& EPA
United States
Environmental Protection
Agency
Office Of Air Quality
Planning And Standards
Research Triangle Park, NC 27711
EPA-453/D-99-002A
November 1999
Air
TRIM
Total Risk Integrated Methodology
TREVLFaTE
TECHNICAL SUPPORT DOCUMENT
Volume I: Description of Module
EXTERNAL REVIEW DRAFT
•*«,
Tmiport, A
Eipoiirc Module
(TRIMJ-«TE)
Exposure-Event Module
4 (TRIM Expo)
-------
EPA-453/D-99-002A
TRIM
Total Risk Integrated Methodology
TRIM.FaTE TECHNICAL SUPPORT DOCUMENT
Volume I: Description of Module
U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Bpulevard, 12th Floor
Chicago, JL 60604-3590
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
External Review Draft
November 1999
-------
[This page intentionally left blank.]
-------
Disclaimer
This document is an external review draft. It has not been formally released by the U.S.
Environmental Protection Agency and should not at this stage be construed to represent Agency
policy. It is being circulated for comments on its technical merit and policy implications, and
does not constitute Agency policy. Mention of trade names or commercial products is not
intended to constitute endorsement or recommendation for use.
NOVEMBER 1999
TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
ACKNOWLEDGMENTS
Acknowledgments
As described in this report, the Office of Air Quality Planning and Standards (OAQPS) of
the U.S. Environmental Protection Agency is developing the Total Risk Integrated Methodology.
The principal individuals and organizations in the TRIM.FaTE development effort and in the
preparation of this report are listed below. Additionally, valuable technical support for report
development was provided by ICF Consulting.
Robert G. Hetes, EPA, Office of Air Quality Planning and Standards
Deirdre L. Murphy, EPA, Office of Air Quality Planning and Standards
Ted Palma, EPA, Office of Air Quality Planning and Standards
Harvey M. Richmond, EPA, Office of Air Quality Planning and Standards
Amy B. Vasu, EPA, Office of Air Quality Planning and Standards
Deborah Hall Bennett, Lawrence Berkeley National Laboratory
Rebecca A. Efroymson, Oak Ridge National Laboratory
Steve Fine, MCNC-North Carolina Supercomputing Center
Dan Jones. Oak Ridge National Laboratory
John Langstaff, EC/R Incorporated
Bradford F. Lyon, Oak Ridge National Laboratory
Thomas E. McKone, Lawrence Berkeley National Laboratory & University of California, Berkeley
Rand) Maddalena. Lawrence Berkeley National Laboratory
NOVEMBER 1999 iii TRIM.FATE TSD VOLUME I (DRAFT)
-------
ACKNOWLEDGMENTS
The following EPA individuals reviewed a previous draft of this document.
EPA Models 2000 TRIM Review Team
Robert F. Carousel
National Exposure Research Laboratory
Office of Research and Development
*S. Steven Chang
Office of Emergency and Remedial Response
Office of Solid Waste and Emergency
Response
Ellen Cooter
National Exposure Research Laboratory
Office of Research and Development
Stan Durkee
Office of Science Policy
Office of Research and Development
Harvey Holm
National Exposure Research Laboratory'
Office of Research and Development
John S. Irwin
Office of Air Quality Planning and Standards
Office of Air and Radiation
* Team Leader
Linda Kirkland
National Center for Environmental Research
and Quality Assurance
Office of Research and Development
Matthew Lorber
National Center for Environmental
Assessment
Office of Research and Development
Haluk Ozkaynak
National Exposure Research Laboratory
Office of Research and Development
William Petersen
National Exposure Research Laboratory
Office of Research and Development
Ted W. Simon
Region 4
Amina Wilkins
National Center for Environmental
Assessment
Office of Research and Development
Review by Other Program Offices
Pam Brodowicz, Office of Air and Radiation, Office of Mobile Sources
William R. Effland, Office of Pesticide Programs
John Girman, Office of Air and Radiation, Office of Radiation and Indoor Air
Steven M. Hassur, Office of Pollution Prevention and Toxics
Terry J. Keating, Office of Air and Radiation, Office of Policy Analysis and Review
Russell Kinerson, Office of Water
Stephen Kroner, Office of Solid Waste
David J. Miller, Office of Pesticide Programs
NOVEMBER 1999
IV
TRIM.FATE ISO VOLUME I (DRAFT)
-------
PREFACE
PREFACE
This draft document, the TRIM.FaTE Technical Support Document, is part of a series of
documentation for the overall Total Risk Integrated Methodology (TRIM) modeling system. The
detailed documentation of TRIM'S logic, assumptions, algorithms, equations, and input
parameters is provided in comprehensive Technical Support Documents (TSDs) for each of the
TRIM modules. The purpose of the TSDs is to provide full documentation of how TRIM works
and of the rationale for key development decisions that were made. This report, which
supersedes an earlier version (U.S. EPA 1998a), documents the Environmental Fate, Transport,
and Ecological Exposure module of TRIM (TRIM.FaTE) and is divided into two volumes. The
first volume provides a description of terminology, model framework, and functionality of
TRIM.FaTE, and the second volume presents a detailed description of the algorithms used in the
module.
To date, EPA has issued draft TSDs for TRIM.FaTE (this report) and the Exposure-Event
module (TRIM.Expo TSD, U.S. EPA 1999a). When the Risk Characterization module
(TRIM.Risk) is developed, EPA plans to issue a TSD for it. The TSDs will be updated as needed
to reflect future changes to the TRIM modules.
The EPA has also issued the 1999 Total Risk Integrated Methodology (TRIM) Status
Report (U.S. EPA 1999b). The purpose of that report is to provide a summary of the status of
TRIM and all of its major components, with particular focus on the progress in TRIM
development since the 1998 TRIM Status Report (U.S. EPA 1998b). The EPA plans to issue
status reports on an annual basis while TRIM is under development.
In addition to status reports and TSDs. EPA intends to develop detailed user guidance for
the TRIM computer system. The purpose of such guidance will be to define appropriate (and
inappropriate) uses of TRIM and to assist users in applying TRIM to assess exposures and risks
in a variety of air quality situations.
Comments and suggestions are welcomed. The OAQPS TRIM team members, with their
individual roles and addresses, are provided below.
TRIM Coordination Deirdre L. Murphy
REAG/ESD/OAQPS
MD-13
RTF, NC 27711
[murphy. deirdre@epa. gov]
TRIM.FaTE Amy B. Vasu
REAG/ESD/OAQPS
MD-13
RTP,NC27711
[vasu.amy@epa.gov]
NOVEMBER 1999 v TRIM.FATE TSD VOLUME I (DRAFT)
-------
PREFACE
TRIM.Expo Ted Palma Harvey M. Richmond
REAG/ESD/OAQPS HEEG/AQSSD/OAQPS
MD-13 MD-15
RTF, NC 27711 RTF, NC 27711
[palma.ted@epa.gov] [richmond.harvey@epa.gov]
TRIM.Risk Robert G. Hetes
REAG/ESD/OAQPS
MD-13
RTF, NC 27711
[hetes.bob@epa.gov]
NOVEMBER 1999 vi TRIM.FATE TSD VOLUME I (DRAFT)
-------
ACRONYMS
ACRONYMS
B(a)P Benzo(a)pyrene
CAA Clean Air Act
CalTOX California Department of Toxic Substance Control's Risk Computerized Model
CART Classification and regression tree
CDF Cumulative distribution function
CRARM Presidential/Congressional Commission on Risk Assessment and Risk
Management
DOE United States Department of Energy
EPA United States Environmental Protection Agency
GIS Geographic Information System
HAP Hazardous air pollutant
IEM Indirect Exposure Methodology
I/O API Environmental Decision Support System/Models 3 Input/Output Applications
Programming Interface
ISMCM Integrated Spatial Multimedia Compartment Model
LHS Latin Hypercube Sampling
LSODE Livermore Solver for Ordinary Differential Equations
MC Monte Carlo
MCM Multimedia Compartment Model
MEPAS Multimedia Environmental Pollutant Assessment System
MPE Multiple Pathways of Exposure
NAAQS National ambient air quality standard
NAS National Academy of Sciences
NATA National Air Toxics Assessment
OAQPS EPA Office of Air Quality Planning and Standards
OSWER EPA Office of Solid Waste and Emergency Response
PAH Polycyclic aromatic hydrocarbon
PDF Probability distribution function
RIA Regulatory impact analysis
SAB Science Advisory Board
SMCM Spatial Multimedia Compartment Model
TRIM Total Risk Integrated Methodology
TRIM.Expo TRIM Exposure-Event module
TRIM.FaTE TRIM Environmental Fate, Transport, and Ecological Exposure module
TRIM.Risk TRIM Risk Characterization module
TSD Technical Support Document
WASP Water Quality Analysis Simulation Program
NOVEMBER 1999
Vll
TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
TABLE OF CONTENTS
TABLE OF CONTENTS
Disclaimer i
Acknowledgments Hi
Preface v
Acronyms vii
Table of Contents ix
1. Introduction 1-1
1.1 Goals and Objectives for TRIM 1-2
1.2 TRIM Design 1-4
1.3 TRIM Development 1-8
1.4 Phasing TRIM into OAQPS' Set of Modeling Tools 1-12
2. Introduction to TRIM.FaTE 2-1
2.1 Review of Existing Fate and Transport Models 2-1
2.2 The Need for an Improved Fate and Transport Modeling Tool 2-6
2.3 Novel Capabilities of TRIM.FaTE 2-8
3. Overview of TRIM.FaTE Concepts and Terminology 3-1
3.1 Spatial Terminology 3-1
3.2 Time-related Terminology 3-5
3.3 Biotic Compartment Types 3-6
3.4 Links ' 3-8
3.5 Sources 3-9
4. Conceptual Design and Mass Balance Framework for TRIM.FaTE 4-1
4.1 Conceptual Design 4-1
4.2 Governing Mass Balance Equations 4-1
4.3 Phases 4-7
4.4 Fate, Transformation, and Transport Processes 4-8
5. Application of TRIM.FaTE 5-1
5.1 Structure of a TRIM.FaTE Simulation 5-1
5.2 Problem Definition 5-1
5.3 Determining Links/Algorithms 5-6
5.4 Simulation Setup 5-11
5.5 Simulation Implementation and Analysis of Results 5-13
6. Treatment of Uncertainty and Variability 6-1
6.1 Sensitivity and Screening Analyses 6-2
6.2 The Monte Carlo Approach for Uncertainty and Variability Analyses 6-6
6.3 Presentation of Uncertainty Results 6-12
7. References 7-1
NOVEMBER 1999 ix TRIM.FATE TSD VOLUME I (DRAFT)
-------
TABLE OF CONTENTS
Appendices
A. Glossary
B. Integrating External Models or Measured Data into TRIM.FaTE
C. Determining Scale and Spatial Resolution
D. TRIM.FaTE Inputs
E. Prototypes I - IV
F. TRIM.FaTE Computer Framework
NOVEMBER 1999 x TRIM .FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
1. INTRODUCTION
The Office of Air Quality Planning and Standards (OAQPS) of the U.S. Environmental
Protection Agency (EPA, or the Agency) has the responsibility for the hazardous and criteria air
pollutant programs described by sections 112 and 108 of the Clean Air Act (CAA). Several
aspects of these programs require evaluation of the health risks and environmental effects
associated with exposure to these pollutants.1 In response to these risk-related mandates of the
CAA, and the scientific recommendations of the National Academy of Sciences (NAS) (NRC
1994), the Presidential/Congressional Commission on Risk Assessment and Risk Management
(CRARM) (CRARM 1997), as well as EPA guidelines and policies, OAQPS recognized the need
for improved fate and transport, exposure, and risk modeling tools that:
• Have multimedia assessment capabilities;
• Have human health and ecological exposure and risk assessment capabilities;
• Can perform multiple pollutant assessments (e.g., ability to assess mixtures of pollutants,
ability to track chemical transformations);
• Can explicitly address uncertainly and variability;
• Have the ability to easily perform analyses iteratively. moving from the use of simpler
assumptions and scenarios to more detailed assessments; and
• Are readily available and user-friendly, so that they can be used by EPA. as well as by a
variety of Agency stakeholders.
In 1996. OAQPS embarked on a multi-year effort to develop the Total Risk Integrated
Methodology (TRIM), a time series modeling system with multimedia capabilities for assessing
human health and ecological risks from hazardous and criteria air pollutants.
The main purpose of the TRIM Status Report is to summarize the work performed during
the second developmental phase of TRIM. The first phase, which included the conceptualization
of TRIM and implementation of the TRIM conceptual approach through development of a
prototype of the first TRIM module, TRIM.FaTE (U.S. EPA 1998b), was reviewed by EPA's
Science Advisory Board (SAB) in May 1998 (U.S. EPA 1998c). The second developmental
phase has included refining TRIM.FaTE and developing a model evaluation plan, initiating
development of the second module (TRIM.Expo), and conceptualizing the third module
(TRIM.Risk). In addition, progress has been made on developing overarching aspects, such as
the computer framework and an approach to uncertainty and variability. Consistent with the
integral role of peer review in the TRIM development plan, the current Status Report and
1 Hazardous air pollutants (HAPs) include any air pollutant listed under CAA section 112(b); currently,
there are 188 air pollutants designated as HAPs. Criteria air pollutants are air pollutants for which national ambient
air quality standards (NAAQS) have been established under the CAA; at present, the six criteria air pollutants are
particulate matter, ozone, carbon monoxide, nitrogen oxides, sulfur dioxide, and lead.
NOVEMBER 1999 ~\ TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
Technical Support Documents (TSDs) were subjected to review by representatives from the
major program offices at EPA and an EPA Models 20002 review team prior to this SAB
advisory.
This TSD is divided into two volumes. The first volume provides a description of the
terminology, model framework, and functionality of TRIM.FaTE. Specifically, Chapter 2
provides an overview of the development and features of TRIM.FaTE, Chapters 3 and 4 discuss
the TRIM.FaTE terminology and conceptual design, Chapter 5 provides a general description of
how the conceptual design is implemented in TRIM.FaTE, and Chapter 6 explains the treatment
of uncertainty and variability in TRIM.FaTE. Volume II of this document presents detailed
descriptions of the algorithms used in the TRIM.FaTE module.
1.1 GOALS AND OBJECTIVES FOR TRIM
The TRIM modeling system is intended to represent the next generation of human and
environmental exposure and risk models for OAQPS. For example, TRIM is expected to be a
useful tool for performing exposure and/or risk assessments for the following CAA programs:
the Residual Risk Program (CAA section 112[fJ); the Integrated Urban Air Toxics Strategy
(CAA section 112[k]); studies of deposition to water bodies and mercury emissions from utilities
(CAA sections 112[m] and 112[n]); petitions to delist individual HAPs and/or source categories
(CAA sections 112[b][3] and 112[c][9]); review and setting of the national ambient air quality
standards (NAAQS) (CAA section 109); and regulatory impact analyses (RIA).
The goal in developing TRIM is to create a modeling system, and the components of that
system, for use in characterizing human health and ecological exposure and risk in support of
hazardous and criteria air pollutant programs under the CAA. The goal in designing TRIM is to
develop a modeling system that is: (1) scientifically defensible, (2) flexible, and (3) user-friendly.
(1) Characteristics of the TRIM components important to their scientific defensibility include
the following.
• Conservation of pollutant mass. The modeled pollutant(s)' mass will be conserved
within the system being assessed, wherever appropriate and feasible, including during
intermedia transfers. For pollutants where transformation is modeled, the mass of the
core substance (e.g., mercury for methylmercury as well as divalent mercury) within the
modeling simulation will be preserved.
• Following the report of the Agency Task Force on Environmental Regulatory Modeling (U.S. EPA
1994a), the Agency conducted the Models 2000 Conference in December 1997. This conference has led to renewed
emphasis on Agency-wide coordination of model development and the proposal for the implementation of a Council
on Regulatory Environmental Modeling (CREM) to facilitate and promote scientifically-based, defensible
regulatory computer models. The charter for CREM has been reviewed by SAB and is being updated for
implementation by the Agency.
NOVEMBER 1999 1^2TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
• Ability to characterize parameter uncertainty and variability. For critical
parameters, the impacts of parameter uncertainty and variability on model outputs will be
tracked and, where feasible, differentiated.
• Capability for multiple pollutant, multiple media, multiple exposure pathway
assessment. The TRIM modeling system is being designed to facilitate assessment of
risks posed by aggregate exposures to single or multiple chemicals from multiple sources
and via multiple exposure pathways.
(2) To ensure flexibility, the features of TRIM include the following.
• Modular design. Major components of TRIM will be independent and can be used
individually, with outside information or models, or in combination. Only those model
components necessary for evaluating the particular pollutants, pathways, and/or effect
endpoints of interest need be employed in an assessment.
Flexibility in temporal and spatial scale. Exposure and risk assessments will be
possible for a wide range of temporal and spatial scales, including hourly to daily or
yearly time steps, and from local (10 kilometers (km) or less) to greater spatial scales
(depending on the module).
• Ability to assess human and ecological endpoints. Impacts to humans and/or biota can
be assessed.
(3) To ensure that TRIM will be user-friendly for a variety of groups, including EPA, state
and local agencies, and other stakeholders, TRIM will have the following characteristics.
Easily accessible. The TRIM modeling system will be accessible for use with a personal
computer (PC). The system may be available for download from the Internet and
accessible through an Agency model system framework (e.g.. Models-3 (U.S. EPA
1999c)).
• Well-documented. Guidance materials for use of the TRIM modeling system will be
provided through a user's guide, with a focus on the modular aspects of the modeling
system, limitations of the modeling system, and appropriate uses, user responsibilities,
and user options.
• Clear and transparent. The graphical user interface of the TRIM computer framework
will provide transparency and clarity in the functioning of the TRIM modules, and output
from the risk characterization module will document modeling assumptions, limitations,
and uncertainties.
NOVEMBER 1999 l .3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
1.2 TRIM DESIGN
The current TRIM design (Figure 1-1) includes three individual modules. The
Environmental Fate, Transport, and Ecological Exposure module, TRIM.FaTE, accounts for
movement of a chemical through a comprehensive system of discrete compartments (e.g., media,
biota) that represent possible locations of the chemical in the physical and biological
environments of the modeled ecosystem and provides an inventory, over time, of a chemical
throughout the entire system. In addition to providing exposure estimates relevant to ecological
risk assessment, TRIM.FaTE generates media concentrations relevant to human pollutant
exposures that can be used as input to the Exposure-Event module, TRIM.Expo. In
TRIM.Expo, human exposures are evaluated by tracking population groups referred to as
"cohorts" and their inhalation and ingestion through time and space. In the Risk Characterization
module, TRIM.Risk, estimates of human exposures or doses are characterized with regard to
potential risk using the corresponding exposure- or dose-response relationships. The TRIM.Risk
module is also being designed to characterize ecological risks from multimedia exposures. The
output from TRIM.Risk will include documentation of the input data, assumptions in the
analysis, and measures of uncertainty, as well as the results of risk calculations and exposure
analysis.
An overarching feature of the TRIM design is the analysis of uncertainty and variability.
A two-stage approach for providing this feature to the user has been developed. The first stage
includes sensitivity analyses that are useful in identifying critical parameters, while more detailed
uncertainty and variability analyses using Monte Carlo methods (e.g., for refined assessment of
the impact of the critical parameters) are available in the second stage. The uncertainty and
variability feature augments the TRIM capability for performing iterative analyses. For example.
the user may perform assessments varying from simple deterministic screening analyses using
conservative default parameters to refined and complex risk assessments where the impacts of
parameter uncertainty and variability are assessed for critical parameters.
Additionally, the modular design of TRIM allows for flexibility in both its development
and application. Modules can be developed in a phased approach, with refinements being made
as scientific information and tools become available. Furthermore, the user may select any one
or more of these modules for an assessment depending on the user's needs. For example, when
performing a human health risk assessment for an air pollutant for which multimedia distribution
is not significant, TRIM.Expo may be applied using ambient concentration data or the output
from an air quality model external to TRIM; the output from TRIM.Expo may then be used as
input to TRIM.Risk to perform the desired risk analyses. In the case of a multimedia air
pollutant, such as mercury, the user may choose to run all three TRIM modules to assess both
human and ecological risks posed by multipathway exposures from multiple media.
NOVEMBER 1999 1-4 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER I
INTRODUCTION
Figure 1-1
Conceptual Design of TRIM
o
Environmental Fate,
Transport, and
Ecological Exposure
(TRIM.FaTE)
[media concentrations
relevant to human
exposures]
Exposure Event
(TRIM. Expo)
Risk
Characterization
(TRIM.Risk)
[media and biota
concentrations and biota
pollutant intake rates
relevant to ecological
exposures]
Ecciocica Effects
Assessment
^6 s ?• --- -s :"6"?
assurrotions and mpjl aat
=."a exposure
n, - a" and eco'og cal)
:a n'y and var ao y
NOVEMBER 1999
1-5
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
Overview descriptions of the TRIM modules are provided in Sections 1.2.1 through 1.2.3,
the status and plans for development are presented in Section 1.3, and plans for application
appear in Section 1.4. A summary of the previous SAB comments and OAQPS responses is
presented in Chapter 2 of the TRIM Status Report. The approach for handling uncertainty and
variability in TRIM is described in Chapter 3 of the TRIM Status Report. Certain aspects of the
TRIM.FaTE module are addressed in greater detail in Chapters 4 through 7, and additional
details on TRIM.Expo and TRlM.Risk are provided in Chapters 8 and 9, respectively, of the
TRIM Status Report. Chapter 10 of the TRIM Status Report discusses the computer framework
that is being implemented for the TRIM system. In addition, the TRIM.Expo TSD provides more
detail on TRIM.Expo.
1.2.1 DESCRIPTION OF TRIM.FaTE
The first TRIM module to be developed, TRIM.FaTE, is a spatial compartmental mass
balance model that describes the movement and transformation of pollutants over time, through a
user-defined, bounded system that includes both biotic and abiotic components (compartments).
The TRIM.FaTE module predicts pollutant concentrations in multiple environmental media and
in biota and pollutant intakes for biota, all of which provide both temporal and spatial exposure
estimates for ecological receptors (i.e., plants and animals). The output concentrations from
TRIM.FaTE also can be used as inputs to a human exposure model, such as TRIM.Expo, to
estimate human exposures.
Significant features of TRIM.FaTE include: (1) the implementation of a truly coupled
multimedia model; (2) the flexibility to define a variety of scenarios, in terms of the links among
compartments as well as the number and types of compartments, as appropriate for the desired
spatial and temporal scale of assessment; (3) the use of a transparent approach to chemical mass
transfer and transformation based on an algorithm library that allows the user to change how
environmental processes are modeled; (4) an accounting for all of the pollutant as it moves
among the environmental compartments; (5) an embedded procedure to characterize uncertainty
and variability; and (6) the capability to provide exposure estimates for ecological receptors. The
TRIM.FaTE module is the most fully developed of the TRIM modules at this time, and this
development has produced a library of algorithms that account for transfer of chemical mass
throughout an environmental system, a database of the information needed to initialize these
algorithms for a test site, and a working computer model.
1.2.2 DESCRIPTION OF TRIM.Expo
The TRIM.Expo module, similar to most human exposure assessment models, provides
an analysis of the relationships between various chemical concentrations in the environment and
exposure levels of humans. Because multiple sources of environmental contamination can lead
to multiple contaminated media, including air, water, soil, food, and indoor air, it is useful to
focus on the contaminated environmental media with which a human population will come into
contact. These media typically include the envelope of air surrounding an individual, the water
and food ingested by an individual, and the layer of soil and/or water that contacts the surface of
an individual. The magnitude and relative contribution of each exposure pathway must be
NOVEMBER 1999 1-6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER!
INTRODUCTION
considered to assess total exposure to a particular chemical. Currently, the focus of TRIM.Expo
development is on inhalation and ingestion exposure; however, dermal exposure will be added
later.
The exposure analysis process
consists of relating chemical
concentrations in environmental media
(e.g., air, surface soil, root zone soil,
surface water) to chemical concentrations
in the exposure media with which a
human or population has contact (e.g.,
air, tap water, foods, household dusts,
and soils). The initial prototype for
TRIM.Expo will predict exposure by
tracking the movement of a population
cohort through locations where chemical
exposure can occur according to a
•specific activity pattern. In a typical
application, TRIM.FaTE could be used
to provide an inventory of chemical
concentrations across the ecosystem at
selected time intervals (e.g., days, hours).
For chemicals that are not persistent
and/or bioaccumulative, processed air
monitoring data or air dispersion
modeling results can be substituted for
TRIM.FaTE output data. The
TRIM.Expo module would then use
these chemical concentration data, _^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
combined with the activity patterns of the
cohorts, to estimate exposures. The
movements are defined as an exposure-event sequence that can be related to time periods for
which exposure media concentrations are available (e.g., from TRIM.FaTE, ambient data, and/or
dispersion modeling results). Each exposure event places the population cohort in contact with
one or more environmental media within a specified microenvironment (e.g., inside a home,
along a road, inside a vehicle) in an exposure district for a specified time interval. In addition to
Jhe location assignments, the exposure event would provide information relating to the potential
for pollutant uptake, such as respiration rate and quantity of water consumed. The TRIM.Expo
module is intended to contribute to a number of health-related assessments, including risk
assessments and status and trends analyses.
TRIM.Expo KEY TERMS
Cohort - A group of people within a population with
the same demographic variables who are assumed
to have similar exposures.
Activity pattern - A series of discrete events of
varying time intervals describing information about
an individual's lifestyle and routine. The information
contained in an activity pattern typically includes the
locations that the individual visited (usually
described in terms of microenvironments), the
amount of time spent in those locations, and a
description of what the individual was doing in each
location (e.g., sleeping, eating, exercising).
Microenvironment - A defined space in which
human contact with an environmental pollutant takes
place and which can be treated as a well-
characterized, relatively homogeneous location with
respect to pollutant concentrations for a specified
time period.
Exposure district - A geographic location within a
defined physical or political region where there is
potential contact between an organism and a
pollutant and for which environmental media
concentrations have been estimated either through
modeling or measurement.
NOVEMBER 1999
1-7
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
1.2.3 DESCRIPTION OF TRIM.Risk
Risk characterization is the final step in risk assessment and is primarily used to integrate
the information from the other three key steps (i.e., hazard identification, dose-response
assessment, exposure assessment). Within the TRIM framework, TRIM.Risk, the risk
characterization module, will be used to integrate the information on exposure (human or
ecological receptor) with that on dose-response or hazard and for providing quantitative
descriptions of risk and some of the attendant uncertainties. The TRIM.Risk module will provide
decision-makers and the public with information for use in developing, evaluating, and selecting
appropriate air quality standards and risk management strategies. The purpose of TRIM.Risk is
to integrate information from other TRIM modules and to facilitate the preparation of a risk
characterization. The TRIM.Risk module will, therefore, be able to summarize or highlight the
major points from each of the analyses conducted in the other TRIM modules. Where possible,
the TRIM.Risk module will do so in an automated manner. In general, TRIM.Risk will (1)
document'assumptions and input data, (2) conduct risk calculations and data analysis, and (3)
present results and supporting information.
Current and proposed EPA guidance on risk characterization will guide the development
of TRIM.Risk. The TRIM.Risk module will be developed in a phased approach similar to other
TRIM modules. Ideally, TRIM.Risk will provide all of the information required to prepare a full
risk characterization. However, the type and variability of information needed for this purpose
are vast. Therefore, the type of information generated by TRIM.Risk will evolve over time as the
Agency gains experience and has the resources to implement more flexibility. For example, early
versions of TRIM.Risk will be limited to preparing summaries of input data and results, without
supporting text. However, as the Agency gains experience, it may be possible to incorporate
generic language to more fully describe the information required for a full risk characterization.
Many EPA risk assessments will be expected to address or provide descriptions of (1) individual
risk,3 including the central tendency and high-end portions of the risk distribution. (2) population
risk, and (3) risk to important subgroups of the population such as highly exposed or highly
susceptible groups or individuals, if known. Some form of these three types of descriptors will
be developed within TRIM.Risk and presented to support risk characterization. Because people
process information differently, it is appropriate to provide more than one format for presenting
the same information. Therefore, TRIM.Risk will be designed so that the output can be
presented in various ways in an automated manner (e.g., Chart Wizard in Microsoft® Excel),
allowing the user to select a preferred format.
1.3 TRIM DEVELOPMENT
In the development of TRIM, existing models and tools are being relied upon where
possible. Adopting or incorporating existing models or model components into a tool that meets
OAQPS' needs is preferable as it is usually the most cost-effective approach. Consequently,
review of existing models and consideration of other current modeling efforts is an important
3 The phrase individual risk as used here does not refer to a risk estimate developed specifically for a single
member of a population. Rather, it refers to the estimate of risk for a subgroup of a population that is presented as
an estimate of the risk faced by a person rather than by the population as a whole,
NOVEMBER 1999 N8 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 1
INTRODUCTION
part of TRIM development activities. Reviews of relevant models existing at the initiation of
development activities for each module are described in this document and in the TRIM.Expo
TSD. Additionally, OAQPS is closely following several current activities as they relate to
TRIM.
Current Agency model development activities relevant to TRIM development include the
recently published updated guidance on assessing health risks associated with indirect exposure
to combustor emissions (U.S. EPA 1999d). This guidance, previously referred to as the Indirect
-Exposure Methodology (IBM), is now called the Multiple Pathways of Exposure (MPE) method.
In addition, the multimedia model, FRAMES-HWIR, has recently been developed by the Agency
to support a specific risk assessment need regarding hazardous chemicals released from land-
based waste management units. The FRAMES-HWIR model has been developed as part of a
focused fast-track (two-year) effort to support a risk-based regulation regarding disposal of
hazardous waste (HWIR99).4 Another model of interest for multimedia pollutants is the
Stochastic Human Exposure and Dose Simulation (SHEDS) model (e.g., Ozkaynak et al. 1999).
The OAQPS will be carefully considering the various aspects of MPE, FRAMES-HWIR. and
SHEDS with regard to OAQPS needs, as well as compatibility with or future improvements or
evaluations of TRIM. As TRIM is intended to be a dynamic method, developmental activities
will consider and respond as appropriate to newly available methods and scientific information.
A current major Agency research project involves the design and development of a
flexible software system to simplify the development and use of air quality models and other
environmental decision support tools. This system, called Models-3, is designed for applications
ranging from regulator)' and policy analysis to understanding the complex interactions of
atmospheric chemistry and physics (U.S. EPA 1999c). The June 1999 release of Models-3
contains a Community Multi-Scale Air Quality (CMAQ) modeling system for urban- to regional-
scale air quality simulation of tropospheric ozone, acid deposition, visibility, and fine particles.
The long-term goal is to extend the system to handle integrated cross-media assessments and
serve as a platform for community development of complex environmental models. In
recognition of the availability of Models-3 over the longer term, OAQPS has designed and is
developing the TRIM computer framework to be compatible with the Models-3 system.
1.3.1 INITIAL DEVELOPMENT ACTIVITIES
The first phase of TRIM development included the conceptualization of TRIM and the
implementation of the TRIM conceptual approach through the development of a prototype of the
first TRIM module, TRIM.FaTE (U.S. EPA 1998b). The progress on TRIM.FaTE included the
development of (1) a conceptual design for the module; (2) a library of algorithms that account
for chemical mass transfer throughout the ecosystem; (3) a database to initialize the algorithms
for a test site; and (4) a working prototype in spreadsheet format.
Consistent with Agency peer review policy (U.S. EPA 1998d) and the 1994 Agency Task
Force on Environmental Regulatory Modeling (U.S. EPA 1994a), internal and external peer
4 The FRAMES-HWIR documentation is scheduled for public release in fall 1999.
NOVEMBER 1999 N9 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 1
INTRODUCTION
review are an integral part of the TRIM development plan. Following the first phase of TRIM
development, OAQPS submitted TRIM to SAB under their advisory method of review (U.S.
EPA 1998c). In May 1998 in Washington, DC, the Environmental Models Subcommittee
(Subcommittee) of the Executive Committee of SAB reviewed the TRIM project. The SAB
Subcommittee was charged with assessing the overall conceptual approach of TRIM and the
specific approach of TRIM.FaTE.
The SAB Subcommittee reported that the development of TRIM and the TRIM.FaTE
module was conceptually sound and scientifically based (U.S. EPA 1998c). The SAB
Subcommittee provided specific recommendations related to six specific charge questions. The
SAB recommendations are detailed in Chapter 2 of the TRIM Status Report along with brief
responses, and changes to TRIM.FaTE based in part on the SAB recommendations are
highlighted in Chapter 4 of the TRIM Status Report.
1.3.2 RECENT ACTIVITIES
During the most recent developmental phase of TRIM, progress has been made in many
areas, including a change to the overall modular design of TRIM. As shown in Figure 1-1, the
TRIM design now includes three modules: TRIM.FaTE, TRIM.Expo, and TRIM.Risk. The
design presented to SAB in May 1998 included three other modules (Pollutant Uptake,
Biokinetics, and Dose/Response). In recognition of the flexibility of the TRIM design, which
provides an ability to rely on a variety of input data and outside models, OAQPS decided not to
include the development of these modules in the TRIM design at this time.
In consideration of SAB comments, TRIM.FaTE was refined, including the development
of new and updated capabilities, as well as the development and limited testing of
methodologies for model set-up, uncertainty and variability analysis, and evaluation. In addition,
OAQPS developed a conceptual plan for TRIM.Expo, initiated work on a prototype of
TRIM.Expo (initially focusing on inhalation), and developed a conceptual design for TRIM.Risk.
Furthermore, the overall computer framework for TRIM was designed and implemented in a PC-
based platform, and substantial progress was made in installing TRIM.FaTE into this framework.
Changes and additions to TRIM.FaTE are discussed in more detail in Chapter 4 of the TRIM
Status Report. The development of TRIM.Expo is discussed in Chapter 8, and the conceptual
plan for TRIM.Risk is described in Chapter 9 of the TRIM Status Report. In addition, the
TRIM.Expo TSD provides more details on TRIM.Expo.
The current TRIM documentation has gone through internal Agency peer review, which
involved reviewers across the Agency, including the major program offices, the Office or
Research and Development, and staff involved in the Agency's Models 2000 efforts. The current
SAB advisory will be the second on TRIM development activities.
NOVEMBER 1999 1-10 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
1.3.3 FUTURE ACTIVITIES
Following the 1999 SAB advisory, improvements will be made to the uncertainty and
variability approach, TRIM.Expo prototype, and TRIM.Risk conceptual plan. These revisions
are scheduled to be completed in 2000. As needed, refinements will be made to the TRIM.FaTE
evaluation plan, and completion of the bulk of those activities are also scheduled for 2000. The
Agency has planned for a substantial amount of progress on each of the TRIM modules for 2000
and 2001, as described below.
• TRIM.FaTE. Future work on TRIM.FaTE will include model evaluation activities and
additional development of the module to accommodate additional chemicals. The
TRIM.FaTE module is expected to be available for limited external use late in 2000 and
to be publicly released in 2001.
TRIM.Expo. Future work on TRIM.Expo in 2000 will include the further development
of ingestion algorithms, incorporation of EPA's Air Pollutant Exposure Model (APEX)
coding into the TRIM platform followed by adjustments to APEX to include ingestion
algorithms, a test case of the inhalation pathway, and a test case of inhalation and
ingestion pathways. Over the longer term, addition of the dermal pathway to the module
will be initiated.
