EPA530-D-03-001C
July 2003
SAB Review Draft
/ A Multimedia, Multipathway, and
Multireceptor Risk Assessment
(3MRA) Modeling System
Volume III: Ensuring Quality of the
System, Modules, and Data
-------
EPA530-D-03-001C
July 2003
SAB Review Draft
Multimedia, Multipathway, and
Multireceptor Risk Assessment
(3MRA) Modeling System
Volume III: Ensuring Quality of the
System, Modules, and Data
prepared by
U.S. Environmental Protection Agency
Office of Research and Development Office of Solid Waste
National Exposure Research Laboratory Washington, D.C.
Athens, GA
and
Research Triangle Park, NC
-------
This document is the third volume of a five-volume set. Volume I describes the
conceptual design, scientific rationale, and supporting data that are the foundation for the 3MRA
modeling system. Volume II describes the data developed and used to run the 3MRA modeling
system. This volume describes the approach to quality assurance, including verification and
validation activities ranging from extensive peer reviews to multimedia model comparisons.
Volume IV describes the methodology used to evaluate sensitivity of model parameters and
characterize different types of uncertainty in the 3MRA modeling system. Volume V describes
the technology design and includes a user's guide.
-------
Table of Contents
Table of Contents
Section Page
List of Figures vii
List of Tables ix
1.0 Introduction 1-1
1.1 3MRA Modeling System Development 1-1
1.2 Overview of the 3MRA Modeling System 1-2
1.3 Organization of This Document 1-3
2.0 Overall Approach to Ensuring Quality of the 3MRA Modeling System 2-1
2.1 Verification and Validation of the 3MRA System 2-3
2.2 Verification and Validation of the Modules 2-4
2.2.1 Concept Review and Code Design 2-6
2.2.2 System-Level Control of Module Input and Output Definitions 2-6
2.2.3 Version Control of Module Code 2-7
2.2.4 Internal and External Code Testing 2-8
2.2.5 Peer Review 2-9
2.3 Verification and Validation of the Site-based Data 2-11
2.3.1 Verification of Site-based Data 2-11
2.3.2 Validation of Site-based Data 2-11
2.4 Verification and Validation of the Chemical Properties Data 2-11
2.4.1 Verification of Chemical Properties Models 2-11
2.4.2 Validation of Chemical Properties Models 2-11
3.0 Ensuring Quality of the 3MRA Modeling System 3-1
3.1 Verification of the Modeling System 3-1
3.1.1 Development Team Communications 3-2
3.1.2 Top-Down Design Approach 3-2
3.1.3 Quality Assurance and Testing Strategy 3-2
3.1.4 Software Archiving and Distribution 3-5
3.2 Validation of the System 3-7
3.2.1 Problem Definition 3-7
3.2.1.1 Site selection 3-7
3.2.1.2 Problem statement and modeling goals 3-11
3.2.2 Site Conceptualization 3-14
3.2.3 Data Collection 3-20
3.2.4 Model Execution 3-20
3.2.5 Interpretation of Results 3-20
4.0 Evaluating Quality of the 3MRA Modeling System Modules 4-1
4.1 Wastewater Source Modules 4-1
4.1.1 Module Description 4-1
4.1.2 Major Module Components/Functionality 4-4
in
-------
Table of Contents
Table of Contents (continued)
Section Page
4.1.3 Summary of Verification 4-5
4.1.4 Summary of Validation 4-6
4.2 Land-based Source Modules and Watershed Module 4-8
4.2.1 Module Description 4-8
4.2.2 Major Module Components/Functionality 4-10
4.2.3 Summary of Verification 4-11
4.2.4 Summary of Validation 4-14
4.3 Air Module 4-17
4.3.1 Module Description 4-17
4.3.2 Major Module Components/Functionality 4-18
4.3.3 Summary of Verification 4-18
4.3.4 Summary of Validation 4-20
4.4 Surface Water Module 4-22
4.4.1 Module Description 4-22
4.4.2 Major Module Components/Functionality 4-23
4.4.3 Summary of Verification 4-24
4.4.4 Summary of Validation 4-25
4.5 Vadose Zone and Aquifer Modules 4-27
4.5.1 Module Description 4-28
4.5.2 Major Module Components/Functionality 4-28
4.5.3 Summary of Verification 4-29
4.5.4 Summary of Validation 4-33
4.6 Farm Food Chain Module 4-35
4.6.1 Module Description 4-35
4.6.2 Major Module Components/Functionality 4-36
4.6.3 Summary of Verification 4-37
4.6.4 Summary of Validation 4-38
4.7 Terrestrial Food Web Module 4-39
4.7.1 Module Description 4-39
4.7.2 Major Module Components/Functionality 4-40
4.7.3 Summary of Verification 4-41
4.7.4 Summary of Validation 4-42
4.8 Aquatic Food Web Module 4-42
4.8.1 Module Description 4-42
4.8.2 Major Module Components/Functionality 4-43
4.8.3 Summary of Verification 4-44
4.8.4 Summary of Validation 4-45
4.9 Human Exposure Module 4-45
4.9.1 Module Description 4-46
4.9.2 Major Module Components/Functionality 4-47
4.9.3 Summary of Verification 4-49
4.9.4 Summary of Validation 4-50
iv
-------
Table of Contents
Table of Contents (continued)
Section Page
4.10 Human Risk Module 4-50
4.10.1 Module Description 4-50
4.10.2 Major Module Components/Functionality 4-51
4.10.3 Summary of Verification 4-51
4.10.4 Summary of Validation 4-53
4.11 Ecological Exposure Module 4-53
4.11.1 Module Description 4-53
4.11.2 Major Module Components/Functionality 4-54
4.11.3 Summary of Verification 4-54
4.11.4 Summary of Validation 4-55
4.12 Ecological Risk Module 4-55
4.12.1 Module Description 4-56
4.12.2 Major Module Components/Functionality 4-57
4.12.3 Summary of Verification 4-57
4.12.4 Summary of Validation 4-58
5.0 Verification and Validation of 3MRA Site-Based Data Collection and Processing ... 5-1
5.1 Site-Based Data Collection Methods 5-1
5.1.1 Conduct Pilot Study 5-4
5.1.2 Establish Spatial Framework/Initial Setup 5-4
5.1.3 Delineate Waterbodies and Watersheds 5-7
5.1.4 Place Human Receptors 5-7
5.1.5 Place Ecological Receptors 5-9
5.1.6 Overlay GIS Coverage 5-10
5.1.7 Process Data for 3MRA 5-10
5.2 Data Verification 5-11
5.2.1 Quality Assurance/Quality Control 5-11
5.2.2 Independent Data Testing 5-15
5.3 Data Validation 5-17
5.3.1 Surface Impoundment Study Data 5-17
5.3.2 Development of SIS Data Sets 5-18
5.3.3 Preliminary Comparisons 5-19
6.0 Verification and Validation of Chemical Properties Models 6-1
6.1 SPARC Chemical Properties Estimator 6-1
6.1.1 Model Description 6-2
6.1.2 Major Model Components/Functionality 6-3
6.1.3 Summary of Model Verification 6-5
6.1.4 Summary of Model Validation 6-5
6.1.4.1 Vapor Pressure 6-5
6.1.4.2 Solubility 6-5
6.1.4.3 Octanol/Water Partition Coefficient 6-7
6.1.4.4 Henry's Law Constant 6-7
v
-------
Table of Contents
Table of Contents (continued)
Section Page
6.2 MINTEQA2 Geochemical Speciation Model 6-7
6.2.1 Model Description 6-8
6.2.2 Major Model Components/Functionality 6-8
6.2.3 Summary of Model Verification 6-9
6.2.4 Summary of Model Validation 6-10
7.0 Summary and Conclusions 7-1
8.0 References 8-1
vi
-------
Table of Contents
List of Figures
Figure Page
1-1 3MRA modeling system design 1-4
1-2 Linkages among the source, fate, transport, exposure, and risk modules of the
3MRA modeling system 1-5
1-3 Organization of the 3MRA modeling system and this document 1-6
2-1 Overall approach to ensure the quality of the 3MRA modeling system 2-3
2-2 Overview of 3MRA module development process 2-5
3-1 Overall approach to ensure the quality of the 3MRA modeling system technology ... 3-4
3-2 Developer testing process for the 3MRA modules 3-4
3-3 Independent testing process for the 3MRA modeling system 3-4
3-4 HoitraChem manufacturing site 3-15
3-5 HoltraChem area photo 3-15
3-6 Aerial view of HoltraChem facility 3-16
3-7 Land use and surface water around HoltraChem facility 3-16
3-8 TRIM surface parcels for HoltraChem facility 3-18
3-9 TRIM air parcels for HoltraChem facility 3-18
3-10 3MRA modeling system surface water reach network for HoltraChem facility 3-19
4-1 Overall approach to ensure the quality of the 3MRA science modules 4-2
4-2 Organization of the 3MRA modeling system modules and this section 4-3
4-3 Information flow for the Wastewater Source Modules in the 3MRA modeling system 4-4
4-4 Information flow for the Land-based Source Modules in the 3MRA modeling system 4-9
4-5 Information flow for the Watershed Module in the 3MRA modeling system 4-10
4-6 Information flow for the Air Module in the 3MRA modeling system 4-17
4-7 Information flow for the Surface Water Module in the 3MRA modeling system .... 4-23
4-8 Surface Water Module test waterbody networks 4-25
4-9 Information flow for the Vadoze Zone and Aquifer Modules in the 3MRA
modeling system 4-28
4-10 Information flow for the Farm Chain Module in the 3MRA modeling system 4-36
4-11 Information flow for the Terrestrial Food Web Module in the 3MRA modeling
system 4-39
4-12 Information flow for the Aquatic Food Web Module in the 3MRA modeling system 4-42
4-13 Information flow for the Human Exposure Module in the 3MRA modeling system . 4-46
4-14 Information flow for the Human Risk Module in the 3MRA modeling system 4-50
4-15 Information flow for the Ecological Exposure Module in the 3MRA modeling
system 4-53
4-16 Information flow for the Ecological Risk Module in the 3MRA modeling system . . 4-56
5-1 Overall approach to ensure the quality of the 3MRA site-based data 5-2
5-2 Site-based spatial overlays 5-3
5-3 Overview of watershed and waterbody layout processing 5-8
5-4. Crown Paper, St. Francisville, LA: Comparison of 3MRA modeling system
(0832903, upper) and SIS (3062, lower) site layouts 5-21
5-5. Cenex Refinery, Laurel, MT: Comparison of 3MRA modeling system
(1230111, upper) and SIS (2418, lower) site layouts 5-22
vii
-------
Table of Contents
List of Figures (continued)
Figure Page
6-1 Overall approach to ensure the quality of the 3MRA chemical property data 6-1
6-2 SPARC-calculated vs. observed log vapor pressure for 747 organic molecules
at 25° C 6-6
6-3 Test results for SPARC calculated log solubilities for 260 compounds 6-6
6-4 Test comparing calculated Kow with measured values 6-7
6-5 MINTEQA2 computes the equilibrium distribution of metals 6-8
viii
-------
Table of Contents
List of Tables
Tables Page
2-1 3MRA Modeling System Peer Reviewers 2-9
3-1 Site Sampling Data for HoltraChem Facility and Nearby Areas and Water Bodies . . 3-12
3-2 Model Estimation Endpoints for Initial HoltraChem Simulations 3-14
4-1 General Requirements for Testing the Aerated Tank Module 4-6
4-2 General Requirements for Testing the Surface Impoundment Module 4-6
4-3 General Requirements for Testing the Land Application Unit Module 4-13
4-4 General Requirements for Testing the Landfill Module 4-13
4-5 General Requirements for Testing the Waste Pile Module 4-14
4-6 General Requirements for Testing the Watershed Module 4-14
4-7 General Requirements for Testing the Air Module 4-20
4-8 Coal-Fired Power Plants used for ISCST3 Evaluation 4-21
4-9 General Requirements for Testing the Surface Water Module 4-24
4-10 Exams Calibration/Validation Case Studies 4-26
4-11 Summary of Verification Activities for the 3MRA Vadose Zone and Aquifer
Modules 4-30
4-12 General Requirements for Testing the Vadose Zone Module 4-32
4-13 General Requirements for Testing the Aquifer Module 4-33
4-14 Summary of EPACMTP Validation Activities 4-33
4-15 General Requirements for Testing the Farm Food Chain Module 4-38
4-16 General Requirements for Testing the Terrestrial Food Web Module 4-41
4-17 General Requirements for Testing the Aquatic Food Web Module 4-45
4-18 Default Pathways Considered by Receptor Type 4-48
4-19 General Requirements for Testing the Human Exposure Module 4-49
4-20 General Requirements for Testing the Human Risk Module 4-52
4-21 General Requirements for Testing the Ecological Exposure Module 4-55
4-22 General Requirements for Testing the Ecological Risk Module 4-58
5-1 Site-Based Data Collection for Representative National Data Set: Summary by
Major Activity 5-5
5-2 Comparison of 3MRA Modeling System Representative National Data Set Data
Collection Activities with Pilot Study Approach 5-6
5-3 Primary Human Receptor Data Sets, Date, and Scale 5-9
5-4 Automated QC Performed on Model Input Database 5-13
5-5 Summary of Results of Independent Testing of Site-Based Data 5-16
5-6 Comparison of Industrial D and SIS Survey Data for Common Sites 5-17
5-7 Comparison of Site-based Data Sources: SIS and 3MRA Modeling System
Representative National Data Set 5-19
5-8 Comparison of Waste Management Data—3MRA Modeling System
Representative National Data Set versus SIS 5-20
6-1 SPARC Physical and Chemical Property Estimations used for the 3MRA
Modeling System 6-2
6-2 Summary of MINTEQA2 Validation Studies 6-11
IX
-------
Table of Contents
List of Appendices
Appendix Page
A Verification and Validation of the Air Module A-l
B Quality Assurance Verification and Validation Tests for the Exposure
Analysis Modeling System - Exams B-l
C Verification and Validation of the EPA's Composite Model for
Transformation Products (EPACMTP), and its Derivatives C-l
Attachment C. 1 Vadose Zone Module Verification Results (1993-1994)
Attachment C.2 Module-Aquifer Verification Results (1993-1994)
Attachment C.3 Results (1993-1994)
Attachment C.4 Vadose-Zone Module Verification Results (1997)
Attachment C.5 Aquifer Module Verification Results (1997)
Attachment C.6 Composite Model Verification (1997)
Attachment C.6 Composite Model Verification (1997)
Attachment C.7 3MRA Vadose-Zone Module Verification Results (1997)
Attachment C.8 3MRA Anlayses Module Verification Results (1997)
Attachment C.9 3MRA Pseudo 3-D Aquifer Module Verification Results (1999)
Attachment C. 10 3MRA Vadoze-Zone Module Verification Results (2000)
D Verification and Validation of the SPARC Model D-l
E Review of Validation Studies Concerning theU.S. EPA Geochemical
Speciation Model MINTEQA2 E-l
x
-------
Section 1.0
Introduction
1.0 Introduction
In 1997, the U.S. Environmental Protection Agency's (EPA) Office of Solid Waste
(OSW) and Office of Research and Development (ORD) began working together on the
development of the Multimedia, Multipathway, and Multireceptor Risk Assessment (3MRA)
modeling system. The 3MRA modeling system is intended to be one of EPA's next generation
of multimedia exposure and risk models to support regulatory decisions. EPA designed the
modeling system specifically to meet the needs of OSW programs, but the 3MRA modeling
system has the flexibility to be used for many EPA applications. In particular, it was designed to
provide risk managers with information, at a national level, on the possible exposure and risk to
human and ecological receptors due to the release of hazardous contaminants from the
management of solid wastes.
This document is the third volume of a four-volume set that describes the 3MRA
modeling system.
¦ Volume I provides an overview of the system, including the reasons for its
development, its conceptual design, the modeling approach, and the underlying
science of each of the 3MRA modeling system component modules;
¦ Volume II describes the data developed and used to run the 3MRA modeling
system.
¦ Volume III describes the verification and validation activities and peer reviews to
ensure quality of the modeling system, modules, and data;
¦ Volume IV provides a discussion of uncertainty and sensitivity.
The rest of this section includes a brief description of the 3MRA modeling system
development process, an overview of the 3MRA modeling system, and the organization of the
rest of this volume.
1.1 3MRA Modeling System Development
EPA's Program office scientists and risk managers in OSW collaborated with EPA
scientists in ORD to develop the 3MRA modeling system. A core team put together EPA's
science plan for developing the system (U.S. EPA, 1999a). The science plan identified OSW
needs and the scientific approach to be taken. These needs were then described in terms of goals
and objectives. One overriding theme was to use existing regulatory (legacy) models where
appropriate, thus relying on the science behind them and the level of acceptance associated with
their previous use. The following legacy regulatory models have been incorporated into Version
1 of the 3MRA modeling system: Industrial Source Complex Short-Term Model, Version 3
(ISCST3), for air dispersion and deposition (U.S. EPA, 1995); EPA's Composite Model for
Leachate Migration with Transformation Products (EPACMTP) (U.S. EPA, 1996a,b,c, 1997a)
1-1
-------
Section 1.0
Introduction
for waste constituent subsurface fate and transport; and EPA's Exposure Analysis Modeling
System II (EXAMS II) (Burns et al., 1982; Burns, 1997) for waste constituent surface water fate
and transport. Each of these legacy models was modified somewhat to provide needed
functionality within the 3MRA modeling system. All other source, fate and transport, food web,
exposure, and risk modules were developed specifically for the 3MRA modeling system based
on sound science and established principles.
The 3MRA modeling system represents the integration of more than twenty-five
independent software components developed by five different software development groups
located across the country. Core teams were formed to oversee the development of each set of
modules. These core teams consisted of OSW and ORD staff working together to design each
module; ensure that adequate test plans were developed; and oversee the testing, verification,
and documentation of the modules. All modules (legacy as well as new modules) were
independently reviewed by external national experts. The core teams were assisted by experts
from several nationally recognized organizations: Batelle Pacific Northwest National Lab
(PNNL); HGL, Inc.; Research Triangle Institute (RTI); and TetraTech, Inc. For the sake of
simplicity, this document attributes all model development, verification, and validation activities
to "EPA" rather than specific EPA offices or contractors.
During the development of the 3MRA modeling system, EPA team was fully aware of
the real world limitations that apply to validating the 3MRA modeling system The 3MRA
modeling system team carried out several activities to ensure the quality fo the modules, data,
and the modeling system. These activities included: peer review of the conceptual development
plan (science plan containing the risk assessment strategy); use of legacy models; internal
independent testing of the modules; external testing of the modules; peer reviews of modules;
public comments on the beta version of the 3MRA modeling system and the model results. This
volume briefly describe these activities which tend to increase the confidence in the system
1.2 Overview of the 3MRA Modeling System
The 3MRA modeling system was developed as a tool to provide risk assessment support
for the evaluation of risks from the management of wastes in waste management systems
(WMUs). The OSW applies risk assessment modeling tools in a variety of situations; one
application is the use of tools to conduct site-based national-level risk assessments to support
rule-making for the identification of hazardous wastes. Consequently, EPA developed the
3MRA modeling system to model environmental settings that are representative of the range of
environmental settings found in the United States, and within this broad range of settings, to
simulate the release, fate, and transport of many chemical constituents in waste undergoing a
range of physical and biochemical processes. Because EPA needs to consider the impacts of a
large number of chemicals, EPA requires a modeling tool that encompasses releases from
WMUs to all media, subsequent fate and transport within those media, uptake in terrestrial and
aquatic food webs, and the potential exposure of specific human and ecosystem receptors to the
chemical constituents and contaminated food items.
The 3MRA modeling system was designed to estimate national distributions of human
and ecological risks resulting from long-term (chronic) chemical constituent release from land-
based and WMUs using a site-based approach. The national distribution of risks is constructed
1-2
-------
Section 1.0
Introduction
by performing "site-based" assessments at a statistically significant number of randomly selected
industrial waste site locations across the United States. The 3MRA framework (US EPA, 1999a)
employs a tiered approach for populating data files for each site characterization and evaluation.
The approach is referred to as "site-based" because the assignment of data values for the site
being simulated occurs according to a tiered protocol. Data values are filled first with data at a
site level. When site data are not available, a statistically sampled value from a geographically
relevant regional distribution of values is used. When a representative regional distribution for
the variable is not available, a value from a national distribution is assigned.
The 3MRA modeling system simulates chemical constituent releases from a WMU to the
various media (air, water, soil) based on the chemical/physical properties of the constituent, the
characteristics of the WMU that is modeled, and the environmental setting (e.g., hydrogeological
conditions and the meteorological region) in which the facility is located. Once released from
the WMU, the contaminant is transported through environmental media and into biological
compartments such as produce, beef, and fish. Human and ecological receptors included in the
simulation may be exposed concurrently to contaminated media and food through multiple
pathways and routes of exposure. For each receptor that is included in the simulation, the 3MRA
modeling system performs risk/hazard calculations based on aggregate exposures modeled
through space and time.
Figure 1-1 provides an overview of the 3MRA modeling system design. As shown in
this figure, the system performs three major functions: (1) the site definition, (2) the multimedia,
multipathway simulation, and (3) the exit level processing. The site definition, as shown in the
figure, includes both the selection of site characteristics from three levels of data, as well as the
estimation and selection of chemical properties. The multimedia, multipathway simulation
includes the execution of all of the science modules linked together to predict the behavior of
chemicals from source release through exposure and risk. The linkages among the science
modules are depicted in Figure 1-2. The exit level processing occurs after the simulation is
completed and consists of two-stage processing of the risk outputs followed by the risk
visualization of exit level distributions. At the top of Figure 1-1, the looping structure used to
conduct national scale analyses is summarized, including the site location loop, the WMU loop,
the number of iterations, and the number of chemicals (which are simulated individually).
The 3MRA modeling system was designed specifically to incorporate Monte Carlo
simulation methods to address both uncertainty and variability in the risk outputs. The sites
currently in the database were randomly selected from solid waste management sites across the
United States to represent the national variability in waste management scenarios and locations.
The methodology for selecting the sites allows for measures of protection to be calculated at the
site level and aggregated over all the sites to develop the national distribution of risks.
1.3 Organization of This Document
The 3MRA modeling system comprises an overall system, 17 separate science modules,
processors, and several databases. Figure 1-3 provides a guide to the organization of this
document. Section 2 provides an overview of the verification and validation of 3MRA modeling
1-3
-------
System User Interface (SUI)
/
Waste Management Facility Loop (201 National Sites)
Waste Management Unit Loop (5 Source Types)
Sampled Input Data Iteration Loop (nr)
Chemical Loop (43 Metals & Organics)
C,.,= Waste stream concentration
Cw Loop
Key
I
| User Interface
II
Data File
o
Processor
~
Database
J. Header Info from SUI
Warnings/Errors to SUI
"V
Site Input Data
~^r
Site Definition
"V
Multimedia Multipathway
Simulation
cw Exit Level Processing
List of Sites
Exit
Level
Processor
II
Site-Based
Database
Multimedia
Multipathway
Simulation
Processor
Exit
Level
Processor
I
Site
Definition
Processor
Regional
Database
Risk
Visualization
Processor
National
Database
List of Chemicals
Chemical
Properties
Database
Chemical
Properties
Processor
Metal
Isotherms
MET
Database
Figure 1-1. 3MRA modeling system design.
-------
Ecological
Risk
Terrestrial Food
Web
Surface
Impoundment
Aquatic Food
Web
Aerated
Tank
Ecological
Exposure
Surface
Water
Watershed
Landfi
Human
Exposure
Waste Pi e
Land
Vadose
Zone
Application
Aquifer
Farm FoodChain
Human
Risk
Sources
Transport
Foodchain
Exposure/Risk
The dashed line indicates that soil concentrations for the local (land-based source) and regional watersheds may be added together to estimate
total soil concentrations for areas (e.g., habitats) that include both regional and local watershed components.
Figure 1-2. Linkages among the source, fate, transport, exposure, and risk modules of the 3MRA modeling system.
-------
Site-based Data
(Section 5)
Site-based Data
Collection Methods
(5.1)
Data Verification (5.2)
Data Validation (5.3)
3MRA Modeling System
(Section 3)
Modules
Chemical Properties
(Section 4)
Models
(Section 6)
• Wastewater Source (Surface Impoundment
and Aerated Tank) (4.1)
• Land-based Source (Landfill, Waste Pile,
• SPARC (6.1)
• MINTEQA2 (6.2)
and Land Application Unit) and Watershed
(4.2)
• Air (4.3)
• Surface Water (4.4)
• Vadose Zone and Aquifer (4.5)
• Farm Food Chain (4.6)
• Terrestrial Food Web (4.7)
• Aquatic Food Web (4.8)
• Human Exposure (4.9)
• Human Risk (4.10)
• Ecological Exposure (4.11)
• Ecological Risk (4.12)
Figure 1-3. Organization of the 3MRA modeling system and this document.
-------
Section 1.0
Introduction
system (including system, modules, and data). Section 3 describes the verification and
validation of the modeling system. Section 4 describes verification and validation of the
component modules. Section 5 describes the verification and validation of the site-based data in
the representative national data set included in the 3MRA modeling system. Section 6 describes
the verification and validation of the models (SPARC and MINTEQ) used to develop some of
the chemical properties. Section 7 summarizes the results of the model evaluation process.
Section 8 provides references cited in this document.
1-7
-------
Section 2.0
Overall Approach to Verification and Validation
2.0 Overall Approach to Ensuring Quality of the
3MRA Modeling System
EPA designed the 3MRA modeling system to be scientifically defensible. To address
comments with regard to scientific defensibility, which included recommendations for peer
review, the Agency implemented specific
steps to build confidence in the system and
ensure that the system would produce a
reasonable estimate of nationwide risk for a
national-level assessment. These activities
are described below:
1. Peer Review. The overall
technical approach and each
science-based module
included in the 3MRA
modeling system were peer
reviewed. Teams of peer
reviewers (at least three per
module) provided critical
feedback about the science-
based modules. More than
45 independent experts
reviewed the science
modules to ensure that the
theoretical concepts
describing the processes
within the release, fate,
transport, uptake, exposure,
and risk components were
adequate representations of
the processes to be
evaluated. The peer review
process is described in
Section 2.2.5.
Verification and Validation
Verification refers to activities that are designed to
confirm that the mathematical framework embodied in
the module is correct and the computer code for a
module is operating according to its intended design so
that the results obtained using the code compare
favorably with those obtained using known analytical
solutions or numerical solutions from simulators based
on similar or identical mathematical frameworks. Or
that the data were properly developed based on the
methods selected for their generation. For the module
code, verification activities would include taking steps
to ensure that code was properly maintained, modified
to correct errors, and tested across all aspects of the
module's functionality. For data generation, example
of verification activities include, determining whether
data definitions (distributions, ranges) in the database
match the data definitions in the 3MRA modeling
system technical background documents.
Validation is conducted after the verification step. It
refers to activities that confirm that the design of a
module provides an accurate representation of the
physical processes it is intended to simulate, or that the
data are accurate and complete. For a module,
validation might include comparing simulation results
with field data or results from other established models
if field data are unavailable. For data generation,
validation might include comparing modeling data for
a site with site-specific data for that location obtained
from a site visit or previously collected data.
2. Verification. All software components and databases underwent a series of tests
to verify that the software and data were performing properly. At the heart of this
protocol was the requirement that each component of the modeling system
include a designed and peer-reviewed test plan that was executed by both the
2-1
-------
Section 2.0
Overall Approach to Verification and Validation
module developer and a completely independent modeler (i.e., someone who did
not participate in the original module development). These procedures, test plans,
test packages, and test results are fully documented and available to the public.
The underlying concept of the verification methodology is that if tests are
designed to challenge all subroutines, and if decision points in the module and
calculations can be duplicated independently of the module, then it can be
concluded that the testing is complete and successful. In addition, the
methodology was predicated on the premise that the looping functions need not
be tested on hundreds of values of the looping parameter. Confirmation of
calculations for a small subset of values demonstrates that the looping functions
in the module are working as intended. Sections 4, 5, and 6 summarize the
verification activities for, respectively, the modules, the site-based data, and the
models used to estimate some of the chemical properties.
3. Validation. True validation of the 3MRA modeling system for a national
application would require validation over the full range of environmental settings
that are relevant to the application. However, determining whether the modeling
system is valid for the full range of settings is not possible: data sets exist that
allow components of the 3MRA modeling system to be validated, but none were
found that would allow validation of the complete and integrated multimedia risk
assessment. Individual modules and data sets were validated if appropriate data
were available. However, instead a comparison study is being conducted using
environmental monitoring field data for mercury from an actual industrial site.
These data represent the relationship between contaminant source and
environmental concentrations; however, the data set is incomplete. This
comparison is ongoing; its current status is described in Section 3. The validation
efforts for the individual modules, the site-based data, and the chemical properties
estimation models are described in Sections 4, 5, and 6, respectively.
4. Model Comparison Study. The 3MRA modeling system as a whole is
undergoing a comparison analysis with EPA's Total Risk Integrated Methodology
(TRIM), which is currently under development. The objective of the model
comparison effort is to increase confidence that the 3MRA modeling system
produces estimates consistent with other multimedia models. This comparison is
ongoing; its current status is described in Section 3.
5. Representative National Data Set. A comprehensive data collection approach
was developed to parameterize the modeling system for 201 sites in accordance
with the site-based approach described in Volume II. The data collection plan
described the general collection methodology for the major types of data (for
example, facility location, land use, soil characteristics, and receptor locations),
including quality assurance and quality control procedures and references for data
sources. The resulting data set was used for the modeling system evaluation
efforts. The data collection approach for the representative national data set is
described in more detail in Volume II.
2-2
-------
Section 2.0
Overall Approach to Verification and Validation
6. Sensitivity and Uncertainty Analyses. As the level of assessment complexity
grows, it becomes both more difficult and more important to establish
comprehensive and quantitative expressions of uncertainty. Uncertainty analysis
is a relatively undeveloped scientific discipline for high order models such as the
3MRA modeling system. Thus, a formal program focusing on sensitivity and
uncertainty analysis for high-order modeling systems was initiated by EPA. The
initial focus of this program is the investigation of parameter sensitivities and
system uncertainties within the 3MRA modeling system. A Supercomputer for
Model Uncertainty and Sensitivity Evaluation (SuperMUSE), comprising 160
PCs, was developed to allow exhaustive experimentation with the 3MRA
modeling system in Monte Carlo mode. The modeling system is suited to
facilitate this new generation of uncertainty analysis and tool development. These
efforts are not only laying a sound foundation for this research but also the
sensitivity work effort so far has helped in ensuring quality of the modeling
system. These analyses are ongoing; its current status is described in more detail
in Volume IV.
Figure 2-1 provides an overview of the quality assurance approach for the 3MRA
modeling system. The rest of this section provides an overview of verification and validation for
the overall system (Section 2.1), the modules (Section 2.2), the site-based data (Section 2.3), and
the chemical properties data (Section 2.4).
2.1 Verification and Validation of the 3MRA System
After the development of the components of the 3MRA modeling system, the first step
was the verification and validation of each component individually. After components satisfied
those tests, they were placed in the complete 3MRA modeling system to confirm their
functionality in the context of the whole system. These system integration tests required that
components operate correctly while executing within the full system context, that is, running the
full set of 3MRA modeling system site/waste management unit combinations for the
contaminants of interest and the iterations of the Monte Carlo simulation. The 3MRA
representative national data set of 201 sites, executed within a Monte Carlo simulation, tested the
robustness of science models in a manner not previously possible. Even legacy codes that had
more than a decade of wide use experienced environmental conditions that caused unstable
numerical solutions.
The system integration tests were initially conducted on single PCs. Computer time for
the 3MRA modeling system was typically on the order of a few minutes per site-based
simulation. Thus, a single PC might run continuously for several days before completing a
simulation for all combinations of sites and WMUs related to a single contaminant. When runs
for multiple contaminants were required, several PCs were used, each running a subset of the full
list of required simulations. However, with the ready availability of faster PCs with more and
faster memory, the runtimes are expected to be shorter. Recently, the EPA has developed
software for distributing the computational load across the Super-computer for Model
Uncertainty and Sensitivity Evaluation (SuperMUSE) cluster and retrieving output files. While
the SuperMUSE was intended as a research tool for investigating uncertainty and sensitivity
related to the 3MRA modeling system, it also serves as a final testing ground for the software
2-3
-------
Section 2.0
Overall Approach to Verification and Validation
System
Module Development
Site-based Data
Collection
Chemical Properties Data
Development
Verification
Validation
QA/QC is an integral part of all development and revision activities.
Integrated System
Testing
Sensitivity and
Uncertainty Analysis
Figure 2-1. Overall approach to ensure the quality of the
3MRA modeling system.
itself. Before Version 1.0 of the 3MRA modeling system was released, the SuperMUSE had
successfully executed on the order of one million individual 3MRA modeling system
simulations.
2.2 Verification and Validation of the Modules
Some of the 3MRA modules have been adapted from established legacy codes, which
have undergone extensive public review, been tested and validated, and been used in
rulemakings. They include the Air Module (ISCST3), the Vadose Zone and Aquifer Modules
(EPACMTP), and the Surface Water Module (EXAMS). Other modules were developed based
on sound science, established methodologies, and accepted QA/QC procedures. Some of those
modules also use components of established codes (e.g., the Wastewater Source Modules use
CHEMDAT8 and EPACMTP). All modules were subjected to a consistent verification and
testing process (see Figure 2-2). The verification steps included
1. Concept Review and Code Design. The module conceptual design was
discussed among the project team members, developed, and reviewed. The
conceptual design included a determination of the module's purpose, essential
2-4
-------
Section 2.0 Overall Approach to Verification and Validation
functionality, and inputs and outputs. Pseudocode
was developed based on the conceptual design
and reviewed by senior project team members.
Concept review and code design are discussed in
more detail in Section 2.2.1.
2. System-Level Control of Module
Input and Output Definitions.
Most module data inputs and
outputs are managed by the 3MRA
Framework for Risk Analysis in
Multimedia Environmental
Systems (FRAMES). The system
design imposed specific
requirements for data definition to
ensure that each input and output
variable was defined and used
consistently with respect to data
type, acceptable range, and units.
Input and output control is
discussed in more detail in
Section 2.2.2.
3. Version Control of Module Code.
The program code for each module
went through numerous updates
during development and refinement
to address errors or add
functionality. EPA used version
control software to maintain a
record of program code updates
and to ensure that the latest version
of the code was used to generate
new executable files. Version
control is discussed in more detail
in Section 2.2.3.
4. Preliminary Testing. Once the
code for a module was fully
developed, module developers
conducted preliminary tests on
each module individually to
determine whether the code was functioning as expected. In general, these
internal tests, which were not documented, were used to identify and resolve
errors in the code. Code testing is discussed in more detail in Section 2.2.4.
Conduct system tests to identify and resolve
incompatibilities between modules.
Conduct test runs across all sites, chemicals, and unit
types to identify remaining module errors
Develop computer code, using Microsoft Visual Source
Safe to maintain version control and to track and
document code changes
Conduct internal verification tests
Conduct external verification tests
Develop list of module inputs and outputs, including
parameter names, data types, dimensionality units, and
min/max values
External independent peer review
Develop conceptual design document and
submit it for review.
Derive scientific formulation of model, including
derivation of key equations and development of
pseudo-code to guide code design
Figure 2-2. Overview of 3MRA module
development process.
2-5
-------
Section 2.0
Overall Approach to Verification and Validation
5. Preliminary System Tests and Production Runs. When the code for all
modules and the system-level software were complete and data sets were
available, EPA ran the complete system, noted module errors, maintained a list of
errors generated by this testing, and tracked new versions of modules or data that
resolved these errors. Code testing is discussed in more detail in Section 2.2.4.
6. Internal Verification Testing. For each module, the module developers
developed and implemented a module test plan to rigorously determine whether
the program was functioning as designed. External project team members
reviewed test plans prior to implementation. This testing usually involved
independent verification of calculations. For some modules, internal verification
testing included limited sensitivity analysis to determine if the module responded
as expected to changes in specific input parameters. The inclusion of such
sensitivity analyses varied by module, based on the judgment of the test plan
developer. Code testing is discussed in more detail in Section 2.2.4.
7. External Verification Testing. For each module, external project team members
repeated the internal verification testing to confirm that the modules operated as
expected. Code testing is discussed in more detail in Section 2.2.4.
8. External, Independent Peer Review. EPA documented the conceptual design
and calculations for each module in background documents, which were
independently peer reviewed. The module developers reviewed the peer review
comments; based on these comments, the module developer, in conjunction with
EPA, determined whether a model change was needed to address the comment,
and, if so, revised the module. Where necessary, internal verification testing was
repeated to verify that the module passed all tests and to ensure that test packages
reflected the latest versions of the executable code and of the input and output
data files. The peer review process is discussed further in Section 2.2.5.
The steps related to the module design, input/output control, code version control, code
testing, and peer review are described in more detail below.
2.2.1 Concept Review and Code Design
Initially, EPA developed and documented the conceptual design and scientific basis for
each module for review and discussion by the project team. The documentation generally
included a description of the module, background information supporting the module design, key
assumptions in the module development, and the mathematical formulation of the module and
relevant derivations. The module documentation became the basis for pseudocode that described
how the computer program for the module would be designed. The programmers then used the
pseudocode to develop the operating code for the module.
2.2.2 System-Level Control of Module Input and Output Definitions
The module background documents included information about the input and output
variables for each module. In general, inputs and outputs of each module are managed through
2-6
-------
Section 2.0
Overall Approach to Verification and Validation
the 3MRA modeling system input/output dynamic link library (DLL) routines. The design of
these routines includes several inherent quality assurance/quality control (QA/QC) checks.
Specifically, the routines require that data be read from or written to ASCII flat data files
containing data in a defined format. This file may be either a site specification file (SSF) or a
global results file (GRF). SSFs are input files generated by the 3MRA modeling system from the
system database as part of the site definition process. GRFs are output files from the 3MRA
modeling system modules that are used by either subsequent modules as input files or the system
to generate the risk results. Each SSF and GRF has an associated data dictionary (DIC) file.
Each DIC file lists each parameter in the associated SSF or GRF and specifies the parameter
name, dimensionality, data type (string, floating point, integer, or logical), units, minimum and
maximum allowable values, and other parameters on which the parameter is dimensioned. Each
parameter entry in the SSF or GRF includes the parameter data type, dimensionality, and units.
The 3MRA modeling system reports an error if there is any inconsistency between an SSF or
GRF and the associated DIC file, or if a data value is outside of the range determined by the
minimum and maximum values specified in the DIC file.
The 3MRA modeling system input/output DLLs include functions for reading data from
SSFs and GRFs. The functions are specific to the dimensionality and data type of the parameter,
and the function calls include the parameter name and units. The 3MRA modeling system
reports an error if there is any inconsistency between the input/output function call and the
associated parameter specification in the SSF or GRF and the DIC file.
In a limited number of cases, module data inputs and outputs are not managed through
the input/output DLLs. These cases include the reading of meteorological data files, writing files
used as inputs for ISCST3, and the reading of ISCST3 output files prior to postprocessing these
data to generate the Air Module GRFs. Because these input/output operations do not include the
inherent checks of the 3MRA modeling system input/output DLLs, the module developers
checked these input/output operations manually in informal tests to ensure they were operating
properly.
2.2.3 Version Control of Module Code
EPA wrote the module code in C++ using the Microsoft Visual Studio integrated
development environment. Several programmers developed, tested, and corrected the modules
over a period of months. To maintain version control and a record of changes made to the
modules, EPA maintained code for each module in Microsoft Visual Source Safe (VSS). VSS
maintains project code in a central database on a server. Each developer can download a
complete set of the current code from VSS. In addition, a developer can "check out" individual
project files, make modifications, and check the file back in to VSS. During the time the
developer has the file checked out, other developers are prevented from checking out the file or
modifying it. When a file is checked back in, the developer can include comments with the new
file. VSS maintains a record of the old and new files, and allows the developer to view the
history of all changes made to a file and compare different versions of a file to determine the
differences in the code between them. If necessary, the current version of a file can be discarded
and the project can be "rolled back" to an earlier version of the file. The programmers used VSS
to obtain the most current version of a module, ensure that only one developer at a time was
2-7
-------
Section 2.0
Overall Approach to Verification and Validation
implementing changes to a file, and examine the history of code changes to determine whether
specific changes had been implemented.
2.2.4 Internal and External Code Testing
The overall purpose of internal and external independent testing was to determine
whether the codes for the various modules were operating as designed. This subsection describes
the internal testing and verification process.
Errors in codes were identified in several ways. Errors in code syntax were found and
corrected as the modules were being compiled. In general, once a module or functional
component of a module was developed and successfully compiled, the developer tested the code
informally to determine whether the code was operating as designed. This testing involved a
number of test cases. When anomalies were found, the developer traced through the code in
"debug" mode, tracking the flow of program execution and examining the data values stored in
specific registers to determine the potential source of the error. To facilitate module testing, the
programmers incorporated output statements into the code to save the values of specific internal
program parameters to an output file for review.
The compatibility of the module with the overall 3MRA modeling system was
determined through a series of system test runs. This process identified inconsistencies between
modules, such as different variable names or passing different variables than those expected by
subsequent modules. Once these inconsistencies were resolved, and the entire 3MRA modeling
system was operating consistently, the developers conducted thousands of runs across a wide
range of contaminants, sites, and unit types. The system developers maintained an error list to
track errors uncovered in these test runs. This error list documented the test data that generated
the error and the module in which the error occurred. Errors were addressed on a case-by-case
basis to identify the cause of the error and implement solutions. In some cases, an error in a
module was traced to a problem with the input data from the 3MRA modeling system database
or generated by a prior module. Otherwise, the developers traced and corrected the error in the
module and reran the case that generated the error to ensure that the error had been satisfactorily
resolved. In some cases, the developer conducted a limited number of informal additional runs
to confirm that the change had resolved the error. Once the developer was satisfied that the error
had been resolved, the developer then submitted the new version of the program to the 3MRA
system managers for inclusion in the next version of the system software. The 3MRA system
managers maintained a record of module errors and the program version that resolved each error.
Once the 3MRA modeling system was stable and generally operating without errors, each
3MRA modeling system module was tested individually to determine whether it was operating
as designed. The testing protocol included internal verification tests conducted by the module
developers, followed by independent tests conducted by another organization on the project
team. For the internal tests, module developers wrote a test plan for each module; this test plan
was reviewed by an external reviewer. The test plan specified the inputs and outputs of the
module, the key module operating requirements, and specified tests that would be used to check
the compliance of the module with the specified requirements. Several of the tests involved
comparing module results with those generated independently from hand calculations or
spreadsheet models. Other tests involved comparing module sensitivity to specific input
2-8
-------
Section 2.0
Overall Approach to Verification and Validation
parameters with the behavior expected based on the model design. Internal tests were
documented in an internal verification test report, which described the test and results, provided
supporting documentation to demonstrate that the program had passed the test, and also provided
any supporting comments regarding how the test was conducted.
After a module passed its internal verification test, the developer sent the module and the
test plan to another organization on the team (that was not involved in the development of the
module) for further review and replication of the verification tests to confirm that the module
completed each test successfully. The external verification tests were documented in technical
memoranda.
2.2.5 Peer Review
The peer review process followed EPA's protocol for peer reviews (Science Policy
Council Peer Review Handbook; U.S. EPA, December 2000) and was used to help ensure
scientific defensibility of the 3MRA modeling system and its components. The 3MRA modeling
system components were independently peer reviewed by the reviewers shown in Table 2-1.
The peer reviewers were independently selected from a national database of experts by an
outside firm that was not involved in the development or testing of the 3MRA modeling system
in any way. EPA provided a list of areas of expertise needed to review each module, and the
consulting firm then selected the reviewers from the database, based on their availability.
Flexibility was built into the 3MRA modeling system design in several ways. First, the
modeling system was designed in modular format so any component could be replaced if better
science became available. The implementation of the modeling system allowed for varied
temporal spatial scales to be user-specified through the data input files. This enhanced the
overall efficiency of the development of the system consistent with the best available science
methodologies.
Table 2-1. 3MRA Modeling System Peer Reviewers
Component
Year
Reviewer
Affiliation
Science Plan
1998
Dr. Yoram Cohen
University of California, Los Angeles
Dr. Paul F. Deisler, Jr.
Independent Consultant
Dr. Mitchell J. Small
Carnegie Mellon University
Surface Impoundment and Aerated Tank
Modules
1998
Dr. Patricia Culligan
Massachusetts Institute of Technology
Dr. Wade Hawthorn
Washington State University
Dr. Michael Overcash
North Carolina State University
Landfill, Waste Pile, Land Application Unit,
and Watershed Modules
1999
Dr. Anita R. Bahe
President, LYNX Group, Ltd.
Dr. Kirk Brown
Texas A&M University
Dr. Wlliam Inskeep
Montana State University
Dr. Clyde. Munster
Texas A&M University
Mr. Wlliam Norris
Virginia Dept. Of Environmental Quality
Mr. Robert Wyatt
President, R. J. Wyatt & Associates
(continued)
2-9
-------
Section 2.0
Overall Approach to Verification and Validation
Table 2-1. (continued)
Component
Year
Reviewer
Affiliation
Air Module (ISCST3)3
1998
Dr. Steven Hanna
George Mason University
Dr. Fred Mogolesko
President, M & L Environmental Consultants
Dr. Bruce Turner
Sr. Tech. Associate, Trinity Consultants
Surface Water Module (Exams ll)a
1999
Dr. Mustafa Aral
Georgia Institute of Technology
Dr. Anthony Donigian
President, AQUA-TERRA Consultants
Dr. Wlbert Lick
University of California, Santa Barbara
Vadose Zone and Aquifer Modules
(EPACMTP)3
1999
Dr. Craig Forster
University of Utah
Dr. M. Akram Hossain
Washington State University
Dr. Carl Mendoza
University of Alberta
Dr. Frank Schwartz
Ohio State University
Farm Food Chain Module
2001
Dr. Donald Mackay
Trent University, Ontario
Dr. Lee Shull
Principal, Montgomery Watson Harza
Dr. Curtis Travis
Principal, Quest Technologies
Terrestrial Food Web Module
2001
Dr. Anne Fairbrother
Sr. Ecotoxicologist, Parametrix, Inc.
Dr. Robert Pastorok
Managing Scientist, Exponent
Dr. Bradley Sample
CH2M Hill
Aquatic Food Web Module
2001
Dr. Lawrence Barnthouse
LWB Environmental Service, Inc.
Dr. Frank Gobas
Simon Fraser University
Dr. Paul Jacobson
Langhei Ecology, LLC
Human Exposure and Risk Modules
2000
Dr. James Butler
Argonne National Laboratory
Dr. Wlliam Kastenberg
University of California at Berkeley
Mr. Stephen Washburn
ENVIRON International Corporation
Ecological Exposure and Risk Modules
2000
Dr. Anne Fairbrother
Sr. Ecotoxicologist, Parametrix, Inc.
Dr. Lawrence Kaputska
President, Ecological Planning & Toxicology
Dr. Robin Matthews
Western Washington University
Dr. Bradley Sample
CH2M Hill, Sacramento, CA
Physical and Chemical Properties
(SPARC)3
1998
Dr. Richard Lee Dickerson
Texas Tech. University
Dr. Ryan DuPont
Utah State University
Dr. Igor Linkov
Menzie-Cura Associates
Aqueous Geochemistry and Sorption
(MINTEQA2)3
1999
Dr. Lynn Dudley
Utah State University
Dr. Kathryn Johnson
Johnson Environmental Concepts
Dr. R. Bruce Robinson
University of Tennessee
Human Health Benchmarks
1999
Dr. Sara Hoover
Golden Associates
Dr. Daland Juberg
ICTM
Dr. Gary Pascoe
Independent Consultant
Dr. Lauren Zeise
California EPA
a Denotes legacy code that forms the basis of the module and which has also been peer reviewed independently.
2-10
-------
Section 2.0
Overall Approach to Verification and Validation
2.3 Verification and Validation of the Site-based Data
2.3.1 Verification of Site-based Data
Data verification includes the activities undertaken prior to and during the data collection
effort to ensure that data of the correct type, amount, and quality (i.e., data that meet data quality
objectives) are provided in the 3MRA modeling system respresentative national dat set. These
activities included
¦ Developing a data collection plan that specified data sources and how data would
be collected from those sources;
¦ Incorporating a quality assurance/quality control (QA/QC) protocols that were
specified as part of the data collection plan, including data entry checks,
independent calculations to verify that data were processed correctly in all
circumstances, and automated checks of critical parameters, formats, and
processes; and
¦ Conducting independent testing of the major site-based data elements.
EPA updated the data collection plan and it is described Volume II, Data volume. Section 5
describes the verification of the site-based data in more detail.
2.3.2 Validation of Site-based Data
EPA validated the accuracy of the site-based data in the representative national data set
by comparing those data with data and model results for two of the sites where more recent data
were independently collected during EPA's 1999 Surface Impoundment Study (SIS) Survey
(U.S. EPA, 2001). Section 5 describes the validation of the site-based data in more detail.
2.4 Verification and Validation of the Chemical Properties Data
Chemical properties were estimated using the SPARC and MINTEQA2 models.
Section 6 describes the verification and validation of the chemical properties models in more
detail.
2.4.1 Verification of Chemical Properties Models
For SPARC, EPA has developed quality assurance software that is executed each quarter
on the current version of SPARC. This software runs the various property modules for a large
number of chemicals and compares the results to historical results obtained over the life span of
the SPARC program. For MINTEQA2, EPA tests all code modifications and additions by a
combination of compiler tests and model execution tests before final adoption.
2.4.2 Validation of Chemical Properties Models
For SPARC, each calculation used to estimate a chemical or physical property has been
validated against numerous measured values for the property of interest. For MINTEQ,
validation has been done by conducting tests and studies that show that the geochemical model,
implemented by the combination of the user's input parameters, the thermodynamic database,
2-11
-------
Section 2.0 Overall Approach to Verification and Validation
and the computer code, provides an acceptable representation of reality or that it produces an
outcome that is an acceptable representation of reality.
2-12
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
3.0 Ensuring Quality of the 3MRA Modeling System
This section addresses the ensurance of the quality of the 3MRA Modeling System at the
system level. Figure 3-1 provides a broad overview of the system verification and validation.
3.1 Verification of the Modeling System
Integrated System Testing
Version Control
Verification of the 3MRA modeling
system requires that the software components of
the system function according to their individual
and collective design. Verification of the 3MRA
modeling system implies that the science-based
modules and data processors are codified in
software form correctly. This section describes
the key aspects of the software development and
verification protocols followed for the 3MRA
modeling system.
The 3MRA modeling system represents
the integration of more than twenty-five
independent software components developed by
five different software development groups
located across the country and possessing
different levels of expertise with respect to
system software development. Because of this
technically diverse and geographically dispersed
development team, there was a need to formulate
a comprehensive and rigorous approach for
executing and managing software development
and quality assurance. Four key elements
contributed to the successful development,
testing, and delivery of the 3MRA modeling
system: (1) open and regular team
communications, (2) a documented top-down
approach to software design, (3) a rigorous
implementation of software quality assurance and testing, and (4) a sound strategy for archiving
the software. The following sections describe the manner in which each of these elements was
implemented for the 3MRA modeling system.
Sensitivity an
Ana
d Uncertainty 1
ysis 1
r
Model Validation through 1
TRIM Comparison 1
Figure 3-1. Overall approach to ensure the
quality of the 3MRA modeling system
technology.
3-1
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
3.1.1 Development Team Communications
The day-to-day management of the 3MRA modeling system development required
communication methods that bridged the geographic distribution of team members on a regular
basis. A combination of e-mail, tele- and video- conferencing, and face-to-face meetings were
used to satisfy the communication requirements. Rapid reporting of problems and transfer of the
science modules was facilitated by the e-mail communications. Regular periodic and on as-
needed basis teleconferences were scheduled as e-mail traffic indicated the need for a team
consultation. During the critical stages of development, teleconferences were held as often as
twice a week. A detailed agenda was developed and distributed to team members before each
call. Project status, technical problems, and technical decisions were discussed.
Periodically (i.e., on the order of two to three times per year), an all-hands meeting was
held to address issues that required extensive and focused discussion. For example, early in the
development process, after the system design had been documented, teleconference discussions
indicated inconsistent interpretations of the design. Attempts to resolve these issues through
teleconferences were deemed unsuccessful when efforts to link the first components failed due to
data interchange conflicts. Therefore, an all-hands meeting was held when multiple groups were
ready to attempt linkages between their modules. The convergence to a common understanding
of the system design with respect to the proper operation of the system was facilitated by the
intense discussions occurring during attempts to link modules.
3.1.2 Top-Down Design Approach
The foundation of the technical approach was the "software system life cycle" process
applied with modern object-oriented system design principles. Two elements applied early in
the process proved to be critical to the success of the effort. First was the documentation of the
system design, especially the specification of the science modules and databases. Second was
the relationship among these modeling components, in the form of data representation and
interchange. The use of an Applications Programming Interface (API) that both provided a
standard data representation scheme and established a required protocol for data interchange
proved to be critical. The API, in addition to minimizing the occurrence of data conflicts
between modules, also served as a focal point for technical discussions.
3.1.3 Quality Assurance and Testing Strategy
The objectives of software testing are three-fold and include the following:
1. To uncover errors in requirements, design, and programming;
2. To identify or confirm limitations of the system; and
3. To verify the capabilities of the system.
The single most significant factor in the successful development of the 3MRA modeling
system was the establishment and steadfast adherence to a rigorous testing strategy. Given the
technically diverse and geographically dispersed nature of the 3MRA modeling system
development team and the size of the software system to be developed, it was imperative to
establish a rigorous software testing strategy. The strategy for testing the 3MRA modeling
3-2
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
system included a procedure for components in a standalone mode and a procedure for systems
integration testing.
The software development occurred at the component level first. When sufficient
infrastructure and science-based components (modules and databases) were available, system-
level integration was initiated. When component developers had completed a first version of the
software, they could test it both as a standalone executable and also in the context of the system
software. Often, component software that ran without error in a standalone mode would
experience significant difficulty when integrated into the system. Typically, this was due to data
conflicts with other components that were either sending data to the module or receiving outputs
from the module.
As components were developed, the developers also established and documented formal
test plans. Software testing for the 3MRA modeling system was requirements based; that is,
specific tests were designed to ensure that each of the requirements associated with the software
design was correctly implemented. While the developer had primary responsibility for the test
plan and its successful execution, it was also required that the test plans be reviewed and the
tests re-executed by independent developers (i.e., software engineers and modelers not directly
involved in developing the component software). Thus, all 3MRA modeling system software
was required to undergo two levels of formal testing. Figure 3-2 illustrates the testing protocol
for the developer and Figure 3-3 illustrates the protocol followed for independent testing, at the
component level. To facilitate the independent testing, the developers constructed test packages
that included all documentation, source code, and test files associated with a component. As can
be seen in Figure 3-3, the independent tester first reviewed the test plan for completeness, to
determine if sufficient tests had been designed to verify the implementation of all requirements.
The independent tester was free to add tests to the test plan. Once the test plan was approved,
the independent tester executed all the tests using the developer's executable version of the
software. When this was successful, the independent tester then recompiled the component
software and re-executed the tests a second time. This second level of independent testing
proved invaluable in identifying issues related to compilation and also ensured that the complete
set of software related to the component was provided.
After the 3MRA modeling system components satisfied all tests, it was necessary to
place them into the complete 3MRA modeling system to confirm their functionality in this
context. The system integration tests required that components operate error free while
executing within the full system context (i.e., running the full set of site/WMU combinations for
the contaminants of interest and the iterations of the Monte Carlo simulation). System-level
testing uncovered not only programming errors such as data conflicts, but also several science-
based errors. The 3MRA modeling system representative national dat set of 201 sites executed
within a Monte Carlo simulation provided a testbed that stressed the numerical solutions of
science models in a manner not previously possible. For example, even legacy codes that had
more than a decade of wide use experienced environmental conditions that resulted in unstable
numerical solutions.
3-3
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Requirements
Documentation
YES
NO
YES
NO
Correct
Problem?
Are Results
Acceptable^
Design
Test Cases
Design
Documentation
Test Plan
Modify
Code(s)
Run Test
Cases and
Evaluate
Results
Complete
Testing
Documentation
and Prepare
Testing Packages
for Independent
Processing
Figure 3-2. Developer testing process for the 3MRA modules.
YES
YES
^Are
Results
NO
NO
YES
YES
/Are\
Results
^Are^
Results
NO
NO
Developer
Modifies
Test Plan
Independent
Review
Independent
Recompile
& Relink
Developer
Test Plan
Developer
Source Code
Run Tests
With Developer
EXE
Collect
Problems
and Report
to Developer
Collect
Problems
and Report
to Developer
Run Tests
With Independently
Generated EXE
Collect
Problems
and Report
to Developer
Developer Modifies
Code and Test Plan
as Appropriate
Developer Modifies
Code and Test Plan
as Appropriate
Developer Modifies
Compile/Link Process
as Appropriate and
Re-submits
Figure 3-3. Independent testing process for the 3MRA modeling system.
3-4
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
The system integration tests were initially conducted on single PCs. Computation time
for the 3MRA modeling system was typically on the order of a few minutes per site-based
simulation. Thus, a single PC might run continuously for several days before completing a
simulation for all combinations related to a single contaminant. When runs for multiple
contaminants were required, several PCs were used, each running a subset of the full list of
required simulations. Recently, EPA has constructed a PC-based cluster that includes
approximately 160 PCs networked together and developed software for distributing the
computational load across this network and retrieving output files. While this configuration of
PCs (named the Super-computer for Model Uncertainty and Sensitivity Evaluation
[SuperMUSE]) was intended as a research tool for investigating uncertainty and sensitivity
related to the 3MRA modeling system, it also serves as a final testing ground for the software
itself. Before Version 1.0 of the 3MRA modeling system was released, the SuperMUSE had
successfully executed on the order of one million individual 3MRA modeling system
simulations.
3.1.4 Software Archiving and Distribution
The 3MRA modeling system is a large and complex software system. It is composed of
over 100 binary, configuration, and data files. In addition, the test packages included over 200
source code files and numerous files. More than 15 software engineers and scientists from five
geographically distributed groups contributed to the development of the 3MRA modeling system
and it is likely that additional software engineers will contribute to its future development.
Coordinating all the changes among these entities in order to assure the quality and
reproducibility of the final release required formal archiving and release procedures.
The first step to establishing these procedures was to set up a version control system.
Version control software is used to archive files and track changes. Several commercial
software packages that could handle this task were reviewed. Microsoft's Visual SourceSafe
was selected because it provided a robust functionality, licenses were available, and the internal
development team had previous experience using it. The Visual SourceSafe software runs
locally on a user's machine. However, the actual file repository is kept on a network drive. This
drive is fully backed up every week and incrementally backed up each evening. Two of the
primary functions of the software package are keeping track of file versions and preventing
multiple developers from modifying the same file at the same time. When a developer wants to
update a file, the file is "checked out," which locks the file and prevents others from updating it.
The file is modified and then "checked backed in." This updates the file in the version control
system. The previous version of the file is not lost, as all versions are archived. This allows the
reconstruction of any version of the components or the entire system.
Individual projects in Visual SourceSafe were created for each system component (e.g.,
science module). The baseline version of all components and the system in general is referred to
Version 1.0.
The initial compilation of the Version 1.0 archive was actually the last step of the testing
protocol. For reference, a system component is compiled in the form of a binary file such as an
executable file or a dynamic link library (dll) file. On the Microsoft Windows platform, an
3-5
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
executable is a file that has an ".exe" extension. A dll file has a ".dll" extension. Each
developer was required to provide a package of files that were to be associated with the Version
1.0 release. The package was to include the following:
¦ All necessary source code and libraries needed to compile the specified module;
¦ All test packages, including test data and auxiliary files (e.g. batch files, script
files) required to run the test suites; and
¦ All documentation describing the purpose of each test case and how to execute it
and the expected results.
As the individual module packages were received, they were then compiled.
Compilation is the step whereby source code, computer instructions written in higher-level
languages such as FORTRAN or C++, is converted into machine-readable binary files. There
are a wide variety of compiler types, brands, and versions available. For instance, to compile
C++ code, EPA used the compilers included with Borland C++ Builder and Microsoft Visual
C++. Each of these has six or more versions in existence. The compiler originally used by each
developer was used whenever possible. When not possible, a version as close to the developer's
version as possible was used. This resulted in a standard set of compilers now used by EPA for
archiving.
All tests included in the packages were then executed. Accompanying documentation
provided step-by-step instructions on how execute the individual test suites. In cases where tests
were not successfully executed, the developer was informed and required to resolve any issues
(e.g., out-of-date test package, errors in source code).
After successful completion of all tests, the resulting materials were entered into Visual
SourceSafe. This completed the component level testing and verification process. The next step
was to perform system testing. The PCs in the SuperMUSE are currently configured with
different versions of Microsoft Windows. SuperMUSE has custom software that allows it to run
stand-alone models such as the 3MRA modeling system on the PC cluster. The software allows
for the distribution of various scenarios and subsequent data collection. Running the software on
SuperMUSE enables EPA to complete hundreds of thousands of runs. This would not be
feasible if all system testing were done on individual PCs. Any component-specific errors that
were discovered were referred back to the developer as described above. The development team
addressed system-level errors and the system was retested. Any file changes that were required
to fix component or system errors were updated into Visual SourceSafe.
After final testing was completed and the archive was populated with a complete set of
system software and related documentation, the materials were made available to the public.
Procedures are under development for distributing the software to the public, providing technical
assistance, and accepting new or updated models. It is not currently clear how version control
will work when the 3MRA modeling system is distributed to the public. All materials, including
source code, will be made available. EPA is currently reviewing various open-source licensing
agreements and anticipates including such an agreement requirement whenever the source code
is requested. It is understood that EPA is the designated owner of the "official" 3MRA modeling
3-6
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
system software. EPA keeps all the files necessary to compile, test, document, and execute the
3MRA modeling system. Although the public will be free to modify the software, the altered
software will not be considered official unless it is processed by EPA through the procedures
described above.
3.2 Validation of the System
In the context of a national regulation, a complete approach to demonstrating that the
3MRA modeling system produces results consistent with field observations would necessarily
require field data comparisons of the 3MRA modeling system for numerous contaminant release
scenarios covering the range of environmental conditions where waste disposal is possible.
However, as pointed out earlier, no such data are currently available to validate the system.
Therefore, EPA had to resort to alternative approaches to ensure the quality of the modeling
system.
The approach undertaken here, as a first step, involves the application of the 3MRA
modeling system to a single site and the comparison of results with (1) available field
observations, and (2) compare with the results of the application of a second modeling system to
the same site.
This approach was implemented through a collaborative effort with the modeling team
developing the Total Risk Integrated Methodology (TRIM), a multimedia model designed to
serve a similar purpose in support of EPA's Office of Air Quality Planning and Standards
(OAQPS). Although TRIM supports the development of air regulations, as opposed to the
3MRA modeling system's focus on land-based disposal of solid wastes, the two models share a
need to simulate the fate and transport of contaminants in all media, not just the media in which a
contaminant is first released. The 3MRA modeling system and TRIM take very different
approaches to simulating the multimedia system of physical/chemical/biological processes and
thus a model comparison study provides an opportunity to gain experience applying each
approach and confidence that model predictions are reasonable.
There are five primary steps to modeling a site: problem definition, site
conceptualization, collection/organization of modeling data, model execution, and interpretation
of results. Problem definition includes site selection, a statement of an assessment problem, and
specific modeling goals to be achieved. Site conceptualization is the process of delineating the
physical features of a site in a manner consistent with model design. The inter-connectivity of
the features must also be specified (e.g., watershed connectivity to stream segments). Data
representing contaminant source terms, environmental conditions related to the physical features
of the site, chemical properties (physical/chemical/biological constants/rates), and contaminant
distribution in each of the media must be collected and organized for use by the model. Model
execution is a mechanical step that applies the model to achieve the stated modeling goals.
Interpretation of results for this study includes both a comparison of results with available field
observations as well as an inter-model comparison. The following sections provide a more
detailed description of these steps for this model comparison study.
3-7
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
3.2.1 Problem Definition
3.2.1.1 Site selection. Establishing criteria for site selection is fundamentally a function
of model purpose and design. The purpose of both the 3MRA modeling system and TRIM is to
estimate the potential human and ecological exposure and associated health risks associated with
the release of contaminants from regulated industrial sources. As such, a site selected for this
study must contain a source of contaminant released to the environment. The contaminant, at
least for initial comparisons, should be a multimedia contaminant, that is, subsequent to release it
should be distributed throughout the environment. Further, data must be available that describe
the environmental setting as it is represented in the models. In the case of both TRIM and the
3MRA modeling system, the environmental setting includes all media: air, soil, vadose zone,
ground water, surface water, sediment, and biota. Data that vary as a function of time (e.g.,
meteorology) must be available for a time period relevant to the simulation. For example, if a
facility operated for thirty years, then time series data should be available for those thirty years,
and for a period of time following facility closure. The 3MRA modeling system and TRIM
estimate long-term average exposures extending over periods of several years to a lifetime. Each
model simulates these periods with time steps that may be as short as an hour or as long as a
year; thus, time-varying data must be available for decades and at intervals as short as an hour.
Finally, data must be available that reflects the impact of the source on the health of the human
and ecological receptors occupying an area near the source, generally to two or more miles from
the source. These data include time series of concentrations in each of the modeled media and
intermedia contaminant fluxes. These concentrations and fluxes should be available at several
locations within the area of interest.
In response to comments of the EPA Science Advisory Board related to a review of
TRIM, an extensive review of the literature was conducted in an effort to locate data sets for use
in evaluating the performance of TRIM.Fate. The conclusion reached by the TRIM modeling
team is that "[N]one of the studies identified during EPA's literature review provides complete
and concurrent information on contaminant concentrations in the five major environmental
media (i.e., air, water, sediment, soil, biota) along with the associated source terms(s) and
historical environmental characteristics (e.g., meteorology, hydrology, landscape properties)."
Given this limitation it is necessary to scale back on the site selection criteria and choose a site
where the amount of available information is maximized.
Of the sites reviewed, the TRIM modeling team selected the HoltraChem Manufacturing
site in Orrington, Maine, and selected mercury as a case study contaminant of concern. The
following description of the HoltraChem site is taken from the Maine Department of
Environmental Protection website1.
The HoltraChem Manufacturing Company in Orrington, Maine... is
located on a 235-acre property on the banks of the Penobscot River.
Approximately 50 acres are developed and include the manufacturing facility,
five landfills, a surface impoundment and a waste pile. The immediate plant area
is approximately 12 acres. The property is west of Route 15 and it abuts the
1 http://www.state.nc.us/dep/rwm/holtrachem/index.htm
3-8
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
PERC waste-to-energy plant on the north and west. It opened in 1967 and
manufactured chlorine, caustic soda (sodium hydroxide) and chlorine bleach
(sodium hypochlorite) used by paper mills. The plant also manufactured
hydrochloric acid and the pesticide chloropicrin. The plant closed in September,
2000.
The plant used a chlor-alkali process to separate sodium and chlorine
from salt water. In this process, elemental mercury was used as a cathode to
collect the sodium from the water. The chlor-alkali process is an older
technology and has been replaced by mercury-free production techniques at
newer plants and some converted older plants. When it stopped operations last
fall, HoltraChem was one of 13 chlor-alkali plants left in the country. There were
as many as 30 chlor-alkali plants in the United States at one time.
A site investigation determined that the HoltraChem property, including
parts of the Penobscot River, is contaminated with mercury, chloropicrin and
several volatile organic compounds. Additional investigation will be done into
the presence of additional areas of mercury contamination and polychlorinated
biphenols (PCBs).
Specifically, the following contaminants at a minimum are present on site:
* Soils: chloropicrin, ethylbenzene, xylenes, arsenic, barium,
cadmium, chromium, lead and mercury.
* Ground water: 1,1 dichloroethane, acetone,
bromodichlorome thane, bromoform, carbon disulfide, carbon
tetrachloride, chlorobenzene, chloroform, chloromethane,
chloropicrin, dibromochloromethane, methylene chloride,
trichloroethene, mercury, potassium, and sodium.
* Surface Water: carbon tetrachloride, chloroform and mercury.
* Sediment: mercury
* Biological samples: mercury
The developed part of the property can be subdivided into three
physiographically distinct areas: 1, the bedrock ridge, 2, the plant area, and 3,
the river area. The northeast trending bedrock ridge is north and northwest of the
plant area and separated from the plant by an abrupt scrap. The elevation of the
crest of the ridge varies from 80 ft. to 145 ft. The plant is constructed on a
relatively flat surface about 65 ft. elevation, about 500feet east of the Penobscot
shoreline. The river area slopes steeply from the plant area to the river. The
northwest side of the bedrock ridge is drained directly to the river by a small
intermittent stream. The western part of the plant area drains west through the
river area by way of the Northern Stormwater Drainage Ditch. The eastern part
3-9
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
of the plant area is drained south to the Penobscot by the Southerly Stream. The
Northern Stormwater Drainage Ditch and the Southerly Stream are perennial
streams that flow all year long and derive some of their water from ground water
discharge.
The overburden deposits at the site can be grouped into four types.
Glacial till is a silty sand that contains abundant clay and gravel as well. Till is
deposited directly from glacial ice without appreciable sorting by water. Glacial
outwash is made up of moderately sorted sand and gravel that tends to be poor in
silt and clay. Outwash is quite permeable, and where thick deposits occur below
the water table it can be an excellent aquifer resource. Fine grained marine
sediments in coastal Maine are referred to as the Presumpscot Formation. The
Presumpscot Formation is made up largely of silty clay that was carried out of
the glacier by meltwater and settled out in the ocean. Finally, there are deposits
of fill on the site that include till, outwash and marine sediments that were
redistributed during construction at the plant, and brine and wastewater sludge
stabilized with sand in the hazardous waste landfills.
The geology of the site varies from place to place. The bedrock ridge is
made up of tough metamorphic rock covered by a thin layer of glacial till. There
are thick deposits of man-made fill on the ridge, including three hazardous waste
landfills. The plant area is developed on thick overburden that is mostly compact
glacial till, but a shallow wedge of sandy outwash occurs all along the southern
edge of the bedrock ridge. The shallow sandy outwash thickens west toward the
river so that the river area is underlain by a thick sequence of sand and gravel.
Ground water flows in bedrock fractures on the bedrock ridge. It flows
north directly to the Penobscot on the north side of the ridge, but it flows south
toward the plant on the south side of the ridge. The thick compact glacial till
under the plant area is not very permeable, so most of the ground water under the
plant area flows in shallow fill and outwash sand over the top of the till surface.
Ground water from the southern part of the plant area discharges to the
Southerly Stream. Ground water from the rest of the plant area flows west to the
Penobscot River through the river area.
While ownership has changed several times, Mallinckrodt, Hanlin and
HoltraChem have each worked to advance the cleanup. Several of the actions
which HoltraChem and Mallinckrodt have taken to control and clean up
pollution at the plant are listed below:
* Relined a lagoon to stop waste from discharging to ground water
which flows into the Penobscot River;
* Installed riprap on the riverbank next to one of the plant landfills
to prevent erosion and river contamination;
3-10
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
» Installed a ground water collection system to capture a portion of
contaminated ground water discharging from the site for
treatment, and
* Instituted measures to prevent contaminated sediment from
reaching the river.
3.2.1.2 Problem statement and modeling goals. As described, the HoltraChem site
includes a host of contaminant release problems, ranging from chemical spills to disposal of
waste in landfills, a surface impoundment, and a waste pile to fugitive emissions from chemical
processing. Contaminants present at the site include several metals, volatile organic compounds,
and PCBs.
Mercury will serve as the initial focus of model application and comparison because
(1) mercury is a contaminant of concern to both OAQPS and OSW, (2) mercury is a pollutant
whose fate, transport, and bioaccumulation involve all media, and (3) mercury levels have been
monitored in a number of media and locations at the HoltraChem site. Table 3-1 summarizes the
available mercury monitoring data.
Various limitations in data availability and differences between TRIM and the 3MRA
modeling system constrain the application and the interpretation of results. With respect to data
availability the following limitations exist:
¦ The source term for mercury emissions is based on based on the characteristics of
the chlor-alkali facility operations and not specific measurements at HoltraChem;
¦ Monitoring data provides more of a snapshot of mercury in the surrounding
environment than a continuous measure of its presence during the period of
HoltraChem operations; and
¦ Contributions from other sources of mercury (both on-site and off-site) are not
explicitly accounted for in this initial assessment.
With respect to differences between the 3MRA modeling system and TRIM the following
limitations in the model comparison are noted:
¦ Currently, TRIM simulates only multimedia fate and transport; human and
ecological exposure and risk are currently being added.
¦ The 3MRA modeling system does not simulate subsurface fate and transport for
atmospherically deposited contaminants. Thus, no vadose zone or aquifer
simulation of mercury movement is conducted for the 3MRA modeling system
runs.
¦ TRIM does not simulate land-based disposal units and thus the landfills, surface
impoundment, and waste pile are not simulated.
3-11
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Table 3-1. Site Sampling Data for HoltraChem Facility and Nearby Areas and Water Bodies
All data sampled between 1993 -1997
Minimum
Maximum
Mean
Std Dev
Median
On-site
Surface Soil (THg* mg/kg, dry wgt)
0.18
10.3
5.05
3.47
4.8
Surface Soil (THg mg/kg, dry wgt)
0.14
310
30.1
47.6
13
Surface Soil (THg mg/kg, dry wgt)
0.2
310
5.37
11.4
1.2
Surface Soil (THg mg/kg, dry wgt)
4.5
126.9
23.8
41.8
SubSurface Soil (THg mg/kg, dry wgt)
0.1
80
12.4
16.3
6.2
Deer Mouse (THg mg/kg, ww, whole body) 75%
moisture
0.06
0.198
0.1
0.063
Earthworm (THg mg/kg, ww, 85% moisture)
0.087
2.82
0.982
0.79
Off-site
Ambient Air (THg ng/m3) [1500m SE]
0.834
157
9.96
15.52
Ambient Air (THg ng/m3) [4300m NNW]
0.993
25.8
2.46
2.15
Ambient Air (THg ng/m3) [6400m NNW]
0.565
14.8
1.85
1.66
Surface Water (THg ug/L unfiltered Upstream of
Facility
0.00359
0.00529
0.004
0.001
0.004
Surface Water (THg ug/L unfiltered Downstream
of Facility
0.000646
0.0703
0.034
0.033
0.027
Surface Water (THg ug/L unfiltered River
Adjacent to Facility
0.0041
0.173
0.015
0.0377
Sediment Swetts Pond (THg mg/kg dw)
0.319
Sediment Thurston Pond (THg mg/kg dw)
0.157
Sediment Brewer Lake (THg mg/kg dw)
0.201
Sediment Fields Pond (THg mg/kg dw)
0.132
Juvenile Loon Swetts Pond (blood THg cone ppm
ww)
1.3
Loon Egg Brewer Lake (Hg cone ppm ww)
1.6
1.8
1.73
0.11
Deer Mouse Dorethea Dix Park (THg mg/kg, ww,
whole body) 75% moisture
0.016
0.087
0.0515
0.05
Earthworm Dorethea Dix Park (THg mg/kg, ww,
85% moisture
0.044
0.044
0.044
(continued)
3-12
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Table 3-1. (continued)
All data sampled between 1993 -1997
Minimum
Maximum
Mean
Std Dev
Median
Short Tail Shrew Dorethea Dix Park (THg mg/kg,
0.064
Eel (THg mg/kg ww 80% moisture River
Upstream of Facility)
0.271
0.876
0.53
0.2
Eel (THg mg/kg ww 81% moisture River
Downstream of Facility)
0.259
0.678
0.46
0.14
River Minnow (THg mg/kg dw whole body
Downstream of Facility)
0.447
White Perch Swetts Pond (mg/kg ww)
0.5
1.31
0.98
0.25
White Perch Fields Pond (mg/kg ww)
0.28
0.72
0.45
0.14
White Perch Thurston Pond (mg/kg ww)
0.6
2.2
1.07
0.43
White Perch Brewer Lake (mg/kg ww)
0.32
0.53
0.41
0.08
State of Maine averages
State Avg Loon Egg Hg Cone (ppm ww 43
samples)
0.93
0.55
State Avg Juvenile Loon THg BloodConc (ppm
ww 52 samples)
0.22
0.29
State Avg Adult Male Loon Hg Cone (ppm ww 6
locations)
0.61
3.71
2.62
1.23
State Avg Adult Male Loon Hg Cone (ppm ww
67 samples)
2.5
1.1
State Avg Adult Female Loon Hg Cone (ppm ww
64 samples)
2.1
1.5
*THg = Total mercury.
