TOXICS RELEASE INVENTORY
ENVIRONMENTAL INDICATORS METHODOLOGY
Prepared for:
TRI Environmental Indicators Work Group
Office of Pollution Prevention and Toxics
U.S. Environmental Protection Agency
401:;^ St, SW
Washington, D.C 20460
Prepared by:
Abt Associates, Inc.
4800 Montgomery Lane
Bethesda, MD 20814
September 23,1992
DRAFT REPORT
-------
DISCLAIMER
This is a draft report; it should not be cited or quoted. This document has not been
reviewed and approved for publication by the Office of Toxic Substances, Office of
Pesticides and Toxic Substances, U.S. Environmental Protection Agency. The use of trade
names or commercial products does not constitute Agency endorsement or recommendation
for use.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON. D.C. 20460
OFFICE OF
PREVENTION. PESTICIDES AND
TOXIC SUBSTANCES
March 29, 1993
Dear Colleague:
Please find attaqhed a copy of the draft report Toxics
Release Inventory Environmental Indicators Methodology. Earlier
I had indicated to you that you would not be sent a copy of the
draft as the final version was to be completed in the near
future. Due to unforseen circumstances it has not been possible
to complete the final version in the time frame originally
envisioned. Therefore, I am sending you a copy of the draft as
many of you expressed an interest in applying the methodology to
issues currently at hand.
I hope you find the report useful. If any assistance is
required in its interpretation please feel free to contact me at
202-260-1567, or by writing to the address above, including the
mail stop TS-779. If you feel a final copy would be needed in
the future you may request a copy.
Sincerely,
Nicolaas Bouwes, PhD
Economist
Regulatory Impacts Branch
Economics, Exposure & Technology
Division (TS-779)
Recycled/Recyclable
Printed with Soy/Canola Ink on paper that
contains at least 50% recycled fiber
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY
I. INTRODUCTION 1
II. GENERAL DESCRIPTION OF THE TRI ENVIRONMENTAL
INDICATORS 2
Human Health Chronic Indicator 3
Ecological Chronic Indicator 3
III. SELECTION OF CHEMICALS AND FACILITIES TO INCLUDE IN THE
INDICATORS 4
Selection of Chemicals 4
Inclusion of Facilities 16
IV. METHODS FOR ADJUSTING EMISSIONS FOR TOXXCITY AND
EXPOSURE 18
Toxicitv Weights — Human 20
Quantitative Data Used in Human Toxicitv Weighting 20
Qualitative Data Used in Hvman Toricity Weighting 23
Genep*! Formal for Combining ^^irilt of Evidence and Potency .... 28
Weights Applied to the Categories 28
The Hupian JHfefllth Toxicitv Matrices 33
Carcinogens 33
Noncarcinogens 36
Selecting |he Final Human Health Toxicitv Weight far a Chemical ... 36
Toxicitv Weights - Ecological 38
Data Used in Aquatic Toxtcjty Wejghtinj 38
The Aquatic Toxicitv Wefoh^ng Matrix, 39
Exposure Weights — Human 40
Quantitative Data Used in Exposure Potential Weighting 40
Qualitative Data Used in Exposure Ppfentf al Weighting 43
The Human Exposure Potential ^tJ8ht^ng Matrix 43
Exposure Weights — Ecological 45
Population Size Weights 45
Quantitative Data Used in Population Weyriftpng 46
Qualitative Considerations iii Evaluating Exoosed Peculation Size 46
^fJQ^f^ffQf^^J^^^ggfffBj^Bjf^^^Jjg^g^^^f^^^^^JgiJgfg^'JjI^^JgQJf^fjg^gJ^^^^glJJ^^f^^Jf^^^f^SSSm • • • ~W
Integrating Toxicitv Data. Exposure Data, and Population Data to Obtain
Fflcilitv-Cheniical-Mcdiiim lndicn*ftr^CPlBntS 47
Hnm^p Health Chronic Indicator 47
Ecological ^Tirpp'C Indicator 47
V. MECHANICS OF ESTIMATING EXPOSURE 48
Treatment of Emissions Range Reporting 51
DP AFT HEPORT - DO NOT CITE OR QUOTE
-------
Acknowledgements
This draft report was prepared by Abt Associates under the guidance of the TRI
Environmental Indicators Work Group. Work Group members include:
Nicolaas Bouwes (Work Group Chair)
Nancy Beach
Daniel Bushman
LorenHall
Karen Hammerstrom
Steven Hassur
Sondra HoUister
John Leitzke
Patrick Miller
Nestor Tirado
Sylvon Vonderpool
The Abt Associates project staff include the following:
Susan Egan Keane (project manager)
Brad Firlie (deputy project manager)
Lisa Akeson
Kathy Cunningham
Jonathan Kleinman
Alexandra Varlay
Carol Wagett
Richard Wells
Michael Conti (technical reviewer)
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
Stack and Fugitive Air Releases .................................. 51
Direct Surface Water Releases ................................... 52
Qn.Site Land Releases ......................................... 56
Releases To POTWs ........................................... 61
Off-Site Transfers ............................................ 65
VI. COMPUTATION OF THE INDICATORS .......................... 69
Indicators Computation ........................................ 71
Comonent Scores; ....................... 71
Simple Sum Normalised to a Base Yean ...................... 71
Other Methods of Calculation Considered ..................... 72
Normalizing the Indicators ..................................... 72
VII. ABIOTIC INDICATORS ....................................... 73
Global Wanning ............................................. 73
Acid Rain .................................................. 75
Stratospheric Ozone Denletion ................................... 76
Tropospheric Ozone ........................................... 77
Particle Deposition ........................................... ' 78
VIII. PROPOSED COMPUTER ALGORITHM FOR CALCULATING THE TRI
INDICATORS ............................................... 78
General Data Input ........................................... 79
Modeling Process ............................................. 85
Indicators Summation Process ................................... 93
IX. ISSUES FOR FUTURE CONSIDERATION ........................ . 93
X. CONCLUSION .............................................. 95
XI. REFERENCES .............................................. 96
Appendix A. Survey of Ranking and Scoring Systems ....................... A-l
Appendix B. Options for a TRI Indicator Ranking/Scoring System ............. B-l
Appendix C Available Toxirity Data for TRI Chemicals ..................... C-l
Appendix D. Review of Data Sources ................................... D-l
Appendix E. Waste Volumes by Industry ................................ E-l
Appendix F. Options for Indicator Computation and Normalization ............ F-l
Appendix G. Understanding the TRI Indicator Modelling Approach ........... G-l
r»T» * r*r T»T«iV»t»T r»O VOT r'l'I'L1 OO OT'OTF
-------
EXECUTIVE SUMMARY
In 1989, the Environmental Protection Agency (EPA) initiated an effort to focus
resources on regulatory or other programs that have the greatest potential to achieve
reductions in health or environmental risks. As pan of this effort, the Agency began to
explore ways to evaluate its successes in reducing risks, an effort which includes the
development of indicators of environmental progress.
The Office of Pollution Prevention and Toxics (OPFT) is in the process of developing
indicators of the impacts of chemical emissions on human health and the environment over
time. The indicators being developed by OPFT are based on existing data sources that
reflect multimedia trends in environmental contaminant releases, and designed to be flexible
enough to incorporate additional chemicals and sources as more information becomes
available.
The database selected as the basis for OPPTs environmental indicators is the Toxics
Release Inventory (TRI), which contains annual reports documenting multimedia
environmental releases and off-site transfers for over 300 toxic chemicals from U.S.
manufacturing facilities and which are of concern to the Agency1. Although the TRI
database does not capture all chemicals or industry sectors of concern to both OPPT and
the Agency as a whole, the database is the Agency's single best source of consistently
reported emissions data
The TRI Environmental Indicators may eventually consist of a set of four indicators,
chronic and acute human health impacts, and chronic and acute ecological impacts. The
focus of this report is the development of indicators of chronic effects to human health and
to aquatic life. The development of corresponding acute effects indicators is not feasible
now, since key data to support such indicators are not available. To appropriately evaluate
potential acute effects, one would need to know the distribution of releases over time (peak
release data), and these data are not currently reported to TRL
The ultimate goal of the indicator effort is to devise a .measure that reflects the
impacts of chemical releases and transfers, which can be used to assess progress in reducing
these impacts over time. The indicator methodology should be comprehensible, replicable,
and easily adaptable to a computer program. The methodology should also be sufficiently
flexible to facilitate separate analyses of the relative contribution of indicator components,
such as release media, industry sectors, chemicals, and geographic areas. The methodology
should also have provisions that either provide for uncertainty or provide alternatives when
'The TRI is mandated by the Superfund Amendment Reauthorization Act (SARA) Title
III Section 313.
-------
data are unavailable. The method should be sufficiently flexible to permit updates as new
chemicals/industries are added to TRI. Finally, all assumptions on which the indicator is
based should be easily identifiable and understood.
This report explains how the proposed TRI Environmental Indicators are constructed,
and includes discussions of the conceptual methodology, data sources, and computational
approach. The proposed indicators are based on risk-related scores, and the report
discusses the similarities and distinctions between the risk-based approach of the indicators
and conventional risk assessments.
In developing the proposed indicators, many approaches to assessing and ranking the
potential risk of chemicals were reviewed. Numerous techniques to score the relative
significance of TRI chemicals and facilities have been and continue to be developed. One
objective of this report is to explain the indicator to a variety of agencies and groups that
may wish to use it or to adapt the indicators or the methodologies to their own needs. A
related objective is to describe the benefits of the proposed approach to the indicator in
terms of flexibility, power as an analytic tool, and usefulness as a policy tool
GENERAL DESCRIPTION OF THE TRI ENVIRONMENTAL INDICATORS
The objective of the indicator methodology presented in this report is to construct
a unitless value that is related to the overall impacts of releases and transfers of all included
TRI chemicals by all facilities to each environmental medium at a given point in time. The
indicators are planned to be calculated each year based on the most current TRI data, and
compared to previous years' indicator values. One of the early years of TRI reporting, e.g.
1989 , would be selected as a "base year" and later years' indicator values compared to it
For the base year, the unitless score would be normalized to a round number such as 10,000;
subsequent years' data would be normalized by the same factor to provide a relative
comparison. This will allow assessment of the changes in estimated impacts of TRI releases
and transfers from year to year.
Four components are generally used to compute each indicator (e.g. for human
health chronic impacts). These are:
• the quantity of a chemical released or transferred,
• a tooticity weight,
• a weight reflecting the size of the potentially exposed population, and
• an exposure potential weight.
Separate assessments are made for each unique combination of a chemical, facility,
and release medium. For each of these releases or transfers, one develops an indicator
"element": a unitless value proportional to the potential impact of each specific release or
transfer. The value for each TRI indicator is simply the sum of all the indicator elements.
11
Y»T'TV"»T»TI T»/"» MOT /TI'Ii* f\J> OTTO "IT
-------
It must be emphasized that the TRI Environmental Indicator methodology is not
intended to be a quantitative risk assessment and does not calculate risk estimates. The
proposed methodology utilizes some of the tools often applied in quantitative assessments,
but in a relative way. The TRI indicator is by its nature only intended to reflect the
direction and the general magnitude of the change in releases over time, scaled by factors
(toxicity, exposure potential, receptor population size) that relate to potential risk. As such,
this indicator value has only relative rather than absolute meaning; it can only be used in
comparisons to other values at different points in time, or in identifying the relative size of
contributing factors.
Human Health Chronic Indicator
Each element of the Human Health Chronic Indicator (which is facility-chemical*
medium-specific) consists of three components: toxicity, exposure (which includes release
size), ?n<* the sigf of the exposed population. Components are multiplied to calculate the
element value. It must be reiterated that an element value is not an absolutely meaningful
measure of risk associated with the facility. These values are useful as an approximate
measure that is comparable to approximate measures for other facilities calculated using the
same methods, to identify relative significance.
The current proposal is focused on general populations: individuals, particularly
highly exposed individuals, are not the focus of the ir*^n*tor. Additional indicators based
upon nighty exposed or sensitive subpopulations may be developed in the future.
Ecological Chronic Indicator
For the Ecological (aquatic life) Chrome Indicator, the elements are developed by
combining aquatic toxicity and exposure components for each facility, each chemical, and
each mgHmm. AS with the human health chronic faHigatnr each component contributes
multiplicativery to the overall magnitude of the element value. For the Ecological Chronic
indicator, the population component is not utilized, sine? the f**y of the exposed population
is typically not known, and because of the difficulty in relating impacts to ecological systems
with many exposed populations of organisms.
SELECTION OF CHEMICALS TO INCLUDE IN THE INDICATORS
in order to be as inclusive as possible nnt^ still ™»fa*gfa computational tractability,
a large subset of TRI chemicals should be included in the indicator. Factors considered for
determining whether to include a chemical include the availability of necessary data to
compute elements , fln^ the magnitude of emissions. The following criteria are proposed
to exclude TRI chf-nvfk from the indicator:
Exclude chemicals with no reports in 1989 (56 chemicals) or only reports
listing zero releases and transfers (17 chemicals, for a total of 73);
111
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
Exclude chemicals with no readily available EPA-reviewed toxicity data (the
number of chemicals is not currently known, as toxicity data reviews have not
been completed).
Possibly exclude chemicals with total releases to each medium below some
threshold, e.g. 1000 Ib (about 14 chemicals). The initial list of chemicals
meeting this criterion is being reviewed to identify (and delete) ones which
may exhibit toxic effects at low concentrations.
METHODS FOR ADJUSTING EMISSIONS FOR TOXICITY AND EXPOSURE
To construct indicator elements that are related to risk, the emissions must be
adjusted in a manner that reflects the risks associated with each media-specific release or
transfer of each chemical. The risk potentially posed by a chemical emission depends on
the inherent toxicity of the chemical, the environmental fate and transport of the chemical
in the medium to which it is released, the degree of contact between the contaminated
medium and the human or ecological receptors, and the size of exposed populations.