TRIM.Risk. Development of TRIM.Risk will begin after SAB comments are received
on the conceptual design. Module development will include identification of data needs
and formatting of data outputs. Programming for a TRIM.Risk prototype is expected to
be completed in 2000.
• TRIM computer framework. Further development of the TRIM computer framework,
including incorporation of the TRIM.Expo (inhalation) module, will take place during
2000. Features to be refined during this time frame include limited geographic
information system (GIS) or mapping capabilities. Additionally, long-range
comprehensive GIS planning will occur. Development of user guidance materials is
planned for 2000 (see text box).
In addition to consulting with Agency scientists during future TRIM development (/. e.,
peer involvement), in late 2000 or early 2001, OAQPS will seek both internal and external peer
review of new aspects following the next phase of TRIM development. In addition to the SAB,
which provides the Agency with reviews, advisories, and consultations, other external peer
review mechanisms consistent with Agency policy (U.S. EPA 1998d) include the use of a group
of independent experts from outside the Agency (e.g., a letter review by outside scientists), an ad
hoc panel of independent experts, and peer review workshops. The OAQPS intends to seek the
peer review mechanism appropriate to the importance, nature, and complexity of the material for
review.
NOVEMBER 1999 1-11 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
USER GUIDANCE
Development of the TRIM user's guide is scheduled to begin in 2000, along with a plan for
training activities. The OAQPS recognizes the importance of developing detailed user guidance that
will assist users in defining, for a particular modeling application, the spatial and temporal resolution,
compartments and linkages, and parameters and initial conditions. For example, the TRIM.FaTE
guidance will likely emphasize the value of performing several different preliminary simulations in
verifying the adequacy of the parcel and compartment specifications for the desired application.
Similarly, detailed user guidance will be developed for TRIM.Expo to assist users in defining
cohorts, study areas, exposure districts, and microenvironments, as well as various parameters and
exposure factors.
It also will be important for the guidance to note the responsibility of the user in defining the
simulation as appropriate to the application. For example, in TRIM.FaTE, default values will likely
be made available with the model for a variety of parameters ranging from physiological
characteristics of various biota to physical characteristics of abiotic media; the user will need to
consider appropriateness of these values or others (e.g., site-specific data) for their application.
While the TRIM modules are intended to provide valuable tools for risk assessment, and their
documentation and guidance will identify, as feasible, uncertainties and limitations associated with
their application, the guidance will emphasize that their appropriate use and the characterization of
uncertainties and limitations surrounding the results are the responsibility of the user.
1.4 PHASING TRIM INTO OAQPS' SET OF MODELING TOOLS
As mentioned earlier, TRIM is intended to support assessment activities for both the
criteria and hazardous air pollutant programs of OAQPS. As a result of the greater level of effort
expended by the Agency on assessment activities for criteria air pollutants, these activities are
generally more widely known. To improve the public understanding of the hazardous air
pollutant (or air toxics) program, the Agency published an overview of the air toxics program in
July 1999 (U.S. EPA 1999e). Air toxics assessment activities (National Air Toxics Assessment.
or NATA) are described as one of the program's key components.5 The NATA includes both
national- and local-scale activities. The TRIM system is intended to provide tools in support of
local-scale assessment activities, including multimedia analyses.
5 Within the air toxics program, these activities are intended to help EPA identify areas of concern (e.g.,
pollutants, locations, or sources), characterize risks, and track progress toward meeting the Agency's overall air
toxics program goals, as well as the risk-based goals of the various activities and initiatives within the program, such
as residual risk assessments and the Integrated Urban Air Toxics Strategy. More specifically, NATA activities
include expansion of air toxics monitoring, improvements and periodic updates to emissions inventories, national-
and local-scale air quality modeling, multimedia and exposure modeling (including modeling that considers
stationary and mobile sources), continued research on health effects of and exposures to both ambient and indoor
air, and use and improvement of exposure and assessment tools. These activities are intended to provide the
Agency with improved characterizations of air toxics risk and of risk reductions resulting from emissions control
standards and initiatives for both stationary and mobile source programs.
NOVEMBER 1999 M2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 1
INTRODUCTION
needs for TRIM comes in the Residual Risk
Program, in which there are statutory
deadlines within the next two to nine years
for risk-based emissions standards decisions.
EXAMPLES OF TRIM APPLICATIONS
A j -i. j • *i_ n -j i D- i D tt one or several local sources could be
As described m the Residual Risk Report to performed using all three modules in the
Congress (U.S. EPA 1999f), TRIM is
intended to improve upon the Agency's
ability to perform multipathway human health
risk assessments and ecological risk associated with air emissions of a criteria
0 air pollutant (e.g., ozone) or one or several
assessments for HAPs with the potential for
multimedia environmental distribution.
Another important upcoming use for TRIM is
in exposure assessment in support of the
review of the ozone NAAQS. The TRIM.Risk.
A human health or ecological assessment
of multimedia, multipathway risks
associated with mercury emissions from
TRIM system.
An assessment of human health risks
volatile HAPs in a metropolitan area could
be developed using an external air model
or ambient concentration data from fixed-
site monitors coupled with TRIM.Expo and
TRIM.Expo and TRIM.Risk modules ^—^-^_
augmented with external air quality
monitoring data and models are intended to support this type of criteria pollutant assessment as
well as risk assessments for non-multimedia HAPs.
Consistent with the phased plan of TRIM development, the application of TRIM will also
be initiated in a phased approach. With the further development of the TRIM modules in 2000
and 2001. EPA will begin to use the modules to contribute to or support CAA exposure and risk
assessments. These initial applications also will contribute to model evaluation. The earliest
TRIM activities are expected to include the use of TRIM.FaTE side-by-side (at a comparable
level of detail) with the existing multimedia methodology6 in risk assessments of certain
multimedia HAPs (e.g., mercury) under the Residual Risk Program. As TRIM.Expo is
developed to accommodate inhalation modeling of HAPs and after it has undergone testing,
OAQPS plans to initially run it side-by-side (at a comparable level of detail) with EPA's existing
inhalation exposure model, HEM (Human Exposure Model (U.S. EPA 1986)). When
TRIM.Risk has been completed, it will be used, as appropriate, in both criteria and hazardous air
pollutant risk assessments.
In later years, OAQPS intends to use TRIM and the TRIM modules in a variety of
activities including (1) residual risk assessments using TRIM.FaTE, TRIM.Expo, and
TRIM.Risk, in combinations appropriate to the environmental distribution characteristics of the
HAPs being assessed; (2) urban scale assessments on case study cities as part of the Integrated
Urban Air Toxics Strategy; and (3) exposure and risk assessments of criteria air pollutants (e.g.,
ozone, carbon monoxide) in support of NAAQS reviews.
6 In support of the Mercury Report to Congress (U.S. EPA 1997a) and the Study of Hazardous Air
Pollutant Emissions from Electric Utility Steam Generating Units - Final Report to Congress (U.S. EPA 1998e),
the Agency relied upon the Indirect Exposure Methodology, which has recently been updated and is now termed the
Multiple Pathways of Exposure methodology (U.S. EPA 1999d). This methodology is being used in initial
assessment activities for the Residual Risk Program (U.S. EPA 1999f).
NOVEMBER 1999 Tl3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
2. INTRODUCTION TO TRIM.FaTE
Implementation of the TRIM system began with development of the TRIM
Environmental Fate, Transport, and Ecological Exposure module (TRIM.FaTE), a flexible
multimedia fate and transport model designed to estimate pollutant concentrations in various
environmental compartments (i.e., media and organisms). These media and biota concentrations,
as well as estimates of pollutant intake by organisms, provide measures of ecological exposure in
various biota on a temporal and spatial scale. The media and biota concentrations also provide
temporally and spatially varying inputs for a human exposure model such as TRIM.Expo, which
can model population cohorts through space and time.
Prior to and during the development of TRIM.FaTE, EPA has reviewed the features of
existing multimedia models and approaches in order to build on, rather than duplicate, previous
efforts. In these reviews, the Agency focused on how existing models address the following
characteristics desired for TRIM.FaTE:
• Ability to address varying time steps (of one hour or greater) and provide sufficient
spatial detail at varying scales (site-specific to urban scale);
• Conservation of pollutant mass within the system being assessed;
• Transparency, as needed for use in a regulator)' context; and
• Performance as a truly coupled multimedia model rather than a set of linked single
medium models.
As a result of the Agency's reviews of other models (Section 2.1), OAQPS concluded (as
described in Section 2.2) that in order to meet the Office's needs for assessing human health and
ecological risks of exposure to criteria and hazardous air pollutants, it is necessary to develop a
truly coupled multimedia modeling framework. In developing TRIM.FaTE, the Agency has
incorporated several features that improve upon the capabilities of existing models. These key
features are summarized in Section 2.3.
2.1 REVIEW OF EXISTING FATE AND TRANSPORT MODELS
In 1996, EPA undertook a review of existing models and approaches as an initial step in
the TRIM development effort. The resulting report, entitled Evaluation of Existing Approaches
for Assessing Non-Inhalation Exposure and Risk with Recommendations for Implementing TRIM
(Mosier et al. 1996), examined several multimedia models. Two additional EPA studies
conducted in 1997 (IT 1997a, IT 1997b) have updated the 1996 study.
The initial literature searches identified several models/approaches for multimedia,
multipathway modeling, including EPA's Indirect Exposure Methodology (IEM), the California
Department of Toxic Substance Control's Multimedia Risk Computerized Model (CalTOX), the
NOVEMBER 1999 2-1 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
Dutch model SimpleBOX, the Integrated Spatial Multimedia Compartmental Model (ISMCM),
and the Multimedia Environmental Pollutant Assessment System (MEPAS).
Efforts to assess human exposure from multiple media date back to the 1950s, when the
need to assess human exposure to global radioactive fallout led rapidly to a framework that
included transport through and transfers among air, soil, surface water, vegetation, and food
chains (Wicker and Kirchner 1987). Efforts to apply such a framework to non-radioactive
organic and inorganic toxic chemicals have been more recent and have not as yet achieved the
level of sophistication that exists in the radioecology field. In response to the need for
multimedia models in exposure assessment, a number of multimedia transport and
transformation models have been recently developed.
Thibodeaux (1979, 1996) proposed the term "chemodynamics" to describe a set of
integrated methods for assessing the cross-media transfers of organic chemicals. The first widely
used multimedia compartment modeling approaches for organic chemicals were the "fugacity"
models proposed by Mackay (1979, 1991) and Mackay and Paterson (1981, 1982). Cohen and
his co-workers applied the concept of multimedia compartment modeling as a screening tool
with the Multimedia Compartment Model (MCM) (Cohen and Ryan 1985), followed by the
Spatial MCM (SMCM) (Cohen et al. 1990), and more recently with the Integrated SMCM
(ISMCM). which allows for non-uniformity in some compartments (van de Water 1995).
Another multimedia screening model, called GEOTOX (McKone and Layton 1986). was one of
the earliest multimedia models to explicitly address human exposure. The CalTOX program
(McKone 1993a, McKone 1993b, McKone 1993c) has been developed for the California EPA as
a set of spreadsheet models and spreadsheet data sets to assist in assessing human exposures to
toxic substance releases in multiple media. More recently, SimpleBOX (van de Meent 1993.
Brandes et al. 1997) has been developed for the National Institute of Public Health and the
Environment in the Netherlands to evaluate the environmental fate of chemicals.
A brief summary of key multimedia models evaluated for applicability to the TRIM.FaTE
effort follows. Other models reviewed are documented in the background reports referenced in
the first paragraph of this section.
• Indirect Exposure Methodology (IEM). The IEM consists of a set of multimedia fate
and exposure algorithms developed by EPA's Office of Research and Development that
is a significant current Agency methodology for multimedia, multipathway modeling for
pollutants for which indirect (i.e., non-inhalation) impacts may be important (i.e., organic
and inorganic pollutants that tend to be long-lived, bioaccumulating, non- (or at most
semi-) volatile, and more associated with soil and sediment than with water).
NOVEMBER 1999 2-2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
_ INTRODUCTION TO TRIM.FATE
An interim document describing this methodology was published in 1990 (U.S. EPA
1990), and a major addendum was issued in 1993 (U.S. EPA 1993).' The IEM has
undergone extensive scientific review, including review by SAB, which has been useful
in focusing efforts in the development of TRIM. The SAB identified several limitations
of IEM that are pertinent to its application to the design quals for TRIM (U.S. EPA
1994b). Concurrently with IEM development, EPA has also developed and applied a
closely related set of multimedia models in a variety of dioxin assessments (U.S. EPA
1994c; updated document expected in 2000).
Descriptions of fate and transport algorithms, exposure pathways, receptor scenarios, and
dose algorithms are presented in the IEM documentation. The IEM includes procedures
for estimating the indirect human exposures and health risks that can result from the
transfer of emitted air pollutants to soil, vegetation, and water bodies. The methodology
addresses exposures via inhalation, food, water, and soil ingestion, and dermal contact.
There appear to be several limitations in the IEM approach relative to the TRIM.FaTE
design criteria and OAQPS' needs. For example, IEM, as currently implemented, can be
applied only to chemicals that are emitted to the air. This limits its ability to provide
assessment of media concentrations resulting from air emissions when other pollutant
sources might have a significant impact on the results. However, IEM is an evolving and
emerging methodology that moves EPA beyond analyzing the potential effects associated
with only one medium (air) and exposure pathway (inhalation) to the consideration of
multiple media and exposure pathways. It is crucial in the development of TRIM that a
sense of continuity be maintained between IEM and proposed TRIM methodologies.
The IEM was designed to predict long-term, steady-state impacts from continuous
sources, not short-term, time series estimates. It consists of a one-way process through a
series of linked models or algorithms and requires annual average air concentrations and
wet and dry deposition values from air dispersion modeling external to IEM. As a result,
IEM cannot provide detailed time-series estimation (e.g., for time steps less than one
year) of media concentrations and concomitant exposure, cannot maintain full mass
balance, and, because it is not a truly coupled multimedia model, does not have the ability
to model "feedback" loops between media or secondary emissions (e.g., re-emission of
deposited pollutants). Furthermore, IEM does not provide for the flexibility OAQPS
needs in site-specific applications or in estimating population exposures. Significant site-
specific adjustment must be made to allow for spatially tracking differences in
1 Since OAQPS' initial review and consideration of IEM in 1996, the methodology and its documentation
have undergone several important changes. A draft revised document addressing SAB and public comments on the
1993 Addendum was released for review in 1998 (U.S. EPA 1998f). The IEM2M was derived from IEM and
applied by OAQPS to estimate exposures to mercury for the Mercury Study Report to Congress (U.S. EPA 1997).
The Agency's Office of Solid Waste and Emergency Response (OSWER) has adapted IEM and compiled detailed
information on many of lEM's input parameters and algorithms in the Human Health Risk Assessment Protocol for
Hazardous Waste Combustion Facilities (U.S. EPA 1998g), which has been applied to assess exposures and risks
for many hazardous waste combustion facilities. The most up-to-date version of the general IEM methodology is
scheduled to be published in late 1999 (U.S. EPA 1999d.) The updated documentation no longer refers to the
methodology as IEM; it is now referred to as the Multiple Pathways of Exposure (MPE) methodology.
NOVEMBER 1999 2O TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
concentrations and exposures. Much of the focus of IEM is on evaluating specific
receptor scenarios (e.g., recreational or subsistence fisher) that may be indicative of high-
end or average exposures, but the model is not designed to model the range of exposures
within a population (e.g., IEM cannot estimate population exposure distributions). More
recent advances (Rice et al. 1997) have addressed some of these issues to some degree,
but have not been fully implemented.
Therefore, while IEM may meet its own design criteria quite well (e.g., can adequately
estimate long-term average exposure media concentrations in the vicinity of an air source
for contaminants for which indirect impacts may be important), it does not fully meet the
needs of OAQPS for the reasons noted above.
• California Department of Toxic Substance Control's Multimedia Risk
Computerized Model (CalTOX). First issued in 1993 (McKone 1993a, McKone 1993b,
McKone 1993c) and updated in 1995, with continual enhancements underway, CalTOX
was developed as a spreadsheet model for California's Department of Toxic Substance
Control to assist in human health risk assessments that address contaminated soils and the
contamination of adjacent air, surface water, sediment, and ground water. CalTOX
consists of two component models: a multimedia transport and transformation (i.e., fate
and transport) model, which is based on both conservation of mass and chemical
equilibrium; and a multipathway human exposure model that includes ingestion,
inhalation, and dermal uptake exposure routes. CalTOX is a fully mass balancing model
that includes add-ins to quantify uncertainty and variability.
The multimedia transport and transformation model is a dynamic model that can be used
to assess time-varying concentrations of contaminants introduced initially to soil layers or
for contaminants released continuously to air, soil, or water. The CalTOX multimedia
model is a seven-compartment regional and dynamic multimedia fugacity model. The
seven compartments are: (1) air, (2) surface soil, (3) plants, (4) root-zone soil, (5) the
vadose-zone soil below the root zone, (6) surface water, and (7) sediment. The air,
surface water, surface soil, plants, and sediment compartments are assumed to be in
quasi-steady-state with the root-zone soil, and vadose-zone soil compartments.
Contaminant inventories in the root-zone soil and vadose-soil zone are treated as time-
varying state variables. Contaminant concentrations in ground water are based on the
leachate from the vadose-zone soil.
The multipathway exposure model encompasses 23 exposure pathways to estimate
average daily doses within a human population in the vicinity of a hazardous substances
release site. The exposure assessment process consists of relating contaminant
concentrations in the multimedia model compartments to contaminant concentrations in
the media with which a human population has contact (e.g., personal air, tap water, foods,
household dusts/soils). The explicit treatment of differentiating environmental media
pollutant concentration and the pollutant concentration to which humans are exposed
favorably distinguishes CalTOX from many other exposure models. In addition, all
parameter values used as inputs to CalTOX are distributions, described in terms of mean
values and a coefficient of variation, rather than point estimates (central tendency or
NOVEMBER 1999 24 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
plausible upper values) such as most other models employ. This stochastic approach
allows both sensitivity and uncertainty to be directly incorporated into the model
operation.
As indicated in the literature review reports, the CalTOX model appears to be the most
promising existing model for application to the TRIM effort. Several of the mathematical
concepts and derivations used by the developers of CalTOX can be directly applied to
meet the TRIM goals. However, CalTOX does have several limitations that prevent it
from being entirely imported into the TRIM approach. These limitations result from the
need to go beyond the intended applications for CalTOX; for example, for landscapes in
which there is a large ratio of land area to surface water area, for a limited range of
chemicals (e.g., non-ionic organic chemicals in a liquid or gaseous state). As a result, the
model does not provide adequate flexibility in environmental settings and chemical
classes (e.g., volatile metals such as mercury) to be suitable for OAQPS' needs. The
most significant of these limitations, in terms of application to TRIM, is the fact that the
CalTOX model, as it currently exists, does not allow spatial tracking of a pollutant as is
required in the TRIM approach.
• SimpleBOX. SimpleBOX is a steady-state, non-equilibrium partitioning, mass balance
model (van de Meent 1993, Brandes et al. 1997). It consists of eight compartments, three
of which are soils of differing use and properties. It also produces quasi-dynamic
(non-steady-state) output by using an external numerical integrator. The model was
developed as a regional scale model for the Netherlands, so its default characteristics
represent the Netherlands. SimpleBOX uses the classical concentration concept to
compute the mass balance (van de Meent 1993). While its goals are comparable to TRIM
to the extent that it simulates regional systems, its coarse spatial and temporal complexity
and lack of exposure media concentration estimates cause it to fall short of TRIM'S goals.
• Integrated Spatial Multimedia Compartmental Model (ISMCM). ISMCM has been
under development at the School of Engineering and Applied Science at University of
California Los Angeles for approximately 15 years (van de Water 1995). A newer
version of ISMCM, called MEND-TOX, is currently under evaluation by the EPA Office
of Research and Development's National Exposure Research Laboratory.
The ISMCM considers all media, biological and non-biological, in one integrated system
and includes both spatial and compartmental modules to account for complex transport of
pollutants through an ecosystem. Assuming mass conservation, ISMCM is able to predict
transport based on a sound mechanistic description of environmental processes, including
estimation of intermedia transfer factors.
One of the limiting factors of the ISMCM system, with regard to use in the TRIM system,
is that it is not structured to incorporate uncertainty and variability directly into the model
operation. Another of the limitations of the ISMCM model within the context of the
goals for TRIM (van de Water 1995) is the fact that the links and compartments (spatial
configuration) of this model are predetermined. Thus, ISMCM was apparently not
designed from the start with the flexibility to meet the goals of TRIM.
NOVEMBER 1999 2!TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
Multimedia Environmental Pollutant Assessment System (MEPAS). MEPAS was
developed at the U.S. Department of Energy's (DOE) Pacific Northwest Laboratory to
assess risks from mixed (i.e., chemical and radioactive) wastes at DOE facilities. This
model consists of single-media transport models linked together under appropriate
boundary conditions and considers four primary types of pollutant pathways (ground
water, overland, surface water, and atmospheric) in evaluating human exposure. MEPAS
also contains an exposure and risk module. The model's ability to estimate multipathway
risks for chemicals and radionuclides makes it unique. The nature of its algorithms
makes it a screening tool, rather than a detailed assessment tool. The model is updated
periodically and the latest version of MEPAS (Version 3.1) contains an uncertainty and
variability analysis module (Buck et al. 1995).
The mathematical design of this model does not include mass balance and could not be
integrated into TRIM. As with IEM, MEPAS represents a "linked" model system that
utilizes a one-way process through a sequence of models that individually describe a
specific environmental process or medium. These types of models are not mass
conservative and do not allow for temporal tracking of the pollutants and concomitant
exposure necessary to meet the needs of TRIM.
2.2 THE NEED FOR AN IMPROVED FATE AND TRANSPORT
MODELING TOOL
Current OAQPS fate and transport models for hazardous and criteria air pollutants do not
address multimedia exposures, and current OAQPS HAP models do not adequately estimate
temporal and spatial patterns of exposures. Adopting or incorporating existing models into a tool
that meets OAQPS' needs represents the most cost-effective approach to developing the tools
needed to support regulatory decision-making related to hazardous and criteria air pollutants.
Based on the OAQPS review of existing multimedia models and modeling systems (described in
Section 2.1), there is no single fate and transport model that meets the needs of OAQPS (outlined
in Chapter 1) and that can be adopted as part of TRIM. Most models are limited in the types of
media and environmental processes addressed. Simply, no single model can address the broad
range of pollutants and environmental fate and transport processes anticipated to be encountered
by OAQPS in evaluating risks from hazardous and criteria air pollutants. In addition, it is
unlikely that one individual model could be developed to address this wide range of concerns.
Therefore, the TRIM framework emphasizes a modular design. The lack of a flexible
multimedia fate and transport model was identified as a major limitation and was the focus of the
first phase implementation efforts for TRIM.
Existing multimedia models can be divided into two basic categories: "linked" single
medium model systems and mass-conserving models. Mass-conserving models can be further
classified as fugacity-based, concentration-based, or inventory-based models depending on the
choice of state variable (i.e., fugacity, concentration, or inventory). The linked single medium
and mass-conserving models each have their own strengths and limitations.
"Linked" single medium modeling systems are composed of several independent single
medium models. The linked system typically calculates fate and transport by running a single
NOVEMBER 1999 2^6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER!
INTRODUCTION TO TRIM.FATE
medium model (e.g., an atmospheric model) and using the output from each time step as the
input for the corresponding time step of another single medium model (e.g., a soil or surface
water model). There are several highly sophisticated single media models to choose from when
constructing a linked system. However, the linked design does not assure conservation of mass
because the dynamic feedback loops and secondary pollutant transfers are not treated in a fully
coupled manner. In addition, the level of detail provided by the linked model system is not easily
adjusted to suit the needs of different modeling objectives.
Mass-conserving multimedia models were developed to fully account for the distribution
of mass within a compartmentalized system. The fugacity type multimedia models were
introduced by Mackay (1979, 1991) as screening tools to assess the relative distribution of
chemicals in air, water, sediment, and soil. The fugacity concept provides a convenient method
for quantifying the multimedia fate of chemicals (Cowen et al. 1995). However, models that use
fugacity as the state variable are limited in application only to organic chemicals. Concentration-
based models like Simple Box and inventory-based models like CalTOX can technically handle
inorganic chemicals, but temporal and spatial resolution is limited by the rigid
compartmentalized structure or boxes used to represent the environmental media. Spatial
compartmental models (e.g., ISMCM) represent the closest current models to an integrated
multimedia system. However, as previously described, ISMCM does not meet the TRIM design
criterion for a flexible architecture.
In general, none of the multimedia models existing at the time TRIM development began
was sufficiently coupled to account for inherent feedback loops or secondary emissions or
releases to specific media, or was able to provide the temporal and spatial resolution critical in
estimating exposures. While the degree to which results would differ between existing models
and a truly coupled multimedia model is unknown, non-coupled multimedia models have been
generally considered to lack scientific credibility. Therefore. OAQPS determined it was
necessary to undertake efforts to develop a truly coupled multimedia model.
Another multimedia model, FRAMES-HWIR, has recently been developed by the
Agency to support a specific risk assessment need regarding hazardous chemicals released from
land-based waste management units. FRAMES-HWIR is a framework system which includes,
along with several site-specific databases and processors, a multimedia, multipathway, and
multireceptor simulation processor (MMSP) for fate and transport and exposure modeling.
MMSP is itself made up of 17 individual modules (e.g., air, watershed, human exposure).
FRAMES-HWIR has been developed as part of a focused fast-track (two-year) effort to support a
risk based regulation for disposal of hazardous waste (HWIR99). The development plan
received peer review in late 1998, and the individual modules have been submitted for peer
review upon completion, with the last of those reviews in progress. The FRAMES-HWIR
documentation is scheduled for public release and accompanying public review in Fall 1999.
OAQPS will be carefully considering the various aspects of FRAMES-HWIR and MMSP - as
well as other evolving Agency multimedia modeling methods, including the MPE (formerly
IEM) methodology discussed in Section 2.1 - with regard to OAQPS' needs, as well as
compatibility with and role in future improvements or evaluations of TRIM.
NOVEMBER 1999 2-7 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
2.3 NOVEL CAPABILITIES OF TRIM.FaTE
As mentioned earlier, several key characteristics have been identified as essential to the
design of TRIM.FaTE:
• Ability to address varying time steps (of one hour or greater) and provide sufficient
spatial detail at varying scales (site-specific to urban scale);
• Conservation of pollutant mass within the system being assessed;
• Transparency, as needed for use in a regulatory context; and
• Performance as a truly coupled multimedia model rather than a set of linked single
medium models.
To accommodate these characteristics, the Agency developed a new model framework
that expanded upon the mass balance and compartmental framework used by CalTOX and the
system of equations used in ChemCan2 and SimpleBOX to produce a modeling system that
incorporates a flexible level of spatial and temporal resolution while maintaining a complete
multimedia mass balance. Development of the TRIM.FaTE framework required the TRIM team
to design several features not available in existing multimedia models. These key features.
which are described below, include:
• Implementation as a truly coupled multimedia model framework:
• The adaptability to match a simulation to the spatial and temporal scales needed for a
broad range of pollutants and geographical areas;
• The use of a unified approach to mass transfer, based on an algorithm library that allows
the user to change mass transfer relationships between compartments without creating a
new modeling scenario;
• An accounting of the pollutant mass distributed within, as well as entering and leaving,
the environmental system;
• An embedded procedure to characterize uncertainty and variability; and
• The capability to be used as an exposure model for ecological receptors.
2 ChemCan is a steady-state fugacity balance model, designed for Health Canada, intended to assist in
human exposure assessment. The model estimates average concentrations in air, fresh surface water, fish,
sediments, soils, vegetation, and marine near-shore waters.
NOVEMBER 1999 ?8 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER!
INTRODUCTION TO TRIM.FATE
2.3.1 TRULY COUPLED MULTIMEDIA FRAMEWORK
One of the significant distinguishing features of the TRIM.FaTE methodology is the
attention paid to possible interactions between media. The transfer of chemical mass between
compartments is not restricted to a one-way process, which is common for many "linked"
multimedia models. Instead, TRIM.FaTE allows the user to simulate the movement of a
chemical in any direction for which transfer can occur. Without this functionality, a multimedia
model can never be truly mass conservative and cannot adequately address feedback loops and
secondary pollutant movement (e.g., revolatilization and transport). The lack of a full mass
balance and the functionality to account for feedback loops and secondary pollutant movement
are generally considered significant sources of uncertainty in the application of "linked" models.
The use of a truly coupled multimedia framework for TRIM.FaTE can reduce this important area
of uncertainty.
2.3.2 SCALABLE COMPLEXITY
The current TRIM.FaTE methodology allows the user a great deal of flexibility in the
design of any particular model application, both spatially and temporally. The functionality to
account for varying degrees of temporal resolution is common among multimedia models.
Conversely, the spatial flexibility provided in TRIM.FaTE is unique among multimedia models
because it allows the user to vary the resolution significantly over the modeled region. For
example, initially the user may define only a few homogeneous regions for the model area. After
inspecting the results of the initial analysis, the user could subdivide those regions where more
resolution is desired. This prevents the user from including more resolution than is necessary for
a particular application, resulting in more efficiency in modeling. Although some applications of
TRIM.FaTE may resemble a simple fugacity-based compartmental model, it also can be scaled to
simulate time-series and spatial resolutions that current fugacity-type models could not handle.
2.3.3 FLEXIBLE ALGORITHM LIBRARY
The manner in which the chemical mass transfer algorithms have been implemented in
TRIM.FaTE is unique among multimedia models. Rather than storing the equations only in
computer code, which is not readable by the user at run time, the equations are stored in a form
that allows the user to inspect the equations, variables in the equations, and values for the
variables for almost any calculated term at run time. It is possible for the user to trace the
calculation of almost any of the chemical mass transfers, which can be useful when trying to
explore an unexpected result. For most models, the user cannot be sure how faithfully the
equations documented have been implemented, or how synchronized the documentation is with
the code. With the TRIM.FaTE methodology, these problems can be substantially alleviated.
Another advantage in the algorithm implementation is the potential to choose from a set
of algorithms for each of the types of chemical mass transfers. The primary benefit would be in
performing sensitivity analyses when there are uncertainties regarding the model approach for
some transport or transformation processes. If there were several different algorithms available
NOVEMBER 1999 2-9 TRIM. FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
for a given process, the user could perform analyses using the different algorithms, thus allowing
decision-makers to consider the impact of algorithms selection on predicted values.
2.3.4 FULL MASS BALANCE
One of the design features of TRIM.FaTE that sets it apart from many other multimedia
models is that it incorporates a full mass balance. In order to maintain a full mass balance, all
environmental media need to be modeled simultaneously, rather than sequentially. This allows
the model to properly account for all of the pollutant mass as it moves from within and between
media. This approach is in contrast to the methodology used in a set of linked models. With
linked models, it is difficult to model the time-fluctuating diffusive transport between the various
media. Furthermore, a series of interactions between more than two media is difficult to capture.
With TRIM.FaTE, all of the model compartments are fully coupled such that the exact
amount of mass that travels between compartments is accounted for explicitly and continuously.
Additionally, diffusion between compartments follows the time-dependent mass in each
compartment. As a result, in contrast to many other models, TRIM.FaTE considers time varying
concentration for diffusion and thus can provide a more accurate algorithm for diffusive mass
transfer among multiple compartments. That is, there is a continuous feedback system adjusting
the relative mass exchange among the compartments.
2.3.5 EMBEDDED PROCEDURE FOR UNCERTAINTY AND VARIABILITY
ANALYSIS
The overall TRIM model framework has been developed to allow for probabilistic
modeling such that variability and uncertainty can be explicitly and separately characterized.
This has involved the development of an approach to estimate variability and uncertainty within
TRIM, in a manner that allows for: (1) integration among the four TRIM modules; (2) tracking
the variability and uncertainty through the modules; and (3) feasible computational processing.
The implementation of this approach for uncertainty analysis is integrated within the
TRIM.FaTE module, as opposed to operating as a separate shell around the module. TRIM.FaTE
handles some of the calculations internally, and passes information to the uncertainty system
during a simulation. This close interfacing of the uncertainty software with the model allows for
greater flexibility in terms of what can be tracked and also dramatically reduces the processing
time required.
The key features of this approach to variability and uncertainty analysis are joint and
separate tracking of variability and uncertainty, characterization of variability and uncertainty of
model results with respect to parameter distributions and correlations, and identification of
critical parameters and correlations. In addition to providing information to support decision-
making, analyses of variability and uncertainty in TRIM will help to guide data and model
improvement efforts.
NOVEMBER 1999 2-10 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 2
INTRODUCTION TO TRIM.FATE
2.3.6 EXPOSURE MODEL FOR ECOLOGICAL RECEPTORS
TRIM.FaTE is also unique in its ability to estimate exposure for ecological receptors.
Several measures of ecological exposure are used in exposure-response models: concentrations
of chemicals in environmental media; body burdens or tissue levels of chemicals in the organism
of concern; and doses to the organism of concern (mass of chemical per mass of organism per
unit time). TRIM.FaTE can output chemical mass in all compartments at each time step, thus
providing body burden estimates for ecological receptors. TRIM.FaTE is also designed to divide
the compartmental chemical mass by the volume or mass of a compartment to estimate
concentrations in soil, sediment, water, air, or biota. Additionally, TRIM.FaTE can output
chemical intake for organisms of interest at the desired temporal and spatial scale.
Body burdens or tissue concentrations are useful measures of exposure because they
integrate exposure from all routes. Dietary exposure is already determined for mammals, birds,
and fish by TRIM.FaTE, and exposure to plants from both air and soil is calculated. However, if
body burden-response models are not available for particular pollutants, models may be available
that relate effects to concentrations in environmental media. These concentrations are available
directly from the TRIM.FaTE output as well. Models that relate doses to toxicity may also be
used, and doses may be calculated using any averaging time that is equal to or shorter than the
length of the TRIM.FaTE simulation.
Given the range of ecological exposure measures directly available from TRIM.FaTE, a
user will rarely be limited in the options for exposure-response models that may be used in an
ecological risk assessment. Example exceptions are: TRIM.FaTE does not estimate the
concentrations of chemicals in vertebrate organs, so models that relate toxicity to organ
concentration are not easily implemented: and TRIM.FaTE does not estimate the concentrations
of chemicals in potentially sensitive life stages offish, other than the adult, so using TRIM.FaTE
output with models that relate toxicity to concentration in a fingerling may give highly uncertain
results.
NOVEMBER 1999 2-11 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
3. OVERVIEW OF TRIM.FaTE CONCEPTS AND
TERMINOLOGY
The TRIM.FaTE methodology integrates OAQPS' needs and multimedia modeling
concepts into a unique model that serves as an integral part of the TRIM system. This chapter
provides an overview of the terminology central to the TRIM.FaTE module. An understanding
of the terminology and concepts presented in this chapter is crucial to understanding the
remainder of this report.
Because the terminology used in the world of environmental modeling can have multiple
meanings and implications, it is essential in the initial steps of any model conceptualization that
the terminology is clearly defined within the model framework. The terminology for multimedia
modeling is particularly complicated because multimedia models are, by nature,
multidisciplinary. Thus,
terminology can be especially
confusing because a single term
can have dramatically different
meanings in different disciplines.