For these reasons, the initial model simulations for comparison focus only on the fugitive
emission of divalent mercury over the thirty year period of facility operations (3MRA simulates
divalent mercury, while the surface water module of 3MRA does speciate mercury. The
monitoring data is for total mercury). The criteria for selecting specific endpoints, locations, and
times are designed to ensure that each media is included, as well as a representative cross section
of terrestrial and aquatic species as a function of both food source and trophic level. Locations
are a function of both model site conceptualizations (e.g., habitat delineation) and locations
where monitoring data have been collected. Table 3-2 lists the endpoints for the initial model
runs.
3-13
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Table 3-2. Model Estimation Endpoints for Initial HoltraChem Simulations
Medium
Endpoints
Hg Species
Locations
Times
Atmosphere
ambient
concentrations (in
breathing air),
deposition rates
divalent
Location of ambient air
monitoring stations,
lakes/ponds for
deposition, selected
watersheds/habitats
Annual average
time series (30
years)
Soil
average surface soil
concentrations
divalent
on site, across river just
south of facility, selected
watersheds/habitats
Annual average
time series (30
years),
Surface Water
(water column)
average water
column concentration
divalent, methyl
River (north/south of
facility), Swetts Pond,
Brewers Lake, Fields
Pond, Thurston Lake
Annual average
time series (30
year)
Surface Water
(sediment)
to be determined
(3MRA has 2
sediment layers,
TRIM?)
divalent, methyl
River (north/south of
facility), Swetts Pond,
Brewers Lake, Fields
Pond, Thurston Lake
Annual average
time series (30
years)
Biota (terrestrial)
terrestrial vegetation,
White-tailed Deer
{terrestrial
herbivore}, Deer
Mouse {terrestrial
herbivore}, Short-
tailed Shrew
{terrestrial
invertebrate feeder},
earthworm
divalent, whole body
to be determined based
on TRIM/3MRA site
conceptualizations
year 30
Biota (aquatic)
Largemouth Bass,
white perch
methyl, whole body
Swetts Pond, Brewers
Lake, Fields Pond,
Thurston Pond
Year 30
Biota
(terrestrial/aquatic)
Common Loon
{semi-aquatic
piscivore}, Mink
{semi-aquatic
piscivore}, Raccoon
{semi-aquatic
omnivore}
divalent, whole body
in proximity of Swetts
Pond, Brewers Lake,
Fields Pond, Thruston
Pond
Year 30
The 3MRA modeling system simulates mercury speciation only in the surface
water while TRIM simulates mercury chemistry in all media.
3.2.2 Site Conceptualization
Figures 3-4, 3-5, and 3-6 show aerial photographs of the HoltraChem facility and the
surrounding area. Figure 3-7 displays the land use patterns and waterbodies in the modeling area
of interest. The site conceptualization task requires that site layout features relevant to the
models be delineated and that logical connections between the features (watershed drainage to
surface water) be specified.
3-14
-------
Section 3.0 Verification and Validation of the 3MRA Modeling System
Qrrington
Figure 3-4. HoltraChem manufacturing site.
3 km E of Orrington,
Maine, United States
68W 53' 20"
68W 50' 54"
68W 48" 29"
68W 46" 03"
68 W 43" 37"
508,800.0
512,000.0
515,200.0
518,400.0
521,600.0
44N 45' 52"
44N 45* 51"
4,956,800.0
4,956,800.0
.-jt f J gflB^
v" --CV
44N 44" 09"
44N 44* 08"
4,953,600.0
4,953,600.0
44N 42" 25"
44N 42" 24"
4,950,400.0
4,950,400.0
% r
'. i .•
44N 40' 41"
44N 40" 40"
4,947,200.0
4,947,200.0
68W 53* 20"
68W 50" 55"
68W 48" 30"
68W 46' 04"
68V/ 43' 39"
508,800.0
512,000.0
515,200.0
518,400.0
521,600.0
Figure 3-5. HoltraChem area photo.
3-15
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Figure 3-6. Aerial view of HoltraChem facility.
| Holtrachemfinalwatersheds.shp
A/Reach File,V3 (01020005)
Lbangme
Urban or Built-up Land
¦I Agricultural Land
Rangeland
Forest Land
Water
Wetland
Barren Land
Tundra
Perennial Snow or Ice
10 Miles
Figure 3-7. Land use and surface water around HoltraChem facility.
3-16
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
The number and extent associated with each of the site features depends on the initial
problem statement, which specifies endpoints of concern. In the case of the HoltraChem site
there is interest in the potential transport of mercury to lakes southeast of the site (Brewers Lake,
Thruston Lake, Fields Pond) because these lakes lie in the primary downwind direction from the
facility. The interest in potential mercury impacts in these lakes defined the aerial extent for the
modeling for each model. Within the area of interest, each model simulates contaminant fate and
transport in all media, including watershed soils; streams, rivers, ponds, and lakes; airsheds;
unsaturated zone; aquifer; and biota associated with the terrestrial and aquatic foodwebs. The
delineation of aerial extent and depth associated with each of these features is conceptualized
differently between the two models.
TRIM conceptualizes a site as a set of surface parcels and air parcels. The parcels are 2-
dimensional areas of assumed homogeneous environmental conditions. The air parcels are
placed such that they conform first to a need to estimate ambient air concentrations and
depositions as a function of radial distance from the source and, secondly, to the degree possible,
so they align with the surface parcels. The surface parcels are placed considering together the
combination of watersheds, habitats, surface water segments, vadose zone, and ground water
aquifer. When depth is assigned to each of the parcels, the result is a volume element that
represents the unit of analysis for the TRIM model. Figures 3-8 and 3-9 illustrate the surface and
air parcels that define the basis of the site conceptualization for the HoltraChem surrounding
area.
The 3MRA modeling system conceptualizes a site by delineating the actual 2-
dimensional area associated with each unit of the physical features. Site conceptualization is
achieved by delineating areas directly from GIS-based maps of the site. For the HoltraChem
site, the site features considered in 3MRA modeling system runs include several watersheds, a
surface water network, a single airshed, and several ecological habitats. The unsaturated zone
below the root zone and the ground water aquifer were not modeled.
Figure 3-10 illustrates the surface water network that is modeled by the 3MRA modeling
system for the HoltraChem site. The surface water network consists of the Penobscot River,
tributaries in the vicinity of the HoltraChem facility, wetlands, and lakes. For modeling
purposes, the network was configured as a interconnected series of fourteen reaches. Directly
connected to specific surface water network reaches are individual watershed subbasins.
Watershed subbasins that drain to the segment of the Penobscot river lying between the extents
of the tributaries containing the lakes of interest are included. Figure 3-11 illustrates the
eighteen watershed subbasins included in the 3MRA modeling system runs for the HoltraChem
site. Figure 3-12 illustrates the ecological habitats that are delineated principally as a function of
land use. Twelve habitat areas are defined, including two cropland habitats, two forest habitats,
four lake habitats, one residential habitat, two wetland habitats, and one river habitat. Within the
3MRA modeling system, each habitat type includes a list of species that are both of interest and
native to the area of the country being modeled. Each species within a habitat is assigned a
home range, specifying the sub-area extent occupied by the species.4
3-17
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
September
14, 2001
Watersheds
/\ / Wate rb o d i es
J\/ County boundaries
A / Ro a ds
/^/Surface Parcels
ESE5
Figure 3-8. TRIM surface parcels for HoltraChem facility.
September 24, 2001
NNW3
Figure 3-9. TRIM air parcels for HoltraChem facility.
3-18
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Surface Water Reach Network
Holtrachem Site
] Waterbodynetwork.shp
Reach File, V3 (01020005)
Figure 3-10. 3MRA modeling system surface water reach network for
HoltraChem facility.
HoltraChem Watershed Subasin Layout
I Holtrachemfinalwatersheds.shp
| Waterbodynetwork.shp
Figure 3-11. 3MRA modeling system watershed subbasin delineation for
HoltraChem facility.
3-19
-------
Section 3.0
Verification and Validation of the 3MRA Modeling System
Habitats
1. Residential
2. Lake
3. River
8. Forest
9. Forest
10. Lake
11. Lake
12.Lake
7. Cropland
4. Wetland
5. Wetland
6. Cropland
Land use
Urban or Built-up Land
Agricultural Land
Rangeland
Forest Land
Water
Wetland
Barren Land
Tundra
Perennial Snow or Ice
N
0
10 Miles
W
T
E
S
Figure 3-12. 3MRA modeling system land use and ecological habitat
delineation for HoltraChem facility.
3.2.3 Data Collection
The data collection effort for the TRIM comparison is on-going.
3.2.4 Model Execution
The TRIM comparison has not yet been executed.
3.2.5 Interpretation of Results
The TRIM comparison has not yet been executed.
3-20
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.0 Evaluating Quality of the 3MRA Modeling System
Modules
This section describes verification and validation activities for the 3MRA modeling
system modules. Figure 4-1 provides an overview of the verification and validation process for
the modules. Figure 4-2 shows which subsection describes each module. The five source
modules and the watershed module are covered in two sections to avoid duplication, one for
Wastewater Source Modules (surface impoundments and aerated tanks) and one for the Land-
based Source Modules (landfill, waste pile, and LAU) and the Watershed Module.
4.1 Wastewater Source Modules
This section documents the
verification and validation activities for the
Aerated Tank and Surface Impoundment
Modules. The draft Wastewater Source
Module documentation was peer reviewed by
the external reviewers shown in the box.
4.1.1 Module Description
The Wastewater Source Modules simulate wastewater management. Two Wastewater
Source Modules were developed for the 3MRA modeling system to represent the common
management practices for industrial wastewaters for the purposes of flow equalization, storage,
treatment (typically biological treatment or neutralization), and solids settling (clarification).
These waste management units (WMUs) are
¦ Surface Impoundments, which may be either aerated or quiescent and are used
to treat, store, or dispose of many industrial wastewaters; and
¦ Aerated Tanks, which are aerated or mixed tanks used to treat or store many
industrial wastewaters.
The two Wastewater Source Modules use a common set of algorithms with similar mass
balance and transport equations. Both modules estimate volatile emissions to air. Leaching
losses from the bottom of the unit are modeled only for surface impoundments; tanks are
assumed to have impervious bottoms. Both modules simulate degradation and solids settling,
although solids settling and accumulation are more significant for quiescent units.
Peer Reviewers for Wastewater Source Modules
¦
Dr. Patricia Culligan of Massachusetts
Institute of Technology
¦
Dr. Wade Hawthorn of Washington State
University
¦
Dr. Michael Overcash of North Carolina
State University
4-1
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Module Adaptation from
Legacy Codes
(EPACMTP, ISCST, Exams)
T
Module Development Based on
Sound Science and Established
Methodologies
J
Internal Review and
Testing
Independent External
Testing
Peer Review
Integrated System
Testing
I
Sensitivity and
Uncertainty Analysis
Figure 4-1. Overall approach to ensure the quality of the 3MRA science modules.
4-2
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Wastewater Modules
(Section 3.2)
Module Description (4.2.1)
Major Module Components/Functionality (4.2.2)
Summary of Verfication (4.2.3)
Summary of Validation (4.2.4)
Summary of Peer Review (4.2.5)
Land-Based Source Modules
(Section 3.3)
Module Description (3.3.1)
Major Module Components/Functionality (3.3.2)
Summary of Verfication (3.3.3)
Summary of Validation (3.3.4)
Summary of Peer Review (3.3.5)
Media Fate & Transport
Air Module ^
f Watershed Module ^
{ Surface Water Module i
f Groundwater Module
(Section 4.4)
(Section 4.3)
(Section 4.5)
(Section 4.6)
• Module Description (4.4.1)
• Module Description (4.3.1)
• Module Description (4.5.1)
• Module Description (4.6.1)
• Major Module Components/
• Major Module Components/
• Major Module Components/
• Major Module Components/
Functionality (4.4.2)
Functionality (4.3.2)
Functionality (4.5.2)
Functionality (4.6.2)
• Summary of Verfication
• Summary of Verfication
• Summary of Verfication
• Summary of Verfication
(4.4.3)
(4.3.3)
(4.5.3)
(4.6.3)
• Summary of Validation
• Summary of Validation
• Summary of Validation
• Summary of Validation
(4.4.4)
(4.3.4)
(4.5.4)
(4.6.4)
• Summary of Peer Review
• Summary of Peer Review
• Summary of Peer Review
• Summary of Peer Review
V (4.4.5) j
I (4.3.5) j
^ (4.5.5) J
^ (4.6.5) J
Farm Food Chain Module
(Section 4.7)
Module Description (4.7.1)
Major Module Components/
Functionality (4.7.2)
Summary of Verfication (4.7.3)
Summary of Validation (4.7.4)
Summary of Peer Review (4.7.5)
v_
Food Webs
Terrestrial Food Web Module
(Section 4.8)
1 Module Description (4.8.1)
1 Major Module Components/
Functionality (4.8.2)
1 Summary of Verfication (4.8.3)
1 Summary of Validation (4.8.4)
1 Summary of Peer Review (4.8.5)
Aquatic Food Web Module
(Section 4.9)
Module Description (4.9.1)
Major Module Components/
Functionality (4.9.2)
Summary of Verfication (4.9.3)
Summary of Validation (4.9.4)
Summary of Peer Review (4.9.5)
Human Exposure Module
(Section 4.10)
Module Description (4.10.1)
Major Module Components/Functionality
(4.10.2)
Summary of Verification (4.10.3)
Summary of Validation (4.10.4)
Summary of Peer Review (4.10.5)
Human Risk Module
(Section 4.11)
¦ Module Description (4.11.1)
¦ Major Module Component/Functionality
(4.11.2)
¦ Summary of Verification (4.11.3)
¦ Summary of Validation (4.11.4)
¦ Summary of Peer eview (4.11.5)
Ecological Exposure Module
(Section 4.12)
¦ Module Descriptoin (4.12.1)
¦ Major Module Components/Functionality
(4.12.2)
¦ Summary of Verification (4.12.3)
¦ Summary of Validation (4.12.4)
¦ Summary of Peer Review (4.12.5)
Ecological Risk Module
(Section 4.13)
¦ Module Description (4.13.1)
¦ Major Module Components/Functionality
(4.13.2)
¦ Summary of Verification (4.13.3)
¦ Summary of Validation (4.13.4)
¦ Summary of Peer Review (4.13.5)
Exit Level Processors
(ELPs)
Figure 4-2. Organization of the 3MRA modeling system modules and this section.
4-3
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
The two Wastewater Source Modules were designed to provide estimates of annual
average volatilization rates to air, which are used by the Air Module. In addition, the Surface
Impoundment Module outputs annual average infiltration rates and leachate contaminant flux
rates, which are used by the Vadose Zone and Aquifer Modules, and annual average surface
impoundment water concentrations, which are used by the Ecological Exposure Module to
estimate exposure to wild animals that may drink or consume organisms from the impoundment.
Figure 4-3 shows the information flow for the Wastewater Source Modules to the other modules.
Key Data Inputs
Operating lifetime
Unit depth
Number of impellors
Henry's law constant
Surface
Impoundment
and Aerated
Tank Modules
Emission Rates
(volatile only)
Chemical Fluxes Infiltration Rates
Air Modu e
(SI only)
Vadose Zone
and Aquifer
Modules
Wastewater Concentrations
(SI only)
Ecological
Exposure
Module
Figure 4-3. Information flow for the Wastewater Source Modules in the 3MRA
modeling system.
4.1.2 Major Module Components/Functionality
The Wastewater Source Modules have six major functions, as follows:
1. Calculate contaminant concentrations within the unit. The Wastewater
Source Modules use a mass-balance, temperature-adjusted approach to estimate
contaminant concentrations in the WMU. This approach considers contaminant
diffusion between wastewater and sediments, and contaminant removal by
volatilization, biodegradation, hydrolysis, partitioning to solids, solids settling,
and, for surface impoundments only, infiltration through the bottom of the unit.
2. Calculate solids concentrations within the unit. The Wastewater Source
Modules use a mass-balance approach to estimate solids concentrations in the
WMU.
3. Calculate volatile emission rates. The modules calculate volatile emission rates
for both aerated and quiescent surfaces.
4-4
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4. Estimate resuspension, sedimentation, and burial velocities within the unit.
The modules account for these processes within the tank or surface impoundment.
5. Estimate contaminant release in leachate. The Surface Impoundment Module
calculates infiltration rates and contaminant leachate flux rates for use in the
Vadose Zone and Aquifer Modules. The Aerated Tank Module does not calculate
leachate release, as tanks are assumed to have impervious bottoms.
6. Adjust for temperature effects. The modules account for the effect of
temperature on air viscosity and density, water viscosity, chemical properties, and
sediment biodegradation rates.
4.1.3 Summary of Verification
The Wastewater Source Modules were verified through a series of verification tests,
which are described in detail in the following documents:
¦ 3MRA Surface Impoundment Module Test Plan (RTI, 2002d)
¦ 3MRA Surface Impoundment Module Test Documentation (RTI, 2002c).
¦ 3MRA Aerated Tank Module Test Plan (RTI, 2002b)
¦ 3MRA Aerated Tank Module Test Documentation (RTI, 2002a)
The volatile emissions component of the Wastewater Source Modules is based on
CHEMDAT8, which has been verified in previous work. Independent engineering review was
conducted on early versions of the CHEMDAT model to ensure that the equations were properly
coded. Sample calculations are provided in the CHEMDAT8 model documentation, and those
sample calculations verify that the model yields the same results as the sample calculations (U.S.
EPA, 1994a). Additionally, the CHEMDAT series of emission models has been publically
available for more than 10 years and has been scrutinized by a variety of industries during rule-
making efforts. CHEMDAT8 and Water8 represent the culmination of several revisions and
modifications made to the emission model over a 5-year period.
EPA developed separate verification test plans for the Surface Impoundment and Aerated
Tank Modules. Tables 4-1 and 4-2 summarize the functional requirements and number of test
cases executed to verify these two modules. All test cases were completed for both source
modules by an independent engineer. Under one of these test cases, the individual and overall
volatilization mass transfer coefficients were compared with those calculated using
CHEMDAT8. This comparison verified that the mass transfer correlations had been correctly
coded in the Wastewater Source Modules. The results of these internal verification efforts are
documented in reports for each source module (RTI, 2002a,c).
4-5
-------
Section 4.0 Verification and Validation of the 3MRA Modeling System Modules
Table 4-1. General Requirements for Testing the Aerated Tank Module
Step
Description
Number of
Test Cases
1
Correct operation of the volatilization mass transfer rate algorithm
13
2
Correct operation of the sediment mass balance
6
3
Correct operation of the annual averaging algorithm
1
4
Correct operation of system requirements
6
Table 4-2. General Requirements for Testing the Surface Impoundment Module
Step
Description
Number of
Test Cases
1
Correct operation of the volatilization mass transfer rate algorithm
15
2
Correct operation of the sediment removal algorithm
8
3
Correct operation of the infiltration rate algorithm
6
4
Correct operation of the SI type assignment algorithm
4
5
Correct operation of the annual averaging algorithm
6
6
Correct operation of system requirements
7
4.1.4 Summary of Validation
The volatilization and infiltration rate components of the Wastewater Source Modules are
based on well-documented and tested models (CHEMDAT8 and EPACMTP, respectively). The
validation of these components is important because the components directly influence the
primary outputs that are used in subsequent modules (i.e., the volatile emissions for both the
Surface Impoundment and Aerated Tank Modules, and the infiltration and contaminant leaching
rate for the Surface Impoundment Module). Therefore, the verification and validation efforts
conducted on these module components help support the validity of the overall modules.
The volatilization mass transfer rate equations used in the Wastewater Source Modules
are based on generally accepted mass transfer correlations. These correlations have been
developed and selected over the past 20 years. Several EPA emission models use these same
volatilization mass transfer rate equations (e.g., CHEMDAT8, Water8, Water9 [U.S. EPA,
1994a]), and these equations have been used to support impact estimates for numerous EPA
regulations. They have been peer reviewed and publically available for more than 10 years.
GCA Corporation performed the initial evaluation and selection of mass transfer rate equations
for use in EPA emission models (GCA, 1982). Over time, a few additional mass transfer
correlations have been developed (e.g., Springer et al., 1984). Each individual mass transfer
correlation was developed from individual studies and data evaluations reported in the literature.
The discussion of the validation efforts provided below focuses on the CHEMDAT8 model
4-6
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
because most of the validation efforts performed on the Wastewater Source Modules'
volatilization rate equations and emission flux estimates are based on benchmarking with the
CHEMDAT8 model. However, the validation of these mass transfer rate equations are not
limited to activities performed on CHEMDAT8.
Initial validation of the volatilization mass transfer rate equations and the resulting
emission estimates compared measured and predicted fraction emitted from a pilot-scale and
several full-scale wastewater treatment units, including both aerated and quiescent units (RTI,
1985; this report also includes sample calculations for the mass transfer correlations and the
fraction emitted).
The CHEMDAT8 model was used to develop baseline emissions and emission impact
estimates used to support air emission standards developed under RCRA (40 CFR Part 264,
Subpart CC). The background information document for the air emissions standard includes
summary documentation of the emission model and presents data used to validate the model
(U.S. EPA, 1991). The comparisons to measured emissions data for five surface impoundments
and both aerated and quiescent tanks indicated that the emission model was relatively unbiased
(generating emission estimates higher than some measurements and lower than others). The
estimated emission rates agreed with all measured emissions within an order magnitude.
The CHEMDAT8 model equations have been used to develop baseline emissions and
emission impact estimates to support several other air emission standards, including the Benzene
Waste Operations National Emissions Standard for Hazardous Air Pollutants (NESHAP)
(40 CFR Part 61 Subpart FF) and the Hazardous Organic NESHAP (HON) (40 CFR Part 63,
Subparts F, G, H, and I). These equations have been publically available, and the emission
estimates that result from the application of these equations have been scrutinized. Comments
received on the CHEMDAT8 model and the emission estimates are a part of the public record in
the dockets of these rule makings. Although no model will perfectly predict the fate of all
contaminants from all types of units because of the variability in operating conditions and
microbial populations, these equations have consistently provided reasonable estimates of air
emissions over a wide range of aerated tank and surface impoundment operating characteristics.
In addition to previous validation of CHEMDAT8, the Wastewater Source Modules were
validated by comparing outputs to those calculated by CHEMDAT8. During internal testing of
the Wastewater Source Modules, the emissions fluxes calculated by the Surface Impoundment
Module were compared with those calculated using CHEMDAT8. Although these models
contain the same volatilization mass transfer rate equations, they employ different solids balance
algorithms, adsorption correlations, and biodegradation rate models. In addition, CHEMDAT8
does not consider infiltration/leachate flux from the surface impoundment. The Surface
Impoundment (and Aerated Tank) Module output agrees well with the CHEMDAT8 module
output. This validation effort is documented in RTI (1998).
The infiltration rate equations used in the Surface Impoundment Module are based on
EPACMTP (U.S. EPA, 1996b). This model has been peer reviewed (including a Scientific
Advisory Board [SAB] review) and publically available for more than 5 years. See Section 4.5.4
for a more detailed discussion of the validation of EPACMTP.
4-7
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.2 Land-based Source Modules and Watershed Module
This section documents the verification and validation activities for the Landfill, Waste
Pile, Land Application Unit (LAU), and Watershed Modules. The Landfill, Waste Pile, LAU,
and Watershed Modules are discussed collectively because of the overall similarity of their
functionality and their software components. Indeed, with respect to fundamental theory and
underlying calculations, there are relatively few differences among these four modules. The
primary differences are in specific parameter values (e.g., anaerobic biodegradation in the
landfill and waste pile versus aerobic biodegradation in the LAU), specific particulate emissions
processes modeled (e.g., vehicular activity emissions are not relevant to the waste pile), or
hydrology-related items (e.g., the landfill is not subject to erosion/runoff), rather than in
underlying mathematical models or software.
The Landfill, Waste Pile, LAU, and
Watershed Modules and associated data were
peer reviewed by the external reviewers
shown in the box. The peer-review
comments are provided in ERG (1999).
4.2.1 Module Description
The Land-based Source Modules
simulate partitioning and emission of
contaminants from land-based waste
management units (WMUs). Three Land-based Source Modules were developed for the 3MRA
modeling system to represent the major management practices where wastes are put into or on
the land for recycling, recovery, reuse, treatment, or disposal. These modules simulate waste
management practices in the following types of WMUs:
¦ Landfills, which are a common disposal unit for many nonliquid industrial
wastes;
¦ Waste piles, which are temporary storage areas on the ground for nonliquid
industrial waste, such as ash or slag; and
¦ Land application units (LAUs), which are used to reuse, treat, or dispose of
industrial waste in liquid, semiliquid, or solid form. Some wastes are used as a
soil amendment, which is a reuse practice; some wastes are applied to land for
treatment through biological degradation; and some wastes are applied to land as
a disposal method.
Each of the three Land-based Source Modules provides some similar and some different
features in terms of the ways contaminants of concern can be released to the environment. All
three modules have the potential to release contaminants to the air by volatilization or particle
entrainment, and to the vadose zone and ground water by leaching. Waste piles, because they
are elevated and have more surface area exposed, are assumed to have a greater potential for air
emissions than the other two source types. Waste piles and LAUs have the additional release
Peer Reviewers for Land-based Source Modules
and Watershed Module
¦ Dr. Anita Bahe of LYNX Group, Ltd.
¦ Dr. Kirk Brown of Texas A&M University
¦ Dr. William Inskeep of Montana State
University
¦ Dr. Clyde Munster of Texas A&M University
¦ Mr. William Norris of VA Department of
Environmental Quality
¦ Mr. Robert Wyatt of R.J. Wyatt and Associates
4-8
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
mechanism of erosion and runoff to the surrounding watershed and the nearest stream or other
waterbody. Landfills are assumed to be below grade and thus do not release contaminants
through erosion or runoff.
The three Land-based Source Modules were designed to provide estimates of annual
average contaminant concentrations in surface soil and contaminant mass emission rates to air,
surface water, and ground water, and to maintain mass balance between the source and all
release routes. The emission rate estimates are then used in the 3MRA modeling system, which
links source modules with environmental fate and transport modules. Figure 4-4 shows the
information flow for the Land-based Source Modules.
Key Data Inputs
• Precipitation rate
• Waste fraction
• Depth to aquifer
• Kd
Emission Rates
Air Module
Vadose Zone and
Aquifer Modules
Chemical Fluxes
Infiltration Rates
Surface Water
Module
Chemical Loadings, Soil Loadings
(LAU, WP only)
Landfill,
Waste Pile,
LAU
Modules
Farm Food Chain
Module
Soil Concentrations
Terrestrial Food
Web Module
Soil Concentrations
Soil Concentrations
Human Exposure
Module
(LAU, WP only)
Ecological
Exposure Module
Soil Concentrations
(LAU, WP only)
Figure 4-4. Information flow for the Land-based Source Modules in the
3MRA modeling system.
The Watershed Module estimates contaminant concentrations in soil resulting from aerial
deposition of contaminants throughout the area of interest (AOI) around each modeled site and
the resulting contaminant loadings to surface waterbodies from runoff and erosion. It also
estimates some hydrological inputs for the Surface Water Module (flows, eroded soil loads) and
the Vadose Zone and Aquifer Modules (infiltration rates). Figure 4-5 shows how the Watershed
Module fits into the 3MRA system.
4-9
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Key Data Inputs
• Annual infiltration rate
• Number of subbasins
• Erosivity factor
Surface
Water
Module
Chemical Loads, Flows
Eroded Soil Loads
Aquifer
Module
Infiltration Rates
Human
Exposure
Module
Watershed
Module
Deposition
Soil Concentrations
Air Module
Rates
Farm Food
Chain
Module
Soil Concentrations
Terrestrial
Food Web
Module
Soil Concentrations
Figure 4-5. Information flow for the Watershed Module in the 3MRA modeling
system.
4.2.2 Major Module Components/Functionality
The Land-based Source Modules contain the following four models:
1. The Generic Soil Column Model (GSCM) was developed to describe the
contaminant fate and transport in a porous medium, such as soil or waste. It
provides the vertical concentration profile in the soil/waste column at various
times.
2. The Local Watershed Model (for waste piles and LAUs) is based on mass
balances of solids and contaminants in the runoff and the top layer of the soil
column modeled by the GSCM. The "local watershed" comprises the land area
between the waterbody and the top of the hillside containing the WMU and may
include an upslope area, the WMU, and a downslope or buffer area between the
WMU and the waterbody.
3. The Particulate Emissions Model was designed to provide estimates of the
annual average emission rate of contaminant mass adsorbed to particulate matter
less than 30 |im in diameter. The release mechanisms considered differ for each
WMU, but may include wind erosion, vehicular activity, unloading operations,
tilling, and spreading/compacting operations.
4-10
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4. The Hydrology Model is a water balance algorithm that performs a daily water
balance for the surface soil layer. Moisture inputs are daily precipitation and
beginning soil moisture; outputs are daily runoff, evapotranspiration, percolation
to deeper layers, and changes in soil moisture. Runoff estimates are based on the
U.S. Department of Agriculture's (USDA's) Soil Conservation Service Curve
Number method (USDA, 1986). Evapotranspiration is based on the Hargreaves
equation (Hargreaves, 1975)—to estimate potential evapotranspiration—and a
function of the soil's current moisture, wilting point, and field capacity to
estimate actual evapotranspiration. Remaining soil moisture in excess of the field
capacity is released as percolation.
The Watershed Module includes one additional component:
5. The Regional Baseflow Model estimates waterbody baseflow (dry weather flow,
i.e., flow not derived from direct surface runoff) based on regional regression
models. Baseflow was assumed to be represented by the 30Q2 flow—the
minimum 30-day average flow with a return period of 2 years—which was
estimated for each of 18 Hydrologic Unit Code (HUC) areas in the conterminous
United States from STOrage and RETrevial Database (STORET) flow data. The
regional baseflow regression models were then developed by regressing the
estimated 30Q2 flow onto the drainage area (watershed area) tributary to the flow
gages for all flow gages within the HUC. The Watershed Module then estimates
baseflow on a watershed-specific basis by selecting the regression model for the
HUC region corresponding to the site being simulated and estimating baseflow as
a function of watershed surface area for each watershed at the site.
4.2.3 Summary of Verification
Internal verification activities and tests were performed for each of the four modules and
are documented in the following verification testing reports:
¦ Landfill Module Verification Testing (RTI, 2000d)
¦ Wastepile Module Verification Testing (RTI, 2000f)
¦ Land Application Unit Module Verification Testing (RTI, 2000b)
¦ Land Application Unit Module Verification Testing: ERRATA (RTI, 2000c)
¦ Watershed Module Internal Verification Testing (RTI, 2000g).
Following internal verification, independent external peer review and verification were also
performed on each module. The external verification is documented in the following reports:
¦ Independent Tests for Landfill Module (Tetra Tech, 2000d)
¦ Independent Tests for Waste Pile Module (Tetra Tech, 2000f)
¦ Implementation of the Test Plan for the HWIR99 Land Application Unit Module
(HGL, 2000)
¦ Independent Tests for the Watershed Module (PNNL, 2000).
4-11
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Finally, responses by the module developers to questions or issues raised by the independent
testers are documented in the following memoranda. All verification tests were considered to be
completed successfully.
¦ Response to Independent Reviewer Comments on Landfill Module Verification
Testing (Little and Baskir, 2000b)
¦ Response to Independent Reviewer Comments on Waste Pile Module Verification
Testing (Little and Baskir, 2000c)
¦ Response to Independent Reviewer Comments on LA UModule Verification
Testing (Little and Baskir, 2000a)
¦ Response to Independent Reviewer Comments on Regional Watershed Module
Verification Testing (Little and Baskir, 2000d).
The design of the verification tests followed an initial assessment of each module's
software "requirements," or desired functionality. Requirements included correctly performing
individual calculations in specific code sections, such as simulating contaminant fate and
transport in a soil column (the GSCM); overall module mass balance conservation; and system-
level requirements such as correctly writing warning and error messages and proper
communication with other modules. For each identified requirement, one or more test cases
were then proposed that specifically verify that requirement. Some test cases served to verify
more than one requirement.
Testing to verify internal calculations (as opposed to, for example, testing to verify that
the program is correctly reading or writing inputs and outputs) was the most demanding aspect
of the verification activities because of the quantitative standard set by the 3MRA modeling
system team. In addition to "verifying" an internal calculation (or set of calculations) by
qualitative, sensitivity-type tests, testing ensured that each calculation was "hand-reproducible,"
either by a hand calculation or replication of the calculation in another program, such as a
spreadsheet or independent program. For purposes of verifying the Land-based Source Modules
and Watershed Module, the definition of "calculation" was interpreted as including any line of
code that involved two or more variables and also performed any type of mathematical
operation. Most of the "hand" calculations were performed using a spreadsheet. One part of the
modules, the GSCM, was too complex to allow either a true hand verification or replication in a
spreadsheet. Therefore, a separate computer program was written in BASIC to independently
solve the GSCM governing equations. The BASIC programmer took the underlying GSCM
equations and independently developed a program to solve those equations, without considering
how they had been implemented in the 3MRA modeling system. Once that independent program
was completed, the outputs from the two codes were compared to verify the GSCM.
As previously explained, much of the functionality of the Land-based Source Modules
and the Watershed Module is common to all four modules, with code that is also common and
called by each module. It was only necessary to verify the common functionality and code for
one of these modules. EPA used the LAU Module to verify the common functionality, as well as
any functionality unique to the LAU Module. Once the LAU Module verification was
completed, verification of the remaining three modules was limited to functionality specific to
those modules. Tables 4-3 through 4-6 present a summary of the functional requirements and
4-12
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
number of test cases executed to verify these four modules. Detailed descriptions of the
requirements and test cases are provided in the verification testing reports (RTI, 2000b,c,d,f,g).
Table 4-3. General Requirements for Testing the Land Application Unit Module
Step
Description
Number of
Test Cases
1
Correctly read input data
1
2
Correct presimulation calculations
1
3
Correct operation of the GSCM
9
4
Correct year-end, post-processing calculations
1
5
Correct simulation-end, leachate flux postprocessing calculations
1
6
Correct operation of the subarea coupling processes
3
7
Correct operation of the water balance algorithm
4
8
Correct operation of the soil erosion algorithm
1
9
Correct operation of the particulate emissions algorithm
1
10
Correct writing of outputs
27
11
Correct operation of system requirements
7
12
Robustness
1
Table 4-4. General Requirements for Testing the Landfill Module
Step
Description
Number of
Test Cases
1
Correctly read input data
1
2
Correct presimulation calculations
1
3
Correct operation of the GSCM
8
4
Correct intrayear changes in modeled spatial domain for cell 1
1
5
Correct cell 1, year-end, postprocessing calculations
1
6
Correct cell 1, simulation-end, leachate flux postprocessing calculations
1
7
Correct end-of-simulation LF cell aggregation calculations
1
8
Correct operation of the water balance algorithm
4
9
Correct operation of the particulate emissions algorithm
2
10
Correct writing of outputs
17
11
Correct operation of system requirements
6
12
Robustness
1
4-13
-------
Section 4.0 Verification and Validation of the 3MRA Modeling System Modules
Table 4-5. General Requirements for Testing the Waste Pile Module
Step
Description
Number of
Test Cases
1
Correctly read input data
1
2
Correct presimulation calculations
1
3
Correct operation of the GSCM
10
4
Correct year-end, postprocessing calculations
1
5
Correct simulation-end, leachate flux postprocessing calculations
1
6
Correct operation of the subarea coupling processes
3
7
Correct operation of the water balance algorithm
4
8
Correct operation of the soil erosion algorithm
1
9
Correct operation of the particulate emissions algorithm
1
10
Correct writing of outputs
7
11
Correct operation of system requirements
7
12
Robustness
1
Table 4-6. General Requirements for Testing the Watershed Module
Step
Description
Number of
Test Cases
1
Correctly read input data
1
2
Correct presimulation calculations
1
3
Correct operation of the GSCM
4
4
Correct year-end, postprocessing calculations
1
5
Correct operation of the water balance algorithm
3
6
Correct operation of the soil erosion algorithm
1
7
Correct writing of outputs
12
8
Correct operation of system requirements
5
9
Robustness
1
4.2.4 Summary of Validation
Validation activities for the Land-based Source Modules and Watershed Module include
both implicit and explicit validation. Several of the modules' software components are based on
empirical data. Because these components are based on observed data rather than theoretical
models, they can be considered to be implicitly validated by definition if the code is verified as
operating correctly. Empirical components include aspects of the hydrology model (runoff
calculations based on the curve number methodology and the Hargreaves evapotranspiration
equation), the soil erosion calculations (based on the Universal Soil Loss Equation [USLE]), and
the particulate emissions calculations.
4-14
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Explicit validation involves comparing predicted results to monitored data or the results
of other models that are regarded as credible. Monitored data are preferred. EPA validated the
hydrology model component by comparison with EPA's Hydrologic Evaluation of Landfill
Performance (HELP) model and the LAU Module by comparison of simulated half-lives of
dioxin in soil with monitored half-life data. In addition, EPA validated the LAU Module using
data from a large soil-column study. These three validation activities are described below.
HELP Model Comparison. The HELP model benchmarking tests are fully described in
Test Case 7.3 of the LAU Module verification testing report (RTI, 2000b). The benchmarking
tests compared predicted runoff and infiltration for six sites throughout the country (San Diego,
CA; Miami, FL; Madison, WI; Denver, CO; Dallas, TX; and Seattle, WA) using 5 years of data
per site. The hydrology model in the Land-based Source and Watershed Modules is not
completely functionally equivalent to the HELP model, so close agreement between the two was
not necessarily expected. Rather, EPA expected long-term average results to be in reasonable
agreement. The comparative results were mixed. At some sites, the two models' predictions
were quite similar. At other sites, the predictions showed relatively large differences. With
regard to differences in infiltration, no consistent bias in the 3MRA modeling system hydrology
model predictions versus the HELP model predictions was apparent. With regard to differences
in runoff, the 3MRA modeling system hydrology model predicted more runoff than the HELP
model for all tested sites. To determine whether this apparent bias in runoff prediction was of
concern, EPA compared runoff estimates from both models at the six sites to long-term average
observed runoff as reported in the Water Atlas (Geraghty et al., 1973). The 3MRA modeling
system hydrology model's predictions were in closer agreement with observed runoff than were
the HELP model's predictions; however, this conclusion should be regarded as tentative because
of the relatively limited number of sites compared and uncertainty about the relevance of the
specific comparisons made. In summary, the benchmarking analysis suggested that the 3MRA
modeling system hydrology model's results were adequate for the 3MRA modeling system's
national screening-level purposes.
LAU Dioxin Half-life Comparisons. The LAU dioxin half-life comparisons were
performed as part of an application of the LAU Module to simulate the persistence of dioxin
compounds in sewage sludge applied to agricultural lands. That work, including the half-life
validation comparisons, is fully described in Exposure Analysis for Dioxins, Dibenzofurans, and
Co-Planar Poly chlorinated Biphenyls in Sewage Sludge (RTI, 2001). The LAU Module was
used in that project to provide LAU concentrations of dioxins in agricultural soils, and those
concentrations were then used in a probabilistic (Monte Carlo) human health risk assessment.
For the half-life comparison, the output LAU soil concentrations associated with specific
percentiles (10th, 20th, 50th, 75th, 90th, 95th, and 99th) of the final risk distribution (based on 3,000
Monte Carlo simulations) were used to calculate the half-life for each percentile. Each risk
distribution percentile represents a somewhat different physical environment, which gives rise to
different half-lives. Half-lives were estimated from the module's outputs by finding the peak
concentration in soil—corresponding to the end of the sludge application period—and then
counting the years until the peak concentration had been reduced by one-half. The resulting
range of simulated half-lives was compared to measured half-lives as reported in the technical
literature. The range of half-lives over the selected percentiles was 20 to 48 years, which is in
reasonable agreement with observed half-lives at several monitored sites. Thus, EPA concluded
4-15
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
that the monitored data corroborated the modeled results for highly persistent contaminants, at
least in a broad sense.
Soil-Column Study Data. The LAU Module has been validated using data from a large
soil-column study conducted to investigate the behavior of organic chemicals (including
halogenated aliphatic hydrocarbons, substituted benzenes, and phenols) during infiltration of
municipal wastewater into soil by the Ground Water and Ecosystem Restoration Division
(GWERD) of EPA's National Risk Management Research Laboratory (NRMRL). The
experiment design and conditions were very close to the assumptions and application conditions
of the LAU Module, making these data suitable for use in evaluating the LAU Module.
The evaluation on LAU source module included:
¦ Whether volatilization of organic chemicals could be correctly accounted for;
¦ Whether the conceptualizations on different attributes of the module (such as
boundary conditions and the assumption of first order decay for contaminant
transformation) are adequate;
¦ Whether the quasi-analytical approach employed for solving the mathematical
model to describe the flow, fate and transport is appropriate; and
¦ Whether the LAU thickness and temperature parameters have significant effects
on the amount of volatilization of organics.
A stand-alone LAU program was obtained by modifying the LAU Module from the
3MRA modeling system to achieve the small intervals of time and space needed to compare the
results to the experimental lab data. This modified LAU program was compared to the LAU
Module and verified to be consistent with the original module. The necessary input parameters to
the LAU program were obtained from the lab experiment design and literature review. The
simulated and the observed volatilization rates were compared to validate the LAU Module.
Sensitivity analyses were implemented to examine the effect of thickness and temperature
parameters of the LAU Module on the evaluation. Furthermore, the evaluation was performed
based on the chemical categories and the volatility of organic compounds.
Overall, the volatilization rate modeled by the LAU program is in the right order of
magnitude for all categories of compounds involved in the experiment, although the simulated
volatilization is consistently lower than the observed volatilization for highly volatile organic
compounds. Moreover, sensitivity analyses indicated that the model outputs of LAU program are
not sensitive to the thickness parameter for volatilization of organic chemicals, but show certain
sensitivity to changes in temperature.
4-16
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.3 Air Module
This section documents the
verification and validation activities for the
Air Module. The Air Module was reviewed
by the external reviewers listed in the box.
Peer Reviewers for Air Module
¦ Dr. Steven Hanna of George Mason
University
¦ Dr. Fred Mogolesko of M&L
Environmental Consultants
¦ Dr. Bruce Turner of Trinity Consultants
4.3.1 Module Description
The Air Module estimates the annual average air concentration of dispersed contaminants
and the annual deposition rates of vapors and particles at various receptor points in the area of
interest (AOI). This module simulates the transport and diffusion of contaminants in the form of
volatilized gases or fugitive dust emitted from area sources into the air. The predicted air
concentrations are used to estimate bio-uptake into plants, and human exposures due to direct
inhalation. The predicted deposition rates are used to determine contaminant loadings to farm
crops and soils, watershed soils, and surface waterbodies. Figure 4-6 shows the information
flow for the Air Module.
Key Data Inputs
1 Source height
1 Wind speed
1 Annual precipitation rate
Land-based
Source
Modules
Surface
Impoundment/
Aerated
Tank Source
Modules
Emission
Rates
Emission
7
Air Concentrations
Air
Module
Rates
Deposition Rates
Deposition Rates
Air Concentrations
Deposition Rates
Farm Food
Chain Module
Watershed
Module
Human
Exposure
Module
Surface Water
Module
Air Concentrations
Deposition Rates
Terrestrial Food
Web Module
Figure 4-6. Information flow for the Air Module in the 3MRA modeling system.
4-17
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.3.2 Major Module Components/Functionality
The Air Module performs four major functions, as follows:
1. Characterize source-specific parameters. For each AOI, the Air Module
characterizes emission sources based on waste management unit (WMU)
dimensions, wastes being managed, and the site-layout. As an option, the module
can calculate the source-specific long-term average particulate mass fraction
distribution from the outputs of the Land-based Source Modules (Landfill, Waste
Pile, or Land Application Unit [LAU] Module). All sources are assumed to be at
ground level except for waste piles and aerated tanks. The height of these units
are imputed from the WMU area and waste loading data.
2. Calculate receptor locations (polar grid or site specific). The Air Module
provides the option to model directly to all site-specific output coordinates needed
by the 3MRA modeling system or to model to a fixed set of polar coordinates and
then use a two-dimensional (2-D) cubic spine method to interpolate from the
polar set to the larger set of interest. The spline interpolation is used to reduce the
ISCST3 run time.
3. Calculate receptor-specific concentration and deposition estimates. The Air
Module calculates annual average air concentration and deposition rates for each
receptor location specified. Concentrations and deposition rates calculated include
- Air concentration of vapors,
- Air concentration of particles,
- Wet deposition rate for vapors and particles, and
- Dry deposition rate for particles.
4. Calculate contaminant-specific annual average concentrations and
deposition rates. The Air Module converts the receptor-specific concentrations
and deposition rates based on unit emission rates (e.g., 1 g/m2-s) to contaminant-
specific estimates by multiplying the values by the contaminant-specific emission
rate for each year.
ISCST3 may be run by the 3MRA modeling system during a run, or it may be run outside
of the 3MRA modeling system and the results of those runs used when the Air Module runs.
4.3.3 Summary of Verification
EPA has used the ISCST model to simulate sources of nonreactive pollutants since 1979
(Bowers and Anderson, 1981). Over the years, EPA has updated ISCST as needed to address
particular applications. A history of the development of ISCST may be found in Irwin (2002).
Throughout ISCST's development, EPA has undertaken numerous verification efforts to show
that the FORTRAN program accurately solves the intended equations. These efforts are
described in more detail in Appendix A, and include the following:
4-18
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Creation of ISCST2. EPA performed a significant overhaul of the program in
1992 to produce ISCST2. During this effort, EPA did a substantial amount of
recoding to make the code more modular. EPA developed and followed a
detailed test plan to ensure that the re-coded model produced results equivalent to
the original model (U.S. EPA, 1992a).
Area source algorithm. A comparison of area source algorithms used by various
dispersion models (U.S. EPA, 1989) showed that the finite line source approach
used in ISCST to model area sources predicted unrealistic concentrations for
receptors located in or near the area source. In 1992, EPA implemented and
tested a new area source algorithm that uses an integrated line source algorithm
(U.S. EPA, 1992b; U.S. EPA 1992c). EPA released this new area source
algorithm to the public in draft form as AREA-ST.
Dry deposition velocity of particles. The need for better estimates of the
intermedia transfer of pollutants from the atmosphere to land, water, and
vegetation to support multimedia environmental analyses prompted a comparison
of existing algorithms for calculating the dry deposition velocity of particles (U.S.
EPA, 1994b). This study also compared plume depletion algorithms. In addition,
the deposition algorithm in the original ISCST2 model was not designed for small
particles. As a result, EPA implemented new deposition velocity and plume
depletion algorithms and released them to the public in draft form as DEP-ST.
Wet and dry deposition and terrain. In 1993, the EPA Administrator
announced that risk assessments including indirect exposure would be required
for permitting hazardous waste incinerators and industrial furnaces. No
regulatory models capable of quantifying wet and dry deposition in all terrains
existed, so EPA Region 5 sponsored further development of the ISCST2 model to
address this need. In this effort, EPA made several revisions to ISCST. First,
EPA combined the AREA-ST and DEP-ST versions of ISCST2 into a single
version referred to as ISC-COMPDEP. EPA performed testing to demonstrate
that ISC-COMPDEP produced equivalent results to AREA-ST and DEP-ST.
Second, EPA added the methodology used in the COMPLEX I model for
modeling point sources in complex terrain to ISC-COMPDEP and performed tests
to ensure that the results were equivalent to those of the original COMPLEX I
model. Third, EPA selected and implemented wet deposition and depletion
algorithms in ISC-COMPDEP. The development and testing of the
ISC-COMPDEP model, including tests to establish equivalent results with the
component models, is documented in Strimaitis et al (1993).
ISC-COMPDEP was renamed ISCSTDFT, and proposed as part of Supplement C to the
Guideline on Air Quality Modeling. As part of the rulemaking process, the model and
associated documentation was subject to public review and comment. Supplement C was
promulgated on August 9, 1995 (60 FR 40465). At that time, the model was renamed ISCST3
and released to the public for use in regulatory applications.
4-19
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
ISCST3 was further modified to meet some of the needs and constraints of the proposed
3MRA application. A major area of concern for application in the 3MRA modeling system was
ISCST3's runtime. EPA made two modifications to ISCST3 to address the runtime issue.
¦ Sampled Chronological Input Methodology. EPA added an option called the
Sampled Chronological Input Methodology (SCIM) to sample from the
meteorological data at regular intervals rather than run each consecutive hour.
This option is used to calculate annual values for concentration and deposition.
EPA tested the use of this option over a range of sources, climate regimes, and
sampling rates (U.S. EPA, 1998d). Based on the results of that testing, EPA
recommended a sampling interval for the 3MRA modeling system. The 3MRA
modeling system does not require the use of the SCIM option, nor does it require
the use of the recommended sampling interval if the SCIM option is used.
¦ Plume depletion algorithm. The plume depletion algorithm for area sources is
numerically intensive, contributing to the long runtime for area sources. EPA
conducted a comparison of available depletion schemes and implemented and
tested a new algorithm (Venkatram, 1998) to replace the existing scheme (Horst,
1983).
In addition to the testing of the individual model components as described above, the
ISCST model, preprocessors, and postprocessors that comprise the Air Module have undergone
internal and independent testing as part of the software development process supporting the
3MRA application. Table 4-7 summarizes the functional requirements and the number of test
cases executed to verify the Air Module.
Table 4-7. General Requirements for Testing the Air Module
Step
Description
Number of
Test Cases
1
Correctly manage files generated by the pre- and postprocessor
14
2
Correctly detect when the core model does not need to run
1
3
Read normalized concentration and deposition values from the air database correctly
and output annualized values correctly
13
4
Correctly implement the spline option and produce a reasonable spline surface
3
5
Correctly implement the massfrax option
3
6
Correctly estimate plume depletion due to dry and wet deposition.
3
7
Ensure that the use of the SCIM option provides estimates close to those of running
the full meteorological record
2
4.3.4 Summary of Validation
ISCST3 provides point estimates of concentration, dry deposition, and wet deposition.
Although numerous studies have compared concentrations predicted by ISCST3 (or its
4-20
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
predecessors) with observed data, the deposition estimates are harder to validate because field
studies of dry and/or wet deposition flux from point or area sources are seldom found in the
literature. The subsections that follow discuss validation studies that have been performed on
the various components of the ISCST3 model with the intent to show that the modeling approach
adequately represents the physical processes in the atmosphere. Appendix A describes the
validation of the Air Module; a summary is provided here.
4.3.4.1 Concentration estimates. The ISCST model (Bowers and Anderson, 1981)
evolved from earlier plume models such as CRSTER (U.S. EPA, 1977), inheriting the same
dispersion algorithms and adding building downwash algorithms. Therefore, evaluation studies
done for CRSTER (and its derivatives) are also applicable to ISCST and subsequent revisions to
it. Concentration estimates from CRSTER were extensively evaluated for a number of point
sources using databases for coal-fired power plants. These studies are listed in Table 4-8.
Table 4-8. Coal-Fired Power Plants used for ISCST3 Evaluation
Database
Location
Stack height(s)
Terrain type
Reference
Clifty Creek
Indiana
3 stacks; all 208 m
low ridges and rolling hills
Londergan et al (1982)
Muskingum
Ohio
2 stacks; all 252 m
low ridges and rolling hills
Cox and Moss (1985)
Paradise
Kentucky
3 stacks; 183 m,
183 m, 244 m
flat terrain surrounded by rolling
hills
Cox etal (1987)
Kincaid
Illinois
1 stack; 187 m
flat terrain
Cox etal (1986)
Separate studies were performed to evaluate the concentrations predicted for area
sources. Field studies of area sources, particularly ones measuring impacts near and within the
source, are scarce. Therefore, EPA used alternative methods for evaluating the algorithm before
recommending it for regulatory modeling applications. As described above, EPA conducted a
comparison of area source algorithms used by various dispersion models (U.S. EPA, 1989) to
select an algorithm for inclusion in ISCST2. In this study, EPA developed a set of prediction
scenarios for testing and comparing the algorithms. A follow-on study examined the sensitivity
of the selected area source algorithm across a range of source characteristics and compared the
results to those from the original ISCST model algorithm. Finally, EPA compared results from
the selected algorithm to data from a wind tunnel study (U.S. EPA, 1992d).
4.3.4.2 Dry deposition estimates. ISCST3 calculates particle dry deposition flux by
multiplying the air concentration by the deposition velocity. Studies of particle deposition flux
attributable to individual sources are difficult to find. In the absence of flux studies to be used
for validation, EPA has focused on validating the algorithm to estimate particle deposition
velocity and relied on the previously noted validations of the concentration algorithms. To select
the dry deposition algorithm to be added to ISCST3, EPA evaluated and compared a number of
deposition velocity algorithms and implemented the "most appropriate approach" (U.S. EPA,
1994b). EPA evaluated the algorithms on their ability to parameterize important physical
processes while requiring only readily available meteorological, chemical, and physical input
data. In addition, Schwede and Paumier (1997) did sensitivity tests that exercised the algorithms
and demonstrated the ability of the model to produce estimates within expected ranges.
4-21
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Another aspect of the deposition calculation is accounting for the material lost from the
plume as a result of deposition. This is referred to as depletion. For use in 3MRA, EPA changed
the depletion algorithm in ISCST3 to one that was shown to be faster and more robust
(Venkatram, 1998). That algorithm was validated against a full numerical solution to the eddy
diffusivity equation (Venkatram, personal communication).
4.3.4.3 Wet deposition estimates. ISCST3 is used in 3MRA to calculate the wet
deposition of both particles and gases. Wet deposition is dependent on the air concentration, the
scavenging coefficient, and the precipitation rate. For wet deposition of particles, the scavenging
coefficient is specified for ISCST3 by particle size category. For wet deposition of vapors, the
scavenging coefficient is contaminant-specific. To reduce the number of model runs required for
a 3MRA application, EPA configured the Air Module to use a single vapor-phase scavenging
coefficient value for all contaminants that causes them to be scavenged as if they were small
particles. The general approach for calculating the wet deposition flux and resulting depletion
was proposed by Maul (1980) based on an analysis of ambient data.
4.4 Surface Water Module
The Surface Water Module is based on EPA's legacy model Exposure Analysis
Modeling System (Exams II), which has been thoroughly verified and validated in numerous
applications. Exams II was extended for use in the Surface Water Module primarily through
development of the pre- and postprocessor, Exams 10. Exams 10 is the interface between
Exams II and the rest of the 3MRA modeling system. It reads data from other modules and
3MRA modeling system databases, builds the Exams input files describing the waterbody
environment and chemical properties, builds a command file that specifies the contaminant
loading history, and controls the Exams simulation. Exams 10 passes control to Exams II,
which conducts the simulation and produces intermediate results files. Exams 10 then processes
the intermediate files and passes the output data back to the proper 3MRA modeling system
database. While the Exams component of the Surface Water Module is a fully functional model
independent of the 3MRA modeling system, it is driven and constrained in various ways by
Exams 10 and the 3MRA modeling system databases.
The external peer reviewers for the
Surface Water Module are listed in the box.
4.4.1 Module Description
The Surface Water Module simulates
contaminant concentrations in surface
waterbodies throughout the AOI around each
site modeled. Inputs to the Surface Water
Module include contaminant loadings from direct air deposition onto surface waters,
contaminant loadings from runoff and soil erosion from land areas associated with sources (LAU
and waste pile only), contaminant loadings from contaminated ground water plumes that are
intercepted by surface waters, contaminant loadings from runoff and soil erosion from
watersheds in the AOI, and hydrological inputs (flows, soil loads) from watersheds. Surface
Water Module outputs include water column and sediment concentrations, which are then used
Peer Reviewers for Surface Water Module
¦
Dr. Mustafa Aral of Georgia Institute of
Technology
¦
Dr. Anthony Donigian of AQUA-TERRA
Consultants
¦
Dr. Wilbert Lick of University of California,
Santa Barbara
4-22
-------
Section 4.0 Verification and Validation of the 3MRA Modeling System Modules
by the Aquatic Food Chain Module, Farm Food Chain Module, and Ecological Exposure
Module. All inputs and outputs are annual average time series. Figure 4-7 shows the
information flow for the Surface Water Module.
Key Data Inputs
• Flow Rate
• Total suspended solids
Deposition Rates
Water Column and
Sediment Concentrations
Chemical Loadings
Water Column
Soil Loadings
Concentrations
Chemical Loads, Flows,
Water Column and
Eroded Soil Loads
Sediment Concentrations
Water Column and
Chemical Loads from
Sediment Concentrations
Groundwater Interception
Surface
Water
Module
Farm Food
Chain
Module
Land-based
Source
Modules
Vadose Zone
and Aquifer
Modules
Ecological
Risk
Module
Aquatic Food
Web Module
Watershed
Module
Ecological
Exposure
Module
Air Module
Figure 4-7. Information flow for the Surface Water Module in the 3MRA modeling system.
4.4.2 Major Module Components/Functionality
The major tasks performed by the Surface Water Module to simulate contaminant
concentrations throughout the surface waterbody network in the AOI are as follows:
1. Construct waterbody network. For each AOI, the Surface Water Module
identifies the streams, wetlands, ponds, lakes, and bay reaches to be modeled and
assigns lengths, areas, depths, and volumes to them.
2. Route hydraulic flow and solids through the waterbody network. The Surface
Water Module conducts water and solids balances for each waterbody in each
year of the simulation.
3. Construct and solve the mass balance equations describing contaminant fate
and transport throughout the waterbody network. The Surface Water Module
calculates the contaminant concentration in each waterbody for each year.
Outputs include total water column concentration and dissolved concentration.
4-23
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.4.3 Summary of Verification
The Surface Water Module was subjected to a series of tests to verify that it accurately
performs its prescribed computations. These tests were intended to verify that the Surface Water
Module correctly implements EXAMS within the 3MRA modeling system. In addition, some
verification testing was performed to ensure that the Surface Water Module accurately
reproduces available analytical solutions for a set of simplified test cases. The verification
testing is described in detail in the EPA document, Quality Assurance Verification and
Validation Tests for the Exposure Analysis Modeling System—EXAMS (U.S. EPA, 2002a). A
summary of that testing is presented here; the complete document is included as Appendix B.
The Surface Water Module must automatically construct and execute a simulation for
each waterbody system at a site using the databases that are created by each Monte Carlo
iteration in a 3MRA modeling system implementation. As mentioned previously, the general
steps in a simulation are to construct a proper waterbody network, conduct water and solids
balances, and calculate contaminant transport and fate. These general steps also formed the
structure of the module testing program. The general requirements of the Surface Water Module
are presented in Table 4-9 along with the number of verification test cases examined for each
requirement.
Table 4-9. General Requirements for Testing the Surface Water Module
Step
Description
Number of Test Cases
1
Construct the water body network
3
2
Construct dispersive exchanges
3
3
Conduct the water balance
3
4
Calculate solids transport
3
5
Calculate conservative contaminant transport
3
6
Calculate ionic speciation for ionizing organic chemicals
2
7
Calculate partitioning to solids and DOC
2
8
Calculate volatilization loss
13
9
Calculate contaminant transformation
5
10
Test for robustness
16
All tests in Steps 1 through 9 were performed using one of three simplified waterbody
networks at a hypothetical site (Figure 4-8). The first two networks are simple one-reach water
bodies. Network 1 is a pond with outflow, and network 2 is a lake with no surface outflow.
Network 3 is composed of four reaches: two headwaters reaches (1 and 2, a pond and a
wetland), which are connected to reach 3 (a stream), which is connected to exiting reach 4 (a
lake).
Within each reach, the surface-water modeling system disaggregates the waterbody into a
set of "compartments." The Surface Water Module constructs a differential equation for each
compartment, and then solves the resulting system of equations representing interactions among
4-24
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
WBN 1:
Pond
WBN 2:
Lake
Reach 2: Wetland
Reach 3: Stream
WBN 3:
Simple Network
Reach 4: Lake or Bay
Reach 1: Pond
WBN = waterbody network
Figure 4-8. Surface Water Module test waterbody networks.
all modeled compartments. For example, for the pond reach-type, the Surface Water Module
simulates three compartments: the littoral zone, the surficial benthic layer, and the underlying
benthic layer. For the lake reach-type, it simulates four compartments: the epilimnion (water
column above the thermocline), the hypolimnion (water column below the thermocline), the
surficial bethnic layer, and the underlying benthic layer.
All test cases were verified and the module requirements in Table 4-9 were considered to
be met.
4.4.4 Summary of Validation
Since Exams was first released in 1983, it has been applied many times to waterbodies
throughout the world. This section summarizes studies that report model performance against
measured data in either a calibration or validation mode. These studies cover a wide range of
waterbodies, chemicals, and fate processes, including
¦ Environments—small streams, rivers, ponds, rice paddies, and bays;
¦ Chemicals—dyes, herbicides, insecticides, phenols, and other organic chemicals
with a variety of chemical properties; and
¦ Fate processes—advection, sorption, sediment-water exchange, volatilization,
hydrolysis, photolysis, water column and benthic biodegradation, and oxidation.
4-25
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Chemical coefficients were supplied from the open technical literature and, in some
cases, from site-specific experiments. Table 4-10 lists the calibration/validation case studies
Table 4-10. Exams Calibration/Validation Case Studies
Reference
Description
Games (1982)
EXAMS was applied to model the fate of linear alkylbenzene sulfonate (LAS) discharged in
treated sewage effluent in Rapid Creek, SD. The important chemical fate reactions for LAS are
partitioning to solids and biodegradation, both of which vary with the isomer chain length. The
most sensitive model parameters were found to be the sediment-water dispersion coefficient
and the sediment biodegradation rate constant. Because these are two of the least-understood
parameters, the authors suggest that "wide error limits would need to be placed on predicted
LAS concentrations... using EXAMS, and probably any evaluative model in a safety
assessment." Nevertheless, the authors concluded that "given the proper assumptions, EXAMS
can successfully predict the concentration of LAS resulting from its steady-state input to a
flowing stream."
Pollard and Hern
(1985)
EXAMS was tested against phenol data in an 18-mile reach below a steel plant discharge to the
Monongahela River, PA. After calibration, the authors observed a "nearly perfect fit of
observed and predicted concentrations by station."
Schramm et al.
(1988)
EXAMS was tested by simulating Disperse Yellow Dye 42 introduced to an experimental
outdoor pond. The model network included a water column segment and two benthic layers.
Although some departures between predicted and observed concentrations were observed
during the course of the study, the authors characterized the overall model performance as
good.
Kolset and
Heiberg (1988)
EXAMS was calibrated and tested with three contaminants from a kraft mill effluent to a
heavily polluted bay on the east coast of Sweden. Four compounds identified in the effluent
were used to test EXAMS - 2,4,6-trichlorophenol (2,4,6-TrCP), 3,4,5-trichloroquaiacol (3,4,5-
TrCG), tetrachloroguaiacol (TeCG), and tetrachlorocatechol (TeCC). Calculated and measured
concentrations agree well for 2,4,6-TrCP, 3,4,5-TrCG, and TeCG, but not for TeCC. The
authors conclude that EXAMS is capable of predicting the concentrations of some selected
chlorophenolics in the Norrsundet area with reasonable accuracy. EXAMS describes all
important transformation and loss processes except for sedimentation reasonably well.
Woodrow et al.
(1990)
Volatilization flux was measured for three rice herbicides—MCPA, molinate, and
thiobencarb—and one insecticide—methyl parathion—from a laboratory chamber and two
flooded rice fields. The flux measurements were compared to predicted fluxes using EXAMS
with chemical properties and chamber or field conditions as input. The normalized
volatilization flux values predicted by EXAMS compared well overall with the observed
values, within 10%-20% for molinate, a factor of 2 for MCPA acid, and a factor of 3 for
methyl parathion. The calculated flux rate for thiobencarb, however, was low by a factor of 5.
The authors attribute this discrepancy to possibly incorrect vapor pressure and/or solubility
data. The authors conclude that EXAMS "appeared to be promising as a predictive tool for
estimating volatilization, when the appropriate chemical properties and environmental
conditions were used as input data."
Tynan et al.
(1991)
EXAMS was tested against data for a variety of contaminants in a 7 km reach below a sewage
treatment works (STW) effluent to a small, unnamed lowland river in England. In general,
EXAMS predictions are described as being reasonably close to measured values, with "fairly
close predictions" for styrene, m-dichlorobenzene, and p-dichlorebenzene. The authors
conclude that EXAMS produces quantitative predictions that compare well with observations
for those contaminants for which reliable rate data exist.
(continued)
4-26
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Table 4-10. (continued)
Reference
Description
Cousins et al.
(1995)
EXAMS was tested against aniline and lindane data in a 7 km reach below an STW effluent to
the large lowland River Calder in West Yorkshire, England. EXAMS predicted very slight
losses for both contaminants, a result consistent with two surveys. Fairly good correlations
were achieved between measured and predicted dissolved water column concentrations, with
predictions falling within a factor of 2 of station means. Measured suspended particulate
concentrations in the river were significantly higher than predicted, probably because of
limitations in the equilibrium partitioning assumption. The measured high particulate
concentrations from the STW appeared to be either irreversibly or strongly bound with a slow
desorption rate. By contrast, predicted bed sediment concentrations were "a little lower than
measured," all falling within an order of magnitude of observations. From this study, the
authors conclude that EXAMS is useful for predicting the fate of non-ionic organic chemicals
in rivers, provided that adequate physicochemical and environmental data are available.
Armbrust et al.
(1999)
EXAMS was calibrated to bensulfuron methyl (BSM) and azimsulfuron (AZM) data taken in
experimental lysimeters that contained 5 cm of water overlying 50 cm of paddy soil. EXAMS
successfully predicted the partitioning and degradation reactions that led to observed
contaminant half-lives in paddy water. Predicted water column concentration responses
generally matched the observed data. EXAMS overestimated soil concentrations by factors of
2 to 4. The authors conclude that a "definitive characterization of the rate of degradation and
mobility of the two contaminants at a specific site would require additional information on
environmental parameters and site-specific soil-chemical interactions."
performed to date in chronological order. These case studies are described in more detail in U.S.
EPA (2002a). EXAMS requires a combination of environmental, chemical, and loading data in
order to properly specify the model parameters. Erroneous, uncertain, or missing data can result
in improper model parameterization, which leads to errors in model predictions. Overly simple
process equations can also lead to errors in model predictions. The case studies referenced here
highlight both parameter uncertainty and model uncertainty. However, despite these sources of
uncertainty, it appears that Exams, and thus the Surface Water Module, is able to predict the
concentrations of most organic chemicals within a factor of 2 or better in the water column, and
within an order of magnitude in the sediment.
4.5 Vadose Zone and Aquifer Modules
The Vadose Zone and Aquifer Modules simulate the migration of chemical constituents
in the subsurface and were extracted from EPA's Composite Model for Leachate Migration with
Transformation Products (EPACMTP) (U.S. EPA, 1996a,b,c; 1997a). This model is used in
EPA regulatory efforts by OSW and has been subject to extensive peer review and public
comment. EPACMTP is the best currently
available tool to predict potential ground
water pathway exposure at a downstream
receptor well for regulatory development
purposes.
Peer Reviewers for Subsurface Module
¦ Dr. Craig Forster of University of Utah
¦ Dr. M. Akram Hossain of Washington State
University
¦ Dr. Carl Mendoza of University of Alberta
¦ Dr. Frank Schwartz of Ohio State University
4-27
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
The Vadose Zone and Aquifer Modules were reviewed by the external reviewers listed in
the box.
4.5.1 Module Description
The Vadose Zone and Aquifer Modules simulate the subsurface movement of chemical
constituents contaminants in leachate from surface impoundments, landfills, waste piles, and
land application units (LAUs) to downgradient drinking water wells and waterbodies. The
modules are not used for aerated tanks, because tanks are assumed not to leak. Figure 4-9 shows
the information flow for the Vadose Zone and Aquifer Modules.
Key Data Inputs
• Well location
• Fraction organic carbon
Chemical Loads from
Subsurface Interception
Infiltration Rates
Ground Water
Concentrations
Chemical Fluxes
Infiltration Rates
Subsurface
Concentrations
Surface
Water
Module
Source
Modules
Human
Exposure
Module
Farm Food
Chain
Module
Watershed
Module
Vadose Zone
and Aquifer
Modules
Figure 4-9. Information flow for the Vadoze Zone and Aquifer Modules in the 3MRA
modeling system.
The Vadose Zone and Aquifer Modules simulate the fate and transport of dissolved
contaminants from a point of release at the base of a WMU, through the underlying soil (the
vadose zone), and through a surficial aquifer (or ground water source). Module outputs include
ground water contaminant concentrations in wells, which are used by the Human Exposure
Module to estimate exposures through drinking water and showering and by the Farm Food
Chain Module to estimate contaminant concentrations in beef and milk from farm well use; and
contaminant fluxes into waterbodies, which are used by the Surface Water Module, along with
contaminant fluxes from atmospheric deposition and overland flow, to estimate contaminant
concentrations in streams, lakes, and wetlands.
The Vadose Zone and Aquifer Modules are used by the 3MRA modeling system only if
there are wells or downgradient streams, lakes, or wetlands at a site. Waterbodies are
downgradient if they are in the direction of ground water flow away from the WMU.
4.5.2 Major Module Components/Functionality
The Vadose Zone and Aquifer Modules perform the following functions:
4-28
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
1. Model vadose zone flow and transport. The one-dimensional (1-D) Vadose
Zone Module simulates infiltration and dissolved contaminant transport, by
advection and dispersion, from the WMU through the soil above the water table
(i.e., the vadose zone) to estimate the contaminant and water flux to the
underlying ground water.
2. Model ground water flow and transport. The quasi-3-D Aquifer Module
simulates ground water flow and contaminant transport, by advection and
dispersion, from the base of the vadose zone to estimate contaminant
concentrations in drinking water wells and contaminant discharge fluxes to
waterbodies.
3. Model subsurface chemical reactions. Both the Vadose Zone and Aquifer
Modules simulate sorption to soil or aquifer materials and biological and
chemical degradation, which can reduce contaminant concentrations as
contaminants move through soil and ground water. In cases where degradation of
a contaminant yields other contaminants that are of concern, the Vadose Zone and
Aquifer Modules can account for the formation and transport of up to six different
daughter and granddaughter degradation products. The modules use sorption
isotherms for metal contaminants, which allow adjustment of sorption behavior to
account for varying metal concentrations and geochemical conditions.
4.5.3 Summary of Verification
Verification of EPACMTP began in 1991. These activities compared numerical and
analytical results from EPACMTP to analytical and numerical solutions from other verified
sources. Initially, the flow and transport mechanisms in the vadose zone and saturated zone
(aquifer) models were verified. Testing continued in 1999 on the portions of EPACMTP
extracted for use in the Vadose Zone and Aquifer Modules. In 2000, EPA conducted
comprehensive verification efforts on all components of the Vadose Zone and Aquifer Modules.
These verification efforts are summarized in "Verification and Validation of the EPA's
Composite Model for Transportation Products (EPACMTP) and its Derivatives" (HGL, 2003)
That summary is included here as Appendix C. A synopsis of this summary is provided here and
in Table 4-11. For additional information, please refer to the original source documents.
EPA's ORD conducted its first verification of EPACMTP in 1992. EPA tested 14
components of the model, including aquifer flow, vadose-zone transport, multiple species
transport with decay, and full-3D aquifer flow and transport. The tests were conducted to
confirm the developer's verification results using the same data sets and to provide an
independent verification of the model using alternate test criteria. Analysts used EPA's
verification activities to identify some technical limitations of EPACMTP. As a result, the
model's code was modified to expand its capabilities.
4-29
-------
Table 4-11. Summary of Verification Activities for the 3MRA Vadose Zone and Aquifer Modules1
Name of Verification
Activity (Year) (Original
Source Document)
Component Tested
Verification Tools
Number
of Test
Cases
Results
U.S. EPA Office of Research
and Development (1992-
1993) (U.S. EPA 1992e;
HGL, 1993)
Composite model—flow and
transport
Reverification of original test problems and data
files; independent verification using alternative
test criteria
21
Some technical limitations
identified; modifications
were conducted
Module-Level Verification
(1993-1994) U.S. EPA
1996d, U.S. EPA 1996e)
Vadose Zone Module—Infiltration
and transport
Compared to analytical and numerical solutions
from VADOFT, FECTUZ, HYDRUS, MOB1,
Ogata and Banks (1961), van Genuchten and
Alves (1982), Shamir and Harleman (1967),
Hadennann (1980), and Hodgkinson and Maul
(1985)
10
verified
Vadose Zone Module—metals
transport with nonlinear sorption
Compared to total mass input or analytical and
numerical solutions from HYDRUS and Ogata
(1970)
4
verified
Saturated Zone—flow and
transport
Compared to analytical and numerical solutions
from MNDXYZ, EPACMTP, MOB1, FECTUZ,
DSTRAM, VAM2D, VAM3D, and Sudicky et
al. (1991)
7
verified
Module- and Model-Level
Verification as EPACMTP
(1997) (U.S. EPA 1997b)
Vadose Zone—flow, infiltration,
and transport
Compared to analytical and numerical solutions
from STAFF3D, EPACMTP, and VAM2D
4
verified
Saturated Zone—flow and
transport
Compared to analytical and numerical solutions
from MNDXYZ, EPACMTP, VAM3DF, and
STAFF3D
3
verified
Composite Model—flow and
transport
Compared to analytical and numerical solutions
from VAM3DF
2
verified
(continued)
-------
Table 4-11. (continued)
Name of Verification
Activity (Year) (Original
Source Document)
Component Tested
Verification Tools
Number
of Test
Cases
Results
3MRA's Ground water
Pathway Module Verification
(1999) "(U.S. EPA 1999b;
U.S. EPA 1999c; U.S. EPA,
1999d)
Vadose and Saturated
Zones—extracted flow and
transport mechanisms
Compared to results from EPACMTP
12
verified
Vadose Zone—1-D flow and
transport
Compared to results from MODFLOW-
SURFACT
2
verified
Saturated Zone—pscudo 3-D flow
and transport
Compared to Darcy's Law, Ogata (1970), and
3MRA
3
verified
Final Verification of 3MRA
Ground water Pathway
Module (2000) (U.S. EPA,
2000c; U.S. EPA, 2000d)
Vadose Zone—all components
NA2
40
verified
Saturated Zone—all components
NA
69
verified
'Adapted from HGL (2003)
2Not available
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Further testing was conducted in 1993 and 1994 to verify the model's flow and transport
components in the vadose and saturated zones. Researchers compared the results of these tests
to analytical and numerical solutions from several simulators with comparable frameworks. Ten
cases verified the model's ability to calculate infiltration and transport mechanisms in the vadose
zone. Another four test cases were conducted to examine metals transport mechanisms with
non-linear sorption in the vadose zone. The cases verified the model's ability to calculate these
mechanisms. The flow, transport, and sorption of contaminants in the saturated zone were
verified in seven test cases.
Verification activities for EPACMTP continued in 1997. These activities focused on
single problem geometry and were developed in accordance with the ASTM "Standard Guide
for Developing and Evaluating Ground-Water Modeling Codes'' (ASTM, 1996). The first four
test cases verified the flow, infiltration, and contaminant transport mechanisms in the model's
vadose zone module. The 3-D flow and transport components of the saturated zone module were
verified in the next three test cases. Finally, two test cases were used to examine the composite
module. These cases verified the model's ability to calculate composite flow and transport
mechanisms in the subsurface.