Differences in toxicity among chemicals, as well as differences in environmental fate and in
the size and characteristics of populations potentially exposed influence the relative
contribution each emission makes to each indicator. Transfers to ofiEtite locations such as
sewage treatment plants (POTWs) require an additional estimate of the impact of treatment
technology on the emissions.
In order to incorporate all of these factors into comparable indicator elements, each
element's value is based on the release data, adjusted by weights reflecting the chemical's
toxicity relative to other chemicals, the potential exposure caused by the emission, and the
size of the potential exposed population. There are different equations for the human
health and the ecological indicators, because each has different types of data available and
different requirements for comparing the effect of releases of TRI chemicals. Both human
and ecological indicator elements incorporate estimates of the uncertainties associated with
component weights. By using toxicity, exposure potential, and population size adjustments
to the emission, the changes in the indicators over time should be more closely related to
changes in risk than changes in emission quantities alone.
The toxicity and exposure factors that are used to adjust the releases are referred to
as weights. Briefly, the weights used to construct the TRI Environmental Indicators are
derived by classifying a release in terms of ranges of toxicity, exposure, and population
exposed, and a«igning a weight value to each range. The weight assigned to each range
increases proportionally as toxicity or exposure increase. The methodology for deriving
weights is discussed in more detail below. It must be reiterated that while the resulting
indicators are intended to relate to risk by considering toxicity, exposure and population
factors, the indicators do not provide actual risk estimates.
IV
•oT-T»r»t>T _ r»n VOT t "rrfc* OP OTYVTF
-------
Toxicity Weights - Human
For both carcinogenic and noncardnogenic chronic effects, the proposed methodology
for developing human health toxicity weights is based on scoring systems that include both
qualitative and quantitative elements. There is uncertainty inherent in determining the
potency of chemicals in causing human health effects, thus any quantitative potency data
must be considered in the context of a qualitative classification of the uncertainty associated
with that data. Therefore, the toxicity of each chemical is classified both quantitatively using
potency data and qualitatively using weight-of-evidence (WOE). Both kinds of data are
considered when deriving the score, and so both kinds of data must be generated for the
TRI chemicals.
For human health, a number of quantitative measures of potency can be used for
toxicity weighting. Commonly used quantitative measures include the Oj", which describes
cancer potency; reference doses (RfDs) for non-carcinogens; and reportable quantities
(RQs). The q,"s, and RfDs are toxicity measures that have been developed for EPA risk
evaluation purposes. RQs have been developed using similar toxicity data reviews to
mandate reporting of releases to EPA at various release volumes. One or more of these
values are available for 247 TRI chemicals. There are some TRI chemicals for which there
are currently no formal EPA toxicity evaluations; for these chemicals, potential toxicity will
be evaluated based on toxicity data available from other data sources, as well as quantitative
structure activity relationship methods, to supplement the formal available EPA toxicity
values. For certain chemicals of extraordinary interest, special resources may have to be
made available to perform an adequate toxicity evaluation. Chemicals for which little if any
toxicity data exist, nn^ thus no evaluation of toxicity is possible, will be excluded from the
The second giemgnt of the h"«tian health toxicity weighting is the qualitative weight-
of-evidence (WOE) classification schemes. There is an exception to this method, however.
For chemicals with RfDs, WOE is considered in the development of an RfD; thus,
consideration of WOE is already imbedded in the RfD value. Therefore, WOE is not used
in this methodology for ««ignmg toxicity weights for chemicals with RfDs. WOE
information must be generated for carcinogens and for noncarcinogens without RfDs. The
WOE schemes proposed in this report have been adopted from several sources. For cancer,
the EPA Cancer Risk Assessment Guidelines2 and the TSCA Chemical Scoring System3
'U.S. Environmental Protection Agency (EPA). 1986a. GnidgKqfff fey ^prcinoeen Risk
Assessment. 51 Federal Register 33992 (September 24, 1986), currently being revised.
'O'Bryan, T.R. and RJL Ross. 1988. Chemical Scoring System for Pyflird flad Exposure
Identification. Journal of Toxicology and Environmental Health, 1:119-134.
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
WOE schemes are used. Non-carcinogenic WOE guidelines from the TSCA Chemical
Scoring System and from various EPA noncancer risk assessment guidelines are proposed
for use for non-carcinogens without RfDs.
Using the quantitative toxicity data and qualitative weight of evidence information,
chemicals are assigned to toxitity ranges and WOE categories. A matrix, with WOE
categories on one axis and toxicity ranges on the other, are used to assign weights to each
chemical Separate matrices for carcinogens and noncardnogens are used for human health
toxicity. Proportional weights - reflecting the magnitude of differences in toxicity between
chemicals — are assigned to the matrix cells. Weights increase by an order of magnitude for
each order of magnitude increase in toxicity and for each increase in WOE category. For
a chemical that causes multiple toxic effects, the TRI indicator will weight a chemical based
on the endpoint associated with the lowest dose at which an effect occurs, but will not apply
additional weights for severity relative to other health effects or for the number of effects
caused.
Toxicity Weights - Ecological
For ecological effects, the Indicator will focus on aquatic life impacts. Very little
data are available for most chemicals on effects to terrestrial or avian species, and the
human health indicator will provide some predictor of these. For aquatic effects, WOE
qualifications of potency data are not used because direct quantitative evidence of toxicity
is available in the form of ambient water quality criteria (AWQQ, LC^s (lethal
concentration, SO percent), no observable adverse effect levels (NOAELs), and aquatic RQs.
Direct measures of bioaccumulation potential may be available in the form of laboratory-
derived bioconcentration factors (BCFs) or bioaccumulation factors (BAFs). Where
laboratory data are unavailable, one of two indicators of bioaociTgnn*nt*CTifi potential ran be
used: either a measure of the potential for organic chemicals to partition between organic
material and water, or the water solubility of the chemical.
For aquatic effects, two matrices will be used to develop separate weights for toxicity
and bioaccumulation. Weights from each matrix are then multiplied for a final weight.
Both matrices assign weights that increase by increments of an order of magnitude as
toxicity ?nd hiftmTM*1111*0**"" potential increase.
Exposure Weights — Human
For human exposure weighting, as with toxicity weighting, both quantitative and
qualitative (uncertainty) elements are considered when weighting exposure potential. Three
steps are involved in deriving the human exposure weight The first step is to categorize
some measure of exposure potential into numerical ranges. The second step is to categorize
the levels of uncertainty associated with the measures of exposure potential The final step
uses' a matrix, with numerical exposure potential ranges arrayed along one axis and
VI
A TT vrrowT . no VOT PITT OP OTTOTir
-------
uncertainty categories arrayed along the other, to assign increasing weights to higher-
certainty/higher exposure potential releases.
For the first step of this process, quantitative measures of exposure potential must
be estimated. To do so, several existing scoring systems take the approach of ordinally
ranking the volume of each release by the some physical measure of the chemical's ability
to move through the environmental medium into which it is released. However, because
the exposure potential rankings would have different physical meanings for different
pathways, comparisons among releases to different media would be difficult, and weighted
releases from different pathways could not be added to obtain a single indicator value.
In this methodology, comparisons across media can be made because a common
quantitative exposure measure for each medium is derived: an estimate of "surrogate dose" •
. a measure related to the amount of chemical contacted by an individual per kg body
weight per day. Although limited facility-specific data and the use of generic models
(described below) prevent the calculation of an actual dose, the surrogate dose measure
developed from the available data and from generic modeling is designed to be, in concept,
comparable for all media.
To estimate the magnitude of exposure potential from TRI releases, the methodology
uses generic environmental fate and transport models. A separate exposure evaluation is
conducted for each release pathway (e.&, air, land, surface water). The exposure
evaluations combine data on media-specific g™«n«n volumes, physicochemical properties,
and where available, site characteristics with generic exposure models to determine an
estimate of the magnitude of the ambient concentration of iwitamtfiant jn the medium into
which the chemical is released. (Again, the use of submitter-estimated TRI emission data
and generic models with many default assumptions make this only a surrogate related to
actual concentration). For the tni™«"» health indicator, the ambient tn*Hia concentrations
are then combined with standard H"»*M»" exposure •«BmiptimiK to es**"10** the magnitude
of the surrogate dose.
It must be emphasized that while this methodology uses the EPA exposure
assessment paradigm to evaluate exposure potential, the results should not be construed as
an actual absolute numerical estimate of dose resulting from TRI releases. Instead, the
purpose is to obtain an order of magnitude estimate of surrogate dose resulting from release
of TRI chemicals relative to the surrogate dose resulting pom other releases included in the
indicator, so that these releases can be weighted appropriately in the indicator.
The modeling approach proposed by this methodology to obtain surrogate ambient
concentrations and surrogate dose measures for each pathway is summarized below.
Stack aiflfl Fugitive Air Emissions! The Gaussian dispersion algorithm from
the Industrial Source Complex Long Term (ISCLT) model, with assumed
stack height, Stability Array (STAR) weather data, and specified distances
v
TVT» ATTP T»Trnr»l>T T»r> *M"»T / 'I'l'l1 r»t> OTTOTT7
-------
from the source is used to estimate the surrogate ambient concentrations of
contaminants. For human exposure, the surrogate dose measure is estimated
by combining population-weighted average air concentration with standard
assumptions for inhalation rate and body weight
Direct Surface Water Releases: Stream flow data from the GAGE database
is combined with release data and chemical half-lives to derive surrogate
concentrations of chemicals in surface water. For human exposure, the
concentration is combined with data on populations served by drinking water
utilities with downstream intakes, bioconcentration factors, standard water
ingestion rates, fish ingestion rates, and body weight assumptions to derive the
surrogate dose.
Tand Releases* Volatilization and groundwater pathways were
considered in deriving the indices. Volatilization release data are reported
by the facilities with other air release data and are thus modeled with other
fugitive air releases from the facility. Groundwater contamination is modeled
only for disposal of TRI chemicals to onsite nonhazardous solid waste
management units. (Chemicals contained in solid waste streams classified as
hazardous are subject to additional disposal provisions which are assumed to
significantly limit potential for groundwater release and human exposure).
Leachate from landfills is modeled as a function of estimated concentration
of contaminant in the landfill, the soil/water partition coefficient, and the bulk
density of landfill material The magnitude of dilution and attenuation of
contammants as they move toward drinking water wells are estimated
according to Mating EPA dilution and attenuation factors (DAFs). The
surrogate dose measure is estimated for the population of well-water drinkers
within a Specified dfctange of the landfill site.
Transfers to POTWs: The fate of chemicals entering POTWs is determined
according to existing estimates of POTW removal efficiencies and within-
POTW partitioning rates (Le., the amount of contaminant that biodegrades,
volatilizes, or is adsorbed to sewage sludge). Hie exposure to chemicals that
are not removed by the POTW are modeled according to the methodology for
direct surface water releases, while volatilized chemicals are modeled as area-
source air releases.
Other Off-Site Tnipsfe"'- For off-site transfers, methodologies for modeling
exposure are available for two disposal technologies: waste incineration and
landfiiling. Air releases from incineration are modeled in the same manner
as other air releases, using removal efficiencies for nonhazardous waste
incinerators. For landfills, groundwater exposures are modeled in the same
planner as on-site disposal faculties. Unlike volatilisation emissions from
onsite facilities, volatilization emissions from off-site facilities arc net
viu
OTTOT1?
-------
reported, and must be estimated, then modeled for exposure potential. Once
estimated, volatile emissions are then modeled in the same way as other air
releases.
Investigations are continuing to examine methods for assessing underground injection,
both on- and off-site, as well as other treatment methods used for off-site transfers. For the
initial calculations of the TRI Environmental Indicators, these pathways will likely be
considered as contributing negligible amounts to the exposure weight
For the second step of the exposure weighting process, uncertainty categories are
created by considering the relative uncertainties associated with modeling different exposure
pathways, each of which will have a different level of modeling precision and dependability
of site-specific data associated with it. This method uses three qualitative categories of
increasing uncertainty to weight exposure estimates. The more uncertainty associated with
an estimate, the less weight given the exposure estimate, with an order of magnitude
difference between the categories with the highest and lowest estimates. (Hie middle
uncertainty category is given a weight twice that of the category with the most uncertainty.)
Exposure Weights - Ecological
For aquatic tife exposure to TRI chemicals, a weight is •««jgMH to each one of
several categories of exposure, with order of •Mprftmte ranges of ambient water
concentrations of the chemical resulting from a release used to define potential exposure.
The weights assigned to the ranges increase by an order of magnitude with each mer*»$ii}g
range of contaminant concentrations.
Aquatic exposure weights for categories of increasing estimated water concentration
do not need to reflect exposure relative uncertainty, since all estimates are derived using the
same methodology. The weights simply increase by an order of magnitude for each
category.
Population Size Weight!
This methodology proposes several rules for the use of population size data in the
exposure evaluation. First, a minimum population vahie is assumed to all releases to avoid
undervaluing sparsely distributed or rural populations. Second, both qualitative data (on the
uncertainty of the population size estimate) and quantitative data are used when evaluating
the SITP of the population potentially exposed to TRI releas
To reflect the uncertainty associated with the size of exposed human populations,
population sizes are assigned to one of three qualitative categories of uncertainty, and are
reduced by an order of magnitude for each level of uncertainty. All population sizes are
rounded up to the nearest 1,000 persons, thus creating a minimum exposed population.
IX
-------
Because of the major difficulties in estimating sizes of the populations of ecological
receptors, the TRI Ecological Indicator does not include a population weight In effect, this
approach assumes that all aquatic emissions occur to equally vulnerable locations.
COMPUTATION OF THE INDICATORS
Each indicator's value is generated by a computer algorithm which links input files
of releases and other input data required to mathematical models of media-specific pathway
releases, which are then linked to toxicity and exposure weighting factors to produce
indicator elements. The program generates up to 500,000 such elements , which are
summed to produce the values for each of the four indicators.
The elements can also be combined and evaluated in other ways than annual
comparisons. The detailed calculations used to create elements allow computation of
subindicators for individual chemicals, geographic regions, industry sectors, facilities, or
exposure pathways. These subindicators, like the overall indicator, cannot be compared to
some absolute level of concern, but could help identify the relative contribution to various
components to the overall estimate of emission impacts.