Two general modeling terms are
TRIM.FaTE, in the adjacent text
box to provide a consistent basis
GENERAL MODELING TERMS
Scenario: A specified set of conditions (e.g., spatial,
temporal, environmental, source, chemical)
used to define a model set-up for a particular
, r. , f .1 r simulation or set of simulations
defined, for the purposes of
Simulation: A single application of a model to estimate
environmental conditions, based on a given
.... . scenario and any initial input values needed
for the discussion in this section
and the remainder of the
document.
The primary objective of the TRIM.FaTE module is to estimate the fate and transport of a
chemical pollutant through a modeled environment. Because the term "pollutant" can have
various meanings, the modeled unit of chemical mass in TRIM.FaTE is referred to as a chemical.
Within the context of TRIM.FaTE, a chemical is simply defined as a unit whose mass is being
modeled by TRIM.FaTE. A chemical can be any element or compound, or even group of
compounds, assuming the necessary parameters (e.g., molecular weight, diffusion coefficient in
air) are defined. Examples of chemicals that may be modeled in TRIM.FaTE are PAHs,
methylmercury, elemental mercury, and benzene.
3.1 SPATIAL TERMINOLOGY
In the TRIM.FaTE module, chemicals are contained within compartments. The term
"compartment" is an extension of what is referred to as "medium" in environmental fate and
transport modeling literature. The term "medium" was considered too limited in its scope
because it generally invokes images of abiotic systems such as soil or air, while TRIM.FaTE
includes both abiotic and biotic systems. Therefore, the term compartment was adopted for
TRIM.FaTE because it captures the flexibility of the TRIM.FaTE module in that it refers to both
abiotic and biotic systems. A compartment is defined as a homogeneous unit of space
characterized by its physical composition and within which it is assumed, for modeling purposes,
NOVEMBER 1999 ~] TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
that all chemical mass is homogeneously distributed and is in phase equilibrium. Multiple
chemicals can exist within a compartment, and the various phases that compose a compartment
(gases, liquids, solids) are assumed to be in equilibrium with respect to chemical partitioning.
For example, within an air compartment it can be assumed that the air molecules are in
equilibrium with the molecules of water vapor. Compartments can be either biotic, such as a
deer compartment, or abiotic, such as a stream compartment. Furthermore, two compartments
could have identical compositions and only be distinguished by their location in the modeled
environment; they are still separate compartments. It is important to note that biotic
compartments do not refer to an individual organism, but instead to the population of that
organism within a specified volume.
The term compartment type is used to denote a specific kind of compartment, such as an
air compartment type or a surface water compartment type. Compartment types are distinguished
from each other by the way they exchange chemical mass with other compartment types.
Compartments of the same type are distinguished from each other by their location and
sometimes also by the values that define their composition at a given location. For example, two
different surface soil compartments may have organic carbon contents of 0.015 and 0.01.
respectively, but they are both described by the compartment type called "surface soil."
Compartment types are classified as either abiotic or biotic. An abiotic compartment type
is a compartment type consisting primarily of a non-living environmental medium (e.g., air, soil)
for which TRIM.FaTE calculates chemical masses and concentrations. Abiotic compartment
types may also contain biota, such as the microorganisms responsible for chemical
transformation. A biotic compartment type is a
compartment type consisting of a population or
community of living organisms (e.g., bald eagle.
benthic invertebrate), or in the case of terrestrial
plants, portions of living organisms (e.g.. stems.
leaves), for which TRIM.FaTE calculates chemical
, ^ .. -TM ,. . , ., Air Surface Water
masses and concentrations. The adjacent text box
lists the abiotic compartment types included in
TRIM.FaTE. A list of the biotic compartment
types included in TRIM.FaTE is provided in
Section 3.3.
ABIOTIC COMPARTMENT TYPES IN
TRIM.FaTE
Root Zone Soil Sediment
Surface Soil Ground Water
Vadose Zone Soil
Each compartment is contained within a volume element. A volume element is a
bounded three-dimensional space that defines the location of one or more compartments. This
term is introduced to provide a consistent method for organizing objects that have a natural
spatial relationship. Typically, only one type of abiotic compartment is contained within a
volume element, although this is not a requirement (e.g., a volume element composed
predominantly of a water compartment could also contain a sediment compartment). Volume
elements are often identified by this abiotic compartment (e.g., surface soil volume element,
ground water volume element) and may contain numerous biotic compartments. All biotic
compartments within a volume element are implicitly associated with an abiotic compartment in
the volume element (i.e., a fish compartment is implicitly associated with a surface water
NOVEMBER 1999 3-2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
compartment), but abiotic compartments do not necessarily have to be associated with any biotic
compartments.
The size and shape of volume elements for a given TRIM.FaTE application depends on
the needs of the user. For example, if the user is most interested in the range of impacts .of a
chemical over a given water body, the water body could be divided into a number of volume
elements with depth, length, and width. Typically, the higher the desired resolution, the greater
the number and the more complicated the shapes of the volume elements.
Figure 3-1 shows the basic spatial relationships between chemicals, compartments, and
volume elements. This figure shows that chemicals are contained within compartments, and
compartments are contained within volume elements. Figure 3-2 demonstrates how multiple
compartments can exist within a single volume element. Because the air compartment is the only
abiotic compartment within the volume element in Figure 3-1, this volume element is referred to
as the air volume element. Likewise, the volume element in Figure 3-2 is referred to as the water
volume element. Figure 3-3 applies the concepts presented in Figures 3-1 and 3-2 by dividing a
hypothetical environment into volume elements and compartments.
Figure 3-1
Simple TRIM.FaTE System3
Volume Element
Compartment (Air)
Chemical
Chemicals shown in this figure, and all subsequent similar figures, are units of mass of the same chemical,
instead of multiple chemicals.
Figure 3-2
Multiple Compartments within a Single Volume Element
I\— Volume Element
Chemical
Compartment (Water)
Chemical
Compartment (Fish)
NOVEMBER 1999
3-3
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
Figure 3-3
"Real Life" Example (Multiple Volume Elements, Multiple Compartments)
Volume
Element I ,
Compartment ,
(Air) —^
Compartment
(Bird)
Compartment
(Deer)
Volume
Element
Compartment
(Water)
Compartment
(All Fish)
Compartment
(All Aquatic
Plants)
Compartment
(Vegetation)
Volume Element
VI
Compartment
(All Worms)
Compartment
(Surface Soil)
Volume Element
V
Compartment
(Vadose
Zone Soil)
Volume Element
IV
Compartment
(Groundwater)
Compartment Volume Element
(Sediment) III
V indicates Water Table
C8o021-schem2
NOVEMBER 1999
3-4
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTERS
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
3.2 TIME-RELATED TERMINOLOGY
Because TRIM.FaTE is a time-varying model, it is important that the temporal
terminology is clearly defined and consistently used. There are three time-related terms that are
central to understanding how TRIM.FaTE relates input data, fate and transport calculations, and
model outputs: simulation period, simulation time step, and output time step. Definitions of
these terms for the purposes of TRIM.FaTE are given below.
For a given TRIM.FaTE simulation, the simulation period is the entire length of time for
which the model is run and compartment masses and concentrations are calculated - in other
words, the time period from the beginning of the simulation until the end. The simulation period,
which is usually one or more years, is always greater than or equal to the length of time between
the first and last source emissions modeled. Thus, source emissions can occur for either all or
part of the simulation period.
For a given model simulation, the simulation time step is the time increment at which
the model calculates (and re-calculates iteratively throughout the simulation period) a new
inventor)' of compartment masses and concentrations. The simulation time step must be less
than or equal to the simulation period - typically, the simulation time step is much less than the
simulation period. Within a given simulation time step, the environmental conditions in each
compartment are assumed to remain constant.
Because time increments for input data can vary greatly across different model inputs, the
simulation time step may change within a simulation depending on how the inputs are changing
with time. For example, meteorological data (e.g, wind speeds, precipitation measurements)
may be available at hourly increments and source emission rate data may be available at monthly
increments. TRIM.FaTE accommodates these different input data increments by assuming that
values are constant within a given input data time increment.
When the simulation time step is small relative to the simulation period, it is often useful
to reduce the amount of output data by outputting results at periods longer than the simulation
time steps. The output time step, which is user specified, is defined as a length of time over
which the compartment masses and concentrations calculated at each simulation time step are
summarized and reported by the model.1 More than one output time step may be useful for a
given TRIM.FaTE simulation (e.g., one year, seventy years). Averaging values over appropriate
output time steps also can be useful for inputting output concentrations into exposure models.
Figure 3-4 presents hypothetical values for each of the TRIM.FaTE time-related terms
and demonstrates the magnitudes of each term relative to the others.
1 Even when the smallest output time step is specified as longer than the simulation time step, TRIM.FaTE
retains the data calculated for each time step for possible future use, rather than discarding it.
NOVEMBER 1999 3^5 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
Figure 3-4
Example of TRIM.FaTE Time-related Terminology"
simulation period
output time step
simulation time step |2y" | 4yre \ 3y» |
-------
CHAPTER 3
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
Table 3-1
Biotic Compartment Types Defined for TRIM.FaTE
Compartment Type
(Trophic Functional Group)'
Algae
Macrophyte
Water column herbivore
Water column omnivore
Water column carnivore
Benthic invertebrate (herbivore)
Benthic omnivore
Benthic carnivore
Terrestrial omnivore
Semi-aquatic piscivore
Semi-aquatic predator/scavenger
Terrestrial insectivore
Semi-aquatic herbivore
Terrestrial predator/scavenger
Semi-aquatic insectivore
Terrestrial herbivore
Semi-aquatic omnivore
Terrestrial ground-invertebrate feeder
Flying insect
Soil detritivore
Plant leaf
Plant leaf surface
Plant stem
Plant root
Representative Subgroup or Species
Generalized algal species
Elodea densa
Bluegill
Channel catfish
Largemouth bass
Mayfly
Channel catfish
Largemouth bass
White-footed mouse
Common loon
Mink
Belted kingfisher
Bald eagle
Black-capped chickadee
Mallard
Red-tailed hawk
Long-tailed weasel
Tree swallow
White-tailed deer
Mule deer
Black-tailed deer
Meadow vole
Long-tailed vole
Raccoon
Short-tailed shrew
Trowbridge shrew
Mayfly
Earthworm
Soil arthropod
Plant leaf
Plant leaf surface
Plant stem
Plant root
1 Plant parts constitute different compartment types even though they are not different trophic groups.
NOVEMBER 1999
3-7
TRIM.FATE TSD VOLUME i (DRAFT)
-------
CHAPTER 3
OVERVIEW OF TR1M.FATE CONCEPTS AND TERMINOLOGY
currently compartment types in TRIM.FaTE because the fate of persistent pollutants in woody
components of vegetation is not well understood.
3.4 LINKS
Chemical mass can be transferred between compartments within the same volume
element, as well as between compartments in adjacent volume elements. In addition, chemical
mass can be transformed within a single compartment. In order to estimate these processes, the
relationships, or links, between the compartments must be determined. A link is defined as a
"connection" that allows the transfer of chemical mass between compartments. Each link is
implemented by an algorithm or algorithms that mathematically represent the mass transfer.
Figure 3-5 expands on the concepts presented in Figure 3-2 by showing links between
compartments in different volume elements. The figure demonstrates that chemicals in the air
compartment can be transferred to the water compartment via a link.
Links do not necessarily exist between all adjacent compartments. This concept is
demonstrated in Figure 3-6 where chemical mass is transferred between the water and air
compartments and between the water and fish compartments, but not between the fish and air
compartments. A link may also exist between two chemicals within the same compartment in
the form of a transformation process (see Figure 3-7).
Figure 3-5
Two Linked Compartments in Separate Volume Elements
Volume Element —
Compartment (Air)
Chemical
Volume Element
Compartment (Water)
Chemical
Figure 3-6
Three Linked Compartments in Two Volume Elements
Link
Volume Element
Compartment (Air)
Chemical
Chemical
Compartment (Water)
—Volume Element
Link
Chemical
Compartment (Fish)
NOVEMBER 1999
3-8
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTERS
OVERVIEW OF TRIM.FATE CONCEPTS AND TERMINOLOGY
Figure 3-7
Transformation Links Between Chemicals within a Compartment
Chemical A
Transformation Link
Chemical B
Compartment (Water)
—Volume Element
Compartment (Fish)
In the TRIM.FaTE framework, links are a qualitative concept and represent the aggregate
of the algorithms and constants used to describe chemical mass transfer among compartments.
They can represent relatively simple processes, such as diffusion, or more complicated processes,
such as advection. For a given chemical, different links may represent different processes having
unique properties. For example, a link between two particular soil compartments may contain
information on the advective flow from one soil compartment to another while the shrew-to-soil
compartment link would contain information on the ingestion rate of soil by shrews. Figure 3-8
adds the concept of links to a portion of the hypothetical environment presented in Figure 3-3.
3.5 SOURCES
The set of all compartments in a TRIM.FaTE modeling scenario is assumed to contain all
of the chemical mass within the system being modeled, excluding sources. A source is an
external component that introduces chemical mass directly into a compartment. Examples of
sources would include the factory emissions of a chemical into an air compartment, the influx of
chemical into a river compartment as the result of a spill, or the influx of chemical mass from
outside the modeling boundaries into an air compartment as the result of a distant power plant.
TRIM.FaTE is designed to accommodate single or multiple source scenarios. Figure 3-9
expands upon Figure 3-6 by adding a source emitting into the air compartment. Figure 3-10
presents the "real life" TRIM.FaTE system shown in Figure 3-8 with the addition of sources.
Figure 3-9
Linked TRIM.FaTE System with Source
Link
Volume Element
Source ,
Chemical
Compartment (Air)
Chemical
Compartment (Water)
—Volume Element
Link
Chemical
Compartment (Fish)
NOVEMBER 1999
3-9
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
I
O
U
"O
00 .
w E
QJ W
!. X
s U
H
U.
s
z
i
a:
H
D
Z
<
e/3
in
U
§
U
U
Qi
fc S
<^
u o
<,
Q
LU
5
O
D
H
tu
5
H
L
c
-------
u
of x
ua O
iS
of
H
Q
Z
C/3
H
C-
LLi
u
§
u
u
u.
3
O
.
E
«
x
U
3
R
fa.
Q
UJ
2
o
Q
p
tu
u_
2
o
a:
u>
m
-------
[This page intentionally left blank.]
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
4. CONCEPTUAL DESIGN AND MASS BALANCE
FRAMEWORK FOR TRIM.FaTE
This chapter, building on the definitions and spatial and temporal concepts discussed in
Chapter 3, presents the overall logic implemented in TRIM.FaTE for expressing transport and
transformation of chemicals in a multimedia environment. Specifically, this chapter discusses
and illustrates the mass balance approach and describes the processes simulated in TRIM.FaTE.
The actual algorithms used to implement the approach are documented in Volume II.
4.1 CONCEPTUAL DESIGN
TRIM.FaTE calculates, given an initial mass inventory and mass inputs over time from
one or more sources, the mass of one or more chemicals being modeled in each compartment in
the modeled system for each simulation time step. With the volume and estimated chemical
mass of each compartment, TRIM.FaTE can then calculate the concentration of each chemical in
each compartment at each time step.
The development of TRIM.FaTE began with a conceptual diagram of the relationships
and processes that affect chemical transport within the environment. The current version of this
diagram is shown in Figure 4-1. In this figure, biotic compartments are represented by rectangles
and abiotic compartments are represented by ovals. The various lines illustrate the potential
chemical transfers between each of the components of the ecosystem.
4.2 GOVERNING MASS BALANCE EQUATIONS
TRIM.FaTE has been developed with an emphasis on conserving chemical mass. This
means that the entire quantity of a chemical input to the system is tracked throughout the system
being modeled. When applied to a specific compartment (e.g., soil, or a mouse population), this
implies that, over a given time period, the amount of the chemical in the compartment at the end
of the period is equal to the amount of the chemical in the compartment at the beginning of the
period, plus the gains of the chemical that occurred during the time period, minus the chemical
that was lost from the compartment during the time period.
To date, the mass balance approach has been implemented primarily for first-order linear
processes. Therefore, this discussion is limited to algorithms of this type. It is important to note
that higher order non-linear methods can also be implemented within this structure.
First-order transfers between compartments are described by transfer factors, referred to
as T-factors. In most cases, T-factors are in units of inverse time. Technically, the units of a T-
factor depend on the sending and receiving chemicals, as what is actually being preserved across
the exchange is the amount of "core" chemical present in all transforming chemicals. The T-
factor is the instantaneous flux of this "core" compound per amount of the compound in the
sending compartment. The definition of the "core" compound depends on the particular
chemicals being considered (e.g., for the test case the core compound is an Hg atom).
NOVEMBER 1999 4-1 TRIM.FATE TSD VOLUME I (DRAFT)
-------
03
V3
t/1
1_ UJ
E. u
U U
o
UJ
£
U.
2
£2
_c
T5
^a!
§
o
3
•o
o
o
c
CO
L-
a.
£•
S
U
C
CO
o.
E
o
CJ
"co
c
CJ
CJ
o
>-,
a
E
-o
o
OS _C
tn «
o to
•c o.
S c
i s
C o
•i-s
I =
fc o
2
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
In most cases, it is expected that the molecular weights of the inter-transforming
chemicals will be similar, and so preservation of the "core" compound will be essentially
equivalent to preservation of mass. Significant differences between the preservation of "core"
compound and the preservation of mass occur only if the sending and receiving chemicals have
very different molecular weights. This has not been the case to date, but may occur as the model
is applied to other transforming chemicals.
The preservation of "core" compound and the preservation of mass would be identical if
the time-dependent masses of the reaction products were estimated simultaneously for each
chemical of interest. Given the current computation and logistical demands of modeling
transformation in many compartments, this is not seen as a practical general solution.
A simplification of a first-order transfer process is shown in the top part of Figure 4-2 for
a system of one chemical, two compartments, and two transformation sinks, where
transformation is treated as an irreversible loss. Denoting by Na(t) and Nb(t) the mass of chemical
in compartments a and b, respectively (in units of mass), it can be seen that:
Chemical gains for compartment a = Sa + TbaNh
Chemical losses for compartment a = TabNa + RaNa
(1)
(2)
and
where:
Na
^
Sa
Tn>
Chemical gains far compartment b - TabNa
Chemical losses for compartment b - TbaNb + RbNb
mass of chemical in compartment a, units of mass
mass of chemical in compartment b, units of mass
chemical source outpurting to compartment a, units of mass/time
transfer factor for movement of chemical from compartment a to
compartment b during simulation time step, units of /time
transfer factor for movement of chemical from compartment b to
compartment a during simulation time step, units of /time
reaction loss of chemical in compartment a, units of '/time
reaction loss of chemical in compartment b, units of /time.
(3)
(4)
NOVEMBER 1999
4-3
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
Figure 4-2
Example of First-order Transfer Processes for Two Compartments, One Chemical
(Transformation Treated as Irreversible Sink)
Source
Compartment A
Mass of chemical in
compartment = ty
dNjdt
dNbldt
dSink Idt
dSinkJdt
T
ab
R
0
o o"
0 0
0 0
0 0
X'
^
Sinka
Sinkb
"T"
S
a
0
0
0
(5)
NOVEMBER 1999
4-4
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
The constraint that mass balance must be preserved means that, over any time interval,
the change in mass in a compartment is equal to the gains minus the losses in mass over the time
interval. The instantaneous change in mass with respect to time is the derivative with respect to
time, denoted by dN/dt. Thus, the mass balance constraint, when applied to the simple system
discussed here, yields a system of two linked differential equations:
rf/V
—2. = S + T, N, - (R + T . W (6)
j^ a ba b ^ a ab' a \"/
dt
JK
T JV' - (Rk + T.)N. (7)
7 ao a ^ D DO' D
Additional terms are needed to properly account for the chemical mass. In particular, the
fate of the chemicals after reacting must be tracked. For this reason, two additional
compartments are added to the system, and serve as the repository of the chemicals after reaction.
These are referred to as "sinks," since after the chemical is transferred into these compartments,
it no longer moves to any other compartments. While the chemical would continue to move in
its altered form throughout the system, this movement is not of interest in this example.
Denoting by Sinka and Sinkh, the mass in the reaction sinks for compartments a and b,
respectively, the complete system is:
dN
a — C _i_ T" Ar (D _i_ T" \ \r
-£- - sa + ^A - (*fl + W. (g)
dN.
—- = T ,N - (R, + T, W, IQ\
i ab a ^ b ba' b \ '}
dSink
a- = R N (10)
dt
dSinkb i ] | j
_____ _ RbNb
This system of equations is shown in matrix form in Equation (5) at the bottom of Figure 4-2.
If the fate of the transformed chemical is of interest, and the necessary algorithms and
input data are available for the transformed chemical, then the mass balance approach can be
modified accordingly. Figure 4-3 shows a generalization of the previous example, including the
matrix form of the system [Equation (12)], to the case where the transformed chemical is
modeled in addition to the chemical being transferred. In this case, transfer factors are added for
the transformed chemical to account for additional possible transfers.
NOVEMBER 1999 4-5 TRIM.FATE TSD VOLUME! (DRAFT)
-------
UJ
C£
C\
S
<
£*
CO
s/2
§
i
(U
SH
«
r c
a. c
S 2
o ^
> u
£ 8
*? ££
-* £ c
« <« c
i; w '^
3 t» r3
.BF 8 E
a.
i_
>
u
O _c
o „
•£ J2
"5.
i
i
c ^
co Z!
0 II
E -
cc
c —
7S z
0 i
E -
01 C
.c a
" E
75 £
(/3 C
S E
k
CM
2
CJ
DC
fo
L
^
,
II -*
?z
0 KT
u_ S
i—
r7
s z
O CM
o: -s
h-
H "^
s^
o ^
1—
11 ~ra
S^
O -
t-
^
CM
i
CO
u
a>
0
it.
u
c —
-2z
O ^
03 ^£,
•a?
T_
CO
o
E
0)
c —
a ^
.y u
E -
Q) C
u E
'° S
U) O
a E
•? °
n
c
o
o
CO
m
DC
c —
75 Z
y i
E -
a> c
^= a
" E
o «
in c
m *-
2L
Z
E
*-
i_
CO
a.
E
o
o
rN -.
QS
"S "xs
a>
u
o
tO
2.
Of
t
<
0 ul
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
Applying this same approach to a general system with M compartments (including all
sinks), and allowing the transfer factors and source terms to depend on time as well, results in a
system of linked differential equations of the form:
^N = A(t)N(f) + *(0,
dt
(13)
where:
N(i) = an A/-dimensional vector whose rth entry is the mass in the rth
compartment
A(t) = an M x M time-dependent matrix
5(/) = an M-dimensional vector accounting for the source terms in each
compartment.
The matrix A(t) is referred to as the transition matrix for the system. This term is
borrowed from Markov theory (Schneider and Barker 1989), although the model is not strictly a
Markov process. The vector 5(0 accounts for pollutant sources located within specific
compartments. The vector A'0 is the initial distribution of mass among the compartments.
4.3 PHASES
There are multiple environmental phases
within many of the compartments in TRIM.FaTE.
The most common phases are liquid, gas, and solid,
which are assumed to be at chemical equilibrium
within a compartment in this model unless otherwise
specified. Other phases may include biotic phases
(e.g., algae in surface water). The adjacent text box
lists the phases currently implemented in
TRIM.FaTE for each medium.
In any compartment, the total amount of
chemical present is made up of the sum of the
amounts in the different phases. Because chemical
equilibrium among phases is assumed, the ratios of
the concentrations in the individual phases are
constant for a given chemical. The fraction of the
chemical that is in each phase in a compartment can
easily be calculated. The chemical mass in each
phase is tracked in TRIM.FaTE because transfer factors are sometimes phase dependent (i.e., the
transfer factor for particle deposition from air is dependent on chemical mass in the particle
PHASES CURRENTLY
IMPLEMENTED IN TRIM.FaTE (listed
by medium)
Air
• vapor
• suspended paniculate
Soil
• soil pore water
• vapor
• soil solids
Surface water
• suspended solids
• water
• algae
Sediment
• sediment pore water
• sediment solids
NOVEMBER 1999
4-7
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
phase of air). The mathematical details related to implementation of phases in TRIM.FaTE are
presented in Chapter 2 of Volume II of the TRIM.FaTE TSD (U.S. EPA 1999d).
4.4 FATE, TRANSFORMATION, AND TRANSPORT PROCESSES
In TRIM.FaTE, the following processes are addressed and implemented as first-order
processes for the modeling of the transfer and transformation of chemicals.
Advective processes;
Diffusive processes;
Dispersive processes;
Biotic processes; and
Reaction and transformation.
More detailed explanation of the mathematical representation of these processes and
documentation of all of the currently implemented algorithms are presented in Volume II of the
TRIM.FaTE TSD (U.S. EPA 1999d).
4.4.1 ADVECTIVE PROCESSES
An advective process is one in which a chemical is transported within a given medium
that is moving from one compartment to another. Mackay (1991) refers to this as a "piggyback"
process, in which a chemical is "piggybacking" on material that is moving from one place to
another for reasons unrelated to the presence of the chemical. Advective processes are modeled
using first-order methods in TRIM.FaTE. Mathematically, all that is required to calculate the
advective flux is the velocity of the moving phase and the amount of the chemical that is in the
moving phase. Examples of advective processes considered for transport of a chemical are:
erosion from a surface soil compartment to a surface water compartment, runoff from a surface
soil compartment to a surface water compartment, and advective transport from one air
compartment to another due to the wind field.
4.4.2 DIFFUSIVE PROCESSES
In a diffusive process, a chemical is transported from one compartment to another as a
result of the magnitude and direction of the concentration differences between the two
compartments at the interface between the two locations. Examples of diffusive processes
considered include exchange between air compartments and soil or surface water compartments.
exchange between benthic sediment compartments and surface water compartments, and
exchange between air compartments. Models for diffusion frequently use non-first-order
methods; however, these are often approximated with first-order methods. All diffusive
processes are currently modeled in TRIM.FaTE using first-order methods. Diffusion rates are
based on the compartment concentrations at the beginning of each simulation time step.
NOVEMBER 1999 4-8 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 4
CONCEPTUAL DESIGN AND MASS BALANCE FRAMEWORK FOR TRIM.FATE
4.4.3 DISPERSION
Dispersion refers to the "spreading out" of a chemical during advective transport, and
may result in movement perpendicular to the direction of advective flow. In TRIM.FaTE,
dispersion is explicitly addressed (as a first-order process) in transfers between surface water
compartment types. For surface water dispersion, the methods in the Water Quality Analysis
Simulation Program (WASP) water transport model are used (Ambrose et al. 1995).
4.4.4 BIOTIC PROCESSES
The transport of chemicals to biota (i.e., into biotic compartments) consists of diffusive
and advective processes, though the latter term is rarely used by biologists. Chemicals diffuse
into plant leaves from air; chemicals deposit onto plant leaves with particles in air, an advective
process. The uptake of chemicals from soil or water in soil by plant roots or earthworms is
treated as diffusion, though water carries the chemical into the plant (advection). Similarly,
chemicals are assumed to enter algae, macrophytes, and benthic invertebrates by diffusion. The
major advective process for animals is food intake by fish, birds, and mammals.
The only transport process within biota that is included in TRIM.FaTE is transport
between roots and leaves through the plant stem in xylem and phloem fluids. The distribution of
chemicals among organs in fish and wildlife is not a feature of TRIM.FaTE.
4.4.5 REACTION AND TRANSFORMATION
Reaction and transformation processes include biodegradation, photolysis, hydrolysis,
oxidation/reduction, and biotic metabolism. These are processes that transform a chemical
species into another chemical species. Reaction and transformation are modeled in TRIM.FaTE
as reversible reactions using first-order reaction/transformation rates (or, equivalently,
transformation half-lives). The first-order transformation rates may incorporate more than one of
the processes involved. Depending on the algorithm and compartments involved, the mass of
chemical transformed may be either lost from the system (i.e . transferred to a sink), or tracked as
a different chemical.
NOVEMBER 1999 4-9 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
CHAPTER 5
APPLICATION OF TR1M.FATE
5. APPLICATION OF TRIM.FaTE
This chapter describes the application of the concepts presented in Chapters 3 and 4 by
briefly explaining each of the main steps necessary to set up and perform a simulation with
TRIM.FaTE. It explains the methods associated with key steps in the modeling process, provides
a general sense of the level of effort associated with performing a TRIM.FaTE simulation, and
summarizes the inputs and outputs of the model. This chapter is not intended to be a "user's
guide" to the model. OAQPS recognizes the importance of developing detailed user's guidance
material for TRIM.FaTE that will assist users in defining the spatial and temporal modeling
resolution, compartments and linkages, as well as parameters and initial conditions. Such
material will be developed during the next phase of TRIM development activities. The intended
purpose of this chapter, however, is simply to provide an understanding of how TRIM.FaTE
works by explaining in general terms how it is applied to model the fate and transport of air
emissions from one or multiple sources.
5.1 STRUCTURE OF A TRIM.FaTE SIMULATION
One of the strengths of TRIM.FaTE is that it is designed to be an iterative and flexible
model. When the modeling process begins, there is a general sequence that typically is followed.
After the initial step, however, there is no fixed order in which the modeling steps are necessarily
performed (although some steps must be completed before others begin). This process is shown
in Figure 5-1. The boxes on the left side of the figure represent a partitioning of the modeling
sequence into five broad areas. These areas include: basic problem definition, specification of
links, simulation setup, simulation implementation, and analysis of results. The particular
division into five such areas is somewhat arbitrary, and in an actual application the progression
may not be quite as linear as that shown in the figure. However, all of these steps are necessary.
The vertical arrows between these boxes represent the typical order of events in the modeling
process. The arrows on the left side of the boxes indicate the iteration that may be necessary or
desired.
The shapes under the heading "Primary' Tools" represent the main tools used in the
modeling process. The arrows from these shapes to the flow boxes indicate where in the
modeling process these tools would be used. To focus on key aspects of the TRIM.FaTE
.approach, only selected tools are shown. There are other tools that may be necessary that are not
included in this figure. Such tools include pre-/post-processing software, which may automate
some aspects of the process, uncertainty and variability analysis software, and general user
interface software.
5.2 PROBLEM DEFINITION
The first step in the TRIM.FaTE modeling process is a clear statement of the problem
definition. Through the formulation of this problem definition , the chemical(s) and source(s) to
be modeled, the initial spatial features of the ecosystem to be modeled, and the simulation period
NOVEMBER 1999 5-1 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
Figure 5-1
TRIM.FaTE Modeling Sequence
PROCESS FLOW
PRIMARY TOOLS
o
Q.
O.
re
Define Problem
Specify modeling region
Specify volume elements
Specify compartments
Specify simulation period
Specify chemicals and
sources to be modeled
Specify Links Between
Compartments
Specify links
For each link specify
algorithm to use from
library
Set Up Run
Set initial conditions/
background concentrations
Specify inpul data
Sei ouipui time step (s)
Perform Simulation
Call algorithm hbrar\ for
each link to determine
transfer factors
Calculate mass distribution in
system of compartments at
requested output time step(s)
Convert mass to
concentration
Calculate concentrations for
output time step(s)
DATA
Spatial data
Meteorological data
Environmental setting
data
Chemical properties
Source terms
ALGORITHM LIBRARY
GENERAL CALCULATION
TOOLS
Differential equation solver
Partial differential equation solver
Analyze Results
NOVEMBER 1999
5-2
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
are determined. Using the nomenclature presented in Chapter 3, the volume elements and
compartments within the volume elements are specified.
5.2.1 DETERMINING SCALE AND SPATIAL RESOLUTION
This section introduces the
A parcel is a planar (i.e., two-dimensional)
geographical area used to subdivide the
modeling region. Parcels, which can be
virtually any size or shape, are the basis for
defining volume elements and do not change
for a given scenario. There can be separate
parcels for air and for the land surface (surface
soil or surface water).
considerations for defining the overall
modeling scale and the level of spatial
complexity (i.e., location, size, shape, and
number of parcels) in a TRIM.FaTE
analysis. After the initial scenario is
constructed and a simulation has been
completed, the preliminary results need to be
evaluated to confirm that an appropriate
level of resolution has been used. An
example of a general approach for determining appropriate scale and spatial resolution as well as
suggestions for defining compartments are included in Appendix C. At some point, EPA may
consider automating the process of determining spatial resolution to enhance user consistency.
5.2.1.1 Specifying the Modeling Region
The first step in determining the scale and spatial resolution of a TRIM.FaTE scenario is
to determine the modeling region (i.e., spatial boundaries of the analysis). In this step, the user
specifies the extent of the area to be modeled. A user should consider factors such as mobility of
the modeled chemical(s), location of source(s). and background concentrations of the
chemical(s). In regions where the predominant wind direction is variable, the modeling region
may be defined to include and extend beyond the region of interest to account for the possibility
of pollutant mass leaving and re-entering the system.
5.2.1.2 Specifying Parcels
The next step in determining the scale and spatial resolution of a TRIM.FaTE scenario is
to specify the modeling parcels. Parcels are planar geographical areas of any size or shape that
are the basis for defining volume elements and do not change for a given scenario. The higher
the number of parcels in a given scenario, the higher the spatial resolution. However, more
parcels generally correspond to greater resource requirements, both in terms of input data
collection and model completion time.
Beyond resource considerations, there are three principal technical considerations for
determining the parcels for a TRIM.FaTE scenario: the likely pattern of transport and
transformation of each chemical of concern (i.e., where significant concentration gradients are
likely to be), locations of natural boundaries, and locations of the targets. The target is defined as
a receptor, either human or ecological, or landscape component (e.g., lake, wetland, agricultural
plot) of interest.
NOVEMBER 1999 5-3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
For the chemical(s) of interest, the three most important factors for determining the
appropriate modeling scale and resolution are the atmospheric transport of chemical mass,
rapidity of chemical transport, and degradation of the chemical(s) in the environment.
Understanding the atmospheric transport of the chemical(s) of interest is useful in developing
both the modeling scale and spatial resolution. Because air pollutants travel more rapidly in air
than any other medium, insight into atmospheric transport can provide the user with a general
idea of the extent of chemical transport and thus can be useful in determining the modeling scale.
Furthermore, this information can provide the user with a general picture of the path of chemical
transport, helping the user determine where higher spatial resolution may be beneficial.
Information about the mobility and degradation of the chemical(s) of interest, when
combined with land use data, can provide additional insight into the transport in media other than
air. It can be helpful in refining the scale of the scenario as well as providing additional input to
help determine the spatial resolution of the scenario.
Natural boundaries are also an important consideration in developing parcels. These
natural boundaries may include areas such as an air shed that can be identified by a combination
of geographic and meteorologic conditions or a watershed or valley. An air shed can include
large valleys such as the Sacramento Valley (CA) where, due to inversion layers and diurnal
wind patterns, the air mass is confined and well mixed throughout the area for a large portion of
the time. Air shed boundaries can also include smaller valleys when meteorological conditions
produce long residence time for the air mass in the bounded region. Air shed boundaries are
useful in providing information about the scale of the model region (i.e.. external boundaries of
the system).
Watersheds are also useful in determining the scale of the system as well as the size and
location of parcels within the system, especially if the concentration in a particular lake or
wetland is of interest. Watershed boundaries can be identified or approximated from
topographical maps by tracing ridgelines and noting the origin and direction of flow for streams
and rivers. The size and location of a watershed can influence the transfer of chemical to water
bodies within the basin.
