e., modeled as an area
source). Each pollutant for each area source was classified either as a
solvent or nonsolvent. Solvents were assumed to be 100 percent emitted
while nonsolvent emissions were estimated using an modeled area source
emission factor based on the ratio of total ESD point source emissions to
total chemical consumption.
3.1.4 Uncertainties and Limitations
Although the ESD Approach has been used for previous preliminary
source assessments, and is based on a large amount of emissions data, some
inherent limitations and uncertainties exist. First, this approach can be
used only for organic chemical for which published production/consumption
data exist. Second, the emission factors and rates presented in Table 3
represent averages taken across entire source category groupings. The
result is an emission factor that represents the average level of pollutant
emission control for the entire source category grouping. Because some
pollutants are controlled to a greater degree than the industry average,
(due either to the pollutant toxicity, or to existing state regulation), it
is expected that emissions at some sources may be overestimated using the
ESD Approach. Also, these data are several years old, and the baseline
emissions may have changed.
Third, the variability of year-to-year production/consumption data
adds uncertainty, since much of the emission estimation procedure depends
on production/consumption data. Although the most recent production/
consumption data may be used, fluctuations may affect the accuracy of
emission estimates using this approach, Finally, estimating solvent
20
-------
emissions from modeled area sources as being equal to process or
consumption rate ignores possible destruction or release to other media,
such as wastewater and solid waste. In the absence of a more refined
methodology, however, assuming that 100 percent of solvent consumed is
emitted to the air provides a conservative and consistent methodology for
estimating solvent emissions from modeled area source categories.
3.2 NATIONAL EMISSIONS DATA SYSTEM (NEDS)/SPECIATION APPROACH
The NEDS Approach involves applying speciation profiles to
particulate matter (PM) and volatile organic compound (VOC) emissions data
provided in NEDS (U. S. EPA, 1990). NEDS is an EPA database of reported
emissions from sources emitting more than 100 tons per year of any criteria
pollutant (U.S. EPA, 1988). Speciation profiles divide specific VOC or PM
streams into individual pollutant streams. Thus, pollutant-specific
emissions can be estimated for all reported sources.
3.2.1. Point Sources
Point emission sources are classified by Source Classification Codes
(SCC) within NEDS. An SCC is an 8-digit code divided into 4 levels of
identification signifying: 1) the category process; 2) the major industry
group; 3) the major product; and 4) different operations at the point
source. A list of the approximately 4,000 defined SCCs, along with
descriptions, can be found in Criteria Pollutant Emission Factors for the
1985 NAPAP Emissions Inventory (U. S. EPA, 1987b).
Speciation profiles have been developed or assigned to each SCC
contained in NEDS. These speciation profiles are an estimate of the
chemical species breakdown of the total VOC and total PM. The speciation
profiles may be used to develop pollutant-specific emission estimates from
the VOC or PM emission rates provided in NEDS. Speciation profiles
[developed by EPA within the national Acid Precipitation Assessment Program
(NAPAP)] vary in quality. First, there are three categories of speciation
profiles: 1) original, 2) assigned, and 3) average. Original profiles are
speciation profiles based on emissions test data representing one or more
SCC. Assigned profiles represent assignment of an original profile where
no actual emissions test data are available for the SCC, but the process
type is similar to an SCC that has an original profile. Average profiles
21
-------
represent average speciation profiles across a large industry cross-section
and are assigned to SCCs where no original or assignable profiles are
available.
Each speciation profile has an associated "Data Quality Rating" ("A"
through "E"). These ratings are based on the following:
Data Quality A: Data set is based on a composite of several tests
using analytical techniques such as GC/MS, and can
be considered representative of the total
population.
Data Quality B: Data set is based on a composite of several tests
using analytical techniques such as GC/MS, and can
be considered representative of a large percentage
of the total population. Profiles from the VOC
field sampling program are assigned to data quality
"B."
Data Quality C: Data set is based on a small number of tests using
analytical techniques such as GC/MS and can be
considered reasonably representative of the total
population.
Data Quality D: Data set is based on a single source using
analytical techniques such as GC/MS; or data set is
taken from a number of sources where data are based
on engineering calculations.
Data Quality E: Data set is based on engineering judgment and/or
has no documentation provided; may not be
considered representative of the total population.
The uncertainty associated with the "E" quality profiles is such that they
are substantially less reliable than the other profiles.
The first step in estimating emissions using the NEDS/Speciation
Approach is to identify all SCCs containing one or more of the pollutants
of interest that have speciation profiles of specified quality. These SCCs
are then assigned to source categories based on the descriptions of each
SCC given in Criteria Pollutant Emission Factors for the 1985 NAPAP
Emissions Inventory. The description may also be used to assign to each
SCC a stack/fugitive (S/F) designation similar to those given for the ESD
Approach.
22
-------
3.2.2. Modeled Area Sources
NEDS also contains VOC and PM national emissions estimates for
modeled area sources. Along with typically recognized area sources (e.g.,
mobile sources), NEDS includes anthropogenic emissions from plants that
emit 100 tons/year and from point sources that emit 25 tons/year or which
are too difficult to inventory individually. As with the point sources,
the NEDS emission estimates can be speciated to obtain pollutant-specific
emission estimates.
3.2.3. Limitations and Uncertainties
There are two basic limitations to the NEDS/Speciation Approach.