ISSUES FOR FUTURE CONSIDERATION
There are two general types of issues: specific methodological questions for the
indicators developed to date, and development of additional indicators. The methodological
questions associated with the indicators developed to date that still need to be resolved
include the following:
• How to address additions/deletions to the TRI chemical list, and whether to
create new indicators when significant changes are made;
• How to include deposition of TRI chemicals into underground injection wells
in the indicators;
• For noncarcinogenic chemicals without RfDs, what specific WOE systems
should be used for neurotoxicity, reproductive toxicity, and other chronic
toxicity endpoints;
• For ofErite transfers, how to better match TRI transfers to particular
treatment practices (e.g., which TRI chemicals are sent to hazardous or
nonhazardous waste management facilities; "***- of specific treatment practices
at POTWs; and how to estimate the potential impact of non-landfill, non-
incineration treatment types); and
• Using initial results of indicator computations, »«»tniti» the impact of various
exposure weights and their WOE modifiers.
TIP A FT PFPOWT - DO NOT CITE OR QUOTE
-------
The possible development of indicators for acute human health effects and acute
aquatic effects depends on improved toricity and other data. In addition, EPA considered
expanding the TRI Environmental Indicators to reflect indirect health and environmental
risks from TRI chemicals, such as global climate change, acid deposition, stratospheric ozone
depletion, tropospheric ozone formation, and paniculate deposition. While many of these
impacts have health-related effects, the complexity and uncertainty in modeling them may
make it impossible to incorporate them into the present set of indicators.
CONCLUSION
As an indication of improvements in environmental quality over time, the Human
Health and Ecological indicators provide the EPA with a valuable tool to measure general
trends related to impacts of TRI chemicals. They do not capture all environmental releases
of concern, and thus changes in these indicators over time cannot be used to evaluate
changes in other sources of environmental pollution. These indicators do, however,
generally relate changes in releases to relative changes in chronic human health and
ecological risks from a large number of toxic chemicals of concern to the Agency.
Importantly, the indicators also provide an ability to analyze the relative contribution of
source categories to important changes.
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
I. INTRODUCTION
In 1989, EPA Administrator William Reilly outlined the goals and processes for
establishing a strategic planning and accountability process at the Agency. Underlying this
approach is the Agency's desire to set priorities and shift resources to areas with the greatest
opportunity to achieve health risk and environmental risk reductions. As pan of this
initiative, the Administrator set forth a plan to develop indicators of environmental progress
in achieving the Agency's goals. Tracking this progress will allow the Agency to measure
its successes in implementing environmental protection and pollution prevention programs,
and to formulate strategic plans for improving the course of future environmental progress.
Within the Office of Pollution Prevention and Toxics (OPPT), environmental
indicators were identified as a component of the Toxic Substances Program Strategy 1992 -
1995. The strategy described an indicator that would reflect environmental exposures as the
Dow Jones Average reflects, the behavior of the New York Stock Exchange. The strategy
envisioned that the indicator will provide an estimate of emissions reductions across media,
including reductions that are voluntary on the part of industry and those that are a result
of government actions.
In general terms, such an indicator will allow EPA to track changes in human health
and environmental impacts from emissions over *i*"g As such, the Agency will need to take
advantage of y*fcrinp flats sources that reflect ***"ititn*Hip trends in environmental
contaminant releases. One such database, the Toxic Release Inventory (TRI), is the OPPTs
most relevant source of data for developing an annual indicator of environmental progress.
TRI is ma-nAttt^A foy the Supexfund AmMiflmem Reauthorization Act (SARA) Title m
Section 313 and reouires that U.S. ^nam^t^ftnTmo facilities file annual reports documenting
multimedia environmental releases and off-site transfers for over 300 chemicals which are
of concern to the Agency. Therefore, any indicator built from TRI data will be primarily
an indicator for the manufacturing sector. However, despite the fact that the TRI database
does not capture all chemicals, industrial sectors, or releases of concern to both OPPT and
the Agency as a whole, the database is the Agency's single best source of consistently
reported release data. Moreover, SARA Section 313 explicitly provides for the expansion
of TRI to cover additional chemicals and industries, and a major effort is currently
underway to add key chemicals and industries to TRL Furthermore, other program offices
in the Agency are in the process of developing indicators that will complement the TRI
indicator.
The TRI "indicator" may eventually consist of a set of four indicators. The focus of
this report is the development of indicators of chronic human health impacts and aquatic
life impacts. The development of corresponding acute effects indicators is not feasible now,
since the data to support such indicators are not available. To appropriately evaluate
potential acute effects, one would need to know the distribution of releases over time (peak
release data), and these data are not currently reported to TRL However, possible changes
TIP AUT PFPOPT - DO NOT < TIT OP OTTOTF
-------
in reporting requirements may allow for the development of separate acute indicators for
human and ecological effects.
The ultimate goal of the indicator effort is to devise a measure that reflects the
impacts of chemical releases and transfers, which can be used to assess progress in reducing
these impacts over time. Hie indicator methodology should be comprehensible, replicable,
and easily adaptable to a computer program. The methodology should also be sufficiently
flexible to facilitate separate analyses of indicator components, including specific media,
industrial sectors, chemicals, and geographic areas. The methodology should also have
provisions that either provide for uncertainty or provide alternatives when data are
unavailable. The method should be sufficiently flexible to permit updates as new
chemicals/industries are added to TRI. Finally, all assumptions on which the indicators are
based should be easily identifiable and understood.
This report explains how the proposed TRI Environmental Indicators are constructed,
and includes discussions of the conceptual methodology, data sources, and computational
approach. The proposed indicators are based on risk-related scores, and the report
discusses the similarities and distinctions between the risk-based approach of the indicators
and conventional risk assessments.
In developing the proposed indicators, many approaches to "«*«"*c ***& ranking the
potential impact of chemicals were reviewed. Numerous techniques to score the relative
significance of TRI chemicals and facilities have been and continue to be developed. One
objective of this report is to explain the indicators to a variety of agencies and groups that
may wish to use or adapt the indicators or the methodologies to their own needs. A related
objective is to describe the benefits of the proposed approach to the indicators in terms of
flexibility, power as an analytical tool, and usefulness as a policy tool
II. GENERAL DESCRIPTION OF THE TRI ENVIRONMENTAL INDICATORS
The objective of the indicator methodology presented in this report is to construct
a unitless vahie related to the overall impacts of releases and transfers of all included TRI
chemicals by all facilities to each environmental medium at a given point in time. The
indicators are planned to be calculated each year based on the most current TRI data, and
compared to previous years' indicator values. One of the early years of TRI reporting, e.g.
1989, would be selected as a "base year" and later years' indicator values compared to it.
For the base year, the unitless score would be normalized to a round number such as 10,000;
subsequent years' data would be normalized by the same factor to provide a relative
comparison. This will allow assessment of the changes in estimated impacts of TRI releases
and transfers from year to year.
-------
Four components are generally used to compute each indicator (e.g. for human
health chronic impacts). These are:
• the quantity of a chemical released or transferred,
• a toxicity adjustment,
• an adjustment reflecting the size of the potentially exposed population, and
• an exposure potential adjustment
Separate assessments are made for each unique combination of a chemical, facility,
and release medium. For each of these releases or transfers, one develops an indicator
"element": a unitless value proportional to the potential impact of each specific release or
transfer. The value for each TRI indicator is simply the sum of all the indicator elements.
It must be emphasized that the TRI Environmental Indicator methodology is not
intended to be a quantitative risk assessment and does not calculate risk estimates. The
proposed methodology utilizes some of the tools often applied in quantitative assessments,
but in a relative way. Hie TRI Environmental Indicators are by their nature only intended
to reflect the direction and the general magnitude of the change in releases over time,
scaled by factors (toxicity, exposure potential, receptor population size) that relate to
potential risk. As such, an indicator value has only relative rather than absolute
it can only be used in comparisons to other values at different points in time, or in
identifying the relative size of contributing factors.
Human Health Chronic Indicator
Each element of the Human Health Chronic Indicator (each of which is facility-
chemical-medhim-specific) consists of three components: tenacity, exposure (which includes
release size), and the size of the exposed population. Components are multiplied to
calculate the element It must be reiterated that an element value is not an absolutely
meaningful measure of risk associated with the facility. These values are useful as an
approximate measure that is comparable to apnrmrimate measures Cor other facilities
calculated using the same methods, to identity relative significance.
The current proposal is focused on general populations: individuals, particularly
highly exposed individuals, are not the focus of the indicator. Additional indicators based
upon highly exposed or sensitive subpopulations may be developed in the future.
Ecological Chronic Indicator
For the Ecological (aquatic life) Chronic Indicator, the elements are developed by
combining aquatic toxicity and exposure components for each faculty, each chemical, and
each medium. As with the Human Health Chronic Indicator, each component contributes
multiplicativety to the overall magnitude of the element value. For the Ecological Chronic
Indicator, the population component is not utilized, since the size of the exposed population
-------
is typically not known, and because of the difficulty in relating impacts to ecological systems
with many exposed populations of organisms.
There are several factors that must be considered in constructing the indicators: these
include selecting chemicals for inclusion, designing a mechanism for adjusting ^missions to
reflect exposure and toritiry, and developing algorithms for calculating these adjustments.
The following sections describe the elements of the proposed TRI indicators. A conceptual
flow chan of the proposed TRI Environmental Indicator methodology is presented in Figure
1.
III. SELECTION OF CHEMICALS AND FACILITIES TO INCLUDE IN THE
INDICATORS
Selection of Chemicals
In order to be as inclusive as possible and still «M»«*«fa computational tractability,
a large subset of TRI chemicals should be included in the indicators. However, several
factors must be considered when determining whether to include a chemical in the
indicators calculation. The number of chemicals included in the indicators could be limited
by the availability of data characterizing the chcrnifal or the magnitude of emissions.
The TRI Environmental ^nrT|'n>t^r Work Group initially identified the following list
of criteria for selecting chemicals to exclude from the TRI indicators;
• Criterion lr Exclude che*"icajswift_norcPQr^HE_or_2pro reporting, For 1989
TRI data, these criteria alone would exclude 73 chemicals and over 2000
fayjii^tfft from flfMMJdgrafiqn^ leaving 244 rfuMiifoaiy and over 20.000 faculties
included in. the TRI indicators nna*yfffc The chemicals excluded by these
criteria are listed in Table 1. These criteria would allow the Agency to
concentrate resources (such as the research on toxicity and environmental fate
data) on thostf c^yn^ic?** that could actually ^rosf & potential impact* One
problem with this approach is that the potential future increase of emissions
of these chemicals could contribute to health and environmental impacts.
However, reporting of increased emissions of chemicals could result in their
addition to the indicators. This issue is covered in a later discussion of the
normalization of the indicators. Because it is possible that some facilities may
: incorrectly report zero releases, this criterion could be limited to excluding
chemicals with no reporting. In this case, the criterion would exclude 53
ghemiryic fmm the indicators analysis, but no faculties, y*np<> if chemicals have
no reports, presumably no facility in the database would have only these
chemicals on its form R.
-------
Figure 1
Methodology for Calculating Human Health TRI Indicator
o
1
s
§
offsite transfers
{POTW transfers
1
ofTslte
Incineration
direct air
emissions
.»
POTW sludge
r iitiiaaaitlnii
IIPUIPC1 •Uvu
^
1
fugitive air
emissions
1
Air *-
i
onslte land disposal;
Includes
underground Injection
•i Volatilisation
i
^
'i
|
i
1
i
Ground
Water
offciU
andergroi
offsite transfers
POTW transfers
>land
;lncludc9
ind Injection
POTW sludge
land disposal
direct water
discharges
1
Sur
i
Exposure Evaluation
i
c
face Water *~
POTW
transfers
•OTW
ffluenl
population data
toxicity data
Exposure Weight
Population Weight
Toxldty Weight
1
Air
Index
ftcDhr,
chemical
Groundwater
Index fae|m
1
Surface Water
Inde* heUHr,
chemical
Sum over all facilities and chemicals
TRI Indicator
-------
Table 1
Chemicals With No Reports or No Reported Releases (73 Total)
Shaded Chemicals Were Reported but With,No Releases (20 Total)
Total number of chemicals(73) and facilities (>2000)
1989 Toxics Release Inventory Data
CAS Chemical
S1752 NITROGEN MUSTARD
53963 ACETYLAMINOFLUORENE4
55185 NITROSODIETHYLAMINE.N
51125 CYANIDE
57578 PROPIOLACTONEJETA
59892 NTTROSOMORPHOLINE.N
60093
60117
60355
62759
81889
92671
95954
96128
97563
99592
100754
119904
AMJNOAZOBENZENE.4
DIMETHYLAMINOAZOBENZENE.4
ACETAMIDE
NTTROSODIMETHYLAMtNEJI
68768 TRIAZIQUONE
75274 . DICHLOROBROMOMETHANE
79447 DIMETHYLCARBAMYL CHLORIDE
CJ. FOOD RED 15
J
82280 1-AMIN02-ME1HYL-ANTHRAQUINONE
91598 NAPHTHYLAMINE.BETA
AMINODSPUENYL.4
J
92875 BENZIDINE
92933 NITROBIPHENYM
TRICOLORO/PHENOL, 2,4,5 -
DIBROMOCHLOROPROPANE (DBCB)
CJ. SOLVENT YELLOW3
NTTROOANISIDmE.5
NTTROSOPIPERIDINEJf
101804
104949
107302
114261
115531
117793
DlAMtHODtPHEtrrLEIOBR. 4.4
AMSW1NEJ'
CHLOROMETHYL METHYL ETHER
PROPOXUR
DICOFOL
AMINOANTHRAQUINONEJ
DIMETHOXYBENZID1NE, 3,3
J
119937 DIMETHYLBENZIDINE,33
122667 DIPHENYLHYDRAZINE, U
126727 TRIS(2,3DIBROMOPROPYL)PHOSPHATC
128665 CL VAT YELLOW 4
134292 ANISIDINEHYDROCHLORIDE.O
135206 CUPFERRON
139651 TfflODIANILINE,4,4
309002 ALDRIN
334883 DIAZOMETIIANE
492808 AURAMINE
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
Table 1
Chemicals With No Reports or No Reported Releases (73 Total)
Shaded Chemicals Were Reported but With No Releases (20 Total)
Total number of chemicals(73) and facilities (>2000)
1989 Toxics Release Inventory Data
CAS
505602
510156
532274
542881
615054
621647
636215
680319
684935
759739
842079
924163
989388
1120714
1335871
1464535
1836755
1937377
2234131
2303164
2602462
2832408
3118976
3761533
4549400
4680788
7440280
8001352
16071866
. 16543558
20816120
39156417
Chemical
MUSTARD GAS
CHLOROBENZELATE
CHLOROACETOPHENONE
BISfCHLOROMETVYL)ETHER
DIAMINOAN1SOLEA4
NirROSODInPROPYLAMINEJ*
TOLUIDINE HYDROCHLORIOE
HEXAMETHYLPHOSPHORAMIDE
NITROSONMEIHYLUREAJt
NTTROSONETHYLUREA.N
CJ. SOLVENT YELLOW 14
NTTROSODInBUTYLAMINEJ*
CJ. BASIC RED 1
PROPANE SULTONE
HEXACHLORONAPHTHALENE
DffiPOXYBUTANE
NTTROFEN
CJ. DIRECT BLACK 38
OCTACHLORONAPHTAHLENE
D1ALLATE
CJ. DIRECT BLUE 6
CJ. DISPERSE YELLOWS
CJ. SOLVENT ORANGE 7
.CJ. FOOD RED 5
NrTROSOMETHYLVINYLAMINEJ*
CJ. ACID GREEN 3
THALLIUM
TOXAPHENE
CJ. DIRECT BROWN 95
NITROSONORNICOTINEjr
OSMIUM TETROXIDE
DIAMINOANISOLESULFATB. 2.4
T»TIIV"»I>T T»O TWIT f'I't'V OP OTTOTF
-------
Criterion 2: Exclude chemicals with relatively low reporting (between 1 and
1.000 Ibs to each medium: air, water, land, off-site transfers, transfers to
POTWg). Excluding these chemicals would allow us to focus the indicators
on chemicals causing the greatest impacts. The danger of this criterion is the
possibility of excluding a high-toxicity but low-volume release that may
contribute substantially to health and environmental impacts. Another
problem with this approach is the potential future increase of emissions of
these chemicals that could contribute to risk. This criterion would exclude 12
chemicals but fewer than 5 faculties from consideration. Those chemicals that
would be excluded by this criterion are found in Table 2. Some Work Group
members suggested increasing the cutoff for off-site transfers to 25,000 Ibs,
since such transfers may undergo treatment before release to the environment.