Another important consideration in developing parcels is the location of the target(s).
The location of the target(s) is important because it allows the user to focus the analysis on the
area(s) of interest, thus allowing resolution to be coarse in areas that are not expected to affect
the chemical concentrations near the targets and resolution to be finer in the areas that may have
an impact on the targets. Evaluations to determine the extent to which these nonessential area(s)
can be simplified without significantly changing the model outcome are ongoing.
The illustrative approach to specifying parcels described in Appendix C generates a
starting point for any given analysis objective for which TRIM.FaTE is designed. The approach
is intended to impart some consistency and transparency into the scenario setup process.
Additionally, after a scale has been chosen, one must determine if that scale is appropriate when
compared to other sources of model uncertainty.
NOVEMBER 1999 5-4 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TR1M.FATE
5.2.1.3 Determining Volume Elements
After the parcels have been determined for a scenario, the volume elements
corresponding to those parcels are specified. This step involves determining the appropriate
number of volume elements and specifying the appropriate depth for each one. Whereas parcels
only represent the modeling region in two dimensions, volume elements add the component of
depth, thus representing the modeling region in three dimensions (the location and two-
dimensional planar shape of a volume element corresponds exactly to the relevant parcel). The
volume elements are determined from a general knowledge of mixing heights in air, average
depth of water bodies or approximate levels of stratification, and typical demarcations in the soil
horizon. The development of volume elements represents the final step in specifying the spatial
resolution of the modeling region.
5.2.1.4 Determining Compartments
Abiotic
Abiotic compartment types are determined by the predominant abiotic medium in the
volume element within which they are contained. At least one abiotic compartment must be
contained within each volume element and, although not typically utilized, the model framework
does support multiple abiotic compartments within a volume element. In most cases, the
determination of abiotic compartments is an implied step because they are simply defined by the
predominant abiotic media within the volume element. For example, if a given volume element
is composed predominantly of surface soil, a surface soil compartment would be included in the
volume element.
Biotic
Of the available biotic compartment types, the user is only required to run the model with
those that significantly influence the overall mass balance of the chemical in TRIM.FaTE. In
applying the model to PAHs, for example, the plant biomass significantly influences the mass
balance in the system. Thus, it would not be appropriate to run a PAH application of the model
without the plant compartment types, even if the particular application was only interested in the
concentrations in abiotic compartments. A user can choose to exclude terrestrial or aquatic biotic
compartment types from the analysis if they are not large reservoirs or significant sinks for
chemical mass. The only vertebrate or invertebrate compartment types in TRIM.FaTE that are
required for model runs are those that are part of the food chain for a trophic group of concern.
A user can perform a TRIM.FaTE assessment for a whole trophic group if a
representative species is chosen, particularly if distributions of input parameters are used. In
addition, a user can choose particular animal species of concern (e.g., threatened or endangered
populations) and parameterize the model for those species.
NOVEMBER ] 999 5-5 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
5.2.2 DETERMINING SOURCE(S) AND CHEMICAL(S) TO BE MODELED
In determining the sources and chemicals to be modeled, the user must consider the
objective of the modeling exercise and the effects endpoints of concern. The user would need to
assess the modeling region and determine the sources and chemicals that can be expected to have
an impact on the endpoints of concern. The user must then decide, given the resource constraints
of the analysis, which sources and chemicals should be included in the modeling analysis.
Ideally, this would be part of the problem definition process performed earlier in the analysis
because the chemical(s) and location of the source(s) influences how the parcels are initially
layed out.
5.2.3 DETERMINING SIMULATION PERIOD
After determining the sources and chemicals to be modeled, the user must determine the
appropriate simulation period by considering the modeling objective, the lifetime of the modeled
source(s), the persistence and, in some cases, mobility of the modeled chemical(s), and the
effects endpoints of concern. In addition, the user should again consider the resource limitations
when determining the simulation period because the selected simulation period directly affects
the necessary computing time.
5.3 DETERMINING LINKS/ALGORITHMS
The second step shown in Figure 5-1 is to specify the links between all adjacent
compartments specified for a given scenario (the compartment types currently available in
TRIM.FaTE are listed in Chapter 3). The system of links is one of the most critical components
of TRIM.FaTE. This component is critical because the links determine how the processes that
drive chemical transfer and transformation will be approximated in TRIM.FaTE. By specifying a
link between two adjacent compartments, it is assumed that some method exists by which to
estimate the transfer of chemical through the link. If more than one method is already available
in the algorithm library, then it is necessary to specify which of the algorithms to use. If an
algorithm is not in the algorithm library, then it must be "added" so that it can be accessed by the
underlying software. Methods for adding additional algorithms to the library will be included in
a future version of the model. Table 5-1 presents the currently implemented links between abiotic
compartment types, and Table 5-2 presents the currently implemented links between biotic
compartment types and abiotic compartment types, and between biotic compartment types.
TRIM.FaTE also has the flexibility to use model results from single-medium models
(e.g., ISC) in place some of the internal links and algorithms. In this case, the output from the
external model would replace the calculations of fate and transport within the specific medium.
A description of how external models can be integrated with TRIM.FaTE is presented in
Appendix B.
NOVEMBER 1999 5-6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
Table 5-1
Links and Processes Addressed for Abiotic Compartment Types
Links Between Compartment Types
Receiving
Air
Surface Soil
Root Zone Soil
Vadose Zone Soil
Surface Water
Sediment
Ground Water
Air Advection Sink
Surface Water Advection Sink
Sediment Burial Sink
Sending
Air
Surface Soil
Surface Water
Surface Soil
Root Zone Soil
Air
Root Zone Soil
Surface Soil
Vadose Zone Soil
Vadose Zone Soil
Root Zone Soil
Surface Water
Surface Soil
Air
Sediment
Surface Water
Surface Water
Vadose Zone Soil
Air
Surface Water
Sediment
Processes Addressed
Bulk Advection
Diffusion
Resuspension
Diffusion
Diffusion
Erosion
Runoff
Diffusion
Diffusion
Dry Deposition
Wet Deposition
Diffusion
Percolation
Diffusion
Percolation
Diffusion
Diffusion
Percolation
Diffusion
Percolation
Bulk Advection
Dispersion
Erosion
Runoff
Dry Deposition
Wet Deposition
Diffusion
Resuspension
Pore Water Diffusion
Abiotic Solids Settling
Pore Water Diffusion
Recharge
Percolation
Bulk Advection Beyond
Bulk Advection Beyond
System Boundary
System Boundary
Solids Advection Beyond System Boundary
NOVEMBER 1999
5-7
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
Table 5-2
Links and Processes Addressed For Biotic Compartment Types
Links Between Compartment Types
Receiving
Leaf Surface
Surface Soil
Leaf
Air
Root
Stem
Soil Detritivore
Root Zone Soil
Flying Insect
Terrestrial Ground-invertebrate Feeder
Terrestrial Vertebrate Herbivore
Sending
Air (Particulates)
Air (Rain Water)
Leaf
Leaf Surface
Leaf
Terrestrial Ground-Invertebrate Feeder
Terrestrial Vertebrate Herbivore
Terrestrial Omnivore
Terrestrial Insectivore
Semiaquatic Omnivore
Predator/Scavenger
Semiaquatic Insectivore
Semiaquatic Herbivore
Semiaquatic Piscivore
Leaf Surface
Air
Stem
Leaf
Root Zone Soil
Root Zone Soil (Water Phase)
Leaf
Root Zone Soil
Root
Soil Detritivore
Sediment
Soil Detritivore
Surface Soil
Air
Leaf
Leaf Surface
Surface Soil
Air
Processes Addressed
Dry Deposition b
Wet Deposition b
Diffusion/Advection
Particle Washoff "
Litterfall b
Litterfall b
Excretion a
Uptake a
Diffusion/Advection
Uptake a
Uptake "
Uptake 3
Equilibrium Partitioning
Uptake •
Diet"
Inhalation b
Diet"
Inhalation "
NOVEMBER 1999
5-8
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
Links Between Compartment Types
Receiving
Terrestrial Omnivore
Terrestrial Insectivore
Semiaquatic Omnivore
Predator/Scavenger
Semiaquatic Insectivore
Semiaquatic Herbivore
Semiaquatic Piscivore
Surface Water
Sending
Leaf
Leaf Surface
Soil Detritivore
Surface Soil
Air
Soil Detritivore
Air
Benthic Invertebrate
Soil Detritivore
Herbivorous Fish
Omnivorous Fish
Carnivorous Fish
Surface Soil
Air
Terrestrial Vertebrate Herbivore
Terrestrial Omnivore
Terrestrial Insectivore
Soil Detritivore
Benthic Invertebrate (Insect)
Benthic Invertebrate (Insect)
Benthic Invertebrate
Leaf
Terrestrial Omnivore
Terrestrial Herbivore
Terrestrial Insectivore
Herbivorous Fish
Omnivorous Fish
Carnivorous Fish
Semiaquatic Omnivore
Semiaquatic Insectivore
Semiaquatic Herbivore
Semiaquatic Piscivore
Algae
Macrophyte
Water Column Herbivorous Fish
Processes Addressed
Diet b
Inhalation "
Diet"
Inhalation "
Diet"
Inhalation b
Dietb
Dietb
Diet"
Dietb
Excretion
Equilibrium Partitioning'
Equilibrium Partitioning K
Elimination ""
NOVEMBER 1999
5-9
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM FATE
Links Between Compartment Types
Receiving
Surface Water (continued)
Algae
Macrophyte
Benthic Invertebrate
Sediment
Water Column Herbivorous Fish c
Water Column Herbivorous Fish a
Water Column Omnivorous Fish c
Water Column Omnivorous Fish d
Water Column Carnivorous Fish c
Water Column Carnivorous Fish d
Benthic Omnivorous Fish c
Benthic Omnivorous Fish d
Benthic Carnivorous Fish c
Benthic Carnivorous Fish d
Sending
Water Column Omnivorous Fish
Water Column Carnivorous Fish
Benthic Omnivorous Fish
Benthic Carnivorous Fish
Surface Water
Surface Water
Sediment
Benthic Invertebrate
Algae
Algae
Surface Water
Herbivorous Fish
Herbivorous Fish
Surface Water
Water Column Omnivorous Fish
Water Column Omnivorous Fish
Surface Water
Benthic Invertebrate
Benthic Invertebrate
Surface Water
Benthic Omnivorous Fish
Benthic Omnivorous Fish
Surface Water
Processes Addressed
Equilibrium Partitioning ac
Elimination M
Equilibrium Partitioning ac
Elimination M
Equilibrium Partitioning K
Elimination M
Equilibrium Partitioning ac
Elimination bd
Uptake '
Uptake a
Uptake 3
Equilibrium Partitioning a
Diet"
Dietb
Gill filtration a
Dietb
Diet b
Gill filtration a
Dietb
Dietb
Gill filtration '
Dietb
Diet"
Gill filtration a
Diet"
Diet"
Gill filtration a
' Uptake, filtration, or partitioning which includes diffusion, advection, and/or active accumulation by organism.
b Advection processes.
c Equilibrium model for bioaccumulation by fish.
" Bioenergetic model for bioaccumulation by fish.
NOVEMBER 1999
5-10
TR1M.FATE ISO VOLUME 1 (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
5.4 SIMULATION SETUP
The third step shown in Figure 5-1 is the preparation of a simulation after the volume
elements, compartments, and links have been specified. This involves specifying the chemical
properties of each modeled chemical, the initial distribution of chemical mass in the
compartments, the data for each modeled source, all environmental setting data needed by the
selected algorithms, and the output time step(s) of interest. The role that each of these inputs
plays in estimating the fate and transport of chemical mass is briefly explained in this section. A
complete list of all the inputs for the currently implemented algorithms is presented in Appendix
D.
5.4.1 CHEMICAL PROPERTIES
To estimate the fate and transport of chemical
mass through the system, the relevant properties for
each modeled chemical must be specified. The list of
chemical properties that are necessary for a given
simulation varies depending on the chemical (e.g.,
organic chemicals differ from metals) and the media
and biota modeled. Several examples of abiotic and
biotic chemical properties are listed in the adjacent
text box.
5.4.2 INITIAL CONDITIONS
For each compartment in a scenario, the user
must specify the initial conditions, i.e., the initial
inventory of chemical mass. Default values of zero
may be assumed in some compartments for pollutants that have a relatively short half-life or if
the objective of the simulation is to assess the effects of a source (or sources) in the absence of
background, but it is important to have estimates of initial conditions if the pollutant is persistent
and the objective is to assess "cumulative" exposures, or if results of the analysis are to be
compared with monitoring and measurement data.
5.4.3 SOURCE DATA
ILLUSTRATIVE EXAMPLES OF
ABIOTIC CHEMICAL PROPERTIES
• half life (in each environmental
medium)
• Henry's Law constant
• melting point
• molecular weight
ILLUSTRATIVE EXAMPLES OF
BIOTIC CHEMICAL PROPERTIES
• half life (for each modeled species)
• accumulation factor (for modeled
animal species)
• bioaccumulation rate (for modeled
plant species)
• elimination rate (for modeled
animal species)
Source data must be specified for each source
modeled in the scenario. Conceptually, there is no
limit on the number of different sources that can be
modeled. The adjacent text box lists the variables that
must be defined for each source.
5.4.4 ENVIRONMENTAL SETTING DATA
SOURCE INPUTS
source location
emission height
emission rate (for each chemical)
particle size (for each chemical)
gas or liquid (for each chemical)
Whereas initial conditions and source data specify the location and influx of chemical
mass in the system, environmental setting data are needed to estimate the fate and transport of
NOVEMBER 1999
5-11
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
that mass throughout the modeled system. There are two general types of data necessary to
define the links between compartments in TRIM.FaTE: meteorological data and other
environmental setting data. The level of desired refinement in the simulation dictates the
appropriate data (i.e., ranging from site-specific data to default values). Each general type of
input data is briefly described below.
5.4.4.1 Meteorological Data
Meteorological data provide the input data necessary
for many of the transport-related algorithms. For example,
the advection algorithms in air rely on wind data, the
deposition algorithms in air rely on precipitation data, and
the erosion algorithms rely on precipitation data.
Meteorological data can be entered as point estimates or
distributions, depending on the goal of the analysis and the
availability of data. Concurrent, meteorological data are
needed for each calculation time step. Preprocessors should
be used to convert the meteorological data to the time increments
time step. The meteorological inputs needed for TRIM.FaTE are
5.4.4.2 Other Environmental Setting Data
METEOROLOGICAL INPUTS
• horizontal wind speed
• horizontal wind direction
• air temperature
• precipitation
• frost date
• mixing height
• atmospheric stability class
• day/night
equivalent to the calculation
listed in the adjacent text box.
Other environmental setting data define
the characteristics of the biotic and abiotic
compartment types that are needed to estimate
the transport and transformation of chemical
mass in the system. These data may depend on
the characteristics of one or both compartments
in a link. For example, atmospheric dust load
only depends on the characteristics of the air
compartment, whereas the erosion flow rate
between a soil compartment and a water
compartment may depend on the characteristics
of both the soil and water. Input data can be
entered as point estimates or distributions,
depending on the goal of the analysis and the
availability of data. The adjacent text box
presents some examples of both biotic and
abiotic environmental setting data that may be
necessary for a TRIM.FaTE simulation.
5.4.5 OUTPUT TIME STEPS
The final input necessary to begin a TRIM.FaTE simulation is the output time step(s).
This determines the interval at which the mass and/or concentration in each compartment will be
reported as an output. Post-processors may be used to aggregate these results over longer
ILLUSTRATIVE EXAMPLES OF ABIOTIC
ENVIRONMENTAL SETTING DATA
atmospheric dust load (for air
compartment type)
soil density (for all soil compartment types)
current flow velocity (for surface water
compartment type)
ILLUSTRATIVE EXAMPLES OF BIOTIC
ENVIRONMENTAL SETTING DATA
population per area (for all animal
compartment types)
biomass per area (for all plant
compartment types)
food ingestion rate (for all animal
compartment types)
NOVEMBER 1999
5-12
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 5
APPLICATION OF TRIM.FATE
averaging periods. For example, the results using an output time step of one hour may be
averaged to produce the daily (and/or monthly and/or annual) concentrations and/or mass of the
pollutant in each compartment.
5.5 SIMULATION IMPLEMENTATION AND ANALYSIS OF RESULTS
The next step is the actual running of the model, where the movement of the chemical(s)
through the compartments is simulated for each calculation time step for the specified simulation
period. The exact manner in which this is performed depends on the algorithms selected. For
each link, a call is made to the algorithm library to determine the transfer factors that indicate the
potential exchange of chemical mass. If all algorithms involve only first-order processes, then
movement of the chemical will be simulated with a system of linked differential equations, the
solution of which would be found using a differential equation solver (e.g., LSODE). For more
complicated algorithms, other tools would be necessary (e.g., a method of solving partial
differential equations).
The basic TRIM.FaTE outputs are
described in the adjacent text box. The
concentration estimates in the biotic
compartments can be used to estimate
ecological risks (see Section 2.3.6). The
concentration estimates in the abiotic and
biotic compartments can be output to an
exposure model (e.g.. TRIM.Expo) to
estimate human exposure.
After the completion of a simulation. Performance of the model
TYPES OF OUTPUTS
TRIM.FaTE provides several different types of
output to a user The main TRIM.FaTE output
is the mass and concentration in each
compartment at each output time step.
TRIM.FaTE can also output all algorithms
used, all input values, and transfer factors for
each transfer of mass, as well as certain
intermediate calculated values, such as fluxes,
that can be used for evaluating the
the user must interpret the model output. ^^^^^^^^^^^^^^^^^^^^^^^^M
This can be a daunting task because of the
quantity of output data TRIM.FaTE produces. For example, for an analysis that models the fate
and transport of three chemicals in 30 compartments for 30 years, with a simulation time step of
one hour, the model would produce over 23 million mass/concentration values (3 x 30 x 30 x
8,760). If the user wanted to also examine the intermediate model calculations, the output data
set could grow even larger. Because output data from a multimedia fate and transport model can
be used in many ways, such as permitting and development of regulations, different users will
have different needs for the model's output. Automated post-processors can be used to present
the output in forms that are useful to the decision-makers, such as the maximum concentration in
the modeling region or in specific compartments, the average concentration in an environmental
medium or species of wildlife, and long-term time trends of environmental concentrations.
NOVEMBER 1999 5-13 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
[This page intentionally left blank.]
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
6. TREATMENT OF UNCERTAINTY AND VARIABILITY
This chapter summarizes the approach for assessing uncertainty and variability in
TRIM.FaTE, which follows the general approach for TRIM as described in Chapter 3 of the
TRIM Status Report. Additional background on how this method was selected is provided in
Appendix B of the TRIM Status Report. The following text box presents definitions for the key
terms used in this chapter to explain the uncertainty and variability analysis framework for
TRIM.FaTE.
KEY TERMS FOR UNCERTAINTY AND VARIABILITY ANALYSIS
Variability
Variability represents the diversity or heterogeneity in a population or parameter, and is sometimes
referred to as natural variability. An example is the variation in the heights of people Variability
cannot be reduced by using more measurements or measurements with increased precision (taking
more precise measurements of people's heights does not reduce the natural variation in heights)
However, it can often be reduced by a more detailed model formulation (e.g , modeling peoples'
heights in terms of age will reduce the unexplained variability of heights).
Uncertainty
Uncertainty refers to the lack of knowledge regarding the actual values of model input variables
(parameter uncertainty) and of physical systems (model uncertainty). For example, parameter
uncertainty results when non-representative sampling (to measure the distribution of parameter
values) gives sampling errors. Model uncertainty results from simplification of complex physical
systems Uncertainty can be reduced through improved measurements and improved model
formulation
Sensitivity analysis
Sensitivity analyses assess the effect of changes in individual model input parameters on model
predictions This is usually done by varying one parameter at a time and recording the associated
changes in model response. One primary objective of a sensitivity analysis is to rank the input
parameters on the basis of their influence on or contribution to the variability in the model output.
Uncertainty analysis
Uncertainty analysis involves the propagation of uncertainties and natural variability in a model's
inputs to calculate the uncertainty and variability in the model outputs. It can also involve an analysis
of the uncertainties resulting from model formulation. The contributions of the uncertainty and
variability of each model input to the uncertainty and variability of the model predictions are explicitly
quantified
The EPA chose a staged approach for analysis of uncertainty and variability. The use of a
staged approach has advantages for models as complex as TRIM. The first stage consists of
analyses that are comparatively easy to implement, identifying influential parameters and giving
an importance-ranking of parameters, which are useful for narrowing down the number of
parameters to be analyzed in the uncertainty and variability analysis. This first stage is
considered a sensitivity and screening analysis. The second stage involves uncertainty and
NOVEMBER 1999
6-1
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
variability analyses of increasing detail and complexity. Figure 6-1 illustrates this staged
approach for the TRIM.FaTE module and how the functional parts fit together. The first tests for
both stages of analysis have been conducted using a TRIM.FaTE test case based on an actual site.
6.1 SENSITIVITY AND SCREENING ANALYSES
Sensitivity and screening analyses comprise the first stage of the overall analysis of
uncertainty and variability of the TRIM.FaTE model. The sensitivity analysis provides a
quantitative characterization of the sensitivity of the model results to variations in the model
input parameters. The screening analysis is essentially a ranking of sensitivity results. The
purpose of the screening analysis is to make a first-order determination of the most influential
parameters, those that will need to be included in the detailed uncertainty analysis.
Several simulations of the TRIM.FaTE model are performed, with the parameters being
varied singly and in pairs, and the model results are summarized to show the sensitivity to
parameters and to identify the most influential parameters. Nominal values for all parameters
need to be specified to provide a base model run for which the sensitivity analysis is performed.
The results of a sensitivity analysis are applicable to a particular location and for the
range of conditions simulated, and may not apply to conditions outside of this. To broaden the
applicability of the sensitivity analysis, the sensitivity analysis can be performed for a number of
different "nominal" base simulations representing distinct modeling regimes (for example,
summer and winter, or Maine and Louisiana locations).
In addition to the base simulation with parameters at nominal values, sensitivity
simulations are performed with each individual parameter varied by Ap, keeping all other
parameters at their nominal values, where Ap is a small fixed percentage (e.g., one to ten percent)
of the nominal parameter value, or where Ap is a small fixed percentage of a measure of the
spread of values the parameter typically addresses. One can use the standard deviation or a range
of percentiles, for example, the range from the 1st to the 99th percentile. A simulation for each
parameter is required for this locally linear analysis. Thus, 2,000 simulations are needed to
examine 2,000 parameters.
Varying parameters by ±Ap instead of+Ap doubles the number of simulations, but allows
one to calculate the local non-linearity of the effect of varying a parameter on the model results.
These are reported as second order terms in the sensitivity measures to show the extent of local
nonlinearity for parameters. Non-local non-linearities are quantified by increasing Ap to be in
the range of 10 to 100 percent of the nominal values or spread of the parameters.
NOVEMBER 1999 6-2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
Figure 6-1
Uncertainty and Variability Analysis Framework
(Illustrated for TRIM.FaTE Module)
Measurements
Bootstrap
I Expert Elicitation
Input j
Distributions I
STAGE 1
TRIM FaTE j
; Sensitivity Simulations!
Sensitivity and
Screening Analysis
Sensitivity
Elasticity
Sensitivity Score
Analysis of Sensitivity Results
• Selection of
Influential Parameters
STAGE 2
TRIM.FaTE
Monte Carlo Simulations
Process Simulation Results
Rank Correlation
Measures
Distributions
i
Analysis of Results
Note: Results from Stage 2 Monte Carlo analyses can
be used to support other analyses, including response
surfaces, classification and regression trees, generalized
regression analyses, and combinatorial analysis/
probability trees.
NOVEMBER 1999
6-3
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
Sensitivity analyses of pairs and other combinations of parameters can also be performed.
The purpose of varying parameters in pairs (or other combinations) is to identify synergistic
sensitivity effects, where the joint variation of parameters is more (or less) influential than the
combined influence of varying the parameters individually. An additional simulation for each
pair of parameters is required to do this, which greatly increases the number of simulations
performed. On the order of two million additional simulations would be needed to conduct an
all-pairs analysis for 2,000 parameters. However, if there is prior information or knowledge
about which parameters may be synergistic, these parameters should be jointly varied.
The results of these simulations are processed to produce measures of the importance of
the parameters in the sense of how the model results change when the parameters are changed.
The measures of parameter sensitivity and ranking identified to be computed for screening
analyses are the sensitivity, the nominal range sensitivity, the elasticity, and the sensitivity score.
We define these measures following Morgan and Henrion (1990).
The sensitivity of a model output to a parameter is the rate of change of the output with
respect to changes in the parameter. Denoting the parameter as p and the model output as y, the
sensitivity (at a particular value p° of p) is conventionally defined as the partial derivative dy/dp,
evaluated at p°. This measure describes how the model responds to small changes in the
parameter p for values of p that are close to p°, and is referred to as a "local" measure.
We calculate the sensitivity by:
Sensitivity = ^-—=^ ^—!- = -2- (14)
where Ap is a small change in the parameter value and:
Ay = y(p0+Ap)-y(P0) (15)
The nominal range sensitivity is used to assess changes in the model outputs resulting
from large variations in input parameters. The effects on model outputs of varying each input
parameter from the low end to the high end of the range of values for the parameter, are
calculated in essentially the same way as the (local) sensitivity:
Nominal Range Sensitivity = hlgh - ^ (16)
low
NOVEMBER 1999 6-4 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
i
The sensitivity can be interpreted as the slope of the tangent to the response surface y(p)
at the point p° (Figure 6-2). Note that the calculated value of the sensitivity depends both on the
nominal parameter value p° and the amount of change Ap. As long as Ap is small, and the model
is well behaved (bounded derivatives) in the local neighborhood of p°, the effect of changing Ap
should be very small. However, the sensitivity to a parameter can be quite different at different
values p° of the parameter. It can be useful to vary both of these to see how the sensitivity
depends on them.
Figure 6-2
Illustration of Sensitivity in One Dimension
y(p)
Lower
sensitivity
\
Higher
sensitivity
The elasticity is defined as the ratio of the relative change in the model output y to a
specified relative change in a parameter p.
Elasticity = / *
y° P°
(17)
where Ap/p° is a fixed relative change. For example, if the specified parameter change is one
percent (Ap/p° = 0.01), then the elasticity is the percent change in y due to a one percent change
in the parameter p, evaluated at a particular value p° of p.
The sensitivity score is the elasticity weighted by a normalized measure of the variability
of the parameter which takes the form of a normalized range or normalized standard deviation of
NOVEMBER 1999
6-5
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
the parameter. The sensitivity score for the model input parameter p with respect to the model
output y is defined as:
Sensitivity Score = — • — • £—
*P H y°
where:
Ay/Ap = change in output y per change in input p
ol\n = coefficient of variation of p (standard deviation/mean)
p°/y° = ratio of nominal values of the input and output
Other normalized measures of the variation of the parameter can be used in place of the
coefficient of variation, e.g., the range of p divided by the mean.
The sensitivity score was used with an early TRIM.FaTE Prototype using one percent
relative changes, for each input parameter, running the model, and then weighting the Ay's by
estimated standard deviations of the parameters to obtain normalized changes (U.S. EPA
1 998d). The analysis for that particular scenario identified 20 parameters with relatively large
sensitivity scores, out of 400 input parameters. Note that sensitivity analyses are scenario-
specific and parameters identified as influential for one scenario can be different for another
scenario.
6.2 THE MONTE CARLO APPROACH FOR UNCERTAINTY AND
VARIABILITY ANALYSES
A Monte Carlo approach with Latin Hypercube Sampling (LHS) was selected to be the
core method for characterizing and analyzing the uncertainty and variability of the TRIM.FaTE
outputs, with respect to the model inputs and parameters. The primary advantages of Monte
Carlo methods for this type of analysis are the generality with which they can be applied, the lack
of assumptions required, and their computational efficiency. Particular strengths of a Monte
Carlo approach relevant to TRIM uncertainty and variability analyses include the following:
• Monte Carlo (MC) can be used to analyze many parameters.
• MC handles different ways of specifying parameter distributions.
• MC treats correlations and dependencies.
• MC allows for tracking the propagation of uncertainty and variability through model
components at any level.
MC gives estimates of confidence bounds for the estimates of the output distributions.
MC allows precision to be increased easily by performing additional iterations.
NOVEMBER 1999 6^6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
• LHS is an efficient sampling scheme, reducing the number of simulations required. (MC
with LHS has computational complexity linear with the number of parameters or model
inputs that are being analyzed.)
MC handles complex algorithms in the model without increased difficulty.
MC is flexible and will accommodate future additional analyses without major
restructuring.
MC output is compatible with a number of methods for specific analyses of uncertainty
and variability, including response surfaces, regression models, classification and
regression trees (CART), ranking methods, and combinatorial analysis.
• MC is widely used, is generally accepted in the scientific community, and can be
explained to a lay audience.
A significant limitation results from the fact that the analysis of uncertainty and
variability requires estimates of the distributions of parameters, reflecting both the uncertainty
and the variability of the parameters. Of lesser importance, estimates of dependencies
(correlations) between parameters would enable a more detailed analysis to be performed.
However, information on the distribution of parameters is unavailable for most parameters.
When a parameter distribution has been developed, it is rarely separated into components of
uncertainty and variability. This limitation of the Monte Carlo approach can be addressed over
time by developing distributions for parameters the model is most sensitive to. Distributions are
not needed for all parameters.
The implementation of this approach for uncertainty and variability analysis is integrated
with the TRIM.FaTE model to some extent, as opposed to operating as a separate shell around
the model. TRIM.FaTE handles some of the iterations internally, and passes information to the
uncertainty system during a simulation. This close interfacing of the uncertainty software with
the model allows for greater flexibility in terms of what can be tracked and also reduces the
overall processing time.
6.2.1 TWO-STAGE MONTE CARLO DESIGN
Joint uncertainty and variability Monte Carlo simulations are generated based on
sampling from an uncertainty distribution and a variability distribution for each parameter, with
the uncertainty distributions sampled in an outer loop and the variability distributions sampled in
an inner loop. For each uncertainty realization (outer loop sample) there is a specified
distribution of variability (for each parameter) from which several samples are drawn to represent
variability in the inner loop. These several samples represent one variability realization. Figure
6-3 illustrates the structure of this two-stage Monte Carlo design.
As an example, suppose there are Nu samples drawn from the uncertainty distributions,
and that for each uncertainty sample there are Nv variability samples. The cumulative
NOVEMBER 1999 6-7 TRIM.FATE TSD VOLUME I (DRAFT)
-------
Ss
cs
c
c.
tu
j
Q
H
H
U.
2
OS
UJ
00
«
u
O
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
distribution function (of a model output) representing variability for that uncertainty sample can
be estimated from these Nv variability samples and statistics can be calculated (mean, percentiles,
variance, etc.). For each of these statistics, we have Nu values, corresponding to the Nu
uncertainty samples. These then are used to calculate a cumulative distribution function for each
statistic, representing the uncertainty distribution for that statistic.
6.2.2 DISTRIBUTIONS OF INPUT PARAMETERS
The Monte Carlo approach requires specification of probability distributions for each
parameter being analyzed for its role in the overall uncertainty of the model. The distributions
can be specified in any of several standard ways, including sets of data points (discrete
distributions), nonparametric probability distribution functions (PDFs) or cumulative distribution
functions (CDFs), and parametric PDFs or CDFs (analytic functions). For discrete distributions
some additional information about the underlying distribution needs to be provided so that
appropriate samples are selected from the distribution (e.g., whether values between data points
are realizable).
Distributions for parameter variability and for parameter uncertainty are required for
those parameters to be analyzed; we do not use "default" distributions where there is no
information. Parameters without any specification of distributions are treated as if they are
known exactly.
6.2.3 LATIN HYPERCUBE SAMPLING
There are four sampling techniques that are widely used in Monte Carlo methods for
generating random samples from parameter distributions: simple random sampling, Latin
hypercube sampling (LHS), midpoint LHS, and importance sampling. Randomness is an
important feature of these methods for sampling, since it allows one to directly estimate the
precision of the statistics estimated using the Monte Carlo approach.
The sampling technique selected for TRIM is LHS, which employs a stratified random
sampling without replacement scheme, which is very efficient for sampling, especially for
multiparameter models (Iman and Shortencarier 1984, Iman and Helton 1987). Importance
sampling strategies also will be used in conjunction with LHS to obtain better coverage of
distribution tails or extreme values. The strata for LHS are chosen to be intervals partitioning the
range of each parameter, in such a way that the parameter has equal probability of realization
within each interval. Then a sample is selected randomly from each of the intervals. To
illustrate this, say there are k intervals used for each parameter. A random sample is selected
from within each interval, and this is repeated for each parameter, yielding k samples for each
parameter. Then k multivariate samples are constructed by randomly pairing up the samples for
each parameter. These k sets of parameter values (each set containing a value for each
parameter) are referred to as the Latin hypercube sample.
If there are correlations among the parameters, there is a technique for sampling within
the LHS framework so that the sample reflects the correlations (Iman and Conover 1982, Iman et
al. 1985). This treatment of correlation is based on rank-order correlation (Kendall and Gibbons
NOVEMBER 1999 6^9 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
1990) and has desirable properties. It can be used with any distribution and with any sampling
scheme, and it does not change the marginal distributions of the parameters.
6.2.4 TREATMENT OF TAILS OF DISTRIBUTIONS
As noted above, for certain influential parameters an importance sampling technique will
be incorporated to obtain adequate sampling coverage of extreme values of these parameters.
Importance sampling refers to a class of sampling techniques that takes into account the areas of
a distribution that are important to the analysis, providing enhanced detail in these areas.
Importance sampling is often used when increased accuracy in one or both tails of a distribution
is desired.
6.2.5 TRACKING INFORMATION BETWEEN MODULES
There are two levels at which tracking of information related to uncertainty analysis
occurs; the first is from one TRIM module to the next, and the second is within each TRIM
module.
The information passed from one TRIM module to the next (e.g., from TRIM.FaTE to
TRIM.Expo) needs to provide enough detail to allow for continuation of the Monte Carlo
propagation of uncertainty and variability in the next module. Information on the joint
distributions of a TRIM module's inputs and outputs is required to do this, for both uncertainty
and variability.
There is some flexibility as to how this information is summarized or condensed. For
example, each TRIM module could pass the complete distributions for all of the output variables
and parameters tracked in the uncertainty analysis. At the other end of the spectrum, the module
could pass only the means, variances, and correlations of the distributions, for a selected subset
of the parameters and outputs.
At this time, an intermediate approach has been implemented for passing information
from TRIM.FaTE to TRIM.Expo, which will be revised if necessary. The distributional
information is somewhat condensed, while passing enough information about the uncertainty and
variability distributions to allow them to be adequately sampled for the Monte Carlo analysis
within the next module. For the uncertainty and variability distributions of each parameter and
each output, TRIM.FaTE calculates and passes 101 percentiles (minimum, maximum, and every
percent) as well as some additional percentiles at the tails (how many is to be determined).
Features of the joint distributional structure are summarized in a correlation matrix of
correlations between all pairs of parameters and outputs. Although this is a strong condensation
of the joint distributional structure, it is likely to be adequate relative to the lack of information
that is likely to be available to estimate parameter correlations in the first place.