First, NEDS generally contains emissions data only for plants emitting
greater than 100 tons per year of a criteria pollutant. Since NEDS obtains
data submitted by state agencies, quality control is inconsistent in terms
of completeness, accuracy of data, and timeliness. Second, the speciation
profiles have varying degrees of quality and applicability. Some
speciation profiles are based on outdated data, while others contain data
derived indirectly from an average of data from a broader source category.
3.3 TOXIC RELEASE INVENTORY SYSTEM (TRIS) APPROACH
The TRIS Approach is a third method that may be used to estimate
pollutant emissions. The TRIS database contains emissions data reported by
individual industrial facilities as required by Superfund Amendments and
Reauthorization Act (SARA) Section 313 . The most recent (1987) TRIS
database contains data for facilities producing or processing as little as
5 tons per year of any single chemical covered under SARA Section 313.
Emissions data in TRIS are facility-specific and the Standard Industrial
Category Codes (SICs) are given for each facility.
23
-------
4.0 HEALTH EFFECTS
The CAA of 1990 list approximately 189 individual chemicals and
groups of chemicals (e.g., antimony compounds) as hazardous air pollutants.
The toxicological properties of these chemicals were evaluated for
inclusion in the Source Category Ranking System (SCRS). In the SCRS,
measures of toxicological potency are combined with exposure scores to
produce source category scores that may be ordered in various ways to
provide information for use in prioritizing among source categories.
4.1 BACKGROUND
As noted above, the CAA listed individual chemicals as well as
general categories of chemicals. One question that had to be addressed
before toxicological properties could be evaluated for the chemical groups
was specifically which chemicals to include in the listed groups for
evaluation purposes. In order to characterize the general categories of
chemicals, lists of chemicals included in various other toxic chemical
legislation were consulted. These lists included: Superfund Amendments
and Reauthorization Act (SARA) Section 302 (Extremely Hazardous
Substances); Comprehensive Environmental Response, Compensation, and
Liability Act Hazardous Substances (CERCLA) (i.e., Reportable Quantity [RQ]
Chemicals); SARA Section 313 (Toxic Chemicals); and Resource Conservation
and Recovery Act (RCRA) Hazardous Wastes (P and U Lists). With a few
exceptions noted below, compounds on these lists were used to define the
constituents of the general categories.
Six general categories in the CAA that require special mention are:
(1) coke oven emissions, (2) mineral fibers, (3) polycyclic organic matter,
(4) glycol ethers, (5) radionuclides, and (6) cyanide compounds. For coke
oven emissions, benzo(a)pyrene was used as a surrogate. B(a)P is on the
CERCLA and RCRA lists. As for polycyclic organic matter (POM) and glycol
24
-------
ethers, three compounds for which there is ambient air data were selected
for each group. Some cyanide compounds were not evaluated in the SCRS
because there were no emissions data available to link to the corresponding
toxicological data. In these cases, the SCRS evaluation process could not
be completed. The CAA also list fine mineral fibers including glass, rock,
and slag fibers that were evaluated in the SCRS.
Four types of health effects that may be produced by exposure to
chemicals include: cancer, reproductive toxicity, acute lethality and
other short-term and long-term toxicity (e.g., neurotoxicity). Each of
these generalized endpoints was evaluated for each chemical. Where data
were available, the dose of each chemical that produced evidence of
toxicity for each endpoint was estimated. In this manner, a potency-based
score was developed for each endpoint.
The four general endpoints of toxicity were initially derived from
analyses performed during the development of the Modified Hazardous Air
Pollutant Prioritization System (MHAPPS). Data for several health
endpoints developed for the MHAPPS, however, provided information on the
weight of evidence that exposure to a chemical produced the health effects,
rather than providing information of the relative potency of a chemical in
producing effects. In ranking source categories for regulatory analysis,
the SCRS combines exposure scores with estimates of health potency to
produce source category scores. As such, the SCRS uses potency-based
information and not the weight of evidence approach from MHAPPS.
Each health endpoint was considered separately in development of
health data files for use in the SCRS. The process used to develop health
potency scores for each endpoint of toxicity are described in the following
sections.
4.2 CARCINOGENICITY
The relationship between exposure to a chemical and the potential for
an increased risk of cancer is characterized in two steps. In the first,
the hazardous identification step, the weight of the evidence that a
chemical is associated with cancer is evaluated. Using the EPA Guidelines
for Care ;gen Risk Assessment (51 FR 33992), chemicals are associated with
descript 3 terms, such as "known human carcinogen" that reflect the type
and quality of data available to associate increased risk of cancer with
25
-------
chemical exposures. In the second step, the dose-response evaluation, the
potency of a chemical in producing cancer is estimated. The numerical
constant that defines the low exposure-risk relationship used by EPA in its
analysis of carcinogens is called the unit risk estimate (URE). The unit
risk estimate for an air pollutant is defined as the estimated increased
lifetime cancer risk occurring in a hypothetical population in which all
individuals are exposed continuously from birth throughout their lifetimes
(defined as 70 years) to a concentration of one microgram per cubic meter
(ug/m3) of the agent in the air they breathe.
Weight of evidence classifications for carcinogenicity were obtained
from the Integrated Risk Information System (IRIS), the Health Effects
Assessment Summary Tables, Fourth Quarter FY 1989 (OERR 2006-303-89-4), the
Gene-Tox program (Mutation Research 185[1,2]; 1987), the EPA Carcinogen
Risk Assessment Verification Effort and consultation with EPA staff.
