United States Science Advisory EPA-SAB-EC-ADV-99-003 L^'
Environmental Board (1400) December 1998
Protection Agency Washington, DC www.epa.gov/s8/)
f/EFA AN SAB ADVISORY ON THE
TRIM.FaTE MODULE OF THE
TOTAL RISK INTEGRATED
METHODOLOGY (TRIM)
CONDUCTED BY THE ENVIRONMENTAL
MODELS SUBCOMMITTEE OF THE
SCIENCE ADVISORY BOARD
U.S. Environmental Protection
Region 5, Library (PL-12J)
77 West Jackson Boulevard,
Chicago, IL 60604-3590
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
December 31, 1998
EPA-SAB-EC-ADV-99-003
OFFICE OF THE ADMINISTRATOR
Honorable Carol M. Browner SC.ENCE ADVISORY BOARD
Administrator
U.S. Environmental Protection Agency
401 M. Street, SW
Washington, DC 20460
Subject: Advisory on the Total Risk Integrated Methodology (TRIM)
Dear Ms. Browner:
The Environmental Models Subcommittee (EMS) of the Executive Committee
(EC) of the Science Advisory Board (SAB) met in Washington DC May 5 and 6, 1998 to
review the Total Risk Integrated Methodology (TRIM), and in particular to review the
TRIM.FaTE module. TRIM is being developed by the Office of Air Quality Planning and
Standards (OAQPS) of the Office of Air and Radiation (OAR). It is an overall modeling
framework intended to provide a flexible method for integrating the release(s) of
pollutants from single or multiple sources to their multimedia, multipathway movement
in order to predict exposure to pollutants and to estimate human health and ecological
risks.
The Subcommittee appreciates the opportunity to provide an early review and
early advice to the Agency on the TRIM methodology and on the TRIM.FaTE module in
particular. Overall, we found the development of TRIM and the TRIM.FaTE module to
be conceptually sound and scientifically based. EPA has gone beyond the current
available modeling to create an integrated framework for dealing with multimedia
transfers in a more self-consistent, and, hopefully, more useful fashion. It is a very
complex model in terms of interconnections, so care needs to be taken to insure that it
is applied appropriately and produces realistic results.
There were six charge questions before the EMS. As a result of the public
review meeting to evaluate the TRIM.FaTE module in light of the charge questions, and
subsequent writing activities by the Subcommittee, this letter and the accompanying
report have been prepared. The letter summarizes EMS' key findings and
recommendations. The attached report provides a more complete description of the
Subcommittee's advice. In response to requests from OAQPS, the Subcommittee
expects to conduct additional advisory reviews of the other TRIM modules (i.e.,
exposure, uptake, biokinetics, dose/response, and risk characterization) over the next
few years.
(}\j Recycled/Recyclable
O) Printed with Soy/Canda Ink on paper that
!<^_7 contains at least 50% recycled fiber
-------
1. Is the overall conceptual TRIM approach appropriate?
The conceptual approach for TRIM appears to be technically defensible. The
Subcommittee notes that the most important challenge facing the TRIM modeling
enterprise is the lack of available data for fate, transport, exposure and risk processes.
The Subcommittee, therefore, recommends that the Agency identify and acquire
significant additional field data to estimate modeling parameters and to "validate" the
model components and other aspects of the modeling system.
2. Is the spatial compartmental mass-balance approach commensurate with
quantifying uncertainty and variability in a scientifically defensible
manner?
The Subcommittee accepts that at present the TRIM.FaTE development team
has yet to enter into substantial work on the analysis of sensitivity and uncertainty, as it
appears that thus far only a local sensitivity analysis has been implemented under
prototypical, test-case conditions. The Subcommittee recommends that the literature
on sensitivity and uncertainty analysis be reviewed thoroughly prior to making firm
choices on the specific forms for incorporating sensitivity and uncertainty analysis into
TRIM and that the TRIM developers take the necessary time to reflect on what value
the analysis of uncertainty and sensitivity will add to their work and the resulting
framework for modeling. The Subcommittee endorses the notion of incorporating
checks and balances into the model framework and software in order to guard against
future users making inconsistent or illogical choices for numerical parameterization of
the model and its uncertainty analysis.
3. Is the overall conceptual approach represented in the TRIM.FaTE Module
appropriate?
There are several strengths to the TRIM.FaTE methodology. The Subcommittee
finds that the TRIM.FaTE module is conceptually sound and aims at an appropriate
level of complexity. The challenge ahead will be in developing and implementing the
details beyond the linear algorithms incorporated in the prototype version. OAQPS1
emphasis on the steady state distribution of contaminants may prove to be an important
limitation. The Subcommittee recommends that the Agency develop a users guide that
includes a description of the strengths and the limitations of the TRIM.FaTE module as
it is currently constructed or envisioned.
-------
4. The TRIM approach is designed to be flexible and to allow for a tiered
approach, to function as a hierarchy of models.
4a. As implemented at this time, is the TRIM.FaTE Module appropriate from a
scientific perspective?
TRIM.FaTE is an ambitious attempt to model fate and transport of air pollutants
through all aspects of an ecosystem. By design it is very broad and encompasses the
different media, pollutant transformations and exchanges. Although the model predicts
the accumulation of released air pollutants in target organisms throughout the food
chain, there are few data sets that exist to compare with the overall model predictions.
TRIM.FaTE consists of a series of connected hypotheses to simulate the complexities
inherent in a multi-media environment. It is these hypotheses that can be scientifically
tested to elucidate the efficacy and the limitations of the overall model.
The Subcommittee finds that TRIM.FaTE as currently presented and configured
has not been checked against a detailed set of observed, spatially varying real world
environmental concentration data. The Subcommittee recommends that the module be
constructed in a fashion that will permit the component results to be disaggregated and
studied to build confidence in the overall prediction of the model.
4b. Is the TRIM.FaTE Module, as designed, an appropriate tool, when run either
at a screening level or for a more refined analysis, for use in providing
information for regulatory decision-making?
The Subcommittee, absent additional testing and evaluation, cannot provide an
assessment or recommendation regarding the appropriateness of the module as a
decision-making tool. At this point in its development, the module seems to hold
promise as a screening tool.
5. Does the TRIM.FaTE Module, as it has been conceptualized, address some of
the limitations associated with other models?
The Subcommittee finds that the TRIM.FaTE module has the excellent feature of
conserving mass for chemicals undergoing first-order linear mass transfer and
transformation processes. It is, however, uncertain at this time as to how the proposed
methods can be adapted for chemicals subject to non-linear higher-order processes.
The Subcommittee recommends that the Agency consider providing additional methods
and guidance to assist users in selecting the appropriate level of spatial and temporal
resolution necessary to obtain adequate precision and accuracy in the results.
-------
6. Does the TRIM.FaTE Module, as it has been conceptualized and demonstrated
to date, facilitate future integration with appropriate data sources and
applications?
The Subcommittee finds that the TRIM.FaTE module could conveniently and
effectively be integrated with data sources such as GIS, provided the spatial scales in
the GIS are congruent with those of TRIM.FaTE. However, coupling TRIM.FaTE with
other more complex models which generate continuous spatial gradients may be
problematic.
In addition to the above, the subcommittee recommends that EPA seek input
from users before and after the TRIM.FaTE methodology is developed to maximize its
utility, to know how it is being used, and to guard against inappropriate uses; to provide
documentation of recommended and inappropriate applications; to provide training for
users; to test the model and its subcomponents against current data and models to
evaluate its ability to provide realistic results; and to apply terminology consistently.
We congratulate EPA for its initial effort, and look forward to conducting
additional reviews of other TRIM modules over the next few years. We look forward to
the response from the Assistant Administrator for the Office of Air and Radiation.