In 1999, the flow and transport mechanisms in EPACMTP were extracted from the
vadose zone and saturated zone modules for use as the Vadose Zone and Aquifer Modules in the
3MRA modeling system. EPA conducted 12 case studies to ensure that the extracted
components functioned correctly in the new module. Two cases verified 1-D contaminant flow
and transport mechanisms in the vadose zone using the 3-D MODFLOW-SURFACT model.
The flow and transport mechanisms in the 3MRA modeling system's pseudo-3-D model for the
saturated zone were also verified in three test cases.
EPA conducted final verification of the vadose zone model and pseudo-3-D saturated
zone model contained in the Vadose Zone and Aquifer Modules in 2000. Tables 4-12 and 4-13
summarize the functional requirements and number of test cases executed to verify the Vadose
Zone and Aquifer Modules, respectively.
Table 4-12. General Requirements for Testing the Vadose Zone Module
Step
Description
Number of
Test Cases
1
Correctly read and screen source and site-specific input data
5
2
Correctly perform any required presimulation processing of input data
3
3
Correct operation of flow component
2
4
Correct operation of the nonmetals transport component
8
5
Correct operation of the metals transport component
10
6
Correct operation of postsimulation output
2
7
Robustness
24
4-32
-------
Section 4.0 Verification and Validation of the 3MRA Modeling System Modules
Table 4-13. General Requirements for Testing the Aquifer Module
Step
Description
Number of
Test Cases
1
Correctly read and screen source and site-specific input data
10
2
Correctly perform any required presimulation processing of input data
69
3
Correctly calculate the appropriate hydraulic conductivity multiplier for sites with
fracture media
9
4
Correctly calculate the appropriate concentration multipliers to account for effects
due to heterogeneous saturated media
2
5
Correctly read and screen contaminant-specific, biodegradation, and metal-specific
data
20
6
Correctly generate numerical grid for simulation
10
7
Correctly simulate ground water flow subject to various conditions
4
8
Correctly simulate contaminant fate and transport subject to various conditions
21
9
Robustness
29
4.5.4 Summary of Validation
Simulations using EPACMTP and its predecessors have been validated using four sites in
North America (see Table 4-14). Site-specific data were modeled and compared to observed
field data at the sites. These validation activities are summarized in " Verification and Validation
of the EPA 's Composite Model for Transformation Products (EPACMTP) and its Derivatives'"
(HGL, 2003). That document is included here as Appendix C. A synopsis of the information
from this report is provided here.
Table 4-14. Summary of EPACMTP Validation Activities
Site Description and Location
Validation Mechanism
Results
Borden Landfill, Borden, Ontario,
Canada
Compared to observed values and
simulation values from Frind and
Hakkanen (1987)
Accurately predicted plume size and
shape
Agricultural Site, Long Island, New
York
Compared to observed values at the
site
Demonstrated reasonable agreement;
relative error decreased as distance
from source increased
Triasulfuron and bromide spill,
Dodge City, Kansas
Compared to observed values
Demonstrated reasonable agreement;
conservative predictions slightly
underestimated; non-conservative
concentrations overestimated
Manufactured Gas Plant, New York
Compared to observed naphthalene
near to source of contamination
Demonstrated qualitatively similar
results
A predecessor to EPACMTP was compared to field and calculated data from the Borden
Landfill in 1990. A site contained a chloride plume in a glaciofluvial aquifer beneath the
landfill. Contaminants at the site are not transported through the vadose zone because the base
of the landfill is located just above the water table. The results of the simulation were compared
4-33
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
to observed field data and calculated data derived from the procedure described in Frind and
Hokkanen (1987). The model accurately predicted the size and shape of the chloride plume.
In 1990, pesticide concentrations in ground water were modeled using the saturated zone
and vadose zone models from the predecessor to EPACMTP. The results of the simulation were
compared to field data from a pesticide-contaminated potato field on Long Island, New York.
Site-specific data were combined with calculations from EPA's Pesticide Root Zone Model
(PRZM) to provide input for the simulation. Researchers found reasonable agreement between
the simulated and observed ground water concentrations. The relative error decreased as the
distance from the site increased.
EPACMTP was used in 1993 to model a controlled release of triasulfuron pesticide and
bromide. The triasulfuron simulation was conducted using a non-conservative flow and
transport model. Bromide was modeled conservatively. These models were compared to an
actual controlled release of these contaminants at a site in Dodge City, Kansas. Compared to
observed data, the simulated, non-conservative triasulfuron concentrations tended to be
overestimated. The model tended to underestimated the conservative bromide concentrations.
Overall the simulated data demonstrated reasonably good agreement with the observed data.
In 1995, EPACMTP was validated against observed field results from a coal tar
manufactured gas plant in New York. Coal tar had been disposed at the surface and migrated
into the aquifer beneath the site. Naphthalene was the contaminant of concern at the site. The
results from the model were qualitatively similar to the observed ground water concentrations
near the source.
The Vadose Zone Module has also been validated using data from a large soil-column
study conducted to investigate the behavior of organic chemicals (including halogenated
aliphatic hydrocarbons, substituted benzenes, and phenols) during infiltration of municipal
wastewater into soil columns by the Ground Water and Ecosystem Restoration Division
(GWERD) of EPA's National Risk Management Research Laboratory (NRMRL). The
experiment design and conditions were very close to the assumptions and application conditions
of the Vadose Zone Module, making these data suitable for use in evaluating the Vadose Zone
Module.
The evaluation of the Vadose Zone Module included:
¦ Whether modeling without accounting for gas phase could acceptably describe
the contaminant fate and transport processes in vadose zone;
¦ Whether the assumption of first order decay of contaminant transformation is
valid;
¦ Whether the conceptualizations on different attributes of the model (such as
boundary conditions and the mathematical model to describe the flow, fate, and
transport, as well as its solutions) are adequate; and
4-34
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
¦ Whether suggestions on how to appropriately apply the Vadose Zone Module
could be made through the evaluation.
An individual standalone Vadose Zone program was obtained by modifying the Vadose
Zone Module from the 3MRA modeling system to achieve the small intervals of time and space
as well as outputs for any depths along the soil column needed to compare the results to the
experimental lab data. This modified Vadose Zone program was compared to the Vadose Zone
Module and verified to be consistent with the original module. The necessary input parameters
to the Vadose Zone program were obtained from the lab experiment design, literature review,
and parameter calibrations. Concentrations of organic compounds of interest at selected
sampling ports were compared with the model simulation results. The Vadose Zone Module was
examined directly using the parameters from the lab experiment design and literature review. It
was further evaluated by taking the calibrated first-order transformation rates obtained based on
the concentrations from part of sampling ports. Then model verifications were carried out using
the concentration data from the other sampling ports.
Comparisons were also performed between the simulation results from the Vadose Zone
Module and the other tested and accepted models with similar or enhanced functions (such as
CHEMFLO) in order to investigate the compatibility of the Vadose Zone Module with those
models. In addition, the simulation outputs of leachate concentrations from the LAU program
described in section 4.2.4 were used as input to the Vadose Zone Module, the results of which
were used in comparison with the lab experiment data to assess the overall errors from
applications of both LAU and Vadose Zone Modules. Moreover, the evaluation was performed
based on the chemical categories and the volatility of organic compounds.
Generally, the Vadose Zone Module functioned quite well in simulating the fate and
transport of organic chemicals in vadose zones, although noticeable differences between the
simulated and observed results could be observed for highly volatile organics. The Vadose Zone
Module and the CHEMFLO model generate comparable results. In addition, the overall final
outputs from both the LAU and Vadose Zone Modules gave a good estimate of the leachate
concentration for organics undergoing both volatilization and transformation, but slightly
overestimated for organics with high volatility and low transformation rates.
4.6 Farm Food Chain Module
This section describes the verification
and validation activities for the Farm Food
Chain Module. The external reviewers listed
in the box reviewed the Farm Food Chain
Module.
4.6.1 Module Description
The Farm Food Chain Module predicts the accumulation of contaminants in the edible
parts of plants through the uptake of contaminants from soil and the deposition of vapor-phase
and particle-bound contaminants from the air. Concentrations are predicted for fruits and
vegetables that are grown above ground, as well as for root vegetables. In addition, the module
predicts the annual average contaminant concentration in beef and milk products from cattle
Peer Reviewers for Farm Food Chain Module
¦ Dr. Donald Mackay of Trent University,
Ontario
¦ Dr. Lee Shull of Montgomery Watson Harza
¦ Dr. Curtis Travis of Quest Technologies
4-35
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
raised on farms. The concentrations in produce, beef, and milk are used as inputs to the Human
Exposure Module to calculate the applied daily dose to human receptors that consume fruits and
vegetables from home gardens or consume produce, beef, or milk from on a farm. Figure 4-10
shows the information flow for the Farm Food Chain Module.
Key Data Inputs
• Farm area
• Plant uptake factor
• Beef biotransfer factor
Air Concentrations
Deposition Rates
Soil Concentrations
Food Item and
Soil Concentrations
Soil Concentrations
Subsurface
Concentrations
Water Column
Concentrations
Human
Exposure
Module
Vadose Zone
and Aquifer
Modules
Watershed
Module
Land-based
Source
Modules
Surface
Water
Module
Air Module
Farm
Food
Chain
Module
Figure 4-10. Information flow for the Farm Chain Module in the 3MRA modeling system.
The modeling construct for the Farm Food Chain Module is based on recent and ongoing
research conducted by EPA ORD and presented in Methodology for Assessing Health Risks
Associated with Multiple Pathways of Exposure to Combustor Emissions (U.S. EPA, 1998b).
4.6.2 Major Module Components/Functionality
The Farm Food Chain Module performs the following four functions:
1. Calculate contaminant concentrations in plants due to contaminants in air.
The Farm Food Chain Module calculates the contaminant concentration in plants
due to the deposition of particle-bound and vapor-phase contaminant
contaminants onto fruits, vegetables, and feed crops that grow above the ground.
4-36
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
2. Calculate contaminant concentrations in plants due to contaminants in soil.
The Farm Food Chain Module calculates the contaminant concentration in plants
due to uptake and translocation of contaminants from the soil into the edible parts
of fruits, vegetables, and feed crops that grow above the ground.
3. Calculate total contaminant concentrations in plants. The Farm Food Chain
Module sums plant concentrations across relevant mechanisms, including direct
deposition of particle-bound contaminants, vapor-phase uptake, and translocation
of contaminants from soil into the edible parts of plants.
4. Calculate contaminant concentrations in beef and milk. The Farm Food
Chain Module calculates exposures to beef and dairy cattle through ingestion of
contaminated forage, feed crops, soil, and drinking water, and the resulting beef
and milk concentrations.
For each year in the simulation, the module predicts point estimates and spatially
averaged concentrations within the AOI. Point estimate concentrations are used to evaluate
exposures to residential home gardeners that grow and eat fruits and/or vegetables within the
AOI. The point estimates reflect the locations of residential receptors that are used to represent
the populations in various Census tracts throughout the AOI. The spatially averaged
concentrations are used to evaluate exposures to farmers that raise and eat their own produce,
beef, or milk products. The spatial averages reflect the farm boundaries that are delineated in
the site layout that defines all of the characteristics of the AOI.
4.6.3 Summary of Verification
The Farm Food Chain Module was verified through a series of tests, which are described
in detail in the following documents:
¦ Farm Food Chain (FFC) Module—Test Plan (RTI, 2002k)
¦ Farm Food Chain Module Internal Verification Testing (RTI, 2002m)
¦ HWIR Farm Food Chain—Review of Compiled Code Executables and Review of
the Implementation of the Test Plan (Snyder, 2000).
EPA originally completed internal verification testing in 2000, followed by external
verification testing. Since then, EPA has made a few minor changes to the module. Therefore,
EPA repeated internal verification testing on the updated version of the module in 2002 to ensure
that the updated version passed all tests and that the test files reflected the outputs from the
updated version. EPA also updated internal verification test plans and testing documentation to
reflect this second round of testing.
To verify that the module was working properly, EPA designed an Excel spreadsheet
using the equations from the module documentation. The testing included a comparison of the
output concentrations (and intermediate concentrations) calculated by the Farm Food Chain
Module to those calculated using the Excel spreadsheet. The comparison verified that the
4-37
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
concentrations in air, soil, fruits, vegetables, feed, beef, and milk calculated by the module were
correct. Table 4-15 summarizes the functional requirements and number of test cases executed
to verify the Farm Food Chain Module.
Table 4-15. General Requirements for Testing the Farm Food Chain Module
Step
Description
Number of
Test Cases
1
Correctly read results files from the Air, Watershed, Surface Water, and Source
modules
3
2
Correctly identify the constituent type as dioxin-like, organic, metal, mercury, or
special
3
3
Correctly read site layout file
3
4
Correctly calculate the area-averaged soil concentrations at root zone depth and in
the surficial soil for the farm
3
5
Correctly calculate the contaminant concentrations in produce for each constituent
type
1
6
Correctly calculate the contaminant concentrations in beef and milk for each
constituent type
1
7
Correctly calculate intermediate concentrations, including area-averaged air
concentrations and deposition rates onto plants, and concentrations in feed crops for
cattle
3
8
Correctly calculate output concentrations when input data include multiple time
series that do not start or end in the same year or are discontinuous
1
All total produce, beef, and milk concentrations and area-averaged soil concentrations
matched for years 1, 5, 10, 15, and 20, demonstrating that the module handles discontinuous time
series correctly. All area-averaged soil concentrations; concentrations in produce, beef, and
milk; and intermediate concentrations calculated by the Farm Food Chain Module were verified
by the spreadsheet calculations.
The independent testing of the Farm Food Chain Module confirmed the internal testing
results: all tests were successfully completed with the executable provided by the development
team. All tests were also successfully completed with an independently generated executable.
4.6.4 Summary of Validation
The Farm Food Chain methodology is based primarily on EPA's MPE methodology
(U.S. EPA, 1998b). The MPE methodology represents the state of the science with respect to
providing reliable guidance in the proper conduct of assessments of risks that may result from
multimedia and multipathway exposure. The National Center for Environmental Assessment
(NCEA) prepared the MPE methodology as an update to EPA's 1990 indirect exposure
document (U.S. EPA, 1990), which is generally known as the Indirect Exposure Methodology,
or IEM. Most of the revisions in the MPE methodology are based on SAB and public comments
4-38
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
on IEM. Earlier versions of this document have undergone internal Agency and external peer
review. However, this methodology has not been formally validated using field data.
4.7 Terrestrial Food Web Module
This section documents the verification
and validation activities that were performed
for the Terrestrial Food Web Module. The
reviewers listed in the box reviewed the
Terrestrial Food Web Module.
4.7.1 Module Description
The Terrestrial Food Web Module calculates the annual average contaminant
concentrations in terrestrial plants and prey (such as earthworms or small mammals) that are
eaten by wildlife. In addition, the module calculates spatially averaged soil concentrations in
two soil horizons—surficial soil and root zone soil—for the receptor home ranges placed within
each of the habitats delineated within the AOI. These concentrations are used as input to the
Ecological Exposure Module in calculating the applied does to receptors of interest, and the root
zone soil concentration is used by the Ecological Risk Module to predict risks to terrestrial plants
and soil communities. Figure 4-11 shows the information flow for the Terrestrial Food Web
Module.
For each home range delineated within the AOI, the Terrestrial Food Web Module
predicts a time series of annual average concentrations of contaminants in soil, along with
concentrations in specific plant and prey categories. The Terrestrial Food Web Module uses the
same algorithms and contaminant-specific data as the Farm Food Chain Module to calculate
concentrations in plants.
Key Data Inputs
• Root zone depth
• Home range area
• Bioconcentration factor
Air Concentrations
Deposition Rates
Soil
Concentrations
Soil Concentrations
Soil, Plant, and Prey
Concentrations
Soil Concentrations
Ecological
Risk Module
Ecological
Exposure
Module
Terrestrial
Food Web
Module
Land-based
Source
Modules
Air Module
Watershed
Module
Figure 4-11. Information flow for the Terrestrial Food Web Module in the 3MRA
modeling system.
Peer Reviewers for Terrestrial Food Web Module
¦ Dr. Anne Fairbrother of Parametrix, Inc.
¦ Dr. Robert Pastorok of Exponent
¦ Dr. Bradley Sample of CH2M Hill
4-39
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.7.2 Major Module Components/Functionality
The Terrestrial Food Web Module uses both predictive models and empirical data to
calculate contaminant concentration in terrestrial food items. Specifically, the Terrestrial Food
Web Module performs the following four functions:
1. Calculate contaminant concentrations in soil. The module calculates spatially
averaged soil concentrations for each home range in each habitat. The soil
concentrations reported by the Land-based Source Modules and the Watershed
Module are defined by the WMU, and the watershed subbasins, respectively, at
each site. The Terrestrial Food Web Module determines the spatial average based
on the proportion of subbasins and/or WMU that overlaps the home range of
wildlife species assigned to a given habitat.
2. Calculate total contaminant concentrations in plants. The Terrestrial Food
Web Module sums plant concentrations across relevant mechanisms, including
direct deposition of particle-bound contaminants, vapor-phase uptake, and
translocation of contaminants from soil into the edible parts of plants. The
module uses the same approach as described for the Farm Food Chain Module to
calculate total contaminant concentrations in plants. The only significant
difference in the respective modules is that the Terrestrial Food Web Modules
defines plant categories for consumption by wildlife rather than humans.
3. Calculate contaminant concentrations in soil invertebrates. Earthworms and
other soil invertebrates constitute a significant dietary component for many
wildlife species. Contaminant-specific soil-to-biota bioconcentration factors are
used to calculate the contaminant concentrations in soil invertebrates based on the
spatially averaged, root zone soil concentration for each wildlife species' home
range.
4. Calculate contaminant concentrations in vertebrate prey categories. Small to
medium sized terrestrial vertebrates are eaten by larger predators (e.g., fox, black
bear, and red tailed hawk). Contaminant-specific soil-to-vertebrate
bioconcentration factors are used to estimate the tissue concentrations of
contaminants in vertebrate prey categories based on the spatially averaged, root
zone soil concentration for each wildlife species' home range.1 The module
reports the minimum and maximum tissue concentrations in each vertebrate prey
category; these outputs allow the Ecological Exposure Module to sample from the
spatial variability of possible prey species.
The Terrestrial Food Web Module is applied to the margin habitats in freshwater systems
and wetlands and to terrestrial habitats, such as the representative forest or grassland. Simple
terrestrial food webs were constructed for each terrestrial habitat to depict the major functional
1 The root zone, rather than surficial, soil concentrations were used to be more representative of the types of
exposures likely to be reflected in soil-to-organism bioaccumulation factors (e.g., direct ingestion of soil
invertebrates from deeper soil horizons; ingestion of prey that feed primarily on plant matter).
4-40
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
and structural components of healthy terrestrial ecosystems. The components represent major
dietary categories for terrestrial wildlife, and represent a broad range of sizes, feeding guilds, and
taxa typical of terrestrial food webs. In addition, the terrestrial prey categories reflect available
methods and data required to estimate prey concentrations. For example, the category of "small
mammals" was developed specifically to take advantage of recent research (Sample et al., 1998)
on estimating tissue concentrations of contaminants in small mammals as a function of the soil
concentration. This research provides a significant improvement over previous approaches that
aggregated a variety of prey items into a single category of "terrestrial invertebrates."
4.7.3 Summary of Verification
The Terrestrial Food Web Module was verified through a series of verification tests,
which are described in detail in the following documents:
¦ Terrestrial Food Web (TFW) Module—Test Plan (RTI, 2002p)
¦ Terrestrial Food Web Internal Verification Testing (RTI, 2002n)
¦ Independent Tests for Terrestrial Food Web Module (Tetra Tech, 2000e).
EPA originally completed internal verification testing in 2000, followed by external
verification testing. Since then, EPA has made a few minor changes to the module. Therefore,
EPA repeated internal verification testing on the updated version of the module in 2002 to ensure
that the updated version passed all tests and that the test files reflected the outputs from the
updated version. EPA also updated the internal verification test plans and testing documentation
to reflect this second round of testing.
Table 4-16 summarizes the functional requirements and number of test cases executed to
verify the Terrestrial Food Web Module.
Table 4-16. General Requirements for Testing the Terrestrial Food Web Module
Step
Description
Number of
Test Cases
1
Correctly read output files from the Air, Watershed, and Source modules
3
2
Correctly identify the constituent type as dioxin-like, organic, metal, mercury, or special
3
3
Correctly read site layout file
3
4
Correctly calculate depth averaged and surficial soil concentrations for each home range
in both aquatic (e.g., stream margin) and terrestrial habitats
3
5
Correctly calculate the contaminant concentrations in plants for each constituent type
1
6
Correctly report the minimum and maximum tissue concentrations for each prey category
3
7
Correctly loop over home ranges (by receptor)
3
8
Correctly loop over habitats
3
9
Correctly loop over years and ensure correct time series management
2
The testing confirmed that the module correctly calculates concentrations in terrestrial
biota for all years with nonzero concentrations during the simulation period. Biota
concentrations were calculated in a module verification spreadsheet and compared with those
4-41
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
from module runs to confirm that the equations are coded properly and that the module is fully
functional.
4.7.4 Summary of Validation
Although the Terrestrial Food Web Module has not been directly validated, it is based on
accepted science. The Terrestrial Food Web methodology is based primarily on EPA's MPE
methodology (U.S. EPA, 1998b). The MPE methodology represents the state of the science with
respect to providing reliable guidance in the proper conduct of assessments of risks that may
result from multimedia and multipathway exposure. The NCEA prepared the MPE methodology
as an update to EPA's 1990 Indirect Exposure document (U.S. EPA, 1990), which is generally
known as the Indirect Exposure Methodology, or IEM. Most of the revisions in the MPE
methodology are based on SAB and public comments on IEM. Earlier versions of this document
have undergone internal Agency and external peer review.
For contaminant uptake in soil invertebrates, the Terrestrial Food Web Module uses
uptake factors developed by the Oak Ridge National Lab (ORNL) (Sample et al., 1998). ORNL
validated these uptake factors to evaluate their predictive utility. Validation consisted of
separating the available data into two groups, a "model" data set and a "validation" data set.
ORNL applied the uptake factors developed from the model data set to the soil concentration
data in the validation data set, and compared the resulting estimated concentrations in
earthworms with the measured earthworm concentrations in the validation data set. The
difference between the estimated values and the measured values was expressed as a proportion
deviation as follows:
proportion deviation = (measured - estimated)/measured
The proportion deviations for the median uptake factors used in the Terrestrial Food Web
Module ranged from -6.35 for mercury to 0.76 for chromium. A negative proportion deviation
indicates an overestimation; a positive one indicates an underestimation.
4.8 Aquatic Food Web Module
This section describes verification and
validation activities for the Aquatic Food Web
Module. The external reviewers listed in the
box reviewed the Aquatic Food Web Module.
4.8.1 Module Description
The Aquatic Food Web Module calculates steady-state contaminant concentrations in
aquatic organisms (e.g., fish, benthic invertebrates, aquatic plants) that are consumed by human
and ecological receptors. These concentrations are used as input to the Human Exposure and
Ecological Exposure Modules in calculating the applied dose to receptors of interest.
Figure 4-12 shows the information flow for the Aquatic Food Web Module.
Peer Reviewers for Aquatic Food Web Module
¦ Dr. Lawrence Barnthouse of LWB
Environmental Service, Inc.
¦ Dr. Frank Gobas of Simon Fraser University
¦ Dr. Paul Jacobson of Langhei Ecology, LLC
4-42
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Key Data Inputs
Waterbody type
Food web structure
K ,
Surface
Water
Module
Water Column and
Sediment Concentrations
Aquatic
Food Web
Module
Fish
Concentrations
Aquatic Plant and
Human
Exposure
Module
Prey Concentrations
Ecological
Exposure
Module
Figure 4-12. Information flow for the Aquatic Food Web Module
in the 3MRA modeling system.
For each year in the simulation, the Aquatic Food Web Module predicts contaminant
concentrations in aquatic biota for freshwater waterbodies in the AOI considered capable of
supporting fish (referred to as "fishable" waterbodies). The model is flexible enough to be
applied to different types of waterbodies, including stream reaches, rivers, lakes, ponds, and
permanently flooded wetlands. Simple freshwater food webs were constructed for each type of
waterbody to depict the major functional and structural components of "healthy" freshwater
ecosystems. The components of each food web represent major categories of aquatic biota in
freshwater systems: aquatic macrophytes, phytoplankton, periphyton, zooplankton, benthic
detritivores, benthic filter feeders, and various feeding guilds of fish. Some of these
concentrations are used internally to calculate concentrations in fish, while other concentrations
are reported as a time series for use in calculating exposures to wildlife and humans, as well as to
calculate ecological hazard (e.g., hazard to sediment dwellers). Thus, the Aquatic Food Web
Module determines which data are appropriate for use in a given waterbody and calculates
concentrations in the aquatic biota assigned to that waterbody.
4.8.2 Major Module Components/Functionality
The Aquatic Food Web Module performs the following functions:
1. Select food web appropriate for each waterbody. The Aquatic Food Web
Module matches an appropriate food web with each waterbody identified as
fishable within the AOI. Eight freshwater food webs were developed to capture
the variability in freshwater systems. They represent warmwater streams/rivers,
wetlands, ponds, and lakes; and coldwater stream s/ri vers, wetlands, ponds, and
lakes.
2. Construct dietary matrix for food web. The Aquatic Food Web Module uses a
constrained, random prey preference sampling approach that selects preference
4-43
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
fractions at random between the minimum and maximum, assuming a uniform
distribution. This approach allows for the dietary composition to reflect the full
range of variability inherent in the diets of freshwater fish.
3. Calculate contaminant concentrations in food web. The Aquatic Food Web
Module calculates concentrations for the biota assigned to each freshwater food
web. The model will only calculate concentrations for biota that are assigned to a
particular food web and waterbody. The calculations involve mechanistic models
or the use of empirical data on bioaccumulation.
4. Report contaminant concentrations for fish consumed by wildlife and
humans. Food webs for freshwater aquatic systems typically have a single apex
predator species (trophic level 4 [TL4]) and a number of other fish species that
occupy different feeding guilds, such as benthic feeders (e.g., catfish). These
other species are, for the purposes of exposure assessment, often grouped into the
category of trophic level 3 (TL3), indicating that they are both predator and prey
in the food web. To predict exposures for wildlife and humans that eat TL3 fish,
the model calculates an average concentration—both wholebody and filet—for
fish that fall into the category of TL3. In addition, the Aquatic Food Web Module
predicts the tissue concentration for the apex predator fish in each waterbody.
The wholebody fish concentrations are used by the Ecological Exposure Module,
and the filet concentrations are used by the Human Exposure Module.
4.8.3 Summary of Verification
The Aquatic Food Web Module was verified through a series of verification tests that are
described in detail in the following documents:
¦ Aquatic Food Web (AFW) Module—Test Plan (RTI, 2002e)
¦ Aquatic Food Web Internal Verification Testing (RTI, 2002f)
¦ Independent Tests for Aquatic Food Web Module (Tetra Tech, 2000a).
EPA originally completed internal verification testing in 2000, followed by external
verification testing. Since then, EPA has made a few minor changes to the module. Therefore,
EPA repeated the internal verification testing on the updated version of the module in 2002 to
ensure that the updated version passed all tests and that the test files reflected the outputs from
the updated version. EPA also updated internal verification test plans and testing documentation
to reflect this second round of testing.
Table 4-17 summarizes the functional requirements and number of test cases executed to
verify the Aquatic Food Web Module.
4-44
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Table 4-17. General Requirements for Testing the Aquatic Food Web Module
Step
Description
Number of
Test Cases
1
Correctly read output files from the Surface Water Model
4
2
Correctly identify the constituent type as dioxin-like, organic, metal, mercury, or
special
4
3
Correctly read site layout file
1
4
Correctly assign the food web (i.e., biota) types for each waterbody
1
5
Correctly assign prey preferences to all elements in the aquatic food web
1
6
Correctly calculate the contaminant concentrations in each element in the food web
for each constituent type
4
7
Correctly average concentrations in T3 fish
4
8
Correctly loop over aquatic habitats
2
9
Correctly loop over all fishable reaches within each aquatic habitat
2
10
Correctly loop over years and ensure correct time series management
3
11
Correctly read the chemical properties and surface water data associated with
methylmercury when multiple chemical species of mercury are present
1
The testing confirmed that the module correctly calculates concentrations in aquatic biota
for all years with nonzero concentrations during the simulation period. A comparison of the
output biota concentrations in the spreadsheet and those from the module runs ensured that the
equations were coded properly and that the module was fully functional.
4.8.4 Summary of Validation
The Aquatic Food Web Module methodology for calculating tissue concentrations for
organic chemicals for fish and other aquatic biota is based on methods identified in refereed
journals. In addition, the Gobas model (Gobas, 1993) and other major elements of the
calculation framework are currently used by EPA in various guidance documents and rule
makings. The methodology for hydrophobic organics (based on the work by Gobas [1993]) has
been validated for coldwater lakes and has been validated for other aquatic systems as well.
Nevertheless, validation exercises comparing output from the Aquatic Food Web Module with
measured concentrations from field studies have not been performed.
4.9 Human Exposure Module
This section describes the verification and validation activities for the Human Exposure
Module. The Human Exposure Module and associated data (e.g., spatial layout, human exposure
factors) were reviewed by the external reviewers shown in the box. The peer-review charge, a
summary of the reviewer comments, and original comments submitted by each individual
4-45
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
reviewer were included in the peer review of
the Background Document for the Human
Exposure and Human Risk Modules for the
3MRAModel (U.S. EPA, 2000e).
4.9.1 Module Description
The Human Exposure Module
calculates the applied dose to human
receptors from ingestion and inhalation of
contaminated media and food. Figure 4-13 shows the information flow for the Human Exposure
Module. The Human Exposure Module uses media and food concentrations calculated in the
media and food web modules to calculate applied doses. These doses are used by the Human
Risk Module to calculate risk measures.
ZKey Data Inputs
• Food intake rates
• Inhalation rate
• Body weight
Fish Concentrations
Air Concentrations
Soil Concentrations
Doses
Ground Water
Concentrations
Soil Concentrations
Food Item and
Soil Concentrations
Human Risk
Module
Human
Exposure
Module
Aquatic
Food Web
Module
Air
Module
Watershed
Module
Farm Food
Chain
Module
Land-based
Source
Modules
Vadose Zone
and Aquifer
Modules
Figure 4-13. Information flow for the Human Exposure Module
in the 3MRA modeling system.
Peer Reviewers for Human Exposure Module
¦ Dr. James Butler of Argonne National
Laboratory
¦ Dr. William Kastenberg of the University of
California at Berkeley
¦ Mr. Stephen Washburn of ENVIRON
International Corporation
4-46
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.9.2 Major Module Components/Functionality
The purpose of the Human Exposure Module is to calculate the inputs needed by the
Human Risk Module to calculate risk and hazard. For carcinogens, risk is calculated from
applied dose (in mg/kg-d). For noncarcinogens, hazard via the ingestion pathway is also based
on dose, whereas hazard via the inhalation pathway is based on the air concentration to which a
receptor is exposed. Thus, the Human Exposure Module has the following functions:
1. Calculate ambient air concentrations. The Human Exposure Module calculates
areal average ambient air concentrations for farms. Ambient (outdoor) air
concentrations for residents are calculated by the Air Module and used by the
Human Exposure Module.
2. Calculate shower air concentration. The Human Exposure Module calculates
shower air concentration from ground water concentration. Ambient air
concentrations are calculated by the Air Module.
3. Calculate dose from inhalation of carcinogens. The Human Exposure Module
calculates dose from inhalation of ambient (outdoor) air and shower air for
carcinogens. Noncarcinogenic risk for inhalation exposures is based directly on
air concentration, not dose; thus, the Human Exposure Module does not calculate
inhalation dose for noncarcinogens.
4. Calculate dose from ingestion of contaminants in media or food. The Human
Exposure Module calculates dose for carcinogens and noncarcinogens from
ingestion of soil, ground water, produce (fruits and vegetables), beef, milk, and
fish.
5. Calculate dose from ingestion of contaminants in breast milk. The Human
Exposure Module calculates dose to infants from ingestion of contaminated breast
milk. This route of exposure is assessed only for dioxin-like chemicals and only
for infants.
Doses and air concentrations are calculated for each contaminant and site, as well as for
each
¦ Receptor type. The module calculates exposures for two types of human
receptors: residents and farmers. These are distinguished by how they are
located—residents at a single exposure point and farmers on an area representing
a farm. Residents are further defined by various kinds of behavior that lead to
different profiles of exposure, such as home gardening and recreational fishing.
Farmers may also be defined by behavior such as recreational fishing. Human
receptor types considered in the 3MRA modeling system are a function of the
goals of a particular analysis and are defined by the receptor data used for the
analysis.
4-47
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
Age cohort. The human receptors are divided into five age cohorts (infants under
1 year, children aged 1 to 5 years, children aged 6 to 11 years, children aged 12 to
19 years, and adults) that are used to determine the most appropriate exposure
factor data for various pathways, such as body weights, inhalation rates, and
consumption rates for various food products.
Exposure pathway. The Human Exposure Module considers nine exposure
pathways. Depending on the exposure inputs entered into the model, all pathways
could be considered for all receptor types. Table 4-18 shows the pathways
modeled and indicates which receptor types are modeled for each pathway by
default. All pathways are modeled by default for all age cohorts with two
exceptions: the breast milk pathway is modeled only for infants (and no other
pathways are modeled for infants), and shower inhalation is modeled only for
children aged 12 to 19 years and adults.
Table 4-18. Default Pathways Considered by Receptor Type
Pathway
Receptor Type
Resident
Resident
Fisher
Home
Gardener
Home
Gardener
Fisher
Beef
Farmer
Beef
Farmer
Fisher
Dairy
Farmer
Dairy
Farmer
Fisher
Air inhalation
/
/
/
/
/
/
/
/
Shower inhalation
/a
/a
/a
/a
/
/
/
/
Ground water ingestion
/a
/a
/a
/a
/
/
/
/
Soil ingestion
/
/
/
/
/
/
/
/
Produce ingestion
/
/
/
/
/
/
Beef ingestion
/
/
Milk ingestion
/
/
Fish ingestion
/
/
/
/
Breast milk ingestion
/
/
/
/
/
/
/
/
a Ground water and shower pathways are considered for residents and home gardeners only if Census data indicate
the presence of private wells in the Census block group. All farms are assumed to have a private well.
Location. The Human Exposure Module calculates exposures for residents for a
single point, placed in the current data set at the centroid of each Census block in
the area of interest (AOI). The module calculates exposure for farmers at a single
farm in each census block group in the AOI that has Census data for farmers and
agricultural land areas.
Year. The module calculates exposures for each year of the simulation. The
exposures predicted by the Human Exposure Module are reported as a time series
of annual average applied doses or air concentrations. Thus, the equations
presented in this section are applied to input data for each year. All temporal
averaging for exposure durations exceeding 1 year is done by the Human Risk
Module.
4-48
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.9.3 Summary of Verification
The Human Exposure Module was verified through a series of verification steps that are
described in detail in the following documents:
¦ Test Plan for the HWIR Human Exposure Module (RTI, 2000e).
¦ HWIR Human Exposure Module Internal Verification Testing (U.S. EPA, 2000a).
The scope of the testing was to verify all major requirements of the Human Exposure
Module. The internal verification included hand-calculation verification of internal calculations,
as well as visual examination of the GRF to verify that certain switches work as designed.
Table 4-19 summarizes the functional requirements and number of test cases executed to verify
the Human Exposure Module.
Table 4-19. General Requirements for Testing the Human Exposure Module
Step
Description
Number of
Test Cases
1
Correctly recognize the presence of farms and residential receptor areas and
correctly calculate inhalation (non-shower) and ingestion (food, soil, water) doses
5
2
Correctly recognize and differentiate contaminants that are only carcinogenic,
contaminants that are only noncarcinogenic, and contaminants that are both
carcinogenic and noncarcinogenic, output the appropriate variables, and correctly
calculate inhalation (nonshower) and ingestion (food, soil, water) doses
3
3
Correctly calculate and assign exposures per receptor, cohort, and exposure area
5
4
Correctly process continuous and discontinuous time series of varying length per
pathway
1
5
Correctly calculate shower concentrations and inhalation doses by means of a
dynamic solution to a system of differential equations
1
6
Correctly assign shower concentrations and inhalation doses only to adults and child
4 cohorts
1
7
Correctly assign and compute breastmilk ingestion doses only to Child 1 (infant)
cohorts
1
8
Correctly perform random selection of 3 fishable waterbody reaches for fanner
fishers and resident fishers and calculate doses from fish ingestion
2
9
Correctly recognize and respond to error/warning traps
3
A series of test cases was developed to test these requirements. The Human Exposure
Module was tested using a variety of contaminants and site layouts (e.g., sites with and without
fishable reaches, sites without any farms, sites without residential receptors). A variety of
contaminants (e.g., pentachlorophenol, acetonitrile, toluene, 2,3,7,8-TCDD, mercury,
chloroform, carbon disulfide, and benzene) were tested, because the decision to perform certain
calculations depends on whether the contaminant is a carcinogen or a noncarcinogen or has an
4-49
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
inhalation and/or oral health benchmark available; breastmilk ingestion doses are calculated for
infants only for 2,3,7,8-TCDD.
The verification testing of the Human Exposure Module was based on the doses and
concentrations reported to the output file. All food (total vegetable and fruit, beef, milk, fish),
breastmilk, ground water, and soil ingestion doses and ambient air (non-shower) and shower
inhalation concentrations and doses calculated for each appropriate receptor/cohort by the
Human Exposure Module matched the independently calculated values in Excel spreadsheets.
Existing error traps operated as expected.
All decision points, functions, and contaminant-specific switches required of the Human
Exposure Module were adequately tested to confer a high degree of confidence in the module's
ability to correctly calculate ingestion and inhalation doses. The combination of basic
mathematics and the selection of a diverse group of contaminants and site layouts provides a
sufficient basis to conclude that the verification was successful.
4.9.4 Summary of Validation
The Human Exposure Module has not been validated because no data set exists to do so.
However, the module is based on widely accepted state-of-the-science formulations.
4.10 Human Risk Module
This section describes the verification
and validation activities for the Human Risk
Module. The Human Risk Module was
reviewed by the external reviewers shown in
the box.
4.10.1 Module Description
The Human Risk Module calculates risk measures for a given contaminant, waste
concentration, and site. Figure 4-14 shows the information flow for the Human Risk Module.
Key Data Inputs
• Health benchmarks
• Population demographics
• Receptor locations
Exit
Level
Processors I and II
Figure 4-14. Information flow for the Human Risk Module in the 3MRA
modeling system.
Peer Reviewers for Human Risk Module
¦
Dr. James Butler of Argonne National
Laboratory
¦
Dr. William Kastenberg of the University of
California at Berkeley
¦
Mr. Stephen Washburn of ENVIRON
International Corporation
7
Human
Exposure
Module
Doses
Human
Risk
Risks ^
Module
HQs
4-50
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
The Human Risk Module uses the annual average daily doses calculated in the Human Exposure
Module to calculate risk statistics. These risk statistics are used by the Exit Level Processors to
determine national-level risk distributions.
4.10.2 Major Module Components/Functionality
For each contaminant, waste concentration, and site, the Human Risk Module generates
risk estimates for each receptor location in the AOI and then calculates the number of receptors
that fall within a specified risk or hazard range to describe the distribution of risks for the
population at each site. The module also determines the timing of maximum risks. The Human
Risk Module has the following functions:
1. Calculate risk measures. The Human Risk Module calculates cancer risk,
noncancer hazard quotient (HQ), and noncancer margin of exposure (MOE) (for
breastfeeding infants only). Depending on the contaminant, the Human Risk
Module may calculate risk, HQ, or both. MOE is only calculated for
breastfeeding infants for dioxin-like chemicals. These risk measures are specific
to a receptor type, an age cohort, an exposure pathway, a receptor location, and a
specific exposure period (identified by starting year). The Human Risk Module
also aggregates risks and HQs from individual exposure pathways (e.g., ground
water ingestion) to determine risk for groups of pathways (e.g., all ingestion
pathways).
2. Process results for decision making. The Human Risk Module puts exposed
and unexposed population into risk bins to estimate the number of receptors that
experience risk within a specified range. Each risk bin is a range of risks or HQs.
For any given pathway and exposure period, the Human Risk Module uses
Census data on population for each receptor location to determine the number of
people of each receptor type and age cohort that experience risk from the
specified pathway in the specified exposure period at risk levels within that bin.
Population is summed across receptor locations that have risks within the same
bin. For each pathway or pathway group, the Human Risk Module estimates total
risk by multiplying the population at a location by the risk for that location and
uses this to determine the exposure period for which the total risk or HQ across
all receptor types and age cohorts is the greatest. This estimate of total risk is not
intended as a final risk measure but is used to identify the timing of maximum
risk. The exposure period is identified by the year in which the risk averaged
over a specified exposure period starts.
4.10.3 Summary of Verification
The Human Risk Module was verified through a series of verification steps that are
described in detail in the following document:
¦ Draft Test Plan for the HWIR Human Risk Module (U.S. EPA, 2000e)
4-51
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
The scope of the testing was to verify all major requirements of the Human Risk Module.
Table 4-20 summarizes the functional requirements and number of test cases executed to verify
the Human Risk Module.
Table 4-20. General Requirements for Testing the Human Risk Module
Step
Description
Number of
Test Cases
1
Correctly calculate risk, hazard quotient (HQ,) and MOE for appropriate receptor, cohort,
pathway, and distance ring
1
2
Correctly calculate total population for any receptor, cohort, distance, pathway combination
1
3
Correctly find critical year for risk/HQ/MOE for any receptor, cohort, distance, pathway
combination
1
4
Correctly detect whether the contaminant is carcinogenic, noncarcinogenic, or both; whether
the carcinogenic and noncarcinogenic effects are additive; and whether a human risk
assessment is to be done
1
5
Correctly develop risk/HQ/MOE histograms for aggregate pathways
1
6
Correctly respond to NumFarm or NumHumRcp
1
7
Correctly respond to ChemBreastMilkExp switch (MOE)
1
8
Correctly respond to ExDur
1
9
Correctly respond to RegPercentile criterion
1
10
Correctly age child cohorts during exposure periods so that cohort-specific, varying
exposures are considered
1
11
Report an error message if a farm index or a human receptor index is not used in any ring or if
a farm index or human receptor index refers to an undefined farm or human receptor
1
12
Report an error message if an the human receptor point is not defined
1
13
Report an error message if too many or too few arguments are set in the Arguments
environment variable or if the Arguments variable has files out of order
1
14
Report an error message for a number of consistency checks of the site layout information
1
The Human Risk Module was tested using a variety of contaminants and a simple site
layout that included two distance rings and all receptor types. A variety of contaminants were
tested, because the decision to perform certain calculations depends on whether the contaminant
is a carcinogen or a noncarcinogen or has an inhalation and/or oral health benchmark available;
breastmilk MOEs are calculated for infants only for 2,3,7,8-TCDD.
The internal verification included hand-calculation verification of calculations. For some
requirements, EPA created an Excel spreadsheet to reproduce the intended functionality. The
results calculated by hand and by the spreadsheet were compared with those from the module
runs to ensure that all equations were coded properly and that the module was fully functional.
All cumulative frequency distributions calculated by the Human Risk Module matched the
values calculated by hand or using the spreadsheet. Consequently, it was concluded that the
module correctly manages all requirements listed above.
4-52
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.10.4 Summary of Validation
The Human Risk Module can not be validated because no data set on human risk values
exists.
4.11 Ecological Exposure Module
This section provides a summary of
verification and validation efforts relating to
the Ecological Exposure Module. The
Ecological Exposure Module was reviewed
by the external reviewers shown in the box.
4.11.1 Module Description
Peer Reviewers for Ecological Exposure Module
¦ Dr. Anne Fairbrother of Parametrix, Inc.
¦ Dr. Lawrence Kapustka of Ecological
Planning and Toxicology, Inc.
¦ Dr. Robin Matthews of Western
Washington University
¦ Dr. Bradley Sample of CH2M Hill
The Ecological Exposure Module calculates the applied dose (in mg/kg-d) to ecological
receptors that may be exposed to contaminants via ingestion of contaminated plants, prey, and
media (i.e., soil, sediment, and surface water). The Ecological Exposure Module uses input
concentrations from the Surface Impoundment, Surface Water, Terrestrial Food Web, and
Aquatic Food Web Modules. Figure 4-15 shows the information flow for the Ecological
Exposure Module.
Key Data Inputs
• Food consumption rate
• Body weight
• Dietary preferences
SI Concentrations
Water Column and
Sediment Concentrations
Applied
Doses
Soil, Plant, and Prey
Concentrations
Aquatic Plant and
Prey Concentrations
Surface
Water
Module
Ecological
Risk Module
Terrestrial
Food Web
Module
Aquatic Food
Web Module
Ecological
Exposure
Module
Surface
Impoundment
Module
Figure 4-15. Information flow for the Ecological Exposure Module
in the 3MRA modeling system.
4-53
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
The methodology and equations used to calculate the applied dose to mammals and birds
assigned to habitats within the AOI are consistent with the principles and guidelines described in
the Guidelines for Ecological Risk Assessment (U.S. EPA, 1998a). The basic forms of these
equations have been used by OSW and other EPA programs to predict applied doses in a variety
of ecological risk analyses, and they are similar to the exposure equations recommended by other
non-EPA risk assessors (see, for example, the Methods and Tools for Estimation of the Exposure
of Terrestrial Wildlife to Contaminants, Sample et al., 1997).
4.11.2 Major Module Components/Functionality
The Ecological Exposure Module performs the following three functions:
1. Construct a dietary matrix for each receptor for each habitat in the AOI.
The Ecological Exposure Module creates a diet for each ecological receptor based
on dietary preferences.
2. Calculate applied doses for animals in terrestrial habitats. Using the dietary
matrix and the media,2 plant, and prey concentrations calculated by the Terrestrial
Food Web Module, the Ecological Exposure Module calculates applied doses for
each avian and mammalian receptor species assigned to terrestrial habitats in the
AOI.
3. Calculate applied doses for animals in margin habitats (wetland or
waterbody). Using the dietary matrix and the media,1 plant, and prey
concentrations calculated by the Aquatic Food Web and Terrestrial Food Web
Modules, the Ecological Exposure Module calculates applied doses for each avian
and mammalian receptor species assigned to margin habitats in the AOI.
4.11.3 Summary of Verification
The Ecological Exposure Module was verified through a series of verification tests that
are described in detail in the following documents:
¦ Ecological Exposure (EcoEx) Module—Test Plan (RTI, 2002g)
¦ Ecological Exposure Module Internal Verification Testing (RTI, 2002h).
¦ Independent Tests for Ecological Exposure Module (Tetra Tech, 2000b).
EPA originally completed internal verification testing in 2000, followed by external
verification testing. Since then, EPA has made a few minor changes to the module. Therefore,
EPA repeated internal verification testing on the updated version of the module in 2002 to ensure
that the updated version passed all tests and that the test files reflected the outputs from the
updated version of the module. EPA also updated internal verification test plans and testing
documentation to reflect this second round of testing.
2 Contaminant concentrations in surface impoundments may also be used to calculate exposure if the
receptor's home range overlaps the impoundment.
4-54
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
In order to verify that the module processed the data correctly, EPA created an Excel
spreadsheet using the equations from the module documentation. The testing included
confirmation that the module correctly calculates the applied doses to mammals, birds, and
herpetofauna for all relevant years (i.e., years with nonzero concentrations and media, plants, or
prey) during the simulation period. The applied doses calculated with the spreadsheet were
compared to those from the module runs to verify that the equations were coded properly and
that the module was fully functional.
Table 4-21 summarizes the functional requirements and number of test cases executed to
verify the Ecological Exposure Module.
Table 4-21. General Requirements for Testing the Ecological Exposure Module
Step
Description
Number of
Test Cases
1
Correctly read output files from the Terrestrial Food Web, Aquatic Food Web,
Surface Water, and Source modules
3
2
Correctly read site layout file
3
3
Correctly calculate the applied dose to mammals, birds, and herpetofauna in both
aquatic (e.g., stream margin) and terrestrial habitats
3
4
Correctly assign prey preferences to all elements in the food web
3
5
Correctly loop over receptors
3
6
Correctly loop over habitats
2
7
Correctly loop over years and ensure correct time series management
1
8
Correctly select prey concentrations for each receptor
1
9
Correctly read the chemical properties and surface water data associated with methyl
mercury when multiple chemical species of mercury are present
1
4.11.4 Summary of Validation
The Ecological Exposure Module has not been validated because no adequate data set
exists to do so. However, the module is based on generally accepted science-based formulations.
4.12 Ecological Risk Module
This section provides a summary of
verification and validation efforts for the
Ecological Risk Module. The Ecological
Risk Module was reviewed by the reviewers
shown in the box.
Peer Reviewers for Ecological Risk Module
¦
Dr. Anne Fairbrother of Parametrix, Inc.
¦
Dr. Lawrence Kapustka of Ecological
Planning and Toxicology, Inc.
¦
Dr. Robin Matthews of Western
Washington University
¦
Dr. Bradley Sample of CH2M Hill
4-55
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
4.12.1 Module Description
The Ecological Risk Module calculates the HQ for species of mammals, birds, and
herpetofauna and, through inference, for communities of organisms that live in close contact
with soil (including plants and soil invertebrates), sediment (benthic invertebrates), and surface
water (e.g., fish, aquatic invertebrates, algae, and aquatic plants). The HQ is the ratio of the
annual average environmental exposure (in units of concentration or dose) to a benchmark for
ecological effects (in units of concentration or dose) and quantifies the potential for
contaminants to elicit an adverse ecological response once released into the environment. To
calculate the HQs, the Ecological Risk Module uses input concentrations from the Surface Water
and Terrestrial Food Web Modules, as well as applied doses calculated by the Ecological
Exposure Module, and compares those values to ecotoxicological benchmarks (EBs, in units of
dose) and chemical stressor concentration limits (CSCLs, in units of concentration), as
appropriate. The module calculates an HQ for every ecological receptor assigned to habitats
within the AOI for a given site. Figure 4-16 shows the information flow for the Ecological Risk
Module.
Surface
Water
Module
Water Column and
Sediment Concentrations
Key Data Inputs
Ecological benchmarks
Habitat type
Water hardness
Terrestrial
Food Web
Soil
Module
Concentrations
Ecological
HQs ^
Exit Level
Risk
Processors
1 and II
Module
l^1
Ecological
Exposure
Module
Doses
Figure 4-16. Information flow for the Ecological Risk Module
in the 3MRA modeling system.
The conceptual approach to characterizing the potential for adverse ecological effects
depends on the assessment endpoints used. Assessment endpoints are explicit expressions of the
actual environmental values that are to be protected (U.S. EPA, 1998a). Candidates for
assessment endpoints often include threatened or endangered species, functional attributes that
support food sources or flood control, or aesthetic values such as the existence of charismatic
species (e.g., eagles) that have special value to society (U.S. EPA, 1998a). The assessment
endpoints must be defined with respect to the valued ecological entity (e.g., a particular species),
and an attribute of that entity that is to be protected (e.g., reproductive fitness). For the 3MRA
modeling system, the assessment endpoints were selected to represent multiple levels of
biological organization (e.g., populations, communities), key functional elements of natural
4-56
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
communities (e.g., primary producers), and biota throughout the trophic continuum. Although
the 3MRA modeling system does not restrict the selection of assessment endpoints for future
applications, the Ecological Risk Module and supporting databases were developed to evaluate
three primary assessment endpoints:
¦ Survival of species that comprise key structural and functional elements of soil,
freshwater, and benthic (sediment) communities;
¦ Reproductive fitness and survival of mammalian, avian, and herpetofaunal
wildlife populations; and
¦ Growth and survival of primary producers (e.g., plants) in terrestrial and
freshwater systems.
These endpoints are ecologically relevant to the habitat types used to represent ecological
variability in the 3MRA modeling system and are sensitive to a broad range of chemical
stressors.
4.12.2 Major Module Components/Functionality
The Ecological Risk Module performs two major functions:
1. Calculate Hazard Quotients (HQs). The Ecological Risk Module calculates
HQs for each receptor at each site according to spatial, temporal, and
environmental characteristics of the site.
2. Process the HQ Results for Decision Making. The Ecological Risk Module
tracks various attributes such as taxa and habitat type to process the HQ results
for decision making. This processing includes placing results in bins and
determining the timing of maximum risks.
4.12.3 Summary of Verification
The Ecological Risk Module was verified through a series of verification tests, which are
described in detail in the following documents:
¦ Ecological Risk Module—Test Plan (RTI, 2002j)
¦ Ecological Risk Module Internal Verification Testing (RTI, 2002i)
¦ Independent Tests for Ecological Risk Module (Tetra Tech, 2000c).
EPA originally completed internal verification testing in 2000, followed by external
verification testing. Since then, EPA has made a few minor changes to the module. Therefore,
EPA repeated internal verification testing on the updated version of the module in 2002 to ensure
that the updated version passed all tests and that the test files reflected the outputs from the
updated version. EPA also updated internal verification test plans and testing documentation to
reflect this second round of testing.
4-57
-------
Section 4.0
Verification and Validation of the 3MRA Modeling System Modules
To verify that the module processes the data correctly, EPA created an Excel spreadsheet
and an Access database using the equations from the module documentation. The testing
included confirmation that the module (1) correctly calculates HQs for ecological receptors for
all relevant years (i.e., years with nonzero concentrations) during the simulation period, and
(2) correctly creates cumulative frequency distributions for each of the ecological risk attributes
considered by the module (e.g., receptor group, habitat type, distance). The HQs calculated by
the spreadsheet and database were compared with those from the module runs to ensure that the
equations were coded properly and that the module was fully functional.
Table 4-22 summarizes the functional requirements and test cases executed to verify the
Ecological Risk Module.
Table 4-22. General Requirements for Testing the Ecological Risk Module
Step
Description
Number of
Test Cases
1
Correctly read output files from the Surface Water, Terrestrial Food Web, and Ecological
Exposure modules
2
2
Correctly identify the constituent type as dioxin-like, organic, metal, mercury, or special
2
3
Correctly read site layout file
2
4
Correctly calculate average sediment and surface water concentrations for each aquatic
habitat delineated at the site
2
5
Correctly adjust ecological benchmarks according to site-based conditions (e.g., water
hardness) and constituent type
2
6
Correctly calculate the hazard quotients for each ecological receptor and identify the max
HQ and associated risk attributes (e.g., receptor group) for each site
2
7
Place the HQ values in the correct bins by risk attribute and build cumulative frequency
distribution
2
8
Correctly loop over receptors
2
9
Correctly loop over habitats
2
10
Correctly loop over years to ensure correct time series management
1
11
For mercury, correctly select the chemical species based on the receptor, and read the
chemical properties and surface water data associated with that species
2
All cumulative frequency distributions calculated by the Ecological Risk Module
matched the values calculated using the spreadsheet and database. Consequently, it was
concluded that the module correctly manages all requirements listed above.
4.12.4 Summary of Validation
Modules like the Ecological Risk Module are very difficult to validate because of lack of
useful data for validation purposes. We have not found any data set suitable for the validation of
this module.
4-58
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
5.0 Verification and Validation of 3MRA Site-Based
Data Collection and Processing
The 3MRA modeling system allows users to model specific facilities using site-based
data. In the representative national data set, site-based data include spatial data defining the site
layout (watersheds, waterbodies, and receptor types and locations), WMU data, soil data for the
watershed surface soils and the vadose zone, and a few other characteristics based on slope and
land use patterns. Figure 5-1 provides an overview of the steps taken to ensure the quality of the
site-based data. This section summarizes the data collection methods for the representative
national data set (Section 5.1), verification of the data (Section 5.2), and validation of the data
(Section 5.3).
5.1 Site-Based Data Collection Methods
Site-based data are data that are assigned to sites based on characteristics specific to the
site's location. A significant component of the site-based data is the site layout, i.e., the physical
arrangement features at the site. Other components of the site-based data include
¦ WMU characteristics (e.g., WMU area, capacity, and waste volumes), which are
based on site-specific Industrial D survey data,
¦ Soil texture, which is assigned based on the predominant soil texture in a
watershed from the State Soil Geographic (STATSGO) database, and
¦ Regional assignments to a meteorological station, hydrogeologic environment, or
U.S. Geological Survey (USGS) hydrologic region, which are made based on site
location or site-specific data. For example, hydrogeologic environment
assignments are assigned based on the subsurface conditions at the site.
For the representative national data set, EPA collected and processed site-based data for
201 Industrial Subtitle D1 facilities using a combination of geographic information system (GIS)
and database programs. In general, GIS Arc Macro Language (AML) programs (for Arclnfo and
Arc View) were used to establish the spatial frame of reference at each site, collect spatial data
within this frame, and interrelate different spatial data coverages using overlays and spatial
relationships to create the data necessary to populate the site layout variables required by the
1 An Industrial Subtitle D facility is an industrial facility that generates solid wastes that are not defined as
hazardous wastes under Subtitle C of the RCRA. The focus of the 3MRA modeling system representative national
data set is on the management of wastes in WMUs designed to manage nonhazardous Industrial Subtitle D wastes.
5-1
-------
Section 5.0 Verification and Validation of 3MRA Site-Based Data Collection and Processing
Conduct Pilot Study
Finalize Site Data
Adjust Data Collection
Processing Methods
Identification of Suitable
Data Sources
Validation of Data Using
Surface Impoundment
Study Data
QA/QC of Data Entry and
Processing
Figure 5-1. Overall approach to ensure the
quality of the 3MRA site-based data.
3MRA modeling system. EPA used GIS AML programs to create the following spatial data
coverages:
¦ Waterbody and watershed layout,
¦ Human receptor locations, including farms and residences, and
¦ Ecological habitats and receptor home ranges.
Figure 5-2 shows these coverages. GIS AML programs also assign soil map units to each
watershed; assign meteorological stations and hydrologic regions to each site; populate the
human receptor points using U.S. Census data; establish x,y (grid) locations for each spatial
feature; overlay and relate spatial features; and export these site-based spatial data as a series of
Access database tables.
5-2
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Human Receptors
(census and landuse data)
Ecological Habitats and Receptors
(landuse, wetlands, T&E species,
etc., data)
Watersheds
(DEMS)
Waterbodies
Streams. Lakes, and Wetlands
(DEMs, reach files, GIRAS, NWI)
Base Grid, Area of Interest
(AOI)
(specified: 100m xioom grid
cells; 2-km radius AOI)
x = facility centroid
~ = waste management unit
Figure 5-2. Site-based spatial overlays.
Following GIS processing, EPA used database programs (in structured query language
[SQL] and Visual Basic) to perform quality control (QC) checks and process site-based data
from the GIS to meet 3MRA modeling system variable and database specifications and
requirements. These database processing programs produce two primary files:
¦ The 3MRA modeling system input database, including site-based, regional, and
national input data tables, along with data tables for references, variable
correlations, user-defined empirical distributions, and general facility information.
5-3
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
¦ The grid database,2 including six data tables containing x,y coordinates for
watersheds, waterbodies, farms, human receptor points, drinking water wells, and
ecological habitats.
Table 5-1 summarizes the site-based data collection process for the 3MRA representative
national data set, including methodology and data sources. The following sections provide
additional detail on the major steps involved in creating the sample site-based data set.
5.1.1 Conduct Pilot Study
EPA conducted a pilot data collection study for five Industrial Subtitle D sites to provide
information needed to plan the data collection effort. The pilot study had three components:
1. Desktop methods, such as manual delineation of watersheds using USGS
topographic quadrangle maps,
2. Site visits, using global positioning system (GPS) technology, to obtain accurate
facility locations and verify features (e.g., receptor locations) and conditions of
interest around the site through windshield surveys, and
3. State office visits, to collect available permit information on facility location;
WMU location, design, and operation; and hydrogeology (where available).
Table 5-2 summarizes, by data type, the information that was collected during the pilot study,
along with the methods developed and used to collect the data for the 201-site representative
national data set. In many cases, the pilot study enabled EPA to identify, develop, and use
efficient and effective methodologies during the data collection effort.
5.1.2 Establish Spatial Framework/Initial Setup
Given the location (latitude and longitude) of the WMU to be modeled, the AOI for an
assessment is determined in the GIS by drawing a radius (2 km for the representative national
data set) extending from the corners of the WMU (all WMUs are assumed to be square for the
sample data set). This determines the spatial frame of reference for the analysis by defining the
area to be characterized by the site-based data collection effort and modeled by the 3MRA
modeling system. The AOI may be further subdivided into distance rings. For the representative
national data set, these rings were set at 500 m, 1,000 m, and 2,000 m from the edge of the
WMU for human risk and at 1,000 m and 2,000 m for ecological risk.
2 The grid database is used by the 3MRA modeling system Site Layout Processor (SLP) to create the air
modeling points for the Air Module and add them to the 3MRA modeling system input database.
5-4
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-1. Site-Based Data Collection for Representative National Data Set:
Summary by Major Activity
Activity
Methodology Summary
Data Sources
Conduct pilot
study
Desktop methods of data collection
Site visits
State office visits
See Table 5-2
Establish
spatial
framework/
initial setup
Locate site and place WMU
Delineate area of interest and distance rings from WMU edge
Establish base grid
Assign meteorological station, hydrologic region, hydrogeologic
environment
Collect GIS source data
Location: Industrial Subtitle D database,
EnviroFacts
Base grid coverage
Meteorological station coverage
Hydrologic region coverage
Surficial geology and aquifer coverages
Source data: see below
Delineate
waterbodies
Programs use Reach File 3 (RF3) stream networks to adjust digital
elevation models (DEMs) to be consistent with actual stream
networks (automated delineations only)
Use National Wetlands Inventory (NWI), RF3, and Geographic
Information Retrieval and Analysis System (GIRAS) for lakes
and wetlands
Attribute with type, connectivity, length, area (lakes and wetlands)
Stream network: RF3
Lakes: RF3, NWI
Wetlands: NWI, land-use coverage
(GIRAS)
Delineate
watersheds
Delineate using programs or manually
Digitize manual delineations
Attribute with area, slope, soil map units, land-use data, waterbody
connectivity
Waterbody delineation coverage
Topography: DEMs
Soils: STATSGO
Land use: GIRAS
Place
human
receptors
Place receptors at centroid of Census block/ring polygons
Place county median farm in every block group with beef or dairy
fanners and cropland/pasture land use
Place wells at receptor points in block groups with wells
Attribute with Census population data by receptor type, age cohort
U.S. Census (1990)
U.S. Agricultural Census (average of
1987
and 1992)
Land use: GIRAS (1975-1985)
Place
ecological
receptors
Manually delineate habitats using GIS ArcView tool
Place home range bins in habitats
Assign receptors to home range bins by habitat
Attribute with type, area
Waterbody delineation coverage
Topography: DEMs
Land use: GIRAS
Habitat/receptor data: literature
Overlay
GIS
coverages
Overlay and relate GIS coverages
Create data tables on connectivity, area of overlap
Create data tables with x,y coordinates of waterbodies, watersheds,
human receptors, farms, wells, and receptor home ranges
Distance ring coverage
Waterbody delineation coverage
Watershed delineation coverage
Human receptor and farm coverages
Habitat and home-range coverages
Base grid coverage
Process data
for 3MRA
modeling
system
Import GIS data tables and run QC programs and protocols
Create derived values/distributions for soil and land-use variables
Populate home ranges and attribute ecological receptors
Convert connectivity to 3MRA site layout format, including correct
and complete indices
Convert Universal Transverse Mercator (UTM) coordinates to site
coordinates
Run QC programs on completed data set
GIS data tables and grid (coordinate) files
STATSGO soil database
Land use and soil property look-up tables
(from literature)
3MRA data specifications
Citations and detailed descriptions of the data sources and methodologies shown in this table can be found in Volume II, Data
Collection.
5-5
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-2. Comparison of 3MRA Modeling System Representative National
Data Set Data Collection Activities with Pilot Study Approach
Data Type
Approach for 201 Sample
Data Set Sites
Approach for Pilot Study
Facility location
Industrial Subtitle D facility match to
Locational
Reference Tables (LRT) (facility
centroid, front gate, zip code
centroid locations)
Address matching software for
unmatched sites
GPS front gate, facility boundaries, WMUs
State office visits (facility maps, locations)
Waste management
unit
Industrial Subtitle D Screening Survey
data (WMU area, capacity, waste
loading)
National estimates based on model unit
approach for other inputs
State office visits (facility maps; WMU
locations,
dimensions, operation)
Site visits to confirm characteristics where
WMUs are visible
Land use
GIS analysis (GIRAS coverage)
Confirm land use, details within GIRAS area
coverages during site visits
Topographic
Watershed delineation (area, flow
length, slope, streams) using GIS
DEMs or manual methods (flat
sites)
Desktop delineation of watersheds (area,
flow length, slope, streams) on USGS
topographic maps
Confirm general watershed characteristics,
topographic features, man-made
drainage during site visits
Waterbody
Waterbody network delineation using
RF3, NWI, GIRAS and DEM
coverages
Assume lakes, order 4 streams are
fishable
Desktop location on topographical map
During site visits, GPS waterbody locations
(streams, lakes, ponds), confirm general
characteristics and add details on
fishable waterbodies
Aquifer
Hydrogeologic Database (regional
analysis), national distributions for
aquifer properties
Assume water flows downhill, towards
surface water
Desktop data review and compilation from
state office visits and hydrogeologic
setting analysis
During site visits, ground-truth setting and
detail on residential well use
Human receptor
information
U.S. Census data (block centroids,
areal averages within radius of
interest) and other data sources
(county agricultural census,
national home gardener
percentage)
Ground-truth locations (GPS), areal
averages, farms, exposure pathways
Provide detail on home gardeners, fanners,
subsistence activities, maximum
exposed individual
Ecological receptor
information
Habitat within AOI delineated using
GIRAS land-used data,
topographic, other data
Ground-truth habitats, exposure pathways
Once the AOI and rings were established, the site-based source data were collected for
each of the spatial data layers to be created. (The GIS data sources for these data are
5-6
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
summarized in Table 5-1.) Regional assignments, including meteorological station and USGS
hydrologic region, were assigned using GIS overlays during this initial step of the analysis.
Site-based WMU data include the WMU dimensions, capacities, and 1985 waste volumes
obtained from EPA's Screening Survey of Industrial Subtitle D Establishments (Schroeder et al.,
1987) and the size-related WMU variables imputed from the Industrial Subtitle D survey data.
The imputed WMU inputs were derived from the site-specific Industrial Subtitle D data using
national relationships developed from literature, personal communications, and best engineering
judgments.
5.1.3 Delineate Waterbodies and Watersheds
Creating the waterbody and watershed layout requires GIS processing to delineate the
waterbodies and watersheds and to obtain spatially related parameters (Figure 5-3, Steps 1
through 8) and then conducting database processing of the GIS data to provide the exact data for
the 3MRA modeling system (Figure 5-3, Step 9). For the representative national data set,
waterbodies (lakes, streams, and wetlands) within the AOI were delineated using nationally
available coverages, including EPA's RF3, land use/land cover data (GIRAS), and, where
available, the NWI. Watershed subbasins were delineated around each waterbody either
automatically, using DEMs of topography, or manually, using topographic maps created from
the DEMs.
After waterbody and watershed delineations were complete, waterbody type was
attributed using the original GIS coverages. Watershed and surface water connectivity, as well
as stream order for manually delineated sites, were assigned manually. A local watershed3 was
identified for the WMU and divided into two to three subareas, depending on the WMU type and
topography. Because the waterbody layout data were highly variable from site to site, the
resulting waterbody data were visually QC-checked for every site before final processing.
5.1.4 Place Human Receptors
For the representative national data set, EPA used GIS programs to place human
receptors within the AOI using U.S. Census block centroids (or block/ring centroids if a block
was divided by a distance ring) and to randomly place beef and dairy farms based on U.S.
Census block group, agricultural census, and land-use data. Table 5-3 shows the primary data
sets used to derive the human receptor data.
The representative national data set includes two basic receptor types: residential
receptors (residents and home gardeners) and farmers. Residential receptors may be recreational
fishers in addition to being a resident or home gardener. Farmers may be beef or dairy farmers,
and either type of farmer may also be a recreational fisher. Within each of the two basic receptor
types, the 3MRA modeling system considers five age cohorts: infants (aged 0 to 1 year), children
aged 1 to 5 years, children aged 6 to 11 years, children aged 12 to 19 years, and adults (aged 20
years and older).
3 The drainage subbasin that includes the WMU.
5-7
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Step 1:
Preprocess National Data
t
Step 6:
Delineate Local Watershed
Step 2:
^
¦ Create local watershed coverage
for each WMU
Prepare Site-Specific Data
~
~
Step 7:
Overlay Coverages
with Reference Grid
Step 3:
Determine Delineation Method
*
t
Step 4:
Delineate Watersheds
¦ Manual delineation/Semi-
automated DEM delineation
Step 8:
Compile GIS Data Tables
~
¦ Attribute and number watershed
subbasins
¦ Create watershed coverage
i
Step 5:
Delineate Waterbodies
Step 9:
Process Data to Determine
Parameters for Watersheds/
Waterbodies
¦ Create stream network
¦ Create waterbody network
coverage
¦ Attribute reach connectivity
Figure 5-3. Overview of watershed and waterbody layout processing.
5-8
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-3. Primary Human Receptor Data Sets, Date, and Scale
Data/
Receptor Type
Data Set
Date
Scale
Residents,
home gardeners
U.S. Census block data
Summary Tape File (STF) IB attribute data
U.S. Census Topologically Integrated
Geographic Encoding and Referencing
System (TIGER)/Line block coverages
Exposure Factors Handbook (home
gardener)
1990
1:100,000 scale
mapping
Beef and dairy
fanners,
farm size
U.S. Census block group data
STF 3 A attribute data
U.S. Census TIGER/Line block group
coverages
1990
1:100,000 scale
mapping
GIRAS land-use data
1975-1985
1:250,000 scale
mapping
U.S. Census of Agriculture
1987 and 1992 (avg.)
County level
Recreational
fishers
National Survey of Fishing, Hunting, and
Wildlife
1992
State level
Citations and detailed descriptions of the data sources shown in this table can be found in Volume II, Data
Collection.
Following the placement of receptors, each receptor point and farm was populated by
receptor type and age cohort using 1990 U.S. Census data (and other data shown in Table 5-3)
and assuming a uniform distribution of the population across the entire block.
5.1.5 Place Ecological Receptors
The representative national data set includes 12 terrestrial, wetland, and waterbody
margin habitats, each with assigned ecological receptors that were developed to represent
ecological receptors nationwide. For each site, EPA delineated ecological habitats by manually
applying an Arc View GIS tool that included GIRAS land use data, the representative national
data set waterbody and watershed coverages and attributes, and DEM topographic data. Habitats
were delineated by selecting grid cells and coding them with one of the 12 habitat codes.
After the manual habitat delineation was completed, EPA generated receptor home
ranges using automated GIS AML programs. Because placing a home range for each receptor
within every habitat proved inordinately time consuming (even though automated), EPA used
four home range bins to reduce the number of home range placement iterations. Within a
habitat, each receptor was assigned to one of four bins based on its average home range size.
The GIS programs delineated the four home ranges for each habitat by randomly placing the
largest bin within the habitat and then randomly placing subsequent, smaller bins within the next
largest bin.
Following the placement of home ranges, both the habitats and home ranges were limited
to the area within the AOI and overlaid with the coverages of waterbodies, watersheds, local
watersheds, and distance rings to determine the spatial relationships between them. Ecological
5-9
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
receptors were added to the habitat using two look-up tables: receptors by habitat type and
receptors by Bailey's section.
5.1.6 Overlay GIS Coverage
Because the GIS coverages cannot be exported directly to the 3MRA modeling system,
spatial layers are defined in terms of their relationships to each other and in terms of a base grid
composed of 100 x 100 m cells. This resolution roughly corresponds to the minimum resolution
of several of the site-based data sources used to develop the representative national data set. To
create the grid files, GIS coverages for watersheds, waterbodies, farms, and ecological habitats
were converted to UTM coordinates and overlaid on the master 100 x 100 m grid file to create
grid tables containing the x,y UTM coordinates of each grid cell occupied by a feature. The
UTM coordinates for human receptor points and wells (human receptor points with drinking
water wells) were also sent as part of the grid database.
In the 3MRA modeling system, spatial features are related to each other using variables
in the site layout data group. These variables include indices that identify the features that
connect to or overlap with another type of feature (e.g., which watersheds overlap with a farm,
which watersheds contain a human receptor) and the fraction of overlap. EPA used GIS
programs to overlay the coverages (such as farms over watersheds, or receptor points over
distance rings) to determine the spatial relationships between coverages (e.g., watershed
occupied by human receptor or farm, waterbody used by farm, etc.).
5.1.7 Process Data for 3MRA
EPA used an Access database to process GIS and other data to create the 3MRA
modeling system representative national data set input database. This processing database
includes a series of SQL and Visual Basic programs to automatically QC-check and process the
site-based, regional, and national data necessary to run the 3MRA modeling system. For the GIS
data, the programs check the incoming data for completeness and consistency. Programs then
convert the site-based data to the 3MRA modeling system format. This conversion includes
¦ Creating and attributing the waterbody network variables
¦ Connecting watersheds to waterbodies
¦ Providing data for soil- and land-use-derived variables
¦ Matching ecological receptors in each habitat to the appropriate home range bin
¦ Assigning indices and ensuring all indices are correct and complete
¦ For the grid database, converting real-world UTM coordinates to a set of metric
x,y points centered about 0,0 (at the facility centroid).
The database processing performs any necessary calculations and/or data manipulations to
produce the final variables and format required by the 3MRA modeling system.
5-10
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
5.2 Data Verification
Data verification includes the activities undertaken prior to and during the data collection
effort to ensure that data of the correct type, amount, and quality (i.e., data that meet data quality
objectives) are provided in the 3MRA modeling system representative national data set. These
activities included
¦ A data collection plan that specified data sources and how data would be
collected from those sources;
¦ Quality assurance (QA)/QC protocols that were specified as part of the data
collection plan, including data entry checks, independent calculations to verify
that data were processed correctly in all circumstances, and automated checks of
critical parameters, formats, and processes; and
¦ Independent testing of the major site-based data elements.
EPA updated the data collection plan and it is now the documentation for the overall data
collection effort (Volume II). Section 5.2.1 describes the QA/QC protocols that EPA planned
and implemented during the 3MRA modeling system representative national data set data
collection effort. Section 5.2.2 describes independent testing of the data.
5.2.1 Quality Assurance/Quality Control
Prior to data collection, EPA developed a basic QA/QC protocol for each data type and
distributed it to all staff working on the data collection. In addition, EPA developed certain QC
protocols common to many data types (see below). Any necessary deviations from these
protocols during data collection were discussed with and approved by the team leader and the
QA officer. The specific QA/QC protocols used during the data collection effort are discussed in
detail by data type in Volume II, Data Collection. EPA conducted QA to ensure that an
adequate QC methodology was in place and was correctly implemented and recorded.
The common QC protocols outlined in the data collection plan were followed during the
data collection effort with minor changes. Common QC protocols for site-based data include
¦ Conducting a senior review and manually checking 100 percent of data entered
from hardcopy sources.
¦ Recording the name of the staff member performing QC checks and the date QC
checks were performed as part of the QC record.
¦ Maintaining files documenting QC activities. These files were used to track
information such as data sources, data entry, and changes to data, and included
copies of hardcopy data sources.
¦ Keeping metadata electronically for all electronic data sources
5-11
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
¦ Validating the data extraction and processing system for automated import of data
from electronic sources and automated data processing in the electronic data
sources before use through hand checks and calculations. After initial system
validation, a sufficient fraction of the data (usually 5 to 10 percent) was manually
checked to ensure that the data processing system was functioning properly.
