EPA’s Risk-Screening Environmental Indicators (RSEI) Methodology RSEI Version 2.3.2 July 2013


a

^Rlfl^	Environmental Protection

Lsl	Agency

Office of Pollution Prevention
and Toxics

Washington, DC 20460

July 2013

EPA's Risk-Screening
Environmental Indicators
(RSEI) Methodology

RSEI Version 2.3.2

-------
Acknowledgments

Acknowledgments

Many people have contributed to the RSEI project over the years. We would especially like to
acknowledge the contributions of Nicolaas Bouwes and Steven Hassur who were the originators
of the model and instrumental to its ongoing development and improvement.

Current Members of the RSEI team are:

Lynne Blake-Hedges
Richard Engler
Cody Rice

Development support provided by Abt Associates Inc., 55 Wheeler Street, Cambridge,
Massachusetts.

Version 2.3.2	j	1988-2011 TR1 data

July 2013

-------
Table of Contents

Table of Contents

Executive Summary	1

ES-1. Introduction	1

ES-2. General Description of the RSEI Model	1

ES-2.1 Geographic Basis of the Model	2

ES-2.2 RSEI Results	6

ES-2.3 Adjusting RSEI Results for Changes in TRI Reporting	8

ES-2.4 Model Implementation	8

ES-2.5 How the RSEI Chronic Human Health Toxicity Weightings Differ from

EPCRA Section 313 Statutory Criteria	9

ES-3. Important Caveats Regarding the RSEI Model	10

ES-4. New Features in Version 2.3.2	12

1.	Introduction	1

1.1	Background	1

1.2	Model Implementation	2

1.3	Organization of this Document	3

2.	General Description of the RSEI Model	5

2.1	General Description	5

2.2	Summary of the Strengths and Limitations of the RSEI Model	8

2.2.1	Strengths	8

2.2.2	Limitations	9

3.	TRI Emissions Data	10

4.	Methods for Calculating Toxicity Weights	12

4.1	Toxicity Weighting Scheme for Non-carcinogens and Carcinogens	13

4.1.1	Qualitative Data	13

4.1.2	Quantitative Data	14

4.1.3	Algorithm for Calculating Toxicity Weight	16

4.2	Selecting the Final Chronic Human Health Toxicity Weight for a Chemical	17

4.3	Chemical Groups	18

4.4	Sources of Data	19

4.5	How Indicator Toxicity Weightings Differ from EPCRA Section 313 Criteria... 21

5.	Exposure and Population Modeling	24

5.1	Geographic Basis of the RSEI Model	25

5.1.1	The Model Grid Cell System	25

5.1.2	Locating Facilities on the Grid	27

5.1.3	Locating People on the Grid	28

5.2	Pathway-specific Methods to Evaluate Chronic Human Exposure Potential	30

5.3	Modeling Air Releases	31

5.3.1 Stack Air Emissions: Method	31

Version 2.3.2	j	1988-2011 TRI data

July 2013

-------
Table of Contents

5.3.2	Fugitive Air Releases: Method	34

5.3.3	Calculating Surrogate Dose for Air Releases	35

5.3.4	Estimating Population Size for Air Releases	36

5.3.5	Calculating an Indicator Element for Air Releases	36

5.3.6	Stack and Fugitive Air Releases: Data	38

5.4	Modeling Surface Water Releases	43

5.4.1	Surface Water Releases: Methods	43

5.4.2	Calculating the Indicator Element for Surface Water	48

5.4.3	Surface Water Releases: Data	50

5.5	Modeling Transfers to POTWs	55

5.5.1	Transfers to POTWs: Method	55

5.5.2	Transfers to POTWs: Data	59

5.6	Modeling Other Off-site Transfers	60

5.6.1	Off-site Transfers: Method	60

5.6.2	Estimating Population for Off-Site Transfers	60

5.6.3	Off-site Transfers: Data	61

5.7	Modeling On-site Land Releases	62

6.	Calculating Results	63

6.1	Combining Indicator Elements	65

6.2	Accounting for Changes in TRI Reporting	65

7.	Current Implementation of the RSEI Method	67

7.1	RSEI Model	67

7.2	Conclusion	68

8.	References	69

Version 2.3.2
July 2013

11

1988-2011 TRI data

-------
Executive Summary

Executive Summary

ES-1. Introduction

EPA's Risk-Screening Environmental Indicators (RSEI) is a screening-level tool that assesses
the potential impact of industrial chemical releases from pounds-based, hazard-based, and risk-
related perspectives. RSEI uses risk concepts to quickly and easily screen large amounts of
Toxics Release Inventory (TRI) data, saving time and resources. RSEI is particularly useful for
examining trends to measure change, ranking and prioritizing chemicals and industry sectors for
strategic planning, conducting risk-related targeting, and supporting community-based projects.

Using estimates of pounds of chemical releases to investigate potential health and environmental
impacts is limited by the assumptions that all chemicals are equally toxic and all people are
equally exposed. Formal risk assessments are more accurate, but are complicated and time
consuming to prepare, requiring detailed data that is not always available, and the results are
often limited in scope and geographic area. The RSEI approach augments estimates of pounds
released with toxicity and exposure considerations, but does not address all of the potential
factors that a full risk assessment would include.

RSEI considers the following information: the amount of chemical released, the toxicity of the
chemical, its fate and transport through the environment, the route and extent of human
exposure, and the number of people affected. This information is used to create numerical
values that can be added and compared in limitless ways to assess the relative risk of chemicals,
facilities, regions, industries, or many other factors. The values are for comparative purposes
and only meaningful when compared to other values produced by RSEI. It should be
emphasized that the result is not a detailed or quantitative risk assessment, but offers a screening-
level, risk-related perspective for relative comparisons of chemical releases.

The RSEI approach is very flexible and can be implemented in various ways. The use of the
model is not limited to any specific set of chemicals; in principle, the adaptable method can
model any chemical if toxicity characteristics, physicochemical properties, release levels, and
release location are known or can be estimated.

As an indication of improvements in environmental quality over time, RSEI provides a valuable
tool to measure general trends based upon relative risk-related impacts of TRI chemicals.
Although RSEI results do not capture all environmental releases of concern, they generally relate
changes in releases to relative changes in chronic human health impacts from a large number of
toxic chemicals of concern to the Agency. Importantly, RSEI provides an ability to analyze the
relative contribution of chemicals and industrial sectors to human health impacts, and RSEI
results serve as an analytical basis for setting priorities for further risk analysis, pollution
prevention, regulatory initiatives, enforcement targeting, and chemical testing requirements.

ES-2. General Description of the RSEI Model

The RSEI model calculates values that reflect the risk-related impacts on chronic human health
of modeled TRI chemical releases and transfers. These values do not provide absolute measures

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

of risk and can only be interpreted as relative measures to be compared with other such values
(reflecting the direction and the general magnitude of changes at different points in time when
analyzing trends, or identifying the relative contribution of variables in a comparative analysis).

The model uses the reported quantities of TRI releases and transfers of chemicals to estimate the
risk-related impacts associated with each type of air and water release or transfer by every TRI
facility. The risk-related impacts potentially posed by a chemical are a function of chemical
toxicity, the fate and transport of the chemical in the environment after it is released, the pathway
of human exposure, and the number of people exposed.1

The RSEI model starts with TRI releases. For each exposure pathway associated with each
chemical release, the model generates an "Indicator Element". For instance, a release of the
chemical benzene to air via a stack from the "ABC" Facility in 1999 is an "Indicator Element".
Each Indicator Element is associated with a set of results, including pounds-based, hazard-based,
and risk-related results, also called scores. The risk-related score is a unitless value proportional
to the potential risk-related impact of each element.

Each Indicator Element can be combined and compared with other Indicator Elements. There
are countless ways that Indicator Elements can be summed together to assess chronic human
health impacts. For example, all of the RSEI Indicator Elements can be aggregated for each year
to allow an assessment of trends in estimated impacts, or the Elements can be grouped to allow
users to compare results for facilities, regions, chemicals, and any combinations of these and
other variables. RSEI does not perform a detailed or quantitative risk assessment, but offers a
screening-level, risk-related perspective for relative comparisons of chemical releases. The
model does not estimate actual risk to individuals. RSEI results are only meaningful when
compared to other results produced by RSEI.

The current version of the model calculates risk-related results for the air and surface water
pathways only. For other pathways, and in instances where information needed to model a
release is not available, only pounds-based and hazard-based perspectives are available. In cases
where toxicity weights are not available, only pounds-based results can be viewed.

ES-2.1 Geographic Basis of the Model

The model relies on the ability to locate facilities and people geographically, and to attribute
characteristics of the physical environment, such as meteorology, to areas surrounding the
facilities once they are located. To locate the facilities and the attribute data to those facilities,
the model describes the U.S. and its territories2 as an 810m by 810m grid system. For each cell
in the grid, a location "address" in terms of (x,y) coordinates is assigned based on latitude and
longitude (1 at/long).

1 The method is focused on general populations; individuals, particularly highly exposed individuals, are not the
focus of the model.

2 The model also includes Puerto Rico, the U.S. Virgin Islands, Guam, American Samoa, and the Northern Mariana
Islands. 1990 U.S. Census data were provided by GeoLytics, Inc., East Brunswick, NJ.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

In order to estimate potential exposure, TRI facilities and the U.S. population must be
geographically located on the model grid. TRI facilities are located using the facilities' lat/long
coordinates. To locate population, the model uses decennial U.S. Census data for 1990, 2000,
and 2010 at the block level. These data3 are used to create detailed age-sex-defined population
groups for each of the census blocks in the U.S. for 1990, 2000, and 2010. The following
population groups are used in the model:4

• Males Aged 0 through 9 years

• Males Aged 10 through 17 years

• Males Aged 18 through 44 years

• Males Aged 45 through 64 years

• Males Aged 65 Years and Older

• Females Aged 0 through 9 years

• Females Aged 10 through 17 years

• Females Aged 18 through 44 years

• Females Aged 45 through 64 years

• Females Aged 65 Years and Older

Because the Census block boundaries change between decennial Census years, each set of
Census block level data is first transposed onto the model grid, which is unchanging, using an
area-weighted method. Once populations for 1990, 2000, and 2010 are placed on the grid
system, the model uses a linear interpolation for each grid cell to create annual estimates of the
population sizes for each year between 1990 and 2000, and again between 2000 and 2010. The
straight-line plot between 1990 and 2000 is extrapolated backward to estimate population for
1988-89.

Once facilities and people are located on the model's grid system, three main components are
used to compute risk-related impacts in the model. These components are:

• The quantity of chemicals released or transferred,

• Adjustments for chronic human health toxicity, and

• Adjustments for exposure potential and population size.

These components and the method used to combine them are described in the following sections.

3 For 1990, not all of the variables were available at the block level for the Continental U.S, Alaska and Hawaii.
For those variables that were only available at the block group level, block group ratios were calculated and applied
to the data at the block level. For 2000 and 2010, all of the required data were available at the block level. For the
U.S. Virgin Islands and the territories, data from larger geographic units (block groups or county-equivalents) were
used. For Puerto Rico, block group data were used for 1990 and block-level data for 2000 and 2010.

4 Not all of the population groups listed are used in viewing results. Model results can only be viewed for the
following groups: Children Under 10, Children 10 through 17, Males 18 through 44, Females 18 through 44, and
Adults 65 years and Older.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

Chemical Releases and Transfers. The model uses information on facilities' chemical releases
and transfers from these facilities to off-site facilities (such as sewage treatment plants and
incinerators) to model risk-related impacts. These releases are reported by facilities to the TRI,
as mandated by the Emergency Planning and Community Right-to-Know Act. As of the 2011
reporting year, there are 620 TRI chemicals and chemical categories listed. Users can view
pounds of chemicals released per year (pounds-based results) for any combination of variables
included in the model.

Adjustments for Chronic Human Health Toxicity. The model is based on current EPA
methodologies for assessing toxicity. The method EPA has chosen for assigning toxicity weights
to chemicals is clear and reproducible, based upon easily accessible and publicly available
information, and uses expert EPA-wide judgments to the greatest extent possible. RSEI reflects
the toxicities of chemicals relative to one another using a continuous system of numerical
weights. Toxicity weights for chemicals increase as the toxicological potential to cause chronic
human health effects increases. Toxicity-adjusted releases are called "hazard-based results" and
provide an alternative perspective to pounds-based or full risk-related results, and are especially
valuable when necessary data for risk-related modeling are not available.

Values developed by EPA experts are used to differentiate the degrees and types of toxicity of
chemicals, and rank them in a consistent manner. Values called Oral Slope Factors and
Inhalation Unit Risks5 provide information pertaining to toxicity for chemicals that may cause
cancer. Reference Doses (RfDs) and Reference Concentrations (RfCs) provide toxicity
information related to noncancer effects.6 Where these values are not available from EPA, other
data sources may be used.

The following data sources are used, in the order of preference:

• EPA's Integrated Risk Information System (IRIS);

• EPA Office of Pesticide Programs' Acute and Chronic Reference Doses Table,
List of Chemicals Evaluated for Carcinogenic Potential, and Pesticide
Reregi strati on Eligibility Documents (OPP);

• Final, published Minimum Risk Levels (MRLs) from the Agency for Toxic
Substances and Disease Registry (ATSDR);

• Final published toxicity values from California Environmental Protection
Agency's Office of Environmental Health Hazard and Assessment (CalEPA);

• EPA's Provisional Peer Reviewed Toxicity Values (PPRTVs), which include
toxicity values that have been developed by EPA's Office of Research and

5 The Oral Slope Factor represents the upper-bound (approximating a 95% confidence limit) estimate of the slope of
the dose-response curve in the low-dose region for carcinogens. The units of the slope factor are usually expressed
as (mg/kg-day)_1. The Inhalation Unit Risk is the upper-bound excess lifetime cancer risk estimated to result from
continuous exposure to an agent at a concentration of 1 (ig/m3 in air.

6 RfDs and RfCs are estimates (with uncertainty spanning perhaps an order of magnitude) of daily exposure [RfD],
or continuous inhalation exposure [RfC], to the human population (including sensitive subgroups) that is likely to be
without an appreciable risk of deleterious noncancer effects during a lifetime.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

Development/National Center for Environmental Assessment/Superfund Health
Risk Technical Support Center (STSC);

• EPA's Health Effects Assessment Summary Tables (HEAST); and

• Final Derived/Interim Derived Toxicity Weights estimated by EPA's Office of
Pollution Prevention and Toxics (Derived).

The toxicity scoring method separately evaluates exposure routes (inhalation and oral) and
classes of effects (cancer and noncancer). For each exposure route, chemicals are scored based
on their single most sensitive adverse effect; if a chemical exhibits both cancer and noncancer
effects, the higher of the two weights is assigned as the final weight for that route. The
algorithms used to assign toxicity weights are shown below in Exhibit ES.l.

Exhibit ES.l

Algorithms for Assigning Toxicity Weights

Oral Pathway

Inhalation Pathway

Non-Carcinogens:

1 / RfD (mg/kg-day)

3.5 / RfC (mg/m3)

Carcinogens

(WOE categories
A and B):

Oral Slope Factor (risk per
mg/kg-day)/ 1.0 x 10"6

Inhalation Unit Risk (risk
per mg/m3)/ 2.8 x 10"7

Carcinogens (WOE
category C):

Oral Slope Factor (risk per
mg/kg-day)/ (1.0 x 10"6 *
10)

Inhalation Unit Risk (risk
per mg/m3)/ (2.8 x 10~7 *
10)

The distribution of toxicity values for TRI chemicals corresponds to a range of toxicity weights
of approximately 0.02 to 1,400,000,000. However, toxicity weights are not bounded.
Continuous toxicity weights are expressed as values with two significant figures.

There are 620 chemicals and chemical categories on the 2011 TRI Chemical List. Toxicity
weights are available for 435 of these chemicals and chemical categories.7 Chemicals with
toxicity weights account for over 99% of the reported pounds for all on-site releases in 2011.

Adjustments for Exposure Potential and Population Size. Quantitatively, exposure potential
is estimated using a "surrogate" dose. To estimate the surrogate dose, a separate exposure
evaluation is conducted for each pathway-specific chemical emission. The exposure evaluations
use models that incorporate data on pathway-specific chemical releases and transfers,
physicochemical properties and, where available, site characteristics, to estimate the ambient
chemical concentration in the medium into which the chemical is released or transferred. The
ambient concentrations are combined with human exposure assumptions and estimates of
exposed population size specific to age and sex.

7 TRI lists metal compounds and their elemental forms separately,e.g., "lead" and "lead compounds" are two
categories in TRI. RSEI combines compounds and elemental forms into one category, as "lead and lead
compounds," for instance.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

The algorithms for calculating surrogate doses rely on the ability to locate facilities and people
geographically on the 810m by 810m grid cell system described earlier. While this method uses
the EPA exposure assessment paradigm to evaluate exposure potential, the results should not be
construed as an actual numerical estimate of dose resulting from TRI releases, because limited
facility-specific data and the use of models that rely on default values for many parameters
prevent the calculation of an actual dose. Instead, the purpose of the methodology is to generate
as accurate a surrogate dose as possible without conducting an in-depth risk assessment. The
estimates of surrogate doses from releases of TRI chemicals are relative to the surrogate doses
resulting from other releases included in the model. Please note that not all pathways are
currently modeled.

ES-2.2 RSEI Results

Because of the multi-functional nature of the model, a variety of results can be created. All
RSEI results are based on the Indicator Element, which is a unique combination of facility,
chemical, release pathway, exposure pathway, and year.8 Each Indicator Element has a set of
associated results:

Exhibit ES.2
Description of RSEI Results

Risk-related results

Surrogate Dose x Toxicity Weight x Population

Hazard-based results

Pounds x Toxicity Weight

Pounds-based results

TRI Pounds released

Risk-related results. The pathway-specific toxicity weight, surrogate dose, and population
components are multiplied to obtain a risk score for the Indicator Element. The surrogate dose is
determined through pathway-specific modeling of the fate and transport of the chemical through
the environment, combined with subpopulation-specific exposure factors. The score is a unitless
measure that is not independently meaningful, but is a risk-related estimate that can be compared
to other estimates calculated using the same methods. If the Indicator Element cannot be
modeled, because of a lack of data needed for modeling or because the release pathway is not
currently modeled, then the risk-related score is zero. The model calculates risk-related results
for the entire population and also for the following subpopulations: children under 10, children
aged 10 to 17, males aged 18 to 44, females aged 18 to 44, and adults aged 65 and older. In
addition, the model also calculates "Modeled Pounds," which is simply the number of pounds
that can be modeled for risk-related scores. Modeled pounds are the pounds to which fate and
transport modeling and exposure assumptions have been applied.

8 Several related Indicator Elements may be associated with certain release and exposure pathways (e.g., direct water
releases may be associated with exposure from drinking water intakes, as well as fish ingestion from recreational
fishing and from subsistence fishing).

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

Hazard-based results. Each Indicator Element is also associated with a hazard-based result
("Hazard"), calculated by multiplying the pounds released by the appropriate chemical-specific
toxicity weight (the toxicity weight also depends on the exposure pathway). The inhalation
toxicity weight is used for releases or transfers to fugitive air, stack air, off-site incineration, and
off-site incineration-no fuel value. The oral toxicity weight is used for releases to direct water
and transfers POTWs. For releases that are not modeled (because the pathway is not modeled or
because other necessary data, such as physicochemical properties, are lacking), the higher
toxicity weight is used. For these results, no exposure modeling or population estimates are
involved. If there is no toxicity weight available for the chemical, then the hazard score is zero.

The model also calculates "Modeled Hazard," which is the chemical- and pathway-specific
toxicity weights multiplied by the Modeled Pounds (as described above), and "Modeled Hazard
* Pop," which multiplies modeled hazard by the potentially exposed population, but without the
fate and transport modeling (and application of exposure assumptions) that would be found in
risk-related results.

Pounds-based results. These results ("TRI Pounds") reflect only the number of pounds released
or transferred that are reported to TRI, and are available for virtually all Indicator Elements. The
model also provides "TRI Pounds with Toxicity Weights," which simply sums the pounds for
chemicals that have toxicity weights in RSEI.

Once results are calculated for each Indicator Element, they can be combined in many different
ways. All of the results are additive, so a result for a specific set of variables is calculated by
summing all the relevant individual Indicator Element results, as follows:

This method is very flexible, allowing for countless variation in the creation of results. For
example, results can be calculated for various subsets of variables (e.g., chemical, facility,
release pathway) and compared to each other to assess the relative contribution of each subset to
the total potential impact. Or, results for the same subset of variables for different years can be
calculated, to assess the general trend in pounds-based, hazard-based, or risk-related impacts
over time.

It must be reiterated that while changes in results over the years would imply that there have
been changes in hazard- or risk-related environmental impacts, the actual magnitude of any
specific change or the reason may not be obvious. Although the value itself may be useful in
identifying facilities or chemicals with the highest potential for hazard or risk, the score does not
represent a quantitative estimate or provide an exact indication of the magnitude of individual
hazard or risk associated with that facility or chemical.