Increasing the exclusion criterion for offsite transfers to less than 25,000 Ibs.
would exclude 19 chemicals and about 10 facilities.
Criterion 3: Exclude chffnNCTiif with relatively low reporting fbeywycfl 1 pnd
1.000 Ib to all media combined! air, ^flter. lftnd. off-site transfer* transfers to
POTWsi. Similar to the previous criterion, this would allow us to concentrate
on chemicals causing the greatest impacts, but may exclude high toxicity,'low
volume releases and encourage future emission increases that would
contribute to environmental impacts. This criteria would exclude 7 chemicals
but fewer **""* 5 facilities from consideration. Those chemicals excluded by
this criterion are found in Table 3.
Criterion 4: Exclude chemicals having BP tpricitv data. As discussed briefly
above, the TRI Environmental Indicators will require incorporation of data
on the relative toxirity of chemicals. Chemicals must be scored based on
available toxicity data If there is insufficient information upon which to
evaluate the toxicity of a chemical, the chemical would be excluded from the
indicators. It should be emphasized that exclusion from the indicators would
be a last resort, used only after all available mechanisms for closing data gaps
(including structure activity relationship methods and best professional
judgment) are exhausted. While *Hi« criterion presents a danger of excluding
risky but poorly studied chemicals, the risk-related nature of the indicators
does not allow for inclusion of these chemicals. Table 4 presents the TRI
chemicals for which formal EPA toxicity data are not available in the
Integrated Risk Information System (IRIS) or in the Health Effects
Assessment Summary Tables, and for which specific health-based (chronic,
carcinogenic) Reportable Quantities have not been developed under
CERCLA. As shown in this table, there are 70 chemicals that have no readily
available toxicity data. Thirty-three of these 70 have zero or no reported
releases. Excluding these 70 chemicals would exclude about 250 facilities.
This table will be circulated among other OPPT divisions and other Agency
offices (such as the Pesticides office) to solicit additional available toxicity
8
-------
Table 2
Chemicals With TRI Releases Less Than 1,000 Pounds in Each Media
(Air, Water, Land, Off-site Transfers, Transfers to POTWs)
Total Number of Chemicals is 12 and Number of Facilities is Less than 5
1989 Toxics Release Inventory Data
CAS CHEMICAL
55210 BENZAMIDE
60344 METHYL HYDRAZINE
7243S METHOXYCHLOR
7SSS8 PROPYLENEIMINE
88891 PICRIC ACID (2,4,6TRINITROPHENOL)
94S97 SAFROLE
96093 STYRENE OXIDE
134327 NAPHTHYLAMINE, ALPHA
156105 NITROSODIPHENYLAMINE,P
593602 VINYL BROMIDE
10034932 HYDRAZINE SULFATE
12122677 ZINEB
VOt AFT REPORT - DO NOT CITE OR QUOTE
-------
Table 3
Chemicals With TRI Releases Between 1 and 1,000 Pounds Across All Media
(Air, Water, Land, Off-site Transfers, Transfers to POTWs)
Total Number of Chemicals is 7 and Number of Facilities is Less Than S
1989 Toxics Release Inventory Data
CAS CHEMICAL
60344 METHYL HYDRAZINE
75558 PROPYLENEIMINE
88891 PICRIC ACID (2.4.6TRINITROPHENOL)
94597 SAFROLE
134327 NAPHTHYLAMINE. ALPHA
156105 NTTROSODIPHENYLAMINE, P
10034932 HYDRAZINE SULFATE
10
-------
Table 4
C sals With No Widely Available EPA Toricity Data (IRIS, HEAST. human health RQs)
TL uumber of chemicals (70) and facilities (about 250)
Shaded chemicals have zero or no reported releases (31)
CAS
20075
SS210
55630
59892
60093
60355
64675
67630
68768
748S1
78842
81889
82280
87627
88891
90040
90948
91087
92671
92933
94360
9S636
96093
97S63
99S92
101688
101779
101804
104949
111422
115071
117793
120718
120809
123386
1237281
Chemical
COBALT COMPOUNDS
BENZAMIDE
NITROOLYCERIN
NITROSOMORPHOUNEJf
AMINOAZOBENZENE. 4
ACBTAMIDE
DIETHYL SULFATE
ISOPROPYL ALCOHOL FOR STRONG ACID PURPOSES
TRIAZIQUONB
ETHYLENE
ISOBUTYRALDEHYDE
C.I. FOOD RED 15
1-AMINO 2-METHYL-ANTHRAQUINONE
XYUDINEA6
PICRIC ACID (2.4.6TRINrrROPHENOL)
ANISIDINE.O
MICHLERS KETONE
TOLUENE2.6DIISOCYANATE
AMINODIPHENYL.4
NITROBIPHENYI,4
BENZOYL PEROXIDE
TRIMETHYLBENZENE. 10,4
STYRENE OXIDE
C.L SOLVENT YELLOW 3
NITROOANISIDINE. 5
METHYLENEBIS(PHENYLISOCYANATE)
METHYLENEDIANILJNE.4,4
DIAMDfODIPHENYLETBER. 4.4
ANISIDINEiP
DIETHANOLAMDfE
PROPYLENE (PROPENE)
AMINOANTHRAQUINOKEO
CRESIDINEJ
CATECHOL (IJ. DIHYDROXYBENZENE)
PROPIONALDEHYDE
BUTYRALDEHYDE
Comment
re: potential data
teratogen
NTP carcinogen
NTP eareinof en
potential carcinogen
potential carcinogen
potential carcinogen
potential carcinogen
. DO NOT CITE OR QUOTE
11
-------
Table 4
Chemicals With No Widely Available EPA Toxicity Data (IRIS, HEAST, human health RQs)
Total number of chemicals (70) and facilities (about 250)
Shaded chemicals have zero or no reported releases (31)
CAS
128665
132649
134292
135206
139139
139651
141322
156105
1S6627
334883
463S81
541413
969642
615054
680319
842079
989388
1313275
1314201
1335871
1836755
2234131
2832408
3118976
3761533
4680788
6484522
7440484
7664939
7783202
10049044
16543558
39156417
Chemical
C.I. VAT YELLOW 4
DIBENZOFURAN
ANISID1NE HYDROCHLORIDE.O
CUPFERRON
NITRILOTRIACETIC ACID
THIODIANILINE.4.4
BUTYL ACRYLATE
NITROSODIPHENYLAMINE. P
CALCIUM CYANAMIDE
DIAZOMETHANE
CARBONYL SULFIDE
ETHYL CHLOROFORMATE
C.L BASIC OREEN 4
DIAMINOANISOLEA4
HEXAMETHYLPROSPHORAMIDE
C.L SOLVENT YELLOW 14
CJ. BASIC RED 1
MOLYBDENUM TRIOXIDE
THORIUM DIOXIDE
HEXACHLORONAPHTHALENE
NFTROFEN
OCTACHLORONAPHTAHLENE
C.L DISPERSE YELLOW 3
C.I. SOLVENT ORANGE 7
C.L FOOD RED 5
C.L ACID OREEN 3
AMMONIUM NITRATE
COBALT
SULFURICACID
AMMONIUM SULFATE
CHLORINE DIOXIDE
NITROSONORNICOTINEJi
DIAMINOANISOLE SULFATE. 2.4
Comment
re: potential data
CAA Section 111
CAA - Sectioa 111 and TSCA 8a
CAA - Scetioa 111 and pettieide
CAA - Section 112
12
-------
data. For several of these chemicals (indicated in the comment column in
Table 4), toxicity data are likely to be available within OPTS. Remaining
data gaps could be filled with additional research and application of
professional judgment. Again, chemicals would be excluded based on this
criterion only after all means to fill data gaps had been exhausted. If new
toxicity data became available, the chemical could later be added to the
indicators (see Section VI for a discussion of normalizing the indicators with
the addition of new chemicals).
chemicals with low toxicity relative to other TRI chemicals allows us to focus
on thoroughly evaluating those chemicals with the highest potential for
contribution to the indicators. Criteria would be defined for "relatively low
toxicity," based on specific toxicity threshold values. This criterion may
exclude certain chemicals that have relatively low toxicity but have very high
volume releases. Table 5 lists chemicals that would be excluded if low human
health toxicity were defined as follows: (a) smallest reportable quantity equals
5,000 pounds; (b) oral and inhalation reference doses are greater than 0.1
mg/kg-day; or (c) oral and inhalation cancer potencies are less than 0.01 kg-
day/mg. These criteria would exclude 37 chemicals nT|^ about 3000 faculties
from consideration.
Exclude chemicals ^feting two, or rpore of these criteria. A combination of
the previous criteria would remove those chemicals with both relatively low
toxicities and low releases, thus avoiding the loopholes discussed, above.
Possible combinations include: yh»™«iiy with low reporting to each media
and low toxicity (Criterion 6); chemicals with low releases to all media
combined and low toxicity (Criterion 7); and chemicals with no reports or
zero releases and low toxicity (Criterion 8). This step would exclude those
chemicals of ™™»i*i concern; these chemicals would be subject to addition
in later years, if their reported releases increase above the established criteria.
Based on 1989 TRI data, however, no chemicals would be eliminated using
Criteria 6 and 7. Criterion 8 would exclude one chemical but no facilities
from consideration.
Possibly exchide "non-TSCA* yhgffyyflls OPPT may wish to focus its
indicators on those chemicals over which it has direct regulatory authority,
e.g., those chemicals listed on the TSCA inventory (40 CFR §372.65). By
doing so, the indicators could track OPPT program effectiveness; changes in
TSCA chemical releases would not be masked by increases in releases of
chemicals over which OPPT has no regulatory control For example, there
are a number of pesticides on the TRI roster (listed in Table 6), which may
exclusively have uses that are not included on the TSCA inventory. This
criterion would exchide few chemicals, since almost all chemicals in TRI are
13
A «r tf-FonoT . no NOT CITE OR QUOTE
-------
Table 5
Chemicals With Low Toricity
Total number of chemicals(37) and facilities (about 3000)
CAS Chemical
67561 METHANOL
67641 ACETONE
71363 BUTYL ALCOHOL, N
75003 CHLOROETHANE (ETHYL CHLORIDE)
75070 ACETALDEHYDE
75252 BROMOFORM (TRIBROMOMETHANE)
75274 DICHLOROBROMOMETHANE
75569 PROPYLENE OXIDE
78875 DICHLOROPROPANE, 1,2
78922 BUTYL ALCOHOL, SEC
79061 ACRYLAMIDE
79107 ACRYLIC ACID
84662 DIETHYL PHTHALATE
85449 PHTHALIC ANHYDRIDE
85687 BUTYL BENZYL PHTHALATE
86306 NITROSODIPHENYLAMINEJf
90437 PHENYLPHENOL.2
CUMENE
BENZAL CHLORIDE
100425 STYRENE
103231 BIS(2ETHYLHEXYL)ADIPATE
106898 EPICHLOROHYDRIN
107211 ETHYLENE GLYCOL
108101 METHYL ISOBUTYL KETONE
108316 MALEIC ANHYDRIDE
108883 TOLUENE
108952 PHENOL
120127 ANTHRACENE
131113 DIMETHYL PH1HALATE
133062 CAFTAN
140885 ETHYL ACRYLATE
1582098 TRIFLURAUN
7440508 COPPER
7440666 ZINC
7647010 HYDROCHLORIC ACID
7664382 PHOSPHORIC ACID
7782505 CHLORINE
Note: This list includes only t
that have EPA tozidty data.