When possible, automated checks were also built into the system to detect data
inconsistencies
¦ Validating the 3MRA modeling system input data processor system by manually
checking a portion (usually 5 percent or more) of the processed data for each
variable to ensure that the system is functioning properly. The system also
included automated checks to identify inconsistencies
¦ Using the 3MRA modeling system Site Definition Processor (SDP) to check each
database update for missing data and for consistency with the 3MRA modeling
system input data specifications (i.e., the site specification file dictionary files).
For certain types of problems discovered during the QC of automated processing, it was
possible to create automated checks that would be performed during the processing to catch any
similar errors. For instance, these checks included searching for problems that violated model
specifications, as well as performing other logical checks of the data. Table 5-4 contains a list of
automated checks performed on the final model input database, all of which resulted from errors
found during QC. Checks for certain errors, such as duplicate rows of data, checks of the
indices, and other routine checking could uncover errors resulting from a number of different
processing problems.
During the data collection process, EPA made minor changes to the protocols set forth in
the data collection plan for individual data types. Many of the changes resulted from changes in
the data collection methodology or further QC needs based on the initial checks of the data or
data processing. Changes from the planned QC protocol to those actually employed are
summarized by data type in the following sections.
5.2.1.1 Site/WMU locations. The data collection plan outlined manual and automated
QC activities designed to obtain better locations for the Industrial Subtitle D facilities. EPA
used automated matching of zip codes and addresses between the Industrial Subtitle D database
and EPA's LRT database to identify spurious facility locations and duplicate locations, which
EPA then screened manually to eliminate duplicates and mismatched data. EPA conducted
manual checks of the matched sites to verify the automated matching process. When zip codes
or addresses did not agree, EPA conducted manual verification. Although none of the planned
QC protocols were changed, the need for additional review of site locations was identified
during the initial watershed delineations of the sites. Some locations put WMUs in rivers or
other waterbodies. Other sites ended up in areas of inappropriate land use (e.g., large surface
impoundments in residential land use areas). As a result, each site/WMU location underwent a
visual review and manual relocation as necessary to ensure reasonable location prior to GIS
processing.
5-12
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-4. Automated QC Performed on Model Input Database
Data Element
Automated QC
All data
Data type checked against data type in data dictionary (DIC) file
All data
Units checked against units in DIC file
All data
Central tendency value checked against maximum in the DIC file
All data
Central tendency value checked against minimum in the DIC file
All data
Maximum value checked against maximum in DIC file
All data
Minimum value checked against minimum in DIC file
All site-based data
Checked for duplicate rows in the site-based data table
All site-based data
All indices in the site-based data table checked to make sure that the largest index is
equal to the central tendency of the Num variable to which the index corresponded
based on DIC file
Grid tables and site
centroid
Checked that the average x and >• coordinates from the combined WBNRch and
WSSub grid tables are less than 100 meters
Local watersheds
Checked the number of local watershed subareas to make sure they are correct based
on WMU type
Ecological habitats
Checked to make sure that the number of habitats for different WMUs at the same
facility are the same
Ecological habitats
Checked to make sure that the number of habitats in EcoRing 3 (which covers the
entire site AOI) is equal to the value of NumHab and that the number of habitats in
Rings 1 and 2 does not exceed the number in EcoRing 3
5.2.1.2 Waste Management Unit. The QC plan for site-based WMU data included
10 percent checks of automated data extraction with checks being conducted across all WMU
types to ensure that all calculations in the database were checked and correct. Internal reviews
consisted of senior engineer review of individual parameter values for realism and review of
overall model system designs to ensure that parameter estimates within the model were
internally consistent. External reviews of model facility designs and parameter estimates were
also conducted to ensure that these were representative of typical industry practices. All of the
QC protocols in the data collection plan were implemented during the data collection.
5.2.1.3 Watershed and Waterbodv Layout. The data collection plan outlined a largely
automated methodology for delineating watershed subbasins and waterbodies. However, as EPA
developed programs and reviewed actual coverages, it became apparent that manual interaction
would be required to delineate the watersheds and waterbodies accurately. For some of the sites,
the 1-degree DEMs were not of sufficient quality to be used in the automated watershed
delineation, so EPA developed a semi-manual methodology. Because of these difficulties in the
delineation, EPA greatly expanded the focus of the QC protocols for these data. The QC
protocols in the data collection plan primarily focused on checking the DEMs for problems,
reviewing and verifying programming, and documenting and storing programs and metadata.
EPA modified some of these protocols and added many new ones to accommodate the updated
methodologies. Because the new method of delineating watersheds involved manual interaction
with programs, the QC checks focused more on checking those manual interactions and less on
checking the automated programs, since problems with the programs would typically be
5-13
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
discovered during the manual interactions. Many new QC protocols were added during the data
collection effort because the data were more complex and variable than were expected. QC
protocols were added to check the GIS processing and the database processing that processed the
raw GIS data to fill the variables required by the 3MRA modeling system. The GIS QC
protocols required a GIS staff member, other than the one doing the delineations, to visually
examine each site and run an interactive macro to verify that everything was done correctly. The
database processing QC protocols required that all waterbody layout variables be checked for
each site following processing. The watershed layout variables were more straightforward and
consistent, and, therefore, only about 10 percent of the data were checked following processing.
5.2.1.4 Soil Data. In general, EPA used the common QC protocols for the soil data.
EPA checked 100 percent of manually entered data following review by a senior staff member,
and performed manual checks of automated calculations on processed data. To check data
compilation from STATSGO and CONUS4 tables within the soil database, EPA checked all data
processing operations using hand calculations for soil map units randomly selected from the
database. When special processing rules applied to certain data categories, at least one of each
type was chosen and checked.
5.2.1.5 Hvdrogeologic Environment. The site-based hydrogeologic assignment and
aquifer temperature were transferred in electronic format from the previous EPA models. EPA
spot-checked these data after transfer to ensure that they were processed into the 3MRA
modeling system format without error. For the GIS-derived spatial layout variables generated
automatically, a percentage of the final values were checked manually to ensure that the GIS and
database programs were processing data accurately and consistently. One exception to the
spatial layout variables was the aquifer flow direction, assumed to flow downhill toward
waterbodies, which was checked manually for 100 percent of the data. The scale of the local
watershed at many of the sites made the use of the DEMs somewhat unreliable in determining
the downhill flow direction.
5.2.1.6 Human Receptors. The QC plans for human receptor data included developing
automated programs to check individual block/group population values in the Census coverages
for population outliers or mistakes in the data. The programs compare all Census population-
type values against urban/rural status and polygon size to check the reasonableness of the data.
EPA also conducted random checks against different data sources (e.g., Census CD, a
commercial CD of Census data) and performed manual calculations for a few sites to validate the
programs used in the automated processing. EPA followed all of the QC protocols in the data
collection plan and developed additional protocols based on problems discovered during
processing. Because of the variability of Census data overlaid with land use data, visual
inspection of farm placement at each site was the safest way to ensure correct farm placement at
every site that had farmland use and farmers. To further check the population numbers in the
final representative national data set input database, EPA compared population totals from the
database (summed across all receptor points, receptor types, and age cohorts), to population
totals (within the 2 km AOI) obtained from GIS Census coverages. Additionally, EPA created
an Excel spreadsheet that repeated the calculations of the human farm receptor GIS program.
4 Data set for the conterminous United States.
5-14
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
This spreadsheet expedited QC by enabling staff to enter raw data into the spreadsheet and
calculate expected results for comparison with the output of the GIS programs. Prior to use, the
methodology and logic for the formulas in the spreadsheet were reviewed and checked by senior
staff.
5.2.1.7 Ecological Receptors. The QC performed for ecological receptor placement
focused on the delineation and GIS processing of the habitats and ranges. Because habitat
delineation consisted largely of subjective evaluations and hand delineation of habitat
boundaries, two designated senior ecologists performed all habitat delineations. Limiting the
delineators to two individuals helped limit the degree of variation in interpretation of spatial
data. Both delineators adhered strictly to the crosswalk tables developed for delineation. Both
delineators kept records of any delineations that involved unusual circumstances or conflicting
issues. These notes were reviewed by both delineators to maintain consistency. Following
delineation of the habitats at all sites, the delineators performed checks on 75 percent of each
other's delineations. Before final data processing, automatic QC programs were used to query
the GIS data tables to ensure that each habitat had no more than four home range bins and that
there were no grid cells in the home range that were not also in the habitat containing the home
range.
5.2.1.8 Grid Database. To ensure the accurate transfer of grid information to the
3MRA modeling system grid database, EPA performed the following QC activities during data
collection:
¦ Automatic regeneration of the grid template for a site prior to GIS postprocessing
of spatial data to ensure proper correspondence with the facility centroid;
¦ Generation and visual review of thumbnail images of all spatial data for every site
to ensure accurate registration and collocation of all data layers; and
¦ Visual checks of a subset of sites in the 3MRA modeling system grid database
against the original GIS coverages to ensure accurate data processing and transfer.
5.2.2 Independent Data Testing
EPA conducted independent data testing to assess the accuracy of the site-based data in
the 3MRA modeling system representative national data set. In addition to independent testing
of the data collection methodologies and the 3MRA modeling system representative national
data set, QA/QC procedures performed during data collection were also reviewed. The
independent data testing addressed site-based data collection in the following areas:
¦ Spatial layout
¦ WMU data
¦ Watershed and waterbody layout
¦ Soil data
¦ Human receptor data
¦ Ecological habitats and receptors.
5-15
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Specific items were checked for each data area. Testers carried out the following checks:
¦ Reviewed and compared data collection and model documentation for
consistency.
¦ Reviewed QA/QC history for completeness. Data preparers were contacted to
review information on QA/QC history to identify any potential gaps in the
QA/QC of the data.
¦ Reviewed selected data methodologies, programs, and results. Techniques used
depended on data type, but often included comparing the input data to the original
source, independently re-creating input data from raw data and comparing them to
the model input data, and reviewing and checking QC procedures and records.
Testing confirmed that most of the data were accurate, but some errors were identified
during the testing and later corrected in an updated 3MRA modeling system representative
national data set (in September 2000). Table 5-5 summarizes the results of independent testing
of site-based data. For more detail on the independent data testing, see Independent Data
Testing for the Hazardous Waste Identification Rule—Draft (RTI, 2000a).
Table 5-5. Summary of Results of Independent Testing of Site-Based Data
Data Element
Data Testing
Spatial framework
Coordinates in the grid database were plotted using ArcView for 10 sites and
compared to the jpeg images provided in the data docket. Coordinates from all
tables— AquWell, Farm, Habitat, HumRcp, WBNRch, and WSSub— were
plotted.
Landfills
Landfill data were extracted from the Industrial Subtitle D database, used to derive
the landfill depth, and used in a statistical regression to calculate replacement
capacity values.
Waste piles
Waste pile data were extracted from the Industrial Subtitle D database and used to
derive replacement values for missing waste quantities. A statistical regression
was done using facilities that had waste quantity data and met certain height
criteria.
Land application units
LAU data were extracted from the Industrial Subtitle D database and used in a
statistical regression of facilities that reported waste quantity data and met the
waste application rate constraints.
Surface impoundments
Extracted Industrial Subtitle D surface impoundment data were used to derive the
depth of the surface impoundment. To calculate replacement values, two
statistical regressions were done: one using facilities that reported waste quantity
data and one using facilities that reported capacity data and also met the unit
constraints.
Watershed and waterbody
layout
Watersheds were delineated manually onUSGS 1:24,000 topographic maps using
the same guidelines used in the data collection documentation.
(continued)
5-16
-------
Section 5.0 Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-5. (continued)
Data Element
Data Testing
Soil data
A review was conducted of QA/QC history associated with use of soil parameters
for various projects.
Human receptor data
Review of receptor placement and attribution was conducted using the Census
data and methodologies described in the documentation.
Ecological receptors and
habitats
Habitats were manually delineated and, according to the same methodology,
described in the data collection documentation and compared to the 3MRA
modeling system representative national data set delineations. Receptors were
manually assigned for each of these habitats and compared with the list of
receptors in the 3MRA modeling system database.
Table 5-6. Comparison of Industrial D and SIS Survey Data for Common Sites
Facility Name,1
Address, SIC2
1985 Industrial
Subtitle D Survey Data
1999 SIS Survey Data
Crown Paper Co.
(Crown Zellerbach Corp.)
St. Francisville, LA 70775
2611 (Pulp mills)
0832903:3
impoundments, 121 acres
(1,355,075 metric tons)
total; 1985 waste volume
is 40,824 metric tons per
year
3062: 2 impoundments.
¦ ASB (Aeration Stabilization Basin; aerated
biological treatment), 43 acres (1,158,100 metric
tons); 42,860,356 metric tons per year
¦ WSI (West Sludge Impoundment; sedimentation
and anaerobic biological treatment), 32 acres
(419,323 metric tons), 172,824 metric tons per
year
Cenex Harvest States
Cooperatives; Laurel
Refinery (Farmers Union
Central Exch)
Laurel, MT 59044
2911 (Petroleum refining)
1230111: 3
impoundments, 4.13
acres (33,877 metric
tons) total; 1985 waste
volume is 557,928 metric
tons per year
2418: 2 impoundments Cassressive aerated biolosical
treatment)
¦ North Aerated BioPond, 1.50 acres (11,089 metric
tons); 1,106,074 metric tons peryear
¦ South Aerated BioPond, 1.49 acres (9,476 metric
tons); 1,077,845 metric tons peryear
1 Industrial D Survey name in parentheses
2 SIC = Standard Industrial Classification Code
5.3 Data Validation
EPA validated the accuracy of the site-based data in the representative national data set
by comparing those data with data and model results for two of the sites where more recent data
were independently collected during EPA's 1999 Surface Impoundment Study (SIS) Survey
(U.S. EPA, 2001). The following section describes this validation activity in detail.
5.3.1 Surface Impoundment Study Data
The 1985 Industrial Subtitle D Survey and the 1999 SIS Survey were statistically
designed to characterize the same universe: facilities managing nonhazardous industrial wastes
5-17
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
in onsite WMUs. The two surveys overlap in their coverage of surface impoundments, but differ
in the level of detail of information collected. The Industrial Subtitle D Survey characterized
number of surface impoundments and the total area, capacity, and waste volumes of those
surface impoundments for more than 1,800 facilities with surface impoundments across the
country. The SIS Survey collected very detailed information about waste characteristics, design
and operating conditions, and the surrounding environment (including the location of residences
and drinking water wells) for 220 facilities. Because the SIS survey was statistically designed, it
should be valuable in validating and corroborating the older Industrial Subtitle D data, as well as
the supplementary data collected for the 3MRA modeling system representative national data
set.
Although more than 60 of the SIS facilities overlap with the Industrial Subtitle D survey
sites, only two of the 201 Industrial D sites randomly selected for the 3MRA modeling system
representative national data set are also in the SIS data set. These sites are shown in Table 5-6,
which also compares the waste management information supplied with each survey. For each
site, three impoundments were reported in the 1985 survey, compared with two in the 1999 SIS.
With respect to the impoundment size, area and total capacity compare reasonably well, as do
annual waste volumes for the Montana refinery (1230111). However, the Industrial D 1985
annual waste volume for the Louisiana pulp mill (0832903) appears to be very low when
compared with the SIS data.
5.3.2 Development of SIS Data Sets
The two overlapping sites provided an opportunity to validate the Industrial Subtitle D
survey data used for the representative national data set against independent 3MRA modeling
system data sets developed using the more recent and more detailed SIS information available
for these sites.
Table 5-7 compares the data sources and collection methodologies used by the 3MRA
modeling system sample national data collection effort and SIS. The primary differences are that
the SIS contains more detailed information of surface impoundment operation characteristics and
waste properties and the SIS survey responses provide more accurate human receptor locations
by marking individual residences on topographic maps. The 1999 SIS data are also more recent
when compared to the data sources used for the 3MRA modeling system representative national
data set, which range from the 1985 Industrial Subtitle D Survey to the 1990 U.S. Census.
5-18
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-7. Comparison of Site-based Data Sources: SIS and 3MRA Modeling
System Representative National Data Set
Data Category
3MRA Modeling System Data
Source / Methodology
SIS Data Source / Methodology
Facility location
Zip code centroids (from Industrial D
survey), improved using EPA Envirofacts
preferred locations, address matching, and
visual placement using GIS coverages
Respondent-marked topographic maps and
diagrams showing impoundment locations
and areal dimensions
WMU dimensions
1985 Industrial D survey data: number of
units, total area, total capacity, total 1985
waste volume (totals of all impoundments)
1999 SIS Survey data (by in-scope1
impoundment): area, capacity, average flow
rates (wastewater and sludge), diagrams
(depth)
WMU operating
characteristics
Imputed from Industrial D dimensions using
published and derived engineering
relationships; aeration assumed for all units
1999 SIS Survey detailed data on
impoundment type, function, aerators,
mixers
Waste properties
Assumed using national distributions
1999 SIS Survey data on contaminant
concentrations and other waste properties
(pH, temperature, BOD, COD, TSS,2 etc.)
Human receptor
locations
1990 Census block/ring centroids; wells
placed based on block group data
Respondent-marked topographic maps
showing residences and wells within 2 km
Hydrogeologic
environment
Assigned based on zip code and national
atlas of conditions
Assigned based on SIS survey subsurface
information; national GIS coverages of
aquifers, soils, and surficial geology
1 Impoundments that have contaminants of concern present or a pH below 2 or above 11.
2 BOD = biological oxygen demand; COD = chemical oxygen demand; TSS = total suspended solids
5.3.3 Preliminary Comparisons
Although the SIS-based 3MRA modeling system data sets are not complete, some
comparisons may be made using the underlying data. Table 5-8 compares WMU and waste
property data from the 3MRA modeling system representative national data set and SIS sources.
Significant differences include the following:
¦ Impoundment areas and depths are similar, but the Industrial Subtitle D waste
flows are significantly lower than the SIS data, especially for site 0832903/3062
¦ The degree of aeration assumed in the 3MRA modeling system representative
national data set is significantly higher than reported in the SIS survey, especially
for site 1230111/2418
¦ The SIS data contain actual site-based waste property data compared to the
national distributions or fixed values assumed for the 3MRA modeling system
representative national data set.
5-19
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
Table 5-8. Comparison of Waste Management Data—3MRA
Modeling System Representative National Data Set versus SIS
Laurel Refinery (MT)
Crown Paper (LA)
3 MR A
SIS
3 MR A
SIS
Variable
1230111
2418North
2418South
832903
3062ASB
Number of surface
impoundments
3
2
2
3
2
Dimensions
Area (m2)
5,571
6,056
6,043
163,229
174,015
Depth (m)
2.77
1.52
1.63
2.77
4.55
Operating Volumes
Q_wmu (m3/s)
0.0059
0.035
0.034
0.0004
1.36
d_setp'
0.41
0.26
0.14
0.57
0.31
Aeration Characteristics
d_imp (cm)
61
14.2
14.2
61
49.5
w_imp (rad/s)
126
377
377
126
126
02eff
distribution
NA4
NA4
distribution
0.85
n_imp2
8
4
2
55
32
Power (hp)
556
40
20
5,000
2,400
F_aer3
0.52
0.034
0.017
0.17
0.072
Waste Properties
PH
distribution
8.09
7.64
distribution
7.6
Temp (C)
8.64
35.3
18.2
19.4
41
BOD (g/cc)
distribution
0.000006
0.00001
distribution
0.000243
Density
0.998
1.001
1.001
0.998
NA
TSS (g/cc)
distribution
0.000084
0.00001
distribution
0.000133
foe
distribution
0.55
0.55
distribution
0.74
1 fraction of impoundment occupied by sediments
2 number of impellers
3 fraction aerated
4 NA = not available
Measurement of the impacts of these differences on model results will be possible once the SIS
data sets are ready for use in the 3MRA modeling system. Depending on the results of these
comparative runs, it may be desirable to use the SIS data set to develop more reasonable site-
based data from the existing Industrial D data. These improvements could include more
appropriate operating characteristics and waste properties for the specific industries represented
by the Industrial Subtitle D and SIS surveys.
Figures 5-4 and 5-5 compare the site layouts generated for the 3MRA modeling system
with the SIS data. Preliminary observations include the following:
5-20
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
lHI
sa
HWIR GTS Dnn^tMijn
Receptor Data
for Site 0832903
T.argwt WMTf typp at site: si
Legend
« Sites
I 1A01
] WMU
Coiinly Boundaries
^ Farms
Human Recsptor Points
• Wells
I 1 Block Boundaries
Far laiduse legend ;ee landuse map
PROPERTY LINE
(POTABLE "WATER'' SUPPLY-IN DUS TRIAL USE)
PUBUC WATER1
SUPPLY W-5
Figure 5-4. Crown Paper, St. Francisville, LA: Comparison of 3MRA modeling system
(0832903, upper) and SIS (3062, lower) site layouts. (Red dots indicate
receptor locations.)
5-21
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
wrr' %¦.
^ i \
V I
MM
i§i
ty/Mw,
HWIP. GIS Docket Mip;
Receptor Data
for Site 1230111
Largest WMU type atsitc: si
I .egend
• Sites
I I API
~ WMU
County Boundaries
^ Farms
. Human Receptor Points
- Wells
I I Block Boundaries
Far landusc legend jcc landusc map
Laurel
Property Line
i ti/'—^ I
'North and Souths^
Aerated Bio Ponds
¦River Flow Direction
V if '
Figure 5-5. Cenex Refinery, Laurel, MT: Comparison of 3MRA modeling system
(1230111, upper) and SIS (2418, lower) site layouts. (Red dots indicate
receptor locations.)
5-22
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
¦ The location for the Louisiana site was very accurate, but the 3MRA modeling
system location for Montana was across the river from the actual location.
¦ Receptors were similar for the Montana site, but the SIS data provide more highly
resolved placement of residences within Census blocks than the block/ring
centroids used in the 3MRA modeling system.
¦ For the Louisiana site, receptors placed at block/ring centroids on the east side of
the Mississippi river were not present in the SIS data (i.e., residents in these
blocks are actually outside of the 2 km-radius AOI).
As with WMU characteristics, the SIS data sets will allow for these differences to be quantified
with respect to the 3MRA modeling system's risk estimates. To separate impacts of the WMU
differences from the impacts of site layout differences, hybrid data sets can be prepared by using
the same WMU characteristics for the different site layouts or by using different WMU data at
each site layout.
Aquifer assignments also differed between the 3MRA modeling system representative
national data set and the SIS data set. Based on SIS survey subsurface data and GIS coverages
of surficial geology, soil, and aquifers, both sites were assigned to a river alluvium
hydrogeologic environment (GWClass 6), versus the 3MRA modeling system representative
national data set assignments of sand and gravel (GWClass 4) for the Louisiana site and alluvial
basins (GWClass 5) for the Montana site. This difference can be attributed to the fact that the
3MRA modeling system representative national data set assignments (which date to before 1995)
were based on zip code centroid locations and a fairly low-resolution national atlas of principal
aquifers. As a result, the narrow alluvial aquifer environments (GWClass 6 and 7), which should
be fairly common given that industrial facilities are often located along rivers, are not
represented in the 201-site representative national data set.
5-23
-------
Section 5.0
Verification and Validation of 3MRA Site-Based Data Collection and Processing
[This page intentionally left blank]
5-24
-------
Section 6.0
Verification and Validation of Chemical Properties Models
6.0 Verification and Validation of Chemical
Properties Models
The chemical properties for the constituents currently included in the 3MRA modeling
system database were developed through modeling and literature search. This section describes
the verification and validation for the two models used to develop some of the chemical
properties: SPARC (System Performs Automated Reasoning in Chemistry) and the MINTEQA2
Geochemical Speciation Model. Figure 6-1 provides on overview of the steps taken to ensure
the quality of the chemical properties models.
6.1 SPARC Chemical Properties Estimator
This section documents the verification
and validation activities for the SPARC chemical
properties estimator model. Although not
incorporated into the 3MRA modeling system as i
module, the SPARC model was used to calculate
the thermodynamic organic chemical properties
required by the various modules, notably
solubility, vapor pressure, Henry's law constant,
octanol/water partition coefficient, air diffusivity,
water diffusivity, and ionization potential. The
verification and validation activities described in
this section are those conducted for the SPARC
model in general, but because they confirm the
basic chemical functionality of this model, they
are relevant to its application to the 3MRA
modeling system.
The SPARC model has been extensively
peer reviewed and papers related to the model
published in the literature. The SAB also peer
reviewed the model in 1991.
Verification
SPARC—quarterly QAto compare
results to historical results
MINTEQ—QA, code verification
Validation
Comparison to measured data
Peer F
.eview 1
i
r
Model Updates |
Figure 6-1. Overall approach to ensure the
quality of the 3MRA chemical property data.
For the 3MRA modeling system, SPARC was used to calculate chemical and physical
property values for standard conditions (i.e., 25° C, pH of 7). These values are stored in a
chemical database accessed by the 3MRA modeling system Chemical Properties Processor
(CPP). The CPP contains algorithms that are used to adjust the values developed for standard
conditions using SPARC to the specific temperature and pH needed. For example, the Aerated
Tank Module calls the CPP during each model realization to get property values adjusted for
6-1
-------
Section 6.0
Verification and Validation of Chemical Properties Models
monthly temperature changes within the tank. For organic chemicals that ionize in the
environment, the CPP uses SPARC-generated ionization coefficients (pKa) to adjust partition
coefficients to site-specific pH conditions in soil and aquifer materials.
6.1.1 Model Description
EPA developed the predictive modeling system SPARC to help meet the growing need
for chemical-specific inputs for risk assessment tools such as the 3MRA modeling system.
SPARC calculates values for a large number of physical and chemical parameters from
molecular structure and basic information about the environment (media, temperature, pressure,
pH, etc.). The ultimate goal for SPARC is to model the chemical and physical behavior of
molecules to predict chemical reactivity parameters and physical properties strictly from
molecular structure for the universe of organic molecules. Table 6-1 lists the properties
calculated using SPARC for the 3MRA modeling system, along with testing status and the
reaction conditions that must be specified to calculate a property. Detailed information on
SPARC can be found in U.S. EPA (2003a).
Table 6-1. SPARC Physical and Chemical Property Estimations used for
the 3MRA Modeling System
Property
Testing Status9
Reaction Conditions
Molecular weight
Yes
none
Density
Yes
Temperature
Volume
Yes
Temperature
Vapor Pressure
Yes
Temperature
Boiling Point
Yes
Pressure
Diffusion Coefficient in Air
Mixed
Temperature,
Pressure
Diffusion Coefficient in Water
Mixed
Temperature
Solubility
Yes
Temperature, Solvent
Henry's constant (gas/liquid partition)
Yes
Temperature, Solvent
Octanol/Water Partition (liquid/liquid partition)
Yes
Temperature, Solvent
Ionization phC, in water
Yes
Temperature, pH
a Testing status:
Yes: already tested and implemented in SPARC
Mixed: capability exists, more testing needed (automated and/or extended)
Mathematical models for predicting the transport and fate of contaminants in the
environment require reactivity parameter values; that is, the physical and chemical constants that
govern reactivity. Although empirical structure-activity relationships that allow estimation of
some constants have been available for many years, such relationships generally hold only
within very limited families of chemicals. SPARC avoids this limitation by predicting chemical
reactivity strictly from molecular structure for virtually all organic compounds, using
computational algorithms based on fundamental chemical structure theory to estimate values for
a large array of physical-chemical properties.
6-2
-------
Section 6.0
Verification and Validation of Chemical Properties Models
EPA has used SPARC for several years to estimate chemical and physical property
values for program offices (e.g., Office of Water, Office of Solid Waste and Emergency
Response, Office of Prevention, Pesticides and Toxic Substances) and EPA Regions. SPARC
has been used in other EPA modeling programs such as LENS3 (a multicomponent mass balance
model for application to oil spills) and by state agencies such as the Texas Natural Resource
Commission. The SPARC web-based calculators have been used by many employees of various
government agencies, academic institutions, and private chemical or pharmaceutical companies
throughout the United States. The web version of SPARC performs approximately 50,000 to
100,000 calculations each month (see U.S. EPA, 2003b, for summary web usage statistics).
6.1.2 Major Model Components/Functionality
SPARC analyzes chemical structure to answer a specific reactivity query in much the
same manner as an expert chemist would. Physical-organic chemists have established the types
of structural groups or atomic arrays that affect certain types of reactivity and have described, in
"mechanistic" terms, the effects on reactivity of other structural constituents appended to the site
of reaction. To encode this knowledge base, a classification scheme was developed in SPARC
that defines the role of structural constituents in affecting reactivity. Furthermore, models have
been developed that quantify the various "mechanistic" descriptions commonly used in
structure-activity analysis, such as induction, resonance, and field effects. SPARC execution
involves the classification of molecular structure (relative to a particular reactivity of interest)
and the selection and execution of appropriate "mechanistic" models to quantify reactivity.
The SPARC computational approach is based on blending well known, established
methods such as SAR (Structure Activity Relationships) (Lemer and Grundwald, 1965; Lowry
and Richardson, 1987), LFER (Linear Free Energy Relationships) (Taft, 1987; Hammett, 1970)
and PMO (Perturbed Molecular Orbital) theory (Dewar and Doughetry, 1975; Dewar, 1969).
SPARC uses SAR for structure activity analysis, such as induction and field effects. SPARC
uses LFER to estimate thermodynamic or thermal properties and PMO theory is used to describe
quantum effects, such as charge distribution derealization energy and polarizability of the p
electron network. In reality, every chemical property involves both quantum and thermal
contributions and necessarily requires the use of all three methods for prediction.
SPARC'S predictive methods were designed for engineering applications involving
physical-chemical process modeling. More specifically, these methods provide:
¦ An a priori estimate of the physical-chemical parameters of organic compounds
for physical and chemical fate process models when measured data are not
available;
¦ Guidelines for ranking a large number of chemical parameters and processes in
terms of relevance to the question at hand, thus establishing priorities for
measurements or study;
¦ An evaluation or screening mechanism for existing data based on "expected"
behavior; and
6-3
-------
Section 6.0
Verification and Validation of Chemical Properties Models
¦ Guidelines for interpreting or understanding existing data and observed
phenomena.
The basic mechanistic models in SPARC are designed and parameterized to be portable
to any type of chemistry or organic chemical structure. This portability affects system validation
and verification in several ways. As the diversity of structures and the chemistry that is
addressable increases, so does the opportunity for error. However, in verifying against the
theoretical knowledge of reactivity, specific situations can be chosen that offer specific
challenges. This is important when verifying or validating performance in areas where existing
data are limited or where additional data collection may be required. This expanded prediction
capability allows one to choose, for exhaustive validating, the reaction parameters for which
large and reliable data sets exist. The SPARC models have been validated on more than 10,000
data points.
Most models that predict a given physical-chemical property (e.g., solubility, boiling
point, etc.) are based directly on experimental data for that property for a limited training set of
chemicals. Model development involves finding the best correlations between various
descriptors of chemical structure and the observed property values. These descriptors are
subsequently used to construct a model that adequately "recalculates" the training (or
calibration) data set. To validate the model, one must demonstrate that the empirical model also
accurately predicts property values for chemicals not included in the training set, but whose
experimental values are known. These data are often called the validation set. In order to
predict a new physical-chemical property (e.g., octanol/water partition coefficient), the entire
process must be repeated, requiring new training and validation data sets for each new property.
With SPARC, experimental data for physical-chemical properties (such as boiling point)
are not used to develop (or directly impact) the model that calculates that particular property.
Instead, physical-chemical properties are predicted using models that quantify the underlying
phenomena that drive all types of chemical behavior (e.g., resonance, electrostatic forces,
induction, dispersion, H-bonding interactions, etc.). These mechanistic models were
parameterized using a very limited set of experimental data, but not data for the end-use
properties that will subsequently be predicted. After verification, the mechanistic models were
used in (or ported to) the various software modules that calculate the various end-use properties
(such as boiling point).
It is critical to recognize that the same mechanistic model (e.g., H-bonding model) will
appear in all of the software modules that predict the various end-use properties (e.g., boiling
point) for which that phenomenon is important. Thus, any comparison of SPARC-calculated
physical-chemical properties to an adequate experimental data set is a true model validation
test—there is no training (or calibration) data set in the traditional sense for that particular
property. The results of validation tests on the various SPARC property models are summarized
in Section 6.1.4 below and discussed in more detail (by each property) in U.S. EPA (2003b).
6-4
-------
Section 6.0
Verification and Validation of Chemical Properties Models
6.1.3 Summary of Model Verification
The unique approach to SPARC modeling impacts the strategy for model verification.
When a mechanistic model is updated or improved by incorporating new knowledge, the impact
on all of the various end-use parameters must be assessed. EPA has developed quality assurance
software that is executed each quarter on the current version of SPARC. This software runs the
various property modules for a large number of chemicals (4,200 data-point calculations) and
compares the results to historical results obtained over the life span of the SPARC program.1
Because all SPARC property modules are driven by the same verified mechanistic models, this
verification approach can be applied after each SPARC improvement to ensure that existing
parameter models still work correctly and that the computer codes for all property and
mechanistic models are fully operational.
6.1.4 Summary of Model Validation
Each SPARC calculation used to estimate a chemical or physical property has been
validated against numerous measured values for the property of interest. A detailed discussion
of each of these validation exercises, including graphical results for many, is provided in U.S.
EPA (2003b). The following examples illustrate the general processes for some of the more
important properties used in 3MRA.
6.1.4.1 Vapor Pressure. The vapor pressure computational algorithm output was
initially verified by comparing the SPARC prediction of the vapor pressure at 25° C to hand
calculations for key molecules. Because the SPARC self-interactions model was developed
initially on this property, the vapor pressure model undergoes the most frequent validation tests.
The calculator was trained on 315 nonpolar and polar organic compounds at 25° C. Figure 6-2
presents the SPARC-calculated vapor pressure at 25° C versus measured values for 747
compounds. The SPARC self-interactions model can predict the vapor pressure at 25° C within
experimental error over a wide range of molecular structures and measurements (over 8 log
units). For simple structures, SPARC can calculate the vapor pressure to better than a factor
of 2. For complex structures such as some pesticides and pharmaceutical drugs for which
dipole-dipole and/or hydrogen bond interactions are strong, SPARC calculates the vapor
pressure within a factor of 3 to 4.
6.1.4.2 Solubility. The solubility calculator spans more than 12 log mole fraction as
shown in Figure 6-3. The root mean square (RMS) deviation was 0.40 log mole fraction, which
was close to the experimental error. SPARC estimates the solubility for simple organic
molecules to better than a factor of 2 (0.3 log mole fraction) and within a factor of 4 (0.6 log
mole fraction) for complicated molecules like pesticides and pharmaceutical drugs. The RMS
deviation for the solids compounds is three times greater than the RMS deviation for liquid
compounds due to the crystal energy contributions.
1 During early stage development, the output of all SPARC modules was compared to hand calculations
with selected chemicals prior to proceeding with further development.
6-5
-------
Section 6.0
Verification and Validation of Chemical Properties Models
E
*•>
O)
o
T3
0)
*•>
TO
3
O
O
6
on
<
Q_
(/)
y = 0.9942X - 0.0117
-5 -3 -1
Observed (log atm)
Figure 6-2. SPARC-calculated vs. observed log vapor pressure for 747
organic molecules at 25° C. The figure includes all vapor
pressure measurements (real, not extrapolated) found in the
literature. The RMS deviation error was 0.15 log atm and R2
was 0.994.
0)
o
E
u>
o
¦a
0)
c
re .2
3
O
TO
o
6
on
<
Q.
W
O
<5
0
-5
y = 0.9944x-0.0431
-10
-15
-15
-10
-5
0
Observed (log mole fraction)
Figure 6-3. Test results for SPARC calculated log solubilities for 260
compounds. The RMS deviation is 0.321 and R2 is 0.991.
For the 119 liquid compounds, the RMS deviation is 0.135
and R2 is 0.997; for the 141 solid compounds, the RMS
deviation is 0.419 and R2 is 0.985.
6-6
-------
Section 6.0
Verification and Validation of Chemical Properties Models
6.1.4.3 Octanol/Water Partition Coefficient. The liquid/liquid partitioning models are the
most extensively tested partitioning models because a large octanol/water data set is available.
Figure 6-4 displays a comparison of the EPA Office of Water recommended observed octanol-water
distribution coefficients versus SPARC and ClogP calculated values. The RMS deviation and R2
values were is 0.18 and 0.996 respectively for SPARC and 0.44 and 0.978 respectively for ClogP
calculated values (Karickhoff and Long, 1995).
Observed log Kow
Figure 6-4. Test comparing calculated Kow with measured values. Squares are
SPARC calculated values, circles are ClogP calculated values. The RMS
deviation and R2 values were 0.18 and 0.996 respectively for SPARC and
0.44 and 0.978 respectively for ClogP calculated values.
6.1.4.4 Henry's Law Constant. The SPARC gas/liquid models have been extensively
tested against observed Henry's constant measurements. In a comparison between measured and.
SPARC-calculated Henry's law constants for 271 organic compounds, the RMS deviation was 0.1
and the R2 was 0.997.
6.2 MINTEQA2 Geochemical Speciation Model
This section documents the verification and validation activities for the MINTEQA2
geochemical speciation model. Although not incorporated into the 3MRA modeling system as a
module, the MINTEQA2 model is used to develop metal sorption isotherms that are used by the
Vadose Zone and Aquifer Modules to provide the pH and concentration- adjusted soil/water
partition coefficients needed to estimate sorption of metal contaminants in the subsurface. The
verification and validation activities described in this section are those conducted over the past
15 years for the MINTEQA2 model in general, but because they confirm the basic geochemical
functionality of this model, they are relevant to its application to the 3MRA modeling system.
6-7
-------
Section 6.0
Verification and Validation of Chemical Properties Models
6.2.1 Model Description
The metal soil/water partitioning coefficients used by the Vadose Zone and Aquifer Modules
are read from sorption isotherms created as a function of pH and metal concentration using the
MINTEQA2 model. MINTEQA2 is a equilibrium speciation model that can be used to calculate the
equilibrium composition of dilute aqueous solutions in the laboratory or in natural aqueous systems.
The model can calculate the equilibrium mass distribution among dissolved species, adsorbed
species, and multiple solid phases (see Figure 6-5) under a variety of conditions, including a gas
phase with constant partial pressure.
Total
Component
Concentrations
The original version of this model (called MINTEQ) was developed in the early 1980s at
Battelle Pacific Northwest Laboratory (Felmy
etal., 1984). The MINTEQ package was
delivered to EPA in 1985 and renamed
MINTEQA1, which was distributed with a
preprocessor program, PRODEFA1, for the
preparation of MINTEQA1 input files. After
more significant revisions were made in the
late 1980s, the name was changed to
MINTEQA2. With further development,
version numbers were used to indicate new
versions, and the model's formal name Rgure ^ MINTEQA2 computes the equilibrium
continues to be MINTEQA2. distribution of metals.
Disso ved
Adsorbed
Precipitated
6.2.2 Major Model Components/Functionality
The MINTEQA2 model includes a comprehensive database that is sufficient for solving a
broad range of problems without the need for additional user-supplied equilibrium constants. The
model uses a predefined set of components that include free ions (e.g., Na+) and neutral and charged
complexes (e.g., H4Si04°, Cr(OH)2+). The database of reactions is written in terms of these
components as reactants. The interactive preprocessor (PRODEFA2) produces the required
MINTEQA2 input files. As implemented to calculate sorption isotherms for the 3MRA modeling
system, MINTEQA2 includes reactions for adsorption of metal ions to hydrous iron oxide and
organic matter surfaces. With these reactions, the dimensionless partition coefficient can be
calculated from the ratio of the sorbed metal concentration to the dissolved metal concentration at
equilibrium. The dimensionless partition coefficient is converted to Kd with units of liters per
kilogram (L/kg) by normalizing by the mass of soil (in kilograms) with which one liter of solution is
equilibrated (the phase ratio). An isotherm is generated when the equilibrium metal distribution
between sorbed and dissolved fractions is estimated for a series of total metal concentrations. For
the 3MRA modeling system, isotherms were generated for a range of pH values and iron oxide and
organic matter conditions designed to capture the national variability of these parameters.
The model and modeling procedure used in estimating metal partition coefficients for the
3MRA Subsurface Module were similar to those described in U.S. EPA (1996e, 1998c), with
several improvements such as expanded and improved thermodynamic and sorption databases,
calculations of isotherms for additional metals, and adjustments to the geochemical modeling
conditions (U.S. EPA, 1999f).
6-8
-------
Section 6.0
Verification and Validation of Chemical Properties Models
6.2.3 Summary of Model Verification
Verification refers to tests and studies that, by design or otherwise, show that the
computations performed by the MINTEQA2 computer code representing a geochemical model are
true to the intent of the conceptual model. Verification tests determine whether the computer code,
compiled into an executable program, arrives at the intended and expected answer for a given set of
input values. For MINTEQA2, "input values" include total concentrations of components, the set of
equilibrium reactions and their thermodynamic constants, perhaps one or more imposed equilibrium
conditions, and the settings of various program flags and options such as ionic strength, method of
estimating activity coefficients, and system temperature. The answer obtained from the computer
code consists of all computed quantities including the equilibrium concentrations of all solution
species and the amounts of solid phases that have dissolved from an initially present solid or
precipitated from the solution. The computed answer also may include the ionic strength and
activity coefficients of solution species.
Verification can be achieved for individual algorithms that make up the entire computer
program. For MINTEQA2, it is impossible to test all program options and features in one program
execution. MINTEQA2 verification rests on comparison of the computed answer with hand
calculations or results from a similar computer code that has itself been verified. In either case, it is
necessary to use the same reactions and thermodynamic data when calculating the results from the
code to be verified as was used in calculating the results for the standard of comparison.
All parts of the MINTEQA2 code have been verified (U.S. EPA, 2002b). As a quality
assurance measure, EPA policy has required testing of all code modifications and additions to
MINTEQA1 and MINTEQA2 by a combination of compiler tests and model execution tests before
final adoption. The compiler tests required that the model be compiled using Fortran compilers
from multiple vendors and that the effect of various compiler options on execution time and
computed results be examined and accounted for. The execution tests for MINTEQA2 consisted of
a series of equilibrium problems for which the answer was known or could be computed by hand
calculations. This quality assurance requirement is a primary basis that supports the assertion that
MINTEQA2 calculations have been verified.
Early verification efforts compared results with those of similar (verified) models. Zachara
et al. (1988) stated that all major code algorithms, including calculation of mass balance, activity
coefficients, and equilibrium speciation, were verified by hand calculations during the model's
initial development. Also, speciation results from test executions, some involving adsorption
reactions, agreed with identical test calculations using other geochemical models (WATEQ3 and
MINEQL). Morrey et al. (1985) compared the results of several equilibrium speciation models,
including MINTEQ, and found that the results from these models are the same when identical
thermodynamic data and program options are used.
EPA performed additional verification tests when new algorithms were added to
MINTEQA2. For example, when the diffuse-layer (generalized two-layer) sorption model was
added in 1989, test problems presented in Dzombak (1987), including computer solutions (using
MICROQL; Westall, 1979) and hand calculated solutions, were used to verify the correct
implementation of the diffuse-layer model in MINTEQA2. In similar manner, the Gaussian
distribution model for computing the complexation of metals with organic matter was verified by
6-9
-------
Section 6.0
Verification and Validation of Chemical Properties Models
hand calculation and using a procedure described in Fish et al. (1986). Results from the formally
implemented Gaussian dissolved organic matter (DOM) model agreed with the "manually
implemented" Gaussian results, excepting a margin of error in the latter inversely proportional to the
number of ligands used to implement it (Allison and Perdue, 1994).
The organized public distribution of MINTEQA2 under EPA's Center for Exposure
Assessment Modeling (CEAM) provided a useful clearinghouse function for reporting suspected
errors in the code or in the thermodynamic database. Especially during the early years of its
distribution, many errors (especially in the preprocessor PRODEFA2) were discovered and
corrected. The confidence that can be placed in MINTEQA2 has been enhanced by its use by the
public and the public's feedback in reporting errors. Modifications made to correct errors were
verified prior to release in new versions.
Finally, the MINTEQA2 code has been verified through the use of the model in teaching
geochemistry and in textbooks. Several universities (e.g., Lindsay and Ajwa, 1995) and EPA's
MINTEQA2 workshops use MINTEQA2, where simple classroom problems in solution equilibria
and redox and sorption reactions use hand calculations to give the student a better appreciation of
the nuances of solving geochemical problems. Prior to the EPA workshops, MINTEQA2 results
were compared with those of MINEQL for the a set of ten instructional problems and were found to
agree. In addition, Drever (1997) used MINTEQA2 to illustrate the solution of problems in ground
water chemistry in The Geochemistry of Natural Waters.
6.2.4 Summary of Model Validation
Validation of MINTEQA2 can be accomplished by conducting tests and studies that show
that the geochemical model, implemented by the combination of the user's input parameters, the
thermodynamic database, and the computer code, provides an acceptable representation of reality or
that it produces an outcome that is an acceptable representation of reality. This presupposes that
there exists a measurement or group of measurements that may be taken as reality and that can be
used as the standard to which the model result is compared.
Validating geochemical models is difficult regardless of whether the model outcome is
compared with measurements on natural field systems or lab systems that mimic the natural
environment. Natural systems are replete with complicating factors that result in imprecise or
uncertain measurements and conditions that fail to correspond to the primary tenet of MINTEQA2-
based geochemical models: that the system reflects equilibrium conditions. Problems and issues
with measurements (analytical methods, sample handling, determining redox status); problems in
incomplete knowledge of the natural environment (true nature of sorption reactions, partial pressures
of gases, rates of reaction, degree of mediation by biota, kinetic effects); and the high degree of
variability in important chemical characteristics of natural systems all serve to complicate
comparisons of model systems with their real counterparts.
Because of these challenges, EPA convened a meeting of geochemists, soil scientists, and
other ground water professionals to discuss the best approach to validating MINTEQA2 for 3MRA.
Opinions ranged from those who believed that comparisons of MINTEQA2 predictions with
measurements made on closely controlled laboratory systems would provide the most relevant
validations to those who believed that laboratory systems could not faithfully represent the real
6-10
-------
Section 6.0
Verification and Validation of Chemical Properties Models
systems that are of interest and preferred a validation exercise closely tied to field sampling.
Statisticians pointed out that because of the natural variability in many important MINTEQA2 input
parameters, the model should be validated at a host of diverse sites. In the 14 years since that
workshop, numerous and varied studies have been performed that reflect these recommendations
and directly or indirectly relate to the validation of MINTEQA2.
The validation MINTEQA2 studies summarized in Table 6-2 show where MINTEQA2
model predictions have (and have not) been borne out by measurements in corresponding real
systems. Details on these studies and their selection and use for validation may be found in U.S.
EPA (2002b). Many of the studies in Table 6-2 were not undertaken specifically to provide
validation support for MINTEQA2; however, each study compares results calculated by
MINTEQA2 with some measure of reality, and studies were included regardless of whether the
comparison was favorable or not.
Table 6-2. Summary of MINTEQA2 Validation Studies
Study Citation
[Metals addressed]
Description of Validation
Results
Simple systems (no sorption or dissolved organic carbon [DOC] complexation)
Unpublished MINTEQA2
workshop problem
[PH]
Comparison of a pH curve from a leaching experiment
showing leachate pH versus concentration of acetic acid in
the leachant with similar curve computed by MINTEQA2
Validated for dolomite lime system tested
Frandsen and Gammons
(2000)
[Zn, Cu, Fe]
Comparison of dissolved metal concentrations predicted by
MINTEQA2 with measured values
Lower modeled dissolved metal
concentrations than measured attributed
to absence of quality metal-sulfide
complex data in model database
Marani et al. (1995)
[Pb]
Comparison of equilibrium mineral phases predicted by
MINTEQA2 with sample mineral phases identified by X-ray
diffraction.
Precipitates in keeping with observations
Fotovatand Naidu (1997)
[Cu, Zn]
Compares free Cu2+ and Zn2+ in solution as determined
using ion exchange procedures versus computed by
MINTEQA2
Speciation measurements and model
predictions in close agreement
Jensen et al. (1998)
[Fe(ll), Mn(ll)]
Compares speciation of Fe(ll) and Mn(ll) in solution as
determined using ion exchange procedures versus
computed by MINTEQA2
Modeled and measured fractions of Fe(ll)
and Mn(ll) were the same
Yu (1996)
[Fe, Al]
Comparison of solid phases predicted to precipitate by
MINTEQA2 versus solid phases identified in field samples
using X-ray diffraction and other analytical methods
MINTEQA2 predicted different iron and
aluminum precipitates than measured;
difference attributed to kinetic inhibitions in
the field
Palmer et al. (1998)
[Cu]
Comparison of Cu2+ activity measured using an ion-
selective electrode with Cu2+ activity computed by
MINTEQA2
Good agreement between modeled and
calculated Cu2+ concentrations for all test
solutions
Davis et al. (1992)
[As, Pb]
Comparison of MINTEQA2-calculated metal solubilities
with measured solubilities
Modeled solubilities compared well with
measurements
(continued)
6-11
-------
Section 6.0
Verification and Validation of Chemical Properties Models
Table 6-2. Continued
Study Citation
[Metals addressed]
Description of Validation
Results
Complex systems (with sorption or DOC compiexation)
Loux etal. (1989)
[Ba, Be, Cd, Cu, Ni, Pb, Tl,
Zn]
Comparison of fraction of metal remaining in solution in
batch equilibrium experiments using aquifer materials
versus fraction of metal dissolved at equilibrium as
calculated by MINTEQA2 using diffuse-layer sorption
model
Agreement between measured and
modeled results sufficient for a number of
elements; behavior of Ba, Be, Cu poorly
described by model (database
subsequently improved)
Jenne (1994)
[Ca, Fe, Mn, Al, Si]
Field study where solid phases predicted to exist at
equilibrium by MINTEQA2 were compared with solid
phases identified using analytical methods
Model predictions reasonably conformed
with observed mineral occurrence
Stollenwerk (1994)
[Al, Fe, Mn, Ca, Cu, Co, Ni,
Zn, pH, S04]
Comparison of dissolved concentrations of metals
measured in a series of wells with solution concentrations
predicted using MINTEQA2 with diffuse-layer model
Combination of dilution and adsorption
accurately simulated ground water
concentrations for several metals
Stollenwerk (1996)
[Al, Cu, Co, Ni, Zn]
Comparison of dissolved concentrations of metals
measured in a column experiment effluent with dissolved
concentrations predicted using MINTEQA2 with diffuse-
layer model
Model-predicted concentrations matched
experimental data reasonably well
Stollenwerk (1995)
[MoOJ
Comparison of dissolved concentration of molybdate in
column experiment effluent with dissolved concentration
predicted using MINTEQA2 with diffuse-layer model
Good match between measured and
modeled adsorption
Doyle etal. (1994)
[As(V)]
Comparison of dissolved concentrations of As in batch and
column tests with dissolved concentrations predicted by
MINTEQA2
Predicted As leachate concentrations
showed good agreement with measured
values
Saunders and Toran (1995)
[Co, Cd, Pb, Sr, U, Zn]
Comparison of dissolved concentrations of metals at
monitoring wells near a disposal pond with dissolved
concentrations predicted by MINTEQA2
Model predictions generally matched field
observations
Christensen and
Christensen (1999)
[Cd, Ni, Zn]
Concentrations of metal-DOC complexes determined in
batch experiments using a resin exchange method were
compared with concentrations of metal-DOC complexes
computed by MINTEQA2
Excellent agreement for Cd; fair
agreement forZn and one Ni sample;
underprediction for one Ni sample. Model
gives useful first approximation of Cd, Ni,
and Zn compiexation by DOC
Christensen et al. (1999)
[Cu, Pb]
Concentrations of metal-DOC complexes determined in
batch experiments using a resin exchange method were
compared with concentrations of metal-DOC complexes
computed by MINTEQA2
Agreement between predicted and
observed compiexation was generally
good
Christensen and
Christensen (2000)
[Cd, Ni, Zn]
Concentrations of metal-DOC complexes determined in
batch experiments using a resin exchange method were
compared with concentrations of metal-DOC complexes
computed by MINTEQA2 over a range of pH values
Poor match with experimental results
because model did not show appropriate
pH response; attributed to lack of lack of
phenolic sites in model representation of
DOC
Khoe and Sinclair (1991)
[Al, Fe, Ca, Mn, Si02, P04,
Pb, U]
Comparison of dissolved metal concentrations predicted by
MINTEQA2 versus concentrations measured in
neutralization experiments
Model predictions for pH, Al, Fe, Ca, Si,
and P04 agreed well with measured
values; agreement not as good for Mn due
to C02 equilibria
Webster and Webster
(1995)
[As(lll), As(V)]
Comparison of dissolved As concentrations measured in
batch experiments with dissolved concentrations predicted
by MINTEQA2 using the diffuse-layer model
Model overpredicted As adsorption;
subsequent study included silica
adsorption in model and obtained much
closer match with experimental results
Woodfine et al. (2000)
[Cu, Ni]
Comparison of Quantitative Water Air Sediment Interaction
(QWASI) model simulation results of average lake water
dissolved metal concentrations with observed values when
MINTEQA2-predicted partition coefficients are used
Model simulations were reasonable given
uncertainties in the input data
Routh and Ikramuddin
(1996)
[Pb, Zn]
Comparison of MINTEQA2-predicted equilibrium solid
phases with solid phases observed by X-ray diffraction and
comparison of predicted water concentrations with
observed values
Measurements confirmed model predicted
sold phases and dissolved concentrations
6-12
-------
Section 7.0
Summary and Conclusions
7.0 Summary and Conclusions
Volume III describes the verification and validation activities to ensure quality of the
3MRA modeling system modules, site-based data, and chemical properties data models. This
section briefly highlights the activities described in the various sections of this Volume.
The 3MRA modeling system science plan and the various science-based modules of the
3MRA modeling system were subjected to external peer reviews by national and international
experts in accordance with EPA's peer review policy. All 3MRA Modeling System modules
were subjected to extensive verification activities, which were documented in module-specific
test plans. Any time a module failed a particular verification test, the code was corrected and the
test redone until the module passed. All verification activities were completed successfully.
Table 7-1 summarizes the validation efforts previously conducted on several of the modules.
These modules demonstrate reasonable agreement with the field and/or laboratory data.
An extensive search for a data set was conducted to obtain appropriate multimedia
environmental data to validate the 3MRA Modeling System. However, in absence of such data, a
model comparison study is undertaken to compare the results of 3MRA model with those
obtained using another multimedia model (TRIM). A comparison of the 3MRA model results
with the multimedia monitoring data used for the TRIM comparison study are also underway.
These monitoring data are limited in nature; however, they are very useful for comparison
purposes in lack of a complete data-set.
The collection of the site-based data for the 3MRA modeling system representative
national data set was carried out under a formal data collection plan. QA/QC protocols
(including many automated checks) were followed. In addition, independent testing of major
data elements was performed. All verification activities for the site-based data collection were
completed successfully. Limited validation of the site-based data was performed for two sites
that were in both the 3MRA modeling system representative national data set and the Surface
Impoundment Study survey data set. Results were mixed. Many parameters demonstrated
reasonable agreement, but some did not (particularly flows). The sample of sites compared
(two) is not large enough to draw definitive conclusions.
All verification activities for the chemical properties estimation model (SPARC) and
speciation of metals model (MINTEQA2) have been completed successfully for numerous
contaminants. Both models continue to be updated and are re-verified whenever updates are
made. Table 7-2 summarizes the outcome of validation for SPARC and MINTEQ. Both models
demonstrate good agreement with data.
The 3MRA Modeling System and the modules were presented and discussed in two
separate national meetings: (1) Annual meeting of the Society for Risk Analysis (SRA); (2) and
7-1
-------
Section 7.0
Summary and Conclusions
the 20th Annual Meeting of the Society for Environmental Toxicology and Chemistry (SETAC).
The papers on the 3MRA methodology and the technology also are being published in
professional refereed journals. A list of professional publications and presentations on the
3MRA Modeling System may be found in Appendix F.
Table 7-1. Summary of Validation of 3MRA Modeling System Modules
Module
Validated?
Outcome/Comment
Wastewater Sources
(Aerated Tank and
Surface Impoundment)
yes
Based on validated models (CHEMDAT8 and EPACMTP); module compares well
to CHEMDAT8
Land-based Sources
(Landfill, Waste Pile, and
Land Application Unit)
and Watershed
yes
Some parts based on empirical models (implicitly validated)
Hydrology model validated using the HELP model; agreement ranged from very
close to moderately large differences; in cases of large differences, comparison to
data from the Water Atlas (Geraghty et al., 1973) tended to support the 3MRA
results
Monitored half-life data for dioxins demonstrated reasonable agreement with
predicted values from the LAU Module
Air (ISCST3)
yes
Several validation studies; demonstrated reasonable agreement for both
concentration and deposition
Surface Water (EXAMS)
yes
Numerous validation studies; demonstrated reasonable agreement for a variety of
contaminants and settings
Vadose Zone and Aquifer
(EPACMTP)
yes
Several validation studies; demonstrated reasonable agreement
SPARC
yes
Validated for all chemical properties with high correlation to observed data for
numerous contaminants.
MINTEQA2
yes
Numerous validation studies demonstrate reasonable agreement with field data for
a variety of contaminants.
7-2
-------
Section 8.0
References
8.0 References
Allison, J.D., and E.M. Perdue. 1994. Modeling metal-humic interactions with MINTEQA2. In
N. Senesi and T.M. Miano, editors, Humic Substances in the Global Environment and
Implications on Human Health. Elsevier Science BV, Amsterdam, Netherlands.
Armbrust, K.L., Y. Okamoto, J. Grochulska, and A.C. Barefoot. 1999. Predicting the
dissipation of bensulfuron methyl and azimsulfuron in rice paddies using the computer
model EXAMS2. J. Pesticide Sci. 24(4):357-363.
ASTM. 1996. Standard Guide for Developing and Evaluating Ground-Water Modeling Codes.
ASTM Subcommittee D-18.21.10, Tracking Number D18.21.92.07.
Bowers, J.F., and A.J. Anderson. 1981. An Evaluation Study for the Industrial Source Complex
(ISC) Dispersion Model. EPA-450/4-81-002. U.S. Environmental Protection Agency,
Research Triangle Park, NC.
Burns, L.A. 1997. Exposure Analysis Modeling System (EXAMS II): User's Guide for
Version 2.97.5. EPA-600/R-97/047. U.S. Environmental Protection Agency, Athens,
GA.
Burns, L. A., D.M. Cline, and R.R. Lassiter. 1982. Exposure Analysis Modeling System
(EXAMS): User Manual and System Documentation. EPA-600/3-83-023. U.S.
Environmental Protection Agency, Environmental Research Laboratory, Athens, GA.
Christensen, J.B., and T.H Christensen. 1999. Complexation of DOC in polluted groundwater:
A comparison of approaches using resin exchange, aquifer material sorption, and
computer speciation models (WHAM and MINTEQA2). Environmental Science &
Technology 33:3857-3863.
Christensen, J.B., and T.H Christensen. 2000. The effect of pH on the complexation of Cd, Ni,
and Zn by dissolved organic carbon from leachate-polluted groundwater. Water
Research 34:3743-3754.
Christensen, J.B., J.J. Botma, and T.H Christensen. 1999. Complexation of Cu and Pb by DOC
in polluted groundwater: A comparison of experimental data and predictions by
computer speciation models (WHAM and MINTEQA2). Water Research 33(15):3231-
3238.
Cousins, I.T., C.D. Watts, and R. Freestone. 1995. Field measurement and modelling the fate of
aniline and lindane in a UK lowland river. Environmental Technology 16:515-526.
8-1
-------
Section 8.0
References
Cox, W.M., and G.K. Moss. 1985. Evaluation of Rural Air Quality Simulation Models,
Addendum A: Muskingum River Data Base. EPA-450/83-003a. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
Cox, W.M., H.W. Rorex, and G.K. Moss. 1986. Evaluation of Rural Air Quality Simulation
models, Addendum C: KincaidSO2 Data Base. EPA-450/83- 003c. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
Cox, W.M., H.W. Rorex, G.K. Moss, and K.W. Baldridge. 1987. Evaluation of Rural Air
Quality Simulation models, Addendum D: Paradise S02 Data Base. EPA- 450/83-003d.
U.S. Environmental Protection Agency, Research Triangle Park, NC.
Davis, A., M.V. Ruby, and P.D. Bergstrom. 1992. Bioavailability of arsenic and lead in soils
from the Butte, Montana, mining district. Environmental Science & Technology
26(3):461-468.
Dewar, M.J.S. 1969. The Molecular Orbital Theory of Organic Chemistry. McGraw Hill, New
York, NY.
Dewar, M.J.S., and R.C. Doughetry. 1975. The PMO Theory of Organic Chemistry. Plenum
Press, New York, NY.
Doyle, T.A., A. Davis, and D.D. Runnels. 1994. Predicting the environmental stability of
treated copper smelter flue dust. Applied Geochemistry 9:337-350.
Drever, J.I. 1997. The Geochemistry of Natural Waters- Surface and Groundwater
Environments, 3rd ed. Prentice-Hall, Upper Saddle River, NJ.
Dzombak, D.A. 1987. A Practical Guide for Use of the Generalized Two Layer Sorption
Model. Prepared for the U.S. Environmental Protection Agency, Athens, GA.
ERG (Eastern Research Group, Inc.). 1999. Peer Review of EPA 's Hazardous Waste
Identification Rule Risk Assessment Model LAU/Waste Pile/Landfill Source Models and
Watershed Module Background and Implementation for the Multimedia, Multipathway,
and Multireceptor Risk Assessment (3MRA) for HWIR99 - Draft. Prepared for the U.S.
Environmental Protection Agency, Office of Solid Waste, under Contract No. 68-W-99-
001.
Felmy, A.R., D.C. Girvin, and E.A. Jenne. 1984. MINTEQ - A Computer Program for
Computing Aqueous Equilibria. Prepared by Battelle Pacific Northwest Laboratories for
the U.S. Environmental Protection Agency, Athens, GA.
Fish, W., D.A. Dzombak, and F.M.M. Morel. 1986. Metal-humate interactions, 2: Application
and comparison of models. Environmental Science & Technology 20:676-683.
Fotovat, A., and R. Naidu. 1997. Ion exchange resin and MINTEQa2 speciation of Zn and Cu
in alkaline sodic and acid soil extracts. Australian Journal of Soil Research 35:711-726.
8-2
-------
Section 8.0
References
Frandsen, A.K., and C.H. Gammons. 2000. Heavy metal complexation with aqueous sulfide in
an anaerobic treatment wetland. In Wetlands and Remediation: An International
Conference. Battelle Press, Columbus, OH.
Frind, E.O., and G.E. Hokkanen. 1987. Simulation of the Borden Plume Using the Alternating
Direction Galerkin Technique. Water Resources Res. 23:918-930.
Games, L.M. 1982. Field validation of Exposure Analysis Modeling System (EXAMS) in a
flowing stream. In K.L. Dickson, A.W. Maki, and J. Cairns, Jr., editors, Modeling the
Fate of Chemicals in the Aquatic Environment. Ann Arbor Science Publishers, Ann
Arbor, MI.
GCA Corporation. 1982. Evaluation and Selection of Models for Estimating Air Emissions from
Hazardous Waste Treatment, Storage and Disposal Facilities. Draft Final Report.
Prepared for Peer Review. Prepared for the U.S. Environmental Protection Agency,
Office of Solid Waste, Washington, DC, under Contract No. 68-02-3168.
Geraghty, J.J., D.W. Miller, F. Van Der Leeden, and F.L. Troise. 1973. Water Atlas of the
United States. Water Information Center Publication, Port Washington, NY.
Gobas, F.A. 1993. A model for predicting the bioaccumulation of hydrophobic organic
chemicals in aquatic food-webs: application to Lake Ontario. Ecological Modelling
69:1-17.
Hadermann, J. 1980. Radionuclide transport through heterogeneous media. Nuclear
Technology 47:311-323.
Hammett, L.P. 1970. Physical Organic Chemistry, 2nd ed. McGraw Hill, New York, NY.
Hargreaves, G.H. 1975. Moisture Availability and Crop Production. Trans. Am. Soc. Agric.
Eng. 18(5).
HGL (HydroGeoLogic, Inc.). 1993. Response to Review Comments on EPACMTP Unsaturated
Zone and Saturated Zone Modules by TetraTech, Inc. HGL, Herndon, VA.
HGL (HydroGeoLogic, Inc.). 2000. Implementation of the Test Plan for the HWIR99 Land
Application Unit Module. Prepared for the U.S. Environmental Protection Agency,
Office of Solid Waste, Washington, DC, under Contract No. 68-W7-0035.
HGL (HydroGeoLogic, Inc.). 2003. Verification and Validation of the EPA 's Composite Model
for Transformation Products (EPACMTP), and its Derivatives. Prepared for the U.S.
Environmental Protection Agency, Office of Solid Waste, Washington, DC.
Hodgkinson, D.P., andP.R. Maul. 1985. One-dimensional Modeling of Radionuclide Migration
Through Permeable and Fractured Rock for Arbitrary Length Decay Chains using
Numerical Inversion of Laplace Transforms. Rep. AERE-R-11880, UK Atomic Energy
Authority, Oxfordshire, England.
8-3
-------
Section 8.0
References
Horst, T.W. 1983. A correction to the Gaussian source-depletion model. In H.R. Pruppacher,
R.G. Semonin, and W.G.N. Slinn, editors, Precipitation Scavenging, Dry Deposition and
Resuspension. Elsevier, New York, NY.
Irwin, J.S. 2002. A Historical Look at the Development of Regulatory Air Quality Models for
the United States Environmental Protection Agency. NOAA Technical Memorandum,
OAR ARL-244. National Oceanic and Atmospheric Administration, Silver Spring, MD.
Jenne, E. A. 1994. Geochemical Model "Validation Reliability of Solubility Equilibria
Calculated with Field Data from and Acidic Metal-rich Plume near Globe, Arizona.
Prepared by Battelle Pacific Northwest Laboratories for the U.S. Environmental
Protection Agency, Athens, GA.
Jensen, D.L., J.K. Boddum, S. Redemann, and T.H. Christensen. 1998. Speciation of dissolved
iron(II) and manganese(II) in a groundwater pollution plume. Environmental Science &
Technology 32(18):2657-2664.
Karickhoff, S.W., and M. Long. 1995. U.S. Environmental Protection Agency Internal Report,
April 10.
Khoe, G.H., and G. Sinclair. 1991. Chemical modeling of the neutralisation process for acid
uranium mill tailings. Proceedings, Hydrometallurgy and Aqueous Processing
Symposium. Annual Meeting of the Metallurgical Society of AIME, New Orleans, LA.
Kolset, K., and A. Heiberg. 1988. Evaluation of the fugacity (FEQUM) and the EXAMS
chemical fate and transport models: a case study on the pollution of the Norrsundet Bay
(Sweden). Wat. Sci. Tech. 20(2): 1-12.
Lemer, J.E., and E. Grunwald. 1965. Rates of Equilibria of Organic Reactions. John Wiley and
Sons, New York, NY.
Lindsay, W. L., and H.A. Ajwa. 1995. Use of MINTEQA2 for Teaching Soil Chemistry. In
R.H. Loeppert, A.P. Schwab, and S. Goldberg, editors, Chemical Equilibrium and
Reaction Models. Soil Science Society of America, Special Publication 42, Madison,
WI.
Little, K., and J. Baskir. 2000a. Response to Independent Reviewer Comments on LAUModule
Verification Testing. Memorandum prepared for the U.S. Environmental Protection
Agency, Office of Solid Waste, Washington, DC, under Contract No. 68-W6-0053.
Little, K., and J. Baskir. 2000b. Response to Independent Reviewer Comments on Landfill
Module Verification Testing. Memorandum prepared for the U.S. Environmental
Protection Agency, Office of Solid Waste, Washington, DC, under Contract No. 68-W6-
0053.
Little, K., and J. Baskir. 2000c. Response to Independent Reviewer Comments on Waste Pile
Module Verification Testing. Memorandum prepared for the U.S. Environmental
8-4
-------
Section 8.0
References
Protection Agency, Office of Solid Waste, Washington, DC, under Contract No. 68-W6-
0053.
Little K., and J. Baskir. 2000d. Response to Independent Reviewer Comments on Regional
Watershed Module Verification Testing. Memorandum prepared for the U.S.
Environmental Protection Agency, Office of Solid Waste, Washington, DC, under
Contract No. 68-W6-0053.
Londergan, R.J., D.H. Minott, D.J. Wackter, and D. Bonitta. 1982. Evaluation of Rural Air
Quality Simulation Models. EPA-450/4-83-003. U.S. Environmental Protection Agency,
Research Triangle Park, NC.
Loux, N.T., D.S. Brown, C.R. Chapin, J.D. Allison, and S.M. Hassan. 1989. Chemical
speciation and competitive cationic partitioning on a sandy aquifer material. Chemical
Speciation and Bioavailability 1:111-125.
Lowry, T.H., and K.S. Richardson. 1987. Mechanism and Theory in Organic Chemistry, 3rd ed.
Harper and Row, New York, NY.
Marin, C.M., V. Guvanasen, and Z.A Saleem. 2003. The 3MRA risk assessment framework - a
flexible approach for performing multimedia, multipathway, and multireceptor risk
assessments under uncertainty. International Journal of Human and Ecological Risk
Assessment (in press; scheduled for publication December 2003).
Marani, D., G. Macchi, and M. Pagano. 1995. Lead precipitation in the presence of sulphate
and carbonate: Testing of thermodynamic predictions. Water Research 29(4): 1085-1092.
Maul, P. R. 1980. Atmospheric Transport of Sulphur Compound Pollutants.
MID/SSD/80/0026/R. Central Electricity Generating Board, Nottingham, England.
Morrey, J.R., C.T. Kincaid, and C.J. Hostetler. 1985. Geochemical Models for Solute
Migration, Volume 3: Evaluation of Selected Computer Codes for Modeling Aqueous
Solutions and Solute Migration in Soils and Geologic Media. EA-3417. Battelle Pacific
Northwest Laboratories, Richland, WA.
Ogata, A. 1970. Theory of Dispersion in a Granular Medium. U.S. Geological Survey
Professional Paper 411-1:1-34.
Ogata, A., and R.B. Banks. 1961. A Solution of the Differential Equation of Longitudinal
Dispersion in Porous Media. U.S. Geological Survey Professional Paper 411-A.
Palmer, F.B, C.A. Butler, M.H. Timperley, and C.W. Evans. 1998. Toxicity to embryo and
adult zebrafish of copper complexes with two malonic acids as models for dissolved
organic matter. Environmental Toxicology and Chemistry 17(8):1538-1545.
8-5
-------
Section 8.0
References
PNNL (Pacific Northwest National Laboratory). 2000. Independent Tests for the Watershed
Module. Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Pollard, J.E., and S.C. Hern. 1985. A field test of the Exams model in the Monongahela River.
Environmental Toxicology and Chemistry 4:361-369.
Research Triangle Institute (RTI). 1985. Preliminary Assessment of Aerated Waste Treatment
Systems at Hazardous Waste Treatment, Storage and Disposal Facilities. Final Report.
Prepared for the U.S. Environmental Protection Agency, Office of Research and
Development, Cincinnati, OH, under Contract No. 68-02-3992.
Research Triangle Institute (RTI). 1998. QA/QC Summary for Wastewater Source Modules
(Tanks and Surface Impoundments). Prepared for the U. S. Environmental Protection
Agency, Office of Solid Waste, Washington, DC, under Contract No. 68-W-98-085.
Research Triangle Institute (RTI). 2000a. Independent Data Testing for the Hazardous Waste
Identification Rule—Draft. Prepared for the U.S. Environmental Protection Agency,
Office of Solid Waste, Washington, DC.
Research Triangle Institute (RTI). 2000b. Land Application Unit Module Verification Testing.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC, under Contract No. 68-W6-0053.
Research Triangle Institute (RTI). 2000c. Land Application Unit Module Verification Testing:
ERRATA. Prepared for the U.S. Environmental Protection Agency, Office of Solid
Waste, Washington, DC, under Contract No. 68-W6-0053.
Research Triangle Institute (RTI). 2000d. Landfill Module Verification Testing. Prepared for
the U.S. Environmental Protection Agency, Office of Solid Waste, Washington, DC,
under Contract No. 68-W6-0053.
Research Triangle Institute (RTI). 2000e. Test Plan for the HWIR Human Exposure Module.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2000f. Wastepile Module Verification Testing. Prepared for
the U.S. Environmental Protection Agency, Office of Solid Waste, Washington, DC,
under Contract No. 68-W6-0053.
Research Triangle Institute (RTI). 2000g. Watershed Module Internal Verification Testing.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC, under Contract No. 68-W6-0053.
Research Triangle Institute (RTI). 2001. Exposure Analysis for Dioxins, Dibenzofurans, and
Co-Planar Poly chlorinated Biphenyls in Sewage Sludge: Preliminary Draft. Prepared
for the U.S. Environmental Protection Agency, Office of Water, Washington, DC.
8-6
-------
Section 8.0
References
Research Triangle Institute (RTI). 2002a. 3MRA Aerated Tank Module Test Documentation.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002b. 3MRA Aerated Tank Module Test Plan. Prepared for
the U.S. Environmental Protection Agency, Office of Solid Waste, Washington, DC.
Research Triangle Institute (RTI). 2002c. 3MRA Surface Impoundment Module Test
Documentation. Prepared for the U.S. Environmental Protection Agency, Office of Solid
Waste, Washington, DC.
Research Triangle Institute (RTI). 2002d. 3MRA Surface Impoundment Module Test Plan.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002e. Aquatic Food Web (AFW) Module—Test Plan.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002f. Aquatic Food Web Internal Verification Testing.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002g. Ecological Exposure (EcoEx) Module—Test Plan.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002h. Ecological Exposure Module Internal Verification
Testing. Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002i. Ecological Risk Module Internal Verification Testing.
Prepared for U.S. Environmental Protection Agency, Office of Solid Waste, Washington,
DC.
Research Triangle Institute (RTI). 2002j. Ecological Risk (Ecological Risk) Module—Test Plan.
Prepared for U.S. Environmental Protection Agency, Office of Solid Waste, Washington,
DC.
Research Triangle Institute (RTI). 2002k. Farm Food Chain (FFC) Module—Test Plan.
Prepared for U.S. Environmental Protection Agency, Office of Solid Waste, Washington,
DC.
Research Triangle Institute (RTI). 2002m. Farm Food Chain Module Internal Verification
Testing. Prepared for U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
8-7
-------
Section 8.0
References
Research Triangle Institute (RTI). 2002n. Terrestrial Food Web Internal Verification Testing.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Research Triangle Institute (RTI). 2002p. Terrestrial Food Web (TFW) Module—Test Plan.
Prepared for the U.S. Environmental Protection Agency, Office of Solid Waste,
Washington, DC.
Routh, J., and M. Ikramuddin. 1996. Trace element geochemistry of Onion Creek near Van
Stone lead-zinc mine (Washington, USA) - chemical analysis and geochemical
modeling. Chemical Geology 133:211-224.
Sample, B.E., M.S. Alpin, R.A. Effroymson, G.W. Suter, and C.J.E. Welsh. 1997. Methods and
Tools for Estimation of the Exposure of Terrestrial Wildlife to Contaminants.
Environmental Sciences Division Publication Number 4650, Oak Ridge National
Laboratory, Oak Ridge, TN.
Sample, B.E., J.J. Beauchamp, R.A. Efroymson, and G.W. Suter, II. 1998. Development and
Validation of Bioaccumulation Models for Small Mammals. Prepared for the U.S.
Department of Energy, under Contract No. DE-AC05-840R21400.