Sincerely,
Dr. Ishwar P. Murarka, Chair ^rjoanTv1. Daisey, Chair
Environmental Models Subcommittee £/ Science Advisory Board
Science Advisory Board
-------
NOTICE
This report has been written as part of the activities of the Science Advisory
Board, a public advisory group providing extramural scientific information and advice to
the Administrator and other officials of the Environmental Protection Agency. The
Board is structured to provide balanced, expert assessment of scientific matters related
to problems facing the Agency. This report has not been reviewed for approval by the
Agency and, hence, the contents of this report do not necessarily represent the views
and policies of the Environmental Protection Agency, nor of other agencies in the
Executive Branch of the Federal government, nor does mention of trade names or
commercial products constitute a recommendation for use.
-------
ABSTRACT
The Environmental Models Subcommittee (EMS) of the Executive Committee
(EC) of the Science Advisory Board (SAB) reviewed the TRIM.FaTE Module of the
Total Risk Integrated Methodology (TRIM) being developed by the Office of Air Quality
Planning and Standards (OAQPS) in the Office of Air and Radiation (OAR). TRIM is
designed to provide a method for integrating multimedia, multipathway sources of
pollutants to more accurately estimate exposure to pollutants and effects from
environmental releases. The Subcommittee found the development of TRIM and the
TRIM.FaTE module to be conceptually sound and scientifically based. It is a very
complex model in terms of interconnections, so care needs to be taken to insure that it
is applied appropriately and produces realistic results. Recommendations are made to
seek input from users before and after the methodology is developed to maximize its
utility, to know how it is being used, and to guard against inappropriate uses; to provide
documentation of recommended and inappropriate applications; to provide training for
users; to test the model and its subcomponents against current data and models to
evaluate its ability to provide realistic results; and to apply terminology consistently.
Keywords: TRIM, TRIM.FaTE, Environmental Models
-------
US ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD
ENVIRONMENTAL MODELS SUBCOMMITTEE OF THE
EXECUTIVE COMMITTEE
CHAIR
Dr. Ishwar Murarka, Chief Scientist and President, ISH Inc., Cupertino, CA
MEMBERS
Dr. Steven M. Bartell, Vice President, SENES Oak Ridge, Inc., Oak Ridge, TN
Dr. Calvin Chien, Senior Environmental Fellow, E.I. DuPont Company, Wilmington, DE
Dr. Kai-Shen Liu, Epidemiologist, California Department of Health Services,
Environmental Health Laboratory Branch, Berkeley, CA
Dr. Paulette Middleton, Deputy Director, RAND Environmental Science and Policy
Center, Inc., Boulder, CO
CONSULTANTS
Dr. M. Bruce Beck, Professor & Eminent Scholar, Warnell School of Forest
Resources, University of Georgia, Athens, GA
Dr. Linfield Brown, Professor, Department of Civil and Environmental Engineering,
Tufts University, Medford, MA
Dr. Arthur J. Gold, Professor, Department of Natural Resources Science, University of
Rhode Island, Kingston, Rl
Dr. Helen Grogan, Cascade Scientific, Inc., Bend, OR
Dr. Wu-Seng Lung, Professor, Department of Civil Engineering, University of Virginia,
Charlottesville, VA
Dr. Jana Milford, Associate Professor, Department of Mechanical Engineering,
University of Colorado, Boulder, CO
Dr. Mitchell Small, Head, Department of Civil & Environmental Engineering and
Engineering & Public Policy, Carnegie Mellon University, Pittsburgh, PA
Dr. Thomas Theis, Professor & Chair, Department of Civil and Environmental
Engineering, Clarkson University, Potsdam, NY
111
-------
SCIENCE ADVISORY BOARD STAFF
Dr. John R. Fowle, III, Deputy Staff Director/Designated Federal Officer,
Environmental Protection Agency, Science Advisory Board (1400) 401 M Street,
SW, Washington, DC 20460
Mrs. Priscilla Y. Tillery-Gadson, Program Assistant, Environmental Protection
Agency, Science Advisory Board (1400), 401 M Street, SW, Washington, DC
20460
IV
-------
TABLE OF CONTENTS
1. EXECUTIVE SUMMARY 1
2. INTRODUCTION 8
3. OVERVIEW COMMENTS AND OBSERVATIONS 9
4. RESPONSE TO CHARGE 10
Charge Question 1: Is the overall conceptual TRIM approach appropriate? . . 10
Charge Question 2: The TRIM approach is designed for the explicit
treatment of uncertainty and variability, including both model
uncertainty and parameter uncertainty. Is the spatial
compartmental mass-balance approach commensurate with
quantifying uncertainty and variability in a scientifically defensible
manner? 14
Charge Question 3: The TRIM.FaTE Module is the environmental fate,
transport, and exposure component of TRIM. Is the overall
conceptual approach represented in the TRIM.FaTE Module
appropriate, given the underlying science, EPA policy, and
regulatory needs? 18
Charge Question 4: The TRIM approach is designed to be flexible and
to allow for a tiered approach, to function as a hierarchy of models,
from simple to complex, as needed 20
4a. As implemented at this time, is the TRIM.FaTE Module, with
its 3-dimensional, spatial compartmental mass-conserving
approach to predicting the movement of pollutant mass
over time, appropriate from a scientific perspective? 21
4b. Is the TRIM.FATE Module, as designed, an appropriate tool,
when run either at a screening level or for a more refined
analysis, for use in provid- ing information for regulatory
decision-making? 23
Charge Question 5: Does the TRIM.FaTE Module, as it has been
conceptualized, address some of the limitations associated with
other models? 23
Charge Question 6: Does the TRIM.FaTE Module, as it has been
conceptualized and demonstrated to date, facilitate future integration
with appropriate data sources (e.g., CIS) and applications? 24
5. CONCLUSION 26
6. LITERATURE CITED R-1
-------
1. EXECUTIVE SUMMARY
The Environmental Models Subcommittee (EMS) of the Executive Committee
(EC) of the Science Advisory Board (SAB) reviewed the TRIM.FaTE Module of the
Total Risk Integrated Methodology (TRIM) being developed by the Office of Air Quality
Planning and Standards (OAQPS) in the Office of Air and Radiation (OAR). TRIM is
designed to provide a method for integrating multimedia, multipathway sources of
pollutants to more accurately estimate exposure to pollutants and effects from
environmental releases. The Subcommittee addressed six charge questions:
Charge Question 1: Is the overall conceptual TRIM approach appropriate, given
the underlying science, EPA policy, and regulatory needs (i.e., what are the
strengths and the weaknesses)?
The conceptual approach for TRIM appears to be technically defensible. The
prospect of using TRIM for screening level as well as for more refined analyses offers
an appropriate way to use science to inform regulatory decision-making. However,
because the approach is still evolving, and subject to change, it is understandably
unclear as to how the overall methodology will operate for addressing the spectrum of
regulatory questions. For example, one of the strengths of model design is that it can
be expanded or simplified as appropriate for a particular application. At the same time,
the very flexibility of the system means that care must be taken to guard against
developing unnecessarily complex or site-specific modeling applications that can lead
to inconsistency or unnecessary delays in the implementation of regulatory policy. In
light of this, the Subcommittee recommends that utmost care be exercised to guard
against developing unnecessarily complex or inconsistent modeling applications.
The Subcommittee notes that the most important challenge facing the TRIM
modeling enterprise is the lack of available data for fate, transport, exposure and risk
processes. This lack of data will limit the ability to estimate the parameters for many of
the chemicals being modeled. Insufficient data also hinder validation efforts. The
Subcommittee, therefore, recommends that the Agency identify and acquire significant
additional field data to estimate modeling parameters and to "validate" the model
components and other aspects of the modeling system.