Inhalation unit risk estimates were obtained where possible for chemicals
classified using the EPA Guidelines for Carcinogen Risk Assessment as
possible, probable or known human carcinogens, or classified in Gene-Tox
using the International Agency for Research on Cancer guidelines as having
limited or sufficient positive evidence of carcinogenicity. Inhalation
unit risk estimates were generally obtained from these same sources. Oral
unit risk estimates were converted to inhalation potency estimates based on
surface area equivalency.
In some cases, unit risk estimates were not available from these
sources, and preliminary estimates were developed by dose-response modeling
and/or through regression analysis using Tumor Dose 50 (TD^) data
(Gold et al., Environmental Health Perspectives 58; 1984, p. 9-319). In
some cases, unit risk factors have not been peer reviewed and are subject
to change. In other case, suitable dose-response data were unavailable for
potency estimation, and unit risk estimates were not developed.
4.3 ACUTE LETHALITY
Acute lethality data encompass many relevant health endpoints of
short-term exposure, although short-term release events seldom result in
death. The acute lethality component was also selected because data are
readily available for evaluation.
26
-------
This comf: it consists ;f data on oral, dermal, or inhalation doses
producing leth :y in expos-res of less than or equal to 24 hours. The
routes of exposure considered were limited to oral, dermal, and inhalation
because these routes are responsible for human exposure to air pollutants.
Data for studies using other routes of exposure (e.g., intravenous,
subcutaneous, intraperitoneal) were not considered relevant. An exposure
period of less than or equal to 24 hours was chosen, since this time period
generally covers the range of potential short-term release events.
The potency measures used were the LDgQ or LC50, or if no such
measures were reported, an LD|_Q or LC^Q was considered. An LD50 i-; the
calculated dose which is expected to cause death in half of the
experimental population. An LCg0 is similar, but refers to a calculated
concentration in air. An LDig is the lowest dose of a substance reported
to have caused death in the experimental population, and, similarly, the
L.CLQ refers to the lowest concentration causing death when the test
population is exposed via inhalation. These four values are inversely
related to potency, with pollutants having relatively low LD50 or L.Cr0 (and
LDLO or LC10) values being more potent than pollutants with higher values.
Data on acute lethality were assembled form an on-line search of the
Registry of Toxic Effects of Chemical Substances (RTECS) (on-line database
produced by the U.S. Department of Health and Human Services, National
Institute for Occupational Safety and Health, Cincinnati, Ohio), from which
the lowest LD5Q, LC5Q, LDLQ, and LC^0 values were extracted for each
chemical form the data field for acute toxicity. RTECS is an easily
accessible reference which is concise, is kept current through regular
updates, and contains data on a very large number of toxic chemicals.
RTECS data can be easily interpreted and extracted without the need for
expert judgement.
For all inhalation data given in ppm or mg/m recorded for both
humans and animals, the doses were converted to mg/kg/day, using average
anirr weights and breathing volumes. Data for oral and dermal studies
were ven in the units mg/kg, and assumed to be mg/kg/day. This
con- on process is illustrated in Table 7. Conversion to a common unit
27
-------
TABLE 7.
DOSE CONVERSION METHODOLOGY AND ASSOCIATED SPECIES
WEIGHT AND RESPIRATION DATA1
For converting concentrations (in ppm) to mg/m3:
mg/m3 = ppm x mw (at 70°F and standard pressure 760 mm Hg)
24.45
where mw = the molecular weight of the chemical substance
The molecular weight for the chemical is contained in the RTECS field MW.
For converting mg/m3 to mg/kg/day:
mg/kg/day * mg/m3 x BV x ET/24
SW
where: BV = breathing volume (m3/day)
SW = species weight (kg)
ET = exposure time (hours)
Animal
Human
Mouse
Rat
Guinea Pig
Rabbit
Cat
Dog
Chicken
Monkey (rhesus)
Hamster
Cattle (standing)
Methodf of AfliflH
Tbf VFAW Hind
Minute
Weight
(kg)
70.0
0.0275
0325
0.650
4.0
4.0
20.0
42
2.68
0.092
144.0
,1 Pm^im.m.rioB VoL L William L Gar. ed_
book OB the Cftiv
-------
of measure was necessary so that doses could be compared and the lowest
lethal dose recorded.
RTECS citations for data reported in journals from eastern European
countries were not used unless no other data were available. This was done
due to a concern that the methodology and reporting of data used in these
countries may be inadequate. If there were no data for a chemical or if
the only RTECS citations were from eastern European journals, the Hazardous
Substance Database (HSDB) (an on-line database produced by the Specialized
Information Services of the National Library of Medicine, Bethesda,
Maryland) was reviewed for the potency measures noted above.
There are three main issues that need to be considered to effectively
use this approach: extrapolation issues, missing data, and data errors.
Extrapolation issues arise from the fact that, for the purpose of
comparisons, inhalation doses were converted to mg/kg/day using standard
species weights and breathing volumes and all doses in mg/kg/day were
assumed to be equivalent among species. Comparison of effects across
species does not reflect the potential for differences in sensitivities to
particular chemicals among species. While the equivalent dose among
species is a generally accepted principle, there can be exceptions for
particular chemicals.
As for missing data, chemicals for which no acute lethality research
studies have been reported in RTECS or HSDB will rank lower relative to
chemicals with acute toxicity data reported in RTECS, regardless of the
true measure of acute lethality. Related to this, only the limited data
evaluation of the scientific literature would likely have yielded more
information.