Saunders, J. A., and L.E. Toran. 1995. Modeling of radionuclide and heavy metal sorption
around low- and high-pH waste disposal sites at Oak Ridge, Tennessee. Applied
Geochemistry 10:673-684.
Schramm, K.-W., M. Hirsch, R. Twele, and O. Eutzinger. 1988. Measured and modeled fate of
Disperse Yellow 42 in an outdoor pond. Chemosphere 17(3):587-595.
Schroeder, K., R. Clickner, and E. Miller. 1987. Screening Survey of Industrial Subtitle D
Establishments. Draft Final Report. Preapred for the U.S. Environmental Protection
Agency, Office of Solid Waste, Washington, DC.
Schwede, D.B., and J.O. Paumier. 1997. Sensitivity of the Industrial Source Complex model to
input deposition parameters. J. Applied Meteorology 36:1096-1106.
Shamir, U.Y., and D.R.F. Harlemann. 1967. Dispersion in layered porous media. J. of the
Hydraulics Division, Amer. Soc. Civ. Engr. 93(5):237-260.
Snyder, S. 2000. HWIR Farm Food Chain—Review of Compiled Code Executables and Review
of the Implementation of the Test Plan. Memorandum from Sandra Snyder to Gene
Whelan, June 8.
Springer, C., P.D. Lunney, and K.T. Valsaraj. 1984. Emission of Hazardous Chemicals from
Surface and Near Surface Impoundments to Air. U. S. Environmental Protection Agency,
Solid and Hazardous Waste Research Division, Cincinnati, OH. Project Number
808161-02.
-------
Section 8.0
References
Stollenwerk, K.G. 1994. Geochemical interactions between constituents in acidic groundwater
and alluvium in an aquifer near Globe, Arizona. Applied Geochemistry 9:353-369.
Stollenwerk, K.G. 1995. Modeling the effects of variable groundwater chemistry on the
adsorption of molybdate. Water Resources Research 31(2):347-357.
Stollenwerk, K.G. 1996. Simulations of reactions affecting transport of constituents in the
acidic plume, Pinal Creek basin, Arizona, U.S. Geological Survey Water-Supply Paper
2466:21-49.
Strimaitis, D.G., J.C. Chang, and J.S. Scire. 1993. Development of ISC-COMPDEP and User
Instructions. Sigma Research Corporation, Concord, MA.
Sudicky, E.A., Y.S. Wu, and Z. Saleem. 1991. Semi-analytical Approach for Simulating
Transport of a Seven-Member Branched Decay Chain in 3D Groundwater Systems.
American Geophysical Union Spring meeting, Baltimore, MD.
Taft, R.W. 1987. Progress in Organic Chemistry, Vol.16. John Wiley and Sons, New York,
NY.
Tetra Tech. 2000a. Independent Tests for Aquatic Food Web Module. Memorandum from
Tetra Tech to RTI. June 5.
Tetra Tech. 2000b. Independent Tests for Ecological Exposure Module. Memorandum from
Tetra Tech to RTI. June 8.
Tetra Tech. 2000c. Independent Tests for Ecological Risk Module. Memorandum from Tetra
Tech to RTI. June 5.
Tetra Tech. 2000d. Independent Tests for Landfill Module. Prepared for the U. S.
Environmental Protection Agency, Office of Solid Waste, Washington, DC, under
Contract No. 68-W6-0053.
Tetra Tech. 2000e. Independent Tests for Terrestrial Food Web Module. Memorandum from
Tetra Tech to RTI. June 5.
Tetra Tech. 2000f. Independent Tests for Waste Pile Module. Prepared for the U.S.
Environmental Protection Agency, Office of Solid Waste, Washington, DC, under
Contract No. 68-W6-0053.
Tynan, P., C.D. Watts, A. Sowray, and I. Hammond. 1991. The transport and the fate of organic
pollutants in rivers II - Field measurement and modelling for styrene, xylenes,
dichlorobenzenes and 4-phenyldodecane. In G. Angletti and A. Bjorseth A., editors,
Proceedings of the Sixth European Symposium Organised within the Framework of
Concerted Action Organic Micropollutants in the Aquatic Environment. Cost 641
Working Party 2, Lisbon, Portugal.
8-9
-------
Section 8.0
References
U.S. EPA (Environmental Protection Agency). 1977. User's Manual for Single-Source
(CRSTER) Model. EPA-450/2-77-013. Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1989. Review and Evaluation of Area Source
Dispersion Algorithms for Emission Sources at Superfund Sites. EPA-450/4-89-020.
Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1990. Methodology for Assessing Health Risks
Associated with Indirect Exposure to Combustor Emissions. EPA-600/6-90-003. Office
of Health and Environmental Assessment, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1991. Hazardous Waste TSDF - Background
Information for Proposed RCRA Air Emission Standards. EPA-450/3-89-023a,
Appendix C, and EPA-450/3-89-023b, Appendix F. Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1992a. Summary of Quality Assurance and
Equivalence Tests Performed on the ISC2 Models. Project report for WA No. 68D00124.
Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1992b. Sensitivity Analysis of a Revised Area
Source Algorithm for the Industrial Source Complex Short Term Model. EPA-454/R-92-
015. Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1992c. Summary of the Quality Assurance and
Equivalence Tests Performed on the Modified Area Source Algorithm for the ISCST2
Model. Project report for WA No. 1-27, 68D00124. Research Triangle Park, NC.
U. S. EPA (Environmental Protection Agency). 1992d. Comparison of a Revised Area Source
Algorithm for the Industrial Source Complex Short Term Model and Wind Tunnel Data.
EPA-454/R-92-014. Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1992e. Memorandum dated November 6, 1992
from the Office of Research and Development, Athens, GA, to the Office of Solid
Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1994a. Air Emission Models for Waste and
Wastewater. EPA-453/R-94-080A. Office of Air Quality Planning and Standards,
Research Triangle Park, NC. CHEMDAT8 program can be downloaded at
http ://www. epa.gov/ttn/chief/software/water/water8. html.
U.S. EPA (Environmental Protection Agency). 1994b. Development and Testing of a Dry
Deposition Algorithm (Revised). EPA-454/R-94-015. Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency). 1995. User's Guide for the Industrial Source
Complex (ISC 3) Dispersion Models. Volume II: Description of Model Algorithms. EPA-
8-10
-------
Section 8.0
References
454/B-95-003b. Office of Air Quality Planning and Standards, Research Triangle Park,
NC.
U.S. EPA (Environmental Protection Agency). 1996a. EPACMTP Sensitivity Analysis. Office
of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1996b. EPA's Composite Model for Leachate
Migration with Transformation Products (EPACMTP): Background Document. Office
of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1996c. EPA 's Composite Model for Leachate
Migration with Transformation Products, EPACMTP, Background Document for Finite
Source Methodology for Chemicals with Transformation Products. R09-96-588. Office
of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1996d. Modeling approach for simulating three-
dimensional migration of land disposal leachate with transformation products. Volume I:
Background document for the unsaturated zone and saturated zone modules. Office of
Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1996e. Background Document for Metals; EPA
Composite Model for Leachate Migration with Transformation Products (EPACMTP).
Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1997a. EPA 's Composite Model for Leachate
Migration with Transformation Products (EPACMTP): User's Guide. Office of Solid
Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1997b. Test and Verification of EPA's
Composite Model for Leachate Migration with Transformation Products (EPACMTP).
Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1998a. Guidelines for Ecological Risk
Assessment. EPA/630/R-95/002F. Risk Assessment Forum, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1998b. Methodology for Assessing Health Risks
Associated with Multiple Exposure Pathways to Combustor Emissions. Update to
EPA-600/6-90-003 Methodology for Assessing Health Risks Associated with Indirect
Exposure to Combustor Emissions. EPA-600/R-98-137. National Center for
Environmental Assessment, Cincinnati, OH.
U.S. EPA (Environmental Protection Agency). 1998c. MINTEQA2/PRODEFA2, A
Geochemical Assessment Model for Environmental Systems: User Manual Supplement
for Version 4.0. Prepared by HydroGeoLogic, Inc., under Contract No. 68-C6-0020.
8-11
-------
Section 8.0
References
U.S. EPA (Environmental Protection Agency). 1998d. Testing of the Sampled Chronological
Input Model (SCIM) option in the enhanced ISCST3 Model for Use in the Hazardous
Waste Identification Rule (HWIR99). Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999a. Risk Characterization Report for the
HWIR 99 Multimedia, Multipathway, andMultireceptor Risk Assessment (3MRA).
Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999b. Verification Document for HWIR99
Pseudo-Three Dimensional Aquifer Module. Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999c. Verification Document for HWIR99
Vadose-Zone Module. Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999d. Vadose and Saturated Modules
Extractedfrom EPACMTP for HWIR99. Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999e. Chemical Data Base for HWIR99. Office
of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999f. Changes in theMINTEQA2Modeling
Procedure for Estimating Metal Partition Coefficients in Groundwater. Prepared by
HydroGeoLogic, Inc., Herndon, VA.
U.S. EPA (Environmental Protection Agency). 1999g. A Framework for Finite-Source
Multimedia, Multipathway, and Multireceptor Risk Assessment (3MRA), DRAFT Report.
Office of Solid Waste, Washington, DC.
U. S. EPA (Environmental Protection Agency). 2000a. HWIR Human Exposure Module Internal
Verification Testing. Prepared by Research Triangle Institute, Research Triangle Park,
NC.
U.S. EPA (Environmental Protection Agency). 2000b. Peer Review of EPA's Hazardous Waste
Identification Rule Risk Assessment Model—Background Document for the Human
Exposure and Human Risk Modules for the Multimedia, Multipathway, and Multiple
Receptor Risk Assessment (3MRA) Model. Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 2000c. Test Plan for HWIR99 Vadose-Zone.
Office of Solid Waste, Washington, DC.
U. S. EPA (Environmental Protection Agency). 2000d. Test Plan for HWIR99 Aquifer Module.
Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 2000e. Draft Test Plan for the HWIR Human Risk
Module. National Exposure Research Laboratory, Athens, GA. August.
8-12
-------
Section 8.0
References
U.S. EPA (Environmental Protection Agency). 2001. Industrial Surface Impoundments in the
United States. EPA-530/R-01-005. Office of Solid Waste, Washington, DC.
U.S. EPA (Environmental Protection Agency). 2002a. Quality Assurance Verification and
Validation Tests for the Exposure Analysis Modeling System - EXAMS. National
Exposure Research Laboratory, Athens, GA.
U.S. EPA (Environmental Protection Agency). 2002b. Review of Validation Studies
Concerning the U.S. EPA Geochemical Speciation ModelMINTEQA2. National
Exposure Research Laboratory, Athens, GA.
U.S. EPA (Environmental Protection Agency). 2003a. Prediction of Chemical Reactivity
Parameters and Physical Properties of Organic Compounds from Molecular Structure
Using SPARC. Internal Report. National Exposure Research Laboratory, Athens, GA.
U.S. EPA (Environmental Protection Agency). 2003b. Verification and Validation of the
SPARC Model. Internal Report. National Exposure Research Laboratory, Athens, GA.
USDA (U.S. Department of Agriculture). 1986. Urban Hydrology for Small Watersheds.
TR-55. Soil Conservation Service, Washington, DC.
Van Genuchten, M.T., and W.J. Alves. 1982. Analytical Solutions of the One-Dimensional
Convective-Dispersion Solute Transport Equation. U.S. Technical Bulletin 1661:151.
Venkatram, A. 1998. A Simple Model for Dry Deposition and Particle Settling. Subcontractor
Progress Report 2 (including addendum). EPA Contract No. 68D70002, Work
Assignment No. 1-001.
Webster, J.G., and K.S. Webster. 1995. Arsenic adsorption from geothermal water. In
Sixteenth Annual PNOC Geothermal Workshop Proceedings. Philippine National Oil
Company, Manila, Philippines.
Westall, J.C. 1979. MICROQLII. Computation of Adsorption Equilibria in BASIC. Swiss
Federal Institute of Technology, EAWAG, Deubendorf, Switzerland.
Woodfine, D.G., R. Seth, D. Mackay, and M, Havas. 2000. Simulating the response of metal
contaminated lakes to reductions in atmospheric loading using a modified QWASI
model. Chemosphere 41:1377-1388.
Woodrow, J.E., M.M. McChesney, and J.N. Seiber. 1990. Modeling the volatilization of
pesticides and their distribution in the atmosphere. In D.E. Kurtz, editor, Long Range
Transport of Pesticides. Lewis Publishers, Inc., Chelsea, Michigan.
Yu, J.Y. 1996. Precipitation of Fe and A1 compounds from the acid mine waters in Dogyae
area, Korea: A qualitative measure of equilibrium modeling applicability and
neutralization capacity. Aquatic Geochemistry 2:81-105.
8-13
-------
Section 8.0 References
Zachara, J.M., R.L. Schmidt, and E.A. Jenne. 1988. Feasibility of Field Testing the MINTEQ
Geochemical Code. Prepared for the U.S. Environmental Protection Agency, by Battelle
Pacific Northwest Laboratories, Richland, WA.
8-14
-------
Appendix A
Verification and Validation of the Air Module
-------
This page intentionally left blank.
-------
Appendix A
Verification and Validation of the Air Module
Appendix A
Verification and Validation of the Air Module
A.l Overview of air module
The purpose of the atmospheric modeling for the Multimedia, Mulipathway and
Multireceptor Risk Assessment (3MRA) is to provide an annual average estimate of air
concentration of dispersed constituents and annual deposition rate estimates for vapors and
particles at various receptor points in the area of interest, which is defined by a 2-km radius
measured from the edge of the largest area source at the site. The chemical constituents are
assumed to be in the form of volatilized gases or fugitive dust emitted from area sources. The
atmospheric module simulates the transport and diffusion of the constituent. The simulated air
concentrations are used to estimate bio-uptake from plants and human exposures due to direct
inhalation. The predicted deposition rates are used to determine chemical loadings to watershed
soils, farm crop areas, and surface waters.
The atmospheric concentration and deposition of constituents can be determined in
several ways. The initial design of the air module was strongly influenced by the needs and
constraints of the proposed HWIR application of 3MRA. The requirements for the air model
included: 1) able to produce estimates of concentration, wet deposition, and dry deposition for
particles and gases at distances within 10 km of the source for a variety of time scales; 2) able to
model rectangular area sources to be consistent with other media models being considered for
inclusion; 3) able to use readily available meteorological data for sites across the nation; and 4)
short runtime. After consideration of several models, we selected the Industrial Source
Complex-Short Term (ISCST3) model, a steady-state Gaussian plume model. ISCST3 fit most
of the design criteria with two notable exceptions: 1) the gas deposition algorithms were not
appropriate for most of the chemicals to be modeled and 2) the runtime for modeling area source
was longer than specified, particularly when including the effects of plume depletion due to dry
deposition. We opted to use bio-transfer factors to account for the exchange of gases between
the air and other media. To address the runtime issue, we embarked on a series of investigations
into ways to develop options for reducing the runtime. These are described in more detail in
Section 2.0.
ISCST3 is used as "legacy code" in the 3MRA framework. That is, the model is left
intact and the necessary interfacing to the framework is handled using pre- and postprocessors.
Together, the EPA air quality model (ISCST3) and the pre- and postprocessing code that
integrates ISCST3 into the 3MRA environment are referred to as the Air Module. The pre- and
postprocessing code also provides additional functionality to support other 3MRA framework
requirements. For example, the preprocessor determines whether to execute ISCST3 to model
the site-specific x-y coordinates requested by the 3MRA framework or, alternatively, to model to
A-3
-------
Appendix A
Verification and Validation of the Air Module
a fixed set of polar coordinates and then use a two-dimensional cubic spline method to
interpolate from this polar set to the x-y coordinates of interest.
A.2 Verification History of ISCST3
The ISCST model has had a long history of use, dating back to 1979 (Bowers et al,
1979), by the EPA for modeling sources of nonreactive pollutants. Over the years, ISCST has
undergone a number of updates as needed to address particular applications. A history of the
development is detailed in Irwin (2002). This section details much of the verification efforts that
have occurred during the continued model development to show that the FORTRAN program
accurately solves the intended equations. A significant overhaul of the program occurred in
1992 when the code was modified to produce the ISCST2 model. During this effort, a
substantial amount of recoding was done to make the code more modular. A detailed test plan
was followed to assure that the re-coded model produced results equivalent to the original model
(U.S. EPA, 1992a).
There have been several components of the ISCST2 model that have been revised since
1992. In response to concerns related to emission sources at Superfund sites, a study was
undertaken to compare area source algorithms used by various dispersion models (U.S. EPA,
1989). The results of the study showed that the finite line source approach used in ISC to model
area sources predicted unrealistic concentrations for receptors located within and near the area
source. A new area source algorithm which uses an integrated line source algorithm was
implemented in ISCST2 and tested in 1992 (U.S. EPA, 1992b; U.S. EPA 1992c) and released to
the public in draft form as AREA-ST. The need for better estimates of the intermedia transfer of
pollutants from the atmosphere to land, water and vegetation to support environmental impact
analyses prompted a study to compare existing algorithms for calculating the dry deposition
velocity of particles (U.S. EPA, 1994). The algorithm in the original ISCST2 model was not
designed for small particles. An additional component of this study was to compare plume
depletion algorithms and propose a revised algorithm. A new deposition velocity algorithm and
plume depletion algorithm were proposed and implemented into ISCST2 and released to the
public in draft form as DEPST. In 1993, the EPA Administrator announced that risk
assessments, including indirect exposure, would be required for permitting hazardous waste
incinerators and industrial furnaces. Since there were no regulatory models capable of
quantifying wet and dry deposition in all terrain, EPA Region 5 sponsored further development
of the ISCST2 model to address their immediate need for performing an assessment as well as
the more general need of the Agency. In this effort, the AREA-ST and DEPST versions of the
ISCST2 model were combined into a single version referred to as ISC-COMPDEP. Testing was
performed to demonstrate equivalence between ISC-COMPDEP and AREA-ST and DEPST.
The methodology used in the COMPLEX I model for modeling point sources in complex terrain
was added to ISC-COMPDEP with equivalence tests begin performed against the original
COMPLEX I model. Finally, wet deposition and depletion algorithms were selected and
implemented in ISC-COMPDEP. The development and testing of the ISC-COMPDEP model
including the equivalency test is documented in Strimaitis et al (1993). ISC-COMPDEP was
provided to EPA's Office of Air Quality Planning and Standards, renamed ISCSTDFT, and
proposed as part of Supplement C to the Guideline on Air Quality Modeling. As part of the
rulemaking process, the model and associated documentation was subject to public review and
A-4
-------
Appendix A
Verification and Validation of the Air Module
comment. Supplement C was promulgated on 8/9/95 (60 FR 40465) and model was renamed to
ISCST3 and released to the public for use in regulatory applications.
To meet some of the needs and constraints of the proposed HWIR application, ISCST3
was further modified. One of the major areas of concern for the HWIR application was the
runtime of ISCST3. There were two modifications made to ISCST3 to address this issue. The
first change was the addition of an option to sample from the meteorological data at regular
intervals rather than run each consecutive hour. This option is used to calculate annual values
for concentration and deposition. This use of this option, referred to as the Sampled
Chronological Input Methodology was tested over a range of sources, climate regimes, and
sampling rates (U.S. EPA, 1998). The testing resulted in a recommended sampling interval for
the HWIR application, however the use of the sampling option and the selection of the sampling
interval are not mandated by the 3MRA software system. A second modification to ISCST3
involved an analysis of the plume depletion algorithm since this is a numerically intensive
calculation for area sources. A comparison of available depletion schemes was completed and a
new algorithm was proposed, tested, and implemented (Venkatram, 1998 ) to replace the existing
scheme (Horst, 1983).
In addition to the testing of the individual model components described above, the model,
preprocessors, and postprocessors have undergone internal and independent testing as part of the
software development process supporting the HWIR application(ref). The testing plan addressed
the following specific requirements:
1. The air module should correctly develop the input file needed to run ISCST3, run
the model, archive results for use in subsequent runs, output results in the correct
format for use by other models, write an error messages to the 3MRA system log
file.
2. The air module should be successfully implement the spline option and produce a
reasonable spline surface and output values in the format appropriate for use by
subsequent models.
3. The air module should correctly implement the plume depletion option.
Concentration and deposition values should be higher when depletion is not
considered.
4. The SCIM option should be correctly implemented by the air module. Values
determined when every hour is sampled should approximate those calculated
when the option is not used.
A.3 Validation of the ISCST3
The ISCST3 model provides point estimates of concentration, dry deposition, and wet
deposition. While there have been a number of studies that have compared concentrations
predicted by ISCST3 (or its predecessors) with observational databases, the deposition estimates
are harder to validate since field studies of dry and/or wet deposition flux from point or area
sources are seldom found in the literature. The subsections that follow discuss validation studies
A-5
-------
Appendix A
Verification and Validation of the Air Module
that have been performed on the various components of the ISCST3 model with the intent to
show that the modeling approach adequately represents the physical processes in the atmosphere.
A.3.1 Concentration estimates
The ISCST model (Bowers et al, 1979) evolved from earlier plume models such as
CRSTER (U.S. EPA, 1977), inheriting the same dispersion algorithms and adding building
downwash algorithms. So, evaluation studies done for CRSTER (and its derivatives) are also
applicable to ISCST and subsequent revisions to it. Concentration estimates from CRSTER
were extensively evaluated for a number of point sources using databases for coal fired power
plants. These studies are listed in Table A-l.
Table A-l. Coal-Fired Power Plants used for ISCST3 Evaluation
Database
Location
Stack height(s)
Terrain type
Reference
Clifty Creek
Indiana
3 stacks; all 208 m
low ridges and rolling hills
Londergan et al (1982)
Muskingum
Ohio
2 stacks; all 252 m
low ridges and rolling hills
Cox etal (1985)
Paradise
Kentucky
3 stacks; 183 m,
183 m, 244 m
flat terrain surrounded by rolling
hills
Cox etal (1987)
Kincaid
Illinois
1 stack;
187 m
flat terrain
Cox etal (1986)
Separate studies were performed to evaluate the concentrations predicted for area
sources. Field studies of area sources, particularly ones measuring impacts near and within the
source are scarce. Alternative methods for evaluating the algorithm were used by EPA to
prepare for recommending the algorithm for regulatory modeling applications. As described
above, a comparison of area source algorithms used by various dispersion models (U.S. EPA,
1989) was conducted to select an algorithm for inclusion in ISCST2. In this study, a set of
prediction scenarios was developed for testing and intercomparing the algorithms. A follow-on
study examined the sensitivity of the selected area source algorithm across a range of source
characteristics and compares the results to those from the original ISCST model algorithm.
Finally, the algorithm selected for ISCST2 was compared to data from a wind tunnel study (U.S.
EPA, 1992d).
A.3.2 Dry deposition estimates
Particle dry deposition flux is calculated in the model by multiplying the concentration
by the deposition velocity. Studies of particle deposition flux attributable to individual sources
are difficult to find. In the absence of flux studies to be used for validation, the emphasis has
been placed on validating the algorithm for estimating the particle deposition velocity and
relying on the previously noted validations of the concentration algorithms. To select the dry
deposition algorithm to be added to ISCST3, EPA evaluated and intercompared a number of
deposition velocity algorithms and implemented the "most appropriate approach" (U.S. EPA,
1994). The algorithms were evaluated on their ability to parameterize important physical
A-6
-------
Appendix A
Verification and Validation of the Air Module
processes while requiring only readily available meteorological, chemical and physical input
data. Additionally sensitivity tests were done by Schwede and Paumier (1997) which exercised
the algorithms and demonstrated the ability of the model to produce estimates within expected
ranges. Another aspect of the deposition calculation is the accounting for the material lost from
the source due to deposition. This is referred to as depletion. For use in the HWIR application,
the depletion algorithm in ISCST3 was changed to use one that was shown to be faster and more
robust (Venkatram, 1998). The algorithm was validated against a full numerical solution to the
eddy diffusivity equation (Venkatram, personal communication).
A.3.3 Wet deposition estimates
ISCST3 is used to calculate the wet deposition of both particles and gases for the HWIR
assessment. Wet deposition is dependent on the concentration, the scavenging coefficient
(dimensionless), and the precipitation rate. For particle deposition, the scavenging coefficient is
specified for ISCST3 by particle size category. For wet deposition, the scavenging coefficient is
chemical specific. To reduce the number of model runs frequired for the HWIR application, the
air module was configured to use a single value was used for all gases which causes gases to be
scavenged as if they were small particles. The general approach for calculating the wet
deposition flux and resulting depletion was proposed by Maul (1980) based on an analysis of
ambient data.
A.4 Peer Reviews
The ISCST3 model has had a long history of use by the EPA for modeling sources of
nonreactive pollutants and is designated as a "preferred model" in the Guideline on Air Quality
Modeling. As such, the model has undergone substantial public review and comment over a
number of years as the model was originally promulgated and subsequently revised. Copies of
the comments received by the EPA during these reviews and the Agency's response to them
have been placed in Docket No. A-92-65. The docket is available for public inspection and
copying.
For use in the HWIR application, ISCST3 was modified to reduce the computational
burden and to provide additional outputs needed by other modules. The first modification
involves no change to the actual dispersion algorithms within the model, but simply allows the
user to select an option that causes the model to sample from an extended record of meteorology
at a user specified interval rather than use each hour in the data. The second modification
involves the calculation of plume depletion resulting from dry deposition. The Horst (1983)
method which had been previously used in ISCST3 was replaced by a method developed by
Venkatram (1998) that provided consistent results with faster runtimes. To further reduce
runtimes, ISCST3 was run for the HWIR application using a unit emission rate. Annual varying
estimates of concentration and deposition were obtained by multiplying the ISCST3 annual
average estimates by an emission rate defined by the source module for a particular year.
Additional runtime savings were obtained by using a single scavenging coefficient for all vapor
phase wet deposition calculations rather than running ISCST3 for each chemical individually.
The revised ISCST3 model and its implementation for the HWIR application was peer
reviewed. The runtime considerations and requirements for the HWIR application were not
A-7
-------
Appendix A
Verification and Validation of the Air Module
adequately represented to the peer reviewers which was reflected in their comments. While
computer processor speeds have greatly increased over the last few years, the ISCST3 model
continues to require a great deal of runtime, particularly considering the number of sources and
chemicals in the HWIR application. However, the framework is flexible and allows the modeler
to reuse stored ISCST3 results (provided with the software system), rather than running the
ISCST3 model each time, to decrease overall system runtimes. When running the ISCST3
model, the modeler may select the sampling interval, if any, to use depending on available
computing power, the number of sources to be modeled, etc. The ISCST3 model does allow the
use of time varying emission rates and chemical specific scavenging coefficients, but the
framework does not allow the use of these features at this time. As far as the technical merits of
the modifications, the reviewers felt that the new depletion algorithm was appropriate for HWIR-
type sources and that the algorithm was based on current scientific knowledge. However, the
reviewers felt that the documentation for the new depletion algorithm was not complete enough
to allow them to fully evaluate the new algorithm. The reviewers agreed that sampling the
meteorology at regular intervals should produce reasonable results, particularly with the
inclusion of a separate sampling rate for wet hours. The reviewers suggested that studies be
undertaken to further examine the issues related to developing annually varying concentration
and deposition estimates as well as those related to chemical specific scavenging coefficients.
References
Bowers, J.F., and A.J. Anderson, An Evaluation Study for the Industrial Source Complex (ISC)
Dispersion Model, EPA-450/4-81-002, U.S. Environmental Protection Agency, Research
Triangle Park, NC, January 1981.
Cox, W.M., and G.K. Moss, Evaluation of Rural Air Quality Simulation Models, Addendum A:
Muskingum River Data Base, EPA-450/83-003a, U.S. Environmental Protection Agency,
Research Triangle Park, NC, June 1985.
Cox, W.M., H.W. Rorex, and G.K. Moss, Evaluation of Rural Air Quality Simulation models,
Addendum C: Kincaid S02 Data Base, EPA-450/83- 003c, U.S. Environmental
Protection Agency, Research Triangle Park, NC, March 1986.
Cox, W.M., H.W. Rorex, G.K. Moss, and K.W. Baldridge, Evaluation of Rural Air Qualithy
Simulation models, Addendum D: Paradise S02 Data Base, EPA- 450/83-003d, U.S.
Environmental Protection Agency, Research Triangle Park, NC, January 1987.
Horst, T.W., 1983: A correction to the Gaussian source-depletion model. In Precipitation
Scavenging. Dry Deposition and Resuspension. H.R. Pruppacher, R.G. Semonin, W.G.N.
Slinn, eds., Elsevier, NY.
Irwin, J.S., 2002: A historical look at the development of regulatory air quality models for the
United States Environmental Protection Agency. NOAA Technical Memorandum OAR
ARL-244. National Oceanic and Atmospheric Adminstration, Silver Spring, MD.
A-8
-------
Appendix A
Verification and Validation of the Air Module
Londergan, R.J., D.H. Minott, D.J. Wackter, and D. Bonitta, Evaluation of Rural Air Quality
Simulation Models, EPA-450/4-83-003, U.S. Environmental Protection Agency,
Research Triangle Park, NC, October 1982.
Schwede, D.B. and J.O. Paumier, 1987: Sensitivity of the Industrial Source Complex Model to
Input Deposition Parameters. J. Applied Meteorology, 36, 1096-1106.
Scire, J.S., D.G. Strimaitis, and R.J. Yamartino, 2000: Users's Guide for the CALPUFF
Dispersion Model. Earth Tech, Inc., Concord, MA.
Strimaitis, D.G., J.C. Chang, J.S. Scire, 1993: Development of ISC-COMPDEP and User
Instructions. Sigma Research Corporation, Concord, MA.
U.S. Environmental Protection Agency, 1977: User's Manual for Single-Soource (CRSTER)
Model. EPA-450/2-77-013. U.S. Environmental Protection Agency, Research Triangle
Park, NC.
U.S. Environmental Protection Agency, 1989: Review and Evaluation of Area Source Dispersion
Algorithms for Emission Sources at Superfund Sites. EPA-450/4-89-020. U.S.
Environmental Protection Agency, Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1992a: Summary of Quality Assurance and Equivalence
Tests Performed on the ISC2 Models. Project report for WA No. 68D00124,U.S.
Environmental Protection Agency, Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1992b: Sensitivity Analysis of a Revised Area Source
Algorithm for the Industrial Source Complex Short Term Model. EPA-454/R-92-015.
U.S. Environmental Protection Agency, Research Triangle Park, NC.
U.S. Environmental Protection Agency, 1992c: Summary of the Quality Assurance and
Equivalence Tests Performed on the Modified Area Source Algorithm for the ISCST2
Model. Project report for WA No. 1-27, 68D00124. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
U.S. Environmnetal Protection Agency, 1992d: Comparison of a Revised Area Source
Algorithm for the Industrial Source Complex Short Term Model and Wind Tunnel Data,
EPA-454/R-92-014. U.S. Environmental Protection Agency, Research Triangle Park,
NC.
U.S. Environmental Protection Agency, 1994. Development and Testing of a Dry Deposition
Algorithm (Revised). EPA Publication No. EPA-454/R-94-015. U.S. Environmental
Protection Agency, Research Triangle Park, NC. (NTIS No. PB 94-183100)
U.S. Environmental Protection Agency. 1998. Testing of the Sampled Chronological Input
Model (SCIM) option in the enhanced ISCST3 Model for Use in the Hazardous Waste
Identification Rule (HWIR99). Office of Solid Waste, Washington, DC.
A-9
-------
Appendix A Verification and Validation of the Air Module
Venkatram, A., 1998: A Simple Model for Dry Deposition and Particle Settling. Subcontractor
Progress Report 2 (including addendum), EPA Contract No. 68D70002, Work
Assignment No. 1-001.
A-10
-------
Appendix A
Verification and Validation of the Air Module
This page intentionally left blank.
A-ll
-------
Appendix B
Quality Assurance Verification and Validation Tests
for the Exposure Analysis Modeling System - Exams
-------
This page intentionally left blank.
-------
Appendix B
Verification and Validation Tests for EXAMS
Appendix B
Quality Assurance Verification and Validation Tests
for the Exposure Analysis Modeling System - Exams
-------
Appendix B Verification and Validation Tests for EXAMS
This page intentionally left blank.
-------
Appendix B
Verification and Validation Tests for EXAMS
EPA-600/X-02/xxx
United States
Environmental Protection
Agency
Research and
Development
Quality Assurance Verification
and Validation Tests for the
Exposure Analysis Modeling
System - Exams
Robert B. Ambrose, Jr., P.E.
Lawrence A. Burns, Ph.D.
Ecosystems Research Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Athens, GA
October 2002
5
-------
Appendix B
Verification and Validation Tests for EXAMS
Disclaimer
This document is intended for internal Agency use only. Mention of trade names or commercial
products does not constitute endorsement or recommendation for use.
B-iii
-------
Appendix B
Verification and Validation Tests for EXAMS
Table of Contents
B. 1 Introduction to the Exposure Analysis Modeling System (Exams) B-l
B.l.l Background B-l
B. 1.2 Exposure Analysis in Aquatic Systems B-3
B. 1.3 The Exams Program B-3
B. 1.4 Exams Process Models B-4
B. 1.4.1 Ionization and Sorption B-4
B. 1.4.2 Transformation Processes B-5
B. 1.4.3 Transport Processes B-6
B. 1.4.4 Chemical Loadings B-6
B.1.5 Ecosystems Analysis and Mathematical Systems Models B-6
B. 1.5.1 Exams Design Strategy B-7
B. 1.5.2 Temporal and Spatial Resolution B-8
B. 1.5.3 Assumptions B-10
B.2 Verification of Exams B-l 1
B.2.1 Background B-ll
B.2.2 Water Body Network B-13
B.2.3 Dispersion Parameters B-13
B.2.4 Water Balance B-14
B.2.4.1 Water Body Network 1 B-14
B.2.4.2 Water Body Network 2 B-15
B.2.4.3 Water Body Network 3 B-16
B.2.5 Solids Balance B-16
B.2.5.1 Water Body Network 1 B-17
B.2.5.2 Water Body Network 2 B-17
B.2.5.3 Water Body Network 3 B-18
B.2.6 Chemical Loadings and Conservative Transport B-19
B.2.6.1 Water Body Network 1 B-19
B.2.6.2 Water Body Network 2 B-19
B.2.6.3 Water Body Network 3 B-20
B.2.7 Ionization B-20
B.2.7.1 Organic Acids B-21
B.2.7.2 Organic Bases B-22
B.2.8 Sorption B-22
B.2.8.1 Organic Chemicals B-23
B.2.8.2 Metals B-24
B.2.9 Volatilization B-24
B.2.9.1 Volatilization in Stream Reaches B-25
B.2.9.2 Volatilization in Ponds, Wetlands, and Lakes B-25
B.2.9.3 Effect of Ionization on Volatilization B-27
B.2.10 Transformation Processes B-28
B.2.10.1 Organic Chemical Transformations B-28
B.2.10.2 Mercury Transformations B-30
B.2.11 Test Robustness B-32
B.2.11.1 Screening across Sites B-32
B.2.11.2 Screening across Chemicals B-33
B-iv
-------
Appendix B
Verification and Validation Tests for EXAMS
Table of Contents (continued)
B.2.11.3 Monte-Carlo Screening with Random Parameter Values B-34
B.3 Exams System Validation Case Studies B-40
B.3.1 Overview B-40
B.3.2 Linear Alkylbenzene Sulfonate in a Small Stream B-40
B.3.3 Phenol in the Monongahela River B-41
B.3.4 Disperse Yellow 42 Dye in a Pond B-41
B.3.5 Chlorophenol, Chloroquaiacol, and Chlorocatechol in Norrsundet Bay
(Sweden) B-42
B.3.6 Three Herbicides and an Insecticide in Rice Paddies B-43
B.3.7 Xylenes, Dichlorobenzenes, Styrene, and 4-phenyldodecane in a small
lowland river in U.K B-44
B.3.8 Aniline and Lindane in River Calder B-45
B.3.9 Bensulfuron Methyl and Azimsulfuron in Rice Paddies B-45
B.3.10 Discussion of Case Studies B-46
References B-47
B-v
-------
Appendix B
Verification and Validation Tests for EXAMS
B.l Introduction to the Exposure Analysis
Modeling System (Exams)
B.l.l Background
The HWIR Surface Water Module is one component of the 3MRA software system that
is designed to evaluate the nationwide probability of human and ecological risk from chemicals
placed in various waste management units (WMUs). The HWIR Surface Water Module takes
the loadings calculated by the source, atmospheric, watershed, and groundwater modules, along
with data on meteorology, hydrology, environmental conditions, and chemical reactivity, and
calculates the chemical concentrations throughout the water body network over time. As
illustrated in Figure B-l, the HWIR Surface Water Module consists of the core model Exams II
and the interface module Exams 10 (Ambrose and Burns, 2000; U.S. EPA 1999), which was
developed specifically for the HWIR project. It reads data from other HWIR modules and
databases, and builds Exams input files describing the water body environment and chemical
properties, along with the command file that specifies the chemical loading history and controls
the Exams simulation. ExamsIO passes control to Exams, which conducts the simulation and
produces intermediate results files. ExamsIO then processes the intermediate files and passes
the output data back to the proper HWIR database.
While the Exams component of the HWIR Surface Water Module is a fully functional
model independent of this project, it is driven and constrained in various ways by ExamsIO and
the HWIR databases. For convenience, the HWIR Surface Water Module will be referred to
here as H-Exams. This composite module is documented in an unpublished report (Ambrose
and Burns, 2000), and in the HWIR99 Background document for surface water modules (U.S.
EPA 1999), part of EPA Office of Solid Waste's web site of HWIR documents. While Exams
can be run interactively or as a batch program, H-Exams is implemented solely as a batch
process. H-Exams does not consider transformations due to photolysis or oxidation.
Transformation rate constants for hydrolysis, biodegradation, and reduction are calculated by the
HWIR chemical processor and passed through the batch chemical database to Exams. Internal
Exams algorithms for calculating rate constants are bypassed.
ExamsIO provides the link between other HWIR modules and databases and the Exams
model. For each site simulation, HWIR software generates site simulation files (SSFs) detailing
the chemical properties and the site characteristics. The source and natural media modules are
then executed, followed by the exposure and risk modules. Modules produce time series
calculations and store results in their global results files (GRFs). ExamsIO reads results
generated by the source module, the atmospheric module, the watershed module, and the aquifer
module. Three batch Exams files are generated by ExamsIO - the run file "hwirexp.exa," the
chemical file "hwirexpl.chm," and the environment file "hwirexp.env." If two or more
chemicals are included in one simulation, then separate files are created for each chemical
(hwirexp2.chm, hwirexp3.chm, etc.). Next, EXAMS is executed, generating the temporary results
files hwirexp.xms and (if storage level is set to maximum) report.xms. Finally, ExamsIO reads
hwirexp.xms and produces the properly named and formatted surface water GRF, which is used
as input by the HWIR exposure modules.
B-l
-------
Appendix B
Veri fication and Validation Tests for EXAMS
Site Definition Files:
1 SL
I ssf
MET I
"at |
CPStream
ssf
SW1
ssf
SW2
ssf
i...................
Module Results Files:
l SR l
I grf I
I AR I
I grf I
I ws
I grf
I AQ I
I grf I
Surface Water Results Files:
SW1
SW2
grf
grf
Preproc
*
*
*
hwirexp
hwirexp
hwirexpl
exa
env
chm
\TJ
Exams-II
\
t
hwirexp
report
xms
xms
i
Postproc
Figure B-l. Surface water module components.
The chemical file specifies the relevant chemical properties, including molecular weight,
melting point, Henry's Law constant, vapor pressure, ionization constants, partition coefficients,
solubilities, various rate constants, and reaction yield coefficients when appropriate. The
environment file specifies all relevant aspects of the physical environment to be simulated.
These properties include compartment geometry, advective flow paths, mixing characteristics,
and environmental characteristics that influence chemical reactions, such as pH, DOC, microbial
activity, and reductant concentration. The run file controls the EXAMS simulation. It sets
simulation control variables at the beginning, then specifies yearly inflows and loadings for each
modeled compartment. Yearly changes in various environmental properties are also specified,
including TSS and water temperature.
The major model simplifications made in response to the project constraints include the
use of average-yearly hydrological and loading inputs, the use of national distributions to specify
some site-specific environmental conditions, and the use of a simple solids balance with no
settling and burial. For sites that experience periodic drying, a small positive flow equivalent to
5 mm/year of direct precipitation onto the water body surface is maintained in order to keep the
model functioning.
B-2
-------
Appendix B
Verification and Validation Tests for EXAMS
B.1.2 Exposure Analysis in Aquatic Systems
Exams was conceived as an aid to those who must execute hazard evaluations solely
from laboratory descriptions of the chemistry of a newly synthesized toxic compound. EXAMS
estimates exposure, fate, and persistence following release of an organic chemical into an aquatic
ecosystem.
When a pollutant is released into an aquatic ecosystem, it is entrained in the transport
field of the system and begins to spread to locations beyond the original point of release. During
the course of these movements, chemical and biological processes transform the parent
compound into daughter products. Residual concentrations can be compared to those posing a
danger to living organisms. These "expected environmental concentrations" (EECs), or exposure
levels, in receiving water bodies are one component of a hazard evaluation.
The toxicologist also needs to know which populations in the system are "at risk."
Populations at risk can be deduced to some extent from the distribution or "fate" of the
compound, that is, by an estimate of EECs in different habitats of single ecosystems. Exams
reports a separate EEC for each compartment, and thus each local population, used to define the
system.
B.1.3 The Exams Program
The need to predict chemical exposures from limited data has stimulated a variety of
recent advances in environmental modeling. These advances fall into three general categories:
¦ Process models giving a quantitative, often theoretical, basis for predicting the
rate of transport and transformation processes as a function of environmental
variables.
¦ Procedures for estimating the chemical parameters required by process models.
Examples include linear free energy relationships, and correlations summarizing
large bodies of experimental chemical data.
¦ Systems models that combine unit process models with descriptions of the
environmental forces determining the strength and speed of these processes in
real ecosystems.
The Exams program is a deterministic, predictive systems model, based on a core of
mechanistic process equations derived from fundamental theoretical concepts. The Exams
computer code also includes descriptive empirical correlations that ease the user's burden of
parameter calculations, and an interactive command language that facilitates the application of
the system to specific problems.
Exams "predicts" in a somewhat limited sense of the term. Many of the predictive
water-quality models currently in use include site-specific parameters that can only be found via
field calibrations. After "validation" of the model by comparison of its calibrated outputs with
additional field measurements, these models are often used to explore the merits of alternative
management plans. Exams, however, deals with an entirely different class of problem. Because
B-3
-------
Appendix B
Verification and Validation Tests for EXAMS
newly synthesized chemicals must be evaluated, little or no field data may exist. Furthermore,
EECs at any particular site are of little direct interest. In this case, the goal, at least in principle, is
to predict EECs for a wide range of ecosystems under a variety of geographic, morphometric, and
ecological conditions. Exams includes no direct calibration parameters, and its input
environmental data can be developed from a variety of sources. The EECs generated by Exams
are thus "evaluative" (Lassiter et al. 1979) predictions designed to reflect typical or average
conditions. Exams' environmental database can be used to describe specific locales, or as a
generalized description of the properties of aquatic systems in broad geographic regions.
Exams relies on mechanistic, rather than empirical, constructs for its core process
equations wherever possible. Mechanistic (physically determinate) models are more robust
predictors than are purely empirical models, which cannot safely be extended beyond the range
of prior observations. Exams contains a few empirical correlations among chemical parameters,
but these are not invoked unless the user approves.
B.1.4 Exams Process Models
In Exams, the loadings, transport, and transformations of a compound are combined into
differential equations by using the mass conservation law as an accounting principle. This law
accounts for all the compound entering and leaving a system as the algebraic sum of (1) external
loadings, (2) transport processes exporting the compound out of the system, and
(3) transformation processes within the system that degrade the compound to its daughter
products. The fundamental equations of the model describe the rate of change in chemical
concentrations as a balance between increases due to loadings, and decreases due to the transport
and transformation processes removing the chemical from the system.
The set of unit process models used to compute the kinetics of a compound is the central
core of Exams. These unit models are all "second-order" or "system-independenf'models: each
process equation includes a direct statement of the interactions between the chemistry of a
compound and the environmental forces that shape its behavior in aquatic systems. Thus, each
realization of the process equations implemented by the user in a specific Exams simulation is
tailored to the unique characteristics of that ecosystem. Most of the process equations are based
on standard theoretical constructs or accepted empirical relationships. For example, light
intensity in the water column of the system is computed using the Beer-Lambert law, and
temperature corrections for rate constants are computed using Arrhenius functions.
B.l.4.1 Ionization and Sorption
Ionization of organic acids and bases, complexation with dissolved organic carbon (doc),
and sorption of the compound with sediments and biota, are treated as thermodynamic properties
or (local) equilibria that alter the operation of kinetic processes. For example, an organic base in
the water column may occur in a number of molecular species (as dissolved ions, sorbed with
sediments, etc.), but only the uncharged, dissolved species can be volatilized across the air-water
interface. Exams allows for the simultaneous treatment of up to 28 molecular species of a
chemical. These include the parent uncharged molecule, and singly, doubly, or triply charged
cations and anions, each of which can occur in a dissolved, sediment-sorbed, DOC-complexed, or
biosorbed form. The program computes the fraction of the total concentration of compound that
is present in each of the 28 molecular structures.
B-4
-------
Appendix B
Verification and Validation Tests for EXAMS
B.l.4.2 Transformation Processes
Exams computes the kinetics of transformations attributable to direct photolysis,
hydrolysis, biolysis, and oxidation reactions.
Exams includes two algorithms for computing the rate of photolytic transformation of a
synthetic organic chemical. These algorithms accommodate the two more common kinds of
laboratory data and chemical parameters used to describe photolysis reactions. The simpler
algorithm requires only an average pseudo-first-order rate constant applicable to near-surface
waters under cloudless conditions at a specified reference latitude. To control reactivity
assumptions, the rate constant is coupled to nominal (normally unit-valued) reaction quantum
yields for each molecular species of the compound. This approach makes possible a first
approximation of photochemical reactivity, but neglects the very important effects of changes in
the spectral quality of sunlight with increasing depth in a body of water. The more complex
photochemical algorithm computes photolysis rates directly from the absorption spectra (molar
extinction coefficients) of the compound and its ions, measured values of the reaction quantum
yields, and the environmental concentrations of competing light absorbers (chlorophylls,
suspended sediments, doc, and water itself).
The total rate of hydrolytic transformation of a chemical is computed by EXAMS as the
sum of three contributing processes. Each of these processes can be entered via simple rate
constants, or as Arrhenius functions of temperature. The rate of specific-acid-catalyzed reactions
is computed from the pH of each sector of the ecosystem, and specific-base catalysis is
computed from the environmental pOH data. The rate data for neutral hydrolysis of the
compound are entered as a set of pseudo-first-order rate coefficients (or Arrhenius functions) for
reaction of the 28 (potential) molecular species with the water molecule.
Exams computes biotransformation of the chemical in the water column and in the
bottom sediments of the system as entirely separate functions. Both functions are second-order
equations that relate the rate of biotransformation to the size of the bacterial population actively
degrading the compound (Paris et al. 1982). This approach is of demonstrated validity for at least
some biolysis processes, and provides the user with a minimal semi-empirical means of
distinguishing between eutrophic and oligotrophic ecosystems. The second-order rate constants
can be entered either as single-valued constants or as functions of temperature.
Oxidation reactions are computed from the chemical input data and the total
environmental concentrations of reactive oxidizing species (alkylperoxy and alkoxyl radicals,
etc.), corrected for ultra-violet light extinction in the water column. The chemical data can again
be entered either as simple second-order rate constants or as Arrhenius functions. Oxidations
due to singlet oxygen are computed from chemical reactivity data and singlet oxygen
concentrations; singlet oxygen is estimated as a function of the concentration of DOC, oxygen
tension, and light intensity. Reduction is included in the program as a simple second-order
reaction process driven by the user entries for concentrations of reductants in the system. As
with biolysis, this provides the user with a minimal empirical means of assembling a simulation
model that includes specific knowledge of the reductants important to a particular chemical
safety evaluation.
B-5
-------
Appendix B
Verification and Validation Tests for EXAMS
B.l.4.3 Transport Processes
Internal transport and export of a chemical occur in Exams via advective and dispersive
movement of dissolved, sediment-sorbed, and biosorbed materials and by volatilization losses at
the air-water interface. Exams provides a set of vectors that specify the location and strength of
both advective and dispersive transport pathways. Advection of water through the system is then
computed from the water balance, using hydrologic data (rainfall, evaporation rates, stream
flows, groundwater seepages, etc.) supplied to Exams as part of the definition of each
environment.
Dispersive interchanges within the system, and across system boundaries, are computed
from the usual geochemical specification of the characteristic length, cross-sectional area, and
dispersion coefficient for each active exchange pathway. Exams can compute transport of
synthetic chemicals via whole-sediment bed loads, suspended sediment wash-loads, exchanges
with fixed-volume sediment beds, ground-water infiltration, transport through the thermocline of
a lake, losses in effluent streams, etc. Volatilization losses are computed using a two-resistance
model. This computation treats the total resistance to transport across the air-water interface as
the sum of resistances in the liquid and vapor phases immediately adjacent to the interface.
B.l.4.4 Chemical Loadings
External loadings of a toxicant can enter the ecosystem via point sources, non-point
sources, dry fallout or aerial drift, atmospheric wash-out, and ground-water seepage entering the
system. Any type of load can be entered for any system compartment, but the program will not
implement a loading that is inconsistent with the system definition. For example, the program
will automatically cancel a rainfall loading entered for the hypolimnion or benthic sediments of a
lake ecosystem. When this type of corrective action is executed, the change is reported to the
user via an error message.
B.1.5 Ecosystems Analysis and Mathematical Systems Models
The Exams program was constructed from a systems analysis perspective. The system
environment comprises those factors (or "forcing functions") affecting system outputs over
which the system has little or no control. Examples for an aquatic ecosystem include runoff and
sediment erosion from its watershed, insolation, and rainfall. System resources are defined as
those factors affecting performance over which the system exercises some control. Resources of
an aquatic ecosystem include, for example, the pH throughout the system, light intensity in the
water column, and dissolved oxygen concentrations. Each of the components or "state
variables" of a system can be described in terms of its local input/output behaviors and its causal
connections with other elements of the system. The systems approach lends itself to the
formulation of mathematical systems models, which are simply tools for encoding knowledge of
transport and transformation processes and deriving the implications of this knowledge in a
logical and repeatable way.
A systems model, when built around relevant state variables (measurable properties of
system components) and causal process models, provides a method for extrapolating future states
of systems from knowledge gained in the past. In order for such a model to be generally useful,
however, most of its parameters must possess an intrinsic interest transcending their role in any
B-6
-------
Appendix B
Verification and Validation Tests for EXAMS
particular computer program. For this reason, Exams was designed to use chemical descriptors
(Arrhenius functions, pKa, vapor pressure, etc.) and water quality variables (pH, chlorophyll,
biomass, etc.) that are independently measured for many chemicals and ecosystems.
B.l.5.1 Exams Design Strategy
The conceptual view adopted for Exams begins by defining aquatic ecosystems as a
series of distinct subsystems, interconnected by physical transport processes that move synthetic
chemicals into, through, and out of the system. These subsystems include the epilimnion and
hypolimnion of lakes, littoral zones, benthic sediments, etc. The basic architecture of a computer
model also depends, however, on its intended uses. Exams was designed for use by
toxicologists and decision-makers who must evaluate the risk posed by use of a new chemical,
based on a forecast from the model. The Exams program is itself part of a "hazard evaluation
system," and the structure of the program was necessarily strongly influenced by the niche
perceived for it in this "system."
Many intermediate technical issues arise during the development of a systems model.
Usually these issues can be resolved in several ways; the modeling "style" or design strategy
used to build the model guides the choices taken among the available alternatives. The strategy
used to formulate Exams begins from a primary focus on the needs of the intended user and,
other things being equal, resolves most technical issues in favor of the more efficient
computation. For example, all transport and transformation processes are driven by internal
resource factors (pH, temperature, water movements, sediment deposition and scour, etc.) in the
system, and each deserves separate treatment in the model as an individual state variable or
function of several state variables. The strategy of model development used for Exams
suggests, however, that the only state variable of any transcendent interest to the user is the
concentration of the chemical itself in the system compartments. Exams thus treats all
environmental state variables as coefficients describing the state of the ecosystem, and only
computes the implications of that state, as residual concentrations of chemicals in the system.
Although this approach vastly simplifies the mathematical model, with corresponding
gains in efficiency and speed, the system definition is now somewhat improper. System
resources (factors affecting performance that are subject to feedback control) have been
redefined as part of the system environment. In fact, the "system" represented by the model is
no longer an aquatic ecosystem, but merely a chemical pollutant. Possible failure modes of the
model are immediately apparent. For example, introduction of a chemical subject to alkaline
hydrolysis and toxic to plant life into a productive lake would retard primary productivity. The
decrease in primary productivity would lead to a decrease in the pH of the system and,
consequently, a decrease in the rate of hydrolysis and an increase in the residual concentration of
the toxicant. This sequence of events would repeat itself indefinitely, and constitutes a positive
feedback loop that could in reality badly damage an ecosystem. Given the chemical buffering
and functional redundancy present in most real ecosystems, this example is inherently
improbable, or at least self-limiting. More importantly, given the initial EEC, the environmental
toxicologist could anticipate the potential hazard.
There is a more telling advantage, moreover, to the use of environmental descriptors in
preference to dynamic environmental state variables. Predictive ecosystem models that include
all the factors of potential importance to the kinetics of toxic pollutants are only now being
B-7
-------
Appendix B
Verification and Validation Tests for EXAMS
developed, and will require validation before any extensive use. Furthermore, although
extremely fine-resolution (temporal and spatial) models are often considered an ultimate ideal,
their utility as components of a fate model for synthetic chemicals remains suspect. Ecosystems
are driven by meteorological events, and are themselves subject to internal stochastic processes.
Detailed weather forecasts are limited to about nine days, because at the end of this period all
possible states of the system are equally probable. Detailed ecosystem forecasts are subject to
similar constraints (Piatt et al. 1977). For these reasons, EXAMS was designed primarily to
forecast the prevailing climate of chemical exposures, rather than to give detailed local forecasts
of EECs in specific locations.
B.l.5.2 Temporal and Spatial Resolution
When a synthetic organic chemical is released into an aquatic ecosystem, the entire array
of transport and transformation processes begins at once to act on the chemical. The most
efficient way to accommodate this parallel action of the processes is to combine them into a
mathematical description of their total effect on the rate of change of chemical concentration in
the system. Systems that include transport processes lead to partial differential equations, which
usually must be solved by numerical integration. The numerical techniques in one way or
another break up the system, which is continuously varying in space and time, into a set of
discrete elements. Spatial discrete elements are often referred to as "grid points" or "nodes", or,
as in Exams, as "compartments." Continuous time is often represented by fixing the system
driving functions for a short interval, integrating over the interval, and then "updating" the
forcing functions before evaluating the next time-step. At any given moment, the behavior of the
chemical is a complicated function of both present and past inputs of the compound and states of
the system.
Exams is oriented toward efficient screening of a multitude of newly invented industrial
chemicals and pesticides. Ideally, a full evaluation of the possible risks posed by manufacture
and use of a new chemical would begin from a detailed time-series describing the expected
releases of the compound into aquatic systems over the entire projected history of its
manufacture. Given an equivalently detailed time-series for environmental variables, machine
integration would yield a detailed picture of EECs in the receiving water body over the entire
period of concern. The great cost of this approach, however, militates against its use as a
screening tool. Fine resolution evaluation of synthetic chemicals can probably be used only for
compounds that are singularly deleterious and of exceptional economic significance.
The simplest situation is that in which the chemical loadings to systems are known only
as single estimates pertaining over indefinite periods. This situation is the more likely for the
vast majority of new chemicals, and was chosen for development of Exams. It has an additional
advantage. The ultimate fate and exposure of chemicals often encompasses many decades,
making detailed time traces of EECs feasible only for short-term evaluations. In Exams, the
environment is represented via long-term average values of the forcing functions that control the
behavior of chemicals. By combining the chemistry of the compound with average properties of
the ecosystem, Exams reduces the screening problem to manageable proportions. These
simplified "first-order" equations are solved algebraically in Exams's steady-state Mode 1 to
give the ultimate (i.e., steady-state) EECs that will eventually result from the input loadings. In
addition, Exams provides a capability to study initial value problems ("pulse loads" in Mode 2),
and seasonal dynamics in which environmental driving forces are updated on a monthly basis
B-8
-------
Appendix B
Verification and Validation Tests for EXAMS
(Mode 3). Mode 3 is particularly valuable for coupling to the output of the PRZM model, which
can provide a lengthy time-series of contamination events due to runoff and erosion of sediments
from agricultural lands.
Daily pesticide export values from PRZM are transferred to Exams as "pulse loading"
events at the beginning of the day. The peak concentration on that day is then reported by
Exams as the average of the start-of-day and end-of-day values. This approximation of PRZM
runoff events does not introduce a gross distortion of the facts: For a 1-ha 2-m deep pond with an
initial concentration of 1 mg/L of a pesticide with a 2-day half-life, Exams reports a first-day
peak of 0.854 mg/L. If the same amount of pesticide is permitted to enter the pond as a steady
load over 24 hours, the peak value is 0.849 mg/L, occurring at the end of the 24 hour period. If
the load were to enter over a run-off duration of 4 hours, the pond would achieve a peak
concentration of 0.989 mg/L.
Transport of a chemical from a loading point into the bulk of the system takes place by
advected flows and by turbulent dispersion. The simultaneous transformations presently result
in a continuously varying distribution of the compound over the physical space of the system.
This continuous distribution of the compound can be described via partial differential equations.
In solving the equations, the physical space of the system must be broken down into discrete
elements. Exams is a compartmental or "box" model. The physical space of the system is
broken down into a series of physically homogeneous elements (compartments) connected by
advective and dispersive fluxes. Each compartment is a particular volume element of the
system, containing water, sediments, biota, dissolved and sorbed chemicals, etc. Loadings and
exports are represented as mass fluxes across the boundaries of the volume elements; reactive
properties are treated as point processes within each compartment.
In characterizing aquatic systems for use with EXAMS, particular attention must be given
the grid-size of the spatial net used to represent the system. In effect, the compartments must not
be so large that internal gradients have a major effect on the estimated transformation rate of the
compound. In other words, the compartments are assumed to be "well-mixed," that is, the
reaction processes are not slowed by delays in transporting the compound from less reactive to
more reactive zones in the volume element. Physical boundaries that can be used to delimit
system compartments include the air-water interface, the thermocline, the benthic interface, and
perhaps the depth of bioturbation of sediments. Some processes, however, are driven by
environmental factors that occur as gradients in the system, or are most active at interfaces. For
example, irradiance is distributed exponentially throughout the water column, and volatilization
occurs only at the air-water interface. The rate of these transformations may be overestimated
in, for example, quiescent lakes in which the rate of supply of chemical to a reactive zone via
vertical turbulence controls the overall rate of transformation, unless a relatively fine-scale
segmentation is used to describe the system. Because compartment models of strongly advected
water masses (rivers) introduce some numerical dispersion into the calculations, a relatively fine-
scale segmentation is often advisable for highly resolved evaluations of fluvial systems. In many
cases the error induced by highly reactive compounds will be of little moment to the probable
fate of the chemical in that system, however. For example, it makes little difference whether the
photolytic half-life of a chemical is 4 or 40 minutes; in either case it will not long survive
exposure to sunlight.
B-9
-------
Appendix B
Verification and Validation Tests for EXAMS
B.l.5.3 Assumptions
Exams has been designed to evaluate the consequences of longer-term, primarily
time-averaged chemical loadings that ultimately result in trace-level contamination of aquatic
systems. Exams generates a steady-state, average flow field (long-term or monthly) for the
ecosystem. The program thus cannot fully evaluate the transient, concentrated EECs that arise,
for example, from chemical spills. This limitation derives from two factors. First, a steady flow
field is not always appropriate for evaluating the spread and decay of a major pulse (spill) input.
Second, an assumption of trace-level EECs, which can be violated by spills, has been used to
design the process equations used in Exams. The following assumptions were used to build the
program.
¦ A useful evaluation can be executed independently of the chemical's actual
effects on the system. In other words, the chemical is assumed not to itself
radically change the environmental variables that drive its transformations. Thus,
for example, an organic acid or base is assumed not to change the pH of the
system; the compound is assumed not to itself absorb a significant fraction of the
light entering the system; bacterial populations do not significantly increase (or
decline) in response to the presence of the chemical.
¦ Exams uses linear sorption isotherms, and second-order (rather than
Michaelis-Menten-Monod) expressions for biotransformation kinetics. This
approach is known to be valid for the low concentrations of typical of
environmental contaminants; its validity at high concentrations is less certain.
Exams controls its computational range to ensure that the assumption of
trace-level concentrations is not grossly violated. This control is keyed to
aqueous-phase (dissolved) residual concentrations of the compound: Exams
aborts any analysis generating EECs in which any dissolved species exceeds 50%
of its aqueous solubility (see Section 2.2.2 for additional detail). This restraint
incidentally allows the program to ignore precipitation of the compound from
solution and precludes inputs of solid particles of the chemical. Although solid
precipitates have occasionally been treated as a separate, non-reactive phase in
continuous equilibrium with dissolved forms, the efficacy of this formulation has
never been adequately evaluated, and the effect of saturated concentrations on the
linearity of sorption isotherms would introduce several problematic complexities
to the simulations.
¦ Sorption is treated as a thermodynamic or constitutive property of each segment
of the system, that is, sorption/desorption kinetics are assumed to be rapid
compared to other processes. The adequacy of this assumption is partially
controlled by properties of the chemical and system being evaluated. Extensively
sorbed chemicals tend to be sorbed and desorbed more slowly than weakly sorbed
compounds; desorption half-lives may approach 40 days for the most extensively
bound compounds. Experience with the program has indicated, however, that
strongly sorbed chemicals tend to be captured by benthic sediments, where their
release to the water column is controlled by their availability to benthic exchange
processes. This phenomenon overwhelms any accentuation of the speed of
processes in the water column that may be caused by the assumption of local
equilibrium.
B-10
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2 Verification of Exams
B.2.1 Background
H-Exams was subjected to a series of tests to verify that the software accurately
performs its prescribed computations. Model output was compared with different analytical
solutions for a set of simplified test cases. These verification tests should be distinguished from
model validation, in which model predictions are compared against observed data. The tests
documented here were intended to verify that ExamsIO correctly implements Exams within the
HWIR system. In addition, the tests verify that Exams accurately reproduces the expected
analytical solutions.
H-Exams must automatically construct and execute a simulation for each water body
system at a site using the HWIR databases that are generated by each Monte-Carlo iteration in a
FRAMES-HWIR implementation. The general steps in a simulation are to construct a proper
water body network, conduct a water balance, conduct a solids balance, and calculate chemical
transport and fate. These general steps form the structure of the module testing program. The
general requirements of the surface water module are presented in Table B-l, along with specific
test identification numbers.
Each of these general steps requires that the module read data from one or more of the
HWIR databases, make intermediate calculations, write data to one or more intermediate Exams
batch files, and transfer the intermediate data into a spawned Exams application. Steps 5
through 9 further require EXAMS to make intermediate calculations, conduct a simulation for a
specified period, and transfer simulation results to an intermediate Exams output file. Steps 4, 5,
7, and 9 require the module to read the data in the Exams output file and produce a properly-
averaged and formatted HWIR output file. Finally, Step 10 requires the module to perform all
calculations without fatal error across a variety of sites, waste management units, and chemicals.
Table B-1. General Requirements for Testing the Surface Water Module
Step
Description
Test Number
1
Construct the water body network (section 3)
1.0, 2.0, 3.0
2
Construct dispersive exchanges (section 4)
1.0, 2.0, 3.0
3
Conduct the water balance (section 5)
1.0, 2.0, 3.0
4
Calculate solids transport (section 6)
1.0, 2.0, 3.0
5
Calculate conservative chemical transport (Section 7)
1.0, 2.0, 3.0
6
Calculate ionic speciation for ionizing organic chemicals (Section 8)
1.1, 1.2
7
Calculate partitioning to solids and DOC (Section 9)
CM
CO
CO
8
Calculate volatilization loss (Section 10)
3.3(a,b,c), 3.4(a-h), 3.5(a,b)
9
Calculate chemical transformation (Section 11)
1.3(a-d), 1.4
10
Test for robustness (Section 12)
4.1 -4.6,
5.1 -5.5,
6.1 -6.5
B-l 1
-------
Appendix B
Verification and Validation Tests for EXAMS
All tests in Steps 1 through 9 were done using one of three simplified water body
networks present at a hypothetical site (Figure B-2). The compartment structures for the three
test water body networks are illustrated in Figure B-3. The first two networks are simple one-
reach water bodies. Network 1 is a pond with outflow, and network 2 is a lake with no surface
HWIR Test Water Body Networks
WBN 1:
Pond
WBN 2:
Lake
Rch 2: Wetland
Rch 3: Stream
Rch1:Pond
Rch 4: Lake or Bay
WBN 3: Simple Network
Figure B-2. HWIR test water body networks.
1. L
2. B
3. B
WBN 1:
Pond
4. L
5.
B
6.
CD
EXAMS Com partment
Types
Littoral
Epilimnion
H: Hypolimnion
B: Surficial Benthic
B: Underlying Benthic
1. E
4. B
WBN 2:
Lake
1. L
2. B
3. B
7. L
8. B
9. B
WBN 3: Simple Network
10. E
13. B
Figure B-3. Exams water body compartment networks.
B-12
-------
Appendix B
Verification and Validation Tests for EXAMS
outflow. Network 3 is composed of four reaches. Two headwaters reaches (1 and 2) are
connected to reach 3, which is connected to exiting reach 4. The reaches are designated pond,
wetland, stream, and lake, respectively. All tests on WBN3 were repeated with reach 4
designated "Bay". Results in both cases should be identical. Analytical solutions based on the
mass balance principal were used for loading, transport, and transformation tests. Basic
chemical principals and formulas were used to test speciation and partitioning. The robustness
tests in Step 10 required operation of the surface water module within the HWIR production
system. These tests confirm that ExamsIO converts HWIR site and reach information into the
proper EXAMS compartment structures with valid geometry.
B.2.2 Water Body Network
These tests confirm that ExamsIO converts HWIR site and reach information into the
proper EXAMS compartment structures with valid geometry. The compartment structures for the
three test water body networks are illustrated in Figure B-3. The Exams variables being verified
in Step 1 include the following:
¦ Type Segment, the compartment type
¦ AREA, the compartment surface area
¦ DEPTH, the compartment mean depth
¦ LENGth, the compartment length
¦ WIDTH, the compartment width
The first water body network is a simple pond. The purpose of this test was to verify that
the module constructs the proper water column and benthic compartment structure with valid
geometry. The second water body network is a simple lake with no exiting surface water flow.
The purpose of this test is to verify that the module constructs the proper epilimnion,
hypolimnion, and benthic compartment structure with valid geometry. The third water body
network contains four reaches. A headwaters pond and a headwaters wetland drain into a
common stream reach, which empties into a downstream lake. The purpose of this test is to
verify that the module constructs this reach network containing the proper water column and
benthic compartment structure with valid geometry. Examination of the EXAMS output files
confirmed that the structure and geometry were transferred properly in all three cases.
B.2.3 Dispersion Parameters
The next series of tests is for proper dispersion pathways and coefficients. These tests
confirm that ExamsIO converts HWIR reach connectivity information into the proper EXAMS
dispersive transport fields with proper dispersion coefficients. For the simple HWIR networks,
longitudinal dispersion between reaches is not simulated. Within all reaches, pore water
diffusion is modeled between the two benthic compartments and bulk exchange is modeled
between the surficial benthic compartment and the water column compartment (the lower
hypolimnetic compartment for lakes and bays). For lakes and bays, dispersion across the
thermocline separating the epilimnion from the hypolimnion is also modeled. Surface area is
specified in the site layout file "SL.ssf' for ponds, wetlands, lakes, and bays, and calculated for
stream reaches. Dispersion coefficients are specified in the surface water input files for each
network. The Exams variables being verified in Step 2 include the following:
B-13
-------
Appendix B
Verification and Validation Tests for EXAMS
¦ JTURB, the first exchange compartment number
¦ ITURB, the second exchange compartment number
¦ XSTUR, the exchange cross-sectional area
¦ CHARL, the characteristic mixing length
¦ DSP, the dispersion coefficient
The purpose of this test was to verify that the module constructs the proper dispersion
pathways between the water column and benthic compartments, and that the proper dispersion
coefficients were assigned. The first water body network is a simple pond. The two dispersion
pathways—water column to surface benthic layer and surface benthic layer to underlying benthic
layer—were set up and coefficients were transferred from the HWIR databases appropriately.
The second water body network is a simple lake. The three dispersion pathways—epilimnion to
hypolimnion, hypolimnion to surface benthic layer, and surface benthic layer to underlying
benthic layer—were set up and coefficients were transferred from the HWIR databases
appropriately. The third water body network contains stream, wetland, pond, and lake reaches.
Proper vertical dispersion pathways were set up for each reach, and coefficients were transferred
from the HWIR databases appropriately.
B.2.4 Water Balance
The next set of tests is for proper water balances. These tests confirm that ExamsIO
converts HWIR reach connectivity and inflow information into the proper EXAMS advective
transport fields with proper compartment inflows. For all water body networks at a site, various
HWIR data files specify reach-reach and watershed subbasin-reach connectivity, baseflow and
runoff flow to each reach, annual average precipitation and evaporation, and upstream inflows.
The Exams variables being verified in Step 3 include the following:
¦ RAIN, the total yearly rainfall rate
¦ EVAP, the total yearly evaporation rate
¦ NPSFL, the total yearly watershed runoff flow
¦ SEEPS, the total yearly groundwater seepage inflow
¦ JFRAD, the upstream compartment number for flow path
¦ ITOAD, the downstream compartment number for flow path
¦ ADVPR, the advection parameter for flow path
B.2.4.1 Water Body Network 1
The first water body network is a simple pond with exiting surface water flow. The
purpose of this test is to verify that the module constructs the proper advection pathways through
the water column and benthic compartments, and that the proper inflows are assigned. Rainfall
and evaporation data read from the meteorological file were verified in the Exams output tables.
Watershed runoff and baseflow were mapped from the proper watershed subbasin, and runoff
volumes were passed to Exams with appropriate units conversion. Converting watershed
baseflow into Exams seepage inflow takes extra calculational steps to account for evaporation
losses. These calculations were checked to confirm that seepage flow was determined as
intended.
B-14
-------
Appendix B
Verification and Validation Tests for EXAMS
The advection pathways are set up using reach type and connectivity information in the
site layout file. In this case, the single reach is designated "exiting," so the first pathway was set
up from water column compartment 1 to 0, with an advection parameter of 1.0 (indicating that
all excess flow in compartment 1 leaves by this pathway). In addition, vertical pathways were
set up to route seepage water from lower benthic segment to upper benthic segment, and from
there to the water column. The advection parameters for seepage were also set to 1, indicating
that all excess flow is directed upward. This advective flow connectivity was confirmed by
inspection of the Exams output files. These tests confirm the inflows and advection pathways
set up and transferred by ExamsIO for this water body. For this simple pond, annual reach flow
is the sum of seepage and watershed inflow along with precipitation and evaporation flows.
Exams does not provide tables to allow direct confirmation of calculated internal flows. These
were confirmed indirectly by checking conservative mass balances.
B.2.4.2 Water Body Network 2
The second water body network is a simple lake with no exiting surface water flow. The
purpose of this test is to verify that the module constructs the proper advection pathways through
the water column and benthic compartments, and that the proper inflows are assigned. Rainfall
and evaporation data read from the meteorological file were verified in the Exams output tables.
Watershed runoff and baseflow were mapped from the proper watershed subbasin, and runoff
volumes were passed to Exams with appropriate units conversion. Conversion of watershed
baseflow into Exams seepage inflow was checked to confirm that seepage flow was determined
as intended.
The advection pathways are set up using reach type and connectivity information in the
site layout file. In this case, the single reach is designated "other," indicating no surface exiting
pathway. Excess flow, then, is routed as seepage loss. The first pathway is set up from
epilimnion to hypolimnion, with an advection parameter of 1.0 (indicating that all excess flow in
compartment 1 leaves by this pathway). Additional downward pathways are set up to route
seepage outflow from hypolimnion to upper benthic compartment, from there to lower benthic
compartment, and thence of the system. In addition, upward seepage inflow pathways are set up
from lower benthic compartment to upper benthic compartment, from there to the hypolimnion
compartment, and finally to the epilimnion. The total seepage outflow is the sum of the net
surface inflow (runoff plus precipitation minus evaporation) and the subsurface seepage inflow.
The total recirculating flow crossing the lower benthic interface is the sum of the subsurface
seepage inflow and the total seepage outflow. The advection parameters for downward seepage
are set to the total seepage outflow divided by the total recirculating flow; the complimentary
advection parameters for upward seepage are set to the subsurface seepage inflow divided by the
total recirculating flow. This advective flow connectivity was confirmed by inspection of the
Exams output files. These tests confirm the inflows and advection pathways set up and
transferred by ExamsIO for this water body. For this simple pond, annual reach flow is the sum
of seepage and watershed inflow along with precipitation and evaporation flows. Exams does
not provide tables to allow direct confirmation of calculated internal flows. These were
confirmed indirectly by checking conservative mass balances.
B-15
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.4.3 Water Body Network 3
The third water body network contains four reaches. A headwaters pond and a
headwaters wetland drain into a common stream reach, which empties into a downstream lake.
The purpose of this test is to verify that the module constructs the proper advection pathways
through the water column and benthic compartments, and that the proper inflows are assigned.
Rainfall and evaporation data read from the meteorological file were verified in the Exams
output tables. Watershed runoff and baseflow were mapped from the proper watershed subbasin,
including fractional flow pathways from a watershed to different reaches. Runoff volumes were
passed to Exams with appropriate units conversion. Conversion of watershed baseflow into
Exams seepage inflow was checked to confirm that seepage flow was determined as intended.
The advection pathways are set up using reach type and connectivity information in the
site layout file. First, pathways are determined for surface water flow among compartments 1, 4,
7, and 10. In this case, reach 4 is an exiting reach, and all excess water in its surface water
compartment (10) is routed outside the network (i.e., to 0). Thus a flow pathway from 10 to 0
with an advective fraction 1.0 is set up (see Table B.10 below). WBNRchNumRch indicates that
one upstream reach is connected to reach 4. WBNRchRchlndex specifies that the upstream
reach index is 3, and WBNRchRchFrac specifies that 100% of the excess flow in reach 3 is
routed to 4. Thus a flow pathway from Exams compartment 7 to 10 is set up with an advection
fraction of 1.0. Finally, WBNRchNumRch indicates that two upstream reaches are connected to
reach 3. WBNRchRchlndex specifies that the upstream reach indexes are 1 and 2;
WBNRchRchFrac specifies that 100% of the excess flow in reaches 1 and 2 are routed to 3.
Thus flow pathways from EXAMS compartment 1 to 7 and from 4 to 7 are set up with advection
fractions of 1.0. The four surface water pathways are confirmed by inspection of Exams output
tables.
Next, upward seepage pathways are set up from the lower benthic compartments in each
reach. These pathways route seepage water from the lower benthic compartment to the upper
benthic compartment, and from the upper benthic compartment to the surface water. For lakes,
seepage water entering the hypolimnion is further routed to the epilimnion. Because in this
network all surface water is routed downstream with no seepage loss, the upward advection
fractions are set to 1.0. The nine seepage flow pathways are confirmed by inspection of Exams
output tables.
B.2.5 Solids Balance
The next set of tests is for proper solids transport. These tests confirm that ExamsIO
converts HWIR solids loading information into the proper EXAMS solids concentrations. These
tests also confirm that ExamsIO properly translates HWIR solids properties into Exams
compartment solids properties. For all water body networks at a site, the HWIR data files
provide watershed subbasin-reach connectivity, solids organic carbon fractions, solids erosion
loadings to each reach, upstream inflows, and sediment properties.