-------
Charge Question 2: The TRIM approach is designed for the explicit treatment of
uncertainty and variability, including both model uncertainty and parameter
uncertainty. Is the spatial compartmental mass-balance approach
commensurate with quantifying uncertainty and variability in a scientifically
defensible manner?
In principle, there is no reason why an analysis of uncertainty could not be
conducted at this stage. However, it is not possible to ascertain whether the spatial
compartmentalization will be a significant source of uncertainty in generating
predictions. It is apparent that choosing a coarse compartmentalization could introduce
significant errors. Another important source of uncertainty is the use of linear models
for physical, chemical, and biological processes. Model uncertainty and parameter
uncertainty will depend a great deal on the extent and quality of the data available for
the modeling effort.
The Subcommittee accepts that at present the TRIM.FaTE development team
has yet to enter into substantial work on the analysis of sensitivity and uncertainty, as it
appears that thus far only a local sensitivity analysis has been implemented under
prototypical, test-case conditions. The Subcommittee recommends that the literature
on sensitivity and uncertainty analysis be reviewed thoroughly prior to making firm
choices on the specific forms for incorporating sensitivity and uncertainty analysis into
TRIM. Agency mandates and other considerations notwithstanding, the Subcommittee
recommends that the TRIM developers take the necessary time to reflect on what value
the analysis of uncertainty and sensitivity will add to their work and the resulting
framework for modeling. The Subcommittee endorses the notion of incorporating
checks and balances into the model framework and software in order to guard against
future users making inconsistent or illogical choices for numerical parameterization of
the model and its uncertainty analysis.
The Subcommittee views the issue of "validation" of TRIM to be an intractable
issue because TRIM will never be capable of (in)validation in the classical sense.
However, the Subcommittee recommends that the history matching and qualitative peer
review should not be set aside, and that the Agency watch for novel methods for
quantitatively assuring the quality of models as tools for fulfilling specified predictive
tasks. The Subcommittee endorses the incorporation of checks and balances into the
TRIM software/framework to guard against future inconsistent or illogical choices for
numerical parameterization.
-------
Charge Question 3: The TRIM.FaTE Module is the environmental fate, transport,
and exposure component of TRIM. Is the overall conceptual approach
represented in the TRIM.FaTE Module appropriate, given the underlying
science, EPA policy, and regulatory needs (i.e., what are the strengths and
weaknesses of the approach)?
The Subcommittee finds that the TRIM.FaTE module is conceptually sound and
aims at an appropriate level of complexity. However, the terminology used to describe
the model structure and components is sometimes confusing and contradictory.
Chemical transformation algorithms are not clearly characterized. Since this module is
still under development, the Subcommittee hopes the developers will pay close
attention to resolving this confusion. The Subcommittee further notes that it is difficult
to understand the difference between application of the module in a screening capacity
versus its application in a more in-depth analysis mode, further examples and
clarification are required.
The Subcommittee finds that the TRIM.FaTE module has several strengths. It
meets the requirements of scientific and technical defensibility. It is flexible in
application to different exposure scenarios, and it is able to address exposures relevant
to human health and ecological risk assessment. Further it is user friendly. The
challenge ahead will be in developing and implementing the details beyond the linear
algorithms incorporated in the prototype version. OAQPS' emphasis on the steady
state distribution of contaminants may prove to be an important limitation. The
Subcommittee understands that the Agency is fully aware of the bounds and limitations
of the methodology. The challenge that remains is to establish how accurate and
precise the model results have to be in order to effectively enable human health and
ecological risk assessments to inform regulatory decision-making.
The Subcommittee recommends that the Agency develop a users guide that
includes a description of the strengths and the limitations of the TRIM.FaTE module as
it is currently constructed or envisioned.
Charge Question 4: The TRIM approach is designed to be flexible and to allow for
a tiered approach, to function as a hierarchy of models, from simple to
complex, as needed.
-------
4a. As implemented at this time, is the TRIM.FaTE Module, with its
3-dimensional, spatial compartmental mass-conserving approach to
predicting the movement of pollutant mass over time, appropriate from a
scientific perspective?
TRIM.FaTE is an ambitious attempt to model fate and transport of air pollutants
through all aspects of an ecosystem. By design it is very broad and encompasses the
different media, pollutant transformations and exchanges. The Subcommittee finds that
TRIM.FaTE, as currently presented and configured, has not been checked against a
detailed set of observed, spatially varying real world environmental concentration data.
Furthermore, with its highly aggregate representation of environmental compartments,
it is unlikely that it can be effectively used to address fully variable 3-dimensional
spatial analysis. Structurally, intermedia mass transfer capabilities have been
articulated, but their scientific "validity" awaits bench-marking against observational
data and other media-specific models. TRIM.FaTE consists of a series of hypotheses
that are connected to simulate the complexities inherent in a multimedia environment.
It is these hypotheses that need to be scientifically tested to elucidate the efficacy and
the limitations of the overall model. A number of other process-based models for air,
surface water or ground water systems (e.g., MODFLOW for 2- and 3-dimensional
groundwater flow, Chesapeake Bay Water Quality Model) may offer greater value for
certain applications than does the TRIM.FaTE module, particularly when transport and
concentration gradients within the particular medium dominate the fate and exposure
outcome.
The Subcommittee recommends that the module be constructed in a fashion that
will permit the component results to be disaggregated and studied to build confidence
in the overall prediction of the model. The model needs to include an internal tracking
and accounting system that will permit scrutiny of smaller segments or
sub-components. The Subcommittee identified several important issues that may limit
the applicability and credibility of the current prototype of the module, including the
omission of dispersion phenomena throughout the model, especially turbulent diffusion
in air. Neglecting horizontal diffusion as an intercompartmental transport mechanism in
the prototype appears to be a serious limitation, especially for short-term simulations.
The effect of this assumption is clearly apparent in the results with constant
meteorology, and it also affects the results for the case with variable meteorology.
Such limitations will be particularly important when chemical properties and site-
specific conditions result in significant local concentration gradients and/or temperature
effects.
-------
4b. Is the TRIM.FaTE Module, as designed, an appropriate tool, when run
either at a screening level or for a more refined analysis, for use in
providing information for regulatory decision-making? Given the module
design (i.e., the potentially large number of parameters and associated
uncertainty and variability), is TRIM.FaTE suitable to support regulatory
decisions?
The Subcommittee, absent additional testing and evaluation, cannot provide an
assessment or recommendation regarding the appropriateness of the module as a
decision-making tool. At this point in its development, the module seems to hold
promise as a screening tool providing results to inform decision-making.
Charge Question 5: Does the TRIM.FaTE Module, as it has been conceptualized,
address some of the limitations associated with other models (e.g., non-
conservation of mass, steady state approach, inability to quantify
uncertainty and variability, limited range of receptors and processes
considered)? Are there other limitations that the TRIM.FaTE model should
address?
The Subcommittee finds that the TRIM.FaTE module has the excellent feature of
conserving mass for chemicals undergoing first-order linear mass transfer and
transformation processes. It is however, uncertain at this time as to how the proposed
methods can be adapted for chemicals subject to non-linear higher-order processes.
Since TRIM.FaTE is a compartmental model, there are no explicit vertical or horizontal
dimensions, though these may be added by including additional compartments for a
given medium. The Subcommittee recommends that the Agency consider providing
additional methods and guidance to assist users in selecting the appropriate level of
spatial and temporal resolution necessary to obtain adequate precision and accuracy in
the results.