(Eq. ES.l)

where:

IEc,fp

RSEI result

chemical-facility-pathway-specific Indicator Element result

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

ES-2.3 Adjusting RSEI Results for Changes in TRI Reporting

When a change occurs in the number of, or reporting requirements for, chemicals and facilities
represented in TRI, the numerical value of RSEI results will be altered if no adjustments are
made to the method of calculation to account for the changes respective to trend analyses.
However, such changes would not necessarily represent a large change in actual environmental
impact, but would reflect a broader understanding of the impacts that may have always existed.
To maintain comparability in the weights over time, the results must be adjusted in some manner
when such changes in TRI reporting occur.

A change in the number of chemicals and facilities in the TRI database can occur through several
mechanisms. The addition to or deletion of chemicals from the TRI chemical list will occur as
EPA responds to petitions or initiates its own action through the chemical listing or delisting
process. The largest revision to the list occurred in November 1994, when the Agency added
245 chemicals and chemical categories to the existing TRI chemical list, effective for the
reporting year 1995. Other revisions have occurred since. To allow for meaningful trend
analysis, the model maintains a list of "core" chemicals which have been reported since 1987,
and for which reporting requirements have not changed.

Compliance with TRI reporting has changed over time, which has led to more facilities
reporting. Increases in the number of reporting facilities may also occur as a result of changes in
reporting requirements. For instance, in the first two years of reporting, facilities that
manufactured or processed more than 50,000 pounds were required to report their releases.
However, EPCRA lowered this threshold to 25,000 pounds in 1989. And for reporting year
2000, thresholds and other reporting requirements for 18 Persistent Bioaccumulative Chemicals
(PBTs) have been changed. Effective for reporting year 1998, TRI has enlarged the set of
facilities required to report to include electric utilities, mining facilities, commercial hazardous
waste facilities, solvent recovery facilities, and wholesale chemical and petroleum terminal
facilities. All of these modifications can act to alter the total emissions reported on the TRI and
the model's estimate of the associated relative risk-based impacts.

When deletions from the chemical list of TRI chemicals occur, RSEI's chemical database is
modified to remove all results from previous reporting years. Also, the yearly TRI data for a
given chemical list of chemicals and facilities are the subject of ongoing quality control review
and correction by both EPA and reporting facilities. As a result, yearly comparisons could be
flawed if such revisions in reported data were not included in each previous year's results.
Therefore, the Indicator Elements are recomputed for all years on an annual basis in order to
incorporate chemical deletions and revisions to the reporting data.

ES-2.4 Model Implementation

The RSEI model is currently implemented in a Microsoft Windows-based computer program.
The program allows users to calculate RSEI results for TRI reporting years 1996-2011 (earlier
years are available upon request) and to present the results in various GIS, graphical, and tabular
formats, as well as to save selected data to spreadsheet and database formats (e.g., Microsoft
Excel and databases such as Access). The program includes on-line help for all of the program
functions, as well as a User's Manual in Adobe Acrobat format.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

Users of the model can perform, usually in a matter of minutes, a variety of screening-level
analyses. Previously, such activities would have taken days, weeks, or even months to organize
the relevant information, evaluate that information, and perform the complex and sophisticated
analyses that are necessary to provide a risk-related perspective. Results can be used for
screening-level ranking and prioritization for strategic planning purposes, risk-related targeting,
and trends analyses. Considerable resources can be saved by conducting preliminary analyses
with the model to identify risk-related situations of high potential concern, which warrant further
evaluation.

As noted above, users can evaluate releases using a number of variables, such as chemical,
medium, geographic area or industry. For instance, the following types of questions can be
investigated:

• How do industry sectors compare to one another from a risk-related perspective?

• What is the relative contribution of chemicals within a given industry sector?

• What release pathway for a particular chemical poses the greatest risk-related
impacts?

Users can view pounds-based, hazard-based, and other results, to investigate the relative
influence of toxicity and population components on the risk-related results, which also
incorporate exposure modeling.

Information regarding the RSEI project is available on OPPT's RSEI web site at
http://www.epa.gov/oppt/rsei/. Complete documentation, frequently asked questions, and
contact information are all posted on the site. Periodic updates and troubleshooting information
are also available for users.

ES-2.5 How the RSEI Chronic Human Health Toxicity Weightings Differ from
EPCRA Section 313 Statutory Criteria

As described above, the RSEI model uses TRI chemical reporting data. However, it is important
that the public not confuse the use of the model as a screening-level tool for investigating relative
risk-based impacts related to the releases and transfers of TRI chemicals, with the very different
and separate activity of listing/deli sting chemicals on the TRI using statutory criteria.

The goal of RSEI is to use data reported to the Agency to investigate the relative risk-based
impacts of the releases and transfers of these chemicals on the general, non-worker population.
The model differentiates the relative toxicity of listed chemicals and ranks them in a consistent
manner. The ranking of each chemical reflects its toxicity only relative to other chemicals that
are included in the model. Toxicity is not compared to some benchmark or absolute value as is
required when adding or removing a chemical from the TRI. Furthermore, the model addresses
only the single, most sensitive toxicity endpoint for chronic human health.

In contrast, the EPCRA statutory criteria used for listing and delisting chemicals consider acute
and chronic human toxicity, as well as environmental toxicity, and consider multiple effects and
the severity of those effects. The criteria also address the "absolute" chronic toxicity of
chemicals on the TRI relative to a benchmark value.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

Because of these differences, the toxicity weightings in the model cannot be used as a scoring
system for evaluating listing/deli sting decisions. RSEI does not attempt to reflect the statutory
criteria for these chemicals.

ES-3. Important Caveats Regarding the RSEI Model

The RSEI model is a screening tool that provides a risk-related perspective in assessing the
relative impacts of releases of toxic chemicals. Risk-related results are available for releases and
transfers to air and water, and pounds- and hazard-based results are available for all media.

RSEI combines estimates of toxicity, exposure level, and the exposed population to provide
risk-related comparisons. It does not provide a detailed or quantitative assessment of risk, and is
not designed as a substitute for more comprehensive, site-specific risk assessments. There are a
number of important considerations associated with each component of the model, as described
in the following sections.

Release Component. The following caveat should be considered regarding the release
component of the model:

• RSEI uses facility-reported TRI data, which has been known to contain some
reporting errors. Since facility management must certify reports to be accurate,
the TRI program does not change any reported data until the reporting facility
submits an official correction. Therefore, there are some releases in the TRI data
that are thought to be erroneous but are still included because facilities have not
submitted corrected reporting forms by the time of the annual public data release
that RSEI uses. Some of these releases are associated with large risk-related
impacts. One erroneous release warrants special note: a 2002 fugitive air release
of 184,770 pounds of nickel in Johnstown, PA, probably overstates the release
amount and may be assigned to the wrong media.

Toxicity Component. The following caveats should be considered regarding the toxicity
component of the model:

• Toxicity weights are not designed to (and may not) correlate with statutory
criteria used for listing and delisting chemicals in TRI. RSEI risk-related model
results account for estimated exposure and may not correlate with listing/de-
listing decisions.

• The RSEI model only addresses chronic human toxicity (cancer and noncancer
effects, such as developmental toxicity, reproductive toxicity, neurotoxicity, etc.)
associated with long-term exposure and does not address concerns for either acute
human toxicity or environmental toxicity.

• Toxicity weights are based upon the single, most sensitive chronic human health
endpoint for inhalation or oral exposure pathways, and do not reflect severity of
effects or multiple health effects.

• Estimated Reference Doses and Reference Concentrations for noncancer effects
incorporate uncertainty factors which are reflected in toxicity weights that are
based upon these values.

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

• Toxicity weights for chemicals that are reported to TRI as groups are based on the
toxicity for the most toxic member of the group. One exception to this is
polycyclic aromatic compounds (PACs), the toxicity for which is assumed to be
18% of the toxicity for benzo(a)pyrene, its most toxic member. This is based on
speciation information and follows the method used by EPA's National-Scale Air
Toxics Assessment (NATA) evaluation for polycyclic organic matter.9

• Several significant assumptions are made regarding metals and metal compounds,
because important data regarding these chemicals are not subject to TRI reporting.
Metals and metal compounds are assumed to have the same toxicity weight,
although the chronic toxicity of some metal compounds may be higher. Metals
and metal compounds are assumed to be released in the valence (or oxidation
state) associated with the highest chronic toxicity. There are two exceptions to
this: (1) For chromium and chromium compounds, it is assumed that facilities
may release some combination of hexavalent chromium and trivalent chromium.
SIC-code specific estimates from the 2002 National Emissions Inventory are used
to estimate the fraction of each type.10 As trivalent chromium has a very low
toxicity, only the hexavalent fraction is modeled, using a toxicity weight
specifically for that valence state; and (2) For mercury and mercury compounds,
toxicity for the oral pathway is based on methyl mercury, and toxicity for the
inhalation pathway is based on elemental mercury.

• While the physical form of released metals or metal compounds can affect
toxicity, a reasonable assumption is made regarding the likely form of most
releases (e.g., the non-cancer toxicity weight for chromic acid mists and dissolved
hexavalent chromium aerosols is much higher than for hexavalent chromium
particulates, but releases of these chemicals as acid aerosols are not expected to be
typical so the toxicity weight for cancer based on the inhalation of particulates is
used). Analysts need to consider these assumptions, and whether the gathering of
additional data are warranted, when examining model results for metals and metal
compounds.

Exposure Component. The following caveats should be considered regarding the exposure
component of the model:

• Like other exposure models, RSEI estimates exposure levels. It does not yield
actual exposures. The model provides estimated air concentrations in each grid
cell.

• The model uses some generic assumptions, e.g., default median stack heights,
diameters, and exit gas velocities related to 2- or 3-digit Standard Industrial
Classification (SIC) codes, or a nationwide median, where facility-specific
median stack height, diameter, and exit gas velocity data are unavailable. For

9 Additional information is available in the NATA documentation (http://www.epa.gov/ttn/atw/sab/appendix-h.pdf).
RSEI assumes that PAC emissions reported to TRI are most like NATA's "7-PAH" category.

10 Available from http://www.epa.gov/ttn/chief/net/2002inventory.html

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Executive Summary

large facilities with multiple stacks, the median height for all stacks is used as the
stack height for the entire facility.

•	In the current version of the model, only air and direct surface water exposures
are fully modeled.

•	The model does not account for population activity patterns.

•	The model has greater uncertainty when examining disaggregated results at the
local or facility level. The model does not account for indirect exposure, air
deposition of pollutants to other media, or absorption of pollutants through the
skin.

Population Component. The following caveats should be considered regarding the population
component of the model:

•	Population values for non-decennial years are estimated based on linear
interpolations at the block level between the 1990 and 2000 and between the 2000
and 2010 U.S. Census datasets, and on extrapolation back to 1988 and forward to
2011.

•	Drinking water populations are estimated by using the total drinking water
populations associated with individual downstream drinking water intakes.
Estimated populations for the fish ingestion pathway are based upon U.S. Fish
and Wildlife Service surveys.

•	Because RSEI results reflect changing population size at the local level, a
facility's relative contribution could increase or decrease even without changes in
its releases over time. While the model is designed to reflect the overall risk-
related impacts on the local population, such population changes should be
considered when examining a facility's environmental management practices.

ES-4. New Features in Version 2.3.2

•	Includes TRI Reporting Years 1988-2011 (data for years 1996-2011 are available
in the public release; data for 1988-1995 are available upon request).

•	Toxicity weights have been updated with the most recent toxicity data.

•	Dioxin and dioxin-related compounds are now modeled for stack and fugitive air
releases, direct water releases, and off-site incineration from reporting year 2008
on. Dioxin transfersto POTWs are not modeled, due to the lack of necessary
physic-chemical data.

•	Exposure factors have been updated.

•	Drinking water intake locations have been updated.

Version 2.3.2
July 2013

12

1988-2011 TRI data

-------
Chapter 1: Introduction

1. Introduction

EPA's Risk-Screening Environmental Indicators (RSEI) is a screening-level tool that assesses
the potential impact of industrial releases from pounds-based, hazard-based, and risk-related
perspectives. RSEI uses risk concepts to quickly and easily screen large amounts of TRI data,
saving time and resources. RSEI is particularly useful for examining trends to measure change,
ranking and prioritizing chemicals and industry sectors for strategic planning, conducting risk-
related targeting, and supporting community-based projects.

The RSEI approach is very flexible and can be implemented in various ways. The use of the
model is not limited to TRI chemicals; in principle, the adaptable method can model any
chemical if toxicity characteristics, physicochemical properties, release levels, and release
location are known or can be estimated.

1.1 Background

In 1989, EPA outlined the goals for establishing strategic planning processes at the Agency
(EPA, 1990c). Underlying this approach was the Agency's desire to set priorities and direct
resources to areas with the greatest opportunity to achieve health and environmental risk
reductions. As part of this initiative, the Administrator set forth a plan to develop indicators to
track changes in environmental health impacts over time. Tracking these changes would allow
the Agency to measure its progress in implementing environmental protection and pollution
prevention programs. In addition, comparing the relative contribution of particular chemicals,
industries and geographic regions through the indicators would allow the Agency (and other
users) to establish priorities for improving human health and the environment.

To efficiently track changes in human health and environmental impacts over time, the Agency
should take advantage of existing data sources that reflect multimedia trends in environmental
contaminant releases. The TRI is one of the Agency's most relevant source of continuous data

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Chapter 1: Introduction

for developing indicators of change in environmental impacts over time. The TRI is mandated by
the Emergency Planning and Community Right-to-Know Act (EPCRA) Title III Section 313 and
requires that U.S. manufacturing facilities file annual reports documenting multimedia
environmental releases and off-site transfers for more than 600 chemicals and chemical
categories which are of concern to the Agency.

In response to the need for environmental indicators, and to take advantage of the rich data
source offered by the TRI, the Office of Pollution Prevention and Toxics (OPPT) convened a
workgroup to explore the development of an indicator or indicators based on the TRI that could
track changes in human health and environmental impacts. Specifically, the approach would
integrate toxicity, exposure and population considerations into the risk-related evaluation of
releases. The RSEI model was developed in response to this initiative.

When evaluating impacts of chemicals, it is important to not only consider the number of pounds
of a chemical released to the environment, but also the toxicity of the chemical, its exposure
potential, and the size of the receptor population. RSEI integrates these factors and provides a
relative risk-based perspective of chemical releases and transfers. To the extent possible, the
RSEI model is based on existing EPA approaches, data, and models, to minimize duplication of
effort and to maximize consistency with other Agency efforts to evaluate human health impacts.

The current version of the model tracks changes in chronic human health impacts. Ultimately,
the model may be expanded to track acute human health and chronic and acute ecological
impacts.

This document explains how the RSEI model is constructed, and describes the conceptual
method, data sources, and the computational approach. The aim is to explain the model to a
variety of agencies and groups that may wish to use or adapt the model, or the RSEI
methodology in general, to their own needs. In addition, it describes the advantages of the RSEI
approach in terms of flexibility, power, and usefulness as an analytical and strategic policy-
planning tool.

1.2 Model Implementation

The RSEI model is currently implemented in a Microsoft Windows-based computer program.
Version 2.3.2 of the model contains TRI data from 1996-2011, and allows users to calculate
RSEI results for these years of reporting, and to present the results in various GIS, graphical, and
tabular formats, as well as to save selected data to spreadsheet and database formats (e.g.,
Microsoft Excel and databases such as Access). Data for the previous years of TRI reporting
(1988 through 1995) are available upon request. The program includes on-line help for all of the
program functions, as well as a set of introductory tutorials for first-time users. A User's Manual
is also available.

Users of the RSEI model can perform a variety of screening-level analyses, usually in a matter of
minutes. Previously, such activities would have taken days, weeks, or even months to organize
the relevant information, evaluate that information, and perform the complex and sophisticated
analyses that are necessary to provide a risk-related perspective. Results can be used for
screening-level ranking and prioritization for strategic planning purposes, risk-related targeting,

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Chapter 1: Introduction

and trends analyses. Considerable resources can be saved by conducting preliminary analyses
with the RSEI model to identify risk-related situations of high potential concern, and which
warrant further evaluation.

Users can evaluate releases using a number of variables, such as chemical, medium, geographic
area or industry. For instance, users can investigate the following types of questions: How do
industry sectors compare to one another from a risk-related perspective? What is the relative
contribution of chemicals within a given industry sector? What release pathway for a particular
chemical poses the greatest risk-related impacts? Users can view pounds-based, hazard-based,
and other results, to investigate the relative influence of toxicity and population on the risk-
related results, which also incorporate exposure modeling.

Information regarding the RSEI project is available on OPPT's RSEI web site, at
www.epa.gov/oppt/rsei/.

1.3 Organization of this Document

Chapter 2 of this document gives a brief description of the RSEI method and model, as well as a
discussion of their overall strengths and limitations. Chapter 3 describes the TRI emissions data
used in the model. Chapter 4 describes the methods used to adjust the emissions data for
chemical toxicity, and Chapter 5 provides a discussion of the geographic basis of the model, as
well as pathway-specific descriptions of adjustments made for exposure potential and population
size. Chapter 6 presents the equations for calculating RSEI results, and Chapter 7 describes
issues pertinent to the current implementation of the RSEI method.

There are also six Technical Appendices that accompany this methodology document and
provide additional information on the data used in the model. The Appendices are as follows:

Technical Appendix A - Listing of All Toxicity Weights for TRI Chemicals and Chemical

Category

Weight-of-Evidence

Sufficient evidence from epidemiological studies to support a causal relationship
between exposure to the agent and cancer.

Limited evidence from epidemiological studies and sufficient animal data.

Sufficient evidence from animal studies but inadequate or no evidence or no data from
epidemiological studies.

Limited evidence of carcinogenicity in animals and an absence of evidence or data in
humans.

Inadequate human and animal evidence for carcinogenicity or no data.

No evidence for carcinogenicity in at least two adequate animal tests in different
species or in both adequate epidemiological and animal studies, coupled with no
evidence or data in epidemiological studies.

Source: 51 FR 33996

For noncancer effects, weight-of-evidence is considered qualitatively in the hazard
identification step of determining an RfD or an RfC. The WOE evaluation for noncancer effects
is different from that for carcinogenic effects. The WOE judgment for noncancer effects focuses
on the dose where chemical exposure would be relevant to humans (Dourson, 1993). That is, the
focus of the WOE evaluation and the expression of the level of confidence in the RfD is a
judgment of the accuracy with which the dose relevant to humans has been estimated. The WOE
evaluation is included qualitatively in the RfD, but does not affect its numerical calculation.

Since weight-of-evidence has been considered in developing RfDs, RSEI does not consider
WOE separately for noncancer effects.

4.1.2 Quantitative Data

Quantitative data on the relative potencies of chemicals are needed for toxicity weighting.

These data generally result from analyses done during the third stage of risk assessment, the
dose-response assessment. This stage involves describing the quantitative relationship between

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

the amount of exposure to a chemical and the extent of toxic injury or disease observed. Risk
posed by exposure to a chemical cannot be described without quantitative dose-response data.
Dose-response data are derived from animal studies or, less frequently, from studies in exposed
human populations. There may be many different dose-response relationships for a chemical if it
produces different toxic effects under different conditions of exposure.

For cancer risk assessment, EPA has developed standard methods for predicting the
incremental lifetime risk of cancer per dose of a chemical. EPA quantitatively models the dose-
response function of a potential carcinogen and typically provides estimates of Oral Slope
Factors (OSFs) or Inhalation Unit Risks (IURs). The OSF represents the upper-bound estimate
of the slope of the dose-response curve in the low-dose region for carcinogens, and is a measure
of cancer potency. The units of the slope factor are usually expressed as (mg/kg-day)"1. The
IUR is the upper-bound excess lifetime cancer risk estimated to result from continuous exposure
to an agent at a concentration of 1 [j,g/m3 in air (RSEI toxicity weights are based on this value
when expressed as risk per mg/m3). Although the level of conservatism inherent in the OSFs and
Unit Risks varies by chemical, OSFs and IURs nonetheless are the best readily available values
that allow a comparison of the relative cancer potency of chemicals.