14
TWATT TfVDfMfT . T»T» VOT
OT'OTP
-------
Table 6
Pesticides Listed in TRI (that may have uses not covered
under TSCA)
CAS
000052-68-6
000056-38-2
000057-74-9
000058-89-9
000062-73-7
000063-25-2
000072-43-5
000076-44-8
000082-68-8
000087-86-5
000088-06-2
000094-75-7
000095-95-4
000096-12-8
000106-93-4
000114-26-1
000115-32-2
000133-06-2
000133-90-4
000309—00—2
000510-15-6
000961-11-5
001582-09-8
001836-75-5
001897-45-6
002164-17-2
002303-16-4
008001-35-2
012122-67-7
012427-38-2
TRICHLOROFON
PARATfflON
CHLORDANE
UNDANE
DICHLORVOS
CARBARYL
METHOXYCHLOR
HEPTACHLOR
QUINTOZENE
PENTACHLOROPHENOL
TR1CHLOROPHENOL (2,4,6-)
2,4-D
2.4.5-TRICHLOROPHENOL
DIBROMO-3-CHLOROPROPANE (U-)
DIBROMOETHANE (1,2-)
PROPOXUR
DICOFOL
CAPTAN
CHLORAMBEN
ALDRIN
CHLOROBENZILATE
TETRACHLORVINPHOS
TRUFLURAUN
NTTROFEN
CHLOROTHALONIL
FLUOMETURON
DIALLATE
TOXAPHENE
ZINEB
MANEB
15
TVO A r«r
no WOT
OP OTTOTF
-------
subject (at least theoretically) to TSCA authority. Exclusion of these
chemicals could also diminish the usefulness of the indicators outside of
OPPT if this criterion leads to the exclusion of a chemical of concern to
another office.
Table 7 summarizes the number of chemicals and the number of faculties included
and excluded by Criteria 1 through 8, discussed above. The Work Group also considered
the idea of a selecting a random sample of the TRI chemicals for inclusion in the indicators.
However, it was determined that a large percentage of the TRI roster (over 200 chemicals
out of 317) would have to be selected in order to obtain an adequate representative sample
with a sufficiently moderate coefficient of variation. Therefore, the Work Group concluded
that randomly selecting chemicals would not be a fruitful exercise at this point. If, in the
future, the TRI reporting system is expanded to contain a much larger number of chemicals,
this option may be revisited.
In consideration of these criteria, this method proposes the exclusion of the following
chemicals from the TRI indicators calculation:
• chemicals with either no reports, or only reports listing zero releases and
transfers;
• chemicals with no available, EPA-reviewed toxicity data (but flag these); and
• possibly exclude chemicals with total releases to each ir*^"1*" below some
threshold (e.g^ 1000 pounds).
Note that chemicals of special interest to various offices in the Agency (such as
chemicals undergoing regulatory activity) may inadvertently be excluded from the indicators
based on one or more of these criteria. Once the final list of excluded chemicals is created,
it can be examined to ensure that chemicals of special concern to the Agency are not
excluded. In addition, other offices will be asked if there are chrmicfilif that we have
included that should in fact be excluded, such as pesticides that have not been nor will be
reregistered for use in the U.S.
Inclusion of Facilities
certain chemicals from the indicators will eliminate some facilities that
release these chemicals from the indicators. There may be additional reasons to exclude
certain facilities from the indicators. For example, the reliability of reporting from certain
facilities may be questionable. There may also be concerns about the resource and
computing requirements for including all facilities in the indicators. For these reasons, the
Work Group considered options where a subset of faculties would be selected for inclusion
in the indicators. Several statistical methods for selecting representative facilities were
considered. Ultimately, the Work Group decided to include all (acuities emitting the
16
-------
Table 7
J
i
j
j
*
5
j
ni> OTTriTF
• • VBW^^VWV* -**m m »«*• mmm • «*• m^tummmm ummmm »••• K ^* ••**••• •*»*• H>
Criteria
1 Chemicals with no or zero
reporting of TRI releases
2 Chemicals with TRI releases less
than 1,000 pounds in each media
3 Chemicab with TRI releases less
than 1,000 pounds across all media
4 Chemicals with no EPA toxicity data
5 Chemicals classified with low
toxicity
6 Chemicab meeting Criteria 2 and 5
7 Chemicab meeting Criteria 3 and 5
8 Chemicab meeting Criteria 1 and 5
Number of
Chemicab
Excluded
73
12
7
70
37
0
0
1
Number of
Chemicab
Remaining
244
305
247
243
280
317
317
316
Number of
Facilities
Excluded
~2200
<5
<5
~250
-3000
0
0
0
Number of
Facilities
Remaining
P
-20.300
-22,500
-22.500
-22.250
- 19.500
-22,500
-22,500
-22,500
17
-------
chemicals included in the indicators. This approach avoids questions of representativeness
of a selected subset of facilities. The Work Group also judged the inclusion of all facilities
to be feasible within the resource constraints given the proposed approach described below.
IV. METHODS FOR ADJUSTING EMISSIONS FOR TOXICITY AND EXPOSURE
To construct indicator elements that are related to risk, TRI emissions must be
adjusted in a manner that relates to the risks associated with each media-specific release or
transfer of each chemical The risk potentially posed by a chemical emission depends on
the inherent toxicity of the chemical, the environmental fate and transport of the chemical
in the medium to which it is released, the degree of contact between the contaminated
medium and the human or ecological receptors, and the size of exposed populations.
Differences in toxicity among chemicals, as well as differences in environmental fate and in
the size and characteristics of populations potentially exposed influence the relative
contribution each emission makes to each indicator. Transfers to ofisite locations such as
sewage treatment plants (POTWs) require an additional estimate of the impact of treatment
technology on the emissions.
In order to incorporate all of these factors into comparable indicator elements, each
element's value is based on the release data, adjusted to reflect the chemical's toxicity
relative to other chemicals, the potential exposure caused by the emission, and the size of
the potential exposed population. There are different equations for the human health and
the ecological indicators, because each has different types of data available and different
requirements for comparing the effect of releases of TRI chemicals. Both human and
ecological indicator elements incorporate estimates of the uncertainties associated with
component weights. By using toxicity, exposure potential, and population size adjustments
to the emission, the changes in the indicators over time should be more closely related to
changes in risk than changes in emission quantities alone. It must be reiterated that while
the resulting indicators are intended to relate to risk by considering toxicity, exposure and
population factors, the indicators do not provide actual risk estimates.
The toxicity and exposure factors that are used to adjust the releases are referred to
as weights. Briefly, the weights used to construct the TRI Environmental Indicators are
derived by classifying a release in terms of ranges of toxicity, exposure, and population
exposed, and assigning a weight value to each range. The weight assigned to each range
increases as toxicity or exposure increase. The methodology for deriving weights is discussed
in more detail below.
Several offices within the EPA and organizations outside the Agency have developed
systems for scoring or weighting chemicals based on potential toxicity and exposure. The
usual purpose of such exercises is to prioritize chemicals for further study or for closer
regulatory scrutiny. A review of chemical scoring and ranking procedures used by offices
within the Agency and by organizations outside the Agency is found in Appendix A
18
ra> A ITT PFPOPT . no WIT OTE OR OIIOTE
-------
Previous scoring systems have used a variety of methods to weight chemicals. The
actual numerical weights applied to chemicals can 'be qualitative, ordinal, weighted or
calculated, or some combination of these approaches. The relative severity of the effects
posed by chemicals can also be included, as can considerations of the quality of the toxicity
data and the exposure estimates. Based on our review of these scoring systems, several
options for a chemical evaluation methodology emerged. Alternative methods, and their
advantages and disadvantages, were considered by the TRI Indicator Work Group and are
summarized in Appendix B. The following section presents a proposed methodology based
on the research described in Appendices A and B and based on Work Group deliberations.
While the method described below contains elements of the options described in Appendix
B, the draft methodology combines these elements in a manner that is not presented
explicitly in the appendix.
General Format of the Weighting Scheme
Each TRI chemical will fall into one range of values for parameters of interest and
will receive the corresponding weight based upon its characteristics. The methodology
follows this categorization approach for both toxicity weighting and exposure weighting. A
chemical may cause toxic effects to more than one health endpoint; thus, different weights
could be assigned, depending on the endpoint considered. For this method, a chemical's
final toxicity weight will be the highest weight it receives among all endpoints considered
(Le., based on its most sensitive endpoint).
Using categorical weights for toxicity and exposure has several advantages when
compared to calrn^at^tiP specific, mrignft numerical weights for fn^miqyi releases. First,
unique weights would imply that we know the toxicity and exposure potential of the release
with enough precision to distinguish among relatively «**M»IJ differences in these values. In
fact, there are significant uncertainties associated with both the toxicity data and the
methods used to evaluate exposure potential Weighting a release based on the broad
categories of toxicity and exposure potential into which it falls avoids the impression of
precision where such precision does not exist Second, chemicals are likely to remain in the
broad toxicity category to which they are originally a«igng^t unless significant new and
different toxicity data become available. Thus, the weights applied to these chemicals are
not likely to be revised frequently, lending stability to the indicators over
The following sections describe the methods used for toxicity and exposure weighting
in both the human health and ecological TRI indicator calculations. The penultimate
section describes rules for using the size of the exposed population in the TRI human health
indicator calculation. The final section proposes a method for integrating the exposure,
toxicity, and, where appropriate, population values in the TRI indicators calculations.
19
-------
Tmricity Weights - Human
For both carcinogenic and noncarcinogenic chronic effects, the proposed methodology
for developing human health toxitity weights is based on scoring systems that include both
qualitative and quantitative elements. There is uncertainty inherent in determining the
potency of chemicals in causing human health effects, thus any quantitative potency data
must be considered in the context of a qualitative classification of the uncertainty associated
with that data. Therefore, the toxicity of each chemical is classified both quantitatively using
potency data and qualitatively using weight-of-evidence (WOE). Both kinds of data are
considered when deriving the score, and so both kinds of data must be generated for the
TRI chemicals.
The Section 313 criteria list several human toxicity parameters that EPA must
consider when evaluating a chemical for addition to TRI, including carcinogenicity, chronic
toxicity, acute toxicity, reproductive toxicity, heritable gene and chromosomal mutations,
developmental toxicity, and neurotoxicity. These parameters are defined in Table 8. A
release could be weighted based upon the number of these effects that it causes, the relative
severity of the effects it causes, and/or the potency of the ghemkn1, Some chemicals have
toxicity data for only one effect, while others will have evidence of effects for several ''of
these toxicity categories. The following describes how available data on the toxicity of TRI
chemicals will be used to weight releases.
Quantitative Data Used in Human ToxJc
Quantitative data on the relative potencies of chemicals are needed for .toxicity
weighting. For cancer risk assessment, EPA has developed standard methods for predicting
the incremental lifetime risk of cancer per dose of a chemical. EPA generally uses a
linearized multistage model of carcinogenesis to quantitatively model the dose»response
function of a carcinogen. The upper bound of the linear term of this model is called the
Oj". This slope factor is a measure of cancer potency1.
For noncarcinogenic risks, data on dose-response are more limited; generally, the
assessor evaluates dose compared to a Reference Dose. A Reference Dose is defined as
"an estimate (with uncertainty spanning perhaps an order of magnitude) of a dairy exposure
to the human population (including sensitive subgroups) that is likely to be without an
appreciable risk of deleterious effects during a lifetime" (EPA, 1987c). By definition,
exposures below this value are unlikely to produce an adverse effect; above this value, an
exposed individual may be at risk for the effect, but the probability of the effect is not
1 Although the level of conservatism inherent in these potency factors varies by chemical,
q,*s nonetheless are the best readily available values that allow us to compare the relative
potency of chemicals.
20
-------
TABLE 8. Toxlclty Endpoints
3
i
Endpolnt
Hnitahfa fhnelie and
Chromosomal Mutation
Developmental Toxicily
Reproductive Toxicily
Acute Toxicily
Chronic Toxicily
Neurotoxicity
Definition
I oia coxictijr concerns me aoiiiiy 01 a chemical 10 produce cancer in animals or
humans.
^fuMiticate Vuffitch afflinrft Ihia encliminl ran ciuim til Iftintl Ihnm Brnarata* iw^ta*it nf
failure to transmit genetic information: gain or loss of whole chromosomes
addition or deletion of a small number of base pairs (mutagenesis).
Any detrimental effect produced by exposures to developing organisms during
structural abnormalities, altered growth, and functional deficits (reduced
This eodpoint concerns the development of normal reproductive capacity.
Chemicals can effect gonadal function, the estrous cycle, mating behavior.
conception, parturition, lactation, and weaning.
Acute loxidty indicates the potential for a short-term exposure (typically noun
or days) by inhalation, oral, or dermal routes to cause acute health effect or
death.
Chronic toxicity indicate! the potential for any advene effects other than cancer
observed in humans or animals resulting from long-term exposure (typically
months or yean) to e chemical.
This eodpoint concerns the eentnl and/or peripheral nervous system. Changes
to the system may be morphological (biochemical changes in the system or
neurological diseases) or functional (behavioral, electrophysiological. or
neurochemical effects).
-------
known, nor is the severity of the effect known. For our purposes, we will assume that
noncancer risk varies as the ratio of the estimated dose to the Reference Dose.
We have compiled human health effects data on the TRI chemicals from a number
of EPA sources. Appendix C contains tables of available toxicity data for the TRI
chemicals. The primary source of these toxicity data was the Integrated Risk Information
System (IRIS). This computerized data source includes information on EPA evaluations of
chemical toxicity for both carcinogens and noncarcinogens. 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 cancer potency estimates (q,*) and Reference Doses (RfDs) for noncarcinogens and
some carcinogens. Data contained in IRIS have been peer-reviewed and represent accepted
Agency policy. It should be noted that this table contains health effects data for certain
classes of metal compounds. These data represent the health effects data for the parent
element, or, where appropriate, the health data of the most potent compound included in
that category. To be conservative, the releases associated with these compound categories
will be weighted based on these health effects data,
Another source of toxicity data is the Health Effects Assgssmgnt Summary Tables.