The flow model for air transport is highly simplified. The Subcommittee
recommends that the Agency carry out additional evaluation of available air models
and make the appropriate selection of additional process modules or components for
incorporation in the TRIM.FaTE module. In particular, incorporation of turbulent
diffusion processes will be important for chemicals where atmospheric concentrations
vary significantly from source-release to exposure locations.
-------
The predictive capability of the module is limited because of the gross transfer of
mass between sources, receptors and sinks. The Subcommittee recommends that
results from TRIM.FaTE be compared with results obtained from the existing "single-
media linked models" to establish the advantages and limitations of the TRIM multi-
media transport and fate model.
The Subcommittee observes that given the stage of its development, TRIM.FaTE
has yet to be substantially tested and exercised using methods for the analysis of
sensitivity and uncertainty. The Subcommittee recommends that the literature on
sensitivity and uncertainty analysis be reviewed and examined thoroughly by the
developers prior to making choices on the specific forms of parametric and model-
structure sensitivity and uncertainty analysis.
Charge Question 6: Does the TRIM.FaTE Module, as it has been conceptualized
and demonstrated to date, facilitate future integration with appropriate data
sources (e.g., GIS) and applications (e.g., multipathway exposure
assessment for humans)?
The Subcommittee finds that the TRIM.FaTE module could conveniently and
effectively be integrated with data sources such as GIS, provided the spatial scales in
the GIS are congruent with those of TRIM.FaTE. However, coupling TRIM.FaTE with
other more complex models which generate continuous spatial gradients may be
problematic.
The linkage between the TRIM.FaTE module and the Exposure Event Module
has not yet been established. The results produced are not directly useable for human
exposure assessment, because TRIM.FaTE does not generate distributions of indoor
air pollutants, which are the most important input for the Exposure Event Module. The
Subcommittee expects that the Agency will be conducting further work to produce
distributions of pollutants in all major microenvironments and media so that a full,
integrated and reliable exposure assessment for humans becomes possible.
Overall, the Subcommittee found the development of TRIM and the TRIM.FaTE
module to be conceptually sound and scientifically based. It is a very complex model in
terms of interconnections, so care needs to be taken to insure that it is applied
appropriately and produces realistic results. Recommendations are made to seek input
from users before and after the methodology is developed to maximize its utility, to
know how it is being used, and to guard against inappropriate uses; to provide
-------
documentation of recommended and inappropriate applications; to provide training for
users; to test the model and its subcomponents against current data and models to
evaluate its ability to provide realistic results; and to apply terminology consistently.
-------
2. INTRODUCTION
The Environmental Models Subcommittee (EMS) of the Executive Committee
(EC) of the Science Advisory Board (SAB) met on May 5-6, 1998 to review the
TRIM.FaTE Module of the Total Risk Integrated Methodology (TRIM). TRIM is being
developed by the Office of Air Quality Planning and Standards (OAQPS) in the Office
of Air and Radiation (OAR) to provide a method for integrating multimedia,
multipathway sources of pollutants to more accurately estimate exposure to pollutants
and effects from environmental releases.
This initial review of TRIM, which focuses on TRIM.FaTE, the environmental
distribution and fate component of the model, is considered an SAB Advisory since it
provides peer review of an Agency work-in-progress. The goal of an SAB Advisory is
to provide suggestions to the Agency for mid-course adjustments that will refine the
ultimate product. In this case, the intent of an early SAB review of the TRIM.FaTE
module is to facilitate the completion of the TRIM multimedia, multipathway computer
methodology. The Subcommittee expects to conduct Advisory reviews of the other
TRIM modules (i.e., exposure, uptake, biokinetics, dose/response, and risk
characterization) over the next few years.
The materials provided to the SAB for review consisted of:
a) the Total Risk Integrated Methodology: Implementation of the TRIM
Conceptual Design through the TRIM.FaTE Module, and;
b) the Total Risk Integrated Methodology: Technical Support Document for
the TRIM.FaTE Module
The charge to the Subcommittee (Attachment A) contained six questions
focusing on the overall conceptual TRIM approach, the treatment of uncertainty and
variability, the appropriateness of the overall conceptual approach represented in the
TRIM.FaTE module, the appropriateness of the approach to predict the movement of
pollutant mass over time, the suitability of TRIM.FaTE to support regulatory decisions,
the extent to which the TRIM.FaTE module addresses some of the limitations
associated with other models, and the ability of the TRIM.FaTE module to facilitate
future integration with appropriate data sources (e.g., CIS) and applications (e.g., multi-
pathway exposure assessment for humans).
8
-------
3. OVERVIEW COMMENTS AND OBSERVATIONS
The general conceptual approach of TRIM is consistent with the state-of-the-art
of multimedia pollutant fate and transport models. The use of a compartmental mass-
conserving approach to predict the movement of pollutant mass over time is
appropriate from a scientific perspective. The idea of using the model in a hierarchical
mode first for screening and subsequently for more refined analyses is appropriate to
inform regulatory decision-making. The TRIM concept represents a complex and
comprehensive set of interconnections. EPA has gone beyond the current available
modeling to create an integrated framework for dealing with multimedia transfers in a
more self-consistent, and, hopefully, more useful fashion.
The major concerns regarding TRIM.FaTE at this point center around the extent
to which it:
a) represents reality and is scientifically verified
b) can provide defensible information for decision-making
c) is user-friendly
-------
4. RESPONSE TO CHARGE
Charge Question 1: Is the overall conceptual TRIM approach appropriate, given
the underlying science, EPA policy, and regulatory needs (i.e., what are the
strengths and the weaknesses)?
The conceptual approach for TRIM appears to be technically defensible.
However, because the approach is still evolving, and subject to change, it is unclear
how the overall TRIM methodology will operate. TRIM'S strength is its multimedia,
multiexposure pathway design that for atmospheric releases will allow consideration of
the exposure pathways, other than direct inhalation, that are critical for the assessment
of likely ecological or human health risks.
The focus of the TRIM.FaTE conceptual approach is on the impact of
atmospheric releases of persistent chemicals. Currently, the system is being
developed to allow the assessment of organic and inorganic chemicals, specifically air
pollutants; EPA is currently developing model parameters for mercury, arsenic,
benzo(a)pyrene and phenanthrene. A technical strength of the approach is the explicit
and full accounting of the distribution of the mass of the chemicals across the
environment. This is important when accounting for the cycling of chemicals in the
environment that may result in ecological or human health risk via a variety of exposure
pathways. A further strength of the TRIM approach is the flexibility accorded by its
open architecture. The modular design allows assessment of different temporal and
spatial scales, and of different human and ecological endpoints. It allows the system to
be expanded or simplified as appropriate for a particular application. At the same time,
the very flexibility of the system means that care must be taken to guard against
developing unnecessarily complex or site-specific modeling applications that can lead
to inconsistency or unnecessary delays in the implementation of regulatory policy. It is
important to remember that TRIM'S utility is for practical applications; not for research.
Still, the flexible architecture should be utilized to test and incorporate new research
findings as they evolve for the overall improvement of the module and the regulatory
decisions that are informed by it.
TRIM.FaTE aims at an intermediate level of complexity. It does this by
expressing the physical system being modeled as a series of linked compartments, a
simplifying approach, but then adds to it considerable complexity in the form of
pollutant-specific reactions, uptake, and intermedia mass transfer and transport
10
-------
(advective and diffusive) between compartments. It is doubtful that a box-type
compartment model of this type could become much more complex, the next level being
models which incorporate precise physical boundary sizes and shapes, and generate
continuous concentration gradients within model domains.
For most of the hazardous chemicals for which TRIM.FaTE is intended, the
complex and often uncertain mechanisms which control partitioning and transformation
among multiple environmental media are such that only order-of-magnitude accuracy is
possible for predictions of environmental concentrations, at best within a factor of two.