RSEI's oral cancer toxicity weight (OTWc) represents how toxic a substance is relative to a
substance that produces a 1 in 1 million risk15 (above background, over a lifetime) at an average
lifetime daily dose of 1 mg/kg-day. If the OSF is greater than this arbitrary slope factor (i.e. the
substance is more toxic than the arbitrary slope factor), the OTWc is greater than 1.

omc = OSFJ^oyhng_ ^ 4

10 kg-day/mg

For noncancer risk assessment, data on dose-response are typically (though not always) more
limited; generally, a risk assessor evaluates dose compared to a Reference Dose (RfD) or
Inhalation Reference Concentration (RfC). Both the RfD and RfC are defined as "an estimate
(with uncertainty spanning perhaps an order of magnitude) of a daily exposure [or continuous
inhalation exposure the RfC] to the human population (including sensitive subgroups) that is
likely to be without an appreciable risk of deleterious [noncancer] effects during a lifetime"
(EPA, 1988a; EPA, 1990b). The units of RfD are mg/kg-day, while the units of the RfC are
mg/m3. A chemical's RfD or RfC is typically based on a No Observable Adverse Effect Level
(NOAEL) or Lowest Observable Adverse Effect Level (LOAEL), combined with appropriate
uncertainty factors to account for intraspecies variability in sensitivity, interspecies extrapolation,
extrapolation from LOAELs to NOAELs, and extrapolation from subchronic to chronic data. In
addition, a modifying factor can be applied to reflect EPA's best professional judgment on the
quality of the entire toxicity database for the chemical. By definition, exposures below the
RfD/RfC are unlikely to produce an adverse effect; above this value, an exposed individual may

15 EPA programs commonly use a risk management range corresponding to an excess individual lifetime risk of
cancer of 1 in 10-6 to 10-4 (EPA, 1999a).

Version 2.3.2
July 2013

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

be at risk for the effect. Empirical evidence generally shows that as the dosage of a toxicant
increases, the severity and/or incidence of effect increases (EPA, 1988a), but for a given dose
above the RfD/RfC, the specific probability or severity of an effect is not known. For purposes
of the RSEI model, it is assumed that noncancer risk varies as the ratio of the estimated dose to
the RfD/RfC.

RSEI's oral non-cancer toxicity weight (OTWnc) represents how toxic a substance is relative to
an arbitrary dose of 1 mg/kg-day. If the RfD is greater than this arbitrary dose (i.e. the substance
is less toxic than the arbitrary dose), the OTWnc is less than 1.

OTWnc = lmSltg-
-------
Chapter 4: Methods for Calculating Toxicity Weights

Exhibit 4.3
Algorithms for Assigning Toxicity Weights



Oral Pathway

Inhalation Pathway

Non-Carcinogens:

1 / RfD (mg/kg-day)

3.5 / RfC (mg/m3)

Carcinogens

(W OE categories
A and B):

Oral Slope Factor (risk per
mg/kg-day)/ 1.0 x 10"6

Inhalation Unit Risk (risk
per mg/m3)/ 2.8 x 10~7

Carcinogens

(WOE category C):

Oral Slope Factor (risk per
mg/kg-day)/ (1.0 x 10"6 * 10)

Inhalation Unit Risk (risk
per mg/m3)/ (2.8 x 10~7 * 10)

4.2 Selecting the Final Chronic Human Health Toxicity Weight for a
Chemical

Inhalation and oral toxicity weights are developed separately. If values are available for each
route, then separate toxicity weights are assigned to each route. If data are available for only one
route, the same toxicity weight is applied for both routes, provided there is no evidence the
effects are route-specific or limited to the "portal of entry" into the body. In rare instances,
toxicity studies are available to show that a given chemical causes no effects via one route; in
these instances, a toxicity weight is assigned only to the route that results in chronic human
health effects. Although assigning the same weight to both routes is not an ideal method, it is
appropriate for a screening-level tool like the RSEI model.

Although chemicals can cause several types of toxic effects, the model assigns a toxicity weight
to a chemical based on the single most sensitive adverse effect for the given exposure route (oral
or inhalation). If a chemical exhibits both carcinogenic and noncarcinogenic effects, the higher
of the associated cancer and noncancer weights is assigned as the final weight for the chemical
for the given route.

The approach of weighting based on the most sensitive adverse effect does not consider
differences in the type, number or target of the effects posed by the chemicals. For example,
liver toxicity is weighted in the same way that neurotoxicity is weighted; in principle, chemicals
causing a certain type of effect could be assigned additional weight if special concern existed for
that type of effect. However, applying such additional weights would require a subjective
evaluation of the relative severity of the health effects. Also, chemicals with a broad range of
adverse health effects are weighted the same as a chemical causing only one effect. This
approach may appear to under-estimate the risk of chemicals with a broad spectrum of effects
relative to chemicals with one or few effects. However, a chemical may appear to demonstrate
just one adverse effect only because there are no data on other effects; thus, applying an
additional weight based on the number of endpoints may undervalue some poorly-studied but
still hazardous chemicals. For these reasons, the options for applying additional weights based
on number and relative severity of endpoints were not adopted.

Version 2.3.2
July 2013

17

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

4.3 Chemical Groups

TRI collects information for some chemicals as combined groups, such as glycol ethers,
polycyclic aromatic compounds, and metal compounds. For metal compounds, RSEI combines
the elemental form of the metal with its compounds category. This is done to reflect the
uncertainty of the chemical identity of the substance released. For example, TRI has separate
reporting for 'nickel' and 'nickel compounds'. Both reflect the pounds of the parent metal nickel
that is released and, in some cases, the two can be combined as a single report for nickel
compounds. RSEI combines the pounds into one entry listed as 'nickel and nickel compounds'.

For all chemical groups, data for the most toxic member of the group is used to represent the
toxicity of the entire group. There are three exceptions. The first exception is chromium and
chromium compounds, for which it is assumed that facilities may release some combination of
hexavalent chromium and trivalent chromium. SIC-code specific estimates from the 2002
National Emissions Inventory are used to estimate the fraction of each type.16 As trivalent
chromium has a very low toxicity, only the hexavalent fraction is modeled, using a toxicity
weight specifically for that valence state.

The second exception is for polycyclic aromatic compounds (PACs), for which it is assumed that
the overall toxicity for the category is 18% of the toxicity for benzo(a)pyrene, the most toxic
member of the group. This figure is based on the polycyclic organic matter (POM) methodology
used by EPA's National-Scale Air Toxics Assessment (NATA) model, which is based on
emissions factors for representative processes used in industries that emit large amounts of POM.
RSEI assumes that the TRI PACs category is most like the 7-PAH category used in NATA.17

The third exception is for dioxins and dioxin-related compounds. EPA first required reporting of
releases and transfers for this category in 2000. Facilities were required to report total dioxin
releases/transfers (in grams) released/transferred to each medium, as well as the amount of each
of the 17 congeners that comprise the category released/transferred to all media combined. EPA
changed the reporting requirements in 2008, when reporters were required to provide the
congener breakdown for each medium. Toxicity information is only available for one congener,
TCDD, but EPA has determined a toxicity equivalence factor (TEF) for each congener, based on
its toxicity relative to TCDD18. RSEI combines TRI's congener breakdown with EPA's TEFs to
calculate a weighted average TEF for each release. When multiplied by the toxicity weight for
TCDD, this provides a toxicity weight for each dioxin release/transfer. For releases/transfers
where the congener breakdown is blank or invalid, RSEI adopts the mean TEF for all of the
dioxin releases to that medium in the reporting facility's 4-digit NAICS code. If a 4-digit NAICS
code for the reporting facility is not available, the overall mean for the specific medium is used.
Because the congener breakdown at the level of the release/transfer is only available from 2008

16	Available from http://www.epa.gov/ttn/chief/net/2002inventory.html

17	The documentation on modeling polycyclic organic matter (POM) from EPA's NATA model can be found at
http://www.epa.gOv/ttn/atw/nata2002/methods.html#pom.

18	TEFs are consensus estimates of compound-specific toxicity/potency relative to the toxicity/potency of an index
chemical. TEFs are the result of expert scientific judgment using all of the available data and taking into account
uncertainties in the available data. For more detail on the dioxin TEFs, see http://www.epa.gov/raf/files/tefs-for-
dioxin-epa-00 -r-10-005-final. pdf

Version 2.3.2
July 2013

18

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

on, hazard and full modeling results are only available for that period; however users can
examine pounds-only results for dioxins from 2000 on.

Toxicity weights for individual chemicals and chemical groups are presented in Technical
Appendix A.

4.4 Sources of Data

Information regarding the human health effects data on the TRI chemicals is compiled from the
following sources:

IRIS. The primary (and most preferred) source of these data is EPA's Integrated Risk
Information System (IRIS). IRIS is available on the internet (at http://www.epa.gov/iris/), and
includes information on EPA evaluations of chemical toxicity for both cancer and noncancer
effects of chemicals. IRIS provides both background information on the studies used to develop
the toxicity evaluations and the numerical toxicity values used by EPA to characterize risks from
these chemicals. These values include upper-bound OSF or IUR values for chemicals with
carcinogenic effects as well as RfDs or RfCs for chemicals with noncancer effects. Data
contained in IRIS have been peer-reviewed and represent Agency-wide expert judgments. The
peer-review process involves literature review and evaluation of a chemical by individual EPA
program offices and intra-Agency work groups before inclusion in IRIS.

OPP. EPA's Office of Pesticide Programs (OPP) Acute Chronic and Reference Doses Table
lists OPP's evaluations of the noncarcinogenic potential of chemicals that are of interest to OPP.
OPP also publishes the List of Chemicals Evaluated for Carcinogenic Potential, which examines
carcinogens. Both of these lists are updated periodically. Individual Pesticide Reregi strati on
Eligibility Documents (REDs) are also used.

ATSDR. The Agency for Toxic Substances and Disease Registry (ATSDR) is an agency of the
U.S. Department of Health and Human Services, which deals with the effect on public health of
hazardous substances in the environment. ATSDR develops Minimum Risk Levels (MRLs) for
chemicals on the CERCLA National Priorities List. An MRL is an estimate of the daily human
exposure to a hazardous substance that is likely to be without appreciable risk of adverse
noncancer health effects over a specified duration of exposure. RSEI uses data from MRLs
developed for chronic exposure only. MRLs are intended to serve as screening levels only, and
are useful in identifying contaminants and potential health effects that may be of concern at
hazardous waste sites. See http://www.atsdr.cdc.gov/mrls/index.html for more information on
MRLs and specific values.

CalEPA. The California Environmental Protection Agency (CalEPA) Office of Environmental
Health Hazard and Assessment (OEHHA) is responsible for developing and distributing
toxicological and medical information needed to protect public health. RSEI uses final toxicity
values published by CalEPA in the Consolidated Table of OEHHA & California's Air Resources
Board (ARB) Approved Risk Assessment Health Values. The table is continuously updated and
can be found on the internet at http://www.arb.ca.gov/toxics/healthval/healthval.htm.

Version 2.3.2
July 2013

19

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

PPRTVs. EPA's Provisional Peer Reviewed Toxicity Values (PPRTVs) include toxicity values
developed by the Office of Research and Development/National Center for Environmental
Assessment/Superfund Health Risk Technical Support Center (STSC).

HEAST. EPA's Health Effects Assessment Tables (HEAST) are constructed for use in the
Superfund and RCRA programs but do not represent Agency-wide expert judgments. These
tables are publicly available from the Superfund program. The tables include OSFs, IURs, and
WOE categorizations for chemicals with cancer effects, and RfDs and RfCs for noncancer
effects.

Derived Values. For a prioritized group of chemicals for which sufficient data was not found in
the above sources, a group of EPA expert health scientists reviewed other available data to derive
appropriate toxicity weights. Although individual literature searches for toxicological and
epidemiological data for each chemical were beyond the scope of this project, sources such as
the Hazardous Substances Data Base (HSDB), as well as various EPA and ATSDR summary
documents, provided succinct summaries of toxic effects and quantitative data, toxicological and
epidemiological studies, and, in some cases, regulatory status data. When the available data on
chronic human toxicity were sufficient to derive values, a toxicity weighting summary was
developed summarizing the information used to develop each of these values. The summaries
are available in Technical Appendix A. The EPA scientists use a technical approach analogous
to the Agency's method for deriving RfD values, RfC values, cancer risk estimates, and weight-
of-evidence (WOE) determinations. However, it must be emphasized that these derived values
are not the equivalent of the more rigorous and resource-intensive IRIS process and are only
useful for screening-level purposes.

Data from these sources are categorized in three-tiered, hierarchical fashion to give preference to
EPA and consensus data sources, where possible. Data are gathered separately for individual
endpoints; a chemical's RfD may be from IRIS, while its OSF may be from HEAST.

The hierarchy used in toxicity weighting is as follows:

Tier 1. The most recent data from IRIS and OPP is used for each chronic health endpoint. If the
dates are comparable, preference is given to IRIS, since IRIS reflects Agency-wide judgments.

Tier 2. In the absence of data from the above sources for an individual chronic health endpoint,
toxicity data from the most recent entry in ATSDR and CalEPA are used.

Tier 3. In the absence of data from the above sources for an individual chronic health endpoint,
the following data sources, in this order, are used: 1) PPRTVs; 2) HEAST; 3) Derived; and 4)
IRIS values previously used in toxicity weighting, that were withdrawn pending revision.

For chemicals with carcinogenicity risk values, weight-of-evidence (WOE) values are obtained
using the same data source hierarchy. Therefore, preference is given to WOE's from IRIS or
OPP. As a general rule, chemicals with cancer potency factors from IRIS or OPP will also have
WOEs. CalEPA, however, references either EPA or the International Agency for Research on
Cancer (IARC) for WOE designations. Therefore, in the absence of an EPA consensus WOE,
WOE's are obtained from IARC. However, due to the differences in WOE definition, it is not

Version 2.3.2
July 2013

20

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

always possible to translate IARC WOE's into EPA WOE's without examining the toxicity data.
WOE's are matched in the following way:

•	IARC Group 1 = EPA Group A (Human Carcinogen)

•	IARC Group 2A = EPA Group B (Probable Human Carcinogen)

•	IARC Group 2B = EPA Group B or EPA Group C (Possible Carcinogen)

•	IARC Group 3 = EPA Group D (Not Classifiable as to Human Carcinogenicity)

•	IARC Group 4 = EPA Group E (Evidence of Non-Carcinogenicity)

The IARC 2B designation is not easily translated to the EPA designation, because its definition
spans EPA Groups B and C. This is a particularly important distinction because the use of a B2
or C designation will affect the calculation of the toxicity weight. Therefore, for the chemicals
with IARC 2B designations, summaries of the toxicity data used to generate the OSF or IUR are
evaluated to derive WOEs. To date, this approach has been used for chemicals with data from
CalEPA; therefore, the CalEPA "Technical Support Document for Describing Available Cancer
Potency Factors" was used for the background information.

Currently, using all of the available data sources described above, toxicity weights are available
for 435 of 620 chemicals and chemical categories on the 2011 TRI Chemical List. Chemicals
with toxicity weights account for over 99% of the reported pounds for all on-site releases in
2011. The Indicator Elements are recomputed for all years in the TRI database on an annual
basis in order to incorporate revisions to the reporting data. However, only reporting years 1996
through 2011 are available in RSEI Version 2.3.2. Data for reporting years 1988 through 1995
are available upon request.

Toxicity weights for individual chemicals and chemical groups are presented in Technical
Appendix A.

4.5 How Indicator Toxicity Weightings Differ from EPCRA Section 313
Criteria

As noted above, the model uses TRI chemical reporting data. All TRI chemicals included in the
model are listed on the TRI because they meet one or more statutory criteria regarding acute or
chronic human toxicity, or environmental toxicity. The goal of the RSEI model is to use these
data reported to the Agency to investigate the relative risk-based impacts of the releases and
transfers of these chemicals on the general, non-worker population. To achieve this goal, the
model differentiates the relative toxicity of listed chemicals and ranks them in a consistent
manner. The ranking of each chemical reflects its toxicity only relative to other chemicals that
are included in the model. Toxicity is not compared to some benchmark or absolute value as is
required when adding or removing a chemical from the TRI. Furthermore, the model addresses
only the single, most sensitive chronic human health toxicity endpoint.

It is important that the public not confuse the use of the RSEI model as a screening-level tool for
investigating relative risk-based impacts related to the releases and transfers of TRI chemicals,
with the very different and separate activity of listing/deli sting chemicals on the TRI using
statutory criteria. A description of the listing/deli sting criteria and process is described below.

Version 2.3.2
July 2013

21

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

The Emergency Planning and Community Right-to-Know Act of 1986 (EPCRA) section
313(d)(2) sets out criteria for adding chemicals to the list of chemicals subject to reporting under
EPCRA section 313(a). The statutory criteria used for listing and delisting chemicals addresses
the "absolute" chronic toxicity of chemicals on the TRI (e.g., multiple effects or the severity of
effects). For a chemical (or category of chemicals) to be added to the EPCRA section 313(c) list
of toxic chemicals, the Administrator must judge whether there is sufficient evidence to establish
any one of the following:

Acute Human Toxicity §313(d)(2)(A) - The chemical is known to cause or can
reasonably be anticipated to cause significant adverse acute human health effects at
concentration levels that are reasonably likely to exist beyond facility site boundaries as a
result of continuous, or frequently recurring, releases.

Chronic Human Toxicity §313(d)(2)(B) - The chemical is known to cause or can
reasonably be anticipated to cause in humans-

(i)	cancer or teratogenic effects, or

(ii)	serious or irreversible-

(I)	reproductive dysfunctions,

(II)	neurological disorders,

(III)	heritable genetic mutations, or

(IV)	other chronic health effects.

Environmental Toxicity §313(d)(2)(C) - The chemical is known to cause or can
reasonably be anticipated to cause, because of-

(i)	its toxicity,

(ii)	its toxicity and persistence in the environment, or

(iii)	its toxicity and tendency to bioaccumulate in the environment, significant

adverse effect on the environment of sufficient seriousness, in the judgment of the

Administrator, to warrant reporting under this section.

To remove a chemical from the section 313(c) list, the Administrator must determine that there is
not sufficient evidence to establish any of the criteria described above as required by EPCRA
section 313(d)(3).

The EPA examines all of the studies available for a chemical to decide if the chemical is capable
of causing any of the adverse health effects or environmental toxicity in the criteria. Agency
guidelines describe when a study shows such effects as cancer (EPA, 1986a), developmental
toxicity (teratogenic effects) (EPA, 1991b), or heritable genetic mutations (EPA, 1986b). The
review makes a qualitative judgment regarding the potential of each chemical to meet at least
one of the criteria and the chemical is added to the list if this judgment is positive. If a chemical
is on the list and it is not possible to make a positive judgment regarding any of the criteria, then
the chemical can be removed.

There is no correlation between the toxicity criteria and methodology used to make listing
decisions under EPCRA section 313 and the methodology used to assign toxicity weights to
chemicals for the RSEI model. Therefore, these toxicity weights cannot be used as a scoring

Version 2.3.2
July 2013

22

1988-2011 TRI data

-------
Chapter 4: Methods for Calculating Toxicity Weights

system for evaluating listing/deli sting decisions. RSEI also does not attempt to reflect the
statutory criteria for these chemicals.

Version 2.3.2	1988-2011 TRI data

July 2013	23

-------
Chapter 5: Exposure and Population Modeling

5. Exposure and Population Modeling

To estimate the magnitude of exposure potential from TRI releases, a separate exposure
evaluation is conducted for each chemical release pathway. The following pathways are
evaluated:

•	Air: stack and fugitive pathways;

•	Surface Water: drinking water intake and fish ingestion pathways;

•	Publicly-Owned Treatment Works (POTWs): fugitive air, groundwater (not
currently modeled), drinking water intake and fish ingestion pathways;

•	Land: groundwater pathway (not currently modeled), volatilization to air
included in fugitive air emission pathway for on-site releases; and

•	Off-site transfers: groundwater (not currently modeled), volatilization (not
currently modeled) and stack air (from incineration) pathways.

The ideal derivation of a dose would involve a site-specific exposure assessment for each release
and exposure pathway. However, such an effort is well beyond the scope of this project; further,
reporting of extensive site-specific information relevant for exposure modeling is not part of a
TRI data submission. For example, the EPA Form R (Toxic Release Inventory Reporting Form)
does not require submission of data on groundwater flow, soil conditions, and other factors that
affect groundwater contamination from land releases. Although some site-specific data are used
in the model, it is not the intent of this project to gather extensive site-specific data or
measurements that would be needed to perform site-specific calculations. The need to accurately
reflect exposure characteristics in the RSEI model must be balanced by the need for simple and
understandable results that are easily communicated to the public and that are based on currently
available data.

Therefore, in this method, the exposure evaluations combine data on pathway and media-specific
emission volumes, physicochemical properties and, where available, site characteristics, with
models to determine an estimate of the ambient concentration of contaminant in the medium into
which the chemical is released. The ambient media concentrations are then combined with
human exposure assumptions to estimate a "surrogate dose". The term surrogate dose is used
because limited facility-specific data and the use of models that rely on default values for many
parameters preclude the calculation of an actual dose estimate. Instead, the purpose of the
method is to generate as accurate a surrogate dose as possible without conducting an in-depth
risk assessment. The estimates of surrogate doses from releases of TRI chemicals are
comparable only to the surrogate doses resulting from other releases included in the model.