These tables are constructed for use in both the Supcrfund program and in the RCRA
Program but do not represent Agency consensus. However, these tables are publicly
available in the Superfund docket. They *!«** include potency estimates, WOE
categorizations for carcinogens, and Reference Doses for noncarcinogens.
The Superfund program also develops reportable quantities that quantify the relative
toxicity of chemicals. The Reportable Quantity (RQ) of a chemical is the minimum quantity
of a release that requires fr"*"*Hiat* notification to the National Response Center by any
person operating a vessel or facility. Separate reportable quantities are developed for
several toxicity measures, including carcinogenicity, acute human toxicity, chronic human
toxicity, and aquatic toxicity. RQs are also developed based on ignitabflity and reactivity.
The regulatory RQ is then the lowest among these RQ values. For cancer, the reportable
quantity value is assigned based on a combination of potency and WOE considerations. For
the purposes of deriving the RQ, potency is characterized by the chemical's ED,0 value for
cancer. The ED10 is the dose that increases a response 10 percent above the control group
(in this case, the response is cancer). Where specific potency data are not available, EPA
scientists use guidelines provided in the RQ development methodology as well as best
professional judgment to evaluate potential carcinogenic potency. For noncarcinogens, the
RQ is based on an evaluation of the human-equivalent minimum effective dose (MED) and
the severity of the effect produced at that dose. (The ™™™n« effective dose is derived
from the Lowest Observable Adverse Effect Level (LOAEL), with adjustments for
nonchronic exposure and interspeties extrapolation).
22
-------
RQs, RfDs, and q,'s are all toxicity measures that have been developed for EPA risk
evaluation purposes. One or more of these values are available for 247 TRI chemicals. The
70 TRI chemicals for which there are currently no formal EPA toxicity evaluations are
found in Table 4. For these chemicals, potential toxicity will be evaluated based on toxicity
data available from other data sources. These sources are listed in Appendix D. These
sources, as well as quantitative structure activity relationship methods, may have to be used
to supplement the formal available EPA toxicity values. For certain chemicals of
extraordinary interest, special resources may have to be made available to perform an
adequate toxicity evaluation. Chemicals for which little toxicity if any data exist, and thus
no evaluation of toxicity is possible, will be excluded from the indicator calculation.
Qualitative Data Used in Human Toxicity Weighting
For toxicity weighting, the proposed method uses both weight of evidence (WOE) and
potency to assign numerical toxicity weights. Typically, risk assessors use a variety of data
to evaluate the potential toxicity of a chemical to humans, including epidemiological data,
data from acute and chronic animal studies, and in vitro toxicity tests. Together, these data
form a body of evidence regarding the potential for toxic chemicals to cause a particular
health effect in humans.' The risk assessor can qualitatively judge the strengths of this body
of evidence when determining the probability of the occurrence of the effect in humans.
Based on this judgment, the chemical is assigned a WOE classification. Weight-of-evidence
schemes can be designed to demonstrate that a chemical causes a specific health effect in
general, or specifically in humans. The categories for carcinogens are related to likelihood
of effects in humans.
Standard weight-of-evidence classification schemes are widely used in cancer risk
assessment The carcinogenic WOE system presented in this methodology relies primarily
on categorical definitions from two weighting/scoring systems: the EPA Cancer Risk
Assessment Guidelines (EPA, 1986a, currently being revised) and the TSCA Chemical
Scoring System (ODryan and Ross, 1988). The EPA Cancer Risk Assessment Guidelines
define the following WOE categories:
Category Weight of evidence
Sufficient evidence from epidemiological studies to support a causal
relationship between exposure to the agent and cancer.
Bl T Jm*ifA evidence from epidemiological studies and sufficient animal data.
B2 Sufficient evidence from animal studies but inadequate or no evidence or no
data from epidemiological studies.
23
t>rr»ni>T. no WOT rvrv OP OTYVTT.
-------
C Limited evidence of cartinogenitity in animals and an absence of data or
evidence in humans.
D Inadequate human and animal evidence for cardnogenitity or no data.
E 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.
The TSCA Chemical Scoring System also provides specific criteria for placing
chemicals in toririty categories based on WOE. This system can serve as a supplement to
the EPA Risk Assessment Guidelines WOE categories. The TSCA Chemical Scoring
System oncogenicity weight-of-evidence scoring system is similar to the categories in the
EPA Cancer Risk Assessment Guidelines:
Category Weight of evidence
8-9 Evidence of oncogenicity from epidemiological studies and/or
positive results in two or more mammalian species.
6-7 Evidence of oncogenicity in either or both sexes of a single
mammalian species, with or without limited epidemiological
data.
4-5 Suggestive evidence of oncogenic potential from
epidemiological studies, mammalian bioassays, cell
transformation in vitro, or promoter/cocarcinogemc activity.
3 Evidence of genotoric potential.
1-2 'Limited evidence of lack of oncogenic potential.
0 No evidence of oncogenic potential from well-conducted, well-
designed mammalian studies 1U tWO OT mOie animal SpCClCS.
These categories could be further modified to include a category of no evidence, similar to
Group D, above. Many TCI chemicals have already been evaluated using one or both of
these two schemes. Furthermore, there is ongoing work between OPPT and the Oak Ridge
National Laboratory (ORNL) to score additional chemicals using this system. Therefore,
the WOE evaluations using these two schemes are currently available for many TRI
chemicals, and more will be available in the near future.
For noncarcinogens, WOE is routinely considered by EPA risk assessors when
developing Reference Doses (RfDs). Reference Doses are usually developed based on a
24
-------
No Observable Adverse Effect Level (NOAEL) or a Lowest Observable Adverse Effect
Level (LOAEL) combined with uncertainty factors and a modifying factor. The uncertainty
factors are applied to account for (a) extrapolating from animal species to human
populations, (b) extrapolating from less than chronic NOAELs to chronic NOAELs, and (c)
extrapolating from studies on particular human populations to the general human
population, including sensitive subpopulations. In addition, a modifying factor can be
applied to reflect EPA's best professional judgment on the quality of the entire toxitity
database for the chemical Since EPA Reference Doses already include both uncertainty
factors and a modifying factor based on deficiencies in the database, they require no further
consideration of uncertainty in this methodology.
Unfortunately, EPA Reference Doses have not been developed for all TRI chemicals.
When necessary, this methodology will require the evaluation of literature data to assign a
toxirity weight to a chemical In these circumstances, the following weight-of-evidence
guidelines from the TSCA Chemical Scoring System can be used for the qualitative
evaluation of the literature data:
For genotoxicitv:
Category
9
8
5-7
2-4
1
0
Weiht of evidence
Evidence «f mammalia
interaction
with mammalian germ cell DNA, or cpidcmiological data
suggesting genotoxicity in humans.
Evidence of genotoxicity in n«ntnamtnaiia« germ cell assays or
evidence of mammalian dominant lethality.
Evidence of genotoxicity in more than one test system, other
than above.
Limited evidence of genotoxicity, including mixed positive and
negative results.
Limited evidence of nongenotoxicity.
Negative test results indicating lack of known genotoxicity.
25
. nr» VOT rrnr OR OTTOTE
-------
For developmental effects:
Category Weight of evidence
8-9 Evidence of adverse developmental effects in humans or in at
least two other mammalian species
6-7 Evidence of adverse developmental effects in one mammalian
species.
5 Developmental effects at doses accompanied by maternal
toricity or otherwise equivocal test results.
4 Adverse developmental effects in nonmammalian species or in
vitro test systems.
3 Indirect evidence suggesting possible adverse developmental
effects.
2 Indirect evidence of lack of adverse developmental effects.
1 Limited evidence of lack of developmental effects.
•
0 No evidence of developmental toricity potential
Again, categories for no evidence could be added to these scoring systems. EPA has also
developed guidelines for certain noncartinogenic effects. These guidelines can be used in
conjunction with the TSCA Chemical Scoring System for qualitatively evaluating literature
toricity data on non-carcinogens. Guidelines for nmtagenitity risk assessment (EPA, 1986b)
describe general categories for WOE of nmtagenicity, to which we have assigned letters for
the purposes of this methodology:
For mutagenicity:
Category Weight of evidence
A Positive data derived from human germ cell mutagenicity
studies.
B Valid positive results from studies on heritable mutation events
jn mammalian perm ^fjly,
C Valid positive results from mammalian germ cell chromosome
aberration studies that do not include an intergeneration test.
26
rvt»
-------
D Sufficient evidence for a chemical's interaction with mammalian
germ cells, together with valid positive mutagenitity test results
from two assay systems.
E Suggestive evidence for a chemical's interaction with
mammalian germ cells, together with valid positive mutagenitity
evidence from two assay systems.
F Positive mutagenitity test results of less strength than defined
in D, combined with suggestive evidence for a chemical's
interaction with mammalian germ cell
G Valid negative test results for all endpoints of concern.
H Inadequate evidence bearing on either mutagenicity or chemical
interactions with mammalian germ cells.
The Agency's "Proposed Amendments to the Guidelines for the Health Assessment 'of
Suspect Developmental Toxicants" (EPA, 1989) delineates its categories with roughly the
same criteria as that of the TSCA Chemical Scoring System.
To implement this method, similar weighting systems will need to be developed for
neurotoxitity, reproductive toxitity and other chronic toxitity endpoints. The TSCA
Chemical Scoring System defines scoring criteria for subchronic/chronic toxitity as a
function of both a "severity score" and a "dose score." These criteria are:
Severity score Severity of effects
3 Life-threatening or severe effects
2 Moderately serious effects
1 Mild effects
0 No observed effects
Dose score Effects Guideline}}
3 At low doses 50 mg/kg-day
These criteria can be adopted and/or modified for use in this methodology.
27
OP OTTOTF
-------
General Format for Combining Weight of Evidence and Potency
This methodology uses matrices to evaluate a chemical based on WOE and potency
simultaneously. Rows and columns form matrix cells to which a toxicity weight is assigned.
The rows of the matrix are defined by the weight-of-evidence categorization, while the
columns of the matrix are defined by ranges of the measure of cancer potency (such as q*)
or the measure of noncarcinogenic threshold doses (such as NOAELs). More than one kind
of quantitative measure can be used to weight chemicals. By taking advantage of a range
of data sources available to weight chemicals, more TRI chemicals can be weighted and
included in the indicator. For example, for the cancer potency, one chemical may have a q/
value, while another chemical may have only an ED10 value. The toxicity weighting matrix
for cancer is constructed so that either measure can be used to assign a weight If both
measures are available, the value that results in a higher weight will be used. The heading
of a matrix column reflects the range of values that fall into that column. Ranges of
different toxicity measures that appear in the same column heading reflect a roughly
comparable magnitude of toxicity. Similarly, WOE descriptions appearing in the same row
label represent roughly the same category of WOE.
The exception to this weighting format is chemicals for which RfDs are available. As
discussed above, WOE is considered in the development of RfDs and is already
incorporated in the numerical value. Therefore, the matrix approach is unnecessary. For
these chemicals, weights will be applied based on ranges of RfD values alone.
Weights Applied tfl fte Categories
Either ordinal or proportional weights could be assigned to the matrix cells. Ordinal
weights delineate the relative toxicity rank among releases and are useful for setting
priorities. They do not, however, provide information on the tMgmtnHg of the toxicity of
chemicals relative to one another. For example, an ordinal rank of 3 for chemical A and
1 for chemical B does not *»*•« chemical A is three *""*» worse *IMMI chemical B. Since
ordinal weights do not reflect proportional differences in toxicity, the ability of the indicator
to reflect changes in health ?"d environmental impacts could be limited if ordinal weights
are used. In fact, if ordinal weights are used, it is possible that the indicator could decrease
over a period when actual risk increases. An example of this possibility is illustrated in
Figure 2.
Unlike ordinal systems, proportional scoring systems use numerical scores that reflect
the magnitude of difference between the impacts associated with chemical releases. Figure
3 shows how the indicator vahie developed in Figure 2 would change if proportional rather
than ordinal weights are assigned to the categories. Because of these considerations, the
matrices used hi this methodology assign proportional weights to matrix cells. Weights
increase by an order of ipag™*"*1* for each order of nMgnfrydy increase in toxicity and for
each increase in WOE category.
28
-------
Figure 2
Assume that the following scoring system is used to calculate the TRI indicator. The indicator
traces the releases of carcinogens to air:
qt* Value
SO or greater
5
-------
The corresponding scores would be:
Year
1
1
2
2
Plant
1
2
1
2
Chemical
vinyl
chloride
benzene
vinyl
chloride
benzene
Toxicity
Score
4
2
4
2
Population
Score
100,000
1,000,000
100,000
1,000,000
Exposure Score
2
2
3
1
Overall
Score
8
40
12
20
The overall TRI indicator for year one is 48, while for year two it is 32. Thus, from the
indicator, it would appear as if health risks have decreased. However, the acutal number of
expected cancer cases has increased dramatically, from roughly 1,600 to over 18,500.
30
-------
Figure 3
Assume that the following scoring system is used to calculate the TRI indicator. The indicator
traces the releases of carcinogens to air:
q,* Value
SO or greater
5
-------
The corresponding scores would be:
Year
1
1
2
2
Plant
1
2
1
2
Chemical
vinyl
chloride
hffcfiMift^
vinyl
chloride
benzene
Toxicity
Score
10,000
100
10,000
100
Population
Score
100,000
1,000,000
100,000
1,000,000
Exposure Score
100
100
1,000
10
Overall
Score
100
10
1,000
1
Using proportional weighting, the overall TRI indicator for year one is 110, while for year two
it is 1,001. Thus the increase in risk protrayed by the indicator successfully reflects the increase
in the estimated number of cancer cases.
32
no
-------
The Human Health Toxicity Matrices
The preceding discussion presented the general framework for weighting the toxicity
of TRI releases. This section describes and explains the matrices developed from this
framework. Two separate toxicity weighting matrices for carcinogens and noncartinogens
are discussed.