4.4 REPRODUCTIVE AND DEVELOPMENTAL TOXICITY
This component for reproductive and developmental effects uses the
lowest TDLO or TCLQ value from the RTECS field for reproductive effects.
The TDLO used here is the lowest dose reported to produce reproductive or
developmental effects in humans or animals. Similarly, the TCLQ is the
lowest concentration when the substance is in air. This factor includes
paternal and maternal effects, effects on fertility, embryo and fetal
effects, developmental abnormalities, tumorigenic effects related to the
reproductive system (e.g., testicular, prostate, ovarian, and uterine
29
-------
tumors; transplacental tumorigenesis; and other reproductive system
tumors), and effects on newborns. Data extracted for the SCRS consisted of
oral, dermal, and inhalation doses, the routes of exposure responsible for
human exposure to toxic air pollutants.
As with the acute lethality component, data were extracted from
RTECS, unless the only studies cited were from eastern European journals.
In such cases, HSDB was consulted and relevant measures extracted. HSDB
was also consulted if there was no information in RTECS for a particular
chemical.
Data errors can be another problem with this process. On occasion,
values given in the databases are reported incorrectly and do not reflect
the lethal dose reported in the journal cited. Since the original journal
articles were not reviewed in this initial data gathering effort, such data
errors would not be identified.
All dose levels extracted from the databases were converted to units
of mg/kg/day. The lower the potency measure (TD.Q) of a chemical, the
lower that chemical would rank within the reproductive and developmental
toxicity component, relative to the other chemicals under consideration.
Like the acute lethality component of SCRS, the issues of extrapolation,
missing data, and data errors are also associated with this component for
reproductive and developmental effects. In addition, there could be
concern over lack of consistency in the severity of the toxic effect. The
different reproductive and developmental effects cited in RTECS are not
equally severe, and therefore, for example, the dose level of a chemical
resulting in decreased fertility might be compared with a dose level of a
chemical resulting in fetal death.
4.5 OTHER TOXICITY
The component in SCRS known as "other toxicity" is included to
address health effects other than carcinogenicity that result from long-
term exposure to toxic air pollutants. Endpoints of this component include
effects on the following: respiratory system, digestive system, nervous
system (including peripheral nerves and sense organs such as nose and
eyes), immunological system, biochemical balances, blood characteristics,
skin, and behavior.
30
-------
The other toxicity component consists of TD.Q and TC.Q data from the
RTECS field for acute toxicity. Data on LD5Q LC5Q, LDLQ were used if the
exposure time was greater than 24 hours. (This was the case for only seven
chemicals.) The TD,Q used here is the lowest dose reported to produce
effects other than cancer, reproductive or developmental effects in humans
or animals. Similarly, the TCLQ is the lowest concentration when the
substance is in air.
The routes of exposure considered were limited to oral, inhalation,
and dermal, since human exposure to toxic air pollutants is limited to
these routes. All dose values were converted to the units of mg/kg/day.
The lowest potency measure was recorded for each chemical and chemicals
were ranked according to these values. The lower the dose, the higher the
chemical ranked.
Data on other toxicity were assembled from RTECS. If no data were
found or if the only data cited in RTECS were from eastern European
sources, then HSDB was reviewed and relevant data extracted.
Like the acute lethality and reproductive/developmental components,
the other toxicity component is also potentially affected by extrapolation
issues, missing data, and data errors. In addition, since the effects for
the other toxicity component are not as clearly defined as in the case of
lethality, there is the concern over the consistency in the severity of the
toxic effect. For instance, it is possible for a chemical with the effect
of eye irritation to rank higher than a chemical that causes liver damage,
if the dose for eye irritation is lower. With this other toxicity
component, there is also an issue of ambiguities surrounding the
definitions of the toxic effects.
31
-------
5.0 EXPOSURE SCORES
The exposure score is not to be interpreted as an exposure
assessment. A scientifically defensible exposure assessment requires more
information, higher quality data, and a more rigorous methodology than that
used in the SCRS to calculate the exposure score. An exposure score is a
product of two components: ground-level pollutant concentration estimates
and limited population data. The ground level pollutant concentration
estimates are calculated by inputting emissions data into simplified air
dispersion algorithms that contain may assumed parameters and fixed
constants. The population data incorporated is either national average
population density, county population density, or an assumed maximally
exposed individual. Specific items included in the data set, incorporated
into the air dispersion algorithm constants, or entered as a variable are
listed in Tables 7 and 8. This section describes the methodology by which
the exposure scores are developed and explains the different techniques
used for area, fugitive, and point sources.
Table 8. SCRS EMISSIONS DATA
Emissions estimates:
Atmospheric dispersion parameters:
Emission source data:
rate
type
(kg/yr)
point', area, fugitive)
release height
release temperature
release flow rate
release vent diameter
pollutant
plant identification
source category
state/county code
32
-------
Table 9. SCRS POPULATION DATA
County code
County population
County population density
Simplified dispersion algorithms were developed for the SCRS to
calculate pollutant emission rates and dispersion characteristics.
Algorithms also were developed to calculate short- and long-term maximum
exposure scores, long-term aggregate exposure scores, and the sum of short-
term maximum exposure scores. This part of the SCRS produces a source
category data file containing four exposure scores for each pollutant in
the source category.