6.2.6 COMPUTATIONAL RESOURCES
Although the Monte Carlo technique is very efficient, Monte Carlo simulations of
TRIM.FaTE require substantial computer processing time, especially when treating more than a
NOVEMBER 1999 6^10 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
few parameters. The available computational resources can be a limiting factor in the scope of
the analysis performed. The more detailed analyses may have to restrict their scope to small
numbers of parameters being jointly varied, for example.
Computer processing time for both the uncertainty propagation and tracking and the
TRIM.FaTE model depends on the definition of the TRIM.FaTE modeling scenario, in terms of
the numbers of compartments, time steps, length of simulation, chemicals, and so forth. It also
depends on the number of parameters and number of model outputs analyzed, the sizes of the
Monte Carlo samples (which relates to the number of simulations), and the level of detail of the
analysis.
6.2.7 SPATIAL AND TEMPORAL RESOLUTION AND AGGREGATION
Estimation of the effects of spatial and temporal aggregation on uncertainty and
variability will be accomplished by sensitivity analyses of Monte Carlo results. For analysis of
spatial aggregation, EPA will set up a small number (three to five) of TRIM.FaTE scenarios with
increasing levels of spatial resolution (decreasing levels of aggregation), and run the same set of
simple Monte Carlo simulations for each scenario. Comparison of the Monte Carlo results for
the scenarios will show the impact of the aggregation on uncertainty and variability for the
scenarios modeled. Similarly, the effects on model output uncertainty of temporal aggregation
will be assessed by comparing uncertainty results from scenarios with and without seasonal
aggregation. Initial efforts have included evaluation of the effects of explicitly modeling
seasonal variability, and treatment of other temporal scales can be evaluated similarly.
6.2.8 SPECIFICATION OF PROBABILITY DISTRIBUTIONS AND CORRELATIONS
OF MODEL INPUTS
The need for distributions for the input parameters is discussed above. Implementation of
this Monte Carlo approach employs a data file that specifies the distributions of uncertainty and
variability for each parameter. For each parameter this file contains the distribution name (e.g.,
lognormal) and the parameters or data that complete the specification of the distribution. There
needs to be a set of distributions for variability and a set of distributions for uncertainty for each
parameter.
The Monte Carlo method can also handle the joint distributions of the dependent
parameters. However, information to estimate full joint distributions is not presently available.
In addition to the marginal distributions, the current implementation only requires the correlation
structure of the set of parameters, which is specified by (N2-N)/2 estimates of pairwise rank
correlations for the set of N parameters.
There are often physical constraints on values of parameters and intermediate quantities
in the model; for example, mass is always non-negative. These can have implications for how
parameter distributions are set. The specified distributional forms should satisfy the physical
constraints as well as reflect the distributions indicated by the available data.
NOVEMBER 1999 6-11 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
6.3 PRESENTATION OF UNCERTAINTY RESULTS
When a model has many inputs and is complex, as TRIM.FaTE is, the analyst will make
use of methods that are simple and give a first-order picture of uncertainty, as well as more
complex methods giving a more refined, detailed analysis of uncertainty. There are several ways
to form summary measures and present the uncertainty and variability of a modeling system.
Loosely speaking, "measures" are low-dimensional statistics (one or a small number of
descriptive statistics), for example, the sensitivity score or the 10th, 50th, and 90th percentiles of a
distribution. In addition to measures, ways of presenting the results include graphs of
distributions, tree diagrams, other graphs, and tables of statistics.
In presenting results, the objectives of the uncertainty and variability analysis should be
clearly stated, and it should be shown how the objectives were met (or not met). It is best to use
a reasonable number of significant digits when reporting results, lest the audience be misled as to
the accuracy of the results (the use of fewer digits also improves readability). The key results
should be presented clearly and not obscured by concomitantly reporting several less-
comprehensible statistics. If it is necessary to report extensive statistics, they can be documented
in an appendix. Tables and figures need to be labeled clearly and completely, so they can more
or less stand alone, reducing misleading impressions if seen out of the context of a report.
The uncertainty results from TRIM.FaTE will be presented in a tiered style, with three
tiers of increasing detail and complexity. The first tier is designed to be meaningful to a wide
audience, including the public, decision-makers, and risk analysts, and will present the main
findings in easily understandable charts, tables, and descriptive text.
The first tier results will be derived from the more comprehensive second tier results.
The second tier will present more detailed results requiring the user to have some familiarity' with
risk analysis statistics, such as probability distribution functions (PDFs), cumulative distribution
functions (CDFs), and graphs of results. For example, a tier 1 table might present the probability
that a predicted risk exceeds a given risk cutoff level, for three cutoff values. A corresponding
tier 2 graph would plot the curve of exceedance probabilities for a range of cutoff values. To
assist in interpretation of PDFs and CDFs, the PDF and CDF of a distribution (or family of
distributions) will always be presented as a pair of graphs, one above the other, with the same
horizontal scales. The mean of the distribution as well as other relevant points (e.g., the 95th
percentile; the deterministically predicted point value) will be indicated on both graphs. The
second tier will also include an overview of the distributions of uncertainty and variability of the
input parameters and an explanation of dependencies and correlations of both the input
parameters and the model results.
The third presentation tier will have the most detailed graphs and tables, and will usually
be referred to for details supporting specific pieces of the analysis. These results can be lengthy,
and will be relegated to an appendix. Also part of the third tier, another appendix will be
prepared which documents the probability distribution of each input parameters. This document
will include a discussion of the data used to estimate the distributions, how much data were used,
representativeness of the data, whether the distributions characterize uncertainty, variability or
NOVEMBER 1999 6-12 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
(both, and how the distributions were estimated from the data. Graphs comparing the fitted
distributions and measured data will be presented along with goodness of fit statistics.
There are a number of first-order measures of the importance, or influence on modeled
results, of model input parameters, which are quite straightforward to compute. These
sensitivity/screening measures (sensitivity, sensitivity score, elasticity) are important in the
context of selecting parameters for detailed analysis of uncertainty and variability, and are
described above in Section 6.1. In addition to these, the results of the TRIM.FaTE uncertainty
and variability analysis are described using distributions, dependencies, and joint distributions
and confidence bounds. Specifically, the core presentation will include the following:
Measures. Graphs, and Tables
• Sensitivity
• Sensitivity score
• Elasticity
• Probability density functions
• Cumulative distribution functions
• Confidence intervals
• Tables of statistics
Rank order correlation
• Ranking groups of parameters
• Correlation matrix
• Scatter plots, scatter plot matrix
These items are described in the remainder of this section.
Probability density functions can be depicted by smoothed histograms. Care must be
taken when using histograms in order to avoid inaccurate representations of peaks and valleys
(caused by too tight a spline fit on histograms with too many bins), to keep the area under the
distribution equal to 1.0, and to avoid over-smoothing, which suppresses features of the
distribution.
Cumulative distribution function graphs lack the intuitive appeal of graphs of the
probability distribution function, but don't have the difficulties of smoothing that probability
distribution function graphs do. The cumulative distribution function can be generated directly
from the data points. Probabilities of exceeding given values or of being within a given range
can be read off graphs of the cumulative distribution function or the probability distribution
function. Overlays of distributions can be informative using either of these graphs.
Confidence bands and vertical confidence intervals on graphs of probability distribution
functions and cumulative distribution functions provide information about the uncertainty of the
distributions. Vertical confidence intervals ("whiskers") tend to be easier to interpret visually
than bands. These graphs are effective for presenting uncertainty and variability together, where
the distribution of variability is graphed with bands or intervals representing the spread of the
uncertainty distributions.
NOVEMBER 1999 6^13 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
Tables of selected statistics provide another useful way to present summaries of
distributions. Descriptive statistics such as percentiles, means, standard deviations, and
coefficients of variation can concisely describe several distributions in a single table. Table 6-1
shows how six statistics summarizing a distribution can be presented to describe a family of
distributions.
Table 6-1
Example Table Summarizing Distributions
Stratification
Variable
Location 1
Location 2
Location 3
Variability of the Concentration of Arsenic in Soil (pg/m1)
Mean
1 *%ile
10th %ile
Median
90th %ile
99th %ile
Tables presenting uncertainty results in relation to cutoff values of concentrations,
exposure, or risk are useful to decision-makers using the results of TRIM modeling. These tables
aim to present in more directly meaningful ways the information presented in PDFs and CDFs.
which can be difficult to interpret if one is not familiar with their use. For example, the
likelihood that a pollutant concentration in a media or specific location is greater than a fixed
concentration cutoff, with a given level of confidence, can be tabulated for a series of cutoff
values (Table 6-2). This type of table will be presented taking into account only the variability of
input parameters, and also taking into account both variability and uncertainty of the parameters.
Another type of table, useful for assessing data requirements, tabulates the effectiveness of
reducing the uncertainty in parameter values towards reducing the uncertainty in model results to
acceptable values. Different users of the model and different types of decisions influenced by the
model results will have differing requirements as to how much uncertainty is acceptable.
Combinations of parameter uncertainty reductions which result in bringing model uncertainty
down to a specified level can be tabulated. If the cost of collecting additional data is taken into
account, then least-cost options addressing this can be presented.
NOVEMBER 1999
6-14
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
Table 6-2
Likelihood that the Concentration of Lead in Catfish Exceeds Concentration Cutoff Values
The
concentration cutoff
value Yc
0.1
1.0
10.0
100.0
Probability the concentration
will exceed Yc (%)
12
0.013
4.1X1Q-6
2x1 0-7
Correlation measures such as Spearman's rank order correlation, Kendall's tau-b, and
Pearson's product-moment correlation of model outputs with inputs are useful for identifying
influential parameters. Spearman's rank order correlation and Kendall's tau-b are based on
ranks. Rank correlation methods have the advantages of robustness to outliers and availability of
distribution-free tests. A good account of rank order correlation methods is given by Kendall and
Gibbons (1990). The Pearson correlation makes the restrictive assumption that the model
input/output relationships are linear (rank order correlation does not), but it does provide
information about how uncertainty in a parameter contributes linearly to uncertainty in the
outputs.
Correlations for several parameters
can be graphed as a tornado chart, where
the x-axis is the correlation (between
parameters and a specific model output
variable) and the y-axis is categorical,
listing the parameters (Figure 6-4).
Horizontal bars extend out from the y-axis,
with the length of the bars equal to the
correlation. The y-axis parameters are
ordered from highest correlation at the top
to lowest at the bottom. It is useful to
include vertical bands indicating where
correlations are not statistically different
from zero (at a specified confidence level
and sample size). Inspection of a tornado
chart can show three classes of parameters:
independent parameters (changing the
parameter within the range does not change
the model results), parameters with small
influence (parameter changes do not affect
the results much), and influential
parameters (results change significantly).
Figure 6-4
Tornado Chart
Pcrcmeter
NOVEMBER 1999
6-15
TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 6
TREATMENT OF UNCERTAINTY AND VARIABILITY
Ranking groups of parameters, in addition to ranking individual parameters, is a way of
ranking that takes into account strong correlations between parameters. Parameters are grouped
that are highly correlated with each other, are influential parameters when looked at individually,
but are not influential when the effects of any other parameter in the group are already accounted
for (small partial correlation). Canonical correlation (Mardia et al. 1979) can be used to identify
the parameter groups.
Correlation matrices provide a way to compactly present the correlations among a set of
variables (inputs, outputs, or both) as a matrix of the correlations between pairs of variables.
Table 6-3 illustrates a correlation matrix for three variables A, B and C. The notation p(A,B)
indicates the correlation between variables A and B. The correlation p(A,B) is the same as
p(B,A), hence the correlation matrix is symmetric and only half of the matrix needs to be filled.
For correlations between parameters and model outputs or between different model outputs, the
Pearson correlations are calculated as well as the rank order correlations.
Table 6-3
Correlation Matrix
Variable
A
B
C
A
1
P(A,B)
P(A,C)
B
1
P(B,C)
C
1
Scatter plots visually illustrate correlations between parameters and also indicate other
features of the relationships between parameters, such as non-linearity, regions with more or less
scatter or correlation, points that drive the correlation, and potential outliers. A scatter plot
matrix for illustrating correlations between parameters is similar to the correlation matrix but
with a scatter plot instead of a number in each "table cell." This is cumbersome for more than a
few parameters and is more effectively used with subsets of parameters that exhibit significant
correlation.
NOVEMBER 1999
6-16
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER 7
REFERENCES
7. REFERENCES
Ambrose, R.A., Jr., T.A. Wool, and J.L. Martin. 1995. The Water Quality Analysis Simulation
Program, WASPS. Part A: Model documentation. Athens, GA: U.S. EPA National Exposure
Research Laboratory, Ecosystems Division.
Brandes, L. J., H. den Hollander, and D. van de Meent. 1997. SimpleBOX v 2.0. Netherlands:
National Institute of Public Health and the Environment (RIVM).
Buck, J.W., G. Whelan, J.G. Droppo, Jr., D.L. Strenge, KJ. Castleton, J.P. McDonald, C. Sato,
and G.P. Streile. 1995. Multimedia Environmental Pollutant Assessment System (MEPAS)
application guidance. Guidelines for evaluating MEPAS input parameters for Version 3. PNL-
10395. Richland, WA: Pacific Northwest Laboratory.
CRARM. 1997. Commission on Risk Assessment and Risk Management. Risk assessment and
risk management in regulatory decision-making. Final report, Volume 2. Washington, DC.
Cohen, Y., W. Tsai, S.L. Chetty, and G.J. Mayer. 1990. Dynamic partitioning of organic
chemicals in regional environments: A multimedia screening-level approach. Environmental
Science and Technology. 24:1549-1558.
Cohen. Y. and P.A. Ryan. 1985. Multimedia modeling of environmental transport:
Trichloroethylene test case. Environmental Science and Technology. 9:412-417.
Cowen. E.C.. D. Mackay, T.C.J. Feihtel. D. van de Meent, A. DiGuardo, J. Davies and N.
Mackay. 1995. The multi-media fate model: A vital tool for predicting the fate of chemicals.
Pensacola. FL: SETAC Press.
Iman. R.L. and J.C. Helton. 1987. An investigation on uncertainty and sensitivity analysis
techniques for computer models. Risk Analysis. 8(1):71.
Iman, R.L., M.J. Shortencarier, and J.D. Johnson. 1985. A FORTRAN 77 program and user's
guide for the calculation of partial correlation and standardized regression coefficients.
SAND83-0044. Albuquerque, NM: Sandia National Laboratories.
Iman, R.L. and M.J. Shortencarier. 1984. A FORTRAN 77 program and user's guide for the
generation of latin hypercube and random samples for use with computer models. NUREG/CR-
3624. SAND83-2365. Albuquerque, NM: Sandia National Laboratories.
Iman, R.L. and W.J. Conover. 1982. A distribution-free approach to inducing rank correlation
among input variables. Communications on Statistics: Simulation and Computing. 11 (3):311 -
334.
NOVEMBER 1999 7-1 TRIM.FATETSD VOLUME I (DRAFT)
-------
CHAPTER 7
REFERENCES
IT. 1997a. IT Corporation. Overview and key features of the Integrated Spatial Multimedia
Compartmental Model (ISMCM). An addendum to evaluation of existing approaches for
assessing non-inhalation exposure and risk with recommendations for implementing TRIM.
Contract No. 68-D-30094, Work Assignment 3-11. Prepared for U.S. Environmental Protection
Agency, April 1997.
IT. 1997b. IT Corporation. Model comparison update for TRIM. Contract No. 68-D-30094,
Work Assignment 4-18, Task 2 Report. Prepared for U.S. Environmental Protection Agency,
November 1997.
Kandall, M. and J.D. Gibbons. 1990. Rank correlation methods. New York, NY: Oxford
University Press.
Mackay, D. 1991. Multimedia environmental models: The fugacity approach. Chelsea, MI:
Lewis Publishers.
Mackay, D. and S. Paterson. 1982. Fugacity revisited. Environmental Science and Technology.
16:654-660.
Mackay. D. and S. Paterson. 1981. Calculating fugacity. Environmental Science and
Technology. 15:1006-1014.
Mackay, D. 1979. Finding fugacity feasible. Environmental Science and Technology. 13:
1218-1223.
Mardia. K.V.. J.T. Kent and J.M. Bibby. 1979. Multivariate analysis. London: Academic
Press.
McKone, T. E. 1993a. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part I: Executive summary. Laboratory.UCRL-CR-111456PtI. Livermore, CA: Lawrence
Livermore National.
McKone, T. E. 1993b. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part II: The dynamic multimedia transport and transformation model. UCRL-CR-111456PtII.
Livermore, CA: Lawrence Livermore National Laboratory.
McKone, T. E. 1993c. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part III: The multiple-pathway exposure model. UCRL-CR-111456PtIII. Livermore, CA:
Lawrence Livermore National.
McKone, T.E. and D.W. Layton. 1986. Screening the potential risk of toxic substances using a
multimedia compartment model: Estimation of human exposure. Regul. Toxicol. Pharmacol.
6:359-380.
Morgan, G.M. and M. Henrion. 1990. Uncertainty: A guide to dealing with uncertainty on
quantitative risk and policy analysis. New York, NY: Cambridge University Press.
NOVEMBER 1999 7^2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER?
REFERENCES
Mozier, J.W. and T.R. Johnson. 1996. Evaluation of existing approaches for assessing
non-inhalation exposure and risk with recommendations for implementing TRIM. 68-D-30094,
Work Assignment 2-9. Prepared by IT Corporation for U.S. Environmental Protection Agency,
April 1996.
NRC. 1994. National Research Council. Science and judgment in risk assessment.
Washington, DC: National Academy Press.
Ozkaynak, H., M.. Zufall, J. Burke, J. Xue, and J. Zidek. 1999. A probabilistic population
exposure model for PM10 and PM2 5. Presented at 9th Conference of the International Society of
Exposure Analysis. Athens, Greece, September 5-8, 1999.
Rice, G., R. Hetes, J. Swartout, Z. Pekar, and D. Layland. 1997. Methods for assessing
population exposures to combustor emissions. Presentation at the 1997 Society for Risk
Analysis Annual Meeting, Washington, DC.
Schneider. H. and G.P. Barker. 1989. Matrices and Linear Algebra. 2nd ed. New York. NY:
Dover Publications, Inc.
Thibodeaux, L. J. 1996. Environmental chemodynamics: Movement of chemicals in air, water,
and Soil. 2nd ed. New York, NY: J. Wiley & Sons.
Thibodeaux. L. J. 1979. Chemodynamics, environmental movement of chemicals in air, water.
and soil. New York. NY: John Wiley and Sons.
U.S. EPA. 1999a. U.S. Environmental Protection Agency. The Total Risk Integrated
Methodology: TRIM.Expo Technical Support Document. External Review Draft. EPA-453/D-
99-001. Research Triangle Park, NC: Office of Air Quality Planning and Standards. November.
U.S. EPA. 1999b. U.S. Environmental Protection Agency. The Total Risk Integrated
Methodology: Status report. Draft. EPA-453/D-99-010. Office of Air Quality Planning and
Standards. November.
U.S. EPA. 1999c. U.S. Environmental Protection Agency. Systems Installation and Operation
Manual for the EPA Third-Generation Air Quality Modeling System (Models-3 Version 3.0).
Atmospheric Modeling Division, National Exposure Research Laboratory, Research Triangle
Park, NC and EPA Systems Development Center (A contractor operated facility), Science
Applications International Corporation, Arlington, VA. June.
U.S. EPA. 1999d. Methodology for assessing health risks associated with multiple pathways of
exposure to combustor emissions. National Center for Environmental Assessment. EPA 600/R-
98/137. Office of Research and Development.
U.S. EPA. 1999e. U.S. Environmental Protection Agency. National Air Toxics Program: The
Integrated Urban Strategy. Federal Register 64: 38705-38740. July 19.
NOVEMBER 1999 7-3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
CHAPTER 7
REFERENCES ___^________
U.S. EPA. 1999f. U.S. Environmental Protection Agency. Residual Risk Report to Congress.
Office of Air Quality Planning and Standards, Research Triangle Park, NC. March.
U.S. EPA. 1998a. U.S. Environmental Protection Agency. The Total Risk Integrated
Methodology: Technical support document for the TRIM.FaTE Module. Draft. EPA-452/D-98-
001. Office of Air Quality Planning and Standards.
U.S. EPA. 1998b. U.S. Environmental Protection Agency. The Total Risk Integrated
Methodology: Implementation of the TRIM conceptual design through the TRIM.FaTE Module.
Status Report. Draft. EPA-452/R-98-001. Office of Air Quality Planning and Standards.
U.S. EPA. 1998c. U.S. Environmental Protection Agency. Advisory on the Total Risk
Integrated Methodology (TRIM). EPA-SAB-EC-ADV-99-003. Science Advisory Board.
U.S. EPA. 1998d. U.S. Environmental Protection Agency. Risk characterization handbook.
Draft. EPA 100-B-98-OOX. Science Policy Council.
U.S. EPA. 1998e. U.S. Environmental Protection Agency. Study of Hazardous Air Pollutants
from Electric Utility Steam Generating Units - Final Report to Congress. EPA 453/R-989-004a.
February.
U.S. EPA. 1998f. U.S. Environmental Protection. Methodology for assessing health risks
associated with multiple exposure pathways to combustor emissions. External Review Draft.
Update to EPA/600/6-90/003. NCEA-C-0238. National Center for Environmental Assessment.
U.S. EPA. 1998g. U.S. Environmental Protection Agency. Human health risk assessment
protocol for hazardous waste combustion facilities. Peer Review Draft. EPA530-D-98-001 A.
Office of Solid Waste and U.S. EPA Region 6, Multimedia Planning and Permitting Division.
U.S. EPA. 1997. U. S. Environmental Protection Agency. Mercury study report to congress
(Volume I - VIII). EPA-452/R-97-005. Office of Air Quality Planning and Standards and Office
of Research and Development.
U.S. EPA. 1994a. U.S. Environmental Protection Agency. Report of the Agency Task Force on
environmental regulatory modeling. Guidance, support needs, draft criteria and charter. EPA
500-R-94-001. Washington, DC: Office of Solid Waste and Emergency Response.
U.S. EPA. 1994b. U.S. Environmental Protection Agency. Review of Draft "Addendum to the
methodology for assessing health risk associated with indirect exposure to combustor emissions".
EPA-SAB-1AQC-94-009B. Washington, DC: Science Advisory Board.
U.S. EPA. 1994c. U.S. Environmental Protection Agency. Estimating exposure to Dioxin-like
compounds. Volume II. Site-specific Assessment Procedures. External Review Draft.
EPA/600/6-88/005Cc.
NOVEMBER 1999 7-4 TRIM.F ATE TSD VOLUME 1 (DRAFT)
-------
CHAPTER?
REFERENCES
U.S. EPA. 1993. U.S. Environmental Protection Agency. Addendum to methodology for
assessing health risks associated with indirect exposure to combustor emissions. External
Review Draft. EPA/600/AP-93/003. Washington DC: Office of Health and Environmental
Assessment.
U.S. EPA. 1990. U.S. Environmental Protection Agency. Methodology for assessing health
risks associated with indirect exposure to combustor emissions. Interim Final. EPA/600/6-
90/003. Washington DC: Office of Health and Environmental Assessment.
U.S. EPA. 1986. U.S. Environmental Protection Agency. Users Manual for the Human
Exposure Model (HEM). Research Triangle Park, NC: Office of Air Quality Planning and
Standards. EPA-540/5-86-001. June.
van de Meent, D. 1993. SIMPLEBOX: A generic multimedia fate evaluation model. Report
No. 672720 001. Bilthoven, Netherlands: National Institute of Public Health and Environmental
Protection (RIVM).
van de Water, R.B. 1995. Modeling the transport and fate of volatile and semi-volatile organics
in a multimedia environment. M.S. Thesis. Los Angeles, CA: University of California.
Whicker, F.W. and T.B. Kirchner. 1987. PATHWAY: A dynamic food-chain model to predict
radionuclide ingestion after fallout deposition. Health Phys. 52:717-737.
NOVEMBER 1999 7-5 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
APPENDIX A
GLOSSARY
APPENDIX A
Glossary
Abiotic Compartment Type
Advective Process
Biotic Compartment Type
Chemical
Compartment
Compartment Type
Diffusive Process
Dispersion
Fugacity
A compartment type consisting primarily of a non-living
environmental medium (e.g., air, soil) for which TRIM.FaTE
calculates chemical masses and concentrations; it may also
contain biota, such as the microorganisms responsible for
chemical transformation (see also compartment type).
A process by which a chemical can be transported within a
given medium that is moving from one compartment to
another.
A compartment type consisting of a population or community
of living organisms (e.g., bald eagle, benthic invertebrate), or in
the case of terrestrial plants, portions of living organisms (e.g.,
stems, leaves), for which TRIM.FaTE calculates chemical
masses and concentrations (see also compartment type).
A unit whose mass is being modeled by TRIM.FaTE. A
chemical can be any element or compound, or even group
of compounds, assuming the necessary parameters (e.g..
molecular weight, diffusion coefficient in air) are defined.
A homogeneous unit of space characterized by its physical
composition within which it is assumed, for modeling
purposes, that all chemical mass is in equilibrium.
A specific kind of compartment, such as an air compartment
type or a mule deer compartment type. Compartment types are
distinguished from each other by the way they exchange
chemical mass with other compartment types.
A process by which a chemical is transported from one
compartment to another as a result of the magnitude and
direction of the concentration differences between two
compartments at the interface between the two locations.
The "spreading out" of a chemical during advective transport.
May result in movement of the chemical perpendicular to the
direction of advective flow.
A measure of the tendency of a substance to escape by some
chemical process from the phase in which it exists.
NOVEMBER 1999
A-l
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX A
GLOSSARY
Link
Model Evaluation
Modeling Region
Output Time Step
Parameter
Parcel
Scenario
Sensitivity
Simulation
Simulation Period
Simulation Time Step
Source
A connection that allows the transfer of chemical mass between
any two compartments. Each link is implemented by an
algorithm or algorithms that mathematically represent the mass
transfer.
The broad range of review, analysis, and testing activities
designed to examine and build consensus about a model's
performance.
The region of space through which the transport and
transformation of the modeled chemical(s) is estimated.
A length of time over which the compartment masses and
concentrations calculated at each simulation time step are
summarized and reported by the model.
A model input that defines a variable in an algorithm (e.g..
emission rate, half-life, biomass).
A planar (i.e., two dimensional) geographical area used to
subdivide a modeling region. Parcels, which can be virtually
any size or shape, are the basis for defining volume elements.
There can be air, land, and surface water parcels.
A specified set of conditions (e.g., spatial, temporal,
environmental, source, chemical) used to define a model setup
for a particular simulation or set of simulations.
The rate of change of the model output with respect to changes
in an input parameter.
A single application of a model to estimate environmental
conditions, based on a given scenario and any initial input
values needed.
The entire length of time for which the model is run and
compartment masses and concentrations are calculated - in
other words, the time period from the beginning of the
simulation until the end.
The time increment at which the model calculates (and re-
calculates iteratively throughout the simulation period) a new
inventory of compartment masses and concentrations.
•
An external component that introduces chemical mass directly
into a compartment.
NOVEMBER 1999
A-2
TRIM.FATE ISO VOLUME 1 (DRAFT)
-------
APPENDIX A
GLOSSARY
Uncertainty The lack of knowledge regarding the actual values of model
input variables (parameter uncertainty) and of physical systems
(model uncertainty).
Variability The diversity or heterogeneity in a population or parameter;
sometimes referred to as natural variability.
Volume Element A bounded three-dimensional space that defines the location of
one or more compartments. This term is introduced to provide
a consistent method for organizing objects that have a natural
spatial relationship.
NOVEMBER 1999 A-3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
APPENDIX B
INTEGRATING EXTERNAL MODELS OR MEASURED DATA INTO TR1M.FATE
APPENDIX B
Integrating External Models or Measured Data into
TRIM.FaTE
At some point, it may be desirable to use either measured data or the output of other
models with the TRIM.FaTE model. In either case, there are two basic forms that the data can
take: (1) fluxes into certain compartments (e.g., deposition rates), or (2) calculated/specified
concentrations or chemical mass in certain compartments. The first case is the simpler of the
two, as it requires only adding additional source terms to the affected compartments. The second
case has more of an impact on the system of equations used in the modeling, and is the focus of
this Appendix.
B.I Compromises that Must be Made in Order to Use Externally Derived
Compartment Concentrations
The incorporation of externally derived compartment concentrations into any multimedia
model dictates that compromises be made with regard to preserving chemical mass balance. The
basic problem is the loss of chemical mass from the compartments that send a chemical to the
compartment for which an externally derived compartment concentration is used. Since the
receiving compartment will not receive the chemical, it is effectively lost from the system. The
only way to avoid this loss of chemical mass is to modify all of the links to the "constant"
compartments so that this exchange does not take place (e.g., disable resuspension from surface
soil to the air domain if the results of an air model are to be used).
While we will know mathematically how much mass has been lost through these
processes, the chemical lost will not be allowed to participate in any further exchanges with other
compartments. Whether this is acceptable or not depends on the attitude of the user, but this
compromise is unavoidable if they are to use externally derived compartment concentrations.
B.2 Implementation Details in the Case of First-order Transfers - Case of
Constant Inputs
In this section, explicit details are provided for incorporating externally derived
compartment concentrations for selected compartments in the case where these concentrations
are constant for the time period of interest. The general case where the values from externally
derived compartment concentrations change with time can be addressed by sequentially using
this method over the different intervals over which the values from the externally derived
compartment concentrations are constant.
NOVEMBER 1999 B-1 TRIM .FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX B
INTEGRATING EXTERNAL MODELS OR MEASURED DATA INTO TR1M.FATE
For a given time interval in which the parameters are constant with time, the system of
differential equations to be solved is:
— = AN + s (1)
dt
where N is the vector of the mass of chemical(s) in each compartment, given by:
N =
«2(0
"„(')
(2)
where n,ft) is the mass of a particular chemical in a compartment/chemical pair, m is the total
number of compartment/chemical pairs; A is the matrix of transfer factors describing transport
between compartment/chemical pairs:
A =
a
n
°
22
a
(3)
and s is th£ vector of sources terms for each chemical in each compartment.
The fact that chemical mass balance is preserved implies that the matrix A satisfies two
basic conditions:
a > 0, if / #j and a,, <0
(4)
fly
(5)
Using externally derived compartment concentrations is equivalent to fixing the
concentration/chemical mass in some compartments. This can be done by solving a reduced
system of differential equations, using constant values for the applicable terms n,(t). This can be
conceptualized as using "virtual sources" for the relevant compartment/chemical pairs, with each
(time-dependent) virtual source set so that the mass of chemical is constant.
For example, if there is only one chemical being considered, and we want the
concentration/mass to be fixed in the first compartment, say n,(t)=Mh then dn,/dt=0, and the
original system becomes:
NOVEMBER 1999
B-2
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX B
INTEGRATING EXTERNAL MODELS OR MEASURED DATA INTO TRIM.FATE
d
dt
M{-
«2(0
«„(')
_
a,, a,2 fl,3 ... a,w
a21 a22 a23 ... a2m
aml °m2 Qm3 "mm
M}'
«2(r)
V').
+
*i
*2
*•».
(6)
.Since the derivative of a constant is zero, examining the first row of the above system shows that:
o =
(7)
i.e., the virtual source s,(t) in the first compartment is given by:
(8)
= -M*n -2. «,(')«„
The terms «/r> for i>l can be determined by solving the system of differential equations
obtained by eliminating the first row, and using n,(t)=M:
d_
dt
2m
M
n2(t)
n (0
m^ -^
(9)
m2 m3 '
n2(t)
This system of differential equations is of the same form as the original equation, and can
be solved using the same solver used for the original equation. However, it can also be rewritten
as a system of the same size as the original system by adding the differential equation dn,/dt=0,
n,(0)=M', this results in the system (with initial condition):
NOVEMBER 1999
B-3
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX B
INTEGRATING EXTERNAL MODELS OR MEASURED DATA INTO TRIM.FATE
d
dt
«,(>)'
*2(r)
v>
=
00 0 ... 0
0 a22 a23 ... a2m
0 °m2 "ml - Qmm
«,(0'
«2(0
v>
+
0
s2 + Ma2)
^ + ^v
5
«,(0)
«2(0)
"JO)
=
A/
»2(0)
w (0)
mv '
(10)
Note that the mass lost from the system to the compartments which are to be held
constant is accounted for, as condition (5) is still satisfied for the diagonal elements of the matrix
in equation (10), where the coefficients flyare used in the sum.
When more than one of the «,'s is constant, this same method can be used. In general, if
the kih compartment/chemical pair is to be constant (say with value Mk), then one puts zeros in
the kth row and klh column, and adds the term Mk ank to the «th row of the source term vector.
This is done for every compartment/chemical pair which is to be constant. For example, applied
to equation (2), one would obtain the following:
d
dt
«,(')'
n2(t)
"*-!<')
"*(')
"*-,(')
.»-<')
000 000 0
o *:: fla ,„., o «„., *2m
0 fl, , -. Q, , -~ a, , , , 0 Q, i , , fli ,
A 1 _ Al-' A - 1,A - 1 A"],*-] A-lm
000 000 0
0 ai-\i °<-n °*-u-i ° a*-i.*-i a*-i.m
o «.., ^ «.,., o ^,, Qmm
»,(') '
«2(0
"*-](')
"*(')
«*..(')
.„(/,
«,(0)'
«2(0)
"*-.(«)
«(0)
=
A/
14(0)
«*-,(»)
"JO).
(11)
The method described in this section fits seamlessly into the general process of
calculating transition matrices and source terms prior to calling LSODE. In general, one uses all
zeros for any row that is to be constant, adds the extra flux terms to the source term vector, and
calls LSODE as is done normally.
NOVEMBER 1999
B-4
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL
RESOLUTION
This appendix is an example of a detailed approach for defining the level of spatial
complexity (i.e., location, size, and number of parcels) in a TRIM.FaTE analysis. There are
many factors that influence scenario complexity, including the characteristics of the pollutant, the
environmental setting, the exposed population, the impact of interest, the available data, and
available computer resources. A clear analysis objective offers a starting point for setting up the
scenario with an appropriate level of complexity. The TRIM.FaTE modeling framework uses a
system where the analysis objective is classified into one of three basic types. For each type of
objective, a series of questions can be used to identify the natural and artificial boundaries of the
system. A preliminary decision tree is developed for each objective type to assist the user in
determining which boundary or set of boundaries is appropriate for a given modeling objective.
The decision trees are used to provide a standard approach for setting up a simulation.
After the initial scenario is constructed and a simulation has been completed, the
preliminary results need to be evaluated to confirm that the most appropriate scale has been used.
The methodology for determining appropriate scale and spatial resolution as well as suggestions
for defining compartments are included in this chapter.
There are several questions that need to be answered before the appropriate level of
scenario complexity can be determined. As with any modeling exercise, the first and foremost
step in a TRIM.FaTE analysis is to clearly state the objective of the analysis. The objective
should identify the chemical(s) of concern, the exposed population (individual, species,
population, cohort, or environmental compartment), and the health endpoint (chronic or acute) to
be assessed. The exposed population, either human or ecological, or landscape component (e.g.,
lake, wetland, agricultural plot), is referred as the target. This section presents a tiered approach
that incorporates all of these objectives to define the appropriate scale and resolution of a given
TRIM.FaTE scenario.