Carcinogens. The toxicity weighting matrix for carcinogens is found in Figure 4. The
rows of the matrix qualitatively classify carcinogens into two general categories: known/
probable and possible. Weight-of-evidence categories A, 61 and 62 of the EPA Cancer
Risk Assessment Guidelines and categories 6-9 of the TSCA Chemical Scoring System
cancer WOE scheme are placed in the "known/probable" category. Class C carcinogens and
TSCA Chemical Scoring categories 3-5 are placed in the "possible" category.
The columns of the matrix describe quantitative measures of toxicity. Two alternative
measures are presented: the EPA q,* and the ED10. Furthermore, the development
methodology for cancer RQs (EPA, 1988a) provides criteria for assigning a chemical to a
potency group in the absence of an EDIO; these criteria can be used for approximating the
range of the ED,0 for purposes of chemical weighting in the current methodology. The
particular ranges of q/ values selected to represent each category of cancer potency were
chosen to correspond to the ranges presented in EPA's Hazard Ranking System (55 Federal
Register 51532). (Corresponding ranges for comparable ED,0 values will need to be
developed.) The Hazard Ranking System (HRS) is a multipathway scoring system "used to
assess the threat associated with actual or potential releases of hazardous substances at sites"
(Federal Register, op at). The HRS score determines whether a site wfll be included on
the National Priorities List (NFL). Part of the HRS scoring system rates the inherent
toxicity of chemicals based on measures of cancer potency, RfDs, and/or acute toxicity.
Ranges of toxicity values that differ by an order of magnitude are assigned scores that differ
by an order of magnitude. The actual numerical weights assigned to the matrix cells in
Figure 4 correspond to the scores assigned in the HRS to these ranges. In certain cases,
ranges presented in our matrix extend beyond those presented in the HRS. However, the
basic logic of «««g™'"E toe weights to these ranges remains the same: ranges that differ by
an order of magnitude are assigned weights that differ by an order of magnitude. Figure
5 shows the distribution of q/ values in these ranges for the TRI chemicals that have these
values.
The cells in the first row of the matrix (that is, the row that corresponds to the
"known/probable" WOE category) were assigned the weights based on the HRS values.
Weights in the lower row (Ltn the lower WOE category) were assigned by dividing the
weights in the row above by a factor of 10.
33
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
S3
2
HI
n
e=a
H
PI
-------
Figure 5
Distribution of Toxicity Values
8
2:
O
H
s
O
O
3
Value
RfDs
Vilue
M
Nunber
Number
Value
Value
Number
It M
Number
5«
35
-------
Noncarcinogens. For noncarcinogens that have established RfDs, the following
toxicity weights will be used:
RfD Range (mg/ke-davl
RfD < 0.00005 100,000
0.00005 < RfD< 0.0005 10,000
0.0005 < RfD < 0.005 1,000
0.005 < RfD < 0.05 100
0.05 < RfD < 0.5 10
0.5 < RfD 1
This weighting system is taken directly from the HRS, with the exception of the
highest weighting category. The weight assigned to this category is logically consistent with
the HRS scoring system: RfDs less than 0.00005 are assigned a weight that is an order of
magnitude greater than RfDs between 0.00005 and 0.0005. Figure 5 shows the distribution
of RfD values in all of these ranges for those TRI chemicals that have RfDs.
For noncarcinogens without RfDs, a toxicity weighting matrix is presented, as shown
in Figure 6. The rows of this matrix also qualitatively classify WOE into two general
categories: known/probable and possible. Categories 5-9 of the TSCA Chemical Scoring
System WOE schemes for noncarcinogens are placed in the "known/probable" category, and
categories 2-4 are placed in the "possible" category. For mutagens, EPA mutagen risk
assessment guideline categories A-E are placed in the "known/probable" category and F in
the "possible" category.
The gfthinmc of this matrix are labeled with the numerical ranges of values that
define each category of noncancer toxicity. Measures that can be used to quantitatively
weight toxicity include the human-adjusted NOAEL (the No Observable Adverse Effect
Level), human-adjusted Lowest Observable Adverse Effect Level (LOAEL), and the human-
adjusted mjnitmmi effective dose (MED). The ranges for the NOAEL values are defined
to roughly correspond to the ranges of RfDs presented above. Corresponding ranges for
LOAEL and MED values still must be developed.
The cells in the first row of the matrix (that is, the row that corresponds to the
"known/probable" WOE category) were assigned the same weights as those assigned to the
RfD ranges. Weights in the lower row (Le., the lower WOE category) were assigned by
dividing the weights in the row above by a factor of 10.
Selecting tfrf Fjjn?j Human Health Toxicity Weight for a fhemigal
Chemicals can cause several types of toxic effects. Tne TRI Human Health Chronic
Indicator method will weight a chemical based on the endpoint associated with the lowest
dose at which an effect occurs. Under this regime, the same weights are assigned to similar
36
-------
o
-------
cells in each category of toxicity so that chemical toxicity scores are comparable across
different types of toxic effects. For example, a highly potent genotoxin with strong WOE
would get the same score as a highly potent carcinogen with strong WOE.2
The approach of weighting based on the endpoint with the lowest dose does not
consider differences in the severity of the effects posed by the chemicals. For example,
cancer is weighted in the same manner that genotoxicity is weighted. Also, chemicals with
a broad range of effects are weighted the same as a chemical with only one effect The TRI
Indicator Work Group considered the option of applying an additional factor to the toxicity
weight based on the severity of the effect Applying such weights across categories of toxic
endpoints would require a subjective evaluation of the relative severity of the health effects.
The Work Group also considered the option of applying an additional weight based on the
number of endpoints for which the chemical demonstrates effects. However, a chemical may
appear to demonstrate just one effect only because there are no data on other effects; thus,
applying a weight based on the number of endpoints may undervalue some poorly studied
but still risky chemicals. For these reasons, the Work Group rejected these options; the
methodology will weight chemicals based on the most severe endpoint without applying
additional weights for severity or number of effects caused.
Toxicirv Weights — Ecological
For ecological effects, the indicator will focus on aquatic life impacts. Very little data
are available for most chemicals on effects to terrestrial or avian species, and the human
health indicator will provide some predictor of these.
Aquatic toxicity weighting differs from human toxicity weighting in two important
respects. First, WOE is not considered a factor in the weighting scheme, since direct
evidence, of chemical toxicity is available from tests on aquatic species. Second, the aquatic
toxicity weighting scheme simultaneously considers toxicity and bioaccumulation potential.
Both of these measures are considered important when evaluating impacts on aquatic
ecosystems.
Data Used in Aquatic Toxicitv Weighting
Common numerical aquatic toxicity data include the Acute or Chronic Ambient
Water Quality Criteria (AWQC), developed by the Office of Water, which may serve as the
basis for water quality standards, as well as values from the literature, such as the lethal
concentration, 50 percent (LCjo) - the chemical concentration in water at which 50 percent
'The toxicity weighting may be pathway-specific for chemicals that demonstrate different
effects by different pathways. For example, some chemicals are carcinogens by the
inhalation pathway, but not by ingestion. Thus, the toxicity weighting may be different for
different release pathways.
38
TTO AFT PFPOPT . DO NOT CITE OR OUOTE
-------
of test organisms die • or life-cycle or chronic No Observable Adverse Effect Levels
(NOAELs). Aquatic RQ values are also available -for some TRI chemicals; aquatic RQs
are determined based on the aquatic LC^ of the chemical. Appendix C lists the AWQC
and aquatic RQs that have been developed for TRI chemicals. Additional aquatic toxicity
data will be obtained from the Aquatic Toxicity Retrieval System (AQUIRE), an aquatic
toxicity database maintained by EPA. These data will be reviewed by OPFT before use in
the indicator.
The preferred measures of bioaccumulation potential are the bioconcentration factor
(BCF) or bioaccumulation factor (BAF). These values are derived from laboratory tests
that compare the contaminant concentration in the environmental medium to the
concentration in the tissues of a test organism (usually fish). Several researchers have found
that for organic contaminants, the BCF or BAF can be approximated as a function of the
log of the octanol water partition coefficient (log K^). The K*, is a physicochemical
property that describes the partitioning of organic chemicals between an organic solvent
(octanol) and water. Finally, when neither of these measures are available, the
bioaccumulation potential can also be approximated by the water solubility of the chemical.
Generally, the less soluble a chemical, the greater its potential for bioaccumulation. Values
for all of these measures of bioaccumulation potential are available from a variety of
sources, including the AQUIRE database, as well as a number of EPA Office of Water
references, the FFn chemical properties data base and standard chemical reference books.
The Aquatic Toxicitv Weighting Matrix
The aquatic toxicity weight assigned to a chemical is a function of both its aquatic
toxicity values and bioaccumulation potential values. Separate weights are assigned based
on each of these measures; the chemical's final toxicity weight is the product of these
individual weights. The individual weights assigned based on the measures of
bioaccumulation potential are the following:
Water Solubility
(rag/1)
> 1,500
500-1500
25-500
. .
.
<25
LogKow
<0.8
0.8-2
2-32
32-45
4.5-5.5
5.5-6.0
BCFQ/kg)
<1
1-10
10-100
100-1,000
1,000-10,000
> 10,000
Weight
0.5
5
50
500
5,000
50,000
39
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
Individual weights based on aquatic toxicity measures are the following:
Life Cycle or
Chronic
NOAEL
(Mg/1)
MOO
10-100
MO
0.1-1
<0.1
LC» G,g/l)
> 1,000
100-1,000
10-100
MO
<1
Acute AWQC or
AALAC Oig/1)
> 100,000
10,000-100,000
1000-10,000
100-1,000
<100
Chronic
AWQC or
AALAC
(Mg/D
> 1,000
100-1,000
10-100
MO
<1
Weight
1
10
100
1,000
10,000
As with the human health toxicity weighting, the quantitative measures used to represent
aquatic toxicity, the value ranges used to define the categories of toxicity, and the numerical
weights assigned to each category were for the most part taken from the Hazard Ranking
System. Hie only exception is the ranges for life cycle or chronic NOAELs, which were
taken from the TSCA Chemical Scoring System.
Figure 7 presents the combined toxicity weighting system for aquatic toxicity. The
rows of the matrix are defined by the bioaccumulation potential categories anH the columns
of the matrix are defined by the aquatic toxicity categories. The cells of the matrix are the
product of the chemical's bioaccumulation potential and aquatic toxicity weights.
Exposure Weights - Human
As with toxicity weighting, both qualitative and quantitative elements are considered
when weighting exposure potential Three steps are involved in deriving the human
exposure weight The first step is to categorize some measure of exposure potential into
numerical ranges. The second step is to categorize the levels of uncertainty associated with
the measures of exposure potential. The final step uses a matrix, with numerical exposure
potential ranges arrayed along one axis and uncertainty categories arrayed along the other,
to assign increasing weights to higher-certainty/higher exposure potential releases.
Quantitative Data Used in Exposure Potential Weighting
For the first step of this process, quantitative measures of exposure potential must
be estimated. To do so, several existing scoring systems take the approach of ordinally
ranking the volume of each release by the some physical measure of the chemical's ability
to move through the environmental medium into which it is released. However, because
the exposure potential rankings would have different physical "^Epfrg* for different
40
ATT T>T>nru>T . r»O WVT PITT Ol? OT'OTE
-------
Ufe Ode or Chi oak NOAEL
Acute AWQC or AALAC
I
HI
n
t»«s
H
O
-------
pathways, comparisons among releases to different media would be difficult, and weighted
releases from different pathways could not be added to obtain a single indicator value.
In this methodology, comparisons across media can be made because a common
quantitative exposure measure for each medium is derived: an estimate of "surrogate dose" -
- a measure related to the amount of chemical contacted by an individual per kg body
weight per day. Although limited facility-specific data and the use of generic models
(described below) prevent the calculation of an actual dose, the surrogate dose measure
developed from the available data and from generic modeling is designed to be, in concept,
comparable for all media.
To estimate the magnitude of exposure potential from TRI releases, a separate
exposure evaluation is conducted for each release pathway. The ideal derivation of an
exposure weight would involve a site-specific exposure assessment for each release medium
and for each exposure pathway. However, such an effort is well beyond the scope of this
project and well beyond the intended use of the TRI data. These data are frequently
estimates of emissions, not precise measured values. Notably, they are not estimates of
environmental concentrations that result from the emissions from the plant Furthermore,
extensive site-specific information relevant to exposure modeling is not pan of a TRI data
submission. For example, the TRI forms do not require submission of data on groundwater
flow, soil conditions, and other factors that affect groundwater contamination from land
releases. It is not the intent of this project to gather additional data or measurements that
would be needed to perform these calculations. The need to accurately reflect exposure
characteristics in the indicator must be balanced by the need for a simple and
understandable indicator that is easily communicated to the public and that is based on
currently available data. Therefore, for this methodology, the exposure evaluations combine
data on media-specific emission volumes, phyacochemical properties, and where available,
site characteristics with site-specific or generic exposure models to determine an estimate
of the ambient concentration of contaminant in the medium into which the chemical is
released. (Again, the use of submitter-estimated TRI emission data and generic models with
many default assumptions make this only a surrogate related to actual concentration). For
the human health chronic indicator, the ambient media concentrations are then combined
with standard human exposure assumptions to estimate the magnitude of the surrogate dose.
It must be emphasized that while this methodology uses the EPA exposure
assessment paradigm to evaluate exposure potential, the results should not be construed as
an actual absolute numerical estimate of dose resulting from TRI releases. Instead, the
purpose is to obtain an order of magnitude estimate of surrogate dose resulting from release
of TRI chemicals relative to the surrogate dose resulting from other releases included in ihe
indicator, so that these releases can be weighted appropriately in the indicator.
The exposure evaluation methods used for each type of release are specific to that
type of release and depend on the models and data available to evaluate that pathway. In
some cases, models will be combined with some site-specific data to estimate exposure; m
42
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
other cases, generic worst-case models may be used in the absence of any site-specific data.
Section V of this report presents an extensive discussion of the specific mechanisms for
evaluating exposure for each type of release or transfer.