Estimates of the surrogate dose for each potentially exposed person are combined with estimates
of the number of people potentially exposed. Potential exposure is determined by the geographic
location of the population, as identified by the decennial U.S. Census for 1990, 2000, and 2010.
The size of the exposed population is calculated separately for each pathway. The model
assumes continuous exposure, and does not account for the activity patterns of the people
potentially exposed. However, population estimates do consider changing demographic patterns
(total population, as well as subpopulations by age and sex).

Version 2.3.2
July 2013

24

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

The methods used to model each type of release are specific to that type of release and depend on
data available to evaluate that pathway. In some cases, models are combined with some site-
specific data to estimate exposure; in other cases, generic reasonable worst-case models may be
used in the absence of any site-specific data. The physicochemical property data used for the
exposure evaluation are found in Technical Appendix B. It should be noted, however, that
products of decay are not modeled. Exclusion of these decay products from the model may
underestimate or overestimate the risk impact of releases, since the decay product may be more
or less toxic than the parent compound.

Specific pathway calculations are discussed in the sections below. First, Section 5.1 discusses
the geographic basis of the model, and describes the grid cell system underlying the model and
how facilities and people are located on it. This discussion describes how annual grid cell-level
general population data sets are created. From these general population data sets, the model
generates estimates of populations exposed through particular pathways. The next sections
describe the modeling for each pathway, including the estimation of surrogate doses and exposed
population for that pathway.

5.1 Geographic Basis of the RSEI Model

Underlying the RSEI method is the ability to locate facilities geographically, and to attribute
characteristics of the physical environment, such as meteorology, to areas surrounding the
facilities, once they are located. To accomplish the location of facilities and the attribution of
data to those facilities, the model describes the U.S. and its territories as an 810m by 810m grid
system. For each cell in the grid system, a location "address" in terms of grid and (x,y)
coordinates is assigned based on latitude and longitude (lat/long).

5.1.1 The Model Grid Cell System

RSEI uses a standard Albers Equal-Area projection19 to create each of the grids that is used in the
model. The grid system is split into six individual grids which cover the continental U.S.,

Alaska, Hawaii, and the territories. Each unique cell address is composed of (1) the grid number,
and (2) the (x,y) address of the cell in that grid. Exhibit 5.1 below provides the grid number
(used in the model to identify each grid), the grid characteristics that can be used to recreate the
grid in a GIS-based system, and the bounding coordinates for each.

19 Versions 2.2.0 and earlier used a non-standard grid developed specifically for RSEI. The use of a standard
projection will make it easier for users to import RSEI data into GIS applications.

1988-2011 TRI data

Version 2.3.2
July 2013

25

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.1
RSEI Grid Characteristics



Grid Reference Characteristics

Grid
Code

Grid

Latitude
of Origin

Central
Meridian

Standard
Parallel 1

Standard
Parallel 2

Lower Left
x Coord.

(m)

Lower Left
v Coord.

(m)

Columns

Rows

14

Conterminous
U.S.

23°N

96°W

29.5°N

45.5°N

-2,365,605

251,505

5,724

3,618

24

Alaska

50°N

154°W

55°N

65°N

-1,046,115

564,975

3,291

2,505

34

Hawaii

20.5625°N

157.5625°W

19.4375°N

21.2375°N

-287,955

-185,895

739

480

44

Puerto Rico/
Virgin Islands

18°N

66.25°W

17.875°N

18.5°N

-185,895

-40,095

462

129

54

Guam/
Marianas

0°

155°E

12°N

15°N

-1,133,595

1,468,935

203

295

64

American
Samoa

0°

170°W

12°S

15°S

-91,125

-1,578,285

203

36





Bounding Coordinates for Lower-Left (LL), Upper-Right (UR),
Lower-Right (LR), and Upper-Left (UL) Corners

Grid
Code

Grid

LL Long

LL Lat

UR Long

UR Lat

LR Long

LR Lat

UL Long

UL Lat

14

Conterminous
U.S.

118.78°W

22.69°N

65.14°W

48.29°N

74.09°W

22.89°N

128.05°W

48.01°N

24

Alaska

170.07°W

53.95°N

111.99°W

68.54°N

129.76°W

52.41°N

176.63°W

71.23°N

34

Hawaii

160.29°W

18.86°N

154.54°W

22.37°N

154.6°W

18.86°N

160.36°W

22.38°N

44

Puerto Rico/
Virgin Islands

68°W

17.63°N

64.46°W

18.58°N

64.47°W

17.63°N

68.01°W

18.58°N

54

Guam/
Marianas

144.54°E

13.18°N

145.98°E

15.4°N

146.06°E

13.24°N

144.44°E

15.34°N

64

American
Samoa

170.85°W

14.38°S

169.32°W

14.12°S

169.32°W

14.38°S

170.84°W

14.12°S

The (x,y) coordinates used in each grid are defined as:

x = number of cells from the center cell in the East-West direction
y = number of cells from the center cell in the North-South direction.

Version 2.3.2	1988-2011 TBI data

July 2013	26

-------
Chapter 5: Exposure and Population Modeling

They can be calculated as follows:

x = INT (Easting / 810 + Adjust x)

(Eq. 5.1)

y = INT (Northing / 810 - Adjusty)

(Eq. 5.2)

where:

Adjustx
Adjustj

Easting

Northing

projected Albers distance East of the central meridian for the grid
(m)

projected Albers distance North of the central meridian for the grid
(m)

0.5 if Easting >0, -0.5 if Easting<0
0.5 if Northing>0, -0.5 if Northing<0

5.1.2 Locating Facilities on the Grid

Once the grid system for the U.S. is created, each facility must be located on the grid and
assigned to a grid cell. Facilities are projected onto each grid using GIS software and the (x,y)
coordinates of the cell where the facility is mapped are assigned to the facility. Once a grid
cell's (x,y) coordinates are assigned, the facility is assumed to be at the cell's center, for ease of
modeling. For a complete description of the method used to select lat/long coordinates for both
reporting facilities and off-site facilities, see Technical Appendix D.

Reporting Facilities. Because the location of a facility is key to the subsequent exposure
modeling (e.g., facility location will determine which population is assumed to be exposed to its
air releases), it is important that the lat/long coordinates are as accurate as possible. RSEI uses
the best pick coordinates from EPA's Locational Reference Tables (LRT) and EPA's Facility
Registry System (FRS), both of which collect coordinates and related documentation on location
from programs across EPA. The facility lat/long coordinates are projected onto the relevant grid,
and the (x,y) coordinates of the grid cell to which the facility maps is assigned. The facility is
then modeled as being located at the center of its assigned cell.

Off-site Facilities. RSEI also models some potential exposures that may result from
environmental releases of chemicals from "off-site" facilities, that is, facilities that receive
transfers from TRI-reporting facilities. Note that off-site facilities do not report transfers
received from other facilities directly to TRI; instead their names and addresses are reported by
the facilities that transfer chemicals to them.20 Each report of a receiving off-site facility
becomes a separate record in the TRI, even though each off-site facility often receives transfers
from more than one TRI-reporting facility. This produces multiple records of the same off-site
facility; however, because the names and addresses are not standardized, the records are usually

20 Some facilities may be considered both on-site and off-site facilities, if they both receive chemical transfers from
other facilities (as an off-site facility) and emit reportable chemical releases (as an on-site facility). RSEI does not
attempt to account for emissions that may be double-counted in this way.

Version 2.3.2	1988-2011 TRI data

July 2013	27

-------
Chapter 5: Exposure and Population Modeling

slightly different, and so cannot easily be matched to each other. EPA has developed an
approximate text string matching program to identify imperfect matches in order to collapse the
set of off-site facilities to what are considered to be unique off-site facilities. The program
matches values without requiring their exact equality. This approach accommodates misspelled
words and inconsistencies in how a facility might report its identifying information over time.
For example, "DuPont," "Du Pont," and "E.I. DuPont" might all refer to the same facility. A
possible match is identified based on similarity rather than exact equality in the name field, and
then the address fields are examined to determine whether the records match.

For RSEI Version 2.1.3 (RY 2003), all off-site records went through the approximate text
matching program, and were also geocoded (lat/long coordinates were assigned based on street
address). For each group of facilities that were determined to be matches, the record whose
geocoded lat/long was of the highest confidence level was selected. The name, address and
lat/long for this facility record is selected for the master database, and used in the model to
represent all of the records in that matched group. For Version 2.1.3, this resulted in a master
off-site database of approximately 47,000 off-site facilities.

Beginning with Version 2.1.5 (RY 2005), off-site facilities are no longer geocoded. Instead, the
entire set of all years of off-site records (1988-2011) is matched back to the previous years'
master off-site database using an approximate text matching program. Again, this is necessary
because there are no IDs assigned to off-site records in TRI that would allow for direct matching
from year to year. Any records that are not matched back to the previous years' database
(including any new off-site facilities) are added in to the master database, resulting in a new set
of approximately 50,000 records. Additional data from EPA's Facility Registry System (FRS)
are also used where possible to identify lat/longs for off-site facilities. Grid-cell addresses for
off-site facilities are determined in the same manner as for reporting facilities; the facility
lat/long coordinates are projected onto the relevant grid, and the (x,y) coordinates of the grid cell
to which the facility maps is assigned. The off-site facility is assumed to be in the center of its
assigned grid cell.

5.1.3 Locating People on the Grid

In order to estimate potential exposure, the U.S. population must also be geographically located
on the model grid. To match annual TRI emissions and capture the effect of the changing
distribution of the population, RSEI uses detailed annual population datasets at the grid cell
level. The data are based on decennial U.S. Census data, and includes information on
population, age and sex.

The following sections describe how the U.S. Census data are used to generate annual population
estimates, and how the unit of analysis for the U.S. Census (the block) is translated into the unit
of analysis for the model (the grid cell).

U.S. Census Data. The model uses decennial U.S. Census data for 1990, 2000, and 2010 at the
block level.21 Census blocks are the smallest geographic area for which decennial census data are
collected. Blocks are of varying size, formed by streets, roads, railroads, streams and other

21 Some U.S. Census data and block shape files were provided by GeoLytics, Inc.

Version 2.3.2
July 2013

28

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

bodies of water, other visible physical and cultural features, and the legal boundaries shown on
Census Bureau maps. In 1990, there were approximately 7 million census blocks. Due to
boundary changes and increased resolution for highly populated areas, there were approximately
9 million blocks in the 2000 Census, and more than 11 million in 2010. Block-level data from
the three decennial Censuses22 are used to create detailed age-sex population groups for each of
the census blocks in the U.S. for 1990, 2000 and 2010. Because the U.S. Census presents data in
slightly different format, some data processing was necessary to create the following age-sex
population groups used in the model:

•	Males Aged 0 through 9 years

•	Males Aged 10 through 17 years

•	Males Aged 18 through 44 years

•	Males Aged 45 through 64 years

•	Males Aged 65 Years and Older

•	Females Aged 0 through 9 years

•	Females Aged 10 through 17 years

•	Females Aged 18 through 44 years

•	Females Aged 45 through 64 years

•	Females Aged 65 Years and Older

For Puerto Rico, the U.S. Virgin Islands, American Samoa, Guam, and the Northern Mariana
Islands, block-level shapefiles and block-level population data were only available for 2000.23
For 1990, the grid cell-level populations from 2000 were scaled by age-sex specific Census
population estimates for 1990 to create 1990 population estimates. For 2010, block-level
population was only available for Puerto Rico; for the other areas, similar age-sex specific
Census data from 2010 were used to scale the 2000 data to create 2010 population estimates.

Grid cells for Puerto Rico and island areas are mapped in the same way as described below.

Mapping blocks to grid cells. Because the grid cell is the unit of analysis for the model, Census
data must be transposed from blocks to the model grid cells. Census provides the geometry for
each block in the Topologically Integrated Geographic Encoding and Referencing (TIGER)
geographic database, which was used to create shape files for the 1990, 2000, and 2010 Census
years. A corresponding set of shape files for grid cells was created, with each grid cell defined
by its four corner points, calculated from its (x,y) coordinates. The shape files were then
compared, in essence overlaid, and each block was mapped to the cells in the grid that it
overlaid, and the percentage of the block's total area falling within each cell was calculated.24

22	For 1990, not all of the variables were available at the block level. For those variables that were only available at
the block group level, block group ratios were calculated and applied to the data available at the block level. For
2000, all of the required variables were available at the block level.

23	For 2010, block-level shapefiles were available, but block-level population data were not released in time for
incorporation.

24	Due to irregular, invalid block shapes, some of the block percentages did not sum to 100 percent. For these
blocks, the boundary overlay process was not used; instead, the whole block was assigned to whatever grid cell

Version 2.3.2
July 2013

29

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

The process described above was performed separately for 1990, 2000, and 2010, as the block
boundaries change between the Censuses. This process resulted in three tables, each with four
fields: the Xcoordinate and the Y coordinate which identify the grid cell, the block identifier
assigned by the Census, and the percent of that block assigned to the grid cell.

Calculating Populations. For each block assigned to a grid cell, the block populations were
multiplied by the percent of that block assigned to that grid cell. Those values were then
summed over each grid cell. This process was performed separately for 1990, 2000, and 2010,
resulting in three grid cell level datasets, each containing the ten age-sex population groups listed
above.25 For 1990, there were 11,083,291 populated grid cells; for 2000, there were 9,399,819;
and for 2010 there were 9,258,679 populated grid cells.

To create annual datasets for 1991 through 1999, a straight-line interpolation at the grid cell-
level is performed within the model between the 1990 and 2000 datasets; annual datasets for
2001 through 2009 are created using a straight-line interpolation between the 2000 and 2010
datasets. The 1990-2000 line is extrapolated backward to create annual datasets for 1988 and
1989 and the 2000-10 line is extrapolated forward to 2011.

5.2 Pathway-specific Methods to Evaluate Chronic Human Exposure
Potential

The following sections describe the algorithms for modeling exposure for each of the following
exposure pathways: (1) stack and fugitive air releases, (2) direct surface water releases, (3)
transfers to POTWs, (4) off-site transfers, and (5) on-site land releases. An overview of the
pathways and methods used to evaluate each pathway is presented in Exhibit 2.1.

The following discussions of exposure modeling frequently mention concentration and surrogate
dose. This is not meant to imply that dose can be accurately calculated within this model. The
exposure algorithms are intended to be simple ways to gauge relative risks from releases to
different media in a consistent, defensible way, by modeling and estimating exposure. In some
cases, the modeling is purposely simplified, given the lack of site-specific data.

When possible, exposures are estimated for relevant subpopulations defined by age, sex, or other
factors. Exposure for individual subpopulations is modeled using exposure factors (i.e.,
inhalation rates, drinking water intakes, fish ingestion rates, and body weight) and population
data specific to such subpopulations. For example, ingestion rates specific to recreational and
subsistence fishers are used to estimate exposures for these fishers and their families. Also, age-
and sex-specific inhalation and drinking water ingestion rates are used. The relevant exposure
assumptions for these subpopulations are also described in the following sections.

contained the centroid of the block (an approximate center point defined by the Census).

25 The data processing results in fractional people; populations were rounded to four decimal places for use in
calculations, but are rounded to the nearest integer for display in the model.

Version 2.3.2
July 2013

30

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3 Modeling Air Releases

On-site air releases accounted for approximately 13 percent of TRI releases and transfers by
weight in 2011. Air releases can either be released through stacks or as fugitive releases. Stack
(or point) air releases include releases to air through stacks, confined vents, ducts, pipes, or other
confined air streams, and represent the majority of air releases (87% of on-site air releases).
Fugitive releases to air include all other on-site air releases, including leaks, evaporation from
surface impoundments, and releases for building ventilation systems. These are modeled as two
separate pathways in the model, although the potentially exposed population and human
exposure assumptions are the same for both. The following sections describe the method and
data sources for each pathway.

5.3.1 Stack Air Emissions: Method

Stack air releases are modeled using the American Meteorological Society/EPA Regulatory
Model (AERMOD). AERMOD replaced the Industrial Source Complex (ISC) model as EPA's
preferred regulatory model in 2005. AERMOD is a steady-state Gaussian plume model used to
estimate pollutant concentrations downwind of a stack or area source. The pollutant
concentration is a function of facility-specific parameters, meteorology, and chemical-specific,
first-order air decay rates. The following sections describe the parameters of the AERMOD
model used.26

5.3.1.1 AERMOD

The AERMOD model is specifically designed to support the EPA's regulatory modeling
programs, as specified in the Guideline on Air Quality Models (Revised).27 AERMOD is a
steady-state plume model. In the stable boundary layer (SBL), it assumes the concentration
distribution to be Gaussian in both the vertical and horizontal. In the convective boundary layer
(CBL), the horizontal distribution is also assumed to be Gaussian, but the vertical distribution is
described with a bi-Gaussian probability density function (pdf). Additionally, in the CBL,
AERMOD treats "plume lofting," whereby a portion of plume mass, released from a buoyant
source, rises to and remains near the top of the boundary layer before becoming mixed into the
CBL. AERMOD also tracks any plume mass that penetrates into the elevated stable layer, and
then allows it to re-enter the boundary layer when and if appropriate. For sources in both the
CBL and the SBL, AERMOD treats the enhancement of lateral dispersion resulting from plume
meander. Unlike existing regulatory models, AERMOD accounts for the vertical inhomogeneity
of the planetary boundary layer (PBL) in its dispersion calculations. This is accomplished by
averaging the parameters of the actual PBL into effective parameters of an equivalent
homogeneous PBL.

26	The following description is based on equations and text provided in the AERMOD manuals and documentation.
The most recent AERMOD manuals are available from EPA's SCRAM website at
http://www.epa.gOv/scram001/dispersion_prefrec.htm#aermod.

27	The Guideline on Air Quality Models can be found in 40CFR Part 51, Appendix W or accessed online at:
http://www.epa.gov/ttn/scram/guidance/guide/appw_05.pdf.

Version 2.3.2
July 2013

31

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3.1.2	Model Dispersion Options

AERMOD is used with its regulatory default options28, except for the following: chemical-
specific decay is considered (the TOXICS with SCIM option is used), and flat terrain is assumed.
The non-default option of modeling urban areas with increased surface heating is not used.
Weather data from National Weather Service (NWS) observation stations is used as the
meteorological input (see Section 5.3.1.4 below).

5.3.1.3	Source Parameters

In the RSEI model, the U.S.29 is represented by a grid system composed of 810m by 810m square
grid cells. Facilities are assigned to a particular grid cell in this grid according to their latitude
and longitude coordinates (see Technical Appendix D for details on the coordinates used). To
increase modeling efficiency, a facility is then assumed to be located at the center of the grid
cell, regardless of where its latitude and longitude coordinates place it within the cell

As a result of this assumption, the actual location of a facility may differ from its modeled
location by up to 573 meters, the maximum distance between the center and the corner of the
cell. To simplify the analysis, a facility's point source emissions are modeled as a single stack
located at the facility's geographic center.

RSEI uses facility-specific stack parameters when available. These include stack height, exit-gas
velocity, and stack diameter. Stack exit gas temperature is assumed constant for all stacks (432°
K). For facilities with multiple stacks, the median value for the stack heights and diameters for
that facility is used. For facilities without stack-specific values, a Standard Industrial
Classification (SIC) code-based median stack parameter is assigned to the facility. If no valid
SIC code is available, or no stack data are available for that SIC code, then overall median values
are used. Stack parameters are discussed further in Section 5.3.6.1 and in Technical Appendix E.

Annual stack air releases as reported to TRI are converted to an equivalent constant emission rate
in grams per second according to the following equation:30

453.6 q	(Eq. 5.3)

31,536,000

where:

Q	= pollutant emission rate (g/sec)

q	= TRI annual stack or point air emissions (lb/yr)

453.6	= constant to convert pounds (lb) to grams (g)

31,536,000 = constant to convert years (yr) to seconds (sec) assuming 365 days
per year

28	See AERMOD User's Guide (EPA, 2007b), http://www.epa.gov/scram001/7thconf/aermod/aermod_userguide.zip

29	Including Puerto Rico, the U.S. Virgin Islands, American Samoa, Guam, and the Northern Mariana Islands.

30	Although RSEI can model any chemical air emission that is accompanied by the appropriate locational, chemical,
and toxicity weight information, the model currently uses TRI reporting as the source of chemical release
information.

Version 2.3.2
July 2013

32

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3.1.4	Meteorological Input Data

For a given pollutant source, meteorology around the source affects the dispersion
characteristics. Meteorological factors such as wind speed and direction, air temperature,
stability, turbulence and the height of the mixing layer all have a direct effect on the dispersion
and dilution of air pollution and the resulting magnitude and location of ground level
concentrations of emitted pollutants.

AERMOD is designed to run with a minimum of observed meteorological parameters, and
requires only a single surface measurement of wind, wind direction and ambient temperature.
Like ISC, AERMOD also needs observed cloud cover. However, if cloud cover is not available
(e.g. from an on-site monitoring program) two vertical measurements of temperature (typically at
2 and 10 meters), and a measurement of solar radiation can be substituted. A full morning upper
air sounding is required in order to calculate the convective mixing height throughout the day.
Surface characteristics (surface roughness, Bowen ratio, and albedo) are also needed in order to
construct similarity profiles of the relevant PBL parameters.