For this type of assessment, completely mixed compartment models are wholly
appropriate. Furthermore, the insights needed for many environmental fate and
transport applications are either insensitive to the exact locations and shapes of
boundaries, or the users are relatively unconcerned with the time a certain
concentration of a contaminant reaches an exact location. (There are site-specific
applications where the location and shapes of boundaries, or the need for computing
intradomain continuous gradients may be important, for example oil spills, sea water
intrusion into groundwater, estuaries with significant tidal fluxes, etc.) Thus, although
from a mathematical point of view environmental fate and transport is a boundary value
problem, the use of completely mixed volumes of generally accurate size (boxes), but
inexact dimensions, allows the greatly simplifying step to be taken of expressing
pollutant mass balances as a series of coupled ordinary differential equations instead
of partial differential equations. This is certainly acceptable, as long as the limitations
of the TRIM approach are matched against the desired uses in practice.
Although the basic approach of TRIM.FaTE is reasonably clear from the Status
Report, in particular the figures and examples, the terminology used to describe the
module is sometimes confusing and, in some cases, contradictory. This may be
important as the TRIM program continues to unfold and new participants and users are
added, and as eventual users external to the EPA attempt to understand its basis.
Chapter three introduces a hierarchical set of terms (volume element, domain, domain
type, domain instance, cell, link, etc.) which are used throughout, but other sections
sometimes use these terms differently, and borrow terms from related areas (for
example thermodynamics, systems theory, and numerical methods and programming)
that are used either without definition, or with definitions that seem to vary.
On p. 3-13: the construction and usage of "sinks" in the current prototype of the
TRIM.FaTE module is not really consistent with the use of this terminology in general
environmental modeling. It is clear that there is a need to develop a capability to track
11
-------
the decomposition products of reactive contaminants, particularly if these products are
more toxic than the "parent" compounds. However, simply partitioning these
compounds into a "box" of degradation products might help to maintain mass-balance
but miss the point in constructing relevant exposure scenarios for risk assessment. It is
important to clearly define all terms and use them consistently. A given individual from
an allied but different background may not be comfortable with the definitions, but at
least he/she will know what is meant within the context of the TRIM effort.
The model is consistent with the Agency's Ecological Risk Assessment and
Human Health Risk Assessment Guidelines, but one challenge facing the TRIM
modeling enterprise is the lack of data needed to assess ecological and human health
risk outcomes to accurately inform EPA's environmental decision-making. There are
two aspects to this. First, there are insufficient data to define the parameters relating to
the chemicals to be modeled (i.e., there are 188 hazardous air pollutants that are
subject to regulation, none of which is sufficiently well-characterized). Secondly, there
are insufficient data available to carry out model "validation" efforts. A major effort is
needed to meet these data needs. This problem is not unique to the TRIM approach.
It is very important that data sets other than those used to develop the models be
sought to test the models. It is recommended that a number of "validation" data sets be
identified so that different components and aspects of the system can be tested
thoroughly. This activity should not be confined to the developmental stages and early
application, but should occur on an ongoing basis.
Further, the model does not specify how to develop the mass transport
coefficients for these configurations for surface waters. The model cannot handle 2-D
horizontal and full 3-D spatial configurations. Kinetic processes in the water column
are very limited as well. While the overall structure and formulation of TRIM.FaTE is
scientifically sound, particular mass transport and chemical transformation terms within
the system are not clearly characterized. The extent to which these are included, and
the particular representations used, need to be carefully evaluated and supported. To
support regulatory needs, the configured module must be "validated" by site-specific
data. The TRIM document does not currently address this issue.
Sufficient data are unlikely to be available for any chemical or set of chemicals
to "validate" all aspects of TRIM.FaTE simultaneously. As such, validation of
TRIM.FaTE may only be possible on a piecemeal basis, component by component.
This could, of course, be viewed as a flaw in the module's design. Yet TRIM.FaTE is
no different in this respect from any other large, complex model being used today as a
12
-------
tool for supporting policy development, screening analyses, and pragmatic, decision
support analyses. There is in general a growing realization that the issue of
determining the trustworthiness and utility of such models is not a closed-form problem
with a simple "yes" or "no" answer. This is not to say that TRIM.FaTE can not therefore
be used with confidence. Rather, the challenge is for alternative, supplementary, more
appropriate methodological frameworks to be developed and applied for evaluating the
legitimacy of such policy-forecasting models (as further discussed under Charge 2
below). In this respect, the Subcommittee notes, in particular, previous Agency
experience in coming to terms with this problem, in evaluating models of the fate of
lead in the environment (Oreskes, 1998).
The Subcommittee strongly recommends that substantial guidance and training
be provided to users of TRIM to ensure its efficient and appropriate application and to
facilitate its wide acceptance. Detailed guidance should be provided on the
applicability of the complete mixing assumption, and on the selection of appropriate
time steps and compartmental volumes. The selection of time steps and spatial
resolution also needs to be carefully matched to the available data and parameter
values. The prototype application assumes equilibrium for most processes, but for
some processes this assumption may not be justified. Therefore, detailed guidance is
needed on when equilibrium assumptions are appropriate, and/or when sensitivity
analysis may be needed to evaluate errors associated with neglecting transient effects.
The Subcommittee also recommends that the EPA provide a clear articulation of
the tasks to be accomplished by TRIM to help ensure its appropriate use and
applications. The SAB recommends that TRIM be designed with self-diagnostic
features to assist the user during its application. Careful consideration should also be
given to the method used for visualization of the results as well as their uncertainties.
This should be designed with consideration of the specific assessment goal for each
application, since it will impact the interpretation of the results and how well they inform
the decisions reached.
Besides its use at EPA for evaluating hazardous air pollutants, TRIM may find
utility in a number of other scientific assessment or educational applications. Examples
include evaluations of fate, transport, exposure and risk for ranking or comparing
chemical emissions for product design/life cycle assessments, and teaching graduate
level classes on fate and transport modeling. We expect such uses to be facilitated,
because the model is intended to be operated on a PC and to be user-friendly.
13
-------
Charge Question 2: The TRIM approach is designed for the explicit treatment of
uncertainty and variability, including both model uncertainty and parameter
uncertainty. Is the spatial compartmental mass-balance approach commensurate
with quantifying uncertainty and variability in a scientifically defensible manner?
In responding to this question the Subcommittee adopted two working
definitions:
a) Uncertainty is the term used to acknowledge the fact that the "true" value
of a quantity (or variable) is not available; there are only a number of
observed or estimated values available and the spread of these reflects,
in some sense, the uncertainty associated with the "true" value.
b) Variability captures the familiar idea that a quantity (or variable) may
have different "true" values at different instants in time, at different
locations in space, and/or from one individual to the next. Variability is an
intrinsic characteristic of a quantity; unlike uncertainty, it is not reduced by
better knowledge and further study, only better characterized.
The Agency has acknowledged the need for TRIM.FaTE to account explicitly for
uncertainty and variability and has apparently built this into the software framework for
the module. The Subcommittee did not see a demonstration, or any output from an
analysis of model uncertainty.
The ability to comment on the reliability, uncertainty and applicability of an
assessment determines the extent to which an assessment can provide useful decision
support for decision-making. While uncertainty and sensitivity analysis are part of the
overall system, it is not clear how this analysis will be incorporated and how it will be
presented as part of the overall assessment. This needs to be clearly defined.