5.1 METHODOLOGY
Simplified dispersion algorithms were developed for the SCRS to
estimate the potential ground-level concentration from fugitive and point
sources for individual plants. Source categories for which plant lists
could not be generated were handled as modeled area sources with a separate
simplified dispersion algorithm. These dispersion algorithms provide a
concentration factor that when multiplied by the pollutant emission rate
yields an approximate pollutant concentration.
The SCRS calculates four exposure scores for each emission type:
short-term maximum, short-term aggregate, long-term maximum, and long-term
aggregate. The short-term maximum exposure is the highest concentration
around any plant in the source category to which an individual might be
exposed. The short-term aggregate Is derived by summing the highest short-
term maximums around the plants in the category. The long-term maximum
exposure is computed based on the most exposed individual. The long-term
aggregate exposure is computed using an estimate of the population within a
radius of 50-km of the emission source. This population estimate is
calculated for point sources using the county population density; for
modeled area sources the national average population density is used. The
50-km distance is the upper limit accepted for dispersion models and
consistent with the Guidelines on Air Quality Models (Revised)(U.S. EPA,
1986).
33
-------
Several EPA-recommended air dispersion models were used to develop
the dispersion algorithms. The Industrial Source Complex Short-Term
(ISCST) model was used to develop an algorithm for averaging short-term
maximum concentrations from fugitive sources (U.S. EPA, 1987c). The ISCST
model was executed by screening meteorological data to estimate 1-hour
average concentrations. An algorithm was developed from the ISCST results
for estimating a short-term maximum ground-level concentration. The
Industrial Source Complex Long-Term (ISCLT) model was used to develop the
long-term dispersion algorithms. Annual average concentrations from point
and fugitive sources were determined using the ISCLT model and
representative STability ARray (STAR) data for three areas in the
United States. The use of multiple meteorological data sets provided a
range of concentrations. Because in all cases the concentration results
from the multiple sets agreed within 50 percent of the average
concentration, the average concentration was used in algorithm development.
The methodology used to determine the short-term maximum
concentration from point sources was developed from the procedure for
estimating maximum ground-level concentrations recommended in (U.S. EPA,
1987c). The procedure was modified to account for momentum plume rise from
nonbuoyant plumes. The inconsistency of using different dispersion models
for the short-term/fugitive source and the short-term/point source was
examined. The comparison considered the concentrations predicted using the
above methodology (used for the short-term point source algorithm
development) and the PoinT PLUme (PTPLU) dispersion model (used for short-
term fugitive source algorithm development). The results were that the
1-hour average concentration predicted using the methodology in (U.S. EPA,
1987c) compares favorably with the maximum ground-level concentration
predicted using the EPA PTPLU dispersion model. For the regional area
source concentration estimates, the Hanna-Gifford area source algorithm was
used (Hanna, Gifford, 1973). A wind speed of 5 m/s was assumed as a
representative national average (U. S. Department of Commerce, 1979) for
use in the Hanna-Gifford algorithm.
A description of the dispersion algorithm development and the related
exposure score follows. The long-term aggregate, the long-term maximum,
and the short-term maximum exposure score algorithms are presented for each
emission type.
34
-------
5.2 MODELED AREA SOURCES
The SCRS algorithms for modeled area sources are based on a
simplified version of a national exposure assessment performed using the
Systems Application Human Exposure And Risk (SHEAR) model (Anderson,
Lundberg, 1983). This version was modified on the assumption that
emissions are proportional to the national population distribution.
The SHEAR model was used to calculate both the net long-term exposure
and the maximum long-term exposure score on a nation-wide basis. The
aggregate long-term county exposures were summed to determine the net long-
term exposure scores for the nation. In addition, the highest long-term
maximum exposure score (county with the highest long-term maximum value)
was determined. The results of this analysis were then used to develop the
modeled area source exposure algorithms.
The exposure values are calculated on an emission unit basis, and
this can be scaled if the national emissions estimate is known. Some
examples of source categories that can be generalized with this modeling
approach are wastewater treatment facilities and dry cleaning emission
sources.
The following are the long-term aggregate, the long-term maximum, and
the short-term maximum exposure score algorithms developed for the modeled
area source types:
Long-term 3
aggregate - (736,194,000 /yg/m -person, when assuming 1 kj/yr/person)/
exposure (226,546,000 people in the U.S.) * (Emission rate (kg/yr))
score
Long-term 3
maximum - (440 fjg/m , when assuming 1 kg/yr/person)/
exposure (226,546,000 people in U. S.) * (Emission rate (kg/yr))
score
Short-term » Long-term maximum exposure
maximum * (31 ratio of exposure factor)
exposure
score
The -ly variable used for modeled area sources is the emissions
estimates The 736,194,000 //g/m per person constant is the nationwide
lifetime exposure estimate assuming i kg of emissions per year per person.
The 226 C46,000 constant is the 1980 census estimate of the U.S.
35
-------
population. The emission-rate variable is generated from the emissions
data file. The 440 fjg/m constant is the maximum modeled concentration
assuming 1 kg of annual emissions per person. The 30 factor is used for
the SCRS scoring technique to estimate the relationship between long-term
and short-term exposures. The ratio of the long-term concentration
estimated by ISCLT and the short-term concentration estimated by ISCST is
30 for area emissions sources. Thus, short-term concentrations are based
on long-term data. The short-term maximum and short-term aggregate
exposure scores are identified for modeled area sources.