5.3.1.5	Calculating Pollutant Concentration

RSEI calculates air concentrations at hypothetical "receptors" located within a circle with a
radius of 49 km surrounding each facility. Any cells with centers within the 49 km circle are
included. The model calculates ground-level concentrations for receptors at 5 kilometer
increments for distances from 5 to 49 km away from the modeled facility. The concentration
assigned to a grid cell containing a receptor is determined at the point in that cell which is nearest
to the facility (see Exhibit 5.2).

For grid cells between the receptor points where calculations are made, concentrations are
interpolated using a spatial weighting technique similar to inverse distance-weighted averaging.

Version 2.3.2
July 2013

33

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Modeled Concentration Used
for Ceil #2











..

Cell #1

/

Cell #2











'k Stack Location (Modeled)

Exhibit 5.2

Assigned concentration at point nearest to facility.

To estimate the concentration for the 810m by 810m center cell containing the facility, the model
calculates the interpolated concentration based on calculated concentrations at 441 receptor
points within a kilometer of the cell center. Analyses performed during the development of
RSEI Version 2.0 (which used the ISC model31) indicated that using the concentration in a
surrounding cell as an estimate for the center cell may either significantly over or under-
represent chemical concentrations there (See Part B of Analyses Performedfor the Risk-
Screening Environmental Indicators).

The model estimates concentrations up to 49 km from the facility. To determine the optimal
distance, EPA modeled air concentrations for the 20 most toxic carcinogens and the 20 most
toxic non-carcinogens included the model at various stack heights. These analyses indicated that
extending modeled distances to 50 km32 was necessary to capture potential concentrations of
concern under certain atmospheric conditions. This distance is expected to capture the majority
of the potential impacts from the TRI facilities, including electric utilities, which usually have
taller stacks than other facilities. Details of these analyses can be found in Part B of Analyses
Performedfor the Risk-Screening Environmental Indicators.33

5.3.2 Fugitive Air Releases: Method

As for stack air releases, long-term pollutant concentrations downwind of the facility due to TRI-
reported fugitive air releases are modeled using algorithms from AERMOD.

31	Version 3 of the long-term ISC model (ISCLT3) was used to calculate chemical concentrations in air until Version
2.2.0, the first version in which AERMOD was used.

32	In the final RSEI modeling, 49 km was used instead of 50km due to modeling constraints.

33	These analyses were performed using an earlier version of RSEI that incorporated EPA's Industrial Source
Complex (ISC) model. RSEI now uses AERMOD, which has replaced ISC as the Agency's recommended air
modeling program.

Version 2.3.2
July 2013

34

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3.2.1	Model Dispersion Options

Model dispersion options used in modeling fugitive air releases are the same as those used for
stack air releases, as described in Section 5.3.1.2.

5.3.2.2	Source Options

Fugitive emissions are modeled as an area source which is 10 meters by 10 meters in size,
located at the center of the cell containing the facility. The model assumes a release height of 3
meters.

Fugitive emissions are converted from pounds per year to grams per square meter per second
(g/m2s) according to the following equation:

Q=	453-6q«		(Eq. 5.4)

fl 31,536,000* 102

where:

Qa	= pollutant area emission rate (g/m2s)

qa	= TRI annual fugitive air emissions (lb/yr)

453.6	= constant to convert pounds (lb) to grams (g)

31,536,000 = constant to convert years (yr) to seconds (sec)
102	= conversion factor necessary to convert annual emissions (g/s) to

area emission rate (g/m2s), assuming an area 10 m x 10 m.

5.3.2.3 Calculating Pollutant Concentration

Pollutant concentrations fugitive air releases are calculated using AERMOD, as described above
for stack air releases.

5.3.3 Calculating Surrogate Dose for Air Releases

The calculated air concentrations described earlier are combined with assumptions regarding
inhalation rate and human body weight to arrive at a surrogate dose for a given cell:

DOSE air = Cmr*Imr * ——	(Eq. 5.5)

BW 1000

where:

DOSEair	=	surrogate dose of contaminant from air (mg/kg-day)

Cair	=	air concentration in cell ([j,g/m3)

I air	=	inhalation rate (m3/day)

BW	=	human body weight (kg)

1000	=	constant to convert ([j,g) to (mg)

Version 2.3.2
July 2013

35

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3.4	Estimating Population Size for Air Releases

The population potentially exposed to air releases is assumed to be equal to the population
assigned to the grid cells in the 810m by 810m modeled area, as described above in Section
5.1.3. Exposed population is only considered for grid cells with nonzero pollutant
concentrations.

5.3.5	Calculating an Indicator Element for Air Releases

Exhibit 5.3 provides a graphical overview of the steps for determining the air modeling
component of the model. First, the pollutant concentration in each cell is calculated using TRI
emissions data and the AERMOD algorithms. Then, subpopulation-specific exposure factors are
used to calculate a surrogate dose for each cell. Finally, the surrogate dose is multiplied by the
number of people in each subpopulation in the cell and by the chemical toxicity weight to obtain
an Indicator Element for the grid cell. Then the results for all grid cells are summed. The result
is an Indicator Element for an air release. To calculate the overall Indicator Element for all air
releases, the same steps are followed for each air release, and the results are summed.

Version 2.3.2
July 2013

36

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.3

Calculating the Indicator Element for Air Releases

Air Release
(Ibs/yr)

EPA/AMS Regulatory Model
(AERMOD) algorithm

1

Chemical concentration
in grid cell X
(|ig/m3)

Subpopulation-specific
exposure factors

	i	

Surrogate Dose
(mg/kg-day)

Population data
and

toxicity weights

y

f

Indicator element for
grid cell X





Sum over all 11,289 grid cells
around facility

	~	

Indicator Element for Air Release

Version 2.3.2
July 2013

37

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.3.6 Stack and Fugitive Air Releases: Data

The air pathways use facility-specific values (stack height, stack diameter and exit gas velocity),
meteorology, chemical-specific first-order air decay rates, and exposure assumptions (inhalation
rate and body weight). The values used for these pathways are summarized in Exhibit 5.4.

Exhibit 5.4
Air Modeling Parameters

Parameter

Value

Source/Comment

Pollution emission
rate

Site-specific (lbs/yr)

TRI

Stack height

Varies based on the facility and the availability of
information. If available, the median height of all
stacks at the facility is used. Otherwise one of the
following is used in declining order of preference:
the median stack height for the facility's 3-digit SIC
code, the median stack height for the facility's 2-
digit SIC-code, or the median stack height for all
TRI-reportable SIC codes.*

AFS, NET, NEI,
EPRI (for Electric
Utilities) and
databases for CA,
NY, and WI; these
data are used in the
vertical term of the
model

Stack diameters

Varies based on the facility and the availability of
information. If available, the median diameter of all
stacks at the facility is used. Otherwise one of the
following is used in declining order of preference:
the median stack diameter for the facility's 3-digit
SIC code, the median stack diameter for the
facility's 2-digit SIC-code, or the median stack
diameter for all TRI-reportable SIC codes.

AFS, NET, NEI,
EPRI (for Electric
Utilities) and
databases for CA,
NY, and WI

Exit gas velocity

Varies based on the facility and the availability of
information. If available, the median exit gas
velocity of all stacks at the facility is used.
Otherwise one of the following is used in declining
order of preference: the median exit gas velocity
for the facility's 3-digit SIC code, the median exit
gas velocity for the facility's 2-digit SIC-code, or
the median exit gas velocity for all TRI-reportable
SIC codes.

AFS, NET, NEI,
EPRI (for Electric
Utilities) and
databases for CA,
NY, and WI

Exit gas
temperature

432° K

Based on EPA
(2004b)

Meteorological
data

Site-specific

Processed using
AERMET as
contained in the
HEM-3 data library
(EPA, 2007)

Version 2.3.2
July 2013

38

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.4
Air Modeling Parameters

Parameter

Value

Source/Comment

Decay rate

Pollutant-specific values account for removal by
physical and chemical processes (s1)

SRC (1994-1999)

Area source size

10 m2

Based on EPA
(1992b)

Area source height

3 m

Based on EPA
(1992b)

* As of RY2006, TRI requires facilities to report NAICS codes instead of SIC codes. RSEI maintains historically
reported SIC codes. For new reporters without SIC codes, NAICS codes are matched to their equivalent SIC codes
where possible, using the Census crosswalk available at http://www.census.gov/epcd/naics02/NAICS02toSIC87.xls.

5.3.6.1 Stack Height, Stack Diameter, and Exit Gas Velocity

Stack parameter data (height, diameter, and exit-gas velocity) were obtained from the AIRS
Facility Subsystem (AFS) within the Aerometric Information Retrieval System (AIRS), the
National Emission Trends (NET) Database, the National Emissions Inventory (NEI), and
databases from three individual states (California, New York, and Wisconsin). For each TRI
facility that had stack parameter data in one or more of these sources, the median parameter of all
stacks at the facility is used. For the TRI facilities that had no stack parameter data in these
sources, the median parameter values for all of the facilities in that facility's Standard Industrial
Classification (SIC) code is used instead. The SIC code-based stack parameters are estimated
from data in AFS and the NET Database for facilities in the appropriate 3-digit SIC code, or in
the 2-digit SIC code if the 3-digit SIC code is unavailable. If no 2-digit SIC code is available,
the median of all stack parameters with TRI-reportable SIC codes is used.

The Electric Power Research Institute (EPRI) provided EPA with site-specific data for electric
utilities (electric utility SIC codes were added to TRI for Reporting Year 1998), transmitted in
two databases. These data included stack height, stack diameter, and exit-gas velocity. Of the
approximately 600 TRI facilities classified in one of the three electric utility SIC codes (4911-
Electric Services; 4931- Electric and Other Services Combined; or 4939- Combination Utilities,
not elsewhere classified) in RY 1998, almost 70 percent matched a corresponding facility listed
in one of the EPRI databases; approximately 30 percent of TRI electric utility facilities did not.34
For the 30 percent that did not match specific facilities, the median parameters taken across all of
the coal or oil combusting stacks in the EPRI databases were used.

Statistical analysis of stack heights by SIC code revealed that, for certain SIC codes, no
significant height differences existed between stacks emitting TRI chemicals and those not
emitting TRI chemicals. For these SIC codes, the median stack height was based on stack

34 One TRI facility classified in one of the relevant utility SIC codes that could not be matched to a specific facility
in the EPRI dataset was matched to a specific facility in the AFS database. In this case, the facility-specific
parameters were taken from AFS.

Version 2.3.2
July 2013

39

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

heights for all facilities. For SIC codes in which there were significant height differences
between stacks emitting or not emitting TRI chemicals, only those stacks emitting TRI chemicals
were used to calculate the median stack height for that SIC code.

For both stack diameter and exit gas velocity, the model uses the same data sources, criteria, and
statistical methods described above for stack height data. Specifically, the model uses either the
median value of all stacks for TRI facilities with this information or an SIC code-based median
value for facilities without the appropriate stack data. Exit gas velocity data were obtained from
AFS, NET, NEI, and state-specific databases. Stack diameter data were obtained from AFS,
NET, NEI, EPRI and databases from three individual states (California, New York, and
Wisconsin).

Analyses have been conducted that show air concentrations predicted by the model using a
combination of generic and site-specific data closely match concentrations estimated by using
more complete site-specific data.35 See Technical Appendix E for details on the derivation of
stack data.

5.3.6.2 Meteorology36

The meteorological data used in RSEI are taken from EPA's Human Exposure Model, Version 3
(HEM-3), a model for use in site-specific air toxics risk-assessment. RSEI uses weather data
included in EPA's HEM-3 data library, which has been prepared using AERMOD's
meteorological processor, AERMET. AERMET requires hourly surface weather observations
and the full twice-daily upper air soundings (i.e., meteorological variables reported at all levels).
The surface and upper air stations are paired to produce the data files require for input into
AERMOD: one file consists of surface scalar parameters, and the other file consists of vertical
profiles of meteorological data.

To simplify processing and minimize the amount of quality assurance needed, HEM-3 's
processing was restricted to meteorological data collected prior to the installation of the
Automated Surface Observation System (ASOS). The ASOS has previously been found to omit
the ceiling height for a large percentage of the observations at a number of meteorological
stations. Installation and operation of ASOS equipment began in 1992; therefore, data for 1991
were used. Data were retrieved from products available from the National Climatic Data Center
(NCDC). The surface data for 1991 were retrieved from the Hourly United States Weather
Observation (HUSWO) CD. Upper air soundings were obtained from the Radiosonde Data of
North America CDs produced by NCDC and the Forecast Systems Laboratory (FSL).

Certain surface characteristics must be specified when processing meteorological data using
AERMET, including the surface roughness length, the Bowen ratio (an indicator of surface
moisture), and the albedo (an indicator of surface reflectivity). These surface characteristics are

35	These analyses were performed using an earlier version of RSEI that incorporated EPA's Industrial Source
Complex (ISC) model. RSEI now uses AERMOD, which has replaced ISC as the Agency's recommended air
modeling program.

36	This description is taken from the HEM-3 User's Manual (EPA, 2007), available at
http://www.epa.gov/ttn/fera/data/hem/hem3_users_guide.pdf.

Version 2.3.2
July 2013

40

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

used by AERMET to calculate the level of shear-induced mechanical turbulence generated by
flow over the surface and for the energy balance calculations used in the determination of the
Monin-Obukhov stability parameter and the convective velocity scale. For the HEM-3
meteorological data, the following surface characteristics were used:

•	Surface roughness length = 0.25 m. At the airport meteorological site, the surface
roughness includes runways, terminal buildings and other airport structures. In
addition, off-airport structures often are within 3 kilometers of the measurement
site. This combination of land covers suggests a value of 0.2 - 0.3 meters is
appropriate.

•	Bowen ratio =1.0. Representing an equal partition of the heat fluxes Albedo =

0.15. Representing conditions for all seasons, including winter without continuous
snow cover.

•	The file STNS.TXT located on the HUSWO CD was used for the anemometer
heights required by AERMET. These heights are to the nearest meter and were
deemed appropriate for use in this application.

5.3.6.3 First-Order Air Decay Rates

Pollutants may be removed from the atmosphere by either physical processes or chemical
transformation. The model uses pollutant-specific air decay rates from SRC's Atmospheric
Oxidation Program (AOPWIN), an atmospheric oxidation computer program (SRC, 1994-1998).
AOPWIN estimates the second-order rate constant for the atmospheric, gas-phase reaction
between photochemically produced hydroxyl radicals and organic chemicals.37 The daughter
products of photodegradation are not modeled further, i.e., it is assumed that all chemicals are
photodegraded into nontoxic compounds. AOPWIN data also contains certain empirically
derived air decay rates. For the model, a concentration of hydroxyl radicals of 1.5 x 106
molecules/cm3 is used to convert the second-order rate constant provided in AOPWIN to a first-
order rate constant. Furthermore, the rate is divided by a factor of two to reflect an assumed
average day length of 12 hours:

AOPWIN

Katr =			*1.5xio6*3600	(Eq. 5.6)

where:

Kair	=	air decay rate (hr"1'

AOPWIN	=	second-order rate constant from AOPWIN

1.5xl06	=	hydroxyl radical concentration (molecules/cm3)

3600	=	constant to convert molecules/seconds to molecules/hour

2	=	constant to reflect assumed day length of 12 hours

37 For a few chemicals, other sources were used. See Technical Appendix B for the source used for each chemical.

Version 2.3.2	1988-2011 TRI data

July 2013	41

-------
Chapter 5: Exposure and Population Modeling

5.3.6.4 Human Exposure Data and Assumptions

For the air pathways, sex- and age-specific inhalation rates and body weights are used in the
model. The primary source for all exposure factors used in the model is EPA's Exposure Factors
Handbook (EPA, 2011, hereafter denoted as EFH), which provides a summary of the available
statistical data on various factors used in assessing human exposure. These factors include:
drinking water consumption, soil ingestion, inhalation rates, dermal factors including skin area
and soil adherence factors, consumption of fruits and vegetables, fish, meats, dairy products,
homegrown foods, breast milk intake, human activity factors, consumer product use, and
residential characteristics. In EFH, EPA recommends mean values for the general population
and also for various segments of the population who may have characteristics different from the
general population. RSEI uses inhalation rates and body weights derived from the recommended
factors, except where noted. The Exposure Factors Handbook can be found on the internet at
http://www.epa.gov/ncea/efh/pdfs/efh-complete.pdf.

EFH (EPA, 1997b, Table 5-23, p. 5-24) was used to estimate inhalation rates for the eight RSEI
age-sex groups (ages 0-17, 18-44, 45-64, 65+).38 The inhalation rates recommended by EFH
were not categorized into the same age groups used in RSEI. For children, the RSEI age groups
were broader than the EFH age groups. Therefore, the exposure factor was calculated using a
weighted average of the inhalation rates for all EFH age groups that overlap the RSEI age group
as follows:

* "/)

EF = -i		(Eq. 5.7)

N

where:

EF	= RSEI exposure factor,

IR,	= intake rate for EFH age group z,

zz,	= number of years that EFH age group z overlaps with the RSEI age

group

N	= number of years in RSEI age group.

For adults, EFH provides only one range of recommended inhalation rate for males and females.
The RSEI adult inhalation factors are based on weighted averages calculated from this data,
using Equation 5.15. The RSEI inhalation factors are then divided by age- and sex-specific body
weights, averaged to match the RSEI age groups using data provided in EFH (EPA, 2011, Table
6-1). Exhibit 5.5 provides the range of data used, and Exhibit 5.6 presents the final exposure
factors used in the model. More detail on the derivation of exposure factors can be found in
Technical Appendix C.

38 The RSEI model provides the ability to view risk scores for the 0-9 year old age group, however, there is not a
separate exposure factor for this age group. Instead, the exposure factor for the 0-17 year old age group is used.

Version 2.3.2
July 2013

42

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.5

Range of Data Used to Estimate Exposure Factors

Parameter

Value

Source/Comment

Inhalation rate

3.6-16.3 m3/day
(Varies by age)

EPA (2011)

Body weight

4.6 -90.5 kg
(Varies by age and sex)

EPA (2011)

Exhibit 5.6
Inhalation Exposure Factors (m3/kg-day)

Model Age
Group

Male

Female

0-17

0.315

0.332

18-44

0.185

0.217

45-64

0.173

0.201

>65

0.159

0.187

5.4 Modeling Surface Water Releases

In 2011, approximately three percent of TRI releases and transfers by weight were released on-
site as direct surface water releases. People may be exposed to chemicals released into surface
water in one of two ways: by drinking tap water from a public water system whose water intake
was located in the stream path of a chemical release; or by eating contaminated fish caught in a
water body in the stream path of a chemical release. The following sections first describe the
methods used to calculate the initial stream concentration for both pathways, and then the
different methods used to calculate surrogate dose and population for the drinking water pathway
and the fish ingestion pathway. The data section presents the data used for both pathways and
the human exposure assumptions used.

5.4.1 Surface Water Releases: Methods

5.4.1.1 Locating the Facility Discharge Reach

The first step in assessing surface water emissions is to locate the discharging facility on the
model grid using the lat/long. Facilities are then matched to a receiving stream reach (a linear,
unbranched section of a water body).

The main data source used is EPA's records of discharge permits for facilities, which in most
cases specify the discharge reach by name or reach number. Permit records, including permits,

Version 2.3.2
July 2013

43

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

monitoring data, and locational and descriptive information pertaining to more than 67,000
regulated are maintained in EPA's Integrated Compliance Information System-National
Pollutant Discharge Elimination System (ICIS-NPDES).

Facilities without assigned reaches are assumed to discharge to the nearest reach, as long as that
reach is within four kilometers of the facility and meets minimum criteria for flow and reach
type, as described below in Section 5.4.3.1. If no assigned reach is available, and no acceptable
reach is found within four kilometers, the discharge is not modeled. Reach data are not available
for Alaska, Guam, American Samoa and the Northern Mariana Islands; therefore, no surface
water releases are modeled for these areas.

5.4.1.2 Calculating Pollutant Concentrations

Chemical concentrations in the receiving stream at a distance x from the discharging facility at
time t are estimated by using a simple first-order decay equation. The facility is assumed to
release its annual discharge at a constant rate throughout the year. Annual average
concentrations are then estimated until one of three conditions occurs: (1) the release has traveled
300 km downstream; (2) the release has been traveling downstream for a week; or (3) the
concentration reaches 1 x 10"9 mg/L. Within the initial stream reach, the mass of the release is
assumed to be instantaneously mixed with the flow at the upstream end of that reach. The
calculated concentration at the downstream end of the reach is then converted back to a mass
(after any decay) and the process is repeated in the adjoining reach. Reaches are defined by
intersections with other hydrological features and these "nodes" initiate the next reach segment.
The chemical-specific decay coefficient is predominantly based on abiotic hydrolysis or
microbial biodegradation, but may also include photooxidation. The general form of the first-
order decay equation is as follows:

Cx = Co e

~k H

(Eq. 5.8)

where:

Cx
Co

kwater

t

concentration at distance x meters from the facility release point
(mg/L)

initial concentration (mg/L), which equals chemical release
(mg/day) divided by mean flow
decay coefficient (sec"1)

time at which Cx occurs (sec), which equals x/u, where u is the
water velocity (m/sec)

For surface water releases, the RSEI model estimates chronic human health exposures for two
pathways: drinking contaminated water and eating non-commercial contaminated fish. Methods
used to estimate each of these exposures are described below.