The Subcommittee accepts that at present the TRIM.FaTE development team
has yet to enter into substantial work on the analysis of sensitivity and uncertainty, as it
appears that thus far only a local sensitivity analysis has been implemented under
prototypical, test-case conditions. Thus, the Subcommittee strongly recommends that
the literature on sensitivity and uncertainty analysis be reviewed thoroughly prior to
making strategic choices on the specific forms of such analysis to be incorporated
within TRIM. We say this because this literature is widely dispersed and, perhaps
more importantly, has grown substantially over the past three or four years. In
14
-------
particular, the Second International Symposium on Sensitivity Analysis of Model Output
was held April, 1998 in Venice, Italy. Its proceedings provide a fairly comprehensive
survey of what is a rapidly evolving field, notably in the area of novel techniques for
global sensitivity analysis, especially as applied in the areas of industrial management
systems, reliability engineering, and assessments of the storage of nuclear waste
materials.
In addition, the Subcommittee is concerned about the issue of correlation among
sources of uncertainty, notably among the errors associated with model parameter
(coefficient) estimates. Consideration of correlated errors is important for two reasons.
First, during calibration of the model, an error in the estimate of one parameter may
cancel out the effects of that parameter on another, with the model user being unaware
of this potential source of uncertainty. The Subcommittee is concerned that the
transfer coefficients in TRIM.FaTE, which are understood to be, in fact, linear sums of
several other coefficients, may be especially prone to this kind of error. Second, when
the model is used for the purposes of forecasting, the consequences of accounting for
the effects of correlated errors can be significant, possibly markedly changing the level
of uncertainty associated with the forecast states.
While calibration of a model is usually understood as serving the purpose of
generating estimates of the unknown (or imprecisely known) parameters in a model, it
can also be viewed as a means of reducing the uncertainty attaching to the model, prior
to calibration. When viewed in this light, it is easy to appreciate how good quality data
for calibration, i.e., field observations with low uncertainty, can lead to lower residual
uncertainty in the model after calibration, which, in turn, should facilitate the generation
of forecasts with lower levels of uncertainty.
Where EPA makes comments in anticipation of the analysis of uncertainty/
sensitivity in the development of TRIM, they suggest that incorporating such analysis
will improve the science upon which the models are based. Because uncertainty
analysis often results in extensive numerical output, its use may imply a more
advanced, sophisticated representation of the problem. While sensitivity analysis can
help to identify where further data and research can be most beneficial in filling critical
knowledge gaps, it does not provide a more advanced scientific representation per se,
it only characterizes the current state of the scientific knowledge. The role and
limitations of sensitivity and uncertainty analysis should be clearly recognized and
acknowledged by TRIM developers and users.
15
-------
Mention is made in several places regarding or alluding to "validation" of the
modules of TRIM. In line with the comments made earlier in response to this charge
question, the Subcommittee would like to point out that models such as TRIM will
(arguably) never be capable of (in)validation in the classical sense, of having been
shown to match observed behavior in the past with, at the same time, a unique and well
defined set of estimates for the model's parameters. This issue has been dealt with at
some length in the paper of Beck et al. (1997). While the Subcommittee notes that
there may not yet be consensus on this, we believe that with respect to the use of
science to inform policy, the notion of model "validation" should not be seen as a matter
of classical scientific prediction - of making statements about future behavior that
subsequent observation may reveal to be true (or false) - but instead as a matter of
designing a tool appropriate for the given (predictive) task.
The Subcommittee does not suggest that history matching and (qualitative)
"validation" efforts should be put aside; rather, following developmental work with one
of the Agency's multi-media models, we suggest that novel methodologies may become
available for quantitatively assuring the quality of models as tools for fulfilling specified
predictive tasks (e.g., Chen and Beck, 1998 ). Indeed, given the novelty, and therefore
somewhat speculative character, of this shift in outlook on the notion of quality
assurance of policy and regulatory models, it may be appropriate to consider convening
a small workshop to assess the prospects for its success (there is also a growing
interest in this matter within the European Environment Agency). The authors of the
Status Report on TRIM.FaTE appear to be well aware of the problems of and new
trends in model validation, since they identify model-intercomparison as an essential
component of model confirmation and sensitivity/uncertainty analysis.
The Subcommittee notes the importance of accounting for the errors of model
structure in uncertainty/sensitivity analysis. In a number of disciplines, interest has
deepened beyond only accounting for the uncertainty arising from erroneous estimates
of parameters in a given (uncontested) conceptual structure for the model. There are
now efforts to attempt to quantify and account for the effects of uncertainty in this
structure (i.e., uncertainty about whether the number of state variables in the model
and the mathematical nature of their interactions have been correctly expressed).
However, widely accepted methods have not yet emerged and accounting for the
consequences of this structural error is far from being resolved.
16
-------
In principle, there is no reason why an analysis of uncertainty could not be
conducted with TRIM.FaTE. At this stage, however, it is not possible to indicate
whether the "spatial compartmentalization" of the model (i.e., this specific choice for the
structure of the model itself) will be a significant source of uncertainty in generating
predictions. It is obvious that choosing a coarse compartmentalization of the model's
structure could introduce significant errors into its predictions. However, if a more
refined scale of spatial differentiation were to be adopted, there would then be an
accompanying obligation to specify in much greater detail the spatial distributions of
the model's parameters, but with limited or often no field data by which to do so (thus
introducing other sources of uncertainty). The literature on the analysis of uncertainty
in modeling surface water quality, for example, suggests that little work has been
published on the consequences of such spatial approximation for predictive
uncertainty.
Although the Subcommittee has been charged with considering the specific
implications of spatial compartmentalization for the analysis of uncertainty and
variability, there is a need for this issue to be kept in perspective, relative to other
potential sources of error. For example, in a given application of TRIM.FaTE it could
be that a more important limitation of the model's structure may be its use (to date) of
merely linear chemical/biological interactions (see page 3-14 of the Status Report).
Model uncertainty and parameter uncertainty depend a great deal on the extent
and quality of the data that supports the modeling effort. If site-specific data are
available to independently derive the key model parameters, then the parameter
uncertainty and prediction uncertainty would be reduced. There is no mention about
using site-specific data to develop key model parameters in the documentation of
TRIM.
The Subcommittee endorses the stated intention of the TRIM development team
to incorporate checks and balances into the software/framework in order to guard
against the future user community making inconsistent or illogical choices for numerical
parameterization of the software for analyzing uncertainty.
17
-------
Charge Question 3: The TRIM.FaTE Module is the environmental fate, transport,
and exposure component of TRIM. Is the overall conceptual approach
represented in the TRIM.FaTE Module appropriate, given the underlying science,
EPA policy, and regulatory needs (i.e., what are the strengths and weaknesses of
the approach)?
There are several strengths to the TRIM.FaTE methodology. It appears, at least
conceptually, to meet the requirements of scientific and technical defensibility, flexibility
in application to different exposure scenarios (i.e., prototypes), ability to address
exposures relevant to human health and ecological risk assessment, and accessibility
for use by the regulatory user community.
The modular structure and process flow of TRIM.FaTE (e.g., Figure 3-4 of the
TRIM.FATE Status Report Document) recommends this approach and appears as a
powerful framework for developing the integrated exposure assessment capabilities
required to support human health and ecological risk assessment. Clearly, the
challenge will be in adding the appropriate level of detail to the overall methodology.
The potential weaknesses of the TRIM.FaTE approach are mainly in the
mechanics and not the overall concept of an integrated environmental fate model.
There appears to be a predisposition toward first order, linear "algorithms" in
implementing the module, despite a considerable technical literature on modeling that
identifies complex and nonlinear relationships for a number of the component
processes that determine the transport, distribution, and fate of contaminants in air,
soils, surface waters, groundwater, sediments, and, particularly, in biota. The inability
to incorporate growth of organisms in the characterization of bioaccumulation of
contaminants appears to be a limitation in the current prototype. There needs to be
some guidelines to warn the user when linear algorithms might lead to exposure
estimates that are unrealistic or simply incorrect.