5.3 FUGITIVE SOURCES
The fugitive emission source algorithms are based on the results of
accepted dispersion models. Two widely used dispersion models, the ISCST
and ISCLT models, were used for the short-term and long-term algorithm
development, respectively. The short-term results were calculated for a
1-hour averaging period.
Three constants (C4, C5, and C6) were developed for the algorithm
that was used to predict the maximum estimated concentrations for fugitive
sources. A set of worst-case hourly meteorological conditions was used to
determine the maximum short-term concentration. Annual joint frequency
distribution meteorological data-for three geographical locations (Atlanta,
Georgia; St. Louis, Missouri; and Hartford, Connecticut) were used to
determine the long-term average concentration. These locations were
selected as typical representatives of U.S. geographic regions.
To develop model inputs, fugitive source emission parameters were
assumed to be fairly consistent. Therefore, the models could be run with
the following representative set of emission parameters:
Release height at 2 m,
Release temperature at ambient levels (131'C), and
?
Release area of 3000 m .
An emission rate of 1 kg/yr was used in the analysis so that the results
could be proportioned to other emission rates.
The simplified algorithm for estimating the long-term average ground-
level concentration is based on ISCLT results. The ISCLT model was run
with the above-mentioned emission parameters and meteorological data from
the three STAR data sets. The dispersion results were used to determine
36
-------
the average concentration within 50 km of the emission source. Therefore,
the ISCLT results were used to calculate an area weighted-average
concentration.
The area weighted-average concentration was calculated by determining
the average concentration for a given radial distance from a source. The
distances are consistent with the SHEAR methodology with 16 receptors per
ring and 10 rings extending out to 50 km. An area weighted-average
concentration was calculated for each of the three locations (i.e.,
Atlanta, St. Louis, and Hartford). The concentrations for each of the
16 receptors located on a ring were averaged, and the average concentration
for each of the 10 rings was weighted by the area assumed to be represented
by that concentration. The representative area included the boundary of
adjacent rings. The area weighted concentration at the three areas were
compared and all three values were within 10 percent of the average.
Therefore, the average concentration was used to develop the exposure
algorithm.
The long-term aggregate exposure algorithm was developed from the
dispersion results and includes the estimated population (based on county
population density) within 50 km of the emission source. The population
parameter is estimated from the county population density for each emission
source's county of residence and the area within a 50-km radius of the
facility.
The only variables used to develop long-term aggregate exposure
scores for fugitive sources are estimated annual emission rate and county
population density. The long-term aggregate exposure algorithm developed
for fugitive emission sources is as follows:
Long-term C4 * (Emission Rate (kg/yr)) ?
aggregate = * (County Population Density/km)
exposure * (7853/km )
score
where:
C4 = Average ISCLT results for the average concentration data at
annual averaging periods
- 3.60 x 10"7 //c'.-n3 km2;
37
-------
and
7853 km = the area within a 50-km radius of the facility.
The simplified algorithm for estimating the long-term maximum ground-
level concentration is also based on ISCLT results. The maximum ground-
level concentration was assumed to occur at a 200-m downwind distance. The
ISCLT model was run with the above-mentioned emission parameters using
three STAR meteorological data sets. An average annual concentration value
was assumed to be representative of the results from the three
meteorological data set locations. Therefore, the dispersion factor was
developed based on average annual concentration values.
The long-term maximum exposure score algorithm was developed from the
dispersion results. The exposure estimate is based on the most exposed
(i.e., maximum) individual. Therefore, a population variable is not
required in the long-term maximum exposure score algorithm.
The only variable used to develop long-term maximum exposure scores
for fugitive sources is the estimated annual emission rate. The long-term
maximum exposure algorithm developed for fugitive emission sources is shown
below:
Long-term maximum exposure score - C5 * (Emission rate (kg/yr))
where:
C5 = ISCLT results at annual hour averaging periods
= 1.23 x 10"3 //g/m3/kg
The algorithm for estimating a short-term maximum ground-level
concentration is based on ISCST results. The ISCST model was run with the
following dispersion parameters, which are expected to provide the maximum
estimated concentration for short-term fugitive sources:
F stability class,
Wind speed at 2.5 m/s, and
Downwind distance of 200 m.
A representative set of emission parameters developed for fugitive sources
for release height, release temperature, and release area was assumed.
38
-------
The only variable used to develop short-term maximum exposure scores
for fugitive sources is the estimated annual emissions rate. The short-
term maximum exposure score algorithm developed for fugitive emission
sources is given below:
Short-term
maximum = C6 * (Emission rate in kg/year)
exposure
score
where:
C6 = ISCST results at 1-hour averaging period
= 0.0372 ^g/m3/kg
5.4 POINT SOURCES
The point source exposure score algorithms were developed using two
approaches. The long-term point source exposure score algorithms were
developed using the method of least squares (i.e., curve-fitting) for
correlating ISCLT results with the effective plume height The short-term
point source exposure score algorithms were developed using basic modeling
procedures from Volume 10 U.S. EPA, 1977.
Long-term algorithms were developed for long-term maximum exposure and
long-term aggregate exposure scores. The algorithms were developed from
the results of a representative data set run through ISCLT for the same
three locations selected for the fugitive source algorithm development
(i.e., Atlanta, St. Louis, and Hartford). One data set was selected as
representative of the distribution of stack parameter data from the NEDS
database. The data set contained 65 sampled data points that represented
equal quartiles across the range of NEDS stack heights. The stack height
range was used because stack height is assumed to be the most critical
stack parameter in dispersion analysis.