5.4.1.3 Modeling the Drinking Water Pathway

Surrogate doses from drinking water are calculated using the chemical concentrations in stream
reaches where drinking water intakes are located. Drinking water intake locations were obtained

Version 2.3.2
July 2013

44

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

from the Public Supply Database, a database of drinking water system information developed
and maintained by the U.S. Geological Survey (USGS), based on information in the Safe
Drinking Water Information System (SDWIS). Each intake is assumed to be drawing water
from the stream reach nearest to its plotted location, as long as the reach is within one
kilometer.39 For this exposure pathway, the chemical concentration in drinking water is assumed
to be equal to the stream concentration (calculated at the upstream end of the reach;
conservatively using the highest concentration), up to the level of the Maximum Contaminant
Level (MCL),40 where applicable. (Seventy-nine TRI chemicals had existing MCLs in effect
during one or more years for which TRI data are available;41 this number includes chemical
categories that are treated as their elemental forms). If the stream concentration exceeds the
MCL, the drinking water is assumed to be treated to the level of the MCL for the year of that
release. For each stream reach with a drinking water intake, the chemical concentration is
combined with standard exposure parameters (see Section 5.4.3.6) to yield a surrogate dose:

C	* T

\ s-y 7—¦ 	 water, reach 1 water	s-rj ^

DOSEdw	—		(Lq. s.y)

BW

where:

DOSEdw

C water, reach
Iwater

BW

surrogate dose of chemical in drinking water (mg/kg-day)
average annual chemical concentration in the reach of interest,
calculated at the upstream end of the reach (mg/L)
drinking water ingestion rate (L/day)
human body weight (kg)

5.4.1.4 Estimating Population Size for the Drinking Water Pathway

To estimate the size of the population exposed to TRI releases through drinking water, the model
uses estimates of the population served by each drinking water intake from USGS's Public
Supply Database, which incorporates population estimates from the Safe Drinking Water
Information System (SDWIS)42. (More information about SDWIS can be found at
http://www.epa.gov/enviro/html/sdwis/.) However, this data set only lists the intake location and
the number of people served by the water system. In many cases, there are multiple water

39	One intake, for the Los Angeles Dept. of Water and Power is not modeled, even though there is a reach within one
kilometer (the L. A. River). Although it is clear that this is not the correct source, the actual reach is not known, so
the intake is not currently modeled.

40	Copper and lead have action levels instead of MCLs; however, RSEI models them in the same manner as MCLs.
This also applies to copper compounds and lead compounds, as metal compounds are modeled like their elemental
forms.

41	As MCLs are sometimes revised and new ones have been added since TRI began reporting, RSEI applies MCL
limits for only the years that the MCLs were in effect. For several chemicals for which MCLS were first instituted
in 1976 and then revised in 1991, the original MCL values were not readily available, so the revised values were
also used for the years before the revision. These chemicals are barium, cadmium, chromium, lead, lindane,
mercury, methoxychlor, nitrate, selenium, and toxaphene.

42	RSEI uses SDWIS data that is contained in the USGS Public Supply Database (PSDB), see Section 5.4.3.2.

Version 2.3.2
July 2013

45

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

intakes per water system. In the absence of other data, it is assumed that the total population of
the water system is exposed to the full concentration of the released chemical estimated at the
reach where a water intake is located (calculated at the upstream end of the reach).

The drinking water intake information from SDWIS contains only the number of people served
by each drinking water system; it does not provide demographic or locational information for
those served (the time frame in which this information was collected also varies widely). To
derive demographic information (that is, age and sex breakdowns) for the population served,
RSEI uses the percentage of people in each of the ten age-sex categories for the total population
located in grid cells within an 80-km radius of each reach containing a drinking water intake (this
information is calculated for the fish ingestion pathway - see Section 5.4.1.4). Then, these
percentages are applied to the SDWIS intake population (population served), creating the
subpopulation groups that are used for calculating results.

5.4.1.5 Modeling the Fish Ingestion Pathway

A second potential exposure pathway is through consumption of fish contaminated by chemicals
discharged from TRI reporting facilities. These fish may be consumed by recreational and
subsistence fishers and their families.43 As in the drinking water pathway, chemical
concentrations are calculated until one of three conditions occurs: (1) the release has traveled 300
km downstream; (2) the release has been traveling downstream for a week; or (3) the
concentration reaches 1 x 10"9 mg/L. The chemical concentration in fish is estimated using the
following equation:

where:

C fishreach C water, reach BCF

(Eq. 5.10)

Cfish, reach
Cwater, reach

BCF

concentration in fish in the specified stream reach (mg/kg)
average annual chemical concentration in the reach of interest
(mg/L)

bioconcentration factor for chemical (L/kg)

The chemical concentration in fish in a reach is combined with exposure assumptions to
determine the surrogate dose from this pathway:

D0SEfc =	(Eq- 5.H)

BW

where:

DOSEfc = surrogate dose of chemical c from facility/(mg/kg-day)

Cfish, reach = average annual chemical concentration in fish tissue (mg/kg)

43 Although store-bought fish may also contain pollutants originating from TRI facilities, modeling this exposure
pathway is not currently possible.

Version 2.3.2
July 2013

46

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Ifish.pop	= fish ingestion rate for recreational or subsistence fishers (kg/day)

BW	= human body weight (kg)

5.4.1.6 Estimating Population for the Fish Ingestion Pathway

The model uses several steps to estimate the population within each grid cell that consumes non-
commercial fish. First, a county-level dataset containing the number of fishing or
hunting/fishing combination licenses was created from state fish and wildlife licensing data for
1996 (if 1996 was not available, 1997 was used). The number of fishing licenses in a county is
then divided by the 1990 total population in the county.44 The resulting ratio is multiplied by the
population in each grid cell in 2000 to obtain the number of individuals with fishing licenses
within that cell. To account for family members who also eat fish caught by one member, the
model multiplies the number of fishers by 2.62, the size of the average U.S. household in 1995
(U.S. Census Bureau, 1996). The total population in a grid cell consuming non-commercial fish
is described by the following equation:

T 1C£flS£ s

FishPop cell = TotalPop cell * ——	* FamSize	(Eq. 5.12)

where:

FishPopceii = total fish-eating population in a grid cell in 2000
TotalPopceii = total resident population in a cell (see Section 5.1.3)

Licenses = number of fishing licenses in the county or state
Pop	= total population in the county or state in 1990

FamSize = average family size

Next, the population that consumes fish is then apportioned based on whether fish are eaten
recreationally or for subsistence. Recreational fishers may fish during only certain times of the
year for recreational purposes or to supplement their diet. In contrast, subsistence fishers may
fish throughout the year and a major part of their diets may consist of fish they catch. Data are
lacking on numbers of recreational compared to subsistence fishers; RSEI follows guidance from
EPA's Office of Water (Harrigan, 2000). The model assumes that of the population that eats
non-commercial fish, 95 percent eat fish on a recreational basis, and the remaining 5 percent
subsist on fish. This apportionment is described by the following relationships:

RecPopcell = FishPopcell * 0.95	(Eq. 5.13)

SubsistPop cell = FishPop cell* 0.05	(Eq. 5.14)

44 If no licensing information for a county was available, all of the grid cells in that county are assigned the ratio of
total state licenses to total state population. If no information was available for the state in which the grid cell is
located, ratio for the state closest to that grid cell is assigned.

Version 2.3.2
July 2013

47

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

where:

RecPopceu
SubsistPopcell

recreational fishers (and families) in a cell
subsistence fishers (and families) in a cell

The fishing population in each cell is then assigned to specific stream reaches where they are
presumed to catch fish. This is done in two steps. First, overlapping circles of 80-km radii
associated with each of the two to seven points that describe individual stream reaches are used
to define those grid cells that will be modeled for fishing population in the 48 contiguous states
(i.e., all fishing areas within 80 km of all stream reaches). The distance of 80 kilometers (50
miles) from the reach is chosen based on a finding reported in the 1991 National Survey of
Fishing, Hunting, and Wildlife-Associated Recreation that 65 percent of anglers travel less than
50 miles to fish (U.S. Department of the Interior, 1993). This distance approximates the size of
many counties and corresponds with the use of county-level fishing license data.

Second, all reaches within an 80-km radius of the center of each grid cell from the first selection
are identified. The fish-eating population in the grid cell is apportioned to each surrounding
stream reach based on the ratio of the length of that reach to the total reach kilometers within 80
km of the cell. For example, Reach A and B may be located within 80 km of a given cell. If
Reach A is 15 km in length and Reach B is 5 km in length (and the entire length of each reach is
completely within 80 km of the cell), then a total of 20 km of stream reaches are located within
the specified distance. Because Reach A represents three-fourths (15/20) and Reach B represents
one-fourth (5/20) of total km, the model therefore assumes that three-fourths of the fishing
population in the cell catches fish from Reach A and one-fourth catches fish from Reach B. Note
that the model uses only the portion of the reach's length that is within 80 km of the cell.

Because of the size of the database created, the fishing population data attributed to individual
reaches is summed and stored at the reach level. The percentage of people in each of the ten age-
sex categories for the aggregated total fishing population (reflecting the ratio of the various age
and gender subpopulations in the neighboring grid cells) is also maintained for each reach. The
model then matches the chemical concentration in fish in the appropriate reach (CfIsh,reach) to the
correctly-apportioned population. This is done for all reaches that have modeled chemical
concentrations.

5.4.2 Calculating the Indicator Element for Surface Water

The Indicator Elements for drinking water and fish ingestion are calculated by generating for
each unique combination of chemical release, stream reach, and exposure pathway a surrogate
dose, then multiplying this dose by the toxicity weight of the chemical released and the estimated
population. The RSEI risk-related result (score) for a surface water release from a TRI facility is
calculated by adding the drinking water element and the fish consumption elements (recreational
and subsistence fishing) for each reach and then summing over all reaches affected by the release
(up to 200 km downstream from the facility). Exhibit 5.7 shows the approach for calculating the
three Indicator Elements for surface water.

Version 2.3.2	1988-2011 TRI data

July 2013	48

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.7

Calculating the Indicator Elements for Surface Water

Direct Water Release
(Ibs/yr)

Water volume and velocity
estimates; decay equation

Population Data1
and

Toxicity Weights

Population Data2
and

Toxicity Weights

Sum over all reaches and both pathways

Indicator Elements for
Surface Water Release

1	Estimated using fish license data, household size, and distance traveled to fish

2	Served by drinking water intakes in Reach X (if any)

Version 2.3.2
July 2013

49

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.4.3 Surface Water Releases: Data

A variety of data are required to estimate exposure to chemical discharges to surface waters. The
parameters required for surface water modeling and the data sources are described below and
listed in Exhibit 5.8

Exhibit 5.8
Surface Water Modeling Parameters

Parameter

Value (Units)

Source/Comment

Stream reach location

Lat/long in decimal degrees

NHDPlus (U.S. EPA/USGS,
2009)

Drinking water location and
population served

Lat/long in decimal degrees
Number of persons

Public Supply Database
(2012), based on SDWIS

Water flow

mean flow (million L /day)

NHDPlus (U.S. EPA/USGS,
2009)

Decay rate of chemical in water

chemical-specific (sec-1)

SRC (1994-99)

Chemical concentration in stream

(mg/L)

calculated

Bioconcentration factor

chemical-specific (L/kg)

SRC (1994-99);
Lyman et al. (1990);
EPA (1999b)

Fish tissue concentration

calculated (mg/kg)



Family size

2.62

U.S. Census Bureau (1996)

5.4.3.1 Stream Reaches

Each facility is matched to an EPA-assigned discharge reach, or if no assigned discharge reach is
available, it is assumed to discharge into the nearest stream reach within four kilometers of the
facility. Certain minimum criteria regarding flow and reach type are applied to the set of
potential discharge reaches, as explained below. If no acceptable reach is found within four
kilometers, then the discharge is not modeled. The stream reaches used in the model are linear
sections of streams, lakes, reservoirs, and estuaries that are linked to form a skeletal structure
representing the branching patterns of surface water drainage systems. Non-transport reaches
(i.e., those that do not have an upstream or downstream connection) are excluded from the
model.

RSEI has adopted the National Hydrography Dataset (NHD) system for indexing stream reaches,
which is much more detailed than the Reach File 1 (RF1) system, which was used in RSEI
Version 2.2.0 and before. The NHD is a feature-based database that interconnects and uniquely
identifies the stream segments or reaches that comprise the nation's surface water drainage
system. The NHD provides a national framework for assigning reach addresses to water-related

Version 2.3.2
July 2013

50

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

entities such as industrial dischargers, drinking water supplies, fish habitat areas, and wild and
scenic rivers. Reach addresses establish the locations of these entities relative to one another
within the NHD surface water drainage network in a manner similar to street addresses. Once
linked to the NHD by their reach addresses, the upstream/downstream relationships of these
water-related entities and any associated information about them can be analyzed using software
tools ranging from spreadsheets to Geographic Information Systems (GIS). EPA's Watershed
Assessment, Tracking & Environmental Results (WATERS) system uses the reach codes in
NHD to link multiple databases containing water quality and programmatic information. EPA
has created a version of the NHD called NHDPlus45, which contains information on stream
velocity and flow volume, which are necessary for modeling chemical concentrations in streams
in the RSEI model.

It should be noted that the atomic unit of the NHDPlus data are the "flowline" which is a sub-
section of a reach. RSEI maintains the data at this level; however, the discussion throughout this
document will mostly be in terms of reaches, since much of the other data (such as discharge
reach) are at this level. In cases where a reach needs to be matched to a flowline, the most
upstream flowline in that reach is used.

Certain criteria were applied to the NHDPlus dataset to select the reaches to be used in the
model. Specifically, because RSEI calculates the movement of a chemical release downstream
using flow and velocity data, qualifying reaches must have at least one downstream or upstream
connecting reach46, and have a non-negative flow and velocity. RSEI will not calculate
concentrations for certain types of reaches, such as coastlines, treatment reservoirs, and bays; the
downstream path of any chemical is assumed to stop if one of these types of reach is
encountered. Additionally, some types of reaches are excluded from the set of fishable reaches,
such as pipelines, aqueducts, and certain types of reservoirs. NHDPlus does not separate canals
(presumably fishable) and ditches (presumably not fishable), so RSEI excludes reaches in the
canal/ditch category if the annual mean flow is less than 5 ft3/s. This is an arbitrary minimum,
and is intended primarily to exclude ditches at the point of the facility discharge. For reaches
designated as not fishable in NHDPlus, the chemical is still assumed to travel downstream to the
next reach, which may or may not be fishable.

Because NHDPlus contains a large number of reaches with very small annual mean flows, RSEI
also excludes the very smallest reaches from assignment as discharge reaches. To determine an
appropriate minimum, a national set of EPA-assigned discharge reaches from the ICIS-NPDES
database was matched to the NHDPlus annual mean flow data. For each NHDPlus region, the
non-zero annual mean flows were ranked, and the fifth percentile flow for each region was
selected as the minimum annual mean flow. Assigned regional minimum values ranged from
0.0036 ft3/s to 1.9 ft3/s. For instance, if a facility in a region whose minimum value was selected
as 0.0036 ft3/s had a nearest reach with annual mean flow of 0.0025 ft3/s, this would not be

45	The NHDPlus data was provided by Horizon Systems Corporation, which hosts and maintains the NHDPlus data.
Documentation and data are available at http://www.horizon-systems.com/nhdplus/index.php.

46	In the NHDPlus dataset, topologically connected reaches with known flow are indicated by a "FlowDir" value of
"With Digitized." Only reaches with this value for this field were included in RSEI, which includes 2.6 million out
of 2.9 million flowlines in NHD.

Version 2.3.2
July 2013

51

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

selected as the discharge reach. Instead the next closest reach with a flow greater than 0.0036
ft3/s would be selected as the discharge reach.

5.4.3.2	Drinking Water Intakes and Populations

Drinking water intake locations were obtained from the Public Supply Database, a database of
drinking water system information developed and maintained by the U.S. Geological Survey
(USGS), based on information in EPA's Safe Drinking Water Information System (SDWIS).
SDWIS is a publicly-accessible database that contains the information EPA uses to monitor
public water systems. The database contains information on over 156,000 water systems, which
serve over 96 percent of the U.S. population.47 SDWIS is operated and maintained by EPA's
Office of Water. USGS's Public Supply Database was designed to support USGS analyses of
the water resources used by public drinking water systems. The Public Supply Database includes
SDWIS data, and supplements it with additional information about the water bodies from which
systems draw their water. In addition, multiple quality-assurance checks have been performed
on the data, including the lat/longs.

The version of the Public Supply database used contains SDWIS population data from 2007.
Approximately 11,400 drinking water intakes were included in the database. Several types of
intakes were excluded from the set used for RSEI: 1) if the drinking water system for the intake
closed prior to 2002; 2) if the source water for the intake was something that could not
reasonably be expected to be connected a network of streams (such as an aqueduct or an
infiltration gallery); or 3) if the intake was emergency, interim (peak) or other (rather than
permanent or seasonal). Excluding these cases left 6,215 intakes that are modeled in RSEI.

5.4.3.3	Water Flow and Velocity

RSEI uses NHDPlus estimates of water flow and velocity based on the unit runoff method,48
which was developed for the National Water Pollution Control Assessment Model (NWPCAM).
The unit run-off method calculates average runoff per square kilometer in a watershed (8-digit
HUC) based on gages in the HydroClimatic Data Network (HCDN). These gages are usually not
affected by human activities, such as major reservoirs, intakes, and irrigation withdrawals; thus,
the mean annual flow estimates are most representative of "natural" flow conditions. Based on
elevation and drainage patterns, each square kilometer in a watershed is assigned to a catchment
area, from which the runoff flows to a specific flowline. The runoff from each catchment area is
summed and attributed to its assigned flowline. That flow is assigned to the next downstream
flowline, to which the downstream flowline's catchment runoff is added, and so on down the
stream path. Unit runoff estimates are calibrated for areas west of the Mississippi to account for
water withdrawals and transfers.

NHDPlus velocities are estimated for mean annual flow conditions (using the unit runoff
method) based on the work of Jobson (1996). This method uses regression analyses on hydraulic
variables for over 980 time-of-travel studies, which represent about 90 different rivers in the U.S.
representing a range of sizes, slopes, and channel geometries. Four principal flowline variables

47	http://water.epa.gov/scitech/datait/databases/drink/sdwisfed/upload/epa816rl3003.pdf.

48	NHDPlus also contains estimates developed using the Vogel method, but this is method is considered to have a
narrow band of applicability.

Version 2.3.2
July 2013

52

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

are used in the Jobson methods: drainage area, flowline slope, mean annual discharge, and
discharge at the time of the measurement.49

NHDPlus is based on the medium-resolution NHD dataset, which does not include the Great
Lakes. Therefore, flow estimates of rivers in the Great Lakes area in the NHDPlus dataset do not
include discharges from the Great Lakes. To account for this, RSEI includes adjustment factors
based on flow estimates from either in-stream gages or annual mean flow estimates from the
U.S. Army Corps of Engineers (USACE)50. The flow adjustment is added to the first reach
downstream of each Great Lake, and the additional flow is added to each downstream reach, and
branched as appropriate. RSEI does not model potential risk in the Great Lakes themselves, due
to the complicated hydrology.

5.4.3.4	Water Decay Rates

Water decay rates are required to model downstream chemical concentrations. The primary
sources for water decay values were Syracuse Research Corporation's (SRC's) ChemFate
database, a component of SRC's Environmental Fate Data Base (SRC, 2002a), which contains
experimental data, and SRC's Aqueous Hydrolysis Rate Program, HYDROWIN (part of the EPI
suite of estimation programs (SRC, 1994-1999)), both of which were developed for the
Environmental Protection Agency. The ChemFate database contains environmental fate and
physical/chemical property information for commercially important chemical compounds,
including TRI chemicals. HYDROWIN estimates hydrolysis rate constants for esters,
carbamates, epoxides, halomethanes, and selected alkyl halides. Values of water decay rates can
be found in Technical Appendix B.