TRIM.FaTE is a broad-based multi-media and multi-path model, but because it
lacks spatial dimensionality within compartments, it does not have the ability to handle
many processes. For example, while TRIM.FaTE can handle velocity in any direction,
it can not handle diffusive/dispersive transfer perpendicular to the longitudinal
direction, and it has a very limited configuration for mass transport in lakes which are
inherently multidimensional.
18
-------
The subcommittee recommends that consideration be given to implementing
TRIM.FaTE with more powerful (and standardized) programming languages. Even
though commercial spreadsheets (e.g., Excel) have become increasingly sophisticated
and powerful, the user remains constrained to machines and operating systems that
support Excel. In addition, there is computational (i.e., memory, disk space,
calculations) overhead in using Excel or other such software. For complex, multi-media
risk assessments of the kind that tend to justify the TRIM.FaTE, there may be
requirements for computing power that exceeds the capacity of personal computers
and MS-DOS or Windows. The TRIM.FaTE should be programmed into a "stand-
alone" and portable higher level language (e.g., C++, Fortran-90).
While the overall approach of TRIM.FaTE is appropriate, the module is not yet
ready for regulatory use. Regulating contaminants in each of the three environmental
media requires much more information and data than are currently available. Proper
use of the module is also an important issue. More specifically, the accuracy and
representation of numerical estimates for module input parameters and coefficients for
the specific site is equally important as the model code. For example, derivation of the
BOD deoxygenation rate (or contaminant biodegradation rate) from the field data is
essential. For toxic metals, determination of the partition coefficient is another key
question often raised by regulatory staff. The TRIM.FaTE module has no apparent
information on this.
It remains questionable whether the common risk assessments, either for human
health or ecological impacts, will require fully integrated, multi-media fate and transport
models. This all-encompassing approach, while justifiable from a certain perspective,
might indirectly result in the eventual loss or deemphasis of a considerable amount of
research and development that has produced a reliable set of transport and fate
models for different media and chemical classes (e.g., Jorgensen, 1994). One possible
alternative to the TRIM.FaTE approach would be to construct a distributed system of
environmental fate models with an intelligent user-interface to select the appropriate
model or models depending on the needs of the assessment (e.g., EPA IMES).
However, having made the decision to move toward a fully specified, multi-media
model, the TRIM.FaTE seems a reasonable approach.
A minor point to note is that the review of contaminant fate modeling in the
TRIM.FaTE documentation is rather incomplete and ignores many important
environmental models that provide useful exposure scenarios for risk assessment (e.g.,
EXAMS). TRIM also seems to ignore previous efforts aimed at identifying and
19
-------
organizing the available fate and exposure models into a useful framework for
application in risk assessment (e.g., EPA IMES).
Other limitations are adequately outlined on p. 6-8. The challenge remains to
address these weaknesses and importantly determine how accurate and precise the
model results have to be in order to effectively support human health and ecological
risk assessment.
Charge Question 4: The TRIM approach is designed to be flexible and to allow for
a tiered approach, to function as a hierarchy of models, from simple to complex,
as needed.
The word "tiered" should be removed from any reference to human health or
ecological risk assessment. This is a term that applies properly to architecture, not risk
assessment. In most instances in the status report, "tiered" really refers to a sequence
of increasingly detailed computations.
An overview of the applicability of the TRIM system for different chemicals, and
for various scientific and policy issues, needs to be provided. For example, for
relatively non-reactive chemicals that distribute themselves in a relatively uniform
manner within each media of the environment, the current framework and assumptions
probably would provide a reasonable assessment of pollutant fate and transport. For
other highly reactive chemicals that exhibit significant spatial gradients within a
medium, it may not. It would be useful to develop a "users guide" to the problem area
limitations, as well as the strengths of the system as it is currently constructed or
envisioned.
The multimedia environmental system represented by TRIM.FaTE is very
complex. This complexity is necessary for the level of comprehensiveness achieved by
the model, though much effort is needed and should continue to be placed on making
the system as accessible and user-friendly as possible. In particular, clear direction
and support is needed for users to help them understand the difference between
application of the model in a screening capacity versus in a more in-depth analysis.
This needs to be more clearly presented in the supporting documentation and case
study applications. The degree to which TRIM is truly user-friendly depends upon the
ease of use of the computer code and the clarity of the supporting documentation.
20
-------
4a. As implemented at this time, is the TRIM.FaTE Module, with its
3-dimensional, spatial compartmental mass-conserving approach to
predicting the movement of pollutant mass over time, appropriate from a
scientific perspective?
As noted above, TRIM.FaTE is a zero-dimensional and not a 3-dimensional
model. Since spatial variability is not included in the coupled governing equations,
TRIM.FaTE is a compartment (completely mixed) model, and there are no explicit
vertical or horizontal dimensions in the cell used to represent the various shapes of the
environment. Spatial variability is implicitly incorporated in the time step selected.
Therefore, the selection of the time step will have a significant impact on the results
due to variabilities in both spatial and temporal dimensions. Thus, all environmental
fate and transport problems which are particularly sensitive to true 3-dimensional
geometries should be modeled with great care using TRIM.FaTE. The report
recommends that a carefully chosen automated method be developed to evaluate the
spatial and temporal resolutions, and provide guidance to users on the appropriateness
of TRIM for specific applications.
Structurally, the mass transfer capabilities of the model are articulated, but the
scientific validity of the approach awaits benchmarking against observational data and
other explicit, media-specific, models. From a scientific perspective the
appropriateness of TRIM.FaTE hinges on its capacity to be evaluated and tested.
TRIM.FaTE is an ambitious attempt to model fate and transport of air pollutants through
all aspects of an ecosystem. By design it is very broad and encompasses the different
media, pollutant transformations and exchanges. Although the module predicts the
accumulation of released air pollutants in target organisms throughout the food chain,
there are few data sets that exist to compare with the overall model predictions.
TRIM.FaTE consists of a series of connected hypotheses to simulate the
complexities inherent in a multi-media environment. It is these hypotheses that can be
scientifically tested to elucidate the efficacy and the limitations of the overall model.
While TRIM.FaTE breaks new ground in its scope, TRIM.FaTE is not unique in the
processes modeled. A host of existing, narrowly focused models are key resources for
evaluating TRIM.FaTE (e.g., MODFLOW for 2- and 3-dimensional groundwater flow,
LEACHN for water and solute movement transformations, and plant uptake and
chemical reactions in the unsaturated zone, AGNPS for modeling agricultural nonpoint-
source pollution in a watershed).
21
-------
The Subcommittee recommends that the model be constructed in a fashion that
will permit the results to be separated into modules that can be individually verified to
build confidence in the overall model. As implemented at this time, the TRIM.FaTE
module is not constructed to permit sub-component evaluations.
Tracking and accounting within the TRIM.FaTE module is needed to isolate its
predictions and to permit benchmark comparison with data sets and other models that
have been widely evaluated and that are inherently better for particular modeling
applications. The components of the module need to be evaluated against data and
models that have been developed over a range of settings and spatial and temporal
scales. This will permit scrutiny of TRIM.FaTE transformation algorithms and the
parameters that are used within this component of the TRIM model. Reporting subsets
of the model results will permit evaluations of the efficacy of particular predictions in
response to the varying levels of aggregation of both the processes that compose a
particular transfer function and the input parameters.