The Exams variables being verified in Step 4 include the following:
¦ SUSED, the water column compartment suspended solids concentration
¦ BULKD, the benthic compartment bulk density
B-16
-------
Appendix B
Verification and Validation Tests for EXAMS
¦ PCTWA, the benthic compartment percent water
¦ FROC, the fraction organic solids
In addition, the HWIR output variable WBNTSSWater is being verified in step 4. These
calculations not only verify the solids computations in the surface water module, but also the
internal flow computations.
B.2.5.1 Water Body Network 1
The first water body network is a simple pond with exiting surface water flow. The
purpose of this test is to verify that the module reads the proper pond sediment properties and
solids loadings, and correctly executes a conservative solids mass transport in a single pond
reach. First, basic pond sediment properties were verified, including sediment organic carbon
content along with bulk density and "percent water," which are calculated from porosity and dry
density.
Next, the water column suspended solids concentrations were checked. ExamsIO
calculates the total suspended solids concentration as the sum of in-stream background
contributions from bank erosion and primary productivity, and external contributions from
watershed erosion. The bank erosion contribution and productivity contributions were correctly
set to 1.0 and 0.60 mg/L, respectively. Thus, background solids concentration is 1.6 mg/L.
The annual watershed contribution to solids is calculated by dividing the annual
watershed erosion loading by the annual reach flow. In this test, erosion loadings from
watershed subbasin 5 for the first three years are 100000,10000, and 1000000 g/day. Reach
flows for the first three years are 150, 139, and 160 m3/day, yielding solids concentrations of
667, 71.9, and 6250 mg/L, respectively. Adding background values, the total suspended solids
concentrations for the first three years should be 669, 73.5, and 6252 mg/L. These values are
confirmed by comparison with Exams output tables.
The final check of this section is for the fraction organic carbon of suspended solids.
This is calculated as the weighted average of the organic carbon content of biotic solids and the
organic carbon content of abiotic solids. These calculations give a long-term average abiotic
solids concentration of 2144 mg/L and an average organic carbon fraction of (2144 * 0.05 + 0.6
* 0.25) / 2145, or 0.0501. This organic carbon fraction is confirmed in EXAMS output tables.
B.2.5.2 Water Body Network 2
The second water body network is a simple lake with no exiting surface water flow. The
purpose of this test is to verify that the module reads the proper lake sediment properties and
solids loadings, and correctly executes a conservative solids mass transport in a single lake
reach. First, basic lake sediment properties were verified, including sediment organic carbon
content along with bulk density and "percent water," which are calculated from porosity and dry
density.
Next, the water column suspended solids concentrations were checked. The bank erosion
contribution and productivity contributions were correctly set to 1.0 and 0.60 mg/L, respectively,
giving a total background solids concentration of 1.6 mg/L. The annual watershed contribution
B-17
-------
Appendix B
Verification and Validation Tests for EXAMS
to solids is calculated by dividing the annual watershed erosion loading in WS.grf by the annual
reach flow. In this test, erosion loadings from watershed subbasin 6 are 100000 g/day every
year. Reach flows for the first 3 years are 525, 945, and 2002, yielding solids concentrations of
190, 106, and 50.0 mg/L, respectively. Adding background values, the total suspended solids
concentrations for the first three years should be 192, 107, and 51.6 mg/L. These values are
confirmed by comparison with Exams output tables.
The final check of this section is for the fraction organic carbon of suspended solids.
This is calculated as the weighted average of the organic carbon content of biotic solids and the
organic carbon content of abiotic solids. These calculations give a long-term average abiotic
solids concentration of 100 mg/L and an average organic carbon fraction of (100 * 0.05 + 0.6 *
0.25) / 101, or 0.0512. This organic carbon fraction is confirmed in EXAMS output tables.
B.2.5.3 Water Body Network 3
The third water body network contains four reaches. A headwaters pond and a
headwaters wetland drain into a common stream reach, which empties into a downstream lake.
The purpose of this test is to verify that the module reads the proper lake, river, pond, and
wetland sediment properties and solids loadings, and correctly executes a conservative solids
mass transport in a multiple reach system. First, basic sediment properties were verified,
including sediment organic carbon content along with bulk density and "percent water," which
are calculated from porosity and dry density.
Next, the water column suspended solids concentrations were checked. The bank erosion
contribution is simply set to a minimum background value of 5 mg/L in stream reaches and
1 mg/L in non-stream reaches. The incremental reach bank loadings are equal to the product of
reach flows and these background concentrations. Given the reach flows of 1000, 1000, 3500,
and 3996 m3/day, the corresponding bank erosion loadings are 1000, 1000, 17500, and
3996 g/day. The annual watershed contribution to solids is calculated by dividing the annual
watershed erosion loading by the annual reach flow. In this test, erosion loadings from
watershed subbasinsl through 4 are a constant 100,000 g/day. Loadings from subbasins 1, 2,
and 3 are mapped 100% to reaches 1, 2, and 3, respectively. Subbasin 4 loadings are split, 50%
going to reach 3 and 50% going to reach 4. The resulting total watershed loadings are 100000,
100000, 150000, and 50000 g/day to reaches 1 through 4, respectively.
Total cumulative abiotic solids loadings to reaches 1 through 4 are 101000, 101000,
369500, and 423496 g/day. Dividing by corresponding reach flows yields abiotic solids
concentrations of 101, 101, 105.6, and 106.0 mg/L in reaches 1 through 4. Adding biotic solids
concentrations gives total solids concentrations of 101.6, 103, 106.6, and 107.5 mg/L. These
values are confirmed in Exams output tables.
The final check of this section is for the fraction organic carbon of suspended solids.
This is calculated as the weighted average of the organic carbon content of biotic and abiotic
solids. The weighted average organic carbon fractions are 0.0512, 0.0539, 0.0518, and 0.0527,
and are confirmed in Exams output tables.
B-18
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.6 Chemical Loadings and Conservative Transport
External loadings of a toxicant can enter the ecosystem via point sources, non-point
sources, dry fallout or aerial drift, atmospheric wash-out, and ground-water seepage entering the
system. This Step 5 test verifies that chemical loadings are passed to Exams and transported
through the EXAMS water body compartments properly. Four chemical loading pathways are
employed - watershed runoff, direct source runoff, seepage inflow, and atmospheric deposition.
These tests confirm that ExamsIO converts HWTR chemical loading information into the proper
Exams chemical loadings. These tests also confirm that Exams properly transports conservative
(non-reactive) chemicals through the water body networks, yielding correct concentrations.
The Exams variables verified in Step 5 include the following:
¦ Rainfall, the total atmospheric deposition loading
¦ Drift, the source runoff loading
¦ NPS Loads, the watershed runoff and erosion loading
¦ Seeps, the groundwater seepage loading
In addition, the HWIR output water column concentration were also verified in step 5.
B.2.6.1 Water Body Network 1
The first water body network is a simple pond with exiting surface water flow. The
purpose of this test is to verify that the module reads the proper source, atmospheric, and seepage
loadings, and correctly executes conservative chemical mass transport in a single reach.
Three types of chemical loadings to the first water body network were tested - source
runoff, atmospheric deposition, and groundwater seepage Loading pathways and time variable
loading values were specified, including a constant source runoff for the first 10 years, variable
wet deposition flux for the first 5 years, and variable groundwater seepage loading for years 20
through 60. All loadings were transferred properly as verified in EXAMS output tables.
Given these loadings and the reach flows, water column concentrations for this non-
reactive chemical can be calculated and compared with model output in the surface water output
file. Calculated concentrations for each year should equal the total annual load divided by the
annual flow. These were confirmed by the time history from the surface water output file.
These tests confirmed not only that the loadings are applied correctly within EXAMS, but also
that the mass balance was conducted correctly, and that results were correctly averaged and
transferred by the ExamsIO postprocessor to the HWIR output file for use by exposure and risk
modules.
B.2.6.2 Water Body Network 2
The second water body network is a simple lake with no exiting surface water flow. The
purpose of this test was to verify that the module reads the proper atmospheric, watershed, and
groundwater loadings, and applies them correctly within EXAMS.
B-19
-------
Appendix B
Verification and Validation Tests for EXAMS
Three types of chemical loadings to the second water body network were tested -
watershed runoff, atmospheric deposition, and groundwater seepage. Loading pathways and
time variable loading values were specified, including variable watershed runoff for the first
20 years, constant particle dry and wet deposition fluxes for years 1 through 15, and variable
groundwater seepage loading for years 20 through 60. All loadings were transferred properly as
verified in Exams output tables. Predicted concentrations cannot be easily confirmed because of
the complications due to the two-layer water column.
B.2.6.3 Water Body Network 3
The third water body network contains four reaches. A headwaters pond and a
headwaters wetland drain into a common stream reach, which empties into a downstream lake.
The purpose of this test was to verify that the module reads the proper source, atmospheric,
watershed, and groundwater loadings, and correctly executes a conservative chemical mass
transport in a simple network.
Four types of chemical loadings to the third water body network were tested - source
runoff, atmospheric deposition, watershed runoff, and groundwater seepage. Split and multiple
loading pathways were specified for this test. Two source runoff pathways were specified, with
constant loadings over the first 10 years. Four constant vapor wet, particle dry, and particle wet
deposition fluxes were specified to reach 4 for years 1 through 10. Constant watershed runoff
loadings over 15 and 20 years were specified from sub-basins 1 through 4, with runoff from sub-
basin 4 directed to two separate reaches. Finally, two groundwater plumes were specified for
years 20-60, with 50% of the loading from plume 2 directed to one reach. All loadings were
transferred properly as verified in EXAMS output tables.
Calculated concentrations for each year should equal the total annual load to a reach
divided by the total annual flow. These are confirmed (with minor deviations in the third
significant figure) in the time history from the surface water output file for reaches 1 through 3,
which have a single water column layer. For reach 4, Exams calculates separate concentrations
in the epilimnion and the hypolimnion, and reports volume-weighted average concentrations in
the output file. These average concentrations should generally be a bit lower than the total load
divided by the total flow, depending on the relative rates of mixing across the thermocline versus
outflow. This is confirmed in the output file.
These tests confirm not only that the loadings were applied correctly within EXAMS, but
also that the mass balance was conducted correctly, and that results were correctly averaged and
transferred by the ExamsIO postprocessor to the HWIR output file for use by exposure and risk
modules.
B.2.7 Ionization
Ionization of organic acids and bases are treated by Exams as thermodynamic properties
or (local) equilibria that alter the operation of kinetic processes. For example, an organic base in
the water column may occur in a number of molecular species (as dissolved ions, sorbed with
sediments, etc.), but only the uncharged, dissolved species can be volatilized across the air-water
interface. Exams allows for the simultaneous treatment of up to 28 molecular species of a
chemical. These include the parent uncharged molecule, and singly, doubly, or triply charged
B-20
-------
Appendix B
Verification and Validation Tests for EXAMS
cations and anions, each of which can occur in a dissolved, sediment-sorbed, DOC-complexed, or
biosorbed form. The program computes the fraction of the total concentration of compound that
is present in each of the 28 molecular structures (the "distribution coefficients," a).
A series of simulations was run with the simple pond (water body network 1) to confirm
that water body pH and chemical dissociation constant values are passed properly to Exams, and
that Exams calculates ionic speciation correctly. Simulation output was compared with
theoretical calculations based on equations presented in section B.2.4.1 of the documentation
report (Ambrose and Burns, 2000; U.S. EPA, 1999).
The Exams variables being verified in Step 6 include the following:
¦ Chemical Species (0) and (-1), the ratio of the first ionization for an organic acid
¦ Chemical Species (-1) and (-2), the ratio of the second ionization for an organic
acid
¦ Chemical Species (0) and (+1), the ratio of the first ionization for an organic base
¦ Chemical Species (+1) and (+2), the ratio of the second ionization for an organic
base
B.2.7.1 Organic Acids
Test 1.1 is a series of simulations run for organic acids, as outlined below. The purpose
of this series of tests is to verify that the module reads the proper pH and pKa data and correctly
executes the indicated dissociation reactions, giving the proper proportions of the neutral
molecule RH3, the singly charged anion RH2", and the doubly charged anion RH2". The (negative
log) dissociation constants pKal and pKa2 were set to 6 and 8, respectively, and a series of
simulations were run with water body pH varying from 9 to 5. For each simulation, results are
verified by inspection of the EXAMS summary output file report.xms,
Ratios of chemical species by valency in any segment can be compared directly to
theoretical species ratios. For the series of simulations in this test, Exams chemical species
concentrations are converted to percentages of the total concentration, and presented along with
theoretical species ratios in the following table (Table B-2). These tests confirmed that Exams
is receiving and processing the organic acid ionization data properly.
Table B-2. Speciation Calculations for Organic Acids
PH
Theoretical Species Ratios for
pKa1=6, pKa2=8
Exams Speciation Calculations, percent
rh3/rh2
rh2/rh2
RH3
rh2
RH2
9
10"3
10"1
0.0091
9.1
91.
8
10"2
10°
0.50
50.
50.
7
10"1
101
8.3
83.
8.3
6
10°
102
50.
50.
0.50
5
101
103
91.
9.1
0.0091
B-21
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.7.2 Organic Bases
Test 1.2 is a series of simulations run for organic bases, as outlined below. The purpose
of this series of tests is to verify that the module reads the proper pH and pKa data and correctly
executes the indicated base dissociation reactions, giving the proper proportions of the neutral
molecule RH3, the singly charged cation RH4+, and the doubly charged cation RH52+. The
(negative log) dissociation constants pKal and pKa2 were set to 8 and 6, respectively, and a series
of simulations were run with water body pH varying from 9 to 5. For each simulation, results
are verified by inspection of the Exams summary output file report.xms.
Ratios of chemical species by valency in any segment can be compared directly to
theoretical species ratios. For the series of simulations in this test, Exams chemical species
concentrations are converted to percentages of the total concentration, and presented along with
theoretical species ratios in the following table (Table B-3). These tests confirm that Exams is
receiving and processing the organic base ionization data properly.
B.2.8 Sorption
Test 3.1 is a series of simulations run for hydrophobic organic chemicals, as outlined
below. Two simulations were run with water body network 3 to confirm that chemical partition
coefficients and related environmental properties are passed properly to Exams, and that Exams
calculates phase distribution correctly. Simulation output was compared with theoretical
concentration ratios based on equations presented in the documentation report. H-Exams uses
partition coefficients to compute the sorption of a chemical to abiotic solids, biotic solids, and
DOC. Coefficients for organic chemicals are processed in a different manner than those for
metals, and are tested separately.
Table B-3. Speciation Calculations for Organic Bases
PH
Theoretical Species Ratios for
pKb1=8, pKb2=6
Exams Speciation Calculations, percent
rh3/rh4+
RH47RH52+
RH3
rh4+
RH52+
9
103
101
100.
0.10
0.010
8
102
10°
98.
0.98
0.98
7
101
10"1
48.
4.8
48.
6
10°
10"2
0.98
0.98
98.
5
10"1
10"3
0.010
0.10
100.
The Exams variables being verified in Step 7 include the following:
¦ Chemical Concentrations for Sediments and Dissolved, the ratio by compartment
¦ Chemical Concentrations for Biota and Dissolved, the ratio by compartment
¦ Chemical Concentrations for DOC Complexed (x 10"3) and Dissolved, the ratio by
compartment
B-22
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.8.1 Organic Chemicals
The purpose of this test is to verify that the module reads the proper solids organic carbon
fractions and DOC concentrations, as well as Koc and Kow values, and correctly executes the
phase partitioning reactions giving the proper proportions of the dissolved, sediment-sorbed,
DOC-bound, and benthos concentrations. The octanol-water partition coefficient and the
organic carbon partition coefficient were set to 400 and 200, respectively.
For organic chemicals, the organic carbon partition coefficient, Koc, is used along with
the organic carbon fraction of solids, foc, to calculate the effective partition coefficient to solids
KPS. Given foc values of 0.05, 0.5, 0.005, and 0.05 for reaches 1 through 4, Kp^ should be equal
to 10, 100, 1 and 10. These values are confirmed in the table (Table B-4) below. Exams
internally calculates partition coefficients for biotic solids and for DOC based on empirical
equations. From these equations, KPB should be 100, and KDOC should be 8 Lw /kg. The unitless
partition coefficient K'DOC relating DOC-bound concentrations in mg/Lw to the dissolved DOC
concentrations is the product of KDOC and the reach variable DOC concentration (with a units
adjustment). DOC concentrations for streams, ponds and lakes, and wetlands were set to 1.25,
12.5, and 125, respectively, giving values of K DOC of 10"5, 10"4, and 10"3, respectively.
These partition coefficients are summarized in the table below (Table B-4), along with
the calculated concentrations at the end of the first year from Exams. The relationships between
Table B-4. Organic Chemical Partitioning in WBN 3
Reach,
Compartment
Partition Coefficients
Exams Concentrations
K doc
L/L
Kps
L/kg
Kpb
L/kg
cD
mg/L
CDoc
mg/L
Cs
mg/kg
cB
mg/kg
1: Pond
focsed = 0.05, DOC =12.5 mg/L
water
10"4
10
6.00e-2
6.00e-6
6.00e-1
-
upper sed
10"4
10
102
5.49e-3
5.49e-7
5.49e-2
5.48e-1
lower sed
10"4
10
6.90e-7
6.90e-11
6.90e-6
-
2: Wetland
focsed = 0.50, DOC = 125 mg/L
water
10"3
10
5.00e-2
5.00e-5
5.00e0
upper sed
10"3
102
102
9.49e-3
9.49e-6
9.49e-1
9.49e-1
lower sed
10"3
102
2.34e-6
2.34e-9
2.34e-4
3: Stream
focsed = 0.005, DOC = 1.25 mg/L
water
10"5
10
5.54e-2
5.54e-7
5.54e-2
upper sed
10"5
1
102
2.03e-2
2.03e-7
2.03e-2
2.03e0
lower sed
10"5
1
4.30e-5
4.30e-10
4.30e-5
4: Lake
focsed = 0.05, DOC = 12.5 mg/L
epilimnion
10"4
10
5.53e-1
5.53e-5
5.53e0
hypolimnion
10"4
10
5.48e-1
5.48e-5
5.48e0
upper sed
10"4
10
102
3.65e-1
3.65e-5
3.65e0
3.65e+1
lower sed
10"4
10
8.94e-4
8.94e-8
8.94e-3
B-23
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.8.2 Metals
Test 3.2 is a series of simulations run for metals, as outlined below. The purpose of this
test was to verify that the module reads the proper DOC concentrations and chemical partition
coefficients, and correctly executes the phase partitioning reactions giving the proper proportions
of the dissolved, sediment-sorbed, DOC-bound, and benthos concentrations. In this metals
partitioning test, KPW was set to 1000, KPSed was set to 100, KPB was set to 10, and KDOC was set
to 80 L/kg. Given DOC concentrations of 1.25, 12.5, and 125 for streams, ponds/lakes, and
wetlands, the calculated values of K DOC arelO"4, 10"3, and 10"2, respectively. These partition
coefficients are summarized in the table (Table B-5) below, along with the calculated
concentrations from Exams. The relationships between the dissolved concentration and the
chemical concentrations associated with DOC, solids, and biota exactly match the computed
partition coefficients, verifying the treatment of sorption for metals.
Table B-5. Metals Partitioning in WBN 3
Partition Coefficients
Exams Concentrations
Reach,
K doc
Kps
Kps
cD
CDoc
Cs
cB
Compartment
L/L
L/kg
L/kg
mg/L
mg/L
mg/kg
mg/kg
1: Pond
DOC = 10 mg/L
water
10"3
103
5.44e-2
5.44e-5
5.44e+1
-
upper sed
10"3
102
10
2.69e-2
2.69e-5
2.69e0
2.69e-1
lower sed
10"3
102
3.37e-6
3.37e-9
3.37e-4
-
2: Wetland
DOC = 100 mg/L
water
10"2
103
4.47e-2
4.47e-4
4.447+1
upper sed
10"2
102
10
8.43e-3
8.43e-5
8.43e-1
8.43e-2
lower sed
10"2
102
2.08e-6
2.08e-8
2.08e-4
3: Stream
DOC = 1 mg/L
water
10"4
103
5.00e-2
5.00e-6
5.00e+1
upper sed
10"4
102
10
4.89e-2
4.89e-6
4.89e0
4.89e-1
lower sed
10"4
102
6.18e-5
6.18e-9
6.18e-3
4: Lake
DOC = 10 mg/L
epilimnion
10"3
103
4.97e-1
4.97e-4
4.97e+2
hypolimnion
10"3
103
4.93e-1
4.93e-4
4.93e+2
upper sed
10"3
102
10
4.68e-1
4.68e-4
4.68e+1
4.68e0
lower sed
10"3
102
4.02e-4
4.02e-7
4.02e-2
B.2.9 Volatilization
A series of simulations was run with water body network 3 to confirm that Henry's Law
constant and related environmental properties are passed properly to Exams, and that Exams
calculates volatilization loss rates correctly. Simulation output was compared with calculations
based on equations presented in the documentation report (Ambrose and Burns, 2000; U.S. EPA,
B-24
-------
Appendix B
Verification and Validation Tests for EXAMS
1999). H-Exams uses the Whitman two-resistance model to calculate volatilization loss rate
constants. Exams bases the gas transfer term on the wind-driven water vapor exchange velocity,
which is calculated in the same way for all water body types. The liquid phase transfer term is
based on the oxygen exchange (reaeration) velocity. Different sets of equations are used to
calculate the reaeration velocity for flowing and stagnant water bodies. The volatilization
equations are tested by manipulating input data for test water body network 3 and examining
intermediate calculations for stream reach 3 and pond reach 1.
The Exams variables being verified in Step 8 include the following:
¦ K02, the reaeration velocity
¦ WIND, the wind speed at 10 cm height
¦ Local pseudo-first-order process half-lives for Volatil(ization)
B.2.9.1 Volatilization in Stream Reaches
Test 3.3 is a simulation run for volatile organic chemicals, as outlined below. The
purpose of this test was to verify that the module reads the proper volatilization parameter values
and correctly executes the volatile loss algorithms in stream reaches.
To fully test the range of computations in the volatilization equations, the Henry's Law
constant was varied between 10"2, 10"4 and 10"6atm-m3/mole, and flow between 3500, 106, and
107 m3/day. Wind speed averaged 5 m/sec and water temperature was held constant at 20°C.
Hydraulic width coefficients ahydw and bhydw were set to 10 and 0.25, respectively; and
hydraulic depth coefficients ahydd and bhydd were set to 0.5 and 0.4. Molecular weight was set
to 278. For each simulation test, several volatilization properties were hand-calculated from the
original equations and verified by inspection of the EXAMS summary file. For H of 10"6, the gas
phase resistance should control the overall volatilization rate, while for H of 10"2, the liquid
phase resistance should control; at 10"4, the gas and liquid phase resistances should contribute
almost equally to the overall rate. The results of these tests, summarized in Table B-6 below,
verify that the H-EXAMS stream volatilization routines are performing as expected.
Table B-6. Calculated and Simulated Stream Volatilization Half-lives
Flow,
m3/day
Depth,
meters
H = 10"2 atm-m3/mole
H = 10"4 atm-m3/mole
H = 10"6 atm-m3/mole
Calculated
*1/2. hr
Exams t1/2,
hr
Calculated
*1/2. hr
Exams t1/2,
hr
Calculated
*1/2. hr
Exams t1/2,
hr
3.5x103
0.084
1.5
1.5
4.7
4.7
320
320
10s
1.33
28
28
58
58
3080
3080
107
3.35
70
70
147
146
7747
7740
B.2.9.2 Volatilization in Ponds, Wetlands, and Lakes
Volatilization in stagnant water bodies is driven by wind speed. Exams proportions the
liquid-phase transfer velocity to the reaeration velocity KQ2. To calculate reaeration in stagnant
water bodies, the theoretically-based equations from O'Connor (1983), summarized in Ambrose
and Burns (2000), were used in H-Exams. Ponds and wetlands were assumed to be in the
B-25
-------
Appendix B
Verification and Validation Tests for EXAMS
intermediate size class, while lakes and bays were assumed to be in the large size class. Test 3.4
is a series of simulations run for volatile organics, as outlined below. The purpose of these tests
is to verify that the module reads the proper values, and correctly executes the non-stream
volatile loss algorithms.
The Henry's Law
constant (ChemHLC)
was set to 0.01 atm-
m3/mole. Several
simulations were run
with different average
wind speeds. For each
simulation test,
calculated reaeration
velocities in reach 1
(pond), reach 2 (wetland)
and reach 3 (lake) were
extracted from the
Exams summary output
file. The complex
calculations are checked
against two simpler
empirical
formulas—Banks (1975) Figure B-4. Comparison of wind-driven reaeration formulas
and Baca and Arnett
(1976). For purposes of this comparison, the reaeration velocity K02 was derived from this
formula assuming that the mixed surface layer is 1 meter in depth. K02 values computed from
these independent formulas for a range of wind speeds are summarized in the adjacent chart,
along with the values calculated for ponds and lakes using the O'Connor equations implemented
in H-Exams (with units converted from cm/hr to m/day for this comparison). Inspection of the
chart (Figure B-4) shows that H-Exams reaeration values are within the range of the simpler
formulas, providing a general verification of this model component. Most HWIR sites have
average annual wind speeds between 3 and 6 m/second. In this range, the O'Connor equations
track the Banks equations closely, as summarized in the table (Table B-7) below.
Table B-7. Wind Speed Effect on Reaeration
Wind Speed,
m/sec
Calculated Reaeration Velocity, m/day
H-Exams Pond
H-Exams Lake
Banks
Baca and
Arnett, low
Baca and
Arnett, high
1
0.19
0.24
0.36
0.091
0.87
2
0.37
0.48
0.51
0.18
1.7
3
0.55
0.71
0.63
0.26
2.6
4
0.74
0.97
0.72
0.35
3.5
5
0.93
1.2
0.81
0.43
4.3
6
1.5
1.9
1.0
0.52
5.2
7
2.2
2.8
1.4
0.61
6.1
Wnd-Driven Reaeration
0 10 20 30
VMnd Speed, rrfsec
40
¦ HExamsRrd
(COrncr, 1963)
¦ HExsms Lake
(COrncr, 1963)
Banks (1975)
Baca axl Anett
(1976) iQA/erd
Baca aid Anett
(197E>) hicfiend
B-26
-------
Appendix B
Verification and Validation Tests for EXAMS
B.2.9.3 Effect of Ionization on Volatilization
Test 3.5 is a series of simulations run for volatile organic acids, as outlined below. The
purpose of this test was to verify that speciaton parameters from the HWIR chemical and site
databases were affecting the volatilization rate properly. In this test, Henry's Law constant was
set to 0.01 atm-m3/mole, and average wind speed was set to 5 m/sec. When no ionization is
specified, the volatilization half-lives for the neutral species in reaches 1 through 4 are calculated
to be 60, 24, 1.5, and 46 hr, respectively.
For organic acids, Test 3.5a sets up one dissociation with pKal equal to 9. In subsequent
runs, the value is set to 8, 7, 6, and 5. pH is held constant at 7. For each simulation, the internal
speciation and volatilization results were examined from EXAMS. First, by inspection, the ratio
of the charged species (valency of -1) to the neutral species (valency of 0) compares exactly with
theoretical expectations. Next, the fraction neutral (RH3) was calculated from the individual
species concentrations. This fraction neutral at different pKa values should also equal the ratio
of the volatilization rate constants under neutral and ionizing conditions. This ratio can be
checked by taking the reciprocal of the half-lives under neutral and ionizing conditions in
Exams. Because half lives are reported to only two significant figures, this comparison does not
always match precisely. Nevertheless, to the available accuracy, these results verify that Exams
is receiving the speciation parameters and producing correct speciation and volatilization results
for organic acids (see summary Table B-8 below).
For organic bases, Test 3.5b sets up one dissociation with pKbl equal to 9. In subsequent
runs, the value is set to 8, 7, 6, and 5. pH is held constant at 7. The results verify that EXAMS is
receiving organic base speciation parameters and producing correct speciation and volatilization
results (see summary Table B-9 below).
Table B-8. Speciation and Volatilization of Volatile Organic Acids
PKa1
Theoretical Species Calculations
for pH = 7
Exams Speciation and Volatilization Calculations
RH3/RH2"
fraction RH3
fraction RH3
fraction of neutral kv by reach
9
102
0.990
0.990
1.0
1.0
1.0
.98
8
101
0.909
0.909
.91
.92
.88
.90
7
10°
0.500
0.500
.50
.51
.50
.50
6
10"1
0.0909
0.0909
.091
.092
.088
.091
5
10"2
0.0099
0.0099
.0099
.0100
.0098
.0099
B-27
-------
Appendix B
Verification and Validation Tests for EXAMS
Table B-9. Speciation and Volatilization of Volatile Organic Bases
pKb1
Theoretical Species
Calculations for pH = 7
Exams Speciation and Volatilization
Calculations
rh3/rh4+
fraction RH3
fraction RH3
fraction of neutral kv by reach
9
102
0.99
0.990
1.0
1.0
1.0
.98
8
101
0.91
0.909
.91
.92
.88
.90
7
10°
0.50
0.500
.50
.51
.50
.50
6
10"1
0.0909
0.0909
.091
.092
.088
.091
5
10"2
0.0099
0.0099
.0099
.0100
.0098
.0099
B.2.10 Transformation Processes
A series of simulations was ran with the simple pond (water body network 1) to confirm
that chemical transformation constants along with ancillary environmental information are
passed properly to Exams, and that Exams calculates transformation losses correctly. All
transformation tests were run on the simple pond, water body network 1. Simulation output was
compared with theoretical calculations based on equations presented in the documentation report
(Ambrose and Burns, 2000; U.S. EPA, 1999).
The Exams variables being verified in Step 9 include the following:
¦ Local pseudo-first-order process half-lives for Hydrol(ysis)
¦ Local pseudo-first-order process half-lives for Biolysis (here, aerobic
biodegradation for water column compartments and anaerobic biodegradation for
benthic compartments)
¦ Local pseudo-first-order process half-lives for Reduct(ion)
¦ Total Chemical Concentrations in the Water Column
¦ Dissolved Chemical Concentrations in the Benthic Sediments
These are found in Tables 12, 13, and 15 in report.xms. In addition, chemical
concentrations are confirmed in SWl.ssf.
B.2.10.1 Organic Chemical Transformations
Test 1.3 is a series of simulations run for organic chemicals. The purpose of this series of
tests is to verify that H-EXAMS properly sets ancillary environmental properties to enable the
first-order transformation reactions and transfers transformation rate constants from the HWIR
chemical database to Exams. Further, this series tests whether Exams correctly executes the
indicated transformation reactions, giving the proper chemical half-lives and water body
concentrations.
B-28
-------
Appendix B
Verification and Validation Tests for EXAMS
In each test, one of the HWIR transformation reactions - hydrolysis, aerobic
biodegradation, anaerobic biodegradation, and reduction - was enabled by specifying a value for
the rate constant in the water body chemical database. As a result, chemical transformation half-
lives calculated by Exams should be equal to (24 • 0.693Ik) hours, where k is the transformation
rate constant in day"1. Steady-state chemical concentrations in any EXAMS compartment should
be given by the following equation:
Ct " Q * k-fD-v (B_1)
where Cx is the total chemical concentration, in mg/L, Lx is the total chemical loading, in g/day,
Q is the total flow, in m3/day, k is the transformation rate constant, in day"1, fD is the chemical
dissolved fraction, and V is the water column volume, in m3.
Benthic-water column exchange is set to 0 for this test in order to better approximate the
simple analytical solution. Three Exams transformation rates used here—water column
biodegradation, sediment biodegradation, and reduction—use mixed, second-order kinetics. The
net loss rate constant is the product of a second-order rate constant and an appropriate
environmental property. In this HWIR implementation, the net loss rate constant for each
transformation reaction is supplied in the chemical file. The surface water model interface
ExamsIO transfers the value of this first-order rate constant to the Exams second-order rate
constant and sets the associated environmental property to unity.
In this example pond, the volume is 1000 m3 (Section 3.1.2). Forthefirst3 years, the
total chemical loading is a relatively constant 10 g/day (Section 7.1.2), and flows are 150, 139,
and 160 m3/day (Section 5.1.2). No partitioning information is specified, so the dissolved
chemical fraction is 1.0. When rate constants for water column reactions (hydrolysis and aerobic
biodegradation) are set to 0.693 day"1, half-lives should be 24 hours. From Eq. B-l, water
column total chemical concentrations should equal 0.01186 mg/L, 0.01202 mg/L, and
0.01172 mg/L in years 1, 2, and 3. These are confirmed in the EXAMS output tables and in the
HWIR surface water results file.
Eq. B-l can also be applied to dissolved pore water concentrations in benthic
compartments. In the HWIR application, two sediment layers are simulated for each reach.
Groundwater flow and loadings enter the underlying layer and are advected upward through the
surficial layer and into the water column. H-EXAMS reports chemical concentrations in the upper
sediment layer. To test those calculations, Eq. B-l can be modified and set up in series:
r = LSeep
DM Qs«p + k'fD'r,L-nBL (">
B-29
-------
Appendix B
Verification and Validation Tests for EXAMS
(B-3)
where CDBL and CDBU are the dissolved pore water concentrations in the lower and upper
sediment layers, in mg/L, LSeep is the seepage loading in g/day, QSeep is the seepage flow, in
m3/day, VBL and VBU are the total volumes of the lower and upper sediment layers, in m3, and nBL
and nBU are the porosities of the lower and upper sediment layers, in Lw/L. For the test pond,
Qseep = 5015, VBL = 250, VBU = 60, nBL = 0.40, and nBU = 0.50. LSeep = 40.09 g/day in years 21 -
40, and 50.116 g/day in years 41 - 60. When fD is 1 and k is 0.693, CDBL should equal
0.3356 mg/L in years 21 - 40, and 0.4196 mg/L in years 41 - 60. CDBU should equal
0.2372 mg/L in years 21 - 40, and 0.2966 in years 41 - 60. When the benthic-water column
exchange coefficient E_sw is set to 0 (to eliminate interference from the surface water column),
the surface water output file should report dissolved benthic concentrations equal to CDBU.
The reduction and anaerobic biodegradation half lives of 24 hours (corresponding to rate
constants of 0.693 day"1) are confirmed for the benthic compartments in the Exams output tables,
and the dissolved benthic concentrations are confirmed in both in Exams tables and in the
HWIR surface water output files.
These and similar tests with different rate constants verify that H-EXAMS processes
organic chemical transformation reactions accurately.
B.2.10.2 Mercury Transformations
Test 1.4 is a simulation run for mercury. The purpose of this test was to verify that H-
Exams properly sets up and executes the mercury transformation reactions, including the
transfer of mass to reaction products. Mercury is simulated in H-EXAMS with three interacting
chemical components representing inorganic divalent mercury (Hgll), methyl mercury (MeHg),
and elemental mercury (HgO). Atmospheric and watershed loadings are predominantly in the
form of Hgll. In the water body, Hgll can be reduced to HgO and methylated to MeHg. HgO can
be oxidized back to Hgll and volatilized to the atmosphere. MeHg can be demethylated back to
Hgll and to HgO.
To implement these mercury reactions, H-EXAMS sets the number of chemicals to 3,
assigning component Hgll to chemical 1, MeHg to chemical 2, and HgO to chemical 3. As an
organic chemical model, Exams handles volatilization but does not include process modules for
the mercury transformation reactions. These are handled by passing appropriate mercury
transformation rate constants and yield coefficients to the Exams biodegradation and reduction
reactions for each mercury component.
These tests first confirmed that all chemical properties, rate constants, and reaction
coefficients were transferred to Exams properly. Reaction processes include partitioning,
volatilization, water column methylation, benthic methylation, water column reduction, water
column demethylation, benthic demethylation, water column Mer demethylation, and water
B-30
-------
Appendix B
Verification and Validation Tests for EXAMS
column oxidation. All rate constants, reaction products, and yield coefficients were confirmed in
Exams output tables.
Mercury test simulations were run using steady hydrologic and solids input. All
environmental properties were checked for proper transfer to Exams. Loadings of divalent
mercury from atmosphere deposition and watershed runoff were read and converted to yearly
Exams loadings. In the atmospheric file, 25 years of steady vapor wet deposition at 5.48e-8
g/m2-day (20 ug/m2-year) and particulate dry deposition at 2.74e-8 g/m2-day (10 ug/m2-year) are
specified. In the watershed output file, 20 years of steady runoff loading at 10"3 g/day is
specified. These loadings imply a rainfall mercury concentration of 25 ng/L and a runoff
mercury concentration of 10 ng/L. Given the soil erosion loading of 10s g/day (103 mg/L in
runoff water), the runoff mercury concentration is consistent with a background level soil
mercury concentration of about 5 to 10 ng/g and a soil mercury partition coefficient of 103 to
104 L/kg.
Transformation rate constants are supplied for the dissolved and solids-sorbed phases.
The DOC-complexed phase is nonreactive. To calculate the reaction process half-lives, the
fraction of chemical in the DOC-complexed phase must be calculated. From partitioning
equations and the specified chemical and environmental data, the dissolved plus solids-sorbed
fraction, freactive, of Hgll is calculated to be 0.9855in the water column and 0.99999 in the upper
benthic layer. For methyl mercury, the resulting freactive is calculated to be 0.99206 in the water
column and 0.99963 in the benthic layers.
For a transformation reaction with a first-order rate constant k, the half-life should be
0.69315/(k x freactive). Based on the rate constants and the reactive fractions calculated above,
reaction half-lives should be 33,770 hours for water column methylation, 166,300 hours for
benthic methylation, 1687 hours for divalent mercury reduction, 3359 hours for water column
demethylation, 33,340 hours for benthic demethylation, 16,770 hours for Mer demethylation,
and 16620 hours for oxidation. These transformation half-lives were confirmed in EXAMS output
tables.
Output from EXAMS and in SWl.grf show that mercury concentrations are at steady state
levels in year 20. The distribution of divalent and methyl mercury among competing phases in
the water column and benthic sediments was confirmed against the equilibrium partition
coefficients. While the simulated mercury concentrations cannot be verified exactly against
analytical solutions, they can be checked approximately. The steady-state (year 20) total water
column concentrations for divalent, methyl, and elemental mercury in the output file are
6.318 ng/L, 0.529 ng/L, and 0.131 ng/L, respectively. The total concentration of 6.98 ng/L is
just below the 7.2 ng/L calculated for a conservative chemical as total load divided by total flow.
This makes sense for a simulation where the only chemical losses come from advective export
and volatilization of a small fraction of the total chemical (elemental mercury comprises just
1.88% of the total water column mercury).
To refine this comparison, Eq. B-l can be used for total mercury. The total loadings
come to 1.082xl0"3 g/day, and the flow is 150 m3/day. The elemental mercury volatilization loss
rate of 0.326 day"1 can be calculated from the reported half-life of 51 hours in Table 12 in
report.xms. The elemental mercury fraction is 1.88%. Applying the steady-state analytical
equation with a total mercury loss rate constant of 0.326x0.0188 gives a total mercury
B-31
-------
Appendix B
Verification and Validation Tests for EXAMS
concentration of 6.93 ng/L, which is very close to the year 20 total here of 6.98 ng/L. This
comparison provides a good bottom-line check on many of the mercury calculations in the
surface water module.
B.2.11 Test Robustness
Steps 1 through 9 test individual processes or combinations of processes on three sites.
While these sites represent a range of normal environmental and chemical conditions, they do
not test extreme conditions that might be encountered in the operation of this software for
national risk assessment. It is possible that some combinations of parameter values could cause
unforseen errors in the surface water module. The robustness tests outlined in this section are
designed to screen for module failures due to various combinations of parameter values across
sites, chemicals, and waste management units.
Test series 4, 5, and 6, described in separate sections below, comprise the robustness
testing program for the surface water module. These tests are conducted within the HWIR
production system, as outlined in Section B.2.2. Two different computers were used for test
series 4 and 5. Computer 1 is a Dell Optiplex GS1 with 128 MB RAM and a Windows 98™
operating system. Computer 2 is an IBM Personal Computer 300 PL with 128 MB RAM and a
Windows 95™ operating system. The full HWIR atmospheric database was installed on
computer 1, allowing Monte-Carlo iterations. Test series 6, which requires this feature, was run
on computer 1 but not computer 2.
B.2.11.1 Screening Across Sites
This series of tests involves simulations of selected chemicals in various types of waste
management units across all sites in the database. The purpose of this series of tests is to screen
for fatal errors that might be generated in the surface water module due to extreme combinations
of environmental parameters interacting with different classes of chemicals. The waste pile
(WP) and land application unit (LAU) represent runoff, leaching, and volatile release pathways,
while the landfill (LF) and surface impoundment (SI) are restricted to leaching and volatile
release. The aerated tank (AT) is subject to volatile release only. Low and high waste levels are
chosen for lead to test for problems that might be generated by very small as well as very large
calculated concentrations. The suggested site screening tests are summarized in the table
(Table B-10) below.
Table B-10. Suggested Site Screening Tests
Test
Chemical
Site
WMU
Waste Level
4.1
benzene
all
all
5
4.2
thiram
all
SI, LAU
5
4.3
benzo(a)pyrene
all
LAU, WP
5
4.4
2,4-D
all
SI, LF
5
4.5
lead
all
LAU
1,5
4.6
mercury
all
LAU, WP, LF
5
B-32
-------
Appendix B
Verification and Validation Tests for EXAMS
Benzene was selected to represent volatile organic chemicals, with a Henry's Law
constant of about 0.003 m3-atm/M. Thiram is one of the most reactive chemicals, with an
alkaline hydrolysis loss rate constant that can reach 1 day"1 at high aquatic pH levels close to 9
(even higher rate constants are possible in the waste management units where pH can
exceed 12). Benzo(a)pyrene is one of the most hydrophobic chemicals, with an organic carbon
partition coefficient of around 8x 10s L/kg. 2,4-D is an organic acid that ionizes with a pKa of
about 3, so that most of the chemical is in the anionic form in water bodies. Lead is a cationic
heavy metal with a high partition coefficient of about 10s. Finally, mercury is a complex
bioaccumulating metal with transformation reactions linking its three major components.
For more thorough site coverage, each of the tests here was run with all the WMUs
selected. Two failures in the surface water module were found and corrected during this series
of tests. In addition, some failures in other modules and processors were detected and reported
to appropriate members of the HWIR team. Using the updated modules, no surface water
failures occur during these tests.
B.2.11.2 Screening across Chemicals
This series of tests involves simulations of all chemicals in the database in two types of
waste management units at selected sites. The purpose of this series of tests is to screen for fatal
errors that might be generated in the surface water module due to extreme combinations of
chemical parameters interacting with different classes of environments. Copper and cobalt had
to be excluded from this test, because required data were not available in the current version of
the HWIR system. Furthermore, of the three mercury components, only divalent mercury should
be selected for simulation; methyl mercury and elemental mercury are generated in the water
bodies from divalent mercury. Low and high waste levels are chosen to test for problems that
might be generated by very small or very large calculated concentrations. The suggested
chemical screening tests are summarized in the table (Table B-l 1) below.
Table B-11. Suggested Chemical Screening Tests
Test
Chemical
Site
WMU
Waste Level
5.1
all
0930205
SI
1,5
5.2
all
1333001
LAU
1,5
5.3
all
0724909
WP
1,5
5.4
all
1434022
AT
1,5
5.5
all
0730914
LF
1,5
Site 0930205 is located in rural North Carolina near Badin. This site contains two water
body networks. The first has five reaches, including stream, lake, and wetland reaches; the
second is a lake with no exiting surface flow. Site 1333001 is located in rural South Carolina
near Florence. This site also contains two water body networks. The first is a simple first order
stream, while the second is a 7th order river with two adjacent wetlands. One of the wetlands
receives contaminated seepage over a long period of time from an adjacent LAU. Site 0724909
is located in rural North Dakota near Williston. A single water body network has 11 reaches,
including 2nd and 3rd order streams, 3 lakes, and 2 wetlands. This area is somewhat arid, and
B-33
-------
Appendix B
Verification and Validation Tests for EXAMS
some reaches are subject to periodic drying. Site 1434022 is located in urban California near
Torrance. Two small lakes are located near the facility. These water bodies have no outflow
and are subject to high evaporation. Finally, site 0730914 is located in rural Texas near Elgin.
A single water body network has 18 reaches, including 2nd through 4th order streams and 7 small
lakes. These five sites are illustrated at the end of this section.
No new failures in the surface water module were generated by this series of tests. Some
failures in other modules and processors were detected and reported to appropriate members of
the HWIR team. Using the updated modules and processors from test series 4, this series of tests
runs with no surface water module failures.
B.2.11.3 Monte-Carlo Screening with Random Parameter Values
The final series of tests involves multiple simulations of selected chemicals at selected
sites. The purpose of this series of tests is to screen for fatal errors that might be generated in the
surface water module due to extreme combinations of chemical and environmental parameters.
Different combinations of values are generated by selecting a large number of realizations for
each test. At least 100 are recommended (200 realizations were run here). High waste levels are
chosen to test for problems that might be generated by very large calculated concentrations. The
suggested Monte-Carlo screening tests are summarized in the table (Table B-12) below.
Table B-12. Suggested Monte-Carlo Screening Tests
Test
Chemical
Site
WMU
Waste Level
6.1
2,4-D
0930205
WP
5
6.2
lead
1333001
LAU
5
6.3
benzo(a)pyrene
0724909
WP
5
6.4
benzene
1434022
AT,SI
5
6.5
thiram
0730914
WP
5
No new failures in the surface water module were generated by this series of tests. Some failures
in other modules and processors were detected and reported to appropriate members of the
HWIR team. Using the updated modules and processors from test series 4, this series of tests
runs with no surface water module failures.
B-34
-------
Appendix B
Verification and Validation Tests for EXAIS
This site (Figure B-5) contains three waste management unit (WMU) types, including
aerated tank, surface impoundment, and waste pile. The underlying geology is metamorphic and
igneous rock. Two water body networks are present. The first has five reaches, including
streams, a 5.4 ha lake (reach 3), and a 10.1 ha wetland (reach 5). The second network is a 358 ha
lake with no exiting surface flow. Potential runoff and leaching pathways connect the WMUs to
this lake.
HWIE.GIS Dock
-------
Appendix B
Verification and Validation Tests for EXAIS
This site (Figure B-6) contains a single waste management unit type, a land application
unit (LAU). The underlying geology is sand and gravel. Two water body networks are present.
The first is a two-reach 1 st-order stream. The second network is a 7th-order river with two
adjacent wetlands covering 630 ha and 936 ha, respectively. Runoff and leaching pathways
connect the LAU to wetland reach 3.
HWIR OB Eocket Map!
Water body /Water shed
Data for Site 1333001
Largest WMU type at site: lau
Legend
Sites
I I AOI
I W WMU
Lakes and Wetlands
Streams/Rivera
Local Watersheds
~ Watersheds
Inset shows site with HU C and RF1
August j, iyyy
Figure B-6. Site 1333001, located in rural South Carolina near Florence.
B-36
-------
Appendix B
Verification and Validation Tests for EXAIS
This site (Figure B-7) contains a single waste management unit type, a waste pile (WP).
The underlying geology contains unconsolidated and semi-consolidated shallow aquifers. A
single water body network has 11 reaches. A 2nd order stream (reaches 3 and 5) drains directly
to a 341 ha section of a large lake (reach 7). A 94 ha wetland (reach 8) also drains to the large
lake. A 3rd order stream (reaches 1, 3, 4, and 6) ends in a 69 ha lake (reach 9), which spills into
wetland reach 10, and then to the large lake. A small 13 ha lake (reach 11) connects to stream
reach 6. Runoff and leaching pathways connect the WP to stream reach 4.
HWIR GIS Docket Maps
Water body/Water shed
Data for Site 0724909
Largest WMU type at site: wp
Legend
• Sites
I I API
W WMU
Lakes and Wetlands
7\j Streams/Rivers
J Local Watersheds
~ Watersheds
August 3,1999,
Figure B-7. Site 0724909, located in rural North Dakota near Williston.
B-37
-------
Appendix B
Verification and Validation Tests for EXAIS
This site (Figure B-8) contains two waste management unit (WMU) types, including
aerated tank and surface impoundment. The underlying geology is sand and gravel. Two small
lakes (1 and 1.7 ha) with no exiting surface flows are present. No runoff or leaching pathways
are active at this site. Contaminants can reach the lakes by volatilization and atmospheric
deposition.
WBN
WBN
HWIR '3IS Docket Maps
Waterbody/W:ater shed
Data Tor Site 1434022
Largest WMU type at site: si
Legend
• Sites
I I API
fw' WMU
Lakes and Wetlands
Streams/Rivers
Local Watersheds
n Watersheds
AugiM 2,1999 m
Figure B-8. Site 1434022, located in urban California near Torrance.
B-38
-------
Appendix B
Verification and Validation Tests for EXAIS
This site (Figure B-9) is contains four waste management unit (WMU) types, including
aerated tank, landfill, surface impoundment, and waste pile. The underlying geology is bedded
sedimentary rock. A single water body network has 18 reaches, including eleven 2nd through 4th
order streams and seven small lakes, ranging from 1.5 to 12 ha in size. Potential runoff and
leaching pathways connect the WMUs to stream reach 11.
HWIRGIS Docket Maps
1— Water body/Water shed
Data for Site 0730914
Largest WM LI type at site: If
Legend
• Sites
I I API
I wl WMU
H Lakes and Wetlands
A/ Streams,levers
T ocal Watersheds
Watersheds
Inset shnws site witli HT1 f! jtiil RF 1
AnguslS, 1? 99,
Figure B-9. Site 0730914, located in rural Texas near Elgin.
B-39
-------
Appendix B
Verification and Validation Tests for EXAMS
B.3. Exams System Validation Case Studies
B.3.1 Overview
Since Exams was released in 1983, there have been many applications to water bodies
across the world. This chapter summarizes applications that report model performance against
measured data in either a calibration or validation mode. The environments include small
streams, rivers, ponds, rice paddies, and bays. Chemicals include dyes, herbicides, insecticides,
phenols, and other organic chemicals with a variety of chemical properties. Fate processes
include advection, sorption, sediment-water exchange, volatilization, hydrolysis, photolysis,
water column and benthic biodegradation, and oxidation. Chemical coefficients have been
supplied from the open technical literature and, in some cases, from site-specific experiments.
The case studies are reported in chronological order, then discussed in general terms at the end
of this chapter.
B.3.2 Linear Alkylbenzene Sulfonate in a Small Stream
Games, L.M., 1982. Field validation of Exposure Analysis Modeling System (Exams) in
a flowing stream. In K.L. Dickson, A.W. Maki, and J. Cairns, Jr., editors, Modeling the Fate of
Chemicals in the Aquatic Environment. Ann Arbor Science Publishers, Ann Arbor, Michigan.
Exams was applied to model the fate of linear alkylbenzene sulfonate (LAS) discharged
in treated sewage effluent into Rapid Creek, a small mountain stream in South Dakota with
relatively constant regulated flow. The model network was composed of 5 water column
segments, each overlying a benthic segment. Segment lengths were set to end at stream
monitoring stations, and varied from 4.7 to 39.2 km. Water column depths varied between 0.43
and 0.56 m, while sediment depths were set at a constant 3 cm. The important chemical fate
reactions for LAS are partitioning to solids and biodegradation, both of which vary with the
isomer chain length. Average chain length in Rapid Creek is about 12, and so measured reaction
coefficients for this homolog were used in the model runs. Partition coefficients to sediment and
to bacteria were 330 and 3070 L/kg. First order biodegradation rate constants measured below
the discharge at 24 C were 0.021 hr"1 for the aqueous phase, and 0.040 hr"1 for the sediment-
sorbed phase. These rate constants were applied to the first four model segments. Lower
measured rate constants from an upstream station (0.002 hr"1 and 0.01 hr"1, respectively, for
aqueous and sediment phases) were applied to the most downstream model segment. Based on
previous work, a Q10 temperature correction factor of 2 was applied.
The model was applied to data from an October 1980 survey, when stream flow was at a
constant 1 m3/sec and water temperature ranged between 4 C and 7 C. The upstream boundary
concentration was set to the observed LAS value at the first monitoring station. The sediment-
water mixing coefficient was treated as a calibration coefficient. The best fit between model
results and survey data was obtained with a value of 5x 10"5 m2/hr (1.4><10"4 cm2/sec).
Qualitatively, the agreement between predicted and measured concentrations was described as
excellent. The plotted results show predicted values within 10% to 50% of the water column
survey data, and predicted values within 10% to 100% of the sediment data. The most sensitive
model parameters were found to be the sediment-water dispersion coefficient and the sediment
biodegradation rate constant. Because these are two of the least-understood parameters, the
authors suggest that "wide error limits would need to be placed on predicted LAS concentrations
B-40
-------
Appendix B
Verification and Validation Tests for EXAMS
... using Exams, and probably any evaluative model in a safety assessment." Nevertheless, the
authors conclude that "given the proper assumptions, Exams can successfully predict the
concentration of LAS resulting from its steady-state input to a flowing stream."
B.3.3 Phenol in the Monongahela River
Pollard, J.E., and S.C. Hern, 1985. A field test of the Exams model in the Monongahela
River. Environmental Toxicology and Chemistry 4, 361-369.
Exams was tested against phenol data in an 18-mile reach below a steel plant discharge
to the Monongahela River, Pennsylvania. The model network was composed of 5 water column
segments, each overlying a benthic segment. Measured flow and depth during the two early
October field surveys averaged 46.4 m3/sec and 3.8 m. Chemical properties and rate constants
for degradation of phenol were obtained from the literature, which indicates that the primary
environmental pathways for phenol are biodegradation, metal-catalyzed oxidation, and
photooxidation. Measured microbial levels in the river were low, averaging 935 cells/mL in the
water and 3 x 1010 cells/mL in the sediment. Total oxidant concentrations could not be measured,
so runs were made with high and low values from the literature (10"9 to 10"8 mol/L). Phenol
loadings from the steel plant effluent were found to be variable, so the average well-mixed
concentration at the first river station downstream of the plant was used to estimate the average
loading rate for this study.
Survey and model results indicated that little or no phenol is present in the sediment
downstream of the plant, and phenol in the water column experiences rapid decay. Observed
water column phenol concentrations were highly correlated with the high-oxidant predictions
(correlations of 0.93 and 0.99 for the two surveys). Chi-square tests, however, indicated a
significant difference between observed data and predictions for the first survey. Comparing
average concentrations by station, predictions appear to fall within a factor of 1.5 to 3 of
observations in the first survey, and within a factor of less than 2 for the second survey. The
authors observe that rapid phenol decay could be caused by the release of oxidants from the steel
plant itself, which would undergo first-order decay downstream. Specifying spatially-variable
oxidant concentrations in accordance with this scenario resulted in a nearly perfect fit of
observed and predicted concentrations by station.
B.3.4 Disperse Yellow 42 Dye in a Pond
K.-W. Schramm, M. Hirsch, R. Twele, and O. Eutzinger, 1988. Measured and modeled
fate of Disperse Yellow 42 in an outdoor pond. Chemosphere 17(3):587-595.
Exams was tested against Disperse Yellow 42 dye introduced to an experimental outdoor
pond 24.5 m2 in area and 1.5 m deep. The model network included a water column segment and
two benthic layers each 1 cm deep. Laboratory experiments were conducted to measure second
order rate coefficients and environmental parameters characterizing this dye's biological and
photolytic degradation. No data are given for the dye's partition coefficient. A loading of dye
was added to the pond, and the concentration was tracked in the water and sediment layers over a
period of 3000 hours (125 days). During this period, water column concentrations declined from
110 ppb to 10 ppb, while concentrations in the upper benthic layer first increased to 1000 ppb
B-41
-------
Appendix B
Verification and Validation Tests for EXAMS
over the first 12 days, then declined to 40 ppb by 70 days. Dye concentrations in the second
layer increased to about 150 ppb, then declined to 15 ppb by day 50.
Exams water column calculations were close to observations near the beginning of the
experiment, but did not decline as rapidly over time as the prototype. After about day 20,
Exams water column calculations exceeded the observations by about 20% to 80%. In the upper
sediment layer, Exams calculations fell below observed concentrations by about an order of
magnitude at the beginning of the experiment, but declined more slowly than the observed data;
Exams matched the final observation on day 70. In the lower sediment layer, Exams
calculations fell below observations by over 2 orders of magnitude. The authors observe that the
Exams transport to the sediment was not realistically simulated. They did not report the value
use for the key model parameters governing this process, the sediment-water mixing coefficient
and the partition coefficient. Nevertheless, they characterized the overall model performance as
good.
B.3.5 Chlorophenol, Chloroquaiacol, and Chlorocatechol in Norrsundet Bay
(Sweden)
K. Kolset and A. Heiberg, 1988. Evaluation of the fugacity (FEQUM) and the Exams
chemical fate and transport models: a case study on the pollution of the Norrsundet Bay
(Sweden). Wat. Sci. Tech. 20(2): 1-12.
Exams was calibrated and tested with three chemicals from a kraft mill effluent to a
heavily polluted bay on the east coast of Sweden. The model network representing an inner and
outer harbor was composed of 6 water column segments from 2.6 to 25 m deep, each overlying a
10 cm benthic segment. Based on empirical methods, water column segments were linked with
estimated exchange flows ranging from 0.9 to 27 million m3/hr. An inner harbor loop flow was
estimated to be 0.072 million m3/hr, and an outer harbor longshore current was estimated to be
11 m3/hr. Chloroform data taken in November 1983 were used with the Exams to calibrate the
network flows. Exchange flows in the outer harbor were reduced by a factor of 5; in the inner
harboer, one exchange flow was reduced 50%, while the other was unchanged.
Four compounds identified in the effluent were used to test EXAMS-2,4,6-trichlorophenol
(2,4,6-TrCP), 3,4,5-trichloroquaiacol (3,4,5-TrCG), tetrachloroguaiacol (TeCG), and
tetrachlorocatechol (TeCC). Chemical properties and oxidation rate constants were specified
from the literature. 2,4,6-TrCP has a Henry's Law constant of 4x 10"6 atm-m3/mole, a log Kow of
3.61, and an oxidation rate constant of 10"3 hr"1. 3,4,5-TrCG has a log Kow of 4.13 and a Henry's
Law and oxidation rate constants of 0. TeCG has a log Kow of 4.42 and Henry's Law and
oxidation rate constants of 0. TeCC has a log Kow of 4.19, Henry's Law constant of 0, and
oxidation rate constants of 0.5><10"2 hr"1.
Exams was run using the calibrated network flows, measured loading rates, and the
chemical data. Calculated and measured concentrations agree well for 2,4,6-TrCP, 3,4,5-TrCG,
and TeCG, but not for TeCC. For the first three compounds, most predictions lie between the
minimum and maximum measured within each compartment. Linear regressions yield slopes
between 0.6 and 0.75, and correlation coefficients between 0.86 and 0.97. Given the uncertainty
in environmental and chemical input data, the authors conclude that this agreement is good.
B-42
-------
Appendix B
Verification and Validation Tests for EXAMS
For TeCC, predicted concentrations are about 10 times higher than observed
concentrations in the inner harbor. This compound is said to have a particularly high affinity for
suspended particles from the effluent, and a significant degree of settling is likely. Exams does
not account for this loss process, and thus overpredicts TeCC concentrations.
The authors conclude that Exams is capable of predicting the concentrations of some
selected chlorophenolics in the Norrsundet area with reasonable accuracy. Exams describes all
important transformation and loss processes except for sedimentation reasonably well.
B.3.6 Three Herbicides and an Insecticide in Rice Paddies
Woodrow, J.E., M.M. McChesney, and J.N. Seiber, 1990. Modeling the volatilization of
pesticides and their distribution in the atmosphere. Pages 61-81 in: Kurtz, D.E., Editor, Long
Range Transport of Pesticides,. Lewis Publishers, Inc., Chelsea, Michigan.
Volatilization flux was measured for three rice herbicides—MCPA, molinate, and
thiobencarb—and one insecticide—methyl parathion—from a laboratory chamber and two
flooded rice fields. The flux measurements were compared to predicted fluxes using Exams
with chemical properties and chamber or field conditions as input. The chamber measured flux
rates from a sample dish filled with aqueous solutions of the chemicals, with a humidified air
stream velocity of 2.2 m/sec. The rice fields were 37 and 41 hectares in area with water depth
maintained at 15-26 cm. Molinate and thiobencarb were applied to the fields as a dry granular
formulation which sank to the bottom sediment where the herbicides slowly dissolved. MCPA
and methyl parathion were applied as emulsified aqueous suspensions which remained largely in
the rice water after application, followed by slow partitioning and breakdown. Chemical
properties, including water solubility and vapor pressure were specified from the literature.
Henry's Law constant was calculated from these two properties.
The calculated Henry's Law constants imply the following order of decreasing volatility:
molinate > thiobencarb > methyl parathion » MCPA » MCPA (DMA salt)
This order was observed experimentally in the laboratory chamber and predicted by EXAMS.
The normalized volatilization flux values predicted by Exams compared well overall with the
observed values, within 10%-20% for molinate, a factor of 2 for MCPA acid., and a factor of 3
for methyl parathion. The calculated flux rate for thiobencarb, however, was low by a factor
of 5. The authors attribute this discrepancy to possibly incorrect vapor pressure and/or solubility
data.
The model was further compared with measured diurnal variations in flux from the two
rice paddies for molinate and methyl parathion. Using input field conditions (including diurnal
wind speed and water temperature), Exams was able to reproduce the observed variations in
volatilization flux for these two chemicals. The observed and predicted water and sediment
concentrations were not reported in this paper. Neither EXAMS nor the laboratory chamber were
able to predict the volatilization of MCPA from the rice paddies. This is attributed to the
observation that MCPA mainly volatilized from dry deposits on soil and plant surfaces.
B-43
-------
Appendix B
Verification and Validation Tests for EXAMS
The authors conclude that Exams "appeared to be promising as a predictive tool for
estimating volatilization, when the appropriate chemical properties and environmental conditions
were used as input data."
B.3.7 Xylenes, Dichlorobenzenes, Styrene, and 4-phenyldodecane in a small
lowland river in U.K.
Tynan, P., C.D. Watts, A. Sowray, and I. Hammond, 1991. The transport and the fate of
organic pollutants in rivers II - Field measurement and modelling for styrene, xylenes,
dichlorobenzenes and 4-phenyldodecane. In: Proceedings of the Sixth European Symposium
Organised within the Framework of Concerted Action Organic Micropollutants in the Aquatic
Environment, Angletti G. and Bjorseth A. (eds.), Cost 641 Working Party 2, Lisbon, Portugal,
22-24 May 1991.
Exams was tested against a variety of chemical data in a 7 km reach below a sewage
treatment works effluent to a small, unnamed lowland river in England.. The model network
was composed of 9 water column segments, each overlying a 10 cm deep benthic segment. A
survey was conducted following a plug of water downstream. The survey date was not
mentioned in this paper, nor were river flow and water column depths (except for a value of
0.4 m in the sixth reach and the observation that travel time through the 7 km reach is about
7 hours). Reach properties were given for key parameters, including pH (7.6-7.7), water
temperature (15-20 C), suspended solids concentration (6-10 mg/L), water column organic
carbon fraction (6%-36% near the effluent), benthic organic carbon fraction (0.4%—1.1%), and
sediment bulk density (1.9-2.1 g/mL).
A larger than expected mass fraction of each chemical was found to be sorbed to
suspended solids, suggesting non-equilibrium partitioning. This violates the assumption in
Exams of local equilibrium. By contrast, bed sediment concentrations were found to be closer
to equilibrium with the water column. Theoretical calculations indicate that several days would
be required for these chemicals to reach a new sorption equilibrium following discharge to the
river. To handle these observations, the Kow value for each compound was increased by 3 orders
of magnitude, while the bed sediment organic carbon fractions were decreased by 3 orders of
magnitude. This enabled Exams to calculate proper chemical concentrations on both suspended
and benthic solids.
In general, Exams predictions are described as being reasonably close to measured
values, with "fairly close predictions" for styrene, m-dichlorobenzene, and p-dichlorebenzene.
Total water column concentrations illustrated in the figures in the study appear to be
within 50% or better for these compounds at most river stations. Predictions of 4-
phenyldodecane concentrations are within an order of magnitude of observations, but do not
match the rapid downstream decline that was observed. The authors attribute this to a lack of
data on this compound, leading to an unquantified degradation process. They conclude that
Exams produces quantitative predictions that compare well with observations for those
chemicals for which reliable rate data exist.
B-44
-------
Appendix B
Verification and Validation Tests for EXAMS
B.3.8 Aniline and Lindane in River Calder
Cousins, I.T., C.D. Watts, and R. Freestone, 1995. Field measurement and modelling the
fate of aniline and lindane in a UK lowland river. Environmental Technology, 16, 515-526.
Exams was tested against aniline and lindane data in a 7 km reach below a sewage
treatment works effluent to the large lowland River Calder in West Yorkshire, England.. The
model network was composed of 9 water column segments, each overlying a benthic segment.
Surveys were conducted in April and August, 1991. Measured river flow and depth were not
described in the article. The river is described as turbid, with suspended solids concentrations
between 15 and 40 mg/L in the two surveys. Chemical properties for aniline and lindane were
specified from the literature. The loss pathways for aniline are biodegradation and photolysis.
Specified properties were Henry's Law constant of 2.53xl0"4 atm-m3/mole, log Kow of 0.90,
first-order biodegradation rate constant of 6.85x 10"3 hr"1, and first-order surface photolysis rate
constant of 2.57x 10"3 hr"1. The loss pathways for lindane are hydrolysis and biodegradation.
Specified properties were Henry's Law constant of 1.87/10"3 atm-m3/mole, log Koc of 3.43,
first-order hydrolysis rate constant of 4.5 x 10"3 hr"1, first-order biodegradation rate constant of
4.0xl0"4hr"1, and first-order surface photolysis rate constant of 4.0xl0"4 hr"1.
Exams predicted very slight losses for both chemicals, a result consistent with both
surveys. Fairly good correlations were achieved between measured and predicted dissolved
water column concentrations, with predictions falling within a factor of 2 of station means.
Measured suspended particulate concentrations in the river were significantly higher than
predicted, probably because of limitations in the equilibrium partitioning assumption. The
measured high particulate concentrations from the STW appeared to be either irreversibly or
strongly bound with a slow desorption rate. By contrast, predicted bed sediment concentrations
were "a little lower than measured," all falling within an order of magnitude of observations.
From this study, the authors conclude that EXAMS is useful for predicting the fate of non-ionic
organic chemicals in rivers, provided that adequate physicochemical and environmental data are
available.
B.3.9 Bensulfuron Methyl and Azimsulfuron in Rice Paddies
Armbrust, K.L., Y. Okamoto, J. Grochulska, and A.C. Barefoot, 1999. Predicting the
dissipation of bensulfuron methyl and azimsulfuron in rice paddies using the computer model
EXAMS2. J. Pesticide Sci. 24(4), 357-363.
Exams was first calibrated to bensulfuron methyl (BSM) and azimsulfuron (AZM) data
taken in experimental lysimeters that contained 5 cm of water overlying 50 cm of paddy soil.
Initial simulations were run assuming that these pesticides are degraded by hydrolysis and direct
photolysis in the water, and by hydrolysis and biological metabolism in the soil. Initial results
led to the specification of an additional water column degradation pathway, indirect photolysis,
which photochemically produces hydroxyl radicals that degrade the pesticides. An average
hydroxyl radical concentration of 7.6x 10"16 M was inferred from calibration of EXAMS to the
lysimeter studies, as was the water - soil dispersion coefficient of 3.8xl0"6m2/hr.
Exams was then set up for two field sites, Ushiku (0.09 ha) and Nihonmatsu (0.10 ha).
Soil organic carbon content for these paddies was 9.74% and 1.55%, respectively. Water
B-45
-------
Appendix B
Verification and Validation Tests for EXAMS
column and soil depths were 5 cm each. Initial pesticide applications of BSM and AZM were
30 g/ha, and 6 g/ha (active ingredients). Chemical properties of the pesticides were specified,
along with the two calibrated variables. Exams successfully predicted the partitioning and
degradation reactions that led to the observed 3-4 day chemical half-lives in paddy water. The
indirect photolysis rate using calibrated hydroxyl radical concentration accounted for a
significant fraction of the degradation. Predicted water column concentration responses
generally matched the observed data. Exams overestimated soil concentrations by factors of 2
to 4. The authors conclude that a "definitive characterization of the rate of degradation and
mobility of the two chemicals at a specific site would require additional information on
environmental parameters and site-specific soil-chemical interactions."
B.3.10 Discussion of Case Studies
Exams requires a combination of environmental, chemical, and loading data in order to
properly specify the model parameters. Erroneous, uncertain, or missing data can result in
improper model parameterization, which leads to errors in model predictions. Overly-simple
process equations can also lead to errors in model predictions. The case studies here highlight
both parameter uncertainty and model uncertainty. Despite these sources of uncertainty, it
appears that Exams is able to predict the concentrations of most organic chemicals within a
factor of 2 or better in the water column, and within an order of magnitude in the sediment.
Advection and dispersion are transport processes that affect all chemicals. The first step
in a model application is characterizing the water body network, including geometry, flows and
exchanges. Specifying geometry and flows is often straightforward, such as for gaged rivers or
simple ponds. For some water bodies such as bays, however, gaps in the basic hydrogeometry
and transport data present a large degree of uncertainty. In such cases, flows were inferred from
tracer studies or from comparing predicted chemical concentrations to observations. For almost
all aquatic environments, sediment-water column exchange is a poorly-characterized process that
is quantified by uncertain parameters, including dispersion coefficient and sediment mixing
depth. This process is often an important contributor to chemical fate. Best results are achieved
when the coefficients are calibrated with site-specific data.
Sorption is another process that significantly affects the fate of many chemicals. For
many organic chemicals, the assumption of local equilibrium between phases governed by the
organic carbon partition coefficient and the solids organic carbon content is robust. Some case
studies, however, highlight the limitations of this equilibrium partitioning process model. When
the discharge is characterized by strong chemical sorption and the travel time through the water
body is relatively brief, the partitioning model tends to underpredict the sorbed phase
concentration in the water column. This can be described as a model error rather than a
parameterization error, although clever assignment of partition coefficients can improve model
performance satisfactorily.
Volatilization is a process that significantly affects a subset of chemicals characterized by
a high Henry's Law constant. The case studies indicate that Exams is able to predict volatile
fluxes reasonably well (within a factor of 2 or 3) if the relevant chemical parameters are
accurate. These parameters include Henry's Law constant, solubility, and vapor pressure.
B-46
-------
Appendix B
Verification and Validation Tests for EXAMS
Any of the chemical transformation processes can become important if their reaction
half-lives are short compared with the residence time in the water body. Biodegradation was
most often measured and applied in these case studies as a first-order reaction. Measuring and
applying benthic biodegradation rate constants seems to be relatively uncertain. Oxidation rates
are a function of the environmental oxidant concentrations, which can be specified or calculated
by a set of photolysis algorithms. These environmental parameters are relatively uncertain, with
typical values varying perhaps an order of magnitude among water bodies. Consequently, site-
specific calibration seems necessary to improves model performance when this reaction is
controlling chemical fate. First-order hydrolysis and photolysis rate constants were used in some
of these case studies with low apparent uncertainty.
References
Armbrust, K.L., Y. Okamoto, J. Grochulska, and A.C. Barefoot, 1999. Predicting the dissipation
of bensulfuron methyl and azimsulfuron in rice paddies using the computer model
Exams2. J. Pesticide Sci. 24(4), 357-363.
Cousins, I.T., C.D. Watts, and R. Freestone, 1995. Field measurement and modelling the fate of
aniline and lindane in a UK lowland river. Environmental Technology, 16, 515-526.
Games, L.M., 1982. Field validation of Exposure Analysis Modeling System (Exams) in a
flowing stream. In K.L. Dickson, A.W. Maki, and J. Cairns, Jr., editors, Modeling the
Fate of Chemicals in the Aquatic Environment. Ann Arbor Science Publishers, Ann
Arbor, Michigan.
Kolset, K. and A. Heiberg, 1988. Evaluation of the fugacity (FEQUM) and the Exams chemical
fate and transport models: a case study on the pollution of the Norrsundet Bay (Sweden).
Wat. Sci. Tech. 20(2): 1-12.
Pollard, J.E., and S.C. Hern, 1985. A field test of the Exams model in the Monongahela River.
Environmental Toxicology and Chemistry 4, 361-369.
Schramm, K.-W., M. Hirsch, R. Twele, and O. Eutzinger, 1988. Measured and modeled fate of
Disperse Yellow 42 in an outdoor pond. Chemosphere 17(3):587-595.
Tynan, P., C.D. Watts, A. Sowray, and I. Hammond, 1991. The transport and the fate of organic
pollutants in rivers II - Field measurement and modelling for styrene, xylenes,
dichlorobenzenes and 4-phenyldodecane. In: Proceedings of the Sixth European
Symposium Organised within the Framework of Concerted Action Organic
Micropollutants in the Aquatic Environment, Angletti G. and Bjorseth A. (eds.), Cost 641
Working Party 2, Lisbon, Portugal, 22-24 May 1991.
Woodrow, J.E., M.M. McChesney, and J.N. Seiber, 1990. Modeling the volatilization of
pesticides and their distribution in the atmosphere. Pages 61-81 in: Kurtz, D.E., Editor,
Long Range Transport of Pesticides, Lewis Publishers, Inc., Chelsea, Michigan.
B-47
-------
Appendix B
Verification and Validation Tests for EXAMS
This page intentionally left blank
B-48
-------
Appendix C
Verification and Validation of the
EPA's Composite Model for Transformation Products
(EPACMTP), and its Derivatives
-------
This page intentionally left blank.
-------
Appendix C
Verification and Validation ofEPACMTP
Appendix C
Verification and Validation of the
EPA's Composite Model for Transformation Products
(EPACMTP), and its Derivatives
Work Assignment Manager
and Technical Direction:
Dr. Zubair A. Saleem
U.S. Environmental Protection Agency
Office of Solid Waste
Washington, DC 20460
Prepared by:
HydroGeoLogic, Inc.
1155 Herndon Parkway, Suite 900
Herndon, VA 20170
Under Subcontract No.: RMC-B-00-021
Resource Management Concepts, Inc.
46970 Bradley Blvd., Suite B
Lexington Park, MD 20653
Under Contract No.: 68-W-01-004
Work Assignment 0-1
U.S. Environmental Protection Agency
Office of Solid Waste
Washington, DC 20460
and
-------
Appendix C
Verification and Validation ofEPACMTP
ACKNOWLEDGMENTS
Several individuals have been involved with this work. Dr. Zubair A. Saleem of the U.S. EPA, Office
of Solid Waste, provided overall technical coordination and review throughout this work. Mr. John
Hendrick of Resource Management Concepts, Inc. (RMC) provided overall project coordination and
quality control. This report was prepared by Ms. Sarah Frost of HydroGeoLogic, Inc. (HGL). The report
was reviewed by Drs. Varut Guvanasen and Jan Kool (HGL).