5.4.3.5	Bioconcentration Factors

Bioconcentration factor (BCF) is the term used to describe the equilibrium concentration of
chemicals in aquatic organisms living in contaminated water. The BCF is defined as the ratio of
the chemical concentration in the organism (mg/kg) to that in the surrounding water (mg/L). The
term "bioconcentration" refers to the uptake and retention of a chemical by an aquatic organism
from the surrounding water only.51 Experimental BCF values were obtained from SRC's
ChemFate database. Other BCFs were estimated from either log(Kow) values using regression
equations from Lyman et al. (1990), or from the SRC estimation program BCFWIN. See
Technical Appendix B for values and references for the bioconcentration factors used in the
model.

49	For more information, see the NHDPlus User's Guide (US EPA/USGS 2009).

50	USGS' National Water Information System (NWIS) Mapper shows stream gages (available at
http://wdr.water.usgs.gov/nwisgmap/'). and the USACE website has average annual flow data (available at
http://www.lre.usace.army.mil/greatlakes/hh/outflows/historic%20connecting%20channel%20outflows/index.cfm)

51	The BCF can underestimate the accumulation of chemicals that are highly persistent and hydrophobic relative to
the bioaccumulation factor (BAF), which measures the uptake and retention of a chemical by an aquatic organism
from all surrounding media (e.g. water, food, sediment). The bioaccumulation factor (BAF) is defined as the ratio
of the chemical concentration in the organism (mg/kg) to that in the surrounding water (mg/L), in situations where
both the organism and its food are exposed. Due to data limitations at the present time, only BCFs are used in the
RSEI model.

Version 2.3.2
July 2013

53

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.4.3.6 Human Exposure Assumptions
Drinking Water. For the drinking water pathway, the model uses estimates for the amount of
tap water ingested to estimate exposure. As in the stack and fugitive air pathways, data are
acquired from EPA's Exposure Factors Handbook (EFH) (EPA, 2011). EFH recommends mean
tap water intake values for males and females combined from two related studies, Kahn and
Stralka (2008) and Kahn (2008) for children less than three years of age, and EPA's analysis of
NHANES data from 2003-2006 for individuals three years of age and older (as cited in EPA,
2011, Table 3-1). The body weights used in these studies to calculate the exposure factors were
different from those recommended in the 2011 EFH, so the values were adjusted to reflect the
currently recommended sex- and age-specific body weights.

Drinking water intake rates per body weight were calculated for each of the modeled groups
(male/female: ages 0-17, 18-44, 45-64, 65+)52 using the weighted average approach presented in
Equation 5.7. The final drinking water exposure factors are presented in Exhibit 5.9. More
detail on the derivation of exposure factors can be found in Technical Appendix C.

Exhibit 5.9
Drinking Water Exposure Factors

Model Age Group

Exposure
Factors (Male)

Exposure Factors
(Female)

(L/kg-day)

0-17

0.0111

0.0117

18-44

0.0117

0.0137

45-64

0.0117

0.0135

365

0.0128

0.0151

Fish Ingestion. RSEI uses annual estimates of the amount of fish ingested by recreational and
subsistence fishers and their families. However, there are no national data on fish ingestion rates
specific to recreational and subsistence fishers. In the absence of such data, RSEI uses fish
ingestion rates from the 1994-1996 USDA Continuing Survey of Food Intake by Individuals
(CSFII). This survey was conducted by the U.S. Department of Agriculture in 50 states and the
District of Columbia over a three-year period. A total of 15,303 individuals provided two non-
consecutive days of data on dietary intake. Appropriate statistical techniques were used to
extrapolate to the national population. In a 2002 publication, EPA assigned specific fish species
to habitats (freshwater, estuarine, and marine) based on the majority of time the species spend in
those habitats (EPA, 2000). Based on these assignments, EPA estimated a distribution of
uncooked finfish and shellfish ingestion rates specific to freshwater and estuarine fish.53 As

52	The RSEI model provides the ability to view risk scores for the 0-9 year old age group, however, there is not a
separate exposure factor for this age group. Instead, the exposure factor for the 0-17 year old age group is used

53	Consumption of marine fish is not included in the ingestion rates, because marine areas are not modeled in RSEI.

Version 2.3.2
July 2013

54

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

recommended by EPA's Office of Water (Tudor et al., 2000), for environmental assessments the
90th percentile is used to represent ingestion rates for recreational fishers, and the 99th percentile
is used for subsistence fishers. The ingestion rates are reported by age group (<15 years, 15-44
years, 45+ years) and sex (EPA, 2002). These values are roughly similar to ingestion rates
obtained from regional studies of recreational fishers and subsistence fishers, respectively. Fish
ingestion values were estimated for the RSEI age groups using Equation 5.7. These values are
then divided by age- and sex-specific body weights, averaged to match the RSEI age groups
using data provided in EFH (EPA, 2011, Tables 8-4 and 8-5). Exhibit 5.10 presents the fish
ingestion rates used in the model. More detail on the derivation of exposure factors can be found
in Technical Appendix C.

Exhibit 5.10
Fish Ingestion Exposure Factors



Recreational (g/kg-day)*

Subsistence (g/kg-day)1

Model Age Group

Male

Female

Male

Female

0-17

0.0678

0.0288

2.37

1.85

18-44

0.182

0.0862

1.76

1.50

45-64

0.362

0.229

1.85

1.41

>65

0.398

0.255

2.04

1.57

1 Fish ingestion exposure factors are converted to kg/kg-day for the surrogate dose calculation in the model.

5.5 Modeling Transfers to POTWs

In 2011, approximately three percent of TRI releases and transfers were transferred to Publicly-
Owned Treatment Works (POTWs). These transfers are mostly of facility wastewater through
underground sewage pipes to a POTW. Each chemical transfer to a POTW is modeled as
entering as liquid influent. Depending on the chemical's physical properties, some portion of the
chemical release in the influent may be discharged into surface water from the POTW,
potentially resulting in human exposure through drinking water or fish ingestion. The rest of the
chemical release may be removed by the POTW through one or more of the following processes:
1) biodegradation, which is not modeled; 2) volatilization, which is modeled like other area air
releases (see Section 5.2.3); or 3) landfilling of sludge, which is not modeled. The following
sections describe the method and data used to model transfers to POTWs.

5.5.1 Transfers to POTWs: Method

Modeling exposure from TRI-reported transfers to POTWs requires: (1) location of the POTW to
which the chemicals are discharged, (2) location of the reach to which the POTW discharges,
(3) consideration of overall removal efficiencies of POTWs and resulting effluent discharges
from POTWs (the chemical-specific removal rate), and (4) consideration of residuals
management at POTWs (partitioning within the POTW).

Version 2.3.2
July 2013

55

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.5.1.1	Locating the POTW

In order to model releases from POTWs, these facilities must first be located on the model grid.
Like other off-site facilities, POTW names and addresses are reported to TRI by the facility
transferring its waste. Latitude and longitude are not reported. In order to derive lat/long
coordinates, the reported street addresses were geocoded for Version 2.1.3 (coordinates were
assigned based on street address).54 POTWs (and incinerators) were also matched to EPA's
Facility Registry System (FRS), based on name, address, and RCRA identification number,
where possible. Duplicate entries for the same POTW (in the common instance where two or
more reporting facilities have transferred to the same POTW) were collapsed to a single entry
using an approximate string matching program (see Technical Appendix D for details). Once
latitude and longitude for a facility are assigned (from geocoding, the FRS data, or based on zip
code centroids), the data are used to map the facility to the grid cell with the same coordinates.
Substantial data processing was necessary to prepare the set of off-site facilities for use in the
model; see Technical Appendix D for details on the steps that were taken.

5.5.1.2	Locating the POTW Discharge Reach

As with TRI reporting facilities, the POTW's discharge reach must be identified. The main data
source used is EPA's records of discharge permits for POTWs, which in most cases specify the
discharge reach by name or reach number. Permit records, including permits, monitoring data,
and locational and descriptive information pertaining to more than 67,000 regulated facilities are
maintained in EPA's Integrated Compliance Information System-National Pollutant Discharge
Elimination System (ICIS-NPDES).

POTWs were matched to the FRS system based on name and address to obtain the FRS ID for
each POTW. THE FRS ID was then used to access ICIS-NPDES and the assigned discharge
reach for each POTW. Approximately 3,000 records were matched to a discharge reach using
this method.

POTWs not matched to a ICIS-NPDES discharge reach were assumed to discharge to the nearest
reach within four kilometers that meets the minimum flow requirements described above in
Section 5.4.3.1.

5.5.1.3	Overall POTW Removal Rate

POTWs cannot completely remove all of the chemicals that are transferred to the plant from the
TRI facility. Some of the chemical loading in the influent will be discharged as effluent to
surface waters. To calculate the fraction of transferred chemical removed by the POTW, the
typical chemical-specific removal rate is applied to the volume transferred to the POTW from the
TRI facility. See Technical Appendix B for a listing of removal rates and references for each
chemical. The remainder is assumed to exit the POTW in water effluent. This effluent is
modeled for drinking water and fish ingestion using the same methods for surface water releases
described above.

54 Geocoding services were provided by Thomas Computing Services, a commercial firm.

Version 2.3.2
July 2013

56

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.5.1.4 Partitioning within the POTW

Chemical loadings may be removed by the POTW treatment processes through biodegradation,
volatilization, and adsorption to sludge. The amount of the chemical that is removed by each of
these processes is modeled using average partitioning rates (see Technical Appendix B for the
listing of partitioning rates and references for each chemical).

Once the fates of chemicals entering the POTW are estimated, exposures associated with
chemical loadings to each compartment are estimated. Chemicals discharged in the POTW
effluent are modeled using the surface water evaluation methods described above. Chemicals
that biodegrade are assumed to degrade to chemicals that do not pose risk. POTW volatilization
releases are treated like area-source air releases, as described earlier.

For chemicals that partition to sludge, the model used to estimate exposure should ideally depend
on the sludge disposal method employed by the POTW. However, sludge disposal practices at a
POTW receiving a TRI transfer cannot be determined from the TRI database. Therefore, the
model algorithm currently assumes all POTW sludge is landfilled at the POTW, a common
method of sludge disposal. POTWs may in reality use other methods of sludge disposal, such as
incineration of sludge. If sludge were incinerated by a POTW, for example, this would result in
different exposure levels (and a different, larger exposed population). RSEI does not currently
model land releases. A summary of the approach to modeling POTW emissions is found in
Exhibit 5.11.

Version 2.3.2
July 2013

57

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.11
POTW Modeling Approach

Version 2.3.2
July 2013

58

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.5.1.5 Estimating Population Size for POTW Transfers

The population exposed to air releases is assumed to be the population within 49 km around the
POTW. The method used to estimate the population surrounding the POTW that is exposed to
surface water effluent discharges up to 200 km downstream from these facilities is described in
the section on exposed populations for surface water releases (see Sections 5.4.1.2 and 5.4.1.4.)

5.5.2 Transfers to POTWs: Data

Exhibit 5.12 presents data used to estimate exposure from releases to POTWs. In addition to the
parameter data presented here, data from the air release pathway (see Exhibit 5.3), and water
release pathway (see Exhibit 5.7) are also used. Environmental fate and transport data and
exposure factors specific to these pathways are described in the relevant sections above.

Exhibit 5.12

Data Used to Estimate Exposure from Releases to POTWs

Parameter

Value

Source/Comment

Removal efficiencies

chemical-specific

RREL or STPWIN
(SRC, 1994-99)

Partitioning within the
POTW

chemical-specific

RREL or STPWIN
(SRC, 1994-99)

5.5.2.1 POTW Removal Rate Efficiency and Within-POTW Partitioning

Data specific to this pathway include POTW removal efficiencies and within-POTW partitioning
rates. These parameters describe the fate of chemicals during treatment at POTWs. The "POTW
Partition Removal" is the total POTW removal efficiency, or the total percentage of the chemical
removed by the POTW (influent minus effluent). The within-POTW partition values describe the
fate of that portion of the chemical removed, that is, whether the chemical may sorb to sludge
(POTW Partition Sludge), volatilize into the air (POTW Partition Volatile) or be biodegraded by
microorganisms (POTW Partition Biodeg). The within-POTW partitioning values are expressed
as percentages of the total POTW removal efficiency; that is, they sum to 100 percent.

POTW removal efficiencies were obtained from the Treatability Database maintained by EPA's
Risk Reduction Engineering Laboratory (RREL). For any given chemical, the RREL
Treatability Database provides a list of removal efficiencies published in the scientific literature.
Each value is characterized by the technology used, the type of influent, and the scale of the
experiment. For all values associated with activated sediment and full scale experiments, a
geometric mean was derived and used as the POTW removal efficiency. Within-POTW
partitioning values were obtained from two sources. For most organic chemicals, values were
supplied by the EPA's Exposure Assessment Branch within OPPT. Inorganic chemicals, except
for ammonia, were assumed to partition 100% to sludge. For chemicals without data from these

Version 2.3.2
July 2013

59

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

sources, SRC's Sewage Treatment Plant Fugacity Model (STPWIN) was used to estimate total
removal efficiency and within-POTW partitioning values.

5.6 Modeling Other Off-site Transfers

In 2011, approximately 44 percent of TRI emissions were transferred to off-site locations other
than POTWs for storage or disposal. TRI reporters are required to supply the name and address
of the receiving facility, and the treatment method used. Currently, only transfers that are
incinerated at the off-site facility are modeled; waste transfers treated by other methods do not
receive risk-related scores.

5.6.1	Off-site Transfers: Method

To assess the exposure potential associated with off-site transfers, it is important to have
information about the off-site facility location and some of its characteristics. Locations of other
off-site facilities are determined in the same way as the locations of POTWs. The reported street
addresses were geocoded for Version 2.1.3 (coordinates were assigned based on street address).55
Incinerators (and POTWs) were also matched to EPA's Facility Registry System (FRS), based
on name, address, and RCRA identification number, where possible. Duplicate entries for the
same off-site facility (in the common instance where two or more reporting facilities have
transferred to the same off-site facility) were collapsed to a single entry using an approximate
string matching program (see Technical Appendix D for details). Once latitude and longitude for
a facility are assigned (from geocoding, the FRS data, or based on zip code centroids), the data
are used to map the facility to the grid cell with the same coordinates. See Technical Appendix
D for detailed information on locating off-site facilities.

The TRI forms require the reporting facility to indicate the treatment/disposal method used at the
off-site facility. If this information is not reported (despite the requirement), the transfer is not
evaluated in the algorithm, but is flagged as a missing value and assigned a zero.

If the treatment method is incineration, then destruction and removal efficiencies (DREs) are
applied to the transfer amount. Once DREs have been applied, the releases are modeled using
the AERMOD-based air modeling algorithms described earlier.

For off-site landfills, there are two exposure pathways, groundwater and volatilization.

However, as land releases are not currently modeled, there are no risk-related results available
for either pathway. Off-site transfers to underground injection wells are also not modeled.
Exhibit 5.13 summarizes the modeling of off-site transfers.

5.6.2	Estimating Population for Off-Site Transfers

Similar to on-site air releases, the population exposed to air releases from off-site transfers is the
population within 49 km around the off-site incinerator.

55 Geocoding services were provided by Thomas Computing Services, a commercial firm.

Version 2.3.2
July 2013

60

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

Exhibit 5.13
Modeling Approach for Off-site Transfers

5.6.3 Off-site Transfers: Data

5.6.3.1 Destruction and Removal Efficiencies

For organics, the destruction and removal efficiency (DRE) is assumed to be 99 percent (see
Technical Appendix B). The exceptions to the 99 percent removal assumption are PCBs and
dioxin and dioxin-related compounds, which are assumed to have a DRE of 99.9999 percent, as
required by TSCA regulation. For inorganics, values are taken from multiple hearth sludge
incinerator studies (EPA, 1992a).

Version 2.3.2
July 2013

61

1988-2011 TRI data

-------
Chapter 5: Exposure and Population Modeling

5.7 Modeling On-site Land Releases

In 2011, approximately 36 percent of TRI emissions and transfers were released as on-site land
releases. On-site land releases include releases to landfills, surface impoundments, land
treatment units, and underground injection wells. For these releases, two major exposure
pathways are of interest-volatilization to air and leaching into groundwater. Volatilization of
chemicals from on-site land releases is reported to TRI under the fugitive emission estimate for
the facility, and is modeled by RSEI as part of the facility's fugitive air release. For more
information on RSEI modeling of fugitive air releases, see Section 5.3 above. EPA is evaluating
screening-level methodologies which might be used to assess risk-related exposures pertaining to
groundwater exposure from on- and off-site land releases and volatilization from off-site land
releases, so this version of RSEI does not provide risk modeling for reported land releases.
However, RSEI does provide the capability for users to examine the pounds of releases to land
that are reported to TRI, as well as viewing these releases from a hazard-based perspective.

The potential for groundwater contamination from land releases depends on the regulatory status
of the unit in which the chemical is released. For example, chemicals could be deposited in an
on-site RCRA-regulated, subtitle C hazardous waste unit, or in an on-site nonhazardous solid
waste management unit. RCRA standards for hazardous waste units are, by regulation, designed
to include technical controls to prevent release of contaminants into groundwater. If chemicals
are placed in such regulated units, EPA assumes that releases to groundwater are negligible so
RSEI assigns a zero value to the risk-related scores for such releases. If chemicals are placed in
nonhazardous land disposal units (landfills, etc.), there is a potential for exposure. This exposure
pathway and volatilization from off-site landfills are currently under review for inclusion in a
future version of RSEI.

On- and off-site land releases to underground injection will not be modeled for exposure by
RSEI. The hydrogeological, spatial, and temporal considerations that are associated with
exposures to toxic chemicals in underground injection wells are situation- and site-specific, so
RSEI is only able to provide pounds-based and hazard-based perspectives for this type of land
release. Note, however, that under well-managed conditions, Class I wells (there are five classes
of wells) are specifically designed to pose minimal risk to human health or the environment.

Version 2.3.2
July 2013

62

1988-2011 TRI data

-------
Chapter 6: Calculating Results

6. Calculating Results

This section summarizes the computation of the principal types of RSEI results. Because of the
multi-functional nature of the model, a wide variety of results can be created. All of the RSEI
functionality is based upon the Indicator Element, which is a unique combination of chemical,
facility, release and exposure pathway, and year. Each Indicator Element has a set of associated
results:

Exhibit 6.1
Description of RSEI Results

Risk-related results

Surrogate Dose x Toxicity Weight x Population

Hazard-based results

Pounds x Toxicity Weight

Pounds-based results

TRI Pounds released

Risk-related results. The surrogate dose, toxicity, and population components are multiplied to
obtain a risk-related score for the Indicator Element. The surrogate dose is determined through
pathway-specific modeling of the fate and transport of the chemical through the environment,
combined with subpopulation-specific exposure factors. Risk-related scores are unitless, and
each of the components (toxicity weight, surrogate dose, and population) when multiplied
provide scores that are relevant only when compared to each other. The unitless Indicator
Elements are not a physically meaningful measure of quantitative risk associated with the
facility, but are approximate measures of relative risk-related impacts that are comparable to
approximate measures for other facilities (or other chemicals, pathways, etc.) calculated using
the same methods within RSEI. If the Indicator Element cannot be modeled, because of a lack of
data required for modeling, or because the pathway is not currently modeled, then the risk-
related score is zero. The model calculates risk-related results for the entire population and also
for the following subpopulations: children under 10, children aged 10 to 17, males aged 18 to 44,
females aged 18 to 44, and adults aged 65 and older. In addition the model also calculates
"Modeled Pounds," which is simply the number of pounds that can be modeled, before fate and
transport modeling and exposure assumptions have been applied.

Higher component "weights" are associated with higher relative risk-related values (and lower
weights are associated with lower relative risks). For chemicals with cancer effects, multiplying
the weights associated with cancer toxicity and exposure to the chemical seems intuitive, since
this is similar to the calculation of cancer risk with a slope factor or unit risk value and dose or
exposure level. For chemicals with noncancer effects, the multiplicative nature of the toxicity
and exposure weights may not seem intuitive, because in risk assessments, risk is usually
characterized as the estimated exposure divided by the RfD/RfC. However, because of the
manner in which the toxicity weights have been constructed, the product of toxicity weight and
surrogate dose varies in the same direction and degree as the ratio of exposure to RfD/RfC. This
is because the toxicity weight is inversely related to the magnitude of the RfD/RfC. Thus, for a
given exposure level, a chemical with a more stringent (i.e., lower) RfD will receive a higher
Indicator Element value than a chemical with a less stringent (i.e., higher) RfD, as shown in the
following example:

Version 2.3.2
July 2013

63

1988-2011 TRI data

-------
Chapter 6: Calculating Results









Exposure

Toxicity









(i.e., surrogate

Weight *



RfD (mg/kg-

Toxicity

Surrogate dose

dose) /RfD

Surrogate



day)

Weight

(mg/kg-day)

Ratio

Dose

Scenario 1

0.1

5

1

1/0.1 = 10

5*1 = 5

Scenario 2

0.01

50

1

1/0.01 = 100

50*1 =50

Since no adverse effects are expected to occur below the RfD, one could argue that releases that
result in surrogate doses below the RfD should be excluded. However, this approach was not
pursued for the following reasons: first, the estimation of surrogate dose is only a screening-level
approximation for the purposes of comparing one release to another in a relative way, and should
never be considered an actual calculation of exposure. To exclude releases resulting in surrogate
doses below the RfD would incorrectly imply that the method could predict precisely when doses
would occur below the RfD. Second, exposure to the same chemical from multiple facilities, or
multiple chemicals from one or more facilities affecting the same health endpoint could act
additively to pose risk, even if each release individually did not exceed the RfD. Finally, if the
surrogate dose is low, this will be reflected by a correspondingly low score relative to other
releases for that chemical.