For the chemical(s) of concern, the two most important factors for determining the
appropriate modeling scale are how rapidly the chemical moves and how rapidly the chemical
degrades in the environment. The range of mobility for each target of interest also needs to be
considered. In addition to providing information about modeling scale, the mobility
characteristics of the target help determine the appropriate level of resolution of the scenario.
Finally, the endpoint being assessed will provide important information about both temporal and
spatial scale of the scenario.
The approach described in the following section generates a starting point for any given
analysis objective for which TRIM.FaTE is designed, and is intended to impart some consistency
and transparency into the scenario set up process. Additionally, once a scale has been chosen,
one must determine if that scale is appropriate when compared to other sources of model
uncertainty.
NOVEMBER 1999 C-1 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
C.I CLASSIFICATION OF THE ANALYSIS OBJECTIVE
The methodology for the set up of a scenario will depend largely on the mobility of the
target. The mobility of each exposed population is categorized into one of three general classes:
mobile, bounded, and stationary. Each class is described in the following sections and
summarized in Table C-l. Depending on which category best describes the target of interest, the
user is referred to one of three binary decision trees that are described in Section C.2. The
decision trees provide a series of questions that help the user determine the appropriate set of
boundaries and parcels in the analysis. Information on the different boundary types is provided
in Section C.2.
Table C-l
Classification of Modeling Objectives Based on Target of Interest
Target Class
Description
Rationale and Example
Mobile
Highly mobile cohorts,
individuals, animals or
organisms
Concentration in air resulting from point source will
decline with distance traveled from source. If a child
goes to school near a source but lives farther from
the source and plays in a park somewhere else then
one would want to maximize resolution within the
model system
Bounded
Animals with a limited
range or habitat
Red tailed hawks or land mammals in a limited or
bounded habitat that is a fixed distance from the
source. Concentrations or environmental conditions
may vary across the habitat/range but highly
resolved concentration gradients between the
source and the study area are not necessary.
Stationary
Fixed location in space
that may be influenced
by its surroundings but
does not move relative
to source
Forest, pond, agricultural plot, wetlands. Consider
chemical transfer from adjacent areas (watershed,
air parcel).
C.I.I MOBILE EXPOSED POPULATIONS
The first target class is referred to as mobile. This class includes humans and large
animals that are highly mobile and can freely move about the region impacted by the source(s).
Mobile targets require maximum resolution when estimating concentration, especially for areas
where the highest exposure is likely to occur or with a high likelihood of occupancy by the target
(schools, residential areas, wintering grounds). Each scenario focused on mobile targets should
be designed to provide maximum possible resolution, given the constraints of limited computing
resources, model uncertainty, and measurement imprecision.
NOVEMBER 1999
C-2
TRIM.FATE ISO VOLUME 1 (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
C.I.2 BOUNDED EXPOSED POPULATIONS
The second target class is referred to as bounded. This class includes targets that are
expected to have some limits, either natural or artificial, on their habitat or mobility range. For
example, red tailed hawks that hunt and live in a specific orchard or fish confined to a certain
lake or pond would be included in this class. The scale and resolution of a given scenario should
be selected to provide the desired level of detail within the bounded region. Areas outside the
bounded region can be simplified to include only the information that influences transport of
chemical into the region of interest. For example, if one region has a high quantity of vegetation
and the chemical has a high degradation rate in vegetation, this would be considered separately
because it influences the mass balance. The bounded area where the target(s) reside may or may
not be well mixed (i.e., the concentration across the area does not change significantly). If the
range is not expected to be well mixed, more spatial resolution within the range can be included,
provided it can be justified under the constraints of model uncertainty. If the range is well mixed
then additional spatial resolution is not necessary.
C.I.3 STATIONARY EXPOSED POPULATIONS
The third target class is referred to as stationary. This class includes all immobile targets
such as lakes, forests, agricultural plots, or wetlands. Stationary targets also require little
resolution between the source and the target. Only information that relates the point source to the
location of interest is necessary. Such information might include adjacent air parcels and
drainage areas from which water (runoff) and soil (erosion) are transferred to the stationary
target.
When more than one target is considered in the analysis, the scenario should be set up to
satisfy the target that requires the most resolution. For example, if a study was interested in both
the exposure received by a cohort of humans (mobile target) and the maximum concentration in a
local pond (stationary target), the setup should follow the procedure described in the decision tree
for the mobile target. The concentration in the pond will likely be estimated in the process of
characterizing the various human exposure pathways. Thus, the "high resolution" system based
on the mobile target should provide an adequate level of detail for estimating the concentration in
the pond. If the concentration in the pond were not determined by the system for the mobile
target, the user could either add additional parcels to the scenario to account for the pond or set
up an additional analysis based on the stationary target.
C.2 PARCEL BOUNDARY TYPES
The decision trees for each of the target classes use existing information about
boundaries of the modeling system to facilitate the set up of a scenario. There are several types
of these boundaries that can be used independently or in concert. For simplicity, three classes of
boundaries are defined: natural boundaries, physiochemical boundaries, and population
boundaries. Each of the boundary types is summarized below along with a description of how
they are used to define system boundaries and parcels. The boundary types are summarized in
Table C-2. These simple classifications allow the boundary types to be easily referenced from
the binary decision trees.
NOVEMBER 1999 GO TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
Table C-2
Boundary Types Coded for Use in Decision Trees
CODE
1
1a
1b
1c
1d
2
2a
2b
2max
3
3a
3b
BOUNDARY TYPE
Natural
Air shed
Water shed
Lakes and rivers
Homogeneous land use/cover regions
Physiochemical
Characteristic travel distance
Dispersion modeling
External boundary capturing 90 % of
chemical mass (system boundary)
Population
Semi-mobile (range)
Immobile (location)
C.2.1 NATURAL BOUNDARIES
Natural boundaries include air sheds, watersheds, water bodies and homogeneous land
use and land cover regions. When specifying natural boundaries one can refer to satellite images,
topographical maps, or GIS coverage databases.
An air shed can include large valleys such as the Sacramento valley (CA) where, due to
inversion layers and diurnal wind patterns, the air mass is confined and well mixed throughout
the area for a large portion of the time. Air shed boundaries can also include smaller valleys
when meteorological conditions produce long residence time for the air mass in the bounded
region. Air shed boundaries are useful in providing information about the scale of the model
region (i.e., external boundaries of the system).
Watersheds are also useful in determining the scale of the system as well as the size and
location of parcels within the system, especially if the concentration in a particular lake or
wetland is of interest. Watershed boundaries can be identified or approximated from
topographical maps by tracing ridgelines and noting the origin and direction of flow for streams
and rivers. The size and location of a watershed can influence the transfer of chemical to water
bodies within the basin.
NOVEMBER 1999
C-4
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
Locations and sizes of water bodies and information describing land use and land cover
patterns can also indicate important boundaries in the system. If the analysis objective includes
estimating the impact of a source on a particular water body, agricultural plot, or forest, these
boundaries can be incorporated into the scenario setup.
C.2.2 PHYSICOCHEMICAL BOUNDARIES
Physicochemical boundaries are based on the characteristic travel distance and direction
of the chemical of interest. Physicochemical boundaries can help define both internal parcels and
system scale. Physiochemical boundaries can be applied to water bodies, such as rivers and
streams, if advection and diffusion in water is the dominant pathway by which a chemical travels
through the environment, but the default translocation pathway for most chemicals will be
advection in the air.
When determining the scale of the external system boundaries, the user needs to
determine the range over which the contaminant is likely to spread and if it is necessary for the
model system to capture this range for the user to be able to answer the desired question. For
example, chemicals that are highly mobile in the environment will move far from the source. One
might want to include this entire range, for example, if they want to determine the total number
of people exposed to the chemical. If the goal is to determine the exposure to a nearby
population, a smaller system might be appropriate. In this case, a background concentration will
need to be used to account for pollutant mass flowing back into the system when there is a
change in the wind direction. For chemicals that rapidly deposit to the land surface, it is often
easier to model the range over which the chemical is likely to spread and thus is desirable to
model the full range of the chemical.
Travel distance (based on the chemical-specific deposition velocity), local weather data
(e.g., wind speed/direction, rainfall data, temperature), and approximate landscape characteristics
(e.g.. locations of water, forest, and bare soil) are used to provide an estimate of the distance that
a chemical will travel from the source. Ideally, one needs to account for the travel distance in
each of the four major directions (i.e., north, south, east, and west) to account for variations
resulting from land use and changing weather patterns. The travel distance can be used to
estimate changes in atmospheric concentration, thus providing the maximum resolution that can
still be considered statistically significant (can be detected given the uncertainty in model
predictions and imprecision in environmental measurements). Characteristic travel distance can
be used to construct polygons that incorporate advection, dispersion, and physical loss of
chemical from the atmosphere.
If the objective is to track the movement and fate of the pollutant over its lifetime then
one would want to calculate the distance over which 90 percent of the pollution had been
removed from the air by reactions and dry deposition.
The size of a grid cell can then be determined by calculating the length over which X
percent of the mass of the chemical species with the shortest characteristic distance is lost. One
would want to base this on both atmospheric reactions and the most rapid depositional processes
(i.e., assume both wet and dry deposition). The selection of a factor of X = 50 is dependent on
NOVEMBER 1999 O5 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
the precision of the model output and relevant measurements and will likely result in grid cells
that are smaller than necessary. The following equation can be used to approximate the distance,
L, for any specified percent reduction:
L = -In ((100 - % reduction) /100) * characteristic length
where:
% reduction = the specified percent change in chemical mass along a path of length L
The characteristic travel distance for a chemical in the environment can be calculate using
the chemical's estimated residence time in the atmosphere along with speed and direction of the
moving phase. The characteristic travel distance in the atmosphere can be calculated finding the
distance at which the concentration has reached 36 percent of the initial concentration (Bennett et
al. 1998). This can be calculated as:
characteristic length = 0.23 * wind velocity/loss rate
where:
characteristic length = distance at which 63 percent of the mass in the air cell has
been removed
wind velocity = average wind velocity (m/d)
loss rate = loss rate from the atmosphere from transformation and
depositional processes (1/d)
The 0.23 term approximates the effects of dispersion but dispersion and diffusion are not
explicitly modeled in the above calculation. Gifford and Hanna (1973) have shown that the
yearly average concentration in a simple box model is proportional to the source strength in mass
per unit area divided by the wind speed. McKone (1993a,l993b,l 993c) has used the Gifford and
Hanna work with Benarie (1980) to derive the proportionality constant in this relationship.
Multiplying the unidirectional wind velocity by 0.23 accounts for is the changing direction of the
wind; in other words, if you averaged the wind in one direction it would be about 23 percent of
the wind speed at any time. This factor may underestimate the characteristic travel distance in
locations with persistent wind flow in one direction, and as a result, may result in finer grid
spacing.
For transformation losses, the loss rate is equal to the reaction rate in the atmosphere
which, for first order reactions, is given by 0.693 divided by the half life in the atmosphere. For
deposition losses, the loss rate is equal to the deposition velocity divided by the mixing height.
loss rate = (0.693/half life + deposition velocity/mixing height)
where:
NOVEMBER 1999 C4TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
half life = reaction half life of the chemical in the atmosphere (d)
deposition velocity = total deposition velocity including wet and dry deposition
to soil, water, and vegetation (m/d)
mixing height = mixing height of the atmosphere (i.e., the height over which
pollutants can be assumed evenly mixed) (m)
The atmospheric mixing height is a function of climatic conditions and can be calculated
from meteorological data. When determining the characteristic travel distance for a chemical,
one may want to make a calculation for each season since the average weather (mixing height,
precipitation, mean winds, etc.) may vary significantly. The reaction rates and deposition
velocities are chemical specific.
Air dispersion models can also be used to locate system boundaries. By plotting x-y
spatial coordinates along with air concentration (z), a map or map-overlay can be generated and
used in the same way that a topographical map is used for identifying natural boundaries. The
key to using dispersion models or characteristic travel distance for estimating boundaries is
deciding what change in concentration can be considered significant. Even if the model gave
perfect information, variation in field measurements and imprecision in analytical equipment
would likely require a 10 percent to 50 percent change in concentration before the concentration
difference between two places on a map could be considered statistically significant (Eiceman et
al. 1993). This coefficient of variation increases as the environmental concentration approaches
the experimental detection limit of the equipment.
However, it can be safely assumed that multimedia models do not give perfect
information. An international group of expert model developers and model users recently
concluded that a reasonable estimate (admittedly subjective) of model accuracy for multimedia
pollutants was a factor of three. This factor is expected to increased by an additional factor of
two each time the pollutant crossed a compartmental interface (Cowen et al. 1995). Thus, as the
pollutant moves away from the source through adjacent compartments in TRIM.FaTE, the
distance across each compartment should be increased (/'. e., reduce resolution with increased
distance from source). This characteristic is intuitive in that at some distance from the source,
the pollutant will become a regional or global pollutant and one will no longer be able to directly
link the pollutant back to the original source.
C.2.3 POPULATION BOUNDARIES
Population mobility boundaries take into account additional information about the habitat
of the target and the different locations in which the target is likely to be during an exposure
event. Population mobility boundaries are comparable to natural boundaries except that no
physical boundary would be visible on a map. Population boundaries can be used to justify
increased complexity of a landscape parcel within a natural boundary or within the range of a
bounded target (as described in previous sections). For example, an antelope may spend most of
its life foraging in the high desert sage around a munitions storage facility. No natural
boundaries exist and physicochemical boundaries may provide more resolution than is
NOVEMBER 1999 CX7 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
appropriate. If the seasonal foraging area of the antelope can be identified, this information can
be used to construct parcels that encompass the animal's range for each season. If that range is
large, physicochemical boundaries can be used to increase resolution (more parcels) within the
boundaries as necessary.
C.3 REVISITING THE INITIAL MODEL SETUP
This section addresses methods for identifying necessary changes in grid spacing and
changes to the external system boundary after an initial TRIM.FaTE simulation and a basic
uncertainty analysis have been completed. It is important to determine if any changes to the grid
spacing are necessary once an initial run has been completed, keeping in mind the analysis
objective, the type of target, and the variance in model results.
If the uncertainty of the concentrations within a compartment is greater than the
difference in concentration between adjacent compartments, using a finer grid scale will not
increase the information that can be obtained from the model. On the other hand, if the
uncertainty is less than the difference in concentration between compartments providing a
situation where there is a statistically significant difference in concentration between adjacent
compartments of similar composition, then a finer grid size may be appropriate. When
considering combining adjacent parcels it is important to also examine environmental
characteristics of each parcel. An obvious example where adjacent parcels would not want to be
combined is adjacent air parcels over land and water. If the water body is large, then the
atmospheric mixing height would be different and as a result, the parcels should not be
combined.
For instance, a user might find that the initial simulation did not include a large enough
region, and the simulations result in a significant portion of chemical mass leaving one of the
system boundaries. In this case, the user might want to consider increasing the scale, if
warranted by the model objectives, and completing a new simulation. The user would need to
consider whether or not there is a potential exposed population downwind from the site. If there
is a sensitive ecosystem, farmland, or marine-harvesting region just outside the suggested model
range, the user might want to extend the range to include this area as the chemical may
bioaccumulate in the food chain, causing a significant exposure. Also, the user would want to
evaluate whether or not the mass leaving the system will result in concentrations above the
background level. If not, the user might not want to expand the region. If the region was
expanded and no gain in useful information was realized, those parcels could always be removed
for future simulations.
C.4 SETTING UP THE MODELING REGION
This section builds upon the principles important to selecting the scenario scale, as
presented in Sections C.I through C.3, and addresses the details of defining parcels, volume
elements, and compartments.
NOVEMBER 1999 C-8 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
C.4.1 DETERMINING PARCELS
To guide the development of parcels for a given model scenario, a series of decision trees
are presented, each for a different model objective. By following the decision tree, appropriate
parcels can be determined. After the analysis objective is defined, the dominant target class and
chemical(s) of concern can be identified. The target class is then used to select the appropriate
decision tree to aid in defining parcels. Refer to Figure C-l for a mobile target, Figure C-2 for a
bounded target, and Figure C-3 for a stationary target. By following the decision tree down a
yes/no path, the important boundary types are identified in the order of significance until an
adequate level of complexity is attained. The boundary types are listed and coded in Table C-2.
The general pattern within the decision trees consists of a starting point indicated by an
oval containing the target class at the top of the page. Following the decision tree, one will come
to a diamond that contains a simple question that is answered either yes or no. This may lead to
another question or to an action indicated by a rectangle. The action boxes direct the user to one
or more of the coded boundary types listed in Table C-2. For mobile targets, the air parcels are
defined first. For the bounded and stationary targets, the land surface parcels are defined first
followed by the air parcels. When all parcels are defined (no more questions remain on the tree)
the process moves to the simplification stage. This stage includes final smoothing of boundary
lines, combining adjacent parcels with similar composition, and adding additional parcels around
the perimeter of the system.
After the preliminary' map of the parcels has been completed based on the decision tree.
some slight modifications may need to be made to the map of parcels. Conceptual filtering (also
known as best judgment) can be used to transform the curved lines into simple connected
polygons while conserving the area and approximate location relative to the source and adjacent
parcels. The horizontal area of an air or water basin can be can be estimated using a planimeter
by tracing the boundary several times and calculating the average area. The area can then be
used to estimate the parcel size. Alternatively, one can simply use a clear ruler and best
judgment to straighten the lines. Boundaries based on airsheds can be simplified in the same
manner.
Although the major land types (e.g., forests, urban areas) generally should be considered
separately, the actual boundaries of the landscape types may need to be modified to fit the grid
structure. In order to capture differences in landscape regions, the land under an air parcel can be
split into multiple parcels. This can be an advantage for including rivers and lakes that are
narrow or small relative to the air parcel size (differences in atmospheric mixing height over land
and water can be ignored if the water body is small). Also, one could include various land uses
in a single land parcel if transport differences across various land uses are not significant (e.g., a
land parcel may include 90 percent conifer forest and 10 percent deciduous forest). In this case, a
hybrid parcel containing a fraction of each cover would be created.
NOVEMBER 1999 C-9 TRIM.FATE TSD VOLUME I (DRAFT)
-------
O
on
LU
<
5
a.
a:
D-
a.
c
U S
x 5
5 S
UJ r-
0. H
CL P
< D
e
o
'So
TJ
O
C.
3
C
u
o
OL
a
I
Q
f-
H
u-
2
c
6
JJ
a
T3
O
U
a:
UJ
CD
e
I
-------
X !=:
5 5
z ±
c
o
o
O
£
c.
^
c
U
u. •—
u
R
e
h-
•O
v
•o
e
3
o
CQ
fc
<
af
Q
5
I
<
U-
DC
UJ
m
-------
CJ
on
u
u
X
D
Z
u
D_
C.
< D
o
'So
w
tJ
c
c
S C
.SP'S
U. 'S
o
fi
C
5
•5
6JD
«
H
fr
ff5
o
o
Q
l^£
fN
U
re
-o
i
-a
o
CJ
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
C.4.2 DETERMINING VOLUME ELEMENTS
After the parcels have been determined, the volume elements are specified. This step
involves determining the appropriate number of volume elements and specifying the appropriate
depth for each one. Whereas parcels only represent the modeling region in two dimensions,
volume elements add the component of depth, thus representing the modeling region in three
dimensions. The development of volume elements represents the final step in specifying the
spatial resolution of the modeling region.
If the parcel represents a surface water area, surface water and sediment volume elements
may need to be defined. The appropriate depth of the surface water volume element can be
determined, for example, based on the average depth of the surface water within that region. An
upper and lower water volume element may be appropriate if, for example, the water body is very
deep or if different types offish and other aquatic animals live at different depths. The level of
refinement (i.e., number of volume elements used to represent) for a surface water body also
depends on the level of detail necessary to answer the modeling questions.
If the parcel represents a land area, the number, depth, and type of soil volume elements
will need to be determined. For instance, if the region is in the forest, the soil is unlikely to be
tilled and thus the number and depth of soil volume elements would be determined based on the
depth the chemical is likely to penetrate. The number of modeled soil layers depends on the
desired level of detail and objective of the scenario, but typically three soil layers (represented as
volume elements) are considered. For a given land parcel, there is generally a thin volume
element composed predominantly of surface soil, reflecting the depth of soil likely to be
incidentally ingested by wildlife. The root zone soil layer, represented by a separate volume
element, would typically be immediately below the surface soil layer and would reflect the depth
to which plant roots are likely to be in contact with the modeled chemical(s). The vadose zone
would then extend from the bottom of the root zone soil layer to the ground water surface. If the
chemical is likely to penetrate deeply into the soil, a volume element composed predominantly of
ground water may also be included.
If the parcel represents air, the number and depth of air volume elements needs to be
determined. The boundaries of air parcels do not necessarily have to coincide with the
boundaries of the surface water and soil parcels, although to limit the computer resources
required for a simulation, the parcel boundaries may be made identical. The number of vertical
layers (represented as volume elements) modeled for each air parcel is determined based on the
desired level of detail and the modeling objective. For example, if the modeled source has a high
release height and only one volume element is modeled, the results are likely to overestimate the
deposition of the chemical close to the source. In this case, it would be advantageous to model
multiple volume elements representing multiple vertical layers. In contrast, it may be appropriate
to model a source with a low release height with one vertical layer.
NOVEMBER 1999 C-13 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
C.4.3 DETERMINING COMPARTMENTS
Abiotic
Abiotic compartments are determined by the predominant abiotic medium in the volume
element within which they are contained. At least one abiotic compartment must be contained
within each volume element and, although not typically utilized, the model framework does
support multiple abiotic compartments within a volume element. In most cases, the
determination of abiotic compartments is an implied step because they are simply defined by the
predominant abiotic media within the volume element. For example, if a given volume element
is composed predominantly of surface soil, a surface soil compartment would be included in the
volume element.
Biotic
The transport of chemicals to biota consists of diffusive and advective processes, through
the latter term is rarely used by biologists. Chemicals diffuse into plant leaves from air;
chemicals deposit onto plant leaves with particles in air, an advective process. The uptake of
chemicals from soil or soil water by plant roots or earthworms is treated as diffusion, though
water carries the chemical into the plant (advection). Similarly, chemicals are assumed to enter
algae, macrophytes and benthic invertebrates by diffusion. The major advective process is food
intake by fish, birds and mammals.
The only transport process within biota that is included in TRIM.FaTE is transport
through the plant stem in xylem and phloem fluids. The distribution of chemicals among organs
in individual wildlife is not a feature of TRIM.FaTE.
C.5 REFERENCES
Benarie, M.M. 1980. Urban Air Pollution Modeling. Cambridge, MA: MIT Press.
Bennett, D.H., T.E. McKone, M. Matthies, and W.E. Kastenberg. 1998. General formulation of
characteristic travel distance for semivolatile chemicals in a multimedia environment.
Environmental Science and Technology. 32:4023-4030.
Cowen, E.C., D. Mackay, T.C.J. Feihtel, D. van de Meent, A. DiGuardo, J. Davies and N.
Mackay. 1995. The multi-media fate model: A vital tool for predicting the fate of chemicals.
Pensacola, FL: SETAC Press.
Eiceman, G.A., N.S. Urquhart, and G.A. O'Connor. 1993. Logistic and economic principles in
gas chromatography-mass spectrometry use for plant uptake investigations. Journal of
Environmental Quality. 22:167-173.
Gifford, F.A. and S.R. Hanna. 1973. Modeling Urban Air Pollution. Atmospheric
Environment. 7:131-136.
NOVEMBER 1999 C-14 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX C
DETERMINING APPROPRIATE SCALE AND SPATIAL RESOLUTION
McKone, T. E. 1993a. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part I: Executive summary. Laboratory.UCRL-CR-111456PtI. Livermore, CA: Lawrence
Livermore National.
McKone, T. E. 1993b. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part II: The dynamic multimedia transport and transformation model. UCRL-CR-111456PtII.
Livermore, CA: Lawrence Livermore National Laboratory.
McKone, T. E. 1993c. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part III: The multiple-pathway exposure model. UCRL-CR-111456PtIII. Livermore, CA:
Lawrence Livermore National Laboratory.
NOVEMBER 1999 C-15 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
[This page intentionally left blank.]
-------
APPENDIX D
TRIM.FATE INPUTS
APPENDIX D
TRIM.FaTE Inputs
Input Parameter
'~ j:; ' '-^-^^iflHIItei
Location of source
Height of emission source
Emission rate (for each chemical)
Particle size
, - ^'
Background concentration in each compartment
Units ^
^fflgjflffi^^g^JIJ^f^fj^^K* '
Horizontal wind speed
Horizontal wind direction
Vertical wind speed
Air temperature
Precipitation
Mixing height
Relative humidity
m/s
degrees
m/s
°K
m / day
m
unitless
SPATIAL DATA
Corners of each volume element (VE)
Height of each air VE
Surface soil depth (for each surface soil VE)
Root zone depth (for each root zone VE)
Vadose zone depth (for each vadose zone VE)
Ground water layer depth (for each aquifer layer VE)
Surface water depth (for each surface water VE)
Sediment layer depth (for each sediment layer VE)
UTM coordinates
m
m
m
m
m
m
m
•ABIOTIC ENVIRONMENTAL SETTB*&ft^I^:^fe ^.,-;* - -
Air
(assume same for all air compartments)
Atmospheric dust load
Dust density
Dry deposition velocity of air particulates
Washout ratio
Surface area per volume of particles
Junge C
Density of air
Fraction organic matter on particulates
Diffusion coefficient of water in air
Boundary layer thickness in air above soil
kg[dust] / m3[air]
kg[dust] / m3fdust]
m/day
[mass chem/volume rain] / mass chem/volume air]
m2farea] / m3[particles]
m-Pa
g/cm3
unitless
m2/d
m
Surface Soil
(assumed same for all surface soil compartments)
Land use type
unitless
NOVEMBER 1999
D-l
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Water content
Air content
Soil material density
Organic carbon fraction
Air soil boundary thickness
Default depth of runoff water
Fraction of area available for vertical diffusion
Fraction of area available for erosion
Fraction of area available for runoff
Units
volume[water] / volume[compartment]
volume[air] / volumefcompartment]
kgrsoill/m3[soil1
unitless
m
m
m2[area available] / m2[total]
m2[area available] / m2[total]
m2[area available] / m2[total]
Root Zone
(assumed same for all root zone compartments)
Water content
Air content
Soil material density
Organic carbon fraction
volume[water] / volumefcompartment]
volume[air] / volumefcompartment]
kgfsoill / m3fsoil]
unitless
Vadose Zone
(assumed same for all vadose zone compartments')
Water content
Air content
Soil material density
Organic carbon fraction
volumefwater] / volumefcompartment]
volumefair] / volumefcompartment]
kgfsoil]/m3fsoill
unitless
Ground Water
(assumed same for all ground water compartments)
Porosity
Air content
Solid material density in aquifer
Organic carbon fraction
volumeftotal pore space] / volumefcompartment]
volumefair] / volumefcompartment]
kgfsoil] / m3[soil]
unitless
Surface Water
(depends on water body type - values provided have been developed for an initial simple water body scenario)
Flush rate
Suspended sediment concentration
Evaporation of water
Current velocity
Organic carbon fraction in suspended sediments
Suspended sediment density
Boundary layer thickness above sediment
Drag coefficient for water body
Viscous sublayer thickness
pH
Chloride concentration
flushes/year
kgfsediment] / m3[water column]
m'fwater] / m2farea]-day
m / s
unitless
kgfsediment] / m3[sediment]
m
unitless
m
unitless
mg/L
Sediment
(depends on associated water bodv type)
Organic carbon fraction
Solid material density in sediment
Porosity of the sediment zone
Benthic solids concentration
unitless
kgfsediment] / m3[sediment]
volume[total pore space] / volumefsediment
compartment]
kgfsedimentl / m'fsediment compartment]
NOVEMBER 1999
D-2
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter Units
^^^HH^^IHPHIHIIK^
General to all media
Molecular weight
Octanol-water partition coefficient (K01V)
Melting point
Water solubility
Henry's Law constant
Diffusion coefficient in pure air
Diffusion coefficient in pure water
Organic carbon partition coefficient
g/mol
L[water] / L[octanol]
°K
mol / m3
Pa-m3 / mol
m2 / day
m2 / day
L[water] / kg[carbon]
Surface Soil
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
1 /day
1 /day
1 /day
1 /day
Root Zone
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
1 /day
1 /day
1 /day
1 / day
Vadose Zone
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
1 / day
1 /day
1 /day
1 / day
Ground Water
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
1 / day
I/ day
1 / day
1 /day
Surface Water
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
1 /day
1 /day
I/ day
11 day
Sediment
Methylation rate constant for Hg(2) to MHg
Demethylation rate constant for MHg to Hg(2)
Reduction rate constant for Hg(2) to Hg(0)
Oxidation rate constant for Hg(0) to Hg(2)
I/ day
I/ day
I/ day
I/ day
IfS^pSH^pwp***1 • • •' ' '^^.;^'-^v,. . •-.- f&'tagm&BgBB&g.; -^-..m • — - ••"• -••.-•* **;rs*&spp?''»i • .•»»w?!*&-w.:?-- - * ' ^fpHMB'l
Total erosion rate from soil
Erosion rates between soil and soil
Erosion rates between soil and surface water
Total runoff rate from soil
Runoff rates between soil and soil
kgfsoil] / m2[area]-day
kgfsoil] / m2[area]-day
kg[soil] / m2[area]-day
m3[water] / m2[area]-day
m3[water] / m2[area]-day
NOVEMBER 1999
D-3
TRIM. FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Runoff rates between soil and surface water
Percolation rates between soil and soil
Surface water flow between surface water compartments
Recharge from ground water to surface water
Horizontal water flow rate in ground water
Deposition of suspended sediment in the water column to
the sediment bed
Resuspension of sediment from the sediment bed to the
water column
Burial rate of sediment in the sediment bed
^ ; •:',,••• "•>-,, " ijiriTfiMPltftritJrtM-ikyfij'iu
^V-v
Units
m3[water] / m2[area]-day
m3[water] / m2[area]-day
m3[water] / m2[area]-day
m3[water] / m2[area]-day
m3[water] / m2[area]-day
kgfsediment] / m2[area]-day
kg[sediment] / m2[area]-day
kgfsediment] / m2[area]-day
ANIMALS - AQUATIC
Water Column Carnivore - Bass
Body weight (BW)
Fraction lipid weight
Biomass per area
Population per area
Ventilation rate
Fraction of food diet comprised of fish omnivore
Fraction of food diet comprised of fish herbivore
Fraction of food diet comprised offish carnivore
Fraction of food diet comprised of fish mayfly nymph
kg
unitless
km /m2
#/m2
ml / min / kg
unitless
unitless
unitless
unitless
Water Column Herbivore - Bluegill
Body weight (BW)
Fraction lipid weight
Biomass per area
Population per area
Ventilation rate
Fraction of food diet comprised of phytoplankton (algae)
Fraction of food diet comprised of macrophyte
Fraction of diet mayfly
kg
unitless
km/m2
#/m2
ml / min / kg
unitless
unitless
unitless
Water Column Omnivore - Channel Catfish
Body weight (BW)
Fraction lipid weight
Biomass per area
Population per area
Ventilation rate
Fraction of food diet comprised of macrophyte
Fraction of food diet comprised of mayfly nymph
Fraction of food diet comprised of omnivore '
Fraction of food diet comprised of fish herbivores
Benthic Oi
Body weight (BW)
Fraction lipid weight
Biomass per area
Population per area
kg
unitless
km/m2
#/m2
ml / min / kg
unitless
unitless
unitless
unitless
mnivore
kg
unitless
kg/m2
#/m2
NOVEMBER 1999
D-4
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Ventilation rate
Fraction of diet comprised of benthic invertebrates
Units
ml / min / kg
unitless
Benthic Carnivore
Body weight (BW)
Fraction lipid weight
Biomass per area
Population per area
Ventilation rate
Fraction of diet comprised of benthic omnivores
kg
unitless
kg/m2
#/m2
ml / min / kg
unitless
Benthic Invertebrate-Mavflv
Body weight (BW)
Biomass per area
Total biomass of invertebrates per area
kg
kg/m2
kg/m2
PLANTS - AQUATIC
Macrophvte
Biomass per area
Density of macrophytes
kg / nr
kg/m3
Phvtoplankton - Ahzae
Diameter of algae
Average cell density (per vol cell, not water)
Algae growth rate
Algae density in water column
Algae carbon content (fraction)
Algae water content (fraction)
mm
g/mm3
1 /day
g[algae] / Lfwater]
unitless
unitless
ANIMALS - TERRESTRIAL
Soil Detritivore - Earthworm
Density per soil area, deciduous forest
Density per soil area, coniferous forest
Density per soil area, grass/herb
Density per soil area, agriculture
Density
Water content of worm
kgfworm] / m2 [area]
kg[worml / m2 [area]
kg[worm] / m2 [area]
kg[worm] / m2 [area]
kg[worm] / L[volume]
mass fraction
Soil Detritivore - Soil Arthropod
Body weight (BW)
Biomass per area
kg
kg/m2
Terrestrial Ground-Invertebrate Feeder - Black-canoed Chickadee
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of plants
Fraction of food diet comprised of benthic invertebrates
Fraction excretion to soil
kg
#/m2
kg[soil]/kgBW-day
unitless
unitless
unitless
unitless
kgffood] / kg BW-day
unitless
unitless
unitless
NOVEMBER 1999
D-5
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Fraction excretion to water
Units
unitless
Semiaauatic Piscivore - Kingfisher
Body weight (BW)
Population per area
Soil ingestion rate
Water_a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food comprised of water column herbivore
Fraction of food comprised of water column omnivore
Fraction of food comprised of benthic omnivore
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kg[soil]/kgBW-day
unitless
unitless
unitless
unitless
kg[foodl / kg BW-day
unitless
unitless
unitless
unitless
unitless
Semiaquatic Predator/Scavenger - Bald eaele
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of mouse
Fraction of food diet comprised of chickadee
Fraction of food diet comprised of water column herbivore
Fraction of food diet comprised of water column omnivore
Fraction of food diet comprised of water column carnivore
Fraction of food diet comprised of benthic omnivore
Fraction of food diet comprised of benthic carnivore
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kg[soill / kg BW-day
unitless
unitless
unitless
unitless
kgffoodl / kg BW-day
unitless
unitless
unitless
unitless
unitless
unitless
unitless
unitless
unitless
Semiaauatic Piscivore - Common Loon
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of diet comprised of water column herbivore
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kgfsoill/ kg BW-day
unitless
unitless
unitless
unitless
kgf food] /kg BW-day
unitless
unitless
unitless
Semiaauatic Omnivore - Mallard
Body weight (BW)
kg
NOVEMBER 1999
D-6
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM. FATE INPUTS
Input Parameter
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of plant
Fraction of food diet comprised of benthic invertebrate
Fraction excretion to soil
Fraction excretion to water
Units
#/m2
kg[soil] / kg BW-day
unitless
unitless
unitless
unitless
kg[foodl / kg BW-day
unitless
unitless
unitless
unitless
Terrestrial Predator/Scaveneer - Red-tailed Hawk
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of soil arthropod
Fraction of food diet comprised of chickadee
Fraction of food diet comprised of mouse
Fraction of food diet comprised of short tailed shrew
Fraction of food diet comprised of vole
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kg[soil] / kg BW-day
unitless
unitless
unitless
unitless
kg[food] / kg BW-day
unitless
unitless
unitless
unitless
unitless
unitless
unitless
Terrestrial Insectivore - Tree Swallow
Body weight (BW)
Population per area *
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of benthic invertebrate
Fraction excretion to soil
Fraction excretion to water
kg
#/nr
kgtsoil] / kg BW-day
unitless
unitless
unitless
unitless
kg[foodl / kg BW-day
unitless
unitless
unitless
Terrestrial Herbivore - Meadow Vole
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
kg
#/m2
kgfsoil]/ kg BW-day
unitless
unitless
unitless
unitless
NOVEMBER 1999
D-7
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Food ingestion rate
Fraction of food diet comprised of plant
Fraction excretion to soil
Fraction excretion to water
Units
kgf food] / kg BW-day
unitless
unitless
unitless
Terrestrial Herbivore - Long-tailed Vole
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of plant
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kg[soil] / kg BW-day
unitless
unitless
unitless
unitless
kg[food] / kg BW-day
unitless
unitless
unitless
Terrestrial Predator/Scaveneer - Lone-tailed Weasel
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of mouse
Fraction of food diet comprised of vole
Fraction of food diet comprised of shrew
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kgf soil] / kg BW-day
unitless
unitless
unitless
unitless
kgf food] / kg BW-da\
unitless
unitless
unitless
unitless
unitless
Semiaquatic Omnivore - Mink
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of mouse
Fraction of food diet comprised of vole
Fraction of diet comprised of water column herbivore
Fraction of diet comprised of water column omnivore
Fraction of diet comprised of benthic omnivore
Fraction of food diet comprised of benthic invertebrate
Fraction of food diet comprised of chickadee
Fraction excretion to soil
kg
#/m2
kgf soil] / kg BW-day
unitless
unitless
unitless
unitless
kgffood] / kg BW-day
unitless
unitless
unitless
unitless
unitless
unitless
unitless
unitless
NOVEMBER 1999
D-8
TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Fraction excretion to water
Units
unit less
Terrestrial Omnivore - White-footed Mouse
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of worm
Fraction of food diet comprised of plant
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kgfsoil] / kg BW-day
unitJess
unitless
unitless
unitless
kgffood] / kg BW-day
unitless
unitless
unitless
unitless
Terrestrial Herbivore - Mule Deer/Black-tailed Deer
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of plant
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kg[soil] / kg BW-day
L[water] / kg BW-day
Lfwater] / kg BW-day
unitless
unitless
kgf food] /kg BW-day
unitless
unitless
unitless
Terrestrial Herbivore - White-tailed Deer
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of plant
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kgfsoil] / kg BW-day
Lfwater] / kg BW-day
Lfwater] / kg BW-day
unitless
unitless
kgffood] / kg BW-day
unitless
unitless
unitless
Semiaauatic Omnivore - Raccoon
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
kg
#/m2
kgtsoil] / kg BW-day
L[water] / kg BW-day
L[water] / kg BW-day
unitless
unitless
kgffood] /kg BW-day
NOVEMBER 1999
D-9
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Fraction of food diet comprised of benthic invertebrate
Fraction of diet comprised of water column herbivore
Fraction of diet comprised of water column omnivore
Fraction of diet comprised of benthic omnivore
Fraction of food diet comprised of worm
Fraction excretion to soil
Fraction excretion to water
Units
unitless
unitless
unitless
unitless
unitless
unitless
unitless
Terrestrial Ground-Invertebrate Feeder - Short-tailed Shrew
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of worm
Fraction of food diet comprised of soil arthropod
Fraction excretion to soil
Fraction excretion to water
kg
#/m2
kgfsoil] / kg BW-day
L[water] / kg BW-day
L[water] / kg BW-day
unitless
unitless
kg[food] / kg BW-day
unitless
unitless
unitless
unitless
Terrestrial Ground-Invertebrate Feeder - Trowbridge Shrew
Body weight (BW)
Population per area
Soil ingestion rate
Water a
Water b
Inhalation a
Inhalation b
Food ingestion rate
Fraction of food diet comprised of soil arthropod
Fraction excretion to soil
Fraction excretion to water
kg
#/nr
kg[soil] / kg BW-day
Lfwater] / kg BW-day
L[water] / kg BW-day
unitless
unitless
kgffood] / kg BW-day
unitless
unitless
unitless
PLANTS - TERRESTRIAL
Agricultural Leaf
Water content
Lipid content
Correction exponent, octanol to lipid
Volume of wet leaf weight per unit area
Density of wet leaf
Mass of leaf per unit area
Dry mass of leaf per unit area
Leaf wetting factor
1 -sided leaf area index
Vegetation attenuation factor
Particle washoff rate constant
Diffusion coefficient of water in air
Date litterfall begins
unitless
kg / kg wet weight
unitless
m3/m2
kg/m3
kg[fresh leaf] / m2[area]
kg[dry leaf] / m2[area]
m
m2[leaf] / m2[area]
unitless
1 /day
m2/d
MM/DD
NOVEMBER 1999
D-10
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Date litterfall ends
Date of harvest
Fraction of foliage harvested
Plant-air boundary layer thickness
Length of leaf i
Units
MM/DD
MM/DD
unitless
m
m
Deciduous Forest Leaf
Water content
Lipid content
Correction exponent, octanol to lipid
Volume of wet leaf weight per unit area
Density of wet leaf
Mass of leaf per unit area
Dry mass of leaf per unit area
Leaf wetting factor
1 -sided leaf area index
Vegetation attenuation factor (to calc interception fraction)
Particle washoff rate constant
Diffusion coefficient of water in air
Plant-air boundary layer thickness
Length of leaf
unitless
kg / kg wet weight
unitless
m3/m2
kg/m3
kg[fresh leaf] / m2[area]
kg[dry leaf] / m2[area]
m
rn^leaf] / m2[area]
unitless
1 /day
nr/d
m
m
Coniferous Forest Leaf
Water content
Lipid content
Correction exponent, octanol to lipid
Volume of wet leaf weight per unit area
Density of wet leaf
Mass of leaf per unit area
Dry mass of leaf per unit area
Leaf wetting factor (to calc interception fraction)
1 -sided leaf area index
Vegetation attenuation factor
Particle washoff rate constant
Diffusion coefficient of water in air
Plant-air boundary layer thickness
Length of leaf
unitless
kg / kg wet weight
unitless
m3 / m2
kg/m'
kg[fresh leaf] / m2[area]
kg[dry leaf] / m2[area]
m
m2[leaf] / m2[area]
unitless
1 / day
nr/d
m
m
Herb/Grassland Leaf
Water content
Lipid content
Correction exponent, octanol to lipid
Volume of wet leaf weight per unit area
Density of wet leaf
Mass of leaf per unit area
Dry mass of leaf per unit area
Leaf wetting factor
1 -sided leaf area index
Vegetation attenuation factor (to calc interception fraction)
Particle washoff rate constant
unitless
kg / kg wet weight
unitless
m3/m2
kg/m3
kgf fresh leaf] / m2[area]
kgfdry leaf] / m2[areal
m
m2[leaf] / m2[area]
unitless
1 /day
NOVEMBER 1999
D-ll
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Diffusion coefficient of water in air
Plant-air boundary layer thickness
Length of leaf
Units
m2/d
m
m
Root - Nonwoodv Only
Wet density of root
Water content of root
Lipid content of root
Correction exponent for octanol and lipids
Correction exponent for the differences between octanol
and lipids
Total volume of dry roots in domain per unit area
Area! density agriculture
Areal density grass/herb
kg/m3
unitless
kg / kg wet weight
unitless
unitless
m3/m2
kg/m2
kg/m2
Stem - Nonwoodv Onlv
Density
Water content of stem
Lipid content
Volume of wet stem per unit area
Density of phloem fluid
Density of xylem fluid
Volume of wet weight in domain per unit area
Flow rate of transpired water per leaf area
Fraction of transpiration flow rate that is phloem rate
Correction exponent between foliage lipids and octanol
g/cm3
unitless
kg/kg wet weight
m3/m2
kg/m3
kg / cm3
m3/m2
m3 [water] /m2[leaf]
unitless
unitless
TEMPORAL ENVIRONMENTAL SETTING DATA
Site-specific
Day of first frost
Day of last frost
Deciduous Forest
Lirterfall begin date
Litterfall end date
Uptake by leaf, end date
Uptake by root (herb/grass), end date
LAI = 0, date
Uptake by leaf, begin date
LAI = default value, date
Litterfall rate constant
unitless
unitless
and Grassland
unitless
unitless
unitless
unitless
unitless
unitless
unitless
I/ day
Coniferous Forest
Uptake by leaf, end date
Uptake by leaf, end date
Litterfall rate constant
•
ANIM/dtl^!