11
-------
Appendix C
Verification and Validation ofEPACMTP
TABLE OF CONTENTS
Page
C.l Introduction C-8
C.l.l General Background C-8
C.1.2 Definition of Verification and Validation C-8
C.2 Technical Background of the Vadose-Zone and Aquifer Modules C-10
C.2.1 EPACMTP C-10
C.2.2 3MRA Subsurface Flow and Transport Modules C-10
C.3.1 ORD Verification (1992-1993) C-ll
C.3.2 Module-Level Verification (1993-1994) C-12
C.3.2.1 Vadose-Zone Module Verification C-13
C.3.2.2 Aquifer Module Verification C-13
C.3.2.3 Metals Transport Module C-16
C.3.3 Verification of Individual Modules and a Composite Model in
EPACMTP (1997) C-16
C.3.3.1 Vadose-Zone Module Verification C-20
C.3.3.2 Aquifer Module Verification C-20
C.3.3.3 Composite Model Verification C-20
C.3.4 Verification of 3MRA Subsurface Flow and Transport Modules (1999) C-20
C.3.4.1 Vadose-Zone Module Verification C-21
C.3.4.2 Aquifer Module Verification C-21
C.3.4.3 Pseudo-3-D Module Verification C-21
C.3.5 Comprehensive Verification of the 3MRA Vadose-zone Pseudo-3-D
Aquifer Modules (2000) C-24
C.3.5.1 Vadose-Zone Module Verification C-24
C.3.5.2 Aquifer Module Verification C-24
C.4 Validation History C-26
C.4.1 Borden Site C-26
C.4.2 Long Island Site C-26
C.4.3 Dodge City Site C-27
C.4.4 EBOS Site C-27
C.5 Summary C-28
C.6 References C-29
in
-------
Appendix C
Verification and Validation ofEPACMTP
TABLE OF CONTENTS (continued)
Page
Attachment
C.l Vadose-Zone Module Verification Results (1993-1994) C.l-1
C.2 Aquifer-Module Verification Results (1993-1994) C.2-1
C.3 Results (1993-1994) C.3-1
C.4 Vadose-Zone Module Verification Results (1997) C.4-1
C.5 Aquifer Module Verification Results (1997) C.5-1
C.6 Composite Model Verification (1997) C.6-1
C.7 3MRA Vadose-Zone Module Verification Results (1997) C.7-1
C.8 3MRA Analyses Module Verification Results (1997) C.8-1
C.9 3MRA Pseudo 3-D Aquifer Module Verification Results (1999) C.9-1
C. 10 3MRA Vadose-Zone Module Verification Results (2000) C. 10-1
C. 11 3MRA Aquifer Module Verification Results (2000) C. 11-1
C.12 EPACMTP Validation Results C.l 2-1
IV
-------
Appendix C
Verification and Validation ofEPACMTP
LIST OF TABLES
Page
Table
C.3-1 Summary ofEPACMTP Verification by the Office of Research and Development,
U.S. EPA, and Tetra Tech, Inc. in 1992-1993 (from U.S. EPA, 1992) C-12
C.3-2 Verification Cases for the Vadose-Zone Module (1993-1994) C-14
C.3-3 Verification Cases for the Saturated Zone Module (1993-1994) C-15
C.3-4 Verification Cases for Metals Transport in the Vadose-Zone Module (1993-1994) C-17
C.3-5a Verification Cases for the Vadose-Zone Module (1997) C-18
C.3-5b Verification Cases for the Aquifer Module (1997) C-18
C.3-5c Verification Cases for Composite Module (1997) C-19
C.3-6b Verification Cases for the 3MRA Aquifer Module (1999) C-23
C.3-6c Verification Cases for the 3MRA Pseudo-Three Dimensional Aquifer Module (1999) . . C-23
C.3-7a Verification Cases for the 3MRA Vadose-Zone Module (2000) C-25
C.3-7b Verification Cases for the 3MRA Aquifer Module (2000) C-25
v
-------
Appendix C
Verification and Validation ofEPACMTP
This page intentionally left blank.
vi
-------
Appendix C
Verification and Validation ofEPACMTP
C.l Introduction
C.l.l General Background
The U.S. Environmental Protection Agency's (EPA) Office of Solid Waste (OSW) has
developed a probabilistic (Monte Carlo) groundwater flow and transport modeling approach to assess
potential exposure of groundwater to toxic chemical constituents in wastes that are managed in
Subtitle D industrial waste management units under RCRA regulations. The exposure to
groundwater is expressed as concentrations of potential contaminants at a drinking water well
(receptor well) located downgradient of the waste disposal facility. This modeling methodology has
been incorporated into the EPA's Composite Model for Leachate Migration with Transformation
Products Code (EPACMTP) (U.S. EPA, 1996a, b, c, d). The U.S. EPA OSW has applied
EPACMTP, as a general fate and transport model, to establish regulatory levels for concentrations of
chemicals in the Subtitle D industrial waste management units for several proposed rules and listing
determinations.
In 1999, the flow and transport components for the vadose-zone and aquifer modules were
extracted from EPACMTP for U.S. EPA's Hazardous Waste Identification Rule (HWIR99)-
Multimedia, Multipathway, and Multireceptor Assessment (3MRA). The pseudo-3-D module was
developed for the aquifer during this period (U.S. EPA, 1999a). At the same time, three ancillary
modules were developed to include the effects of fractures, heterogeneity, and anaerobic
biodegradation. The technical framework of the flow and transport modules along with the three
ancillary modules were subject to peer and public reviews. A number of comments were received.
The comments have recently been compiled and priorities assigned. One of the high priority issues is
the verification and validation of various 3MRA simulation modules. Prior to performing additional
verification and validation to the 3MRA vadose-zone and aquifer modules, it is necessary to compile
all pertinent information relating to the verification and validation of the 3MRA modules and their
predecessors in the past. Details of the verification and validation exercises conducted in the past
decade are summarized in this document.
C.1.2 Definition of Verification and Validation
For a simulation module to gain credibility that it can be used to simulate natural phenomena
with reasonable accuracy, apart from good documentation and rigorous reviews, it has to undergo a
two-step process: verification, and validation; which are the two most important steps in the quality
assurance program of a module (van der Heijde, 1987). A description of verification and validation
of a module is presented below.
Verification. The objective of the code verification process is two-fold (National Research
Council, 1990): (1) to demonstrate that the computational algorithms can accurately solve the
governing equations and (2) to assure that the computer code is fully operational. A module is said to
be 'verified' when it can be demonstrated that the mathematical framework embodied in the module
is correct. A module may be verified by comparing its simulation results against known analytical
solutions or numerical solutions from simulators based on similar or identical mathematical
frameworks.
Most modules (especially those based on legacy codes) may have been verified according to
the definition above.
C-l
-------
Appendix C
Verification and Validation ofEPACMTP
Validation. Model validation is conducted after the verification step. (Note that the word
'model' is used under the subject of validation instead of 'module.' A model in this case is a
combination of the mathematical framework embodied in the module and data representing a site or
hydrogeologic system of interest.) The objective of model validation is to determine how well the
mathematical representation of the processes describes the actual system behavior in terms of the
degree of correlation between model calculations and actual observed data (National Research
Council, 1990). Ideally, results should be compared to the results of a well-defined field experiment
or a well-conditioned laboratory experiment.
Validation of the predictive capabilities of the model is accomplished through comparison
with experimental data by using independent estimates of the parameters. In principle, this is the
ideal approach to validation. However, unavailability and inaccuracy of field characterization data
often prevent the application of such a rigid validation approach to actual field systems.
Methods that may be used to validate a model include:
¦ Using field data. Typically, parts of the field data are designated as calibration data, and
a calibrated site model is obtained through reasonable adjustment of parameter values.
Other parts of the field data are designated as validation data; the calibrated site model is
used in a predictive mode to generate similar data for comparison. For instance a
groundwater model may be calibrated against water level measurements, and then
validated by comparing predicted against measured contaminant concentrations in down-
gradient wells. Although this procedure will not allow complete validation of a modeling
process, it will provide some insight into potential problems of the model. This approach
is limited because splitting of a (groundwater) data set into two components does not
yield completely independent data sets. Two completely independent sets of data usually
do not exist, so that the verification (calibration) data and validation data are related.
¦ Using synthetic data. At times, the implementation of the above-described validation
approach using field data is not possible nor practical due to lack of adequate, complete
and high-quality field data. Thus, testing of groundwater models is limited to extended
verifications, and code comparisons. In this case, a newly developed model is compared
with established models designed to solve the same type of problems. If the results from
the new code do not deviate significantly from the those obtained with the existing codes,
a relative or comparative validity is established. If code comparison is used to evaluate a
new code, the code should again be validated as soon as adequate data sets become
available.
Absolute validity of a model is never determined. Establishing absolute validity requires
testing over the full range of conditions for which the model is designed, an exercise that is almost
never possible or practical. As stated by the National Research Council (1990), a validated model
should not be applied in a predictive mode beyond its historically observed range or range of
calibration.
C-2
-------
Appendix C
Verification and Validation ofEPACMTP
C.2 Technical Background of the Vadose-zone and Aquifer
Modules
C.2.1 EPACMTP
The U.S. EPA OIW has developed a probabilistic (Monte Carlo) groundwater flow and
transport modeling approach to assess potential exposure of groundwater to toxic chemical
constituents in wastes that are managed in Subtitle D industrial waste management units under
RCRA regulations. The exposure to groundwater is expressed as concentrations of potential
contaminants at a drinking water well (receptor well) located downgradient of the waste disposal
facility. This modeling methodology has been incorporated into EPACMTP (U.S. EPA, 1996 a, b, c,
d). The OIW of the U.S. EPA has applied EPACMTP, as a general fate and transport model, to
establish regulatory levels for concentrations of chemicals in the Subtitle D industrial waste
management units for several proposed rules and listing determinations.
The current subsurface flow and transport components in EPACMTP comprise the following
modules:
¦ 1-D vertical variably saturated flow and transport submodules—collectively referred to as
the vadose-zone module;
¦ 3-D saturated flow and transport submodules—collectively referred to as the 3-D aquifer
module;
¦ Quasi-3-D saturated flow and transport submodules—collectively referred to as the
quasi-3-D aquifer module; and
¦ Areal two-dimensional (vertically averaged) saturated flow and transport submodules—
collectively referred to as the areal two-dimensional aquifer module.
The first module is used to simulate flow and transport of constituents in the vadose-zone.
The following three modules are used to simulate flow and transport in the aquifer beneath the
vadose-zone.
Details of the above modules are provided in the EPACMTP background document (U.S.
EPA 1996a).
C.2.2 3MRA Subsurface Flow and Transport Modules
During the past 5 years, a number of enhancements have been made to EPACMTP and
EPACMTP-derived subsurface fate and transport modules. These include the development of a
computationally efficient pseudo-3-D module for the modeling system (U.S. EPA, 1999c), a new
surface impoundment module for the 3MRA modeling system (U.S. EPA, 1999d), methodologies to
handle fractures, heterogeneity, and anaerobic biodegradation based on a new nation-wide rate
database (U.S. EPA, 1999c, e). In addition, a number of comments from public and peer reviews
regarding features and theoretical aspects ofEPACMTP have been received and evaluated.
Based on the existing aquifer module in EPACMTP, a pseudo-3-D aquifer module was
developed as a component of the 3MRA modeling system, which is an integrated framework
C-3
-------
Appendix C
Verification and Validation ofEPACMTP
consisting of medium-specific pollutant fate, transport, exposure, and risk modules. The 3MRA
modeling system was first applied as part of the U.S. EPA's 1999 Hazardous Waste Identification
Rule (HWIR) notice. The differences between the pseudo-3-D and the EPACMTP modules are listed
below:
¦ Solution schemes. The pseudo-3-D scheme is based on a hybrid numerical-analytical
solution technique (U.S. EPA, 1999c). This solution scheme produces receptor well
concentrations that are somewhat more conservative than those generated by the fully
3-D module. Its computational speed is greater than that of the fully 3-D module by a
factor of approximately 300.
¦ Chemical concentrations at receptor wells. The concentrations at receptors are
reported in complete breakthrough curves. EPACMTP reports peak and averaged-
around-the-peak concentrations.
¦ Proximity of surface bodies. A surface water body into which groundwater discharges
pollutant flux is part of the pseudo-3-D module but not the EPACMTP-based module.
¦ Fractured media. Correction for the magnitude of a really averaged hydraulic
conductivity is allowed for fractured media in the pseudo-3-D module but not the current
module in EPACMTP.
¦ Heterogeneity. Correction for receptor well concentration to account for local variability
in hydraulic conductivity and porosity is allowed for heterogeneous media in the pseudo-
3-D module but not the EPACMTP-based module.
¦ Anaerobic biodegradation. A national database for anaerobic biodegradation was
developed for a number of organic chemicals (U.S. EPA, 1999e). The pseudo-3-D
module generates a value of chemical-specific anaerobic biodegradation rate based on the
probabilistically selected pH regime, temperature range, and redox environment.
¦ Time-dependent infiltration rate. A new module has been developed to simulate a
series of different infiltration rates to represent various stages of surface impoundment
operation. In each stage, the corresponding flow field is assumed to be steady. At this
time, this module has not yet been fully tested.
C.3 Verification History
EPACMTP has been verified extensively by comparing its simulation results against both
analytical and numerical solutions. Numerous verification cases were conducted from 1991-2000.
A summary of the verification cases is provided in the following subsections. The accompanying
figures for selected test cases are presented in the designated appendices.
C.3.1 ORD Verification (1992-1993)
In 1992, a verification analysis of the newly developed EPACMTP was performed by the
Office of Research and Development (ORD) of the U.S. EPA (U.S. EPA, 1992). A list of
verification cases in the verification exercise is listed in Table C.3-1. As shown in the table, two
steps of code verification were conducted: a re-verification of the original test problems and data
files provided by HydroGeoLogic, Inc. and independent verification using alternative test criteria.
C-4
-------
Appendix C
Verification and Validation of EPACMTP
Based on the analysis of Tetra Tech, some technical limitations in the EPACMTP code were
identified. One of the weaknesses, which occurred in the aquifer module, pertained to potential mass
loss of contaminants from the system due to the upstream boundary proximity and conditions in
EPACMTP. The code was modified in response to the comments (HydroGeoLogic, Inc., 1993).
Table C.3-1 Summary of EPACMTP Verification by the Office of Research and Development,
U.S. EPA, and Tetra Tech, Inc. in 1992-1993 (from U.S. EPA, 1992)
Case
Description
Reverification
of HGL Tests
Independent
Verification
1
Steady-state, aquifer flow, single layer
Yes
Yes
2
Steady-state, vadose-zone transport, two layers
Yes
3
Transient vadose-zone transport, single layer - analytical solution
Yes
Yes
4
Transient vadose-zone transport, single layer - numerical solution
Yes
5
Transient vadose-zone transport, single layer, nonconservative
solute—numerical solution
Yes
6
Transient vadose-zone transport, three layers, nonconservative
solute—numerical solution
Yes
7
Transient vadose-zone transport, single layer, nonlinear
adsorption—numerical solution
Yes
8
Multiple species transport; 3-member chain decay; source decay
Yes
9
Steady-state, aquifer flow
Yes
Yes
10
Quasi-3-D aquifer transport—numerical solution
Yes
Yes
11
Nonlinear aquifer transport
Yes
Yes
12
3-species transport, 2-D (x,y)
Yes
Yes
13
7-species transport, 2-D (x,y)
Yes
Yes
14
Full-3-D aquifer flow and transport
Yes
C.3.2 Module-Level Verification (1993-1994)
A module-level verification task was performed between 1993-1994 and reported in
EPACMTP Background Documents (U.S. EPA, 1996 b, d). Numerous components of EPACMTP's
flow and transport sub-modules, in both the vadose and aquifer modules, were verified between
1993-1994 against analytical solutions, and numerical solutions from a number of simulators with
similar mathematical frameworks. Details of the verification are presented below.
C.3.2.1 Vadose-Zone Module Verification
The vadose-zone and the aquifer modules were subdivided into the flow and transport sub-
modules. The ten verification cases for the vadose-zone module are summarized in Table C.3-2 and
are briefly described below. Excerpts of verification results for the vadose-zone test cases are
presented in figures in Attachment C-l. Reference to the figures in Attachment C-l is provided in
Table C.3-2. Additional information regarding the test cases and respective verification results may
C-5
-------
Table C.3-2 Verification Cases for the Vadose-Zone Module (1993-1994)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Steady-state infiltration
Semi-analytical FECTUZ module vs. fully
numerical finite element VADOFT
Figure C-1-1, Attachment C1
2
Steady-state infiltration in a layered soil
Semi-analytical FECTUZ module vs. fully
numerical finite element VADOFT
Figure C-1-2, Attachment C1
3
Steady-state infiltration in a layered soil with a
ponding depth
Semi-analytical FECTUZ module vs. fully
numerical finite element VADOFT
Figure C-1-3, Attachment C1
4
Steady-state transport in a layered soil
Steady-state analytical solution vs. finite element
numerical solution of FECTUZ
Figure C-1-4, Attachment C1
5
1-D transient transport under pulse input
conditions
Semi-analytical solution vs. numerical finite
element module of FECTUZ and the HYDRUS
code
Figure C-1-5, Attachment C1
6
1-D transport of a conservative solute species in a
saturated soil column of semi-infinite length
Numerical solution of FECTUZ vs. analytical
solution of Ogata and Banks (1961)
Figure C-1-6, Attachment C1
7
1-D transport of a conservative and
nonconservative solute species in a saturated soil
column of finite length
FECTUZ vs. analytical solution of van Genuchten
and Alves (1982)
Figure C-1-7, Attachment C1
8
Transport of a conservative species in a layered
soil column
FECTUZ vs. Shamir and Harleman (1967) and
Hadermann (1980)
Figures C-1-8 and C-1-9,
Attachment C1
9
Transient transport under conditions of nonlinear
adsorption with a pulse source
FECTUZ vs. finite difference code MOB1
Figure C-1-10, Attachment C1
10
Multispecies transport with three member, straight
decay chain with a decaying source boundary
condition
FECTUZ vs. analytical solution modified from
Hodgkinson and Maul (1985)
Figure C-1-11, Attachment C1
-------
Appendix C
Verification and Validation ofEPACMTP
be found in U.S. EPA (1996b). The vadose-zone module ofEPACMTP was originally called
FECTUZ. The numerical transport simulation in FECTUZ is no longer part ofEPACMTP.
The first three test cases of the vadose-zone module are for the flow sub-module and focus on
steady-state flow within layered and nonlayered soils. They were verified by comparing the results
of the semi-analytical FECTUZ (U.S. EPA, 1989) module against the numerical finite element
VADOFT model (Huyakorn, et al., 1987). Test Case 1 evaluated steady-state infiltration in a soil.
Test Cases 2 and 3 are similar, both involving steady-state infiltration in a layered soil, whereas Test
Case 3 introduced the surface impoundment boundary condition (ponding depth) to the system.
The last seven test cases, summarized in Table C.3-2, pertain to transport sub-module
verification. Test Cases 4 and 5 tested the analytical steady-state transport module and the semi-
analytical transport solution, respectively. Test Case 4 involved steady-state transport in a layered
soil and verification against the FECTUZ numerical solution, while Test Case 5 evaluated transient
transport with verification against both the FECTUZ numerical solution and the HYDRUS code
(Kool and van Genuchten, 1991).
Test Cases 6 through 10 utilize the FECTUZ numerical solution to examine transport of a
contaminant in a soil column. Case 6 concerns 1-D transport of a conservative solute species and is
verified against the analytical solution of Ogata and Banks (1961). Test Case 7 considers downward
vertical transport of both conservative and nonconservative constituents. The results are compared
against the analytical solution given by van Genuchten and Alves (1982). Test Case 8 concerns 1-D
transport of a conservative solute species in a layered soil column. Two sub-cases with different
dispersivity values were compared with the analytical solutions presented by Shamir and Harlemann
(1967) and Hadermann (1980). Test Case 9 considers solute transport with both linear and nonlinear
adsorption. This is verified against the MOB1 finite element solution (van Genuchten and Alves,
1982). Test Case 10 examines transport of a 3-member, straight decay chain and is verified against
the analytical solution, modified from Hodgkinson and Maul (1985).
C.3.2.2 Aquifer Module Verification
The saturated zone module ofEPACMTP was originally developed on a stand alone basis
and called CANSAZ-3-D. Seven benchmark problems were analyzed to verify the flow and
transport solutions in the CANSAZ-3-D modules; (Sudicky et al., 1990) and are summarized in
Table C.3-3. Excerpts of verification results for the test cases are presented in Attachment C-2.
Reference to the figures in Attachment C-2 is provided in Table C.3-3. Additional information
regarding the test cases and respective verification results may be found in U.S. EPA (1996b). Test
Case 1 was designed to verify the 3-D steady-state groundwater flow solution. For this purpose, the
hydraulic head and groundwater flow velocities obtained from CANSAZ-3-D were compared against
the MNDXYZ analytical solution (Ungs, 1986; Attachment C-2 of U.S. EPA, 1996b). Test Case 2
was designed to compare the analytical and numerical transport solutions for the case of single
species transport in a uni-directional steady-state groundwater flow field. Test Case 3 involved
2-dimensional transport of a 3-member decay chain. The CANSAZ-3-D results for this test problem
were verified against the numerical VAM2D code (Huyakorn et al., 1992). Test Case 4 involved
verification of CANSAZ-3-D against an analytical solution (Sudicky et al., 1991) for a case
involving a complex, seven-member branched decay chain. Test Case 5 was designed to verify the
nonlinear sorption option. This problem involves 1-D flow and transport with a nonlinear Freundlich
isotherm. CANSAZ-3-D was verified against the numerical MOB1 (van Genuchten, 1981) and
FECTUZ. Test Case 6 involves fully 3-D flow and transport. The CANSAZ-3-D solution was
compared against results obtained with the 3-D DSTRAM flow and transport code (Huyakorn and
C-7
-------
Table C-3-3. Verification Cases for the Saturated Zone Module (1993-1994)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Steady-state groundwater flow in a 3-D domain
CANSAZ-3-D vs. analytical solution
of MNDXYZ
Figure C.2-1, Attachment C.2
2
Single species transport in a uni-directional flow field—
analytical and numerical transport modules
CANSAZ-3-D vs. 3-D analytical
solution
Figure C.2-2, Attachment C.2
3
2-D transport of a 3-member decay chain. Steady-state flow
and transient solute transport in an unconfined aquifer
CANSAZ-3-D vs. VAM2D
Figure C.2-3, Attachment C.2
4
2-D transport of a complex, seven-member branched decay
chain with 1-D groundwater flow
CANSAZ-3-D vs. Gaussian source
analytical solution of Sudicky et al.
(1991)
Figure C.2-4, Attaclunent C.2
5
Nonlinear sorption reactions in a 1-D, steady-state flow and
transient transport. Pulse source using a Freundlich isotherm
CANSAZ-3-D vs. MOB1 and
FECTUZ
Figure C.2-5, Attaclunent C.2
6
Steady-state flow and transport modeling of a single
conservative species in 3-D aquifer domain
CANSAZ-3-D vs. DSTRAM
Figure C.2-.6, Attaclunent C.2
7
Steady-state groundwater flow and transient solute transport
in 3-D aquifer domain with a horizontal patch source
CANSAZ-3-D vs. VAM3-D
Figure C.2-7, Attaclunent C.2
-------
Appendix C
Verification and Validation ofEPACMTP
Panday, 1991). Test Case 7 was designed to evaluate the automatic model domain discretization
option for a 3-D flow and transport problem and was verified against the numerical VAM3-D code
(Panday et al., 1993).
C.3.2.3 Metals Transport Module
The major modifications to accommodate metals transport with nonlinear sorption were made
to the vadose-zone module, therefore the verification cases are applicable to this module. Five
verification test cases are summarized in Table C.3-4 and excerpts of verification results are
presented in Attachment C.3. Reference to the figures in Attachment C.3 is provided in Table C.3-4.
Additional information regarding the test cases and respective verification results may be found in
U.S. EPA (1996d). Test Case 1 involved continuous release of a nonsorbing solute to test the linear
adsorption partitioning capabilities. An analytical solution from Ogata (1970) was compared against
the EPACMTP results. Test Case 2 involved nonlinear Freundlich adsorption isotherms. The
Freundlich isotherm was represented by its closed form. Two different source conditions were
utilized: continuous and finite sources. Freundlich exponents greater than and less than one were
examined. The results from EPACMTP were compared with those from HYDRUS. Test Case 3
involves transport of lead in a fully saturated soil column. The verification of this case was
performed by comparing the computed cumulative mass against the total input mass. Test Case 4
involves 1-D transport of a solute, with Freundlich exponents of less than and greater than 1 and was
verified against HYDRUS.
C.3.3 Verification of Individual Modules and a Composite Model in
EPACMTP (1997)
In 1997, a testing plan was developed for EPACMTP code verification (U.S. EPA, 1997), in
accordance with the ASTM, "Standard Guide for Developing and Evaluating Ground-Water
Modeling Codes" (ASTM, 1996). The verification process focused on a single problem geometry,
representative of waste disposal scenarios in terms of spatial dimensionality and climatic/
hydrogeological conditions. The verification process first subdivided the problem setting into
individual hydrogeologic components, assessed their functionality relative to an overall fate and
transport problem, and then compared each component to analytical solutions or other codes.
The vadose-zone module, the aquifer module, and the composite model were verified
following the ASTM standards. The vadose-zone problem geometry was a 1-D column extending
from the land surface to the water table. Boundary conditions for numerical contaminant transport
involved a continuous source on the water table beneath the waste management unit. The region of
the water table outside the source area received constant recharge from the ground surface. Ten test
cases were conducted. These test cases may be subdivided into those for the vadose-zone module,
the aquifer module, and the composite model, and are summarized in Tables C.3-5a, C.3-5b, and
C.3-5c, respectively.
C.3.3.1 Vadose-Zone Module Verification
The vadose-zone module verification is summarized with four cases in Table C.3-5a and
excerpts of verification results are presented in Attachment C.4. Reference to the figures in
Attachment C.4 is provided in Table C.3-5a. Additional information regarding the test cases and
respective verification results may be found in U.S. EPA (1997). Test Case 1 evaluated steady-state
variably saturated flow and Test Case 2 considers infiltration through a clay liner and ponding depth.
C-9
-------
Table C.3-4. Verification Cases for Metals Transport in the Vadose-Zone (1993-1994)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Linear adsorption partitioning with continuous release of a nonsorbing
solute
Analytical solution (Ogata, 1970) vs.
EPACMTP result
Figure C.3-1, Attachment C.3
2
Nonlinear adsorption isotherm. The Freundlich isotherm was
represented by its closed function form. Freundlich isotherms greater
than and less than one were considered for continuous and finite
source conditions
EPACMTP vs. HYDRUS
Figures C.3-2 and C.3-3, Attachment C.3
3
Transport of lead using MINTEQA2-generated isotherms
Cumulative vs. total input mass
Figure C.3-4, Attachment C.3
4
Pulse source and Freundlich exponents of 0.5, 0.8, and 1.5
Analytic solution vs. HYDRUS
Figure C.3-5, Attachment C.3
-------
Table C.3-5a. Verification Cases for the Vadose-Zone Module (1997)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Steady-state variably saturated flow
EPACMTP vs. STAFF3D
Figure C.4-1, Attachment C.4
2
Infiltration through a clay liner from a surface impoundment
EPACMTP vs. STAFF3D
Figure C.4-.2, Attachment C.4
3
Contaminant transport with linear sorption and decay
Numerical EPACMTP vs. analytical EPACMTP
Figure C.4-3, Attachment C.4
4
Contaminant transport with branched chain decay and linear sorption
EPACMTP vs. VAM2D
Figure C.4-4, Attachment C.4
Table C.3-5b. Verification Cases for the Aquifer Module (1997)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
5
3-D steady-state groundwater flow
EPACMTP vs. MNDXYZ analytical solution
Figure C.5-1 and C.5-2, Attachment
C.5
6
3-D contaminant transport with linear sorption and decay
EPACMTP numerical module vs. EPACMTP analytical
module; EPACMTP vs. VAM3DF
Figure C.5-3, Attachment C.5
7
3-D contaminant transport with four species, branched chain decay and linear
sorption
EPACMTP vs. STAFF3D
Figure C.5-4, Attachment C.5
Table C.3-5c. Verification Cases for Composite Module (1997)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
8
Composite flow and contaminant transport
EPACMTP vs. VAM3DF
Figure C-6-1, Attachment C-6
9
Monte Carlo analysis based on composite flow and contaminant transport
EPACMTP vs. VAM3DF
Figure C-6-2, Attachment C-6
-------
Appendix C
Verification and Validation ofEPACMTP
Test Cases 3 and 4 both considered contaminant transport with linear sorption, but Test Case 3
examined linear decay while Test Case 4 evaluated four species with branched chain decay. Test
Cases 1 and 2 were verified against STAFF3D (HydroGeoLogic, Inc., 1995a), Test Case 4 was
verified against VAM2D while Test Case 3 compared the steady-state results from numerical and
analytical transport modules ofEPACMTP.
C.3.3.2 Aquifer Module Verification
The aquifer module verification is summarized with three cases in Table C.3-5b and excerpts
of verification results are shown in Attachment C.5. Reference to the figures in Attachment C.5 is
provided in Table C.3-5b. Additional information regarding the test cases and respective verification
results may be found in U.S. EPA (1997). The 3-D steady-state fully saturated flow module in
EPACMTP was verified against the analytical solution MNDXYZ in Test Case 5. Test Cases 6 and
7 examined contaminant transport and were verified against VAM3DF (HydroGeoLogic, Inc.,
1995b) and STAFF3D, respectively. Test Case 6 involved transport of a contaminant with linear
sorption and decay, while Test Case 7 involved linear sorption and a four species, branched chain
decay.
C.3.3.3 Composite Model Verification
The EPACMTP composite model comprises the following fate and transport modules: a
vadose-zone module, and a aquifer (saturated zone) module. These modules are connected according
to the detailed description in U.S. EPA (1996). The composite model verification is summarized
with two test cases in Table C.3-5c and excerpts of verification results are shown in Attachment C.6.
Reference to the figures in Attachment C-6 is provided in Table C.3-5c. Additional information
regarding the test cases and respective verification results may be found in U.S. EPA (1997). Test
Case 8 considered the composite flow and contaminant transport structure. Test Case 9 assessed the
sensitivity of the geometric assumptions used to develop EPACMTP. A limited Monte-Carlo
analysis was performed to assess the sensitivity of the confined water table assumption to predicting
the probability of exceedance at a monitoring well. Both Test Cases 8 and 9 were verified against
VAM3DF which is a 3-D, variably saturated numerical flow and transport code.
C.3.4 Verification of 3MRA Subsurface Flow and Transport Modules (1999)
In 1999, the flow and transport components for the vadose-zone module and aquifer module
were extracted from EPACMTP to provide the groundwater pathway module for the 3MRA system.
The basic premise for verification of the vadose-zone and aquifer modules was that EPACMTP had
been rigorously verified, so it was sufficient to show that the modules reproduced EPACMTP results.
Therefore, both the steady-state flow and transport sub-modules of the aquifer module (U.S. EPA,
1999c) and the flow and transport sub-modules of the vadose-zone module (U.S. EPA, 1999b,c) were
compared against the numerical results from EPACMTP to ensure that the extracted modules
remained intact. There are two exceptions that will be discussed below. The new saturated zone
pseudo-3-D module was also developed during this period (U.S. EPA, 1999a).
The 18 test cases for the vadose-zone, aquifer, and pseudo 3-D modules are summarized in
Tables C.3-6a, C.3-6b, and C.3-6c, respectively. The figures are presented in Attachment C.7
through C.9. The vadose-zone problem geometry was a 1-D column extending from the land surface
to the water table. Boundary conditions for numerical contaminant transport involved a continuous
source on the water table beneath the waste management unit. The region of the water table outside
the source area was also considered to be a recharge boundary.
C-12
-------
Appendix C
Verification and Validation ofEPACMTP
C.3.4.1 Vadose-Zone Module Verification
There are eight vadose-zone module verification cases (Table C.3-6a). Excerpts of results for
the verification cases are presented in Attachment C-7. Reference to the figures in Attachment C.7 is
provided in Table C.3-6a. Additional information regarding the test cases and respective verification
results may be found in U.S. EPA (1999b,c). All of the cases concern contaminant transport. Test
Case 1 evaluated an exponentially depleting source. Test Case 2 involved transport of a contaminant
with no sorption and no hydrolysis. Test Case 3 examined sorption and hydrolysis with one species,
while Test Case 4 involved two species with chain decay. Test Case 5 examined linear and nonlinear
metal transport using the MINTEQA2 isotherms. Test Case 6 evaluated biodegradation resulting in
chain decay reactions with four species. Test Cases 7 and 8 examined contaminant concentration at a
receptor well and pressure heads at each grid node, respectively. In this instance, both Test Case 7
and 8 were verified against MODFLOW-SURF ACT (HydroGeoLogic, Inc., 1996), a 3-dimensional
numerical groundwater flow and transport code.
C.3.4.2 Aquifer Module Verification
There are seven aquifer module verification cases (Table C-3-6b) with excerpts of
verification results presented in Attachment C-8. Reference to the figures in Attachment C-8 is
provided in Table C.3-6b. Additional information regarding the test cases and respective verification
results may be found in U.S. EPA (1999c). Test Case 1 evaluated an exponentially depleting source.
Test Case 2 involved transport of a conservative contaminant with no sorption and no hydrolysis.
Test Case 3 examined sorption and hydrolysis with one species, while Test Case 4 involved two
species with chain decay. Test Case 5 examined linear and nonlinear metal transport using the
MINTEQA2 isotherms. Test Case 6 evaluated biodegradation resulting in chain decay reactions with
four species. Test Case 7 evaluated the generated Monte Carlo distributions.
C.3.4.3 Pseudo-3-D Module Verification
There are three verification cases for the pseudo-3-D aquifer module (Table C.3-6c).
Excerpts of verification results are shown in Attachment C-9. Reference to the figures in
Attachment C.9 is provided in Table C.3-6c. Additional information regarding the test cases and
respective verification results may be found in U.S. EPA (1999a,c). Test Case 1 examined the
average groundwater specific flow rate determined by the saturated flow sub-module and was
verified using Darcy's Law. Test Case 2 examined the numerical component of the contaminant
transport sub-module and is verified using the analytical solution by Ogata (1970). Test Case 3
verified the combined analytical-numerical contaminant transport sub-module using verification
results of Test Case 2 subject to the analytical portion of the aquifer transport sub-module.
C.3.5 Comprehensive Verification of the 3MRA Vadose-zone Pseudo-3-D
Aquifer Modules (2000)
In 2000, a comprehensive verification was conducted of all of the components in the
extracted aquifer and the vadose-zone modules (U.S. EPA, 2000a,b). For testing purposes, each
component was executed as a stand-alone program outside of the 3MRA Software System
environment.
C-13
-------
Table C.3-6a Verification Cases for the 3MRA Vadose-Zone Module (1999)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Exponentially depleting conservative source with no sorption or hydrolysis
Vadose-Zone Module vs. EPACMTP
Figure C.7-1, Attachment C.7
2
Constant-concentration source pulse with no sorption or hydrolysis
Vadose-Zone Module vs. EPACMTP
Figure C.7-2, Attachment C.7
3
Constant-concentration source pulse with sorption and hydrolysis, one species
Vadose-Zone Module vs. EPACMTP
Figure C.7-3, Attachment C.7
4
Constant-concentration source pulse with sorption and hydrolysis, and chain
decay
Vadose-Zone Module vs. EPACMTP
Figure C.7-4, Attachment C.7
5
Metals: (mercury and lead), with constant-concentration source pulse with
MINTEQ-based sorption and no hydrolysis
Vadose-Zone Module vs. EPACMTP
Figures C.7-5 and C.7-6,
Attachment C.7
6
Constant-concentration source pulse with biodegradation, sorption and chain
decay
Vadose-Zone Module vs. EPACMTP
Figure C.7-7, Attachment C.7
7
1-D contaminant transport between a top boundary at the bottom of the source
zone and the water table with mass loading to the top boundary from the
leachate flux from the source module
Vadose-Zone Module vs. MODFLOW-SURFACT
Figure C.7-8, Attachment C.7
8
1-D variable saturated flow between a top boundary at the bottom of the source
zone and the water table with mass loading to the top boundary from the
leachate flux from the source module
Vadose-Zone Module vs. MODFLOW-SURFACT
Figure C.7-9, Attachment C.7
-------
Table C.3-6b Verification Cases for the 3MRA Aquifer Module (1999)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Exponentially depleting source with no sorption or hydrolysis
Aquifer Module vs. EPACMTP
Figure C.8-1, Attachment C.8
2
Constant-concentration source pulse with no sorption or hydrolysis
Aquifer Module vs. EPACMTP
Figure C.8-2, Attachment C.8
3
Constant-concentration source pulse with sorption and hydrolysis, one species
Aquifer Module vs. EPACMTP
Figure C.8-3, Attachment C.8
4
Constant-concentration source pulse with sorption and hydrolysis, and two species
with chain decay
Aquifer Module vs. EPACMTP
Figures C.8-4 and C.8-5, Attachment C.8
5
Metals: (mercury and lead), with constant-concentration source pulse with sorption
and no hydrolysis
Aquifer Module vs. EPACMTP
Figure C.8-6, Attachment C.8
6
Constant-concentration source pulse with biodegradation, sorption and four species
chain decay
Aquifer Module vs. EPACMTP
Figures C.8-7, C.8-8, C.8-9, and C.8-10,
Attachment C.8
7
Comparison of Monte Carlo saturated zone simulations
Aquifer Module vs. EPACMTP
Figure C.8-11, Attachment C.8
Table C.3-6c Verification Cases for the 3MRA Pseudo-Three Dimensional Aquifer Module (1999)
Case
Description
Verification Method
Excerpts of Verification Results
Presented in
1
Average Groundwater Specific Flow Rate
Aquifer Module vs. Darcy's Law analytical
solution
Figure C.9-1, Attachment C.9
2
Numerical Component of the Contaminant Transport Sub-module
Aquifer Module vs. analytical solution by
Ogata
Figure C.9-2, Attachment C.9
3
Analytical-Numerical Component of the Contaminant Transport Sub-module
Aquifer Module vs. analytical solution
Figure C.9-3, Attachment C.9
rs
o'
s
£
I
o'
£
§
^3
-------
Appendix C
Verification and Validation ofEPACMTP
C.3.5.1 Vadose-Zone Module Verification
There are 40 vadose-zone module verification test cases summarized in Table C.3-7a.
Selected figures for Test Areas 4 and 5, the nonmetals and metals transport components, are
presented in Attachment C. 10. Reference to the figures in Attachment C. 10 is provided in
Table C.3-7a. Additional information regarding the test cases and respective verification results may
be found in U.S. EPA (2000a). The reading and screening of source and site-specific input data was
verified in three cases. Verification of the pre-simulation processing of input data was performed
with two cases. The flow, nonmetal transport, and metals transport components were verified with 1,
5, and 4 cases, respectively. The post simulation processing of output data was verified with two
cases. The robustness testing verified the stability of the simulation when executed with extreme
values for selected parameters. The parameters were selected based on the results of a parameter
sensitivity analysis (U.S. EPA, 1996e). The vadose-zone module's robustness was verified with
13 cases.
Table C.3-7a. Verification Cases for the 3MRA Vadose-Zone Module (2000)
Test
Area
General Requirements
Number of
Verification Cases
Excerpts of Verification
Results Presented in
1
Verification of reading and screening of source
and site-specific input data
3
N/A
2
Verification of pre-simulation processing of input
data
2
N/A
3
Verification of the flow component
1
N/A
4
Verification of the nonmetals transport component
5
Figure C.10-1,
Attachment C.10
5
Verification of the metals transport component
4
Figure C.10-2,
Attachment C.10
6
Verification of post simulation output
2
N/A
7
Verification of the vadose-zone module's
robustness
13
N/A
C.3.5.2 Aquifer Module Verification
There are 69 aquifer module verification cases summarized in Table C.3-7b. Selected figures
for Test Area 8, the fate and transport component, are present in Attachment C. 11. Reference to the
figures in Attachment C. 11 is provided in Table C.3-7b. Additional information regarding the test
cases and respective verification results may be found in U.S. EPA (2000b). The reading and
screening of source and site-specific input data was verified in four cases. Verification of the pre-
simulation processing of input data was performed with 17 cases. The fractured media, and
heterogeneous saturated media components were verified with 3, and 1 cases, respectively. The
reading and screening of chemical-specific, biodegradation and metal-specific data was verified in
six tests. The numerical grid generation was verified in four cases. The flow component was
verified with 4 cases, while the contaminant fate and transport component was verified in 19 cases.
The robustness testing verified the stability of the simulation when executed with extreme values for
selected parameters. The parameters were selected based on the results of a parameter sensitivity
analysis (U.S. EPA, 1996e). The aquifer module's robustness was verified with 11 cases.
C-16
-------
Appendix C
Verification and Validation ofEPACMTP
Table C.3-7b. Verification Cases for the 3MRA Aquifer Module (2000)
Test
Area
General Requirements
Number of
Verification Cases
Excerpts of Verification
Results Presented in
1
Verification of reading and screening of source
and site-specific input data
4
N/A
2
Verification of pre-simulation processing of input
data
17
N/A
3
Verification of the fractured media component
3
N/A
4
Verification of the heterogeneous saturated
media component
1
N/A
5
Verification of reading and screening of chemical-
specific, biodegradation, and metal-specific data
6
N/A
6
Verification of numerical grid generation
4
N/A
7
Verification of the flow component
4
N/A
8
Verification of the contaminant fate and transport
component
19
Figures C.11-1 and C.11-2,
Attachment C.11
9
Verification of the aquifer module's robustness
11
N/A
C.4 Validation History
Validation, as defined previously, may be conducted using actual measured field data. It is
helpful to assess the validity of simplifying assumptions and the predictive capabilities ofEPACMTP
against well documented realistic site data. EPACMTP and its predecessors (from which flow and
transport components in EPACMTP were derived) have been validated based on actual observations
at four sites, although no validation has been performed using the 3MRA vadose-zone and aquifer
zone modules. In 1990, EPACMS (CANSAZ) was validated against the data from the Borden
Landfill site, along with the data from a second agricultural field site on Long Island, New York
(U.S. EPA, 1990). This validation included the combination of the saturated and the vadose-zone
modules in EPACMS. In 1993, the composite model was validated against data from a Dodge City,
Kansas site (Kool et al., 1994). Then, in 1995 EPACMTP was validated against the data from the
EBOS Site 24 in New York (U.S. EPA, 1995). The four validation cases are presented in the
following subsections. Note that all the figures referenced to in the following subsections are
presented.
C.4.1 Borden Site
The Borden landfill is located in Borden, southern Ontario, Canada, and occupies an area of
approximately 4 ha (Figure C. 12-1). The landfill was operational for 36 years and at its closure was
capped with a thin layer of sand. The site overlies 8 to 20 meters of a glaciofluvial sand aquifer,
which overlies a confining silty clay deposit. The chloride plume extends about 700 meters
northward of the landfill and occupies nearly the entire vertical thickness of the aquifer. The waste
material was deposited just above the water table, therefore transport did not occur in the vadose-
zone.
Generally, the flow and transport parameters and the procedure described by Frind and
Hokkanen (1987) were used for the EPACMS simulation. The exception is that Frind and Hokkanen
assigned a higher recharge rate to some areas outside of the source area, but this refinement was not
C-17
-------
Appendix C
Verification and Validation ofEPACMTP
utilized for the EPACMS simulation. A curvilinear grid was utilized to describe the aquifer geometry
and because EPACMS assumes a constant saturated thickness, the VAM3-D-CG code (Huyakorn
and Panday, 1989) was used to perform the groundwater flow simulation. Next, the CANSAZ
(EPACMTP module) module was used to simulate transient transport. CANSAZ utilized the same
finite grid as the groundwater flow simulation, as well as the groundwater velocity distribution from
the VAM3-D simulation.
The chloride concentrations were compared for the observed values (Figure C.12-2), the
CANSAZ simulation values (Figure C.12-3) and the Frind and Hokkanen simulation values
(Figure C. 12-4). The CANSAZ model accurately predicted the extent of the plume and the overall
plume shape compared to both the Frind and Hokkanen model and the field values.
C.4.2 Long Island Site
The site is located on the south shore of the North Fork of Long Island, New York
(Figure C.12-5). The agricultural site was contaminated with the pesticide aldicarb in the 1970's.
The source was a 2.5 ha potato field overlying sandy loam soils with a high infiltration rate. An
unconfined aquifer is located approximately 2 meters below the surface.
Both site-specific data and monitoring data are limited at this site. The site characterization
was obtained from previous studies by INTERA (1980) and Carsel et al. (1985). The EPA Pesticide
Root Zone Model (PRZM) (Carsel et al., 1984) was used to predict the 3-year average recharge rate
and average aldicarb concentration at the base of the root zone as input for the EPACMS vadose-
zone module. The steady-state groundwater flow field was generated using the analytical 2-D
solution on EPACMS, followed by a 3-year transient aldicarb transport simulation.
The simulated concentrations of aldicarb in groundwater with distance from the source were
compared with the observed values (Figure C. 12-6). The agreement between the simulated and
observed concentrations was quite reasonable, with the relative error decreasing with increasing
distance from the source.
C.4.3 Dodge City Site
The Dodge City, Kansas site (Figure C.12-7), located in the Arkansas River valley, is
documented by Ourisson et al. (1992). The source is a controlled release of Triasulfuron pesticide
(nonconservative) and bromide (conservative) which, over a 2-year period, is transported through the
vadose-zone and the aquifer. The site covers an area of approximately 2.3 acres (approximately 1 ha)
overlying one meter of sandy loam soil which overlies a sand and gravel unit. The water table is
located at a depth of 3 meters.
The Dodge City site was well characterized, the source was well defined and the monitoring
data were available for both soil and groundwater. Site-specific values were obtained from Ourisson
et al., (1992), Carsel and Parrish (1988), Gelhar et al. (1985), Carsel et al. (1984), and derived values.
EPACMTP was used to simulate the flow and transport of both conservative and nonconservative
constituents.
The groundwater concentration model predictions were compared against the observed
values. The model tended to underestimate bromide concentrations (Figure C.12-8) slightly and
overestimate Triasulfuron concentrations (Figure C. 12-9). The application of the model to the
C-18
-------
Appendix C
Verification and Validation ofEPACMTP
Kansas field site showed reasonably good agreement between model predictions and groundwater
monitoring results.
C.4.4 EBOS Site
The Electric Power Research Institute (EPRI) research site referred to as EBOS site 24 is a
disposal site for a coal tar Manufacturing Gas Plant located in New York state (Figure C. 12-10).
Initially, the coal tar was disposed of in a trench on the site, then over time migrated downward into
the aquifer (Figure C. 12-11 and C. 12-12). The site is characterized by 15 to 30 feet of typical glacial
outwash sand deposits overlying a clay confining layer. Napthalene was labeled the constituent of
concern, since it was the polyaromatic hydrocarbon (PAH) with the highest concentrations in the coal
tar.
The site-specific parameters were provided by EPRI in 1993 and consisted of both known
and estimated values. EPACMTP was used to simulate the flow and transport of constituents through
the vadose-zone and the aquifer. One point to note was that since the coal tar had moved down into
the aquifer, the constituents could be leached out through direct contact with ambient groundwater.
In the EPACMTP simulation, it was necessary to leach the constituents out of the waste by
infiltration from the vadose-zone.
Napthalene concentrations near the source before (Figure C. 12-13) and after (Figure C. 12-14)
source removal were predicted by EPACMTP. The results from EPACMTP were qualitatively
similar to the observed concentration in terms of groundwater concentrations near the source.
C.5 Summary
EPACMTP, its predecessors (EPACMS, CANSAZ, and FECTUZ), and its derivatives
(3MRA vadose-zone and aquifer modules) have been verified extensively during the past decade at
each of the developmental stages. The model has been verified, in numerous cases, by comparing the
simulation results against both analytical and numerical solutions. Additionally, EPACMTP and its
predecessors have been validated using actual site data from four different sites. The relevant
verification and validation history, discussed in the previous sections of this document, is
summarized below.
The preliminary verification ofEPACMTP was performed by ORD in 1992. Following the
preliminary verification, detailed module-level verification was conducted on the flow and transport
sub-modules of the vadose-zone and the aquifer modules between 1993-1994. The modules were
verified against analytical solutions, and numerical solutions from a number of well-documented
simulators. In 1997, the EPACMTP code was verified utilizing a testing plan developed according to
ASTM standards. The vadose-zone and the aquifer modules, as well as the composite model (based
on the sequentially linked vadose-zone and aquifer modules), were verified against analytical and
numerical solutions. In 1999, the vadose-zone and aquifer modules were extracted from EPACMTP
to be included as part of the 3MRA software system. The flow and transport sub-modules of both
modules were verified against the results from EPACMTP. Additionally, for the 3MRA software
system a pseudo-3-D aquifer module was developed. An exhaustive verification was conducted of
all of the components in the extracted vadose-zone module and the new pseudo-3-D aquifer module
in 2000. The modules were verified against analytical solutions and EPACMTP results.
C-19
-------
Appendix C
Verification and Validation ofEPACMTP
EPACMTP and its predecessors have been validated using field data from four unique sites
from 1990-1995. These sites include: the Borden site, the Long Island site, the Dodge City site, and
the EBOS site 24.
C.6 REFERENCES
ASTM, 1996. Standard Guide for Developing and Evaluating Ground-Water Modeling Codes.
ASTM Subcommittee D-18.21.10, Tracking Number D18.21.92.07, February 9.
Carsel, R.F., C.N. Smith, L.A. Mulkey, J.D. Dean, and P. Jowise, 1984. User's Manual for the
Pesticide Root Zone Model (PRZM): release 1. Environ. Res. Lab., Athens, GA, U.S. EPA,
EPA-600/3-84-109.
Carsel, R.F., L.A. Mulkey, M.N. Lorber, and L.B. Laskin, 1985. The Pesticide Root Zone Model
(PRZM): A Procedure for Evaluating Pesticide Leaching Threats to Groundwater.
Ecological Modeling 30: 4g-6g.
Carsel, R.F., 1988. Developing Joint Probability Distributions of Soil Water Retention
Characteristics. Water Resources Res., 24(5): 755-769.
Frind, E.O., G.E. Hokkanen, 1987. Simulation of the Borden Plume Using the Alternating Direction
Galerkin Technique. Water Resources Res., 23, 918-930.
Gelhar, L.W., A. Mantoglou, C. Welty, and K.R. Rehfeldt, 1985. A Review of Field-Scale Physical
Solute Transport Processes in Saturated and Unsaturated Porous Media. Electr. Power Res.
Inst., Palo Alto, CA, EPRIEA-190.
Hadermann, J., 1980. Radionuclide transport through heterogeneous media, Nuclear Technology,
v. 47, pp. 311-323.
Hodgkinson, D.P. and P.R. Maul, 1985. One-dimensional Modeling of Radionuclide Migration
Through Permeable and Fractured Rock for Arbitrary Length Decay Chains using Numerical
Inversion of Laplace Transforms, Rep. AERE-R. 11880, UK Atomic Energy Authority,
Oxfordshire, U.K.
Huyakorn, P.S., and J.E. Buckley, 1987. VADOFT: Finite Element Code for Simulating One-
Dimensional Flow and Solute Transport in the Vadose-zone. Technical Report Prepared for
the U.S. Environmental Protection Agency, Athens, GA.
Huyakorn, P.S., and S. Panday, 1989. VAM3D-CG: Variably Saturated Analysis Model in Three
Dimensions with Pre-conditioned Gradient Matrix Solvers. HydroGeoLogic, Inc., Herndon,
VA.
Huyakorn, P.S., and S. Panday, 1991. DSTRAM: Density-Dependant Solute Transport Analysis
Finite Element Program, version 3.1. Technical Report. HydroGeoLogic, Inc., Herndon,
VA.
Huyakorn, P.S., J.B. Kool, and Y.S. Wu, 1992. VAM2D - Variably Saturated Analysis Model in
Two Dimensions. NUREG/CR-5352.
C-20
-------
Appendix C
Verification and Validation ofEPACMTP
HydroGeoLogic, Inc., 1993. Response to Review Comments on EPACMTP Unsaturated Zone and
Saturated Zone Modules by TetraTech, Inc. HydroGeoLogic, Inc., 1155 Herndon Parkway,
Suite 900, Herndon, VA 20170.
HydroGeoLogic, Inc., 1995a. STAFF3D: A Three-Dimensional Finite Element Code for Simulating
Fluid Flow and Transport of Radionuclides in Fractured Porous Media. Version 3.0.
HydroGeoLogic, Inc., 1155 Herndon Parkway, Suite 900, Herndon, VA 20170.
HydroGeoLogic, Inc., 1995b. VAM3DF: Variably Saturated Analysis Model in Three-Dimensions
for the Data Fusion System. Version 1.0. HydroGeoLogic, Inc., 1155 Herndon Parkway,
Suite 900, Herndon, VA 20170.
HydroGeoLogic, Inc., 1996. MS-VMS Software (Version 1.2) Documentation. HydroGeoLogic,
Inc., 1155 Herndon Parkway, Suite 900, Herndon, VA 20170.
INTERA Environmental Consultants, 1980. Mathematical Simulation of Aldicarb Behavior on Long
Island. Technical Report, Contract No. 80-6876-02. U.S. EPA, Athens, Georgia.
Kool, J. B., and M.Th. Van Genuchten, 1991. One-Dimensional Variable Saturated Flow and
Transport Model, Including Hysteresis and Root Water Uptake. U.S. Salinity Laboratory,
USDA-ARS, Riverside, CA.
Kool, J. B., P.S. Huyakorn, E.A. Sudicky, Z.A. Saleem, 1994. A composite modeling approach for
subsurface transport of degrading contaminants from land-disposal sites. Journal of
Contaminant Hydrology, 17, 69-90.
National Research Council, 1990. Groundwater Models: Scientific and Regulatory Applications.
National Academy Press, Washington, DC.
Ogata, A., and R.B. Banks, 1961. A Solution of the Differential Equation of Longitudinal Dispersion
in Porous Media, U.S. Geological Survey Professional Paper No. 411-A.
Ogata, A., 1970. Theory of Dispersion in a Granular Medium. U.S. Geological Survey Professional
Paper No. 411-1, p. 134.
Ourisson, P. J., 1992. Small Scale Prospective Groundwater Study of Amber on Winterwheat in
Kansas, Vol. 1. Ciba Geigy Corp., Agric. Div., Greensboro, NC.
Panday, S.M., P.S. Huyakorn, R. Therrien, andR.L. Nichols, 1993. Improved three-dimensional
finite-element techniques for field simulation of variably saturated flow and transport. Journal
of Contaminant Hydrology, 12(1), 1-33.
Shamir, U.Y., and D.R.F. Harlemann, 1967. Dispersion in Layered Porous Media, J. of the
Hydraulics Division, Amer. Soc. Civ. Engr., v. 93, Hy5, pp. 237-260.
Sudicky, E.A., Y.S. Wu, and Z. Saleem, 1991. Semi-Analytical Approach for Simulating Transport
of a Seven-Member, Branched Decay Chain in 3-D Groundwater Systems, EPS Trans. AM.
Geophys. Union, p. 135.
Sudicky, E.A., J.B. Kool and P.S. Huyakorn, 1990. CANSAZ: Combined Analytical-Numerical
Model for Simulating Flow and Contaminant Transport in the Saturated Zone. Version 2.0
C-21
-------
Appendix C
Verification and Validation ofEPACMTP
with Nonlinear Adsorption and Chain Decay Reactions. Technical Report Prepared for U.S.
EPA, Office of Solid Waste, Washington, DC, 20460.
Ungs, M.J., 1986. Steady-state Analytical Solution to the 3-D Groundwater Mounding Problem of
an Unconfined Aquifer (MNDXYZ). Prepared for U.S. EPA, Office of Solid Waste,
Washington, DC, 20460. Also in Appendix B of U.S. EPA (1996b)
U.S. EPA, 1989. Unsaturated Zone Flow and Transport Module: Version 2.0 with Nonlinear
Sorption and Chain Decay Reactions. U.S. EPA, Office of Solid Waste, Washington, DC,
20460.
U.S. EPA, 1990. Response to Comments on the CANSAZ Flow and Transport Model for Use in
EPACMS by the Science Advisory Board (SAB) of the United States Environmental
Protection Agency. U.S. EPA, Office of Solid Waste, Washington, DC, 20460.
U.S. EPA, 1992. Memorandum dated November 6, 1992 from the U.S. EPA Office of Research and
Development, Athens, GA, to the U.S. EPA Office of Solid Waste, Washington, DC, 20460.
U.S. EPA, 1995. Briefing for EPA's Science Advisory Board on EPACMTP, March 8.
U.S. EPA, 1996a. Background Document for EPACMTP: User's Guide. U.S. EPA, Office of Solid
Waste, Washington, DC, 20460.
U.S. EPA, 1996b. Modeling approach for simulating three-dimensional migration of land disposal
leachate with transformation products. Volume I: Background document for the unsaturated
zone and saturated zone modules. U.S. EPA, Office of Solid Waste, Washington, DC, 20460.
U.S. EPA, 1996c. EPA's Composite Model for Leachate Migration with Transformation Products
(EPACMTP) Background Document for Finite Source Methodology. U.S. EPA, Office of
Solid Waste, Washington, DC, 20460.
U.S. EPA, 1996d. EPA's Composite Model for Leachate Migration with Transformation Products
(EPACMTP) Background Document for Metals: Volume 1. U.S. EPA, Office of Solid
Waste, Washington, DC, 20460.
U.S. EPA, 1996e. Sensitivity Analysis of the EPACMTP Monte Carlo Sampling Techniques and
Key Input Parameter Distribution. U.S. EPA, Office of Solid Waste, Washington, DC,
20460.
U.S. EPA, 1997. Test and Verification of EPA's Composite Model for Leachate Migration with
Transformation Products (EPACMTP). U.S. EPA Office of Solid Waste, Washington, DC,
20460.
U.S. EPA, 1999a. Verification Document for HWIR99 Pseudo-Three Dimensional Aquifer Module.
U.S. EPA, Office of Solid Waste, Washington, DC, 20460.
U.S. EPA, 1999b. Verification Document for HWIR99 Vadose-Zone Module. U.S. EPA, Office of
Solid Waste, Washington, DC, 20460.
U.S. EPA, 1999c. Vadose and Saturated Modules Extracted from EPACMTP for HWIR99. U.S.
EPA, Office of Solid Waste, Washington, D.C.
C-22
-------
Appendix C
Verification and Validation ofEPACMTP
U.S. EPA, 1999d. Source Modules for Tanks and Surface Impoundments: Background and
Implementation for the Multimedia, Multipathway, and Multireceptor Risk Assessment
(3MRA) for HWIR99. U.S. EPA, Office of Solid Waste, Washington, D.C.
U.S. EPA, 1999e. Anaerobic Biodegradation Rates of Organic Chemicals in Groundwater: A
Summary of Field and Laboratory Study. U.S. EPA, Office of Solid Waste, Washington,
DC.
U.S. EPA, 2000a. Test Plan for HWIR99 Vadose-Zone. U.S. EPA Office of Solid Waste,
Washington, DC, 20460.
U.S. EPA, 2000b. Test Plan for HWIR99 Aquifer Module. U.S. EPA Office of Solid Waste,
Washington, DC, 20460.
Van der Heijde, P.M.K., 1987. Quality assurance in computer simulations of groundwater
contamination. Environmental Software, 2(1): 19-28.
Van Genuchten, M.Th., 1981. MOB1: A Numerical Model for the Simulation of Chemical Mobility
in Porous Media. Unpublished.
Van Genuchten, M.Th., and W.J. Alves, 1982. Analytical Solutions of the One-Dimensional
Convective-Dispersion Solute Transport Equation, U.S. Technical Bulletin No. 1661, p. 151.
C-23
-------
Appendix C
Verification and Validation ofEPACMTP
this page intentionally left blank.
C-24
-------
Attachment C.l
Vadose Zone Module Verification Results
(1993-1994)
-------
This page intentionally left blank.
-------
Attachment C.l
Vadose Zone Verification Results
LIST OF FIGURES
Page
C. 1-1 Test Case 1: Predicted pressure head distribution in the unsaturated zone.
The solid line represents finite element solution, and data points
represent semi-analytical solution C.l-1
C. 1-2 Test Case 2: Predicted saturation distribution in the unsaturated zone.
The solid line represents the finite element solution, data points represent the
semi-analytical solution C.l-2
C. 1 -3 Test Case 3: Predicted saturation distribution in the unsaturated zone.
The solid line represents the finite element solution, and data points represent
the semi-analytical flow solution C.l-3
C. 1-4 Test Case 4: Transport in layered soil. The solid lines correspond to the
transient solution, evaluated at t=5, 15, 30, and 80 years. Data points
represent the steady-state solution C.l-4
C. 1-5 Test Case 5: Comparison of breakthrough curves for the transient pulse input
transport problem. The solid line represents the solution of the numerical
HYDRUS code, and the data points represent the semi-analytical and numerical transport
solution of FECTUZ C. 1 -5
C. 1-6 Test Case 6: Simulated concentration profiles of solute transport in a semi-
infinite soil column. Solid lines represent the analytical solution. Data points
represent the numerical solution of FECTUZ C. 1-6
C. 1-7 Test Case 7: Simulated concentration profiles of the problem of solute transport
in a soil column of finite length. Sub-case a involves a conservative species,
while sub-case b involves a non-conservative species. The solid lines represent
the analytical solution, and the data points represent the numerical solution
of FECTUZ C.l-7
C. 1-8 Test Case 8, Sub-case 1: Simulated outflow breakthrough curve of the problem
of solute transport in a layered soil column. The dispersivity values are 0.076 cm,
0.174 cm, and 0.436 cm for layers 1, 2, and 3, respectively. The solid line
represents the analytical solution while the data points represent the numerical
solution of FECTUZ C. 1 -8
C. 1-9 Test Case 8, Sub-case 2: Simulated outflow breakthrough curve of the problem
of solute transport in a layered soil column. The dispersivity values are 0.76 cm,
1.74 cm, and 4.36 cm for layers 1, 2, and 3, respectively. The solid line
represents the analytical solution while the data points represent the numerical
solution of FECTUZ C. 1 -9
C. 1-10 Test Case 9: Comparison between FECTUZ and MOB1 for Test Case 9,
(a) linear adsorption and (b) nonlinear adsorption. The solid line represents
the finite element solution of FECTUZ while the data points represent the finite difference
solution of MOB 1 C.l-10
C. 1-11 Test Case 10: Comparison of the problem of transport with a three-member
decay chain with (a) first-type and (b) third-type decaying source boundary
conditions. The solid line represents the solution by FECTUZ and the data points represent
the analytical solution C. 1-11
in
-------
Attachment C.l
Vadose Zone Verification Results
This page intentionally left blank.
iv
-------
Attachment C.l
Vadose Zone Verification Results
3 --
6 --
X
I—
(X
LU
Q
9 --
12 --
15
- Finite Element
~ Semi-Analytical
-15
•12 -9 -5
PRESSURE HEAD (m)
Figure C.l-1 Test Case 1: Predicted pressure head distribution in the unsaturated
zone. The solid line represents finite element solution, and data points
represent semi-analytical solution.
C.l-1
-------
Attachment C.l
Vadose Zone Verification Results
4 --
JE
X
I—
CL
LU
Q
0 --
10
- Finite Element
o Semi
-Analytical
/
0
o\
J u-
1 —j 1 1
0.20
0.40 0.60 0.80
SATURATION
1.00
Figure C.l-2 Test Case 2: Predicted saturation distribution in the unsaturated zone. The
solid line represents the finite element solution, data points represent the semi-
analytical solution.
C.l-2
-------
Attachment C.l
Vadose Zone Verification Results
— = Finite Element. i45 nodes
0 = Semi-analytical
4 --
nz
i—
Q_
tu
~
B --
8 --
10
0.25
0.50 0.75
SATURATION
1.00
Figure C.l-3 Test Case 3: Predicted saturation distribution in the unsaturated zone. The
solid line represents the finite element solution, and data points represent the
semi-analytical flow solution.
C.l-3
-------
Attachment C.l
Vadose Zone Verification Results
4 --
t-5 yrs
yrs
t=30 yrs
— = Transient
0 = Steady State
8 --
tmQO yrs
+
10 f 1—e-i 1 1—
0.00 0.25 0.50 0.75
Relative Concentration
1.00
Figure C.l-4 Test Case 4: Transport in layered soil. The solid lines correspond to the
transient solution, evaluated at t=5,15,30, and 80 years. Data points represent
the steady-state solution.
C.l-4
-------
Attachment C.l
Vadose Zone Verification Results
1.2
ooooo Numerical FECTUZ
noon a Semi—analytical FECTUZ
HYDRUS
1.0
o
O
0.8
O
c
o
0.6
c
Q)
O
c
o
O
0.4
_o
a>
OL
0.2
0.0
0.6
0.5
0.4
0.3
Time (years)
0.2
o.i
o.o
Figure C.l-5 Test Case 5: Comparison of breakthrough curves for the transient
pulse input transport problem. The solid line represents the solution
of the numerical HYDRUS code, and the data points represent the
semi-analytical and numerical transport solution of FECTUZ.
C.l-5
-------
Attachment C.l
Vadose Zone Verification Results
i.0
O 0 0 g>
-©-e-
Analytic
8
FECTUZ
50 hr
6
25 hr
a)
> .4
2
2
300
350
403
200
250
150
50
100
0.
Distance, cm
Figure C.l-6 Test Case 6: Simulated concentration profiles of solute transport in a semi-
infinite soil column. Solid lines represent the analytical solution. Data points
represent the numerical solution of FECTUZ.
C.l-6
-------
Attachment C.l
Vadose Zone Verification Results
Analytic Soln.
O FECTUZ
C
o
20 d
10 d
0.0
0.
4
e.
Di stance
12.
16.
20.
(a) Au = 0 d 1
Analytic Soln.
O FECTUZ
.8
(O
.6
10 d
.2
20 d
0.0
0
4
8.
12.
16.
20.
Distance, m
(b) X = 0.25 d_1
u
Figure C.l-7 Test Case 7: Simulated concentration profiles of the problem of solute
transport in a soil column of finite length. Sub-case a involves a conservative
species, while sub-case b involves a non-conservative species. The solid lines
represent the analytical solution, and the data points represent the numerical
solution of FECTUZ.
C.l-7
-------
Attachment C.l
Vadose Zone Verification Results
1.0
.3
c
_o
-f-•
CS
C .6
(D
O
C
o
O
<1)
> .4 4-
_cd
d)
DC
CASE 1
2 --
0.0
1
0
FECTUZ
till
630
650
700 750
Time, s
800
850
Figure C.l-8 Test Case 8, Sub-case 1: Simulated outflow breakthrough curve of the
problem of solute transport in a layered soil column. The dispersivity values
are 0.076 cm, 0.174 cm, and 0.436 cm for layers 1,2, and 3, respectively. The
solid line represents the analytical solution while the data points represent the
numerical solution of FECTUZ.
C.l-8
-------
Attachment C.l
Vadose Zone Verification Results
.3 --
C
03 .G
CASE 2
Analytic
FECTUZ
3. 3 t iO
400
500
600 700
Time, s
800
900
100^
Figure C.l-9 Test Case 8, Sub-case 2: Simulated outflow breakthrough curve of the
problem of solute transport in a layered soil column. The dispersivity
values are 0.76 cm, 1.74 cm, and 4.36 cm for layers 1, 2, and 3,
respectively. The solid line represents the analytical solution while the
data points represent the numerical solution of FECTUZ.
C.l-9
-------
Attachment C.l
Vadose Zone Verification Results
LINEAR ISOTHERM
10-
2 20-
m
~0
40 J
50
0.0
0.2
0.4
0.6
0.8
1.0
CONCENTRATION
(a)
NONLINEAR ISOTHERM
FECTUZ
MOB1
£ 20-
m
"0
—i
x
o 30-
3
40-
50
o.s
0.6
CONCENTRATION
0.4
0.2
0.0
CONCENTRA"
(b)
Figure C.l-10 Test Case 9: Comparison between FECTUZ and MOB1 for Test Case
9, (a) linear adsorption and (b) nonlinear adsorption. The solid line
represents the finite element solution of FECTUZ while the data points
represent the finite difference solution of MOB1.
C.l-10
-------
Attachment C.l
Vadose Zone Verification Results
1st type source
0.5 -
0.4 -
0.2 -
0.0
150
100
50
Distance (m)
(a)
0.7
3rd type source
•J3 0.5 -
CJ
0.0
150
100
50
0
Distance (m)
(b)
Figure C.l-11 Test Case 10: Comparison of the problem of transport with a three-
member decay chain with (a) first-type and (b) third-type decaying
source boundary conditions. The solid line represents the solution by
FECTUZ and the data points represent the analytical solution.
C.l-11
-------
Attachment C.l
Vadose Zone Verification Results
This page intentionally left blank.
C.l-12
-------
Attachment C.2
Module-Aquifer Verification Results
(1993-1994)
-------
This page intentionally left blank.
-------
Attachment C.2
Module-Aquifer Verification Results
LIST OF FIGURES
Page
C.2-1 Test Case 1: Comparison of CANSAZ-3D and analytical 3D groundwater
flow solution showing hydraulic heads and Darcy velocities in the horizontal
plane at the top of the saturated zone: (a) CANSAZ-3D, (b) Analytical C.2-1
C.2-2 Test Case 2: Comparison of predicted horizontal transverse concentration
profiles at x=40 m and x=75 m from the source C.2-2
C.2-3 Test Case 3: Comparison of CANSAZ-3D (solid contours) and VAM2D
(dashed contours) C.2-3
C.2-4 Test Case 4: Comparison between analytical solution (solid lines)
and CANSAZ-3D (dashed lines) for 7-member decay chain problem at
t 200 yrs C.2-4
C.2-5 Test Case 5: Predicted concentration distributions at t=5 and t=10 days for
the nonlinear sorption option C.2-5
C.2-6 Test Case 6: Comparison of steady-state concentration contours in the
horizontal plane at the top of the saturated zone (z = 15.71 m); predicted
by (a) CANSAZ-3D and (b) DSTRAM C.2-6
C.2-7 Test Case 7: VAM3D (dashed lines) and CANSAZ/EPACMTP (solid lines)
predicted receptor well concentrations; a) y=0, z=0; b) y=0, z=14, c) y=48,
z 0; d) y 48. z 14 C.2-7
-------
Attachment C.2
Module-Aquifer Verification Results
This page intentionally left blank.
iv
-------
Attachment C.2
Module-Aquifer Verification Results
/ T T / /"A-* -*
~fr* Vf» ^
—» ->
T~» I ~> I -->1 "» I
200 300 <400
Horizontal distance (m)
500
600
(A)
250
£ 200
a)
o 150
D
'td 100
tn
c
P „
4
T ? ? j**'*!-* ->
\: i: r i
lO -* •* —•*¦ T>—*
-------
Attachment C.2
Module-Aquifer Verification Results
0.2
c
o
o
¦£0.1
0)
o
c
o
o
Quasi 3D Numerica
__ __ 3D Analytical
Distance
from
centerline
T
(m)
Figure C.2-2 Test Case 2: Comparison of predicted horizontal transverse concentration profiles
at x=40 in and x=75 m from the source.
C.2-2
-------
Attachment C.2
Module-Aquifer Verification Results
SPECIES 1
20 40 60 80 100 120 140 160 180
DISTANCE ALONG FLOW DIRECTION (m)
SPECIES 2
o
s 4
UJ 2
_J
L±J
0
-0 20 40 60 80 100 120 140 160
DISTANCE ALONG FLOW DIRECTION (m)
Figure C.2-3 Test Case 3: Comparison of CANSAZ-3D (solid contours) and VAM2D (dashed
contours).
SPECIES 3
20 40 60 80 100 120 140 160
DISTANCE ALONG FLOW DIRECTION (m)
C.2-3
-------
Attachment C.2
Module-Aquifer Verification Results
Species 1
8
O.I
%
2
0
Distonce along Centerline fm)
Species 2
4
0
10
Species 3
Species 4.
s
e
*
i
o
no
10
Species 5
Species 6
a
a
4
2
0
1
Species 7
Figure C.2-4 Test Case 4: Comparison between analytical solution (solid lines) and CANSAZ-3D
(dashed lines) for 7-member decay chain problem at t = 200 yrs.
C.2-4
-------
Attachment C.2
Module-Aquifer Verification Results
0.8 -
10 Days
CD 0.6 -
5 Days
0.4
CANSAZ
FECTUZ
M0B1
> 0.2 -
CD 0.0
CK
100
80
60
source (m)
20 40
Distance from
Figure C.2-5 Test Case 5: Predicted concentration distributions at t=5 and t=10 days for the
nonlinear sorption option.
C.2-5
-------
Attachment C.2
Module-Aquifer Verification Results
.800
400
300
E
V '
C
o
oos
100
0
500
1000
1500
Distance along flow direction (m)
(A)
500
400
300
5 200
UJ
0.09
006
100
0-25
0
1500
1000
500
Distance along flow direction (m)
(B)
Figure C.2-6 Test Case 6: Comparison of steady-state concentration contours in the horizontal
plane at the top of the saturated zone (z = 15.71 m); predicted by (a) CANSAZ-3D
and (b) DSTRAM.
C.2-6
-------
Attachment C.2
Module-Aquifer Verification Results
Cei ter I In a i-4s ! I e ( Dep th = 1 4m)
10
TC yr•)
Tlm» C yr»)
Off-center We Me (Dspth
Off-c en ter Uie lis ( Dap th
14)
£
U
T( yr •!
Figure C.2-7 Test Case 7: VAM3D (dashed lines) and CANSAZ/EPACMTP (solid lines)
predicted receptor well concentrations; a) y=0, z=0; b) y=0, z=14, c) y=48, z=0; d)
y=48, z=14.
C.2-7
-------
Attachment C.2
Module-Aquifer Verification Results
This page intentionally left blank.
C.2-8
-------
Attachment C.3
Results
(1993-1994)
-------
This page intentionally left blank.
-------
Attachment C.3
Module-Aquifer Verification Results
LIST OF FIGURES
Page
Figure C.3-1 Test Case 1: Comparison of EPACMTP-Metals result with analytical
solution for the unsaturated-zone with linear adsorption isotherm C.3-1
Figure C.3-2 Test Case 2, Finite Source: Comparison of unsaturated zone concentration
profiles of HYDRUS and EPACMTP-Metals for nonlinear adsorption with
Freundlich exponent = 1.5 C.3-2
Figure C.3-3 Test Case 2, finite source: Comparison of unsaturated zone concentration
profiles of HYDRUS and EPACMTP-Metals for nonlinear adsorption with
Freundlich exponent = 0.5 C.3-3
Figure C.3-4 Test Case 3: Cumulative mass in the unsaturated zone and total input mass
for Lead Isotherm #3 corresponding to low organic acids (LOA), low HFO,
low POM, and high pH condition C.3-4
Figure C.3-5 Test Case 4: Comparison of concentration profiles for Freundlich
exponent = 0.5 C.3-5
in
-------
Attachment C.3
Module-Aquifer Verification Results
This page intentionally left blank.
iv
-------
Attachment C.3
Module-Aquifer Verification Results
0.8
C
o
43 0.6
O
L.
-M
c
CD
O 0.4 -
C
o
o
time
20 d
ooooo EPACMTP— metals
Analytical Solution
0.2 -
20
Depth (m)
Figure C.3-1 Test Case 1: Comparison of EPACMTP-Metals result with analytical
solution for the unsaturated-zone with linear adsorption isotherm.
C.3-1
-------
Attachment C.3
Module-Aquifer Verification Results
2.0
Finite Source (10 years)
Freundlich Exponent =1.5
ooooo HYDRUS
EPACMTP—Metals
(Function form)
aaaaa EPACMTP-Metals
(Tabular form)
C 0.5
t i i i | i i i i i i i i i—|—i—i—i—i—j—i—i—i—r
5 10 15 20
Depth (m)
Figure C.3-2 Test Case 2, finite source: Comparison of unsaturated zone
concentration profiles of HYDRUS and EPACMTP-Metals for
nonlinear adsorption with Freundlich exponent = 1.5.
C.3-2
-------
Attachment C.3
Module-Aquifer Verification Results
2.0
Finite Source (10 years)
Freundlich Exponent =0.5
Cn
ooooo HYDRUS
EPACMTP—Metals
(Function form)
a a a & a EPACMTP—Metals
(Tabular form)
C 0.5
0.0
20
0
10
15
25
5
Depth (m)
Figure C.3-3 Test Case 2, finite source: Comparison of unsaturated zone concentration
profiles of HYDRUS and EPACMTP-Metals for nonlinear adsorption
with Freundlich exponent = 0.5.