Hazard-based results. Each Indicator Element also is associated with a hazard-based result
("Hazard"), calculated by multiplying the pounds released by the appropriate chemical-specific
toxicity weight (the toxicity weight also depends on the exposure pathway). The inhalation
toxicity weight is used for releases or transfers to fugitive air, stack air, off-site incineration, and
off-site incineration-no fuel value. The oral toxicity weight is used for releases or transfers to
direct water and POTWs. For releases that are not modeled (because the pathway is not modeled
or because other necessary data, such as physicochemical properties, are lacking), the higher
toxicity weight is used. For these results, no exposure modeling or population estimates are
involved. If there is no toxicity weight available for the chemical, then the hazard score is zero.

The model also calculates "Modeled Hazard," which is the chemical- and pathway-specific
toxicity weights multiplied by the Modeled Pounds (as described above), and "Modeled Hazard
* Pop," which multiplies modeled hazard by the potentially exposed population, but without the
fate and transport modeling (and application of exposure assumptions) that would be found in
risk-related results.

Pounds-based results. These results ("TRI Pounds") reflect only the number of pounds released
or transferred that are reported to TRI, and are available for virtually all Indicator Elements.

The model also provides "TRI Pounds with Toxicity Weights," which simply sums the pounds
for chemicals that have toxicity weights in RSEI.

Version 2.3.2
July 2013

64

1988-2011 TRI data

-------
Chapter 6: Calculating Results

6.1	Combining Indicator Elements

Once results are calculated for each Indicator Element, they can be combined in many different
ways. All of the results are additive, so a result for a specific set of variables is calculated by
summing all the relevant individual Indicator Element results, as follows:

* = XX2X/.,	(Eq 6 1)

where:

R	= RSEI result

IEcip	= chemical-facility-pathway-specific Indicator Element result

This method is very flexible, allowing for countless variation in the creation of results. For
example, results can be calculated for various subsets of variables (e.g., chemical, facility,
pathway) and compared to each other to assess the relative contribution of each subset to the
total potential impact. Or, results for the same subset of variables for different years can be
calculated, to assess the general trend in pounds-based, hazard-based, or risk-related impact over
time.

It must be reiterated that while changes in results over the years would imply that there have
been changes in hazard- or risk-related environmental impacts, the actual magnitude of any
specific change or the reason may not be obvious. Although the value itself may be useful in
identifying facilities or chemicals with the highest potential for hazard or risk, the score does not
represent a quantitative estimate or provide an exact indication of the magnitude of individual
hazard or risk associated with that facility or chemical.

6.2	Accounting for Changes in TRI Reporting

When a change occurs in the number of, or reporting requirements for, chemicals and facilities
represented in TRI, the numerical value of RSEI results will be altered if no adjustments are
made to the method of calculation to account for the changes respective to trend analyses.
However, such changes would not necessarily represent a large change in actual environmental
impact, but would reflect a broader understanding of the impacts that may have always existed.

A change in the number of chemicals and facilities in the TRI can occur through several
mechanisms. First, the addition to or deletion of individual chemicals from the TRI chemical list
will occur as EPA responds to petitions or initiates its own action through the chemical listing or
delisting process. Several additions and deletions to the data set have already occurred since
1987 (the first year of TRI reporting). Furthermore, the Agency added 245 chemicals and
chemical categories to the TRI chemical list in a single year, effective in 1995. The deletion of
chemicals would presumably have a minor effect since such chemicals would be deleted due to
their low hazard. Delisted chemicals are removed from RSEI. To account for changes in the
representation of chemicals in the TRI database, RSEI uses a special identifier called "Core
Chemical" which denotes chemicals that have been listed on the TRI since the first year (1987)
of reporting and which have had no change in their reporting requirements. Another identifier,

Version 2.3.2
July 2013

65

1988-2011 TRI data

-------
Chapter 6: Calculating Results

"Mini Core" denotes chemicals that have been listed on the TRI since 1995 and which have had
no change in their reporting requirements. These identifiers allow users to conduct separate
analyses for the "Core" or "Mini Core" chemicals, and so exclude chemicals whose emissions
changes over time are caused by reporting requirement changes. Similarly, the 245 chemicals
added for the 1995 reporting year can also be analyzed separately.

Facility-level changes can also affect year-to-year scores generated using RSEI. For instance,
compliance with TRI reporting has improved over time, which has led to more facilities
reporting. Increases in the number of reporting facilities may occur as a result of changes in
reporting requirements. For instance, in the first two years of reporting, facilities that
manufactured or processed more than 50,000 pounds were required to report their releases.
However, EPCRA lowered this threshold to 25,000 pounds in 1989. For reporting year 2000,
thresholds and other reporting requirements for 18 Persistent Bioaccumulative Chemicals (PBTs)
have been changed. These modifications can act to alter the total emissions reported to the TRI
and the model's estimates of associated hazard- and relative risk-based impacts. Also, effective
in the 1998 reporting year, certain SIC codes were added to TRI, adding to the universe of
reporting facilities.56 To assist users in separating out the effects of the 1998 expansion, RSEI
allows for the exclusion of facilities in the newly-required SIC codes when doing trend analyses.

The yearly TRI reporting data for a given list of chemicals and facilities are the subject of
ongoing quality control review and revision. As a result, yearly comparisons could be flawed if
ongoing revisions by individual facilities were not included in each year's results. Therefore, the
Indicator Elements are re-computed for all years in the database on an annual basis in order to
incorporate revisions to the reporting data. This annual calculation is based on the corrections
incorporated in annual Public Data Releases available from EPA's TRI program.

56 This facility expansion rule required the affected facilities to report their releases for the 1998 reporting year. The
added SIC codes are: codes 10 (except 1011, 1081, and 1094), 12 (except 1241), industry codes 4911, 4931 and
4939 (limited to facilities that combust coal and/or oil for the purpose of generating power for distribution in
commerce), 4953 (limited to facilities regulated under RCRA), 5169, 5171, and 7389 (limited to facilities engaged
primarily in solvent recovery services on a contract or fee basis) (EPA 1997a).

Version 2.3.2
July 2013

66

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

7. Current Implementation of the RSEI Method

7.1 RSEI Model

The RSEI model is currently implemented in a Microsoft Windows-based computer program.
The program allows users to calculate RSEI results for reporting years 1996-2011 and to present
the results in various GIS, graphical, and tabular formats, as well as to save selected data to
spreadsheet and database formats (e.g., Microsoft Excel and databases such as Access). The
program includes on-line help for all of the program functions, as well as User's Manual in
Adobe Acrobat format.

Users of the model can perform, usually in a matter of minutes, a variety of screening-level
analyses. Previously, such activities would have taken days, weeks, or even months to organize
the relevant information, evaluate that information, and perform the complex and sophisticated
analyses that are necessary to provide a risk-related perspective. Results can be used for
screening-level ranking and prioritization for strategic planning purposes, risk-related targeting,
and trends analyses. Considerable resources can be saved by conducting preliminary analyses
with the model to identify risk-related situations of high potential concern, and which warrant
further evaluation.

As noted above, users can evaluate releases using a number of variables, such as chemical,
medium, geographic area or industry. For instance, the following types of questions can be
investigated:

•	How do industry sectors compare to one another from a risk-related perspective?

•	What is the relative contribution of chemicals within a given industry sector?

•	What release pathway for a particular chemical poses the greatest risk-related
impacts?

Users can view various pounds- and hazard-based results to investigate the relative influence of
toxicity and population components on the risk-related results. However, only the risk-related
results incorporate exposure modeling.

Users should note that, as implemented for the personal computer, RSEI employs a facility-based
approach. All modeled impacts are attributed to the facility originally releasing or transferring
the chemical. For instance, an air release from an off-site incinerator is modeled as exposing the
population around the off-site facility, but the results (pounds, hazard, score, etc.) are attributed
to the reporting facility that transferred the chemical to the off-site incinerator. Similarly, while
impacts may extend beyond geographic boundaries such as zip code, county, or state, the results
are attributed to the geographic entity in which the facility is located. EPA employs the RSEI
methodology to create other databases which are geographic-based. These very large datasets
are operated outside of the user-friendly interface provided by the RSEI model.

RSEI Version 2.3.2 can be downloaded from the RSEI web site at
http://www.epa.gov/oppt/rsei/pubs/get_rsei.html. The installed RSEI model requires
approximately 2 GB of hard disk space. It is designed for operation using a 32-bit operating

Version 2.3.2
July 2013

67

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

system (Windows 95, 98, 2000, XP, and NT4). The program is written using Borland's Delphi
(the software is coded in Object Pascal) and uses the Paradox file format. RSEI Version 2.3.2
contains data for TRI reporting years 1996-2011; data for TRI reporting years 1988-1995 are
available upon request.

Information regarding the RSEI project is available on the RSEI web site.57

7.2 Conclusion

As an indication of improvements in environmental quality over time, RSEI provides EPA and
the public with a valuable tool to measure general trends based upon the relative risk-related
impacts of TRI chemicals. Although RSEI results do not capture all environmental releases of
concern, they generally relate changes in releases to relative changes in chronic human health
impacts from a large number of toxic chemicals of concern to the Agency. Importantly, RSEI
provides an ability to analyze the relative contribution of chemicals and industrial sectors to
human health impacts, and RSEI results serve as an analytical basis for setting priorities for
pollution prevention, regulatory initiatives, enforcement targeting and chemical testing
requirements.

57 The RSEI website is available at www.epa.gov/oppt/rsei.

Version 2.3.2
July 2013

68

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

8. References

General References

Bodek, I., W.J. Lyman, W.F. Reehl, and D.H. Rosenblatt. 1988. Environmental Inorganic
Chemistry. Pergamon Press. New York.

Boubel, R.W., et al. 1994. Fundamentals of Air Pollution. Academic Press. New York.

Canadian Ministry of National Health and Welfare. 1981. Tapwater Consumption in Canada.
Doc # 82-EHD-80. Public Affairs Directorate, Department of National Health and
Welfare, Ottawa, Canada.

Chow, V.T., Open-channel Hydraulics, McGraw-Hill, New York, 1959.

Dourson, M. 1993. Environmental Criteria and Assessment Office, U.S. Environmental
Protection Agency. Personal communication, October 19.

Ershow, A.G. and K.P. Cantor. 1989. Total Water and Tapwater Intake in the United States:

Population-Based Estimates of Quantities and Sources. Life Sciences Research Office,
Federation of American Societies for Experimental Biology.

Harrigan, P. 2000. Office of Water, U.S. Environmental Protection Agency (EPA). Personal
communication. March.

Horn, Marilee. 2008. U.S. Geological Survey (USGS). Written communication and provision
of data via email. April.

Job son, H. E., Prediction of Traveltime and Longitudinal Dispersion in Rivers and Streams, U.S.
Geological Survey Water Resources Investigations Report 96-4013 (1996), U.S.
Geological Survey.

Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt. 1990. Handbook of Chemical Property
Estimation Methods. American Chemical Society. Washington, D.C.

Price, Curtis. 2012. U.S. Geological Survey (USGS). Written communication and provision of
data via email. June.

Syracuse Research Corporation (SRC). 1994-1999. EPIWin v 3.01-EPA. Syracuse, NY.

Syracuse Research Corporation (SRC). 2002a. Chemfate/Environmental Fate Data Base.
Accessed online at http://esc.syrres.com/efdb.htm (now at
http ://www. syrres. com/esc/efdb. htm).

Syracuse Research Corporation (SRC). 2002b. PhysProp Data Base. Accessed online at
http://esc.syrres.com/interkow/physdemo.htm (now at
http://www.syrres.com/esc/chems3.htm).

Version 2.3.2
July 2013

69

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

Tudor, L., H. Jacobs, and P. Harrigan. 2000. Memorandum to EAD, SASD, and HECD.

Subject: Fish Consumption Rates to Use in Environmental Assessments/Benefits
Analyses. May 10.

U.S. Census Bureau. 1996. TABLE 1. Projections of Households by Type: 1995 to 2010, Series
1, 2, and 3. Available at: http://www.census.gOv/prod/l/pop/p25-1129.pdf. May.

U.S. Department of the Interior, Fish and Wildlife Service (DOI, FWS). 1993. 1991 National
Survey of Fishing, Hunting, and Wildlife Associated Recreation. U.S. Department of
Commerce, Bureau of the Census. U.S. Government Printing Office, Washington D.C.

U.S. Geological Survey (USGS). 2012. Public Supply Database (PSDB). Written communication
via email and provision of data by Curtis Price. June.

U.S. Environmental Protection Agency (EPA). 1985. Exposure to Airborne Contaminants

Released from Land Disposal Facilities — A Proposed Methodology. Prepared for the
Office of Solid Waste by Environmental Science and Engineering, Inc. ESE Document
Number 85-527-0100-2140. August.

U.S. Environmental Protection Agency (EPA). 1986a. Guidelines for Carcinogen Risk
Assessment. 51 Federal Register 33992 (September 24, 1986).

U.S. Environmental Protection Agency (EPA). 1986b. Guidelines for Mutagenicity Risk
Assessment. 51 Federal Register 34006 (September 24, 1986).

U.S. Environmental Protection Agency (EPA). 1987. Graphical Exposure Modeling System
(GEMS) User's Guide. Prepared for the Office of Pesticides and Toxic Substances,
Exposure Evaluation Division by General Sciences Corporation under Contract No.
68023970. February.

U.S. Environmental Protection Agency (EPA). 1988a. IRIS Background Document #1.

Reference Dose (RfD): Description and Use in Health Risk Assessments. Integrated Risk
Information System (IRIS). Online. Maintained by Environmental Criteria and
Assessment Office, Cincinnati, OH.

U.S. Environmental Protection Agency (EPA). 1988b. Report to Congress: Solid Waste
Disposal in the United States. Volume 2. April.

U.S. Environmental Protection Agency (EPA). 1988c. National Survey of Solid Waste

(Municipal) Landfill Facilities. Office of Solid Waste EPA/530-SW88-034. September.

U.S. Environmental Protection Agency (EPA). 1988d. Industrial Subtitle D Risk Screening
Analysis Results. Prepared for the Office of Solid Waste by ICF, Inc. December 30.

U.S. Environmental Protection Agency (EPA). 1990a. Exposure Factors Handbook. Office of
Health and Environmental Assessment. EPA/600/8-89/043. March.

Version 2.3.2
July 2013

70

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

U.S. Environmental Protection Agency (EPA). 1990b. Hazard Ranking System: Final Rule. 55
Federal Register 241. pp. 51532-51667.

U.S. Environmental Protection Agency (EPA) 1990c. Science Advisory Board, Relative Risk
Reduction Project. Reducing Risk. EPA SAB-EC-90-021. September.

U.S. Environmental Protection Agency (EPA). 1991a. Toxics in the Community. National and
Local Perspectives. The Office of Pesticides and Toxic Substances. EPA 560/4-91-014.
September.

U.S. Environmental Protection Agency (EPA). 1991b. Guidelines for Developmental Toxicity
Risk Assessment. 56 Federal Register 63798 (December 5, 1991).

U.S. Environmental Protection Agency (EPA). 1992a. Human Health Risk Assessment for the
Use and Disposal of Sewage Sludge: Benefits of Regulation. Prepared for the Office of
Water. January.

U.S. Environmental Protection Agency (EPA). 1992b. User's Guide for the Industrial Source
Complex (ISC2) Dispersion Models. Volume 2. Description of Model Algorithms.
Prepared for the Office of Air Quality, Planning and Standards, Technical Support
Division. March.

U.S. Environmental Protection Agency (EPA). 1993b. User's Guide to the Building Profile
Input Program. Office of the Quality Planning and Standards. October.

U.S. Environmental Protection Agency (EPA). 1999a. "Residual Risk Report to Congress".
EPA-453/R-99-001. United States Environmental Protection Agency, Office of Air
Quality Planning and Standards. March.

U.S. Environmental Protection Agency (EPA). 1999b. Enhanced Reach File Version 1.2

Configured for ARC/INFO. http://water.usgs.gov/GIS/metadata/usgswrd/XML/erfl.xml.

U.S. Environmental Protection Agency (EPA). 1997a. Addition of Facilities in Certain Industry
Sectors. 62 Federal Register 84, pp.23833-23892.

U.S. Environmental Protection Agency (EPA). 2002. Estimated Per Capita Fish Consumption
in the United States: Based on Data Collected by the United States Department of
Agriculture's 1994-1996 Continuing Survey of Food Intake by Individuals. Office of
Water, Office of Science and Technology. August.

http://water.epa.gov/scitech/swguidance/fishshellfish/outreach/upload/2002 08 28 fish
consumption report.pdf

U.S. Environmental Protection Agency (EPA). 2004a. AERMOD: Description of Model

Formulation. Office of Air Quality, Planning and Standards. September. EPA-454/R-
03-004. Accessed online at

http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf.

Version 2.3.2
July 2013

71

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

U.S. Environmental Protection Agency (EPA). 2004b. User's Guide for the AMS/EPA
Regulatory Model - AERMOD. Office of Air Quality, Planning and Standards.
September. EPA-454/B-03-001. Accessed online at
http://www.epa.gov/scram001/7thconf/aermod/aermod_userguide.zip

U.S. Environmental Protection Agency (EPA). 2005. Guidelines for Carcinogen Risk

Assessment. Risk Assessment Forum. March. EPA/630/P-03/001B. Accessed online at
http://www.epa.gov/IRIS/cancer032505.pdf.

U.S. Environmental Protection Agency (EPA). 2007. The HEM-3 User's Guide. HEM-3

Human Exposure Model Version 1.1.0 (AERMOD version). Draft. January. Available
at: http://www.epa.gov/ttn/fera/data/hem/hem3 users guide.pdf.

U.S. Environmental Protection Agency (EPA) and U.S Geological Survey (USGS). 2009.
NHDPlus User's Guide. January 20. Available at ftp://ftp.horizon-
system s. com/NHDPlus/documentati on/NHDPLU S UserGui de. pdf.

U. S. Environmental Protection Agency (EPA). 2011. Exposure Factors Handbook:2011
Edition. Office of Health and Environmental Assessment. Volume 1. EPA/600/R-
090/052F. September. Available at http://www.epa.gov/ncea/efh/pdfs/efh-complete.pdf.

RSEI supporting documentation released by EPA

These documents can be found on the RSEI website (under the Documents and Documents

Archive headings) at http://www.epa.gov/oppt/rsei/documents.html.

User's Manual for RSEI Version 2.3.2. July 2013.

RSEI Technical Appendices:

Technical Appendix A - Listing of All Toxicity Weights for TRI Chemicals and

Chemical Categories
Technical Appendix B - Physicochemical Properties for TRI Chemicals and

Chemical Categories
Technical Appendix C - Derivation of Model Exposure Parameters
Technical Appendix D - Locational Data for TRI Reporting Facilities and Off-site
Facilities

Technical Appendix E - Derivation of Stack Parameter Data
Technical Appendix F - Summary of Differences between RSEI Data and the TRI
Public Data Release

TRI Relative Risk-based Environmental Indicators: Summary of Comments Received on the
Draft 1992 Methodology and Responses to Comments. Prepared for the Office of
Pollution Prevention and Toxics, Economics, Exposure and Technology Division,
Regulatory Impacts Branch. May 1997. Prepared by Abt Associates under Contract # 68-
D2-0175.

Version 2.3.2
July 2013

72

1988-2011 TRI data

-------
Chapter 7: Current Implementation of the RSEI Method

TRI Relative Risk-based Environmental Indicators: Methodology. Prepared for the Office of
Pollution Prevention and Toxics, Economics, Exposure and Technology Division,
Regulatory Impacts Branch. June 1997. Prepared by Abt Associates under Contract #
68-D2-0175.

TRI Relative Risk-based Environmental Indicators: Interim Toxicity Weighting Summary
Document. Prepared for the Office of Pollution Prevention and Toxics, Economics,
Exposure and Technology Division, Regulatory Impacts Branch. 1997. Prepared by Abt
Associates under Contract # 68-D2-0175.

Ground-Truthing of the Air Pathway Component of OPPT's Risk-Screening Environmental
Indicators Model. December 1998.

Estimates of Stack Heights and Exit Gas Velocities for TRI Facilities in OPPT's Risk-Screening
Environmental Indicators Model. June 1999.

Version 2.3.2
July 2013

73

1988-2011 TRI data

-------