Water-column d
Camivore-omnivore partition coefficient
Alpha for carnivore
t.H*.
unitless
unitless
I/ day
iQUATIC
rnivore - Bass
kg /kg
unitless
day
NOVEMBER 1999
D-12
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Assimilation efficiency
Gamma
Units
percent
Water-column Herbivore - Bluegill
Herbivore-algae partition coefficient
Alpha for herbivore
tinha
Assimilation efficiency
Gamma
kg/kg
unitless
day
percent
Water-column Omnivore - Channel Catfish
Omnivore-herbivore partition coefficient
Alpha for omnivore
t,,nha
Assimilation efficiency
Gamma
Benthic Invertebrate (re
Benthic invertebrate-sediment partition coefficient
Alpha for omnivore
W
kg /kg
unitless
days
percent
presented bv Mavfly)
kg /kg
unitless
days
Benthic Carnivore (represented bv Lareemouth Bass')
Carnivore-omnivore partition coefficient
Alpha for omnivore
^alpha
Assimilation efficiency
kg /kg
unitless
day
percent
Benthic Omnivore (represented bv Channel Catfish)
Omnivore-invertebrate partition coefficient
Alpha for omnivore
talnha
Assimilation efficiency
kg /kg
unitless
day
percent
PLANTS - AQUATIC
Macrophvte
Macrophyte- water partition coefficient
Alpha for macrophyte
tatoha
L/g
unitless
days
Phvtoplankton - Algae
D™
Uptake rate
unitless
mm2-d-'-L
ANIMALS - TERRESTRIAL
Soil Detritivore - Earthworm
Earthworm-soil partitition coefficient, dry
t,inh« for worm «-» soil
Alpha for worm «-» soil
mg/kg per rag/kg
day
unitless
Soil Detritivore - Soil Arthropod
Arthropod-soil partition coefficient
t.in>,. for arthropod «-* soil
Alpha for arthropod «-» soil
Terrestrial Ground-Invertebrate F
First-order transformation rate constant for Hg(0)->Hg(2)
kg / kg wet wt
day
unitless
eeder - Black-capped Chickadee
I/ day
NOVEMBER 1999
D-13
TRIM.FATE ISO VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
First-order transformation rate constant for MHg->Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
Units
I/ day
I/ day
1 /day
I/ day
I/ day
unitless
unitless
unitless
Semiaquatic Piscivore - Kingfisher
First-order transformation rate constant for Hg(0)-*Hg(2)
First-order transformation rate constant for MHg-*Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)->MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
/day
/day
/day
/day
/day
/day
unitless
unitless
unitless
Semiaquatic Predator/Scavenger - Bald Eaele
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg->Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-»-MHg
First-order transformation rate constant for Hg(2)->Hg(0)
First-order transformation rate constant for MHg->Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
1 /day
1 /day
1 /day
1 /dav
I/ day
1 / day
unitless
unitless
unitless
Semiaquatic Piscivore - Common Loon
First-order transformation rate constant for Hg(0)-*Hg(2)
First-order transformation rate constant for MHg->Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-»-MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-+Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
1 / day
1 / day
1 /day
1 /day
1 /day
1 /day
unitless
unitless
unitless
Semiaquatic Omnivore - Mallard
First-order transformation rate constant for Hg(0)-+Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for Hg(0)->MHg
First-order transformation rate constant for Hg(2)-»MHg
First-order transformation rate constant for Hg(2)-+Hg(0)
First-order transformation rate constant for MHg-^Hg(O)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
I/ day
1 /day
11 day
1 /day
I/ day
I/ day
unitless
unitless
NOVEMBER 1999
D-14
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
Assimilation efficiency for inhalation for MHg
Units
unitless
Terrestrial Predator/Scavenger - Red-tailed Hawk
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg->Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
I/ day
I/ day
I/ day
I/ day
I/ day
I/ day
unitless
unitless
unitless
Terrestrial Insectivore - Tree Swallow
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
1 /day
1 /day
1 /day
1 /day
1 / day
1 /day
unitless
unitless
unitless
Terrestrial Herbivore - Meadow Vole
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-*-Hg(2)
First-order transformation rate constant for Hg(0)->MHg
First-order transformation rate constant for Hg(2)->MHg
First-order transformation rate constant for Hg(2)->Hg^(0)
First-order transformation rate constant for MHg-*Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
1 /day
1 /day
1 /day
1 /day
1 /day
1 /day
unitless
unitless
unitless
Terrestrial Herbivore - Lone-tailed Vole
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-*Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-+Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
Terrestrial Predator/Scaven
First-order transformation rate constant for Hg(0)-*Hg(2)
First-order transformation rate constant for MHg-*Hg(2)
First-order transformation rate constant for Hg(0)-»MHg
First-order transformation rate constant for Hg(2)-»MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
I/ day
I/ day
1 /day
I/ day
11 day
I/ day
unitless
unitless
unitless
eer - Lone-tailed Weasel
11 day
I/ day
I/ day
I/ day
I/ day
NOVEMBER 1999
D-15
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
First-order transformation rate constant for MHg->Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
Units
I/day
unitless
unitless
unitless
Semiaauatic Omnivore - Mink
First-order transformation rate constant for Hg(0)-*Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for Hg(0)-+MHg
First-order transformation rate constant for Hg(2)-+MHg
First-order transformation rate constant for Hg(2)->Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation of Hg(0)
Assimilation efficiency for inhalation of Hg(2)
Assimilation efficiency for inhalation of MHg
I/ day
I/ day
I/day
I/ day
I/ day
I/ day
unitless
unitless
unitless
Terrestrial Omnivore - White-footed Mouse
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg->Hg(2)
First-order transformation rate constant for Hg(0)-"MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-"Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
I/ day
I/ day
I/ day
I/ day
I/ day
I/ day
unitless
unitless
unitless
Terrestrial Herbivore - Mule Deer/Black-tailed Deer
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg->Hg(2)
First-order transformation rate constant for Hg(0)-»-MHg
First-order transformation rate constant for Hg(2)->MHg
First-order transformation rate constant for Hg(2)-*Hg(0)
First-order transformation rate constant for MHg-*Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
I/ day
I/ day
I/ day
I/ day
I/day
I/ day
unitless
unitless
unitless
Terrestrial Herbivore - White-tailed Deer
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for HgCO^MHg
First-order transformation rate constant for Hg(2)-*MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
I/ day
I/ day
I/ day
I/ day
I/day
I/ day
unitless
unitless
unitless
Semiaauatic Omnivore - Raccoon
First-order transformation rate constant for Hg(0)->Hg(2)
First-order transformation rate constant for MHg-*Hg(2)
I/ day
I/day
NOVEMBER 1999
D-16
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX D
TRIM.FATE INPUTS
Input Parameter
First-order transformation rate constant for Hg(0)-+MHg
First-order transformation rate constant for Hg(2)-»MHg
First-order transformation rate constant for Hg(2)->Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
Units
I/ day
1 /day
11 day
1 / day
unitless
unitless
unitless
Terrestrial Ground-Invertebrate Feeder - Short-tailed Shrew
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-*Hg(2)
First-order transformation rate constant for Hg(0)->MHg
First-order transformation rate constant for Hg(2)-»MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-»Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
11 day
I/ day
I/ day
I/ day
1 /day
1 /day
unitless
unitless
unitless
Terrestrial Ground-Invertebrate Feeder - Trowbridge Shrew
First-order transformation rate constant for Hg(0)-»Hg(2)
First-order transformation rate constant for MHg-»Hg(2)
First-order transformation rate constant for Hg(0)->MHg
First-order transformation rate constant for Hg(2)->MHg
First-order transformation rate constant for Hg(2)-»Hg(0)
First-order transformation rate constant for MHg-*Hg(0)
Assimilation efficiency for inhalation for Hg(0)
Assimilation efficiency for inhalation for Hg(2)
Assimilation efficiency for inhalation for MHg
1 /day
1 /day
1 /day
1 /day
1 /day
1 /day
unitless
unitless
unitless
PLANTS - TERRESTRIAL
Leaf
First-order transformation rate constant for Hg(0)-*Hg(2)
First-order transformation rate constant for Hg(2)->MHg
First-order transformation rate constant for MHg ->Hg(2)
Washout ratio Hg(2) vapor
Washout ratio Hg(0) vapor
Washout ratio Hg paniculate
1 /day
1 /day
1 /day
unitless
unitless
unitless
Root
Alpha for root «-* root-zone soil
talnha
Dry root/root-zone-soil partition coefficient
unitless
day
mg / kg per mg / kg
Stem
Transpiration stream concentration factor
kg / m3 per kg / m3
Leaf Surface
Transfer factor from leaf to leaf surface (Hg)
Transfer factor from leaf surface to leaf (Hg particle)
I/ day
I/ day
NOVEMBER 1999
D-17
TRIM.FATE TSD VOLUME I (DRAFT)
-------
[This page intentionally left blank.]
-------
APPENDIX E
_ PROTOTYPES I - IV
APPENDIX E
Prototypes I - IV
This appendix provides a description of the process of applying the TRIM.FaTE .
methodology (Chapter 4) to cases of increasing complexity (referred to as "prototypes"). Section
1 discusses the computer implementation of the prototypes; Section 2 describes the development
•process for each prototype; Section 3 addresses the features of the prototypes, including the types
of compartments and links simulated; and Section 4 discusses the chemical-specific parameters
and values used in prototype 4. The goals of this appendix are to: (1) illustrate the flexibility of
TRIM.FaTE for application at different levels of spatial and temporal resolution; (2) illustrate
how different multimedia configurations with TRIM.FaTE are set up; and (3) document the
historical development of TRIM.FaTE leading up to Prototype V.
E.I COMPUTER IMPLEMENTATION OF PROTOTYPES
The concepts discussed in Chapter 4 have been implemented in all the prototypes using a
combination of Microsoft Visual Basic™, Fortran, and Microsoft Excel™ software. An
object-oriented architecture was implemented using Visual Basic 5 application environment
imbedded within Excel 97 to model the hierarchy of components of TRIM.FaTE. This hierarchy
includes volume elements, compartment types, compartments, links, and sources. The coding
architecture is not tied to any specific ecosystem configuration. A preliminary algorithm library
that utilized this coding architecture was also implemented.
If all transport processes are simulated as a first-order process, the result is a system of
linear ordinary differential equations, as explained in Section 4.2. This system must be solved to
determine the redistribution of chemical mass as a function of time. For TRIM.FaTE, this
system is solved using the Livermore Solver for Ordinary Differential Equations (LSODE)
(Radhakrishnan and Hindmarsh 1993), a Fortran program freely available via several online
numerical algorithm repositories.
The LSODE subroutine solves systems of first-order ordinary differential equations of the
form (Hindmarsh 1983):
= F(t,y),y(to) =
where y is an n-dimensional time-dependent vector, i.e.,
The system of differential equations can be stiff or non-stiff. In the stiff case, it treats the
Jacobian matrix (Schneider and Barker 1989) as either a full or banded matrix. It uses Adams
(Schneider and Barker 1989) methods (predictor-corrector) in the non-stiff case, and backward
differentiation formula methods in the stiff case. The linear systems that arise are solved by
direct methods. LSODE supersedes the older GEAR and GEARB packages.
NOVEMBER 1999 E-1 TRIM.FATE TSD VOLUME 1 (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
The only restriction on the size of the system of differential equations is that imposed by
computer memory. This code was modified so that it could be accessed by Visual Basic 5 in
Excel 97. Another Fortran code was used, in a similar manner, to determine the steady-state
solution to the system of linear differential equations (Barrodole and Stuart 1981).
Microsoft Excel spreadsheets were used for general preprocessing, postprocessing, and
data storage (additional databases for spatial data were also created using Visual Basic and
accessed by Excel). Excel spreadsheets also served as a convenient interface to the Visual Basic
and Fortran subroutines.
The approach taken for testing the methodology made it possible to investigate the
implications of draft algorithms and to work on the development of a flexible system for
addressing conceptual site models with many compartments. The pre- and postprocessing for the
ultimate implementation of TRIM.FaTE may require a more sophisticated platform. However,
with some modification, much of the Visual Basic code, and all of the Fortran code, can be used
in other computer programming languages.
E.2 PROTOTYPE DEVELOPMENT
Multiple prototypes were developed with increasing complexity to model the movement
of a chemical through an ecosystem. This section describes features of the prototypes in
increasing order of complexity.
E.2.1 PROTOTYPE I
Prototype I (PI) was designed to test the mass transfer methodology (Section 4.2) and the
LSODE utility. Air, surface, soil, ground water, surface water, and fish compartment types were
simulated in PI as illustrated in the conceptual site model shown in Figure E-l. PI includes a
uniform volume source emission of benzene into the air compartment volume. Benzene was
selected because most of its transfer factors were readily available from CalTOX (Maddalena et
al. 1995).
Some transfer factors were derived independently of CalTOX for the air to air sink, soil
to ground water, fish to water, and water to fish transfers. The remaining factors were taken
directly from CalTOX. The dimensions of the terrain were adapted from CalTOX to facilitate
comparison of results. Chemical reaction was not simulated in this prototype.
The runs produced estimates of benzene mass throughout the system, and no problems
were experienced in running the LSODE subroutine. The resulting mass distribution of benzene
in various compartments was examined qualitatively to ensure that the numerical routines were
producing stable and realistic solutions. A quantitative analysis of the results was not performed
because the input parameters were selected only to test the implementation infrastructure. The
results were approximately commensurate with theoretical expectations with no unstable or
anomalous values. These results prompted further testing of the modeling approach on a more
complex ecosystem.
NOVEMBER 1999 E-2 TRIM.FATE TSD VOLUME I (DRAFT)
-------
UJ >
X
Q
2
£. CL
£
c-
e
o
U
o
k.
<
Q^
LU
§.
u.
S
0)
I
LU
T
OI
02
U
O
-------
APPENDIX E
PROTOTYPES 1 - IV
E.2.2 PROTOTYPE II
Prototype II (P2) includes more spatial detail sophistication than PI in both the types and
number of compartments used. Unlike PI, P2 included multiple volume elements for both the
soil and air compartment types and included the use of plant and sediment compartments. In
addition, the links between compartments had multiple-phase (i.e., gas, liquid, and solid) mass
transfers. P2 included a volume source emission of benzo(a)pyrene (B[a]P) into only one of the
air compartment volumes. This made possible a very simple representation of spatial transport.
B(a)P was selected as a test chemical for this and subsequent prototypes because of its
persistence in the environment and because it is a HAP (a chemical of concern in the CAA). The
derivation of the transfer factors is described in detail in the second volume of this document.
The conceptual site model for P2 is shown in Figure E-2.
Multiple-phase (liquid, gas, and solid) transport within a compartment was introduced in
P2. The phases are assumed to be at chemical equilibrium, with the ratios of the concentrations
in the individual phases constant.
P2 was run for four different conditions that included constant source terms under pristine
conditions, an artificially lower organic carbon partioning coefficient (K^) value for B(a)P, a
constant source term with non-pristine conditions in surface water, and a time-varying source-
term condition. In all cases, under steady-state conditions, most of the B(a)P accumulated in the
plants, with minimal penetration into the subsurface. In the water column, most of the B(a)P was
found in the sediment sink, with minimal accumulation seen in the fish compartment. Decrease
of the Koc value resulted in corresponding increase in mass in subsurface soil. Only the air
compartment type seemed to be responsive to the varying source-term condition.
The transfer factors and steady-state outputs of P2 were compared to runs performed on CalTOX
(Maddalena et al. 1995). Most of the transfer factors used in P2 were very similar to those in
CalTOX; the mass distributions of B(a)P were similar in air, soil, and surface water
compartments and differed by three orders of magnitude in plant, sediment, and ground water
(aquifer) compartments. This led to refinement of the TRIM.FaTE algorithms for plant and
sediment compartment types. The difference in the ground water masses was due to the fact that
both TRIM and CalTOX have a simple approximations to model transport in ground water.
E.2.3 PROTOTYPE III
Prototype HI (P3) focuses on code and input data structure refinements because the code
and input data are significantly more complex than either PI or P2. P3 was developed both to
incorporate lessons learned from P2, which has a refined set of abiotic algorithms, and to set up
the TRIM.FaTE model for the case study model run Prototype IV (P4). P3 includes a conceptual
site that approaches the spatial scale (approximately 10-kilometer [km] radius) of the ecosystem
used for the testing the full prototype (P4). The conceptual site model for P3 is shown in Figure
E-3. The vertical dimensions of individual air compartments are not indicated because these
dimensions were allowed to vary with time according to a set of specified meteorological
conditions. The soil and surface water compartments were split into finer grid structures relative
to P2, and several new biotic algorithms were added. The source term simulated in P3 was a
NOVEMBER 1999 &4 TRIM.FATE TSD VOLUME I (DRAFT)
-------
[T t ^
X
5 —
2
LL: u.
c- c-
<£
I
D
(/!
5
f-
.
-
£
&.
re
a
e
e
(S
£
a>
^
SS
\
1
CD
JO
O
£
=6
0)
0)
c5
c
1
•4 ^>
I
-> <-
-> •*-
E
U5
Ov
D£
UJ
m
u
O
2
-------
APPENDIX E
PROTOTYPES I - IV
volume-source emission of B(a)P into only one of the four air compartments. This was used to
make an approximation to a continuous point-source release.
The differences of P3 relative to P2 include:
• Addition of terrestrial earthworm, kingfisher, and mouse compartment types;
• Addition of aquatic food-web system;
• Addition of macrophyte compartment type;
• Addition of compartments with varying heights for air;
• Division of soil compartments horizontally;
• Introduction of "thermoclines" and refinement of mixing for surface water;
• Refinement of plant algorithms;
• Refinement of soil diffusion algorithms;
• Addition of erosion in the soil compartment types;
• Refinement of ground water algorithm;
• Introduction of flexible code design; and
• Introduction of temporal variation for a few key input parameters.
As in the case of P2, several runs were performed for P3. The results showed that the
plant, macrophyte, and sediment compartments are major sinks of B(a)P in the environment. The
model showed that B(a)P mass distribution in the environment is sensitive to total macrophyte
volume in the water column. The model results were extremely responsive in most
compartments to varying source-term conditions. Comparisons of P3 outputs with CalTOX
outputs showed that B(a)P mass distributions in the ecosystem being simulated were in closer
agreement than was seen in the case of P2. This was believed to be a result of refining the
algorithms as previously stated and implied that the prototype was appropriate for application to
a more complicated test case.
E.2.4 PROTOTYPE IV
Whereas PI through P3 used generic inputs and were intended for evaluation simulations.
P4 was designed to be applied to an actual site. PI through P3 were used to develop and test the
TRIM.FaTE algorithms. P4 was developed and used to illustrate and evaluate the likely limits of
TRIM.FaTE with respect to the number of land parcels and length of time steps used. This
prototype had the shortest plausible time step (1 hours), a large number of land units in the planar
view (20 parcels), and 21 different biotic compartment types. This level of detail resulted in
several hundred compartments, including abiotic and biotic compartments, and the sinks
necessary to account for transformation and transport losses outside of the system boundary. To
test the model using a realistic ecosystem, P4 was applied to an area in the northwestern region of
the United States.
NOVEMBER 1999 E-6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
x "7
Z t^
u u:
E >
< H
§
0£
a.
X
XN^
—
C.
o
"o
u
c-
T' £
H -s
g-S
a ®
DC ^^
tu «
•^^
£J
p
U
C
^
X
X
N
S
N
to
* V *
*
^y:
\ ~^
s nr!
^ ^
o <
-^
\ ""•>.
f
I- ""--
D ^ \ ~N-
"1 f
j^ ""x^
at £ a>
J'il
« — £ fe.
Ill
/^ >
,
^ ^ ^
y .
j
N
CC
J K
o a
SO!
o:
DO
-i —
in
-.,
. "-».
£
N
>
Eel
^ M
^"~~
5
CO
1
=
^
w
u.
a
m
**^ ^
_^
-
^
>
CM
^ |
.^
** V ^
§
w
°
X.
tn
C
g
ij
-~,^
^ ^
"e
8
d
~»
I
1
CO
1
1
*c
"•-,
^
0
5
o
-------
APPENDIX E
PROTOTYPES I - IV
E.3 PROTOTYPE FEATURES
The specific features modeled in the prototypes are discussed in this section. Section 3.1
presents the abiotic compartment types modeled; Section 3.2 includes the biotic compartment
types modeled; and Section 3.3 discusses the abiotic and biotic links associated with the
prototypes.
E.3.1 ABIOTIC COMPARTMENTS
In PI (Figure E-l), the air, soil, and surface water each consist of a single volume element
and compartment. Ground water was simulated simply as a sink to the soil compartment. P2, as
shown in Figure E-2, divides the air into four volume elements (two upper air and two lower air
layers); divides the soil into four volume elements (surface soil, root zone, and vadose zones one
and 2); and simulated ground water, surface water, and sediment as single volume elements. In
P3, (Figure E-3) the air consists of six volume elements (two lower air and two upper air over
soil, and a lower air and upper air over surface water); the soil was divided into 32 volume
elements (eight surface soil, eight root zone, eight vadose zone 1, and eight vadose zone 2);
ground water and surface water were both simulated with two volume elements; and sediment
was simulated as a single volume element. P4 simulates 129 abiotic volume elements. Parcels
were defined in P4 and divided vertically based on compartment type. The 129 abiotic
compartments associated with the parcels in P4 are summarized in Table E-l.
E.3.2 BIOTIC COMPARTMENTS
In PI and P2, a single fish species is modeled and only uptake and loss of chemical
through the gills is simulated. In the transition to P3 and P4, the number of biotic water column
compartments was expanded from a single fish species to an aquatic food web represented by
several feeding trophic levels (compartment types). Bioaccumulation by herbivores, as well as
omnivores and carnivores, is accommodated within the P3 and P4 simulations. It is important to
note, however, that the trophic level representations were simplified to reflect primary uptake and
loss from a single representative species from each trophic level.
Both P3 and P4 include terrestrial wildlife as compartments. Wildlife may be exposed to
chemicals through food, soil, and water ingestion, and through inhalation of chemicals in air.
Elimination of chemicals from body tissues may occur through metabolic breakdown of the
chemical and excretion through urine, feces, milk (mammals only), and eggs (birds only).
Terrestrial and semiaquatic biota were not considered in PI and P2. Two species were
introduced in P3: a white-footed mouse (Peromyscus leucopus) and the belted kingfisher (Ceryle
alcyori). These species were selected because they are taxonomically dissimilar (mammal versus
bird) and represent differing compartment types (terrestrial omnivore and semiaquatic piscivore,
respectively). P4 simulated a more complex terrestrial, aquatic, and semiaquatic system, as
summarized in Table E-2.
NOVEMBER 1999 E-8 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
Table E-l
Types of Abiotic Compartments and Number of Volume Elements Modeled
Compartment
Type
Air
Soil
Surface Water
Sediment
TOTAL NUMBER
Number of Volume Elements*
P1
1 -Air
1 - Soil (general)
1 - Ground water
1 - Surface Water
NA
4 Volume Elements
P2
2 - Upper Air Layer
2 - Lower Air Layer
1 - Surface Soil
1 - Root Zone
1 - Vadose Zone 1
1 - Vadose Zone 2
1 - Ground water
1 - Surface Water
1 - Interstitial Water
1 - Sediment
12 Volume Elements
P3
3 -Upper Air Layer
3 - Lower Air Layer
8 - Surface Soil
8 - Root Zone
8 - Vadose Zone 1
8 - Vadose Zone 2
2 - Ground water
1 - Upper Surface Water
Layer
1 - Lower Surface Water
Layer
1 - Interstitial Water
1 - Sediment
44 Volume Elements
P4
20 -Upper Air Layer
20 - Lower Air Layer
14 -Surface Soil
14 -Root Zone
14 - Vadose Zone 1
14 -Vadose Zone 2
14 -Ground water
1 - Upper Lake Layer
1 - Lower Lake Layer
5 - River Segments
6 - Interstitial Water
6 - Sediment
129 Volume Elements
' Reaction and advection sinks are not listed in this table.
Table E-2
Biotic Compartments Modeled
Compartment
Aquatic
Ecosystem
Terrestrial
Ecosystem
Semi-Aquatic
Ecosystem
P1
Single Fish
Species
NA
NA
P2
Single Fish
Species
NA
NA
P3
• Macrophytes (Benthic
Herbivores)
• Aquatic Herbivores
• Aquatic Omnivores
• Aquatic Carnivores
• White-footed Mouse
(Ommvore)
• Earthworm (Soil Detritovore)
• Plant Leaves, Roots, Xylem
and Stem
• Belted Kingfisher (Piscivore)
P4
• Macrophytes (Benthic Herbivores)
• Mayfly (Benthic Herbivores)
• Bluegill (Modeled as Herbivore)
• Channel Catfish (Ommvore)
• Bass (Carnivore)
• Mallard (Herbivore)
• Raccoon (Ommvore)
• Tree Swallow (Insectivore)
• White-footed Mouse (Ommvore)
• Earthworm (Soil Detritovore)
• Black-capped Chickadee (Insectivore)
• Red-tailed Hawk (Predator)
• Long-tailed Weasel (Predator)
• Black-tailed Deer (Herbivore)
• Long-tailed Vole (Herbivore)
• Mink (Piscivore)
• Trowbridge Shrew (Ground
Invertebrate Feeder)
• Insects
• Plant Leaves, Roots, Xylem and Sterr
• Belted Kingfisher (Piscivore)
• Wetland Plant Leaves, Roots, Xylem
and Stem
NOVEMBER 1999
E-9
TRIM.FATE ISO VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
P3 and P4 also simulated pollutant transfer to earthworms. The concentration in
earthworms was assumed to be in equilibrium with the solid, liquid, and vapor-phase
concentrations of the chemical in the root zone compartments.
Plants were introduced to the TRIM.FaTE framework in P2. The plant component of the
ecological model implemented for P2, P3, and P4 is comprised of leaves, roots, xylem, and stem.
Plants are divided into these compartment types because: (1) the literature suggests that
concentrations of non-ionic organic chemicals in foliage are primarily related to those in air and
that concentrations in roots are generally related to those in soil (with stems serving as the
conduit between the two), and (2) herbivores may eat part but not all of a plant. Each
compartment type was assumed to be homogeneously-mixed. The plant algorithms implemented
in P2 through P4 are applicable for mature plants only, and did not address plant growth.
E.3.3 LINKS
If mass can move from one compartment to another compartment without first moving
through intervening compartments, then the two compartments are considered "linked." Each
link is associated with an algorithm that determines the direction and rate of mass flow between
the two compartments. Links may be between compartments in adjacent volume elements or
compartments within a volume element. At a given spatial location, and within a single volume
element, more than one compartment may exist and linkages may exist between these
compartments.