The input matrix to ISCLT included the stack parameters from the
representative NEDS data set and meteorological parameters (i.e., stability
class and wind speed) for the three representative locations. The emission
parameters obtained from the NEDS database were discharge temperature,
discharge velocity, vent diameter, and discharge height. The long-term
maximum exposure score algorithm was developed to correlate the ISCLT
39
-------
maximum concentration result with the effective plume height. The long-
term aggregate exposure score algorithm was developed to correlate the area
weighted-average concentration within 50 km of each source in the data set
with the effective plume height. The effective plume height for both the
long-term maximum and aggregate exposure score is estimated for stability
Class D (i.e., stable conditions) and a wind speed of 5.0 m/s.
The method of least squares provides a fit of the independent
dispersion constants with ISCLT-generated concentration results. The
independent dispersion constants are the emission parameters for
calculating the effective plume height. The constants used were discharge
height, discharge temperature, discharge velocity, and vent diameter. A
Gaussian distribution was assumed for the ISCLT results. The effective
plume height equation is based on the methodology described in U.S. EPA,
1977.
The only variables used to develop long-term aggregate exposure
scores for point (e.g., stacks) emission sources are estimated annual
emissions rate and county population density. The long-term aggregate
exposure score algorithm developed for point emission sources is shown
below:
Long-term- (Al * eB1*lon9-term effective plume height**2j
aggregate = * (Emission rate (kg/yr)) * K 2 .
exposure * (County population density/km ) * (7853 km )
score
where:
A1,B1 « correlations developed from curve-fitting 65 data
points with ISCLT results;
Al = 0.0088
Bl - -0.000197
40
-------
K = 3.171 x 10~5
= the conversion factor from yg/m (assuming a 1 g/sec emission
rate) to ;/g/m (assuming a 1 kg/yr emission rate)
and
2
7853 km = the area within a 50-km radius of the facility.
The only variable used to develop long-term maximum exposure scores for
point emission sources is estimated annual emission rate. The long-term
maximum exposure score algorithm developed for point emission sources is
given below:
maximum"" - (A2 * eB2*lon9-term effective plume height*2}
exposure * (Emission rate (kg/yr)) * K
score
where:
A2,B2 = correlations developed from curve-fitting 65 data points with
ISCLT results
A2 = 20.502
B2 = -0.005
The representative data set (i.e., 65 data points from NEDS) was
modeled for stable, unstable, and neutral stability classes. The maximum
concentration did not occur at stable conditions for any of the 65 data
points. Therefore, the stable class was not considered further in the
dispersion algorithm development. The Point Plume dispersion model was
used to eliminate the need for modeling the stable stability class.
The short-term maximum exposure score is the maximum value estimated
for the unstable stability class with either a looping plume or a coning
plume depending on which scenario results in the maximum concentration.
The short-term maximum exposure score algorithm developed for point
emission sources is shown below:
41
-------
Short-term maximum exposure score (Unstable, Looping Plume) =
(1.0316 * Emission rate (kg/yr))/Critical wind speed (m/s)
* Effective stack height**(-1.5998)
Short-term maximum exposure score (Unstable, Coning Plume) =
(4.2978 * Emission rate (kg/yr))/Critical wind speed (m/s)
* Effective stack height**(-1.9921)
5.5 SOURCE CATEGORY EXPOSURE DATA FILE
The result of the exposure score calculation portion of the Source
Category Ranking System is the Exposure Score File. This source Category
Exposure Score File is one of the primary inputs to the final portion of
the system that calculates a score for each source category and ranks them
accordingly. The Source Category Exposure Score File contains four
exposure values and an estimated emission rate for each pollutant emitted
from each source category. The organization of this file is shown in
Table 10.
42
-------
H-
O
UJ
GO
S
X
UJ
ac
S
UJ
^^
P^^
<
UJ
QC
O
GO
o
UJ
CO
H-
0)
1.
c ^ ^^-
o
I/)
(/I
i
UJ
01
CT
d)
S_
o>
o>
g *~
0)
1
Ol
c
0
_J
E
g
X
E
E »~
4_>
1
Ol
o
^
t
0)
Ol
O)
01
g *"
0)
1
4-1
o
GO
E
3
X
I
g ^
1
i-
o
i- GO
O
Ol i ' i
4-1
<_> c c e
19 19 19
Qf 4^) 4J 4^)
U 3 3 « « 3
V. r- . i
3 f t f
O O O O
GO a. a. at.
CM
J^
O
s»
4J
U
(U
U
3
O
0)
4-1
2
C '^>
0
21
i
Ol
0
^^
aT
0>
£
Ol
Ol
g ~ "*
,
4->
L.
O
to
E*
I
j^
I
g__^ ^
i
L.
O
GO
I i i
e e e
mm m
3 3 » « 3
pm ^^ ^^
O 0 O
a-a. a.
4)
*
w
_ .
0
'v>
1/1
E
UJ
5
Ol
c
Ol
Ol
g
9)
,
Ol
c
o
E
X
S
g
Qt
i
Ol
0
^J
aT
Ol
a>
Ol
Ol
g
0)
1
4-1
O
«jfi
s
i
X
s
g
1
z w
0
w GO
O
ff -^
u e e
m m
O 4-*4-»
U 3 3
S. tr-
3 r-r-
o o o
GO QuO.