As examples, there are considerable data and models to evaluate such internal
sub-component predictions as:
a) Estimates of sediment transport and overland flow between land parcels
and between the watershed and receiving waters at different time
intervals. The Subcommittee recommends that the TRIM.FaTE results be
compared (benchmarked) against results from well-tested site-specific
watershed models and that the comparison include data sets at much
finer spatial resolution than are currently proposed for TRIM.FaTE.
b) The relationship between contaminant loading to a receiving water body
and the mass of the contaminant within phases of the aquatic
environment (i.e., sediment vs. water).
c) The prediction of partitioning between the sediments and the water
column.
d) The prediction of the concentrations of specific pollutants within plants,
fish and wildlife that result from exposure to the mass of contaminant
within the aquatic and/or terrestrial setting.
22
-------
The Subcommittee identified several important simplifications within the current
prototype of the model that will limit its applicability to a number of localized air
pollutant problems and that could hinder its scientific acceptance, including omission of
dispersion phenomena throughout the model, especially turbulent diffusion in air.
Neglecting horizontal diffusion as an intercompartmental transport mechanism in the
prototype appears to be a serious limitation, especially for short-term simulations. The
effect of this assumption is clearly apparent in the results with constant meteorology,
and it also affects the results for the case with variable meteorology.
4b. Is the TRIM.FATE Module, as designed, an appropriate tool, when run
either at a screening level or for a more refined analysis, for use in provid-
ing information for regulatory decision-making? Given the module design
(i.e., the potentially large number of parameters and associated uncertainty
and variability), is TRIM.FaTE suitable to support regulatory decisions?
Although theoretically applicable for both screening and "refined" site-specific
analyses, the module appears most suited to screening analyses for idealized cases
with simple flows (e.g., absent wind shear or complex terrain) and simple geometry.
For more complicated situations, the need for many compartments to represent a given
medium and the fact that the volume elements are not addressed by spatial coordinates
within a compartment may prove excessively cumbersome.
In the absence of additional testing and evaluation, the Subcommittee cannot
provide an assessment or recommendation regarding the appropriateness of the
module as a decision-making tool. Clearly, science informs decisions, but many other
factors shape policies. At this point in its development, the module seems to hold
promise as a screening tool. With increased testing and evaluation, EPA will gain
insight into the acceptable types of applications along with the limitations and situations
where the module is unsuitable for use.
Charge Question 5: Does the TRIM.FaTE Module, as it has been conceptualized,
address some of the limitations associated with other models (e.g., non-
conservation of mass, steady state approach, inability to quantify uncertainty and
variability, limited range of receptors and processes considered)? Are there
other limitations that the TRIM.FaTE model should address?
The TRIM.FaTE module is noteworthy for its dynamic design systems, with
flexibility to simulate a range of temporal scales from hours to several years.
23
-------
TRIM.FaTE has the excellent feature of ensuring conservation of mass. This feature is
possible for both dynamic and steady-state calculations for chemicals with first-order
linear processes since the approach used is not based on linking different models for
different compartments or domain instances (the entire system is presented as a single
informational structure, i.e., one matrix). It is however, uncertain at this point as to how
the proposed methods can be expanded for higher-order non-linear processes,
particularly for the steady state matrix solution procedure.
The system of intercompartmental linkages is one of the most critical
components of the model. These linkages are specified by the users. While this
feature provides flexibility, it also requires a trained user. The development and
inclusion of an expert system is recommended to provide guidance to users.
Guidelines for specifying linkages may be based on physicochemical properties of the
chemicals (which control transfer and transformation processes), and/or characteristics
of the specified environmental landscape and ecosystems. This suggestion is
applicable to all the other flexible features in the model that require specification as
shown in Figure 2-4 of the TRIM.FaTE documentation.
Because only gross transfer of pollutants between sources, receptors, and sinks
is considered, the predictive capability of the module is limited. It is recommended that
results from TRIM.FaTE be compared with those from existing linked systems for
single-media cases, to evaluate the impact of this more aggregate, simplified simple
approach.
Charge Question 6: Does the TRIM.FaTE Module, as it has been conceptualized
and demonstrated to date, facilitate future integration with appropriate data
sources (e.g., GIS) and applications (e.g., multipathway exposure assessment for
humans)?
TRIM.FaTE is conveniently structured to use spatially arranged data sources
such as GIS, provided the spatial scales are congruent with those of TRIM.FaTE.
However, coupling TRIM.FaTE with other more complex models which generate
continuous spatial gradients may be problematic. Such might be the case, for example,
for the use of output data from an air transport model, which would require some
degree of spatial averaging before being input to TRIM.FaTE This raises a somewhat
larger issue regarding model uncertainty. It appears that most of the thought on
uncertainty for TRIM.FaTE revolves around the following factors: inherent parameter
uncertainty, the sensitivity of the output to various parametars, and the sensitivity of the
24
-------
output to variations in input. These are certainly valid concerns. However, less
obvious are uncertainties associated with the averaging process itself. For example,
might the level of uncertainty vary with the overall size of the area to be studied and
with the number of volumes used to represent the system? This issue, in turn, is
related to issues of "validation" and calibration, and the availability of data from actual
sites against which to compare TRIM.FaTE output.
As to the application of the results of the TRtM.FaTE module to the Exposure-
Event module, the link between the two has not been established yet. The inhalation
route has only been coarsely integrated into the TRIM.FaTE module. The key point is
that the TRIM.FaTE module, as it is currently described, will provide a concentration of
pollutants (chemicals) in ambient air (outdoors). Thus, to realistically address the
inhalation route it will be important to ensure that the Exposure Event module accounts
for the differences in air pollutant concentration in the microenvironments where
inhalation occurs. Distributions of indoor air pollutants, which are critical for the
assessment of human exposure to air contaminants, are not now provided by the
TRIM.FaTE module. Further work is needed to produce temporal and spatial
distributions of pollutants in various microenvironments and media, so that an
integrated exposure assessment for humans is possible. This will likely entail
estimating exposures to air pollutants in a cohort of humans by performing exposure
monitoring (e.g., office, living room, bedroom, car) coupled with specific activity
patterns (e.g., working, watching TV, sleeping, driving). Currently, the TRIM.FaTE
module does not produce concentration data for locations where people spend 90% of
their time.
25
-------
5. CONCLUSION
The Subcommittee found the development of TRIM and the TRIM.FaTE module
to be conceptually sound and scientifically based. It is a very complex model in terms
of interconnections, so care needs to be taken to insure that it is applied appropriately
and produces realistic results. The Subcommittee recommends that the Agency seek
input from users before and after the methodology is developed to maximize its utility,
to know how it is being used, and to guard against inappropriate uses; to provide
documentation of recommended and inappropriate applications; to provide training for
users; to test the model and its subcomponents against current data and models to
evaluate its ability to provide realistic results; and to apply terminology more
consistently.
The Subcommittee commends the Agency on its efforts and looks forward to
reviewing the other TRIM modules.
U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Boulevard, 12th Floor
Chicago, JL 60604-3590
26
-------
6. LITERATURE CITED
Beck, M.B., Ravetz, J.R., Mulkey, L.A., and T.O. Barnwell. 1997. On the Problem of
Model Validation for Predictive Exposure Assessments. Stochastic Hydrology
and Hydraulics: 11, 229-254.
Chen, J. and M. B. Beck. 1998. Quality Assurance of the EPA's Multi-Media Model in
Performing a Predictive Screening Task. Technical Report, U.S. Environmental
Protection Agency, EPA/600/R-98/106.
Jorgensen, L.A. and B. Pederson. 1994. Trace Metals in Fish Used for Time Trend
Analysis and as Environmental Indicators. Marine Pollution Bulletin: 28(1), 24-
32.
Oreskes, N. 1998. Evaluation (Not Validation) of Quantitative Models for Assessing
the Effects of Environmental Lead Exposure. Environmental Health
Perspectives Supplements: 106 (S.6), 1453-1460.
R-l
-------
D,STRIBUTIONUST
Administrator
EPA Headquarters Ubrary
Service
------- |