Table E-3 shows examples of generalized links applied in PI through P4. This table is
generic and can be used in conjunction with Tables E-l and E-2 to define a specific link. For
example, in P2 through P4, transfer of a pollutant can occur from an upper air compartment to
adjacent upper air compartments and to a lower air compartment. This is represented in Table E-
3 by the air (sending compartment) to air (receiving compartment) link. A more complex
example is the links associated with the kingfisher from the semi-aquatic ecosystem. As a
receiving compartment, pollutant(s) can transfer to the kingfisher from air (i.e., lower air), soil
(i.e., surface soil), surface water (i.e., upper lake layer), and aquatic (i.e., bluegill) ecosystems.
The links from sending compartments to sinks are not shown in Table E-3. Sinks refer to
the compartments of pollutant mass leaving the modeled ecosystem through a reaction or
physical process(es).
NOVEMBER 1999 E-10 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES J-IV
Table E-3
Examples of Links Associated with Compartments Types
Sending Compartment Types
Air
Soil
Ground water
Surface Water
Sediment
Terrestrial Ecosystem
Aquatic Ecosystem
Semi-aquatic Ecosystem
Receiving Compartment Types
Air
Soil
Surface Water
Terrestrial Ecosystem
Semi-aquatic Ecosystem
Air
Soil
Ground water
Surface Water
Terrestrial Ecosystem
Semi-aquatic Ecosystem
Ground water
Surface Water
Surface Water
Sediment
Aquatic Ecosystem
Semi-aquatic Ecosystem
Terrestrial Ecosystem
Surface Water
Aquatic Ecosystem
Terrestrial Ecosystem
Air
Soil
Aquatic Ecosystem
Semi-aquatic Ecosystem
Terrestrial Ecosystem
Surface Water
Terrestrial Ecosystem
Air
Soil
Surface Water
E.4 PAH-SPECIFIC VALUES USED IN TESTING OF PROTOTYPE IV
This section discusses the testing approach for chemical specific parameters and values.
More detailed descriptions of algorithms associated with many of the parameters discussed in
this section can be found in TRIM.FaTE TSD Volume II.
NOVEMBER 1999
E-ll
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
E.4.1 TRANSFORMATION OF PAHs BY PLANTS
Metabolism in Plants
Few studies of metabolism of organic chemicals in plants exist. Exceptions include
metabolism of: atrazine by poplar trees (Burken and Schnoor 1997); pentachlorophenol in
soybean and spinach (Casterline et al. 1985); trichloroethylene in carrots, spinach, and tomatoes
(Schnabel et al. 1997); PCBs in plants (reviewed in Puri et al. 1997); and bromacil, diclobenil,
nitrobenzene, and 1,3-dinitrobenzene in soybean plants and barley roots. Metabolic rate
constants were only calculated in the first paper. Investigations of the metabolism of polycyclic
aromatic hydrocarbons in plants include: metabolism of phenanthrene and anthracene by tomato
and wheat (Harms 1996), metabolism of anthracene and benz[a]anthracene in bush bean
(Edwards 1988), metabolism of anthracene by soybean (Edwards et al. 1982), metabolism of
anthracene in bush bean (Edwards 1986), and metabolism of various PAHs by bush bean (in
progress, T. McKone, personal communication, August 1997). The first two papers are
somewhat useful for the calculation of a metabolic rate constant, and the ongoing study by
McKone may prove most useful when completed. Unfortunately, the two papers are dynamic
studies with PAH taken up through the soil and air and degraded gradually, perhaps at a first-
order rate, and with metabolites present in the nutrient solution that could also be taken up.
Thus, it is difficult to calculate the metabolic rate constant. Harms (1996) provides
radioactivity (percentage of applied) of parent compound (phenanthrene or anthracene) and
metabolites in culture medium; parent compound, metabolites, and nonextractable residue in
shoots; and parent compound, metabolites, and nonextractable residue in roots after five days of
exposure. If it is assumed that a) non-extractable residues reflect the measured proportion of
parent compound to metabolite, b) metabolites produced in aseptic culture medium were
produced by roots rather than by shoots, c) metabolites did not move between plant organs, and
d) that most of the measured parent compound was in the plant for the majority of the five days
(the rate of uptake may have been rapid because of the application of phenanthrene in
liposomes), a simple calculation of a first-order metabolic rate constant can be made. (Although
these are poor assumptions, it is notable that the order of magnitude variability in rate constants
for metabolism of phenanthrene in shoots of two plant species (below) is probably greater than
errors associated with the above assumptions.)
Thus, a calculation of a lower bound on the first-order metabolic rate constant can be
made. The equation used is In (N/N0) = -kt, where N is the radioactivity of the metabolite pool
after five days and N0 is the sum of the radioactivity of the parent compound pool and metabolite
pool after five days (assumed to be the total radioactivity of the parent compound in the plant
close to the beginning of the experiment). If the calculation is made, the rate constants are:
0.008/d for phenanthrene in tomato leaf and stem, 0.08/d for phenanthrene in wheat leaf and
stem, 0.24/d for phenanthrene in tomato root, and 0.28 for phenanthrene in wheat root. The half-
lives range from 2.5 to 90 days.
Similarly, a calculation of a lower bound on the first-order metabolic rate constant for
benzo(a)pyrene can be made using results from uptake and metabolism of benz(a)anthracene by
bush beans in nutrient solution (Edwards 1988). The PAH was added continually to solution to
NOVEMBER 1999 E^12 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
maintain a constant concentration. In a previous experiment it was determined that most of the
benz(a)anthracene absorbed by roots was taken up within one day. After 30 days 25 percent of
the radioactivity was parent compound and 14 percent was in the form of metabolites; the
distribution of the parent compound and metabolites in the plant is presented in the paper. Using
the same assumptions as above, low estimates of the rate constants are: 0.015/d for
benzo(a)pyrene in root, 0.19/d for the PAH in stem and 0.12/d for the chemical in foliage.
Randy Maddalena and Tom McKone of Lawrence Berkeley Laboratory investigated the
uptake of anthracene, fluoranthene, phenanthrene, and pyrene from air by leaves of bushy beans.
The following calculation is based on a personal communication from Tom McKone in
September 1997. These compounds appear to have reaction rates on the order of 0,1 to 0.3 /day
(half-life of three to 10 days) and thus are somewhat higher than the low estimate of the rate
constant for phenanthrene metabolism in leaves described above.
It is expected that metabolism in plants is estimated within an order of magnitude in
TRIM.FaTE. The parameters in Table E-4 should be used for phenanthrene and benzo(a)pyrene
or as defaults for other PAHs. Different numbers may be chosen in the future as additional
information is obtained. As the root and leaf compartment types are connected, the rate constant
for the stem is likely to change.
Table E-4
First-order Metabolic Rate Constants (d ')
Chemical
Phenanthrene
Benzo(a)pyrene
root
0.3
0.02
stem
0.08
0.2
leaf
02
0.2
Photolysis on the Plant Surface
The process of photolysis on the plant surface was not implemented in the PAH test case
of TRIM.FaTE because the leaf and leaf surface were not separate compartment types. In future
runs of the model for PAHs, photolysis on the leaf surface may be included. Few investigations
of the photolysis of contaminants on plant foliage have been undertaken. An exception is the
photodegradation of 2,3,7,8-tetrachlorodibenzodioxin sorbed to grass foliage (k = 0.0156 hr-1).
It is assumed that photolysis of organic contaminants on the leaf surface occurs at a rate that is
somewhat less than that of PAHs sorbed to paniculate matter in air; PAHs on leaves are probably
exposed to a lower light intensity than those in air. Thus, the rate constant on leaf surfaces is
assumed to be one-half of the rate constant of photolysis of PAHs on particulates in air. Kamens
et al. (1987) provides measurements of the rate constant for benzo(a)pyrene when the chemical is
present at a loading of 30 to 350 ng/mg particulates (0.0211 min"1) and when the PAH is present
at a loading of 1000 to 2000 ng/mg particulates (0.009 min"1). Their more general equation for
determining the rate constant (in min"1) for the 30 to 350 ng/mg loading case is:
In k = -1.355 - 1.279(1 /T) + 0.831(ln(I)) + 0.816(ln[H2O]),
NOVEMBER 1999 E-13 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES I - IV
where:
I = the average solar intensity (cal/cm2/min)
[H2O] = water vapor concentration in g/m3
Kamens et al. (1987) have not investigated photolysis of 3-ringed PAHs such as
phenanthrene. Behymer and Hites (1988) suggest that photolysis is independent of PAH
structure for substrates with a carbon content greater than five percent. In an experiment in
which fifteen fly-ash substrates were irradiated using a mercury vapor lamp (17.6 W/m2), they
investigators measured photolytic rate constants for phenanthrene ranging from <0.00069 hr"1 to
0.0050 hr"1, with a mean of 0.0019 hr"1. The mean rate constant for benzo(a)pyrene was
measured at 0.0035 hr"1. Thus, this measurement is more than an order of magnitude lower than
the numbers in the Kamens study (note that they are presented in min"1).
Without knowledge of solar intensity (and with lots of uncertainty), the following rates
are suggested for photolysis of contaminants on a leaf surface during the daytime hours: 0.03 hr-1
for benzo(a)pyrene and 0.001 hr"1 for phenanthrene.
E.4.2 DISTRIBUTION, ELIMINATION, AND TRANSFORMATION OF PAHs IN
WILDLIFE
The toxicological literature was reviewed to identify models or parameters to describe the
absorption, metabolism, and excretion of phenanthrene in both avian and mammalian species.
No data were found to describe the toxicokinetics of phenanthrene in birds. Although models to
describe the toxicokinetics of phenanthrene in mammals were not found, data suitable for
estimating absorption, metabolism, and excretion rates following oral exposure were available.
These data, and rate estimates developed from them, are outlined below. Phenanthrene appears
to be readily absorbed, metabolized, and eliminated by mammals. Rahman et al. (1986) orally
dosed rats with single one mg dose radiolabeled phenanthrene. Eight hours post dose, 72.74
percent of the initial radio label dose had been recovered in bile or urine, suggesting an
assimilation efficiency of approximately 73 percent.
Chang (1943) orally exposed rats to an experimental diet containing one percent
phenanthrene and by oral gavage of 11 or 13 mg phenanthrene. Amount of parent compound
excreted in feces was measured. Because excretion rates were comparable regardless of the
mode of exposure, results from both dietary and gavage exposure were pooled. Rats excreted
four to seven percent (mean equals 5.75 percent) of the original dose. Conclusions from this
study are limited by the small sample size used in limited description of the methods employed.
Chu et al. (1992) exposed both rats and guinea pigs to doses of radiolabeled phenanthrene
of 10 mg/kg/d via gavage. After 48 hours, rats and guinea pigs had excreted 52 percent and 47
percent of the initial radiolabel. hi rats, 90 percent of the excreted radiolabel was in urine and 10
percent in feces; among guinea pigs, 95 percent of the excreted radiolabel was in urine and five
percent in feces . Of the radiolabel in the urine both species, 95.8 percent and 95.7 percent
consisted of metabolites of phenanthrene and 4.2 percent and 4.3 percent of unmetabolized
phenanthrene in rats and guinea pigs, respectively.
NOVEMBER 1999 E^14 TR1M.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
___^_ PROTOTYPES I - IV
Female rats were orally or dermally exposed to phenanthrene, either as phenanthrene
alone or as phenanthrene adsorbed to sandy or clay soil (Kadry et al. 1995). Absorption was
greatest for pure phenanthrene as compared to phenanthrene adsorbed to soil. Percent absorption
of the initial dose ranged from 55.7 percent to 65.3 percent and 0.7 percent to one percent for oral
and dermal pathways, respectively. After 72 hours, 47.6 percent to 52.4 percent of the initial oral
dose was recovered in urine; 27.8 percent to 22.1 percent was recovered in feces. After 96 hours,
36.2 percent to 48.4 percent of the initial dermal dose was recovered in urine; 8.6 percent to 14
percent was recovered in feces.
The results of these studies are listed and summarized in Table E-5. From these data, the
mean excretion (EJ, metabolic (Em), and absorption efficiencies for phenanthrene are 3.2
percent, 63.4 percent, and 33.8.0 percent, respectively. The first-order rate constants for
metabolism range from 0.1 day"1 to one day"'. Because no data were found for assimilation for
water, soil, or food, assimilation via all pathways is assumed to be equal, e.g., Aa-Aw=As-As.
Because no data were found concerning uptake and elimination of phenanthrene by birds,
parameters developed for mammals should be used. Due to physiological differences between
birds and mammals, use of mammalian values for birds will contribute significant uncertainty to
the final tissue residue estimate.
No studies data were found to enumerate elimination of phenanthrene via lactation (Et) or
elimination via egg production (Ee). However, transfer of contaminants from the diet to milk or
eggs may be estimated using models reported in Travis and Arms (1988) and McKone (1993a,
1993b, 1993c).
E.4.3 UPTAKE OF PAHs BY BENTHIC INFAUNA
Uptake of PAHs is based on the water to benthic infauna transfers presented in Section
7.3.2.1 of TSD Volume II. Uptake of contaminants from water is primarily based on respiratory
processes. (Stehly et al. 1990) have found that the clearance rate of B(a)P and phenanthrene
from water by the mayfly is analogous to the clearance rate of oxygen during respiration. The
uptake of these two PAHs can, therefore, be estimated similarly to the ratio of oxygen clearance
to the volume of water passing over respiratory surfaces. With a known or assumed volume of
water passing over respiratory membranes with known concentrations of B(a)P and
phenanthrene, the extraction efficiency of these PAHs can be calculated. Generic algorithms in
Section 7.4.2.1 of TSD Volume II were adapted from Stehly et al. (1990) for estimating PAH
uptake and loss within the benthic invertebrate, based on the clearance rate driven by the volume
of water cleared and the bioaccumulation factor (BCF). Uptake rates, as measured by a clearance
rate constant, as well as the bioconcentration factor for 30, 60, and 120-day-old mayflies for
B(a)P and phenanthrene, were provided by Stehly et al. (1990).
NOVEMBER 1999 E-15 TRIM.FATE TSD VOLUME I (DRAFT)
-------
I
n
B
05
Q
0
e
'i
es
H ^
o
JD
R
"S
R
R
g
cc
u
<
0)
o
1
£
V
{£
to
(t
o ~
• Metaboli
tant (day
5 ,
UJ ^
Q)
o
e
il
10
ra
0
II
O £
>f Total Dose
nthrene or M
(percent)
ll
CO
OJ
76
c
CO
.c
CO
a:
CO
CO
CO
CD
0>
CD
a,
co
CO
CO
C3
CNJ
r-
a>
"co
'S
3
O
o
CM
00
O)
r-
co
C3
CM
CSI
CM
CD
"to
«3
O
O
CO
ci
!
CM
co
o
CM
^
cn
O3
•
CO
«
£>
TJ
CO
CO
C)
CO
CM
C\i
r-
^
o
CO
in
h-
1
•
CO
0)
to
o
CO
CO
ci
CO
•o
CM
(D
in
•
CO
"53
o
a>
o
§
o
CO
?£
U
c
o
•g
c
CO
o
o
T3
"8
5
X
O — -• « S
"g >v P
^>2
« u o
2 S S
118
• 118-
C _«2 "> so
SI'S 55
•£ 4>
-------
APPENDIX E
^ PROTOTYPES I - IV
E.5 REFERENCES
Barrodale, I. and G. F. Stuart. 1981. ACM transactions on mathematical software. September.
Behymer, T.D., and R.A. Kites. 1988. Photolysis of polycyclic aromatic hydrocarbons adsorbed
on fly ash. Environ. Sci. Technol. 22:1311-1319.
Burken, J.G., and J.L. Schnoor. 1997. Uptake and metabolism of atrazine by poplar trees.
Environ. Sci. Technol. 31:1399-1406.
Casterline, J.L., Jr., N. M. Barnett, and Y. Ku. 1985. Uptake, translocation, and transformation
of pentachlorophenol in soybean and spinach plants. Environ. Res. 37:101-118.
Chang, L.H. 1943. The fecal excretion of polycyclic hydrocarbons following their
administration to the rat. J. Biol. Chem. 151:93-99.
Chu, I., K.M.E. Ng, P.M. Benoit, and D. Moir. 1992. Comparative metabolism of phenanthrene
in the rat and guinea pig. J. Environ. Sci. Health. 827:729-749.
Edwards, N.T. 1988. Assimilation and metabolism of polycyclic aromatic hydrocarbons by
vegetation - An approach to this controversial issue and suggestions for future research. In:
Cooke, M. and A. J. Dennis, eds. Polynuclear aromatic hydrocarbons: A decade of progress.
Tenth International Symposium. Columbus, OH: Batelle Press, pp. 211-229.
Edwards. N.T. 1986. Uptake, translocation and metabolism of anthracene in bush bean
CPhaseolus vulgaris L.). Environ. Toxicol. Chem. 5:659-665.
Edwards, N.T., B.M. Ross-Todd, and E.G. Garver. 1982. Uptake and metabolism of 14C
anthracene by soybean (Glvcine max). Environmental and Experimental Botany.
Harms, H.H. 1996. Bioaccumulation and metabolic fate of sewage sludge derived organic
xenobiotics in plants. The Science of the Total Environment. 185:83-92.
Hindmarsh, A.C. 1983. ODEPACK, A systematized collection of ode solvers. In: R.S.
Stepleman et al., eds. Scientific computing. North-Holland, Amsterdam, pp. 55-64.
Kadry, A.M., G.A. Skowronski, R.M. Turkall, and M.S. Abdel-Rahman. 1995. Comparison
between oral and dermal bioavailability of soil-adsorbed phenanthrene in female rats. Toxicol.
Lett. 78:153-163.
Kamens, R.M., Z. Guo, J. N. Fulcher, and D.A. Bell. 1987. Influence of humidity, sunlight, and
temperature on the daytime decay of polyaromatic hydrocarbons on atmospheric soot particles.
Environ. Sci. Technol. 22:103-108.
NOVEMBER 1999 E-17 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX E
PROTOTYPES 1 - IV
Maddalena, R.L., I.E. McKone, D.W. Layton, and D.P.H. Hsieh. 1995. Comparison of multi-
media transport and transformation models: Regional fugacity mode vs. CalTOX. Chemosphere.
30:869-889.
McKone, T. E. 1993a. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part I: Executive summary. Laboratory.UCRL-CR-111456PtI. Livermore, CA: Lawrence
Livermore National.
McKone, T. E. 1993b. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part II: The dynamic multimedia transport and transformation model. UCRL-CR-111456PtII.
Livermore, CA: Lawrence Livermore National Laboratory.
McKone, T. E. 1993c. CalTOX, A multimedia total-exposure model for hazardous-wastes sites
Part III: The multiple-pathway exposure model. UCRL-CR-111456PtIII. Livermore, CA:
Lawrence Livermore National.
Puri. R. K.. Y. Qiuping, S. Kapila, W. R. Lower, and V. Puri. 1997. Plant uptake and
metabolism of polychlorinated biphenyls (PCBs). hi: Wang, W., J. W. Gorsuch, and J. S.
Hughes, eds. Plants for environmental studies. Boca Raton, FL: Lewis Publishers, pp. 481-513.
Radhakrishnan, K. and A.C. Hindmarsh. 1993. Description and use of LSODE, the Livermore
Solver for Ordinary Differential Equations. LLNL UCRL-ID-113855.
Rahman, A.. J.A. Barrowman, and A. Rahimtula. 1986. The influence of bile on the
bioavailability of polynuclear aromatic hydrocarbons from the rat intestine. Can. J. Physiol.
Pharmacol. 64:1214-1218.
Schnabel, W.E., A.C. Dietz, J.G. Burken, J.L. Schnoor, and P.J. Alvarez. 1997. Uptake and
transformation of trichloroethylene by edible garden plants. Wat. Res. 31:816.
Schneider, H. and G.P. Barker. 1989. Matrices and Linear Algebra, 2nd ed. New York, NY:
Dover Publications, Inc.
Stehly, G.R., P.P. Landrum, M.G. Henry and C. Klemm. 1990. Toxicokinetics of PAHs in
Hexagenia. Environmental Toxicology and Chemistry. 9(2): 167-174.
Travis, C.C. and A.D. Arms. 1988. Bioconcentration of organics in beef, milk, and vegetation.
Environ. Sci. Technol. 22:271-274.
NOVEMBER 1999 E-18 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
APPENDIX F
TRIM.FaTE Computer Framework
The TRIM.FaTE computer framework provides the infrastructure required to conduct and
analyze TRIM.FaTE simulations. The framework allow users to:
Define the issue to be studied, including time period, geographic region, pollutants,
media, and populations of interest;
• Specify and choose algorithms that will be used for simulations;
Select modeling parameters, including emissions sources, characteristics of the
environment (e.g., air temperature and soil permeability), and simulation time step;
• Identify data sets to be used and created;
Execute the simulation;
• Perform sensitivity studies; and
• Export results.
There are two versions of the framework: prototype and Version 1.0. The framework
prototype has served as a testbed for evaluating approaches. It has been designed to allow
changes to be quickly implemented and to allow ideas to be quickly tried. The prototype was
used to conduct the simulations described in this document. Version 1.0 was completed
September, 1999 and will be used for future studies. Version 1.0 incorporates lessons learned
from Prototypes I through V with the addition of features that increase the usefulness of the
system, such as management of multiple modeling scenarios, portability between Windows and
UNIX, and improved ease of use and robustness.
This description of the TRIM.FaTE computer framework generally covers both the
prototype and Version 1.0 with indications where necessary that descriptions apply to only one of
the implementations. Additional information about the architecture and design of TRIM.FaTE
Version 1.0 can be found in Fine et al. (1998a,1998b).
F.I SOFTWARE ARCHITECTURE
Bass et al. (1998) provide the following definition:
The software architecture of a program or computing system is the structure or
structures of the system, which comprise software components, the externally
visible properties of those components, and the relationships among them.
The prototypes and Version 1.0 have different architectures, so they are described separately.
NOVEMBER 1999 F^i TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
F.I.I ARCHITECTURE OF THE PROTOTYPES
The prototypes are implemented in an object-oriented manner, with almost all important
quantities implemented as objects/classes. These include:
• parcels;
• volume elements;
• compartments;
• chemicals;
• links;
• algorithms;
• parameters (input parameters and calculated parameters);
• runs; and
• projects.
In the prototypes, a project is constructed in a hierarchical fashion: first a parcel is
created, then volume elements can be added "to" the parcel, and then compartments can be added
to the volume elements. Links can be created manually or can be automatically determined
based on the spatial adjacency information of the project.
When a run is initiated, the needed transition matrices, source term vectors, and initial
condition vector are constructed from the modeled system. This process utilizes the link
topology and algorithms associated with each link, in addition to the source specified for
particular compartments and the implied source terms calculated based on any boundary air
concentrations specified. The transition matrices and associated source term and initial condition
vectors are used in successive calls to the differential equation solver (i.e., LSODE), after which
the predicted chemical mass in each compartment is available.
An expression evaluator is also included within the design of the prototypes. This is used
to evaluate almost all algorithms and other needed calculated quantities (e.g., distribution
coefficients in soil for organics, which are calculated from properties of the chemical and the soil
compartment). The expressions themselves are stored as strings, using an object-oriented syntax
consistent with the overall object model used. These expressions are "compiled" when a run is
performed, with the objects needed to calculate each expression obtained for subsequent
calculation. This allows flexible naming of variables and the creation of numerous intermediate
terms that can help provide insight into the finer details of a particular run. Further, it
significantly improves the quality of output reports that can be produced. For example, detailed
reports can be generated that show the exact equations used to calculate a given quantity, as well
as the values of the terms used in its calculation. The successful implementation of such a
system in the prototype made it possible to seriously consider, and ultimately decide upon,
implementing a similar capability in the TRIM.FaTE Version 1.0.
F.1.2 VERSION 1.0 ARCHITECTURE
As shown in Figure F-l, the TRIM computer system architecture is complex but flexible,
allowing it to be applied in developing each of the different TRIM modules. The architecture
NOVEMBER 1999 f3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
components used to describe TRIM are classified as those that primarily provide (1) functionality
(rectangles), and (2) those that primarily provide data (ovals). However, each of the components
except for external data sources provide both functionality and data. The architectural
components that are implemented to some degree in Version 1.0 are depicted with shadows.
This figure is designed to represent the relationships within the TRIM computer framework,
rather than the data flow within the system. Therefore, the word along an arrow forms a sentence
where the verb on the arrow connects the two architecture components at the end of an arrow.
For example, in the upper left hand corner of the figure, the TRIM Core "invokes" Analysis and
Visualization Tools. Each of the TRIM components shown in Figure F-l are described below.
F.l.2.1 TRIM Core
The TRIM Core primarily provides services required by multiple TRIM components or to
integrate those components. The following functions are provided by the Core:
• A mapping tool that shows TRIM spatial objects, such as volume elements, and arbitrary
supplemental information supplied by the user, such as soil types. The map display will
also allow users to specify the X-Y extent of TRIM.FaTE volume elements.
• Coordination of TRIM graphical user interface components. This includes allowing the
user to invoke TRIM modules, such as TRIM.FaTE, and maintaining lists of open
windows.
Allowing users to edit and view property values, where a property is an attribute (e.g.,
molecular weight) that describes an entity simulated by a model, such as a compartment
or volume element. Properties include air temperature, scavenging coefficients, and
chemical reaction rates.
• Management of plug-in data importers and exporters.
• Calculation of sensitivity and uncertainty using TRIM models (not supported in Version
1.0).
Utility functions used by TRIM modules, such as routines to assist with data storage and
retrieval.
NOVEMBER 1999 F-3 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
Figure F-l
TRIM Computer System Architecture
External Data
Source*
Analysis and
Visualization
Tools
Time-
Stepped
Data Files
TRIM Cora
Map display
User interface coordination
Property editor
Data Input/Output
Plug-in management
Sensitivity & uncertainty calculations
Utility functions
Importersand
Exporters
Projects
OutdoorEnvironment
OutdoorEnvironment editor
Affected populations
Run characteristics
Uses results from
Shadow indicates that
some functionality would
be present in Version 1.0.
Primarily
Functionality
Note- Each arrow
summarizes either how
TRIM components
interact or how one
component is a special
case of another.
NOVEMBER 1999
F-4
TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
F.I.2.2 Project
All information pertinent to an environmental study is stored in a "project." Each project
is also responsible for displaying the information it contains and allowing the user to change the
information, in some cases relying on a TRIM Core functionality such as the property editor. A
project can contain one or more "scenarios," where each scenario contains a description of the
outdoor environment being simulated, populations being studied, and model parameters, such as
the simulation time step.
F.l.2.3 TRIM Modules
Each TRIM module, such as TRIM.FaTE, provides simulation or analysis functionality.
Where required, they also provide specialized graphical user interfaces to support their
functionality. Version 1.0 includes the TRIM.FaTE module. Future TRIM versions will have
support for additional TRIM modules.
TRIM.FaTE uses a number of algorithms that compute chemical transfer coefficients
between and transformation coefficients within conceptual compartments. As new chemicals
and ecosystems are studied, new algorithms will be required. To address this need, users will be
able to add algorithms, which are stored in libraries and projects. The algorithms that are stored
in libraries can be applied to various projects.
F.I.2.4 Libraries
A substantial amount of relatively static information is required to conduct studies of
multimedia fate and transport and effects on selected populations. For instance, the measured
properties of chemicals change infrequently. Also, the boundaries of a study region might stay
constant for years. Users can store such information in TRIM object libraries. They can then
easily reuse selected information from a library in future projects. When information from a
library is used in a project, a copy is made of the information, which protects the project from
future changes to the library.
F.I.2.5 External Data Sources, Importers, and Exporters
It will be common for TRIM users to access or create data sets beyond TRIM projects.
Some data sets may be too large to be conveniently stored in projects, while other data sets
already exist in non-TRIM formats. TRIM provides several methods for accessing external
information. The TRIM Core accepts user inputs and will read and write data in native TRIM
files. The format of these files is based on the Environmental Decision Support System/Models-
3 Input/Output Applications Programming Interface (I/O API) (Coats 1998). The I/O API format
can be easily read and written from several programming languages, is platform-independent, is
suitable for large data sets, is self-describing (i.e., contains information about variables and time
periods contained in the file), and is computationally efficient.
TRIM also allows users to plug in data importers and exporters. Data importers read non-
TRIM data sets and set appropriate TRIM properties. For example, an importer could read files
NOVEMBER 1999 F^5 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
containing measurements of surface air temperature and set properties in ground-level TRIM air
domains. Data exporters provide TRIM results in a form that is suitable for use by another
program or for interactive review. This could include comma-delimited files that could be
imported into a spreadsheet and tabular results for people to review.
F.I.2.6 Analysis and Visualization Tools
Version 1.0 includes no analysis and visualization tools. Instead, simulation results can
be easily exported to Excel or other analysis packages, In the future, TRIM will include some
analysis and visualization capabilities and might allow additional capabilities to be developed
and plugged-in by users.
F.2 IMPLEMENTATION APPROACHES
The computer framework has been developed using an object-oriented approach. There
has been much discussion in the software engineering literature (e.g., Booch 1993) on the
benefits of this approach, including increased software extensibility, reuse, and maintainability.
The essence of object-oriented software development is that concepts, such as a volume element,
are represented as a unit that contains internal data (e.g., the boundaries of a volume element) and
operations on that data (e.g., compute volume) and that one class of objects (e.g., volume
element with vertical sides) can be a specialization of another class of objects (e.g., volume
element). Being able to specialize classes of objects allows general functionality to be shared by
several specialized classes. TRIM'S view of the outdoor environment (with volume elements
that contain compartments) and the development of associated graphical user interfaces are well
suited for an object-oriented treatment.
TRIM is being developed in an iterative manner. The major components and
responsibilities of a class of objects are understood before implementation, but some details may-
be worked out as implementation proceeds. Graphical user interface mock-ups and significant
new capabilities are shown to potential users before implementation begins. During
implementation, the design is modified as needed. This user-oriented development approach
helps highlight potential problems before undesirable approaches become embedded in the
system. Furthermore, the object-oriented, open-ended structure of TRIM is intended to make
future changes and additions a relatively simple process.
For Version 1.0 of TRIM, simpler and/or more reliable approaches were used in
preference to faster and/or less resource-intensive approaches. In cases where simple approaches
will not have adequate performance or will significantly limit the potential for future changes,
more complex approaches will be used. As time permits, operations that cause noticeable speed
or resource problems will be optimized.
F.3 IMPLEMENTATION LANGUAGE
Due to the different objectives of the framework prototypes and Version 1.0, different
development languages were chosen. The rationale for each choice is described below.
NOVEMBER 1999 F-6 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
F.3.1 PROTOTYPES
Microsoft's Visual Basic was used as the primary tool with which to implement the
prototypes. This was due to a number of factors:
Ease of use with Microsoft Excel, which all members on the team had (for early
prototypes);
• Object-oriented features of language, while limited1, simplify a dynamic, iterative
architecture development cycle; and
Straightforward to call needed Fortran codes (e.g., differential equation solver, linear
equation solver, triangulation).
F.3.2 VERSION 1.0
The Version 1.0 computer framework was developed primarily, but not entirely, in the
Java programming language. Some parts of TRIM.FaTE, such as the differential equation solver,
and other TRIM models, such as the exposure model, are implemented in FORTRAN, and other
parts, such as the polygon overlay algorithm, are implemented in C.
Advantages of using Java include the following.
• Java code is portable across different hardware and operating systems. This is especially
important for graphical user interfaces, which will comprise a large fraction of the TRIM
code and which can be difficult to develop for multiple platforms.
• Java offers a good combination of speed of development, robustness, and support for
object-oriented designs.
• Java is supported by multiple vendors. This often leads to competitive pressures to
improve development tools, and it reduces the likelihood that one vendor's product
strategy or financial problems will cripple TRIM development.
• Java provides built-in support for multithreading, which allows multiple operations to
proceed simultaneously, and networking.
The disadvantages of using Java include the following:
• Java programs typically execute more slowly than programs written in C++ or BASIC.
As the technologies for compiling and executing Java programs advance, the speed
penalty for using Java should decrease.
' The primary limitation is that Visual Basic does not support inheritance. However, it does support
polymorphism (an object/class can implement any number of interfaces), which is utilized to a large degree to
simplify the logic of the programming.
NOVEMBER 1999 F^T TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
Fewer plug-in components (e.g., mapping tools) and libraries (e.g., matrix manipulation)
are available for Java than there are for languages such as C++ or BASIC on Windows,
but the number of Java components available is continuing to grow.
• Java development tools are not as mature as tools for other languages, but that situation is
improving.
F.4 EMBEDDED TRANSFER ALGORITHMS
As described elsewhere, TRIM.FaTE allows users to specify and choose algorithms that
compute chemical transformation and transfer factors. This provides significant flexibility to
describe different pollutants and environmental systems. However, some transfer algorithms are
too complex to be represented as user-entered formulas. These algorithms are described below.
F.4.1 WIND SPEED BETWEEN AIR COMPARTMENTS
The wind speed from one air compartment to another is calculated as the sum of the
transport and dispersive/lateral wind speed. The methods used to calculate these quantities are
implemented in subroutines within the source code, rather than through the use of expressions for
the expression evaluator. Details on these methods can be found in Section 3.1 of the
TRIM.FaTE TSD Volume II.
F.4.2 INTERFACIAL AREA BETWEEN VOLUME ELEMENTS
The interfacial area shared by volume elements is used frequently (e.g., for advective and
diffusive transfers). This is calculated by subroutines in the source code itself. In the prototype.
each side of a volume element that might intersect another volume element is triangulated (in
conjunction with a dynamic link library for triangulation). Next, intersection of the
triangulations is computed. Version 1.0 uses a more specialized but faster approach that takes
advantage of current restrictions on the structure of volume elements (sides must be vertical and
tops and bottoms horizontal). The X-Y projections of side-by-side volume elements are
examined for line segment overlap. The length of the overlap multiplies the extent of vertical
overlap. The interfacial area for volume elements that are stacked vertically is computed by
intersecting the polygons that represent the X-Y projections of the volume elements and then
computing the area of the resulting polygon. When more general shapes are permitted for
volume elements, a more general calculation, such as the triangulation approach, will be
incorporated into Version 1.0.
F.5 REFERENCES
Bass, L., P. Clements, and R. Kazman. 1998. Software architecture in practice. Reading, MA:
Addison-Wesley.
Booch, G. 1993. Object-oriented analysis and design with applications. Redwood City,
California: The Benjamin/Cummings Publishing Company, Inc.
NOVEMBER 1999 F-8 TRIM.FATE TSD VOLUME I (DRAFT)
-------
APPENDIX F
TRIM.FATE COMPUTER FRAMEWORK
Coats, C. 1998. The EDSS/Models-3 I/O API. http://www.iceis.mcnc.org/EDSS/ioapi/.
Fine, S.S., A. Eyth, andH. Karimi. 1998a. The Total Risk Integrated Methodology (TRIM)
Computer System Architecture. Research Triangle Park, NC: MCNC-North Carolina
Supercomputing Center. November.
Fine, S.S..A. Eyth, andH. Karimi. 1998b. The Total Risk Integrated Methodology (TRIM)
Computer System Design. Research Triangle Park, NC: MCNC-North Carolina
Supercomputing Center. November.
NOVEMBER 1999 F-9 TRIM.FATE TSD VOLUME I (DRAFT)
-------
|