__^
^^
*
*
»
1
c
m
4-1
3
r^
F""*
£
43
-------
6.0 SOURCE CATEGORY SCORING AND RANKING
Health effects and exposure scores are combined in the SCRS to
produce the source category scores used to rank the listed source
categories. The flow chart in Figure 3 shows how this final integration
takes place.
Two main data files are brought into the SCRS integration function:
1) Health Effects Score and 2) Source Category Exposure Score. A wide
variety of health effects score for each pollutant is included in the
Health Effects Score File as described in Section 4.0. The Source Category
Exposure Score File includes long-term and short-term exposure score for
each source category-pollutant combination as described in Section 5.0.
Both health effects and exposure scores are scaled in the SCRS so
that the maximum value is one. Scaling is necessary because of the large
differences in ranges of the various data sets and allow the user to vary
the weighting assigned to different exposure types and health effects. For
instance, the long-term aggregate exposure scores are calculated with
population as one of the factors and, therefore, have a potential range of
more than four orders of magnitude greater than the other exposure scores
that have a much smaller range since they do not include population. The
scaling is accomplished by dividing each value in a given data set by the
largest value in the data set. This assures that the largest value in the
scaled data set is 1.0 (i.e., the largest value in the data set divided by
itself) and that the other values are between 0.0 and 1.0. Although both
logarithmic and linear scaling options are included in the system, linear
scaling is most often used.
The user may choose weighting factors to use with health effects data
and with the short- and long-term exposures. This allows for rankings
based on only one type of health effect or on a specified set of health
44
-------
1
I
o
?
"S 9
O o
0)
0)
I
O)
c
re
DC
a
c
CO
O)
o
o
CO
CO
DC
O
CO
CO
1
O)
45
-------
effects. Rankings can also be based on long-term exposures, short-term
exposures, or a combination of both.
The health effects score and exposure scores are multiplied together
using the specified weighting factors to get the scores for a given
pollutant in a source category. The total score for the source category is
accumulated by adding up all the individual pollutant scores for that
category. The source category score is then used to rank the source
categories.
The scores provide a relative measure for ranking the source
categories. However, the scores should not be interpreted as a
quantitative measure of any particular health effect.
46
-------
7.0 REFERENCES
Anderson, G. E., and G. W. Lundyberg. (Systems Applications, Inc.).
User's Manual for SHEAR. Prepared for the U. S. EPA. Durham, North
Carolina. SYSAPP-83/124. May 1983.
Hanna, S. R. and F. A. Gifford. Modeling Urban Air Pollution. Atmospheric
Environment. 7:803-817. 1973.
U. S. Department of Commerce, National Oceanic and Atmospheric
Administration, Climatic Atlas of the United States. Asheville,
North Carolina. 80 pp. 1979.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. Guidelines for Air Quality Maintenance Planning and
Analysis Volume 10 (Revised): Procedures for Evaluating Air Quality
Impact of New Stationary Sources. Research Triangle Park, North
Carolina. EPA 450/4-77-001. 1977.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. Guidelines on Air Quality Models (Revised). Research
Triangle Park, North Carolina. EPA 450/2-78-027R. 1986.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. Summary of Continuous Emissions Data from Seven Source
Categories Producing or Using Hazardous Organic Compounds. Research
Triangle Park, North Carolina. EPA 450/3-87-020. 1987a.
U. S. Environmental Protection Agency, Office of Research and Development.
Criteria Pollutant Emission Factors for the 1985 NAPAP Emissions
Inventory. Research Triangle Park, North Carolina.
EPA 600/7-87-015. 1987b.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. Industrial Source Complex (ISC) Dispersion Model User's
Guide - Second Edition (Revised) Volume 1. Research Triangle Park,
North Carolina. EPA 450/4-88-002a. 1987c.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards, National Air Data Branch. National Emissions Data System.
Research Triangle Park, North Carolina. July 1988.
U. S. Environmental Protection Agency, Office of Air Quality Planning and
Standards, Air Quality Management Division. Volatile Organic
Compound (VOC)/Particulate Matter (PM) Speciation Data System
Documentation and User's Guide. Version l-32a. Contract No. 68-02-
4286. Research Triangle Park, North Carolina. September 1990.
47
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverie before completing)
1. REPORT NO. 2. 3. RECK
EPA-g.
12. SPONSORING AGENCY NAME AND ADDRESS 13. TYP
Director, office of Air Quality Planning arid Standards Fil
U. S. Environmental Protection Agency 14'SPO
Research Triangle Park, North Carolina 27711 EP/
15. SUPPLEMENTARY NOTES
>IENT'S ACCESSION NO
RT DATE
tember 1993
ORMING ORGANIZATION CODE
ORMING ORGANIZATION REPORT NO
GRAM ELEMENT NO.
TRACT/GRANT NO.
-Dl-0117
E OF REPORT AND PERIOD COVERED
ial
MSORING AGENCY CODE
A/200/04
16. ABSTRACT
In developing the source category schedule for emission standards under section 112
of the Clean Air Act, the EPA developed the source category ranking system (SCRS)
to help prioritize source categories. This document explains the methodolooy used
by the SCRS.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS b. IDENTIFIERS/OPEN ENDS
-
18. DISTRIBUTION STATEMENT 19. SECURITY CLASS (This i
Unclassified
20. SECURITY CLASS (This (
Unclassified
D TERMS c. COSATI Field/Group
jage/ 22. PRICE
EPA Form 2220.1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
------- |