Qualitative Data Used in Exposure Potential Weighting
Consideration of uncertainty in the exposure evaluation is necessary for making
comparisons across pathways, since the exposure evaluation methods for various pathways
differ significantly in their possible level of refinement. For the purposes of exposure
weighting, we have defined the following uncertainty categories for use in this methodology:
Category Explanation
A Combines modeling with some generic and some dependable
site-specific data to generate exposure estimates.
B Combines modeling with some generic and some site-specific
data, but identification of appropriate site-specific data subject
to error and will often be filled in with generic assumptions.
C Extrapolates generic exposure estimates from actual data from
other sites to exposure at particular sites (e.g^ groundwater
modeling).
In general, more generic methods for exposure modeling require more assumptions and
extrapolations. These assumptions and extrapolations tend to be conservative, so that more
generic modeling tends to yield overestimates of exposure. The categories defined above
are designed so that a lower category indicates more generic modeling that is likely to
overestimate exposure when compared to the next highest category.
The Human Exposure Potential Weighting Matrix
As with toxicity weighting, a matrix has been developed to weight exposure potential
that allows simultaneous consideration of the estimated level of exposure and the evaluation
of the relative level of uncertainty associated with each estimate. The matrix for exposure
weighting is presented in Figure 8. The exposure value is expressed in mg/kg-day and is
calculated on a pathway-specific basis, as described in Section V. The uncertainty categories
are as defined above. As with toxicity weighting, proportional weights are assigned so that
exposures that differ by an order of magnitude are assigned weights that differ by an order
of magnitude.
As discussed above, the exposure uncertainty categories are designed so that a lower
category indicates that the exposure estimate is probably an overestimate compared to the
next highest category, therefore, for this methodology, the exposure potential score will be
43
TWA FT PFTHTOT - DO NOT CITE OR QUOTE
-------
58
Q
H
n
l=-3
H
PI
O
a
-------
reduced by a factor of 5 if assigned to category B, and by an order of magnitude if assigned
to category C.
Exposure Weights ~ Ecological
The estimated ambient water concentration value is used directly to evaluate
potential exposures to aquatic life. The method for evaluating ambient surface water
concentrations resulting from TRI releases is discussed in Section V. Once ambient
concentrations have been determined, exposure weights are assigned based on the following
ranges:
Estimated water concentration fug/1) Weight
MO 10
10-100 100
100-1,000 1,000
> 1000 10,000
Aquatic exposure weights for categories of increasing estimated water concentration
do not need to reflect exposure relative uncertainty, since all estimates are derived using the
same methodology. The weights simply increase by an order of magnitude for each
category.
Population Size Weights
The Work Group considered several options for including population in the indicator.
One suggestion was to represent populations as ranges rather than using absolute numbers,
especially for small populations. Use of a range may be more appropriate given the
uncertainty about the size of the exposed population for most exposure pathways. Another
reason for this approach is to avoid undervaluing potentially high individual impacts on
small rural populations. Using a range approach, yt««ii rural populations would be assured
of a minimum weighting. On the other hand, if reliable population data are availably an
argument can be may for avoiding the introduction of additional uncertainty in the indicator.
by using ranges for population values.
In light of these considerations, this methodology proposes several rules for the use
of population data in the exposure evaluation. First, a minimum population value is
assigned to all releases to avoid undervaluing rural populations. Second, both quantitative
data and qualitative data (on the uncertainty of the population size estimate) are used when
evaluating the size of the population potentially exposed to TRI releases.
45
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
Quantitative Data Used in Population Weiehtine
The determination of the size of the population exposed to TRI releases will vary
substantially depending on the medium to which the chemical is released. The methods for
estimating the size of the exposed population are discussed for each pathway in Section V.
Methods for estimating exposed population sizes will vary in their accuracy. We may
have good data on the number of persons surrounding a particular release site but know
little about how many of these individuals actually come into contact with contaminated
media. The uncertainty associated with the population estimate will differ by pathway. For
example, the size of the population potentially exposed to contaminated groundwater will
be known with significantly less certainty than the size of the population exposed to air
emissions, because of the greater certainty associated with air modeling (we know who lives
around the sites, and everyone breathes air) than with modeling of groundwater (location
and movement not known, number of persons drinking groundwater not known).
To address this concern, this methodology will consider uncertainty in the evaluation
of the size of exposed populations. Uncertainty categories are denned as follows:
A Population siyf is well*characterized, MM! contact with
contaminated media is believed to affect most or all members
of the exposed population.
B Population is characterized v"n£ crude data that is likely to
overestimate size of exposed population, but contact with
contaminated media is believed to affect most or all members
of the exposed population.
C -Population is characterized using crude data that is likely to
over estimate size of exposed population, and contact with
contaminated media occurs with an unknown fraction of those
potentially exposed.
A lower category indicates that the population estimate is probably an overestimate
compared to the next highest category; therefore, for this methodology, the estimated size
of the population will be reduced by a factor of 10 if assigned to category B, and by two
orders of magnitude if the population estimate is assigned to category C.
All population sizes will also be rounded up to the nearest 1,000 persons. By
rounding, we acknowledge that the methods used to estimate the population sizes are
imperfect and do not justify the use of precise values. Also, this rounding creates a
minimum exposed population size, since populations of 0*999 persons will be counted as
46
r»r» vr»T rvnr OP OT'OTF.
-------
1000. A minimum population size ensures the inclusion of rural populations in the
indicator.
Because of the major difficulties in estimating sizes of the populations of ecological
receptors, the TRI Ecological Indicator does not include a population weight. In effect, this
approach assumes that all aquatic emissions occur to equally vulnerable locations.
Integrating Toxicity Data., Exposure Data, and Population Data to Obtain Facility*
Chemical-Medium Indicator Elements
The previous sections described how each element of the human health chronic
indicator is determined. The following equation combines these components for each
facility, each chemical, and each medium:
Element^ = ToxieityWeight^ • ExposureWeight^ • AdjustcdPopulation^
where
i = subscript for chemical i,
; s subscript for facility j, and
k = subscript for medium k.
The weights am* the population size are multiplied in this setting because each component
(toxidry, exposure, and population) contributes in a multiplicative way to the overall
magnitude of the impact The result of the multiplication of the weights is a facility-
chemical-medium-specific "indicator element" It must be reiterated that this is not a
physically meaningful measure of risk associated with the facility, but is an approximate
measure that is comparable to approximate measures for other facilities calculated using the
same methods. In Section VI, these elements will be combined into a single TRI chronic
human health indicator. These elements can be used for diagnostic purposes when the
indicator is calculated.
Ecolorical Chronic Indicator
The methods for determining aquatic tcoticity and exposure weights are described
above. The following general equation combines these components for each facility, each
chemical, and each medium:
47
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
= ToxicityWeigkty • ExposureWeight^,
where
/ = subscript for chemical i,
y = subscript for facility j, and
k = subscript for medium k.
As with the human health indicator, the weights are multiplied in this setting because each
component (toxicity and exposure) contributes multiplicatively to the overall magnitude of
the impact The result of the multiplication of the weights is a facility-chemical-medium-
specific "indicator element." Again, the elements should not be interpreted as actual
measures of risk. In Section VI, these elements will be combined into a single TRI chronic
ecological indicator.
V. MECHANICS OF ESTIMATING EXPOSURE
The preceding section addressed how exposure and toxicity will be assigned weights
based on the relative magnitude of each and the underlying uncertainty. The purpose of-this
section is to describe how the magnitude of exposure is calculated for each pathway.
Ideally, we would model each TRI release with the rigor of a full risk assessment, relying
on site-specific data. This is impossible, however, because resources and information are
limited. For a set of indicators that are meant to generally track relative year to year
changes in human am* environmental health impacts, it is I»ISQ unnecessary. For our
purposes, it is adequate to approximate the results of such a modeling exercise.
TRI chemical releases are reported for seven different release and transfer pathways.
On-site pathways include direct air (stack and fugitive), UmdfiHing, underground injection,
and surface water releases. Chemicals might also be transferred to Publidy Owned
Treatment Works (POTWs) or Treatment, Storage, and Disposal Faculties (TSDFs). The
release and transfer pathways for which exposure potential must be evaluated are presented
in Figure 9.
To incorporate exposure in the TRI indicators, the TUX Indicator Work Group
considered two fi»"fo»*w*ai approaches. First, releases could be weighted vsing proxy
measures of exposure, such as population density surrounding the facility, and/or
environmental mobility characteristics. Alternatively, releases could be modeled, where
possible, to derive ambient environmental concentrations of contaminants at varying
distances from the release site, and then combined with the corresponding population data.
When considering modeling versus a more simplified approach, the Work Group
noted the fact that some releases (such as air releases) may be more amenable to a
modeling approach than other types of releases (such as ofEate transfers); thus, releases
48
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
FIGURE 9
Overall Approach for Modeling TRI Discharges
8
I
8
O
S
DIRECT WATER
GEMS system
Calculate concentrations at drinking
intakes. Use population served as
exposed population; use a fraction of
this population as fish ingestion
population.
I
Effluent
POTW
Transfers
DIRECT AIR
Suck. Fugitive
Volitilitalion
ISCLT
(Indust. Source Comp. Long Term)
Calculate concentrations at various
ring distances described by TRIAIR.
Use BO/ED for population exposed.
I
| Sludge Disposal
Incincf&t ion
i
Volatilization
DIRECT LAND
(Lindfilling)
Lantiniling
Estimate leachate concentration.
Apply generic DAP.
Use population within 4 sq. km
of source.
Offsile
Incineration
Offsite
Tuisfers
or
lie
Sill
nn
ing
49
-------
from certain media may be evaluated in a more refined way than releases to other media.
Furthermore, TRI release data are generally estimates rather than precise quantifications
of releases, and, since we lack site-specific data, other modeling inputs would by necessity
be generic. It may not be useful to perform modeling if release data are estimated and if
inputs to the model are necessarily generic. All of these points argue for using a more
simple approach to evaluating exposure.
However, even generic modeling has significant advantages. First, while modeling
may not allow precise estimates of exposure due to generic and estimated inputs, it does
allow us to estimate the magnitude of exposure and to consider local conditions where these
data are available. In short, uncertainty should not be introduced where alternatives exist.
Furthermore, to compare and sum exposure from different media, each media contribution
must be expressed in an equivalent form, i.e., estimated "surrogate dose." In each pathway
discussed below, surrogate dose can be estimated via generic modeling of each facility
release and then can be coupled with toxiciry to constitute a facility-chemical-media-specific
increment to the entire indicator. No other method allows for this uniform element
calculation. Without this common denominator, we would be forced to evaluate how a
proxy measure of groundwater exposure (based on solubility) can be compared to a proxy
measure of air exposure (based on photolysis rate). The relative ease of modeling releases
to different media remains a problem; however, it may be appropriate to have better
estimates of exposure for certain pathways if these pathways cause significant impacts and
would contribute disproportionately to the indicators. For example, from a population
perspective, it may be important to better understand air releases than ofiEdte transfers since
air releases will probably contribute far more to population impacts than transfers to
hazardous waste facilities. In general, the Work Group preferred modeling where possible
since it is the common Agency mechanism for incorporating emission data, local conditions,
and population density into an exposure evaluation.
The exposure evaluation scheme presented below does not lend itself to quick
lculation by hand, though hand calculation is not unimaginable. However, its structure
offers unlimited combinations and views of the indicators' components. On one level,
overall computation of the indicators could not be more replicable or verifiable, since it will
be performed completely within one computer program. However, each facility-chemical-
media indicator element could be disseminated in computer format for those wishing
different views of its makeup. Regions, states, or individuals could use these individual
constituents to create their own subindicators.
The following discussions of exposure frequently mention concentration and surrogate
dose. We do not mean to imply that we somehow supplanted the risk assessment process
and can accurately calculate cases. We speak of those terms only in the abstract The
method is a simple way to gauge relative risks from releases to different media in a
congruent, defensible way. In some cases, the modeling will be purposely simple, given our
lack of site-specific data. The differences in the level of refinement of exposure modeling
are addressed by \i&n£ the uncertainty weighting scheme dkfMwd in Section IV.
SO
TM» A*-T VVDOVT . no VOT mr OP OT'OTE
-------
Treatment of Emissions Range Reporting
Emissions data are a fundamental element of the exposure evaluation process. As
discussed above, TRI facilities report emissions to seven pathways. Emissions data can be
presented as numerical point estimates, or, if releases are below 1,000 pounds, as an
estimated range of emissions. Where emissions are reported as a range, we could either use
the midpoint of the range or the upper bound of the range to represent the emissions of
that chemical. To produce a conservative estimate of exposure potential, we will use the
upper bound of the range as our estimate of emissions, since this value is the maximum that
the facility could be emitting.
Stack and Fugitive Air Releases
Ideally, reported stack and fugitive air releases from the TRI database would be
modeled using site-specific data (such as source area or stack height). Since TRI does not
contain such facility-specific information, we must use default values to model TRI facilities
with established EPA air dispersion models.
For this methodology, we will use the Industrial Source Complex Long Term (ISCLT)
model developed by the Office of Air Quality Planning and Standards (OAQPS). ISCLT
is a steady-state Gaussian plume model used to estimate long-term pollutant concentrations
downwind of a source. The concentration is a function of site-specific parameters (stack
height, stack velocity) and chemical specific decay rates.1 The general form of the
concentration equation from a point source at a distance r greater than 1 meter away is4:
.QfSVD
where
C* = concentration at distance r (pg/m1),
Q «= pollutant emission rate (g/s),
f = frequency of occurrence of wind speed and direction,
'Importantly, chemicals with extremely short half-lives in air will not be modelled using
these procedures. Such chemicals will be assumed to degrade before significant exposure
occurs. Products of decay could also be modeled where data permit.
4 This equation is from EPA (1992). The equation is for a specific wind speed,
direction, and category ($fc). Each facility has several combinations of these that must be
added to arrive at a total concentration at that point The equation for area sources is
similar.
51
DRAFT REPORT - DO NOT CITE OR QUOTE
-------
6 = sector width (radians),
S = smoothing function used to smooth discontinuities at sector boundaries,
u = mean wind speed (m/sec),
|