C.3-3
-------
Attachment C.3
Module-Aquifer Verification Results
50-
CD
40-
CO
cn
o 30-
U->
_o
D
E
Z5
Q
20-
10-
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
ooooo Computed Cummulative Mass
Total Input Mass
Depth (m)
Figure C.3-4 Test Case 3: Cumulative mass in the unsaturated zone and total input
mass for Lead Isotherm #3 corresponding to low organic acids (LOA),
low HFO, low POM, and high pH condition.
C.3-4
-------
Attachment C.3
Module-Aquifer Verification Results
1 .0 -^-e-e-e-e-e-ea
~ • Analytical
OM
-------
Attachment C.3
C.3-6
Module-Aquifer Verification Results
-------
Attachment C.4
Vadose-Zone Module Verification Results
(1997)
-------
This page intentionally left blank.
-------
Attachment C.4
Vadose-Zone Module Verification Results
LIST OF FIGURES
Page
Figure C.4-1 Test Case 1: Soil water saturation profile, as a function of elevation, where
the water table is located at 5 meters and source infiltration qR = 0.5 m/y . . C.4-1
Figure C.4-2 Test Case 2: Soil water saturation profile, as a function of unsaturated
zone depth for a surface impoundment scenario C.4-2
Figure C.4-3 Test Case 3: Steady-state contaminant concentration profile, where the
ground surface is located at 0.0 m and the water table is located at 5.0 m . . C.4-3
Figure C.4-4 Test Case 4: Contaminant concentration profiles as a function of depth for
four species involved in branched chain decay at time t=250 days. The
ground surface is located at 0.0 m and the water table is located at 5.0 m . . C.4-4
-------
Attachment C.4
Vadose-Zone Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C.4
Vctdose-Zone Module Verification Results
epacmtp
staff3d
j_i
0.4 0.6
Water Saturation
Figure C.4-1 Test Case 1: Soil water saturation profile, as a function of
elevation, where the water table is located at 5 meters and source
infiltration qR = 0.5 m/y.
C.4-1
-------
Attachment C.4
Vctdose-Zone Module Verification Results
s
a
1)
Q
CD
fl
0
N
w
5
0.0 0.2 0.4 0.6 0.8 1.0
Water Saturation
Figure C.4-2 Test Case 2: Soil water saturation profile, as a function of
unsaturated zone depth for a surface impoundment scenario.
C.4-2
-------
Attachment C.4
Vctdose-Zone Module Verification Results
0
8
numerical, steady state
analytical, steady state
¦ . t » I I .1 1 > I I I t r r i r ¦¦>
2 3 4 5
Depth (m)
Figure C.4-3 Test Case 3: Steady-state contaminant concentration profile, where the
ground surface is located at 0.0 m and the water table is located at 5.0 m.
C.4-3
-------
Attachment C.4
Vctdose-Zone Module Verification Results
cmtp, species 1
cmtp, species 2
cmtp, species 3
'cmtp, species 4
vam2d, species 1
vam2d, species 2
vam2d, species 3
vam2d, species 4
2 3
Depth (m)
Figure C.4-4 Test Case 4: Contaminant concentration profiles as a function of depth
for four species involved in branched chain decay at time t=250 days. The
ground surface is located at 0.0 m and the water table is located at 5.0 m.
C.4-4
-------
Attachment C.5
Aquifer Module Verification Results
(1997)
-------
This page intentionally left blank.
-------
Attachment C.5
Aquifer Module Verification Results
LIST OF FIGURES
Page
C.5-1 Test Case 5: Steady state groundwater flow, calculated by EPACMTP
distributions of Darcy velocity vectors (a) and hydraulic head (b) along
the vertical cross-sectional views (z-x); areal (x-y) distributions of Darcy
velocity vector (c) and Hydraulic head (d) at the top of the saturated zone . C.5-1
C.5-2 Steady state groundwater flow field for Test Case 5 calculated using
MNDXYZ: distributions of Darcy velocity vectors (a) and hydraulic
head (b) along the vertical cross-sectional views (x-z); areal (x-y)
distributions of Darcy velocity vectors (c) and hydraulic head (d) at
the top of the saturated zone C.5-2
C.5-3 Test Case 6: Concentrations breakthrough curves at a receptor well
located at the top of saturated zone 100m downgradient oft he source
for simulations obtained with EPACMTP analytical and numerical
(automatic and uniform grid) solutions and VAM3DF solution
(uniform grid) C.5-3
C.5-4 Test Case 7: Contours of base-10 logarithm concentrations for species 4
of a branched chain decay sequence using EPACMTP at time t=500 days
along a) the top of the saturated zone and b) on a cross-section along the
centerline of the grid. Corresponding concentration contours calculated
using STAFF3D are shown along the same c) plan and d) cross-sectional
views C.5-4
-------
Attachment C.5
Aquifer Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C.5
Aquifer Module Verification Results
Cross-sectional View
0 20 40
h)
60 80 100 120
: II 1 1 1 1 1
i- 9! M ! M !
• 11' il i li 1-L '1 t li ill
I I
\ $ I i \
ill J i 1 i
1 1 11
3 § 5 iM u
I J i li .il l—I. ili J 1 li i
c
o
w
c
©
E
V
N1
20
40
60
80
100
120
40 60 80
X-dimension (m)
Figure C.5-1 Test Case 5: Steady state groundwater flow, calculated by EPACMTP
distributions of Darcy velocity vectors (a) and hydraulic head (b) along the
vertical cross-sectional views (z-x); areal (x-y) distributions of Darcy velocity
vector (c) and Hydraulic head (d) at the top of the saturated zone.
C.5-1
-------
Attachment C.5
Aquifer Module Verification Results
Cross-sectional View
cL
Plan View
JO
% 06
I 0
20
40
60
80
100
120
£
=510
>" 5
IN
! I
;
i i
!
1 1
i
|
1
j \
j
oxc
==t=
•J—u
sab:
i=t
111
dssi
=afc»
Isssk
l-h
—;....
sssb
4-1
i=
4-
b-4-
0
20 40 60 80
X-dimension (m)
100
120
Figure C.5-2 Steady state groundwater flow field for Test Case 5 calculated using MNDXYZ:
distributions of Darcy velocity vectors (a) and hydraulic head (b) along the
vertical cross-sectional views (x-z); areal (x-y) distributions of Darcy velocity
vectors (c) and hydraulic head (d) at the top of the saturated zone.
C.5-2
-------
Attachment C.5
Aquifer Module Verification Results
O
O
O
0.06
0.05
0.04
0.03
0.02
0.01
0.00 ' ' ' ' * 1
Time (years)
Figure C.5-3 Test Case 6: Concentrations breakthrough curves at a receptor well located
at the top of saturated zone 100m downgradient oft he source for simulations
obtained with EPACMTP analytical and numerical (automatic and uniform
grid) solutions and VAM3DF solution (uniform grid).
analytical
cmtp, automatic grid
cmtp, uniform grid
vam3df, uniform grid
C.5-3
-------
Attachment C.5
Aquifer Module Verification Results
X-dimension (m)
Figure C.5-4 Test Case 7: Contours of base-10 logarithm concentrations for species 4 of a
branched chain decay sequence using EPACMTP at time t=500 days along a)
the top of the saturated zone and b) on a cross-section along the centerline of
the grid. Corresponding concentration contours calculated using STAFF3D are
shown along the same c) plan and d) cross-sectional views.
C.5-4
-------
Attachment C.6
Composite Model Verification
(1997)
-------
This page intentionally left blank.
-------
Attachment C. 6
Composite Model Verification
LIST OF FIGURES
Page
C.6-1 Test Case 8: Distribution of base-10 logarithm concentrations calculated using
EPACMTP at time t=500 days along a) the top of the saturated zone (xy plane,
z=10.0 m) and b) on a cross-section along the centerline of the grid (xy plane,
y=0.0 m). Corresponding concentration contours calculated using STAFF3D
are shown along the same c) areal and d) cross-sectional views C.6-1
C.6-2 Test Case 9: Cumulative probability curves for exceeding a specified
concentration at a receptor well located along the centerline of the grid,
100.0 m downstream of the source using a) EPACMTP with a uniform grid of
130x15x10 elements in the saturated zone, b) EPACMTP with an automatic
grid, c) EPACMTP with a quasi-3D grid in the saturated zone, and
d) VAM3DF C.6-2
in
-------
Attachment C. 6
Composite Model Verification
This page intentionally left blank.
iv
-------
Attachment C. 6
Composite Model Verification
Z (m) Y (m)
Z (m) Y (m)
ouioui o cn o cn
O 1111111111111111 o II11111111111111
Q.
X
4 g
CD
Z3
cn
o'
o
o
oi
o
o
o
ocnom o cn o cn
O 1111111111111111 o 111111111111111
O" OJ
cn
o
o
cn
o
o
o
Figure C.6-1 Test Case 8: Distribution of base-10 logarithm concentrations calculated
using EPACMTP at time t=500 days along a) the top of the saturated
zone (xy plane, z=10.0 m) and b) on a cross-section along the centerline
of the grid (xy plane, y=0.0 m). Corresponding concentration contours
calculated using STAFF3D are shown along the same c) areal and d)
cross-sectional views.
C.6-1
-------
Attachment C. 6
Composite Model Verification
1.0
= 0.8
¦8
sz
2 0.6
a.
CD
I OA
Z3
E
E 0.2
=3
o
0.0
1—
-1 I 1 1 ll| 1 1 J—"¦ 1 1 1 1 • 1 »l
-
:
-
-
1 ¦¦ 1 —
-
-
epacmtp, uniform grid -
epacmtp, automatic grid
— - — epacmtp, quasi-3D
-
vam3df
.1
..... I 1 . 1 1 _L.I LLlI 1 1 1 1 1 III
10
,-s
10"4
10
10'z
10
10°
C/Co
Figure C.6-2 Test Case 9: Cumulative probability curves for exceeding a specified
concentration at a receptor well located along the centerline of the grid,
100.0 m downstream of the source using a) EPACMTP with a uniform
grid of 130x15x10 elements in the saturated zone, b) EPACMTP with an
automatic grid, c) EPACMTP with a quasi-3D grid in the saturated
zone, and d) VAM3DF.
C.6-2
-------
Attachment C.7
3MRA Vadose-Zone Module Verification Results
(1997)
-------
This page intentionally left blank.
-------
Attachment C. 7
3MRA Vadose-Zone Module Verification Results
LIST OF FIGURES
Page
C.7-1 Test Case 1: Exponentially depleting source with no sorption and no
hydrolysis. Comparison of VZM and EPACMTP C.7-1
C.7-2 Test Case 2: Constant-concentration pulse with no sorption and no
hydrolysis. Comparison of VZM and EPACMTP C.7-1
C.7-3 Test Case 3: Constant-concentration pulse with sorption and hydrolysis.
Comparison of VZM and EPACMTP C.7-2
C.7-4 Test Case 4: Constant-concentration pulse with sorption, hydrolysis, and
chain decay. Comparison of VZM and EPACMTP C.7-2
C.7-5 Test Case 5a: Mercury transport. Comparison of VZM and EPACMTP C.7-3
C.7-6 Test Case 5b: Lead Transport. Comparison of VZM and EPACMTP C.7-3
C.7-7 Test Case 6: Biodegradation Transport. Comparison of VZM and
EPACMTP C.7-4
C.7-8 Test Case 7: Breakthrough Curve at the Water Table for Vadose Zone
Module and Verification Code (MS-VMS) at site LF0223504 for Benzene
(source was terminated after time = 200 years) C.7-5
C.7-9 Test Case 8: Pressure Head Profile for the HWIR99 Vadose Zone Module
and verification code (MS-VMS) at site LF0223504 for Benzene C.7-6
-------
Attachment C. 7
3MRA Vadose-Zone Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C. 7
3MRA Vctdose-Zone Module Verification Results
Depleting Source with Conservative Chemical
Unsaturated Zone Transport
9.00
i.00
7.00
6.00
^ 5.00
o 4.00
3.00
2.00
1.00
0.00
0
20
40
60
80
100
120
140
160
Time (years)
Figure C.7-1 Test Case 1: Exponentially depleting source with no sorption and no hydrolysis.
Comparison of VZM and EPACMTP.
Pulse Source with Conservative Chemical
Unsaturated Zone Transport
12.00
10.00
i.00
o EPACMTP
VZM
— 6.00
4.00
2.00
0.00
0
2
4
6
8
10
12
14
16
18
Time (years)
Figure C.7-2 Test Case 2: Constant-concentration pulse with no sorption and no hydrolysis.
Comparison of VZM and EPACMTP.
C.7-1
-------
Attachment C. 7 3MRA Vctdose-Zone Module Verification Results
Pulse Source with Chemical Decay and Retardation
Unsaturated Zone Transport
3.50
3.00
2.50
a- 2.00
O EPACMTP
VZM
O 1.50
1.00
0.50
0.00
0
5
10
15
20
25
30
Time (years)
Figure C.7-3 Test Case 3: Constant-concentration pulse with sorption and hydrolysis.
Comparison of VZM and EPACMTP.
Chain Decay
Unsaturated Zone Transport
0.25
0.20
0.15
—I
cr
E
:
=
O
0.10
0.05
0.00
0
200
400
600
800
1000
1200
1400
1600
Time (years)
o EPACMTP-P
~ EPACMTP-C
VZM-P
VZM-C
Figure C.7-4. Test Case 4: Constant-concentration pulse with sorption, hydrolysis,
and chain decay. Comparison of VZM and EPACMTP.
C.7-2
-------
Attachment C. 7
3MRA Vctdose-Zone Module Verification Results
Unsaturated Zone Mercury Transport
0.10
.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Time (years)
Figure C.7-5 Test Case 5a: Mercury transport. Comparison of VZM and EPACMTP.
Unsaturated Zone Lead Transport
o EPACMTP
VZM
500000 1000000 1500000
Time (years)
Figure C.7-6 Test Case 5b: Lead Transport. Comparison of VZM and EPACMTP.
C.7-3
-------
Attachment C. 7
3MRA Vctdose-Zone Module Verification Results
Anaerobic Biodegradation
Unsaturated Zone Transport
J
\
\
I
1
1
I
I
V
'
\
0
1
0 200 400 600 800 1000 1200 1400 1600
Time (years)
O E PACT MP-PC E
~ EPACTMP-TCE
A E PACT MP-DC E
EPACTMP-VC
VZM-PCE
VZM-TCE
VZM-DCE
VZM-VC
Figure C.7-7 Test Case 6: Biodegradation Transport. Comparison of VZM and EPACMTP.
C.7-4
-------
-------
Attachment C. 7
3MRA Vctdose-Zone Module Verification Results
Verification Pressure Profile for HWIR99 Vadose Zone Module and MS-VMS
for Benzene at Site LF0223504
3
Pressure Head (m)
¦ ~ ¦ "MS-VMS —B—Vadose Module
Figure C.7-9. Test Case 8: Pressure Head Profile for the HWIR99 Vadose Zone Module
and Veriviation Code (MS-VMS) at site LF0223504 for Benzene.
C.7-6
-------
Attachment C.8
3MRA Anlayses Module Verification Results
(1997)
-------
This page intentionally left blank.
-------
Attachment C.8
3MRA Analyses Module Verification Results
LIST OF FIGURES
Page
C.8-1 Test Case 1: Exponentially depleting source with no sorption nor hydrolysis.
Comparison ofSZMs and 3-D EPACMTP C.8-1
C.8-2 Test Case 2: Constant-concentration pulse with no sorption nor hydrolysis.
Comparison ofSZMs and 3-D EPACMTP C.8-1
C.8-3 Test Case 3: Constant-concentration pulse with sorption and hydrolysis.
Comparison ofSZMs and 3-D EPACMTP C.8-2
C.8-4 Test Case 4a: Constant-concentration pulse with sorption, hydrolysis, and
chain decay. Comparison of SZMs and 3-D EPACMTP, Parent Chemical C.8-3
C.8-5 Test Case 4b: Constant-concentration pulse with sorption, hydrolysis, and
chain decay. Comparison of SZMs and 3-D EPACMTP,
Daughter Chemical C.8-3
C.8-6 Test Case 5: Mercury Transport. Comparison of SZMs and
3-D EPACMTP C.8-4
C.8-7 Test Case 6a: Constant-concentration pulse with biodegradation and chain
decay. Comparison of SZMs and 3-D EPACTMP. Tetrachloroethylene C.8-5
C.8-8 Test Case 6b: Constant-concentration pulse with biodegradation and chain
decay. Comparison of SZMs and 3-D EPACMTP, Trichloroethylene C.8-5
C.8-9 Test Case 6c: Constant-concentration pulse with biodegradation and chain
decay. Comparison of SZMs and 3-D EPACMTP, Dichloroethylene C.8-6
C.8-10 Case 6d: Constant-concentration pulse with biodegradation and chain decay.
Comparison of SZMs and 3-D EPACMTP, Vinyl Chloride C.8-6
C.8-11 Test Case 7: Monte-Carlo results for landfill waste management unit
using HWIR default distributions comparing 3-D EPACMTP and the
pseudo 3-D SZM (SZM-3D1) C.8-7
-------
Attachment C.8
3MRA Analyses Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C.8
3MRA Analyses Module Verification Results
6.00
O 3.00
1.00
20
40
60
100
120
Time (years)
O EPACMTP
SZM-3D3
X SZM-3D1
140
Figure C.8-1 Test Case 1: Exponentially depleting source with no sorption nor
hydrolysis. Comparison of SZMs and 3-D EPACMTP.
^ EPACMTP
SZM-3D3
X SZM-3D1
10.00
Time (years)
Figure C.8-2 Test Case 2: Constant-concentration pulse with no sorption nor hydrolysis.
Comparison of SZMs and 3-D EPACMTP.
C.8-1
-------
Attachment C.8
3MRA Analyses Module Verification Results
0.20
0.18
0.16
0.14
0.12
0.08
0.06
0.04
0.02
0.00
0
5
10
15
20
25
30
35
40
45
50
Time (years)
Figure C.8-3 Test Case 3: Constant-concentration pulse with sorption and hydrolysis.
Comparison of SZMs and 3-D EPACMTP.
C.8-2
-------
Attachment C.8
3MRA Analyses Module Verification Results
0.00016
0.00012
0.00008
O EPACMTP Parent
SZM-3D3 Parent
X SZM-3D1 Parent
200 400 600 800 1000 1200 1400 1600 1800
Time (years)
Figure C.8-4 Test Case 4a: Constant-concentration pulse with sorption, hydrolysis, and
chain decay. Comparison of SZMs and 3-D EPACMTP, Parent Chemical.
0.00018
0.00016 -
0.0001
0.00008 -
0.00002
0 200 400 600 800 1000 1200 1400 1600 800
Time (years)
O EPACMTP Daughter
SZM-3D3 Daughter
X SZM-3D1 Daughter
Figure C.8-5 Test Case 4b: Constant-concentration pulse with sorption, hydrolysis, and
chain decay. Comparison of SZMs and 3-D EPACMTP, Daughter Chemical.
C.8-3
-------
Attachment C.8
3MRA Analyses Module Verification Results
O EPACMTP
SZM-3D3
X SZM-3D1
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time (years)
Figure C.8-6 Test Case 5: Mercury Transport. Comparison of SZMs and 3-D
EPACMTP.
C.8-4
-------
Attachment C.8
3MRA Analyses Module Verification Results
O EPACTMP-PCE
SZM-3D3 PCE
X SZM-3D1 PCE
800
1000 1200 1400 1600
Time (years)
Figure C.8-7 Test Case 6a: Constant-concentration pulse with biodegradation and
chain decay. Comparison of SZMs and 3-D EPACTMP.
Tetrachloroethylene.
O EPACMTP TCE
SZM-3D3 TCE
X SZM-3D1 TCE
800
1000 1200 1400 1600 1800
Time (years)
Figure C.8-8 Test Case 6b: Constant-concentration pulse with biodegradation and
chain decay. Comparison of SZMs and 3-D EPACMTP,
T richloroethylene.
C.8-5
-------
Attachment C.8 3MRA Analyses Module Verification Results
0.20
0.15
O EPACMTP DCE
SZM-3D3 DCE
X SZM-3D1 DCE
0.05
0.00
0
200
400
600
800
1000
1200
1400
1600
1800
Time (years)
Figure C.8-9 Test Case 6c: Constant-concentration pulse with biodegradation and chain
decay. Comparison of SZMs and 3-D EPACMTP, Dichloroethylene.
mm
4.5
4
3.5
3
O EPACMTP VC
SZM-3D3 VC
X SZM-3D1 VC
2.5
2
1.5
1
0.5
0
0
200
400
600
800
1000
1200
1400
1600
1800
Time (years)
Figure C.8-10 Case 6d: Constant-concentration pulse with biodegradation and chain
decay. Comparison of SZMs and 3-D EPACMTP, Vinyl Chloride.
C.8-6
-------
Attachment C.8
3MRA Analyses Module Verification Results
1
0.8
0.7
0.6
jy
§ 0.5
:l:
CL
EPACMTP- 3D
SZM - 3D1
0.3
0.2
0.1
0 1 1 1 1 1 1 1 1 1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Relative Peak Concentration
Figure C.8-11 Test Case 7: Monte-Carlo results for landfill waste management unit using
HWIR default distributions comparing 3-D EPACMTP and the pseudo 3-D
SZM (SZM-3D1).
C.8-7
-------
Attachment C.8
3MRA Analyses Module Verification Results
this page intentionally left blank
C.8-8
-------
Attachment C.9
3MRA Pseudo 3-D Aquifer
Module Verification Results
(1999)
-------
This page intentionally left blank.
-------
Attachment C.9
3MRA Pseudo 3-D Aquifer Module Verification Results
LIST OF FIGURES
Page
C.9-1 Test Case 2: Centerline Breakthrough Curve for HWIR99 And Verification
Modules. X 257. for Well 58 1-1
C.9-2 Test Case 3: Breakthrough Curves for HWIR99 and Verification Modules
at Well 81 1-2
in
-------
Attachment C.9
3MRA Pseudo 3-D Aquifer Module Verification Results
This page intentionally left blank.
iv
-------
5.00E+01
4.50E+01
4.00E+01
3.50E+01
J 3.00E+01
O)
2.50E+01
2.00E+01
1.50E+01
1.00E+01
5.00E+00
O.OOE+OO
0
5
10
15
20
25
30
35
40
45
50
Time (yr)
Verify ~ Module
Figure C.9-1 Test Case 2: Centerline Breakthrough Curve for HWIR99 And Verification Modules, X = 257, for Well 58.
-------
3.0E-05
O
vo
to
2.5E-05
2.0E-05
01
E.
c
% 1.5E-05
c
ai
o
c
o
o
1.0E-05
5.0E-06
0.0E+00
Time (yr)
-Well 81 Verify -X~Well 81 Module
Figure C.9-2 Test Case 3: Breakthrough Curves for HWIR99 and Verification Modules at Well 81.
-------
Attachment C.10
3MRA Vadoze-Zone Module Verification Results
(2000)
-------
This page intentionally left blank.
-------
Attachment C.10
3MRA Vadoze-zone Module Verification Results
LIST OF FIGURES
Page
C. 10-1 Test Area 4: Breakthrough Curve for Conservative Chemical with
a Pulse Source, the Non-metals Transport Component of the
Vadose-zone Module C.10-1
C. 10-2 Test Area 5: Breakthrough Curve for Metal Transport with Non-Linear
Isotherm, the Metals Transport Component for the Vadose-zone Module . C. 10-2
in
-------
Attachment C.10
3MRA Vadoze-zone Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C.10
3MRA Vctdoze-zone Module Verification Results
Breakthrough Curve for Case 6.2
Conservative Chemical, Pulse Source
Stamm
-HWIR99
-EPACMTP
Time (years)
Figure C.10-1 Test Area 4: Breakthrough Curve for Conservative Chemical with a Pulse
Source, the Non-metals Transport Component of the Vadose-zone Module.
C.10-1
-------
Attachment C.10
3MRA Vctdoze-zone Module Verification Results
Breakthrough Curve for Case 7.4
Non-Linear Isotherm, Mercury
"HWIR99
"EPA CMTP
10 20 30 40 50 60 70 80 90 100
Time (years)
Figure C.10-2Test Area 5: Breakthrough Curve for Metal Transport with Non-Linear
Isotherm, the Metals Transport Component for the Vadose-zone Module.
C.10-2
-------
Attachment C.ll
3MRA Aquifer Module Verification Results
(2000)
-------
This page intentionally left blank.
-------
Attachment C.ll
3MRA Aquifer Module Verification Results
LIST OF FIGURES
Page
C. 11-1 Test Area 8: Predicted concentrations at receptor well number 64 by
the Aquifer module and the analytical solution C.l 1-1
C.ll-2 Test Area 8: Predicted concentrations at the receptor well by the
Aquifer module and EPACMTP for straight-chain (Parent to Child)
decay scenario C.ll-2
in
-------
Attachment C.ll
3MRA Aquifer Module Verification Results
This page intentionally left blank.
iv
-------
Attachment C.ll
3MRA Aqui fer Module Veri fication Results
J 0.04
" A qui fer M odule
" V erification
120 140
Time (years)
Figure C.ll-1 Test Area 8: Predicted concentrations at receptor well number 64
by the Aquifer module and the analytical solution.
C.ll-1
-------
Attachment C.ll
3MRA Aqui fer Module Veri fication Results
«-"V
mtfoi
= 8.00E-05 "
• Aquifer Module - Parent
n Aquifer Module - Child
EPACMTP - Parent
-E-EPACMTP - Child
750 1000
Time (years)
Figure C.ll-2 Test Area 8: Predicted concentrations at the receptor well by the
Aquifer module and EPACMTP for straight-chain (Parent to Child)
decay scenario.
C.ll-2
-------
Attachment C.12
EPACMTP Validation Results
-------
This page intentionally left blank.
-------
Attachment C.12
EPACMTP Validation Results
LIST OF FIGURES
Page
C. 12-1 Location of the Borden Landfill showing the monitoring network.
Cross-section A-A' is along longitudinal plume axis (from Frind and
Hokkanen, 1987) C.12-1
C. 12-2 Observed chloride plume along cross-section A-A' in August 1979
(from Frind and Hokkanen, 1987) C.12-2
C.12-3 The simulated chloride plume obtained by the CANSAZ simulation . . . C.12-3
C.12-4 The simulated chloride plume obtained by Frind and Hokkanen, 1987 . C.12-4
C. 12-5 The plan view of the Long Island field site. Groundwater flow directions
are shown with arrows C.12-5
C. 12-6 Comparison between observed and predicted groundwater aldicarb
concentrations for the Long Island site in December, 1970 and
May, 1980 C.12-6
C. 12-7 Plan view of Dodge City, Kansas site located in the Arkansas River
Valley C.12-7
C. 12-8 Comparison of predicted and observed Bromide breakthrough curves at
the Dodge City, Kansas site C.12-8
C. 12-9 Comparison of predicted and observed Triasulfuron breakthrough curves
at the Dodge City, Kansas site C.12-9
C.12-10 EBOS Site 24 layout and location of source area C.12-10
C.12-11 EBOS Site 24 groundwater sampling locations C.12-11
C. 12-12 Changes in the groundwater Napthalene plume over time a) June 1990:
Before source removal b) October 1992: After source removal C. 12-12
C. 12-13 Comparisons along the plume centerline of groundwater Napthalene
concentrations before source removal C. 12-13
C. 12-14 Comparisons along the plume centerline of groundwater Napthalene
concentrations after source removal C. 12-14
in
-------
Attachment C.12
EPACMTP Validation Results
This page intentionally left blank.
iv
-------
BORDEN LANDFILL SITE
ROAD
O*
• •
O0» LANDFILL \
METRES
100
oo
• WATER TABLE WELL
O multilevel sampler
© BUNDLE PIEZOMETER
Figure C.12-1 Location of the Borden Landfill showing the monitoring network. Cross-section A-A' is along longitudinal
plume axis (from Frind and Hokkanen, 1987).
-------
Borden Landfill Site
230
230
225
225
E 220
i\\
220
215
o
210
210
c
100-
o
205
205
o
>
200
200
UJ
195
195
700
800
900
600
1000
500
4 00
0
100
200
300
D i s t o n c e ( m)
Figure C.12-2 Observed chloride plume along cross-section A-A' in August 1979 (from Frind and Hokkanen, 1987).
6
-------
BORDEN LANDFILL SITE
0 200 400 600 800 1000
220
a
100
O 210
Ezq 200
190
0
200
400
600
800
1000
Distance (m)
Figure C.12-3 The simulated chloride plume obtained by the CANSAZ simulation.
-------
BORDEN LANDFILL SITE
230
225.
220.
2)5.
210.
205
200.
195.
o. >00. 200. 300. *00. 500. 600. 700. 800. 900. 1000.
Distance (m)
215.
in
° 210.
C
o
o 20S.
>
V
^ 200.
Figure C.12-4 The simulated chloride plume obtained by Frind and Hokkanen, 1987.
C.12-8
-------
Attachment C.12
EPACMTP Validation Results
PLAN VIEW OF LONG ISLAND FIELD SITE
ORCHARD
{OLD)
STUDY
POTATO
F IELO
POTATOES
kYE
NURSERY
WE TLANO
200 ft
100 m
Figure C.12-5 The plan view of the Long Island field site. Groundwater flow directions are
shown with arrows.
C.12-9
-------
COMPARISON OF EPACMTP WITH FIELD RESULTS
100
Dec. 1979
May. 1980
Observed, Dec. 1979
, May. 1980
EPACMTP.
80-
n
a.
a.
60-
c
o
15
40-
c
a>
o
c
o
o
20
200
150
100
50
Downstream Distance (m)
Figure C.12-6 Comparison between observed and predicted groundwater aldicarb concentrations for the Long Island site in
December, 1970 and May, 1980.
C.12-10
-------
PLAN VIEW OF THE KANSAS FIELD SITE
East-West Distance (m)
©
Groundwater Monitoring Well
[3
Source Area
87.25
Piezometric Head (ft)
Figure C.12-7 Plan view of Dodge City, Kansas site located in the Arkansas River valley.
C.12-11
-------
Attachment C.12
EPACMTP Validation Results
COMPARISON OF PREDICTED AND OBSERVED
BROMIDE BREAKTHROUGH CURVES
c
O
c
(1)
o
c
o
o
MW-2
FIELD DATA
MODEL
-op"
100 200 300 400 500 600 700
Time (days)
mw-3
100 200 300 400 500 600 700
Time (days)
4
MW-5
3
2
't>q .o.^-
0
100 200 300 400 500 600 700
Time (days)
Figure C.12-8 Comparison of predicted and observed Bromide breakthrough curves at the
Dodge City, Kansas site.
C.12-12
-------
COMPARISON OF PREDICTED AND OBSERVED
TRIASULFURON BREAKTHROUGH CURVES
Fieuo data
MODEL
u 100 200 300400 500 600 700
Time (davs^
100 200 300 400 500 S00 700
Tims fdavsl
MW-5
100 200 300 400 500 600 700
i im2 (davsi
Figure C.12-9 Comparison of predicted and observed Triasulfuron breakthrough curves at the Dodge City, Kansas site.
C.12-13
-------
EBOS SITE 24 LAYOUT AND LOCATION OF SOURCE AREA
JP
^terRow
Figure C. 12-10 EBOS Site 24 layout and location of source area.
C.12-14
-------
Attachment C.12
EPACMTP Validation Results
EBOS SITE 24 GROUNDWATER SAMPLING LOCATIONS
0 200
1 I
Scale in F*«t
Transect
© Fringe Wells 0 Centertine Wells • Background Wells ~ Multilevel Samplers
Figure C.12-11 EBOS Site 24 groundwater sampling locations.
C.12-15
-------
Attachment C.12
EPACMTP Validation Results
CHANGES IN GROUNDWATER NAPTHALENE PLUME OVER TIME
a) Before Source Removal
Source
Area
V
b) After Source Removal
0 200
1 i
Concentration
(mg/L) ¦ 2.0 +
F
0 Sample Location
0.5-1.0 ¦ 0.1-0.5 a 0.01-0.1
Source
Removal
Area
Figure C.12-12 Changes in the groundwater Napthalene plume over time a) June 1990:
Before source removal b) October 1992: After source removal.
C.12-16
-------
^1
r
E
Q.
3
C
o
as
c
Cl>
o
c
o
o
Plume Centerline Concentrations
Before Source Removal
Field
EPACMTP
MYGRT
Distance from Trench (m)
Figure C.12-13
Comparisons along the plume centerline of groundwater Napthalene concentrations before source removal.
-------
Plume Centerline Concentrations
After Source Removal
Field
EPACMTP
MYGRT
E
Q.
Q.
tst
c
o
(0
h_
c
CD
O
c
o
o
300
200
250
50
100
150
Distance from Trench (m)
Figure C.12-14
Comparisons along the plume centerline of groundwater Napthalene concentrations after source removal.
-------
Appendix D
Internal Report
Verification and Validation of the SPARC Model
-------
Appendix D Verification and Validation of the SPARC Model
D-ii
-------
Appendix D
Verification and Validation of the SPARC Model
By
S. H. Hilal
Ecosystems Research Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Athens, GA 30605
January 7, 2003
D-iii
-------
Appendix D
Verification and Validation of the SPARC Model
DISCLAIMER
This document is intended for internal Agency use only. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
D-iv
-------
Appendix D
Verification and Validation of the SPARC Model
SUMMARY
The major differences among behavioral profiles of molecules in the environment are
attributable to their physicochemical properties. For most chemicals, only fragmentary
knowledge exists about those properties that determine each compound's environmental fate. A
chemical-by-chemical measurement of the required properties is not practical because of expense
and because trained technicians and adequate facilities are not available for measurement efforts
involving thousands of chemicals. In fact, physical and chemical properties have only been
measured for about 1 percent of the approximately 70,000 industrial chemicals listed by the U.S.
Environmental Protection Agency's Office of Prevention, Pesticides and Toxic Substances [1],
Although a wide variety of approaches are commonly used in regulatory exposure and
risk assessments, knowledge of the relevant chemistry of the compound in question is critical to
any assessment scenario. For volatilization, sorption and other physical processes, considerable
success has been achieved in not only phenomenological process modeling but also a priori
estimation of requisite chemical parameters, such as solubilities and Henry's constant. Granted
that considerable progress has been made in process elucidation and modeling for chemical
processes, such as photolysis and hydrolysis, reliable estimates of the related fundamental
physicochemical properties (i.e., rate and equilibrium constants) have been achieved for only a
limited number of molecular structures. The values of these latter parameters, in most instances,
must be derived from measurements or from the expert judgment of specialists in that particular
area of chemistry.
Mathematical models for predicting the transport and fate of pollutants in the
environment require reactivity parameter values—that is, the physical and chemical constants that
govern reactivity. Although empirical structure-activity relationships that allow estimation of
some constants have been available for many years, such relationships generally hold only
within very limited families of chemicals. On the other hand, we are developing computer
programs that predict chemical reactivity strictly from molecular structure for virtually all
organic compounds. Our computer system called SPARC (SPARC Performs Automated
Reasoning in Chemistry) uses computational algorithms based on fundamental chemical
structure theory to estimate a large array of physical/chemical parameters. See Table D-l for
current SPARC physical property and chemical reactivity parameter estimation capabilities.
In every aspect of SPARC development, from choosing the programming environment to
building model algorithms or rule bases, system validation and verification were important
criteria. The basic mechanistic models in SPARC were designed and parameterized to be
portable to any type of chemistry or organic chemical structure. This extrapolatability impacts
system validation and verification in several ways. First, as the diversity of structures and the
chemistry that is addressable increases, so does the opportunity for error. More importantly,
however, in verifying against the theoretical knowledge of reactivity, specific situations can be
chosen that offer specific challenges. This is important when verifying or validating
performance in areas where existing data are limited or where additional data collection may be
required. Finally, this expanded prediction capability allows one to choose, for exhaustive
validating, the reaction parameters for which large and reliable data sets do exist to validate
D-v
-------
Appendix D
Verification and Validation of the SPARC Model
against. The SPARC models have been validated on more than 10,000 data points. The
verification and validation of the SPARC models will be presented in this report.
Table D-l. SPARC Current Physical And Chemical Properties Estimation Capabilities
Physical Property & Molecular Descriptor
Status
Reaction Conditions
Molecular Weight
Yes
Polarizability
Yes
Temp
a, (3 H-bond
Yes
Microscopic bond dipole
Yes
Density
Yes
Temp
Volume
Yes
Temp
Refractive Index
Yes
Temp
Vapor Pressure
Yes
Temp
Viscosity
Mixed
Temp
Boiling Point
Yes
Press
Heat of Vaporization
Yes
Temp
Heat of formation
UD
Temp
Diffusion Coefficient in Air
Mixed
Temp, Press
Diffusion Coefficient in Water
Mixed
Temp
Activity Coefficient
Yes
Temp, Solv
Solubility
Yes
Temp, Solv
Gas/Liquid Partition
Yes
Temp, Solv
Gas/Solid Partition
Mixed
Temp, Solv
Liquid/Liquid Partition
Yes
Temp, Solv
Liquid /Solid Partition
Mixed
Temp, Solv
GC Retention Times
Yes
Temp, Solv
LC Retention Times
Mixed
Temp, Solv
Chemical Reactivity
Ionization pKa in Water
Yes
Temp, pH
Ionization pK:i in non-Aqueous Solution.
Mixed
Temp, Solv
Ionization pKa in Gas phase
Mixed
Temp
Microscopic Ionization pK(l Constant
Yes
Temp, Solv, pH
Zwitterionic Constant
Yes
Temp, Solv, pH
Molecular Speciation
Yes
Temp, Solv, pH
Isoelectric Point
Yes
Temp, Solv, pH
Electron Affinity
Mixed
Ester Carboxylic Hydrolysis Rate Constant
Yes
Temp, Solv
Hydration Constant
Mixed
Temp, Solv
Tautomer Constant
Mixed
Temp, Solv, pH
E/2 Chemical Reduction Potential
Mixed
Temp, Solv, pH
Mixed : Some capability exists but needs to be tested more, automated and/or extended.
Yes : Already tested and implemented in SPARC
UD: Under Development at this time
Press : Pressure, Temp: Temperature, Solv: Solvent
a: proton-donating site, (3: proton-accepting site.
D-vi
-------
Appendix D
Verification and Validation of the SPARC Model
TABLE OF CONTENTS
INTRODUCTION D-l
PREVIOUS PEER REVIEWS OF THE SPARC SYSTEM D-l
ISSUES REGARDING VERIFICATION AND VALIDATION OF SPARC D-2
SPARC COMPUTAIONAL APPROACH D-3
SPARC PHYSICAL PROPERTIES MODELS D-4
Validation of the SPARC Refractive Index Model D-6
Validation of the SPARC Molecular Volume Model D-8
Validation of the SPARC Vapor Pressure Model D-9
Validation of the SPARC Boiling Point Model D-10
Validation of the SPARC Activity Coefficient Model D-l 1
Validation of the SPARC Solubility Model D-12
Validation of the SPARC Mixed Solvents D-13
Validation of the SPARC Partition Constants Models D-13
Gas/Liquid Partition Model D-13
Liquid/Liquid Partition Model D-l4
Gas/Solid Partitioning Models D-14
Liquid/Solid Partitioning Model D-l5
Validation of the SPARC Diffusion Coefficient in Air Model D-17
SPARC CHEMICAL REACTIVITY MODELS D-l9
Validation of the SPARC pKa in Water Models D-l 9
Validation of the SPARC Carboxylic Acid Ester Hydrolysis Rate Constant Models . D-20
Validation of the SPARC Electron Affinity D-22
D-vii
-------
Appendix D
Verification and Validation of the SPARC Model
SPARC MONOPOLE MODELS D-23
Validation of the SPARC Monopole Models D-23
QUALITY ASSURANCE D-23
CONCLUSION D-23
APPENDIX D-24
REFERENCES D-25
D-viii
-------
Appendix D
Verification and Validation of the SPARC Model
INTRODUCTION
Recent trends in environmental regulatory strategies dictate that EPA will rely heavily on
predictive modeling to carry out the increasingly complex array of exposure and risk
assessments necessary to develop scientifically defensible regulations. The pressing need for
multimedia, multistressor, multipathway assessments, from both the human and ecological
perspectives, over broad spatial and temporal scales, places a high priority on the development of
broad new modeling tools. However, as this modeling capability increases in complexity and
scale, so must the inputs. These new models will necessarily require huge arrays of input data,
and many of the required inputs are neither available nor easily measured. In response to this
need, researchers at NERL-Athens have developed the predictive modeling system SPARC
which calculates a large number of physical and chemical parameters from pollutant molecular
structure and basic information about the environment (media, temperature, pressure, pH, etc.).
Currently, SPARC calculates a wide array of physical properties and chemical reactivity
parameters for organic chemicals strictly from molecular structure. See Table D-l.
SPARC has been in use in the Agency programs for several years, providing chemical
and physical properties to program offices (e.g., Office of Water, Office of Solid Waste and
Emergency Response, Office of Prevention, Pesticides and Toxic Substances) and Regional
Offices. Also, SPARC has been used in Agency modeling programs (e.g., the Multimedia,
Multi-pathway, Multi-receptor Risk Assessment (3MRA) and LENS3, a multi-component mass
balance model for application to oil spills) and to state agencies such as the Texas Natural
Resource Commission. The SPARC web-based calculators have been used by many employees of
various government agencies, academia and private chemical/pharmaceutical companies
throughout the United States. The SPARC web version performs approximately 50,000-100,000
calculations each month. (See the summary of usage of the SPARC web version in the
Appendix.)
PREVIOUS PEER REVIEWS OF THE SPARC SYSTEM
Over the lifespan of its development, the SPARC computer system has undergone
numerous (and various types of) reviews that have helped to establish its validity. For example,
we have published 10 journal articles on the SPARC computer system, each of which has
undergone extensive peer-review (See references 2-11). Also, the SPARC computer system
underwent an EPA Science Advisory Board (SAB) review in 1991, which was relatively early in
its development stage. This multi-day review gave the SPARC development team the
opportunity to demonstrate the system, and to discuss its modeling philosophy with experts in
environmental science. Their comments on the system were very favorable, and they provided
important input on further system development. Following is a brief excerpt from the SAB's
written report. "For a program still in development, progress is excellent. Resources should be
made available to complete the documentation and conduct extensive testing of the model" [12],
Also, the SPARC computer system has been included, along with other projects at our
Division, in several major peer reviews, once in 1997 and again in 2000. These reviews were
D-l
-------
Appendix D
Verification and Validation of the SPARC Model
conducted by "blue-ribbon" panels of scientists from outside of EPA. Again, the comments on
SPARC were laudatory and they provided important input to model development. For example,
the following is an excerpt from the 1997 peer review panel's report on the SPARC project.
"The review panel is extremely impressed with the quality and productivity of this broad
project, as well as with the presentation, the speaker, and the body of work that he summarized
in such a remarkable fashion. This effort represents a central cog in the entire ERD program in
environmental chemistry, as well as a key component of the Division's programmatic support to
the broader agency. Moreover, this body of work represents a truly impressive service to the
larger environmental chemistry community. EPA must find ways ofproviding permanent, long-
term support for the commitment to this effort. It should also work toward making this service
fully available via the internet. This project truly represents a key component of the
NERL/A thens scientific endeavor " [13],
Furthermore, the SPARC developers have frequently engaged in informal consultation with
leaders in relevant fields of science throughout the SPARC model design. These scientists include
the late Dr. Robert Taft of the University of California, Dr. John Garst and Dr. Bruce King of the
University of Georgia, Dr. Ralph Dougherty of the Florida State University and Dr. Samuel
Yalkowski of the University of Arizona.
In summary, the SPARC system has now been extensively reviewed by many renowned
scientists outside of EPA and in many different peer-review processes. Reviewer comments have
always been favorable and the suggestions of these scientists have always been used to improve
further model development. This type of "open communication" with leaders in various fields of
science improves and helps to establish the validity of the SPARC models.
Although development of the SPARC program has been aimed at use in environmental
assessments, these physicochemical models have widespread applicability in the academic and
industrial communities For example, the SPARC program has been used at several universities as
an instructional tool to demonstrate the applicability of physical organic models to the quantitative
calculation of physicochemical properties (e.g., graduate class taught by the late Dr. Robert Taft at
the University of California). Also, the SPARC calculator has great potential for aiding industry
(such as Pfizer, Merck, Pharmacia & Upjohn, etc.) in the areas of chemical manufacturing and
pharmaceutical and pesticide design.
ISSUES REGARDING VERIFICATION AND VALIDATION OF SPARC
To adequately convey the importance of verification and validation of SPARC models, it
is necessary to first describe briefly, and in general terms, the SPARC modeling approach and
philosophy. Indeed, it is useful to compare and contrast the SPARC approach to that of more
conventional models for predicting physicochemical parameters.
Most models that predict a given physicochemical property (e.g., solubility, boiling point,
etc.) are based, in a very direct way, on experimental data for that property for a limited training
set of chemicals. Model development involves finding the best correlations between various
descriptors of chemical structure and the observed property values. These descriptors are
D-2
-------
Appendix D
Verification and Validation of the SPARC Model
subsequently used to construct a model that adequately "recalculates" the training (or
calibration) data set. Then, to validate, one must demonstrate that the empirical model also
accurately predicts property values for chemicals not included in the training set, but whose
experimental values are known. These data are often called the validation set. In order to
predict a new physicochemical property (e.g., octanol/water partition coefficient), the entire
process must be repeated, requiring new training and validation data sets for each new property.
On the other hand, with SPARC, experimental data for physicochemical properties (such
as boiling point) are not used to develop (or directly impact) the model that calculates that
particular property. Instead, physicochemical properties are predicted using a few models that
quantify the underlying phenomena that drive all types of chemical behavior (e.g., resonance,
electrostatic, induction, dispersion, H-bonding interactions, etc.). These mechanistic models
were parameterized using a very limited set of experimental data, but not data for the end-use
properties that will subsequently be predicted. After verification, the mechanistic models were
used in (or ported to) the various software modules that calculate the various end-use properties
(such as boiling point). It is critical to recognize that the same mechanistic model (e.g., H-
bonding model) will appear in all of the software modules that predict the various end-use
properties (e.g., boiling point) for which that phenomenon is important. Thus, any comparison
of SPARC-calculated physicochemical properties to an adequate experimental data set is a true
model validation test — there is no training (or calibration) data set in the traditional sense for
that particular property. The results of validation tests on the various SPARC property models
are presented below in the sections devoted to each property.
The unique approach to SPARC modeling also impacts our strategy for module
verification. For example, when a mechanistic model is updated or improved by incorporating
new knowledge, the impact on all of the various end-use parameters must be assessed. Toward
this end, we have developed quality assurance software that executes each quarter. This
software runs the various property modules for a large number of chemicals (4200 data-point
calculations) and compares the output to historical results obtained over the life-span of the
SPARC program. (Note that, early in our developmental stage, output of all SPARC modules
were compared to hand calculations with selected chemicals to the extent possible. Satisfactory
results were obtained prior to proceeding with further development). In this way, we ensure that
existing parameter models still work correctly after new capabilities and improvements are
added to SPARC. This also ensures that the computer code for all property and mechanistic
models are fully operational. Since the same approach to verification was taken for all property
modules, and since they are all driven by the same verified mechanistic models, we will not
discuss verification in the following sections devoted to each property.
SPARC COMPUTATIONAL APPROACH
SPARC does not do "first principles" computation; rather, it analyzes chemical structure
relative to a specific reactivity query much as an expert chemist might. SPARC utilizes directly
the extensive knowledge base of organic chemistry. Organic chemists have established the
types of structural groups or atomic arrays that impact certain types of reactivity and have
described, in "mechanistic" terms, the effects on reactivity of other structural constituents
D-3
-------
Appendix D
Verification and Validation of the SPARC Model
appended to the site of reaction. To encode this knowledge base, a classification scheme was
developed that defines the role of structural constituents in affecting or modifying reactivity.
Furthermore, models have been developed that quantify the various "mechanistic" descriptions
commonly utilized in structure-activity analysis, such as induction, resonance and field effects.
SPARC execution involves the classification of molecular structure (relative to a particular
reactivity of interest) and the selection and execution of appropriate "mechanistic" models to
quantify reactivity. In brief, the SPARC model consists of a set of core models describing
intra/intermolecular interactions that are linked by the appropriate thermodynamic relationships
to provide estimates of reactivity parameters under desired conditions such as temperature,
pressure and pH. The details of SPARC computational methods are presented in a companion
U.S. E.P.A report, "Estimation of Physical Properties and Chemical Reactivity Parameters from
Molecular Structure using SPARC" [14], Hence, only an overview will be given here.
For physical properties, intermolecular interactions are expressed as a summation over all
the interaction forces between molecules (i.e., dispersion, induction, dipole and H-bonding).
Each of these interaction energies is expressed in terms of a limited set of molecular-level
descriptors (volume, molecular polarizability, molecular dipole, and H-bonding parameters) that,
in turn, are calculated from molecular structure. For chemical reactivity, molecular structure is
broken into functional units. Reaction centers with known intrinsic reactivity are identified and
the impact on reactivity of appended molecular structure is quantified using mechanistic
perturbation models.
A "toolbox" of mechanistic perturbation models has been developed that can be
implemented where needed for a specific reactivity query. Resonance models were developed and
validated on more than 5000 light absorption spectra [1, 2], whereas electrostatic interaction
models were developed and validated on more than 4500 ionization pKas in water [3-8], Solvation
models (i.e., dispersion, induction, H-bond and dipole interactions) have been developed and
validated on more than 8000 physical property data points on properties such as vapor pressure,
boiling point, solubility, Henry's constant, GC chromatographic retention times, Kow, etc [3, 9,
10], The SPARC computational approach is based on blending well known, established methods
such as SAR (Structure Activity Relationships) [15, 16], LFER (Linear Free Energy
Relationships) [17, 18] and PMO (Perturbed Molecular Orbital) theory [19, 20], SPARC uses
SAR for structure activity analysis, such as induction, resonance and field effects. LFER is used to
estimate thermodynamic or thermal properties and PMO theory is used to describe quantum effects
such as charge distribution derealization energy and polarizability of the n electron network.
SPARC PHYSICAL PROPERTIES MODELS
For all physical properties (e.g., vapor pressure, boiling point, activity coefficient,
solubility, partition coefficients, GC/LC chromatographic retention times, diffusion coefficients,
etc.), SPARC uses one master equation to calculate characteristic process parameters:
Process Interaction Other ^
D-4
-------
Appendix D
Verification and Validation of the SPARC Model
where AGinteraction describes the change in the energy associated with the intermolecular
interactions accompanying the process in question. For example, in liquid to gas vaporization,
AGinteraction describes the difference in the energy associated with intermolecular interactions in the
gaseous phase versus that associated with interactions in the liquid phase. The intermolecular
interaction forces between the molecules are assumed to be additive. The AGother lumps all non-
interaction energy components such as entropy changes associated with mixing or expansion, and
changes in internal molecular (vibrational, rotational) energies. At the present time, the
intermolecular interactions in the liquid phase are modeled explicitly, interactions in the gas phase
are ignored, and molecular interactions in the crystalline phase are extrapolated from the subcooled
liquid state using the melting point. The 'non-interaction' entropy components are process specific
and will be described later, in the vapor pressure and the activity coefficient models. The
intermolecular interactions in the liquid phase are expressed as a summation over all the mechanistic
components:
AG'Interaction AGDispersion AGInduction AGDipole-dipole AGH-Bond 0^
Each of these interaction mechanisms is expressed in terms of a limited set of pure
component descriptors (liquid density-based volume, molecular polarizability, microscopic bond
dipole, and hydrogen bonding parameters), which in turn are calculated strictly from molecular
structure [3, 9],
Dispersion interactions are present in all molecules, including polar and non-polar
molecules. Induction interactions are present between two molecules when at least one of them has a
local dipole moment. Dipole-dipole interactions exist when both molecules have local dipole
moments. H-bonding interactions exist when a; J3j or otj (3; products are non zero, where a represents
the proton donation strength and (3 represents the proton acceptor strength. In SPARC, all the
physical property estimations derive from a common set of core models describing
intra/intermolecular interactions, and require as user inputs molecular structure (both solute and
solvent(s)) and reaction conditions of interest (temperature, pressure, etc.). The self-term, AG,,
(solute-solute) interaction model is used to describe the vapor pressure at 25° C. The self terms, AG;;
and AGjj (solvent-solvent) plus the cross term, AGy (solute-solvent) interactions, are required to
describe the solute, i, activity coefficient in any solvent, j at 25° C.
Like the chemical reactivity models, the AG;;, AGy and AGjj models have been extended and
validated on numerous physical properties under different reaction conditions such as temperature,
pressure and solvents. The self-term interaction model has been tested on a large number of vapor
pressures, boiling points, diffusion coefficients and heat of vaporization. Likewise, the solute-
solvent interaction model has been validated on activity coefficients, solubilities, partition
coefficients and GC/LC chromatographic retention times in any solvent at any temperature.
D-5
-------
Appendix D
Verification and Validation of the SPARC Model
Validation of the SPARC Refractive Index Model
The molecular polarizability and volume can be related to the index of refraction (n)
using the Lorentz-Lorenz equation. For our units of cmVmole for volume (V) and A3/molecule
for polarizability (P), the Lorentz-Lorenz equation can be written as
n2 -1 4^(0.6023^)
n2 +2 ~ 3V (^D~3'>
The refractive index output was initially verified by comparing the SPARC prediction
value to hand calculations for selected key molecules. The refractive index calculator was trained
on 325 non-polar and polar organic compounds at 25° C then validated on 578 organic
compounds at 25° C [9, 10] as shown in Figure D-l. The statistical performance for the SPARC
refractive index calculator is shown in Table D-2. See reference 9 for sample hand calculations.
D-6
-------
Appendix D
Verification and Validation of the SPARC Model
Table D-2. SPARC Physical And Chemical Properties Calculator Statistical Performance
Versus Observations
Property
Units
Total#
Molecule
RMS
R2
Reaction
Conditions
Temp/Solvent
Refractive Index
N/A
578
0.007
0.997
25
Volume
g/cm3
1440
1.97
0.999
25
Vapor Pressure
logatm
747
0.15
0.994
25
Boiling Point
°C
4000
5.71
0.999
0.1-1520 torr
Heat of Vaporization3
Kcal/mole
1263
0.301
0.993
25, Boiling Point
Diffusion Coefficient in Air4
cm2/s
108
0.003
0.994
25
Activity Coefficient
logMF5
491
0.064
0.998
25,41 solvents
Solubility
logMF
647
0.40
0.987
25, 21 solvents
Distribution Coefficient
N/A
623
0.43
0.983
25 Octanol,
Toluene CC14,
Benzene,
Cyclohexane, Ethyl
Ether
Henry's Constant
(mole/L)2
286
271
0.34
0.10
0.990
0.997
25, Water
25, Hexadecane
GC Retention Time2
Kovtas
295
10
0.998
25-190, Squalane,
B18
LC Retention Time
Kovtas
125
0.095
0.992
25, Water/Methanol
Gas pKa3
Non-aqueous pKa3
pKa in water
Kcal
Kcal
Kcal/1.36
400
300
4338
2.25
1.90
0.356
0.999
0.960
0.994
25, Alcohols,
Aceteonitrile,
Acetic acid, DMF1,
THF1, pyridine
25-100, Water
Electron Affinity
e.V.
260
0.14
0.98
Gas
Ester Carboxylic Hydrolysis
Rate
M-y
1470
0.37
0.968
25-130C, Water,
Acetone,
Alcohols, Dioxane,
Aceteonitrile
Tautomer Constant3
Hydration Constant3
Kcal/1.36
Kcal/1.36
36
27
0.3
0.43
0.950
0.744
25 C, Water
E/2 Chemical Reduction3
e.V.
352
0.18
0.95
25, Water,
Alcohols, DMF1
Aceteonitrile,
DMSO1
1 DMF: N,N -dimethylforamide; DMSO: Dimethyl sulfoxide; THF: Tetrahydrofuran.
2 GC retention times in SE-30 and PEG-20M liquid stationary liquid phase is not included in this report.
3 See the companion SPARC report [14],
4 Models were developed after the HWIR exercise.
5 MF: mole fraction.
D-7
-------
Appendix D
Verification and Validation of the SPARC Model
T3
1.52
a>
re
3
1.47
O
re
O
1.42
O
0£
<
CL
1.37
tn
1.32
y = 1.0101X- 0.0152
1.32 1.37 1.42 1.47
Observed
1.52
Figure D-l. SPARC-calculated versus observed refractive index at 25° C.
The RMS (Root Mean Square) deviation was 0.007 and R2 was 0.997.
Validation of the SPARC Molecular Volume Models
The zero order density-based molecular volume at 25° C is expressed as
y°=Wfr"g-A,)
(D-4)
where Vifrag is the volume of the ith molecular fragment and A; is a correction to that
volume based on both the number and size of fragments attached to it. The Vifrag are determined
empirically from measured liquid-density based volumes, and then stored in the SPARC
database. The zero order volume at 25° C is further adjusted for changes resulting from dipole-
dipole and hydrogen bonding intermolecular interactions:
F,. = F„
A
dipole - dipole
ZD,
1
F °
A
a
H - bond
F„
(D-5)
where D, is the weighted sum of the local dipole for the molecule, and a and P are the H-
bonding parameters of potential proton donor and proton acceptor sites within the molecule,
respectively. Adipoie-diPoie and AH-bond are adjustment constants due to dipole-dipole and H-
bonding, respectively. The final molecular volume at any temperature T is then expressed as a
polynomial expansion in (T-25) corrected for H- bonding, dipole density and polarizability
density interactions [9, 14],
The molecular volume can be calculated within 2 cm3 mole"1 for most organic molecules.
Figure D-2 shows the SPARC-calculated versus observed molecular volumes for both polar and
non-polar compounds at 25° C. The statistical performance for the volume calculator is in
Table D-2. See reference 9 for sample hand calculations.
D-8
-------
Appendix D
Verification and Validation of the SPARC Model
Q)
O
E
CO
E
o
¦D
0)
ra
O
6
0£
<
Q.
<0
0 .9 9 6 6 x + 0.2 982
10 0 0
8 0 0
6 0 0
4 0 0
2 0 0
2 0 0
4 0 0
6 0 0
8 0 0
Observed (cm /mole)
10 0 0
Figure D-2. SPARC-calculated vs. observed-liquid density based volume at 25° C for
1440 organic molecules. The RMS deviation was 1.97 cm3 mole"1 and R2 was 0.999.
Validation of the SPARC Vapor Pressure Model
The saturated vapor pressure is one of the most important physiochemical properties of
pure compounds. By 1978, vapor pressure data (as a function of temperature) were available for
more than 7000 organic compounds [21], Despite the frequency of reporting in the published
literature, the number of compounds where the vapor pressure was truly measured and not
extrapolated to 25° C from higher temperature measurements is limited. Most of the measured
25° C vapor pressure data are for compounds that are either pure hydrocarbons or molecules that
have relatively small dipole moments and/or weak hydrogen bonds. There is a pressing need to
predict the vapor pressure of those compounds that have not been measured experimentally. In
addition to being highly significant in evaluating a compound's environmental fate, the vapor
pressure at 25° C provides an excellent arena for developing and testing the SPARC self interac-
tion physical process models.
The vapor pressure, vp°, of a pure solute, i, can be expressed as function of all the
intermolecular interaction mechanisms, A Gn (interaction), as
o
log vpj
- AG.. (Interaction )
LogT + C
(D-
2.303 RT
6)where log (T) + C describes the change in the entropy contribution associated with the volume
change in going from the liquid to the gas phase. The crystal energy term (given in reference
14), CE, must be added to equation 6 for molecules that are solids at 25° C, the CE contribution
becomes important, especially for rigid structures such as aromatic or ethylenic molecules that
have high melting points [14],
D-9
-------
Appendix D
Verification and Validation of the SPARC Model
The vapor pressure computational algorithm output was initially verified by comparing the
SPARC prediction of the vapor pressure at 25° C to hand calculations for key molecules. Since the
SPARC self interactions model, AG;;, was developed initially on this property, the vapor pressure
model undergoes the most frequent validation tests. The calculator was trained on 315 non-polar
and polar organic compounds at 25° C. Figure D-3 presents the SPARC-calculated vapor
pressure at 25° C versus measured values for 747 compounds. The SPARC self-interactions
model can predict the vapor pressure at 25° C within experimental error over a wide range of
molecular structures and measurements (over 8 log units). For simple structures, SPARC can
calculate the vapor pressure to better than a factor of 2. For complex structures such as some of
the pesticides and pharmaceutical drugs where dipole-dipole and/or hydrogen bond interactions
are strong, SPARC calculates the vapor pressure within a factor of 3-4. The statistical
performance for the vapor pressure calculator is shown in Table D-2. See references 9 and 14
for sample hand calculations. The vapor pressure model was also tested on the boiling point and
heats of vaporization [9, 14],
Figure D-3. SPARC-calculated vs. observed log vapor pressure for 747 organic molecules at
25° C. The figure includes all the vapor pressure measurements (real not extrapolated)
we found in the literature. The RMS deviation error was 0.15 log atm and R2was 0.994.
Validation of the SPARC Boiling Point Model
SPARC estimates the boiling point for any molecular species by varying the temperature
at which a vapor pressure calculation is done. When the vapor pressure equals the desired
pressure, then that temperature is the boiling point at that pressure. The normal boiling point is
calculated by setting the desired pressure to 760 torr. Boiling points at a reduced pressure can be
calculated by setting the desired pressure to a different value.
D-10
-------
Appendix D
Verification and Validation of the SPARC Model
SPARC temperature dependence models were developed initially on the boiling point.
The boiling point calculator was trained on 1900 boiling points for a wide range of non-polar and
polar organic compounds. The calculator was validated against 4000 boiling points measured at
different pressures ranging from 0.05 to 1520 torr spanning a range of over 800° C as shown in
Figure D-4.
o
T3
-------
Appendix D
Verification and Validation of the SPARC Model
presents the validation for SPARC-calculated log activity coefficients versus measured values
for 491 compounds at 25° C in 41 different solvents. The SPARC activity coefficient test
statistical parameters are shown in Table D-2. The activity coefficients calculator was also
tested on the solubility in more than 20 different solvents and partition coefficients in more than
18 different solvents. See following sections for more details.
-2 0 2 4 6 8
Observed (log mole fraction)
Figure D-5. SPARC-calculated versus observed log activity coefficients at infinite dilution
for 491 compounds in 41 solvents including water. Only 15% of these compounds
have strong dipole-dipole and/or H-bond interactions. The RMS deviation was
0.064 log mole fraction and with an R2 of 0.998.
Validation of the SPARC Solubility Model
SPARC does not calculate the solubility from first principles, but rather from the infinite
dilution activity coefficient model discussed previously. SPARC first calculates the infinite
dilution activity coefficient, y°°; when log y°° is greater than 2, the mole fraction solubility can be
reliably estimated as xso1 = l/y°°. However, when the log y°° is calculated to be less than 2, this
approximation fails. In these cases, y°° is greater than yso1 and SPARC would underestimate the
solubility using the inverse relationship. In order to overcome this limitation, SPARC employs
an iterative calculation. SPARC sets the initial guess of the solubility as "/"'guess = l/y°°. SPARC
then 'prepares' a mixed solvent that is xsolguess in the solute and (1- xsolguess) in the solvent.
SPARC then recalculates y°° in the 'new' solvent and the corresponding %solgll0ss. This process is
continued until y°° converges to 1 (miscible). The solubility calculator spans more than 12 log
mole fraction as shown in Figure D-6. The RMS deviation was 0.40 log mole fraction, which
was close to the experimental error. SPARC estimates the solubility for simple organic
molecules to better than a factor of 2 (0.3 log mole fraction) and within a factor of 4 (0.6 log
mole fraction) for complicated molecules like pesticides and pharmaceutical drugs. The RMS
deviation for the solids compounds is 3 times greater than the RMS deviation for liquids
compounds due to the crystal energy contributions. For more details see reference 14. The
statistical parameters for calculated log solubility for 647 organic molecules in 21 different
solvents including water at 25° C are shown in Table D-2.
D-12
-------
Appendix D
Verification and Validation of the SPARC Model
Figure D-6. Test results for SPARC calculated log solubilites for 260 compounds.
The RMS deviation is 0.321 and R2 is 0.991. The RMS deviation for 119 liquid
solubilities is 0.135 and R2 is 0.997 while for the 141 solids compounds the
RMS deviation is 0.419 and R2 is 0.985.
Validation of the SPARC Mixed Solvents Model
SPARC can handle solvent mixtures for a large number of components. However, speed
and memory requirements usually limit the number of solvent components to less than twenty on
a PC. The user specifies the name and volume fraction for each solvent component. Each of the
solvent components must have been previously initialized as a solvent. SPARC will allow the
user to specify a name for the mixture so that it can be used later as a 'known' solvent. The
activity coefficients (or solubility) of molecules in binary solvent mixtures have been tested and
appear to work well. Figure D-7 shows the calculated log y in a water/methanol mixture versus
measured values. For more details see reference 14.
Validation of the SPARC Partition Constants Models
All partition (Liquid/Liquid, Liquid/Solid, Gas/Liquid, Gas/Solid) constants are
determined by calculating the activity coefficient of the molecular species of concern in each of
the phases without modification or extra parameterization to the activity coefficient model.
Gas/liquid (Henry's Constant) Model
Henry's constant may be expressed as
Hx = vp° fn (D-8)
where vp,° is the vapor pressure of pure solute i (liquid or subcooled liquid) and yij°° is the
activity coefficient of solute (i) in the liquid phase (j) at infinite dilution. SPARC vapor pressure
D-13
-------
Appendix D
Verification and Validation of the SPARC Model
6
o
y = 0.9535x +0.1422
0
2
3
4
5
Observed (log mole fraction)
Figure D-7. SPARC-calculated versus observed log activities for
120 compounds in water/methanol mixed solvent at 25° C.
The RMS deviation error was 0.18 and the R2 was 0.980.
and activity coefficient models can be used to calculate the Henry's constant for any solute out
of a mixed solute-solvent liquid phase. An application of Henry's law constant for the prediction
of gas-liquid chromatography retention time is given in the companion SPARC report [14],
Liquid/Liquid Partitioning Model
SPARC calculates the liquid/liquid partition constant, such as the octanol/water
distribution coefficient, by simply calculating the activity of the molecular species in each of the
liquid phases as
where the y°°s are the infinite dilution activities in the two phases and Rm is the ratio of
the molecularites of the two phases (Mi/M2). Although octanol/water partition coefficients are
widely used and measured, the SPARC system does not limit itself to this calculation. SPARC
can calculate the liquid/liquid partition coefficient for any two immiscible phases.
Gas/Solid Partitioning Model
SPARC calculates gas/solid partitioning in a manner similar to gas/liquid partitioning.
For the solid phase, the solvent self-self interactions, AGjj, are dropped from the calculation
when one of the phases is solid. This type of modeling will be useful for calculating retention
times for capillary column gas chromatography.
CO
CO
log Kuqimq2 = l°g//(y2 - l°g//(;| + log Rm
(D-9)
D-14
-------
Appendix D
Verification and Validation of the SPARC Model
Liquid/Solid Partitioning Model
SPARC calculates liquid/solid partitioning in a manner similar to liquid/liquid
partitioning. For the solid phase, the solvent self-self interactions, AGy, are dropped from the
calculation.
The gas/liquid models have been extensively tested against observed Henry's constant
measurements. The two largest data sets are air/water and air/hexadecane systems. The
liquid/solid and gas/solid partitioning models are implemented in code but have not been
extensively tested. The liquid/liquid partitioning models are the most extensively tested
partitioning models due to the large octanol/water data sets available. The statistical parameters
for SPARC-calculated partition constants in many solvents at 25° C are shown in Table D-2.
Figure D-8 shows calculated versus observed Henry's constant for compounds dissolved in
hexadecane. Figure D-9 shows the current general performance of SPARC for log Ksoivent/water,
where the solvents are carbon tetrachloride, benzene, cyclohexane, ethyl ether, octanol and
toluene. Figure D-10 displays a comparison of the EPA Office of water (OW) recommended
observed octanol-water distribution coefficients versus SPARC and C log P calculated values.
The RMS deviation and R2 values were is 0.18 and 0.996 respectively for SPARC and 0.44 and
0.978 respectively for ClogP calculated values [22],
O
-8
O)
o
¦D
a)
-8
-3
2
Observed log Kow
7
Figure D-8. Observed vs. SPARC-calculated Henry's constants for 271 organic
• • • *2
comnounds in hexadecane. The RMS deviation was 0.1. while the R was 0.997.
D-15
-------
Appendix D
Verification and Validation of the SPARC Model
^ 2
oS
0 0
•§- -2
¦a
B -4
re
1 "6
8 -s
y = 0.998x - 0.006
-10
Observed (mole/L)
Figure D-9. SPARC-calculated versus observed log distribution
coefficients KsoiVent/waterfor 623 organic compounds in six solvents at
25° C. The RMS deviation was 0.38 and R2 was 0.983.
Observed log Kow
Figure D-10. Test of OW for calculated Koctanoi/water versus measured values.
Squares are SPARC calculate values, circles are ClogP calculate values. The RMS
deviation and R2 values were 0.18 and 0.996 respectively for SPARC and
0.44 and 0.978, respectively, for ClogP calculated values.
D-16
-------
Appendix D
Verification and Validation of the SPARC Model
Validation of the SPARC Diffusion Coefficient in Air Model
Several engineering equations exist that do a very respectable job of calculating
molecular diffusion coefficients in air over wide ranges of temperature and pressure. The
equation most compatible with the SPARC calculator is also the relationship that seems to
perform the best over a wide variety of molecules. This equation is that of Wilke and Lee [23],
which for binary diffusion coefficient is expressed as:
rri3/ 2
Dab = [3.03 - (0.98 / Mab)](10~3) ^ (D-
PMabCJabQd
10)where Dab is the binary diffusion coefficient in cm2/s, T is the temperature in K, MA and MB
are the molecular weights of A and B in g/mol and MAb is 2[(1/MA) + (1/Mb)]"1 and P is the
pressure in bar. The Qd is a complex function of T and has been accurately determined by
Neufeld [24],
SPARC predicts gas phase binary diffusion coefficients at any temperature and pressure
to better than 6% as shown in Figure D-l 1. The statistical parameters are shown in Table D-2.
0.2
T3
a>
re
0.15
y = 0.9762x +0.0021
3
O
re
O
0.1 -
6
£
<
a.
CO
0.05
0 h
<
,
0 0.05 0.1 0.15
0.2
Observed (cm2/s)
Figure D-ll. SPARC-calculated vs. observed
diffusion coefficient.
D-17
-------
Appendix D
Verification and Validation of the SPARC Model
The overall SPARC physical properties training set output is shown in Figure D-12. The
training set includes vapor pressure (as a function of temperature), boiling point (as a function of
pressure), diffusion coefficients (as a function of pressure and temperature), heat of vaporization
(as function of temperature), activity coefficient (as a function of solvent), solubility (as a
function of solvent and temperature), GC retention times (as a function of stationary liquid phase
and temperature) and partition coefficients (as a function of solvent). This set includes more
than 50 different pure solvents (see Table D-3) as well as 18 mixed solvent systems. The
observed measured values for the training and validations sets were from many sources such as
references 26-34. For other SPARC physical properties models such as GC/LC retention time
in polar and non-polar liquid phase, heat of vaporization and diffusion coefficient in water, see
reference 14.
300 n y = 0.9981x + 0.0878
a> 100
1 "100
O
-300
-300
-100
100
300
Observed
Figure D-12. SPARC-calculated vs. 2400 observed training
set physical property values. The aggregate RMS is 0.29
and R2 is 0.997. For more details see text.
Table D-3. Solvents that Have Been Tested in SPARC
Chloroform
1-butanol
1-chloro hexadecane
1-dodecanol
OV-101
1-propanol
butanone
1-nitro propane
2-dodecanone
isopropanol
isobutanol
acetone
2-nitro propane
aceteonitrile
PEG-20M
benzyl ether
benzene
benzylchloride
benzonitrile
SE-30
cyclohexane
decane
bromobenzene
butronitrile
pyridine
cyanohexane
ethanol
dioctyl ether
cyano cyclohexane
water
heptane
hexane
hexadecane
heptadecane
squalane
methanol
nonane
1-butyl chloride
nitrobenzene
1-me naphthalene
nitroethane
octane
nitro cyclohexane
nitro methane
2-me naphthalene
nonanenitrile
squalene
pentadecane nitrile
isoquinoline
m-cresol
quinoline
phenol
1,2,4 trichlorobenzene
hexafluorobenzene
p-xylene
D-18
-------
Appendix D
Verification and Validation of the SPARC Model
SPARC CHEMICAL REACTIVITY MODELS
SPARC reactivity models have been designed and parameterized to be portable to any
chemical reactivity property and any chemical structure. For example, chemical reactivity
models are used to estimate ionization pKa, zwitterionic constant, isoelectric point and speciation
fractions as a function of pH. The same reactivity models are used to estimate gas phase electron
affinity and ester hydrolysis rate constants in water and in non-aqueous solutions.
Validation of the SPARC pKa in Water Models
Like all chemical reactivity parameters addressed in SPARC, molecular structures are
broken into functional units called the reaction center and the perturber in order to estimate pKa
in water. The reaction center, C, is the smallest subunit that has the potential to ionize and lose a
proton to a solvent. The perturber, P, is the molecular structure appended to the reaction center,
C. The pKa of the reaction center is adjusted for the molecule in question using the mechanistic
perturbation models. The pKa for a molecule of interest is expressed in terms of the
contributions of both P and C.
where (pKa)c describes the ionization behavior of the reaction center, and 5p(pKa)c is the
change in ionization behavior brought about by the perturber structure given as
where 8respKa, 5eiepKa and 5S0ipKa describe the differential resonance, electrostatic and
solvation effects of P on the initial and final states of C, respectively.
The SPARC pKa calculator was trained on 2500 organic molecules, then validated on
4338 pKa's (4550 including carbon acid) in water as shown in Figure D-13 and Table D-4. The
calculator was tested for multiple ionization's up to the 6th (simple organic molecules) and 8th (azo
dyes) for molecules with multiple ionization sites. In addition, the pKa models were tested on all
the literature values we found for zwitterionic constants (12 data points), the thermodynamic
microscopic ionization constants, pk,, of molecules with multiple ionization sites (120
measurement data points, the RMS deviation error is 0.5), the corresponding complex speciation
as a function of pH and the isoelectric points (29 measurement data points) in water. The
diversity and complexity of the molecules used was varied over a wide range in order to develop
more robust models during the last few years. Hence, the SPARC pKa models are now very
robust and highly tested against almost all the available experimental literature data.
While it is difficult to give a precise standard deviation of a SPARC calculated value for
any given individual molecule, in general SPARC can calculate the pKa for simple molecules
such as oxy acids and aliphatic bases and acids within ±0.25 pKa units; ±0.36 pKa units for most
other organic structures such as amines and acids; and ±0.41 pKa units for =N and in-ring N
PKa = (PKa)c+SP(pKa)c
(D-ll)
8„(pKa)c = 8elepKa + 5respKa + 8solpKa ...
(D-12)
D-19
-------
Appendix D
Verification and Validation of the SPARC Model
reaction centers and for complicated structures. Where a molecule has more than six ionization
sites (n > 6), the expected SPARC error is ±0.65 pKa units. For more details see reference 14.
0.9 9 2 5
0 18 9
TO
*
Q.
O
a.
<
D.
V)
¦1 2
¦1 2
Observed pKa
Figure D-13. SPARC-calculated versus observed for 4338 pKa's of 3685 organic com-
pounds. The RMS deviation was equal to 0.37. This test does not include carbon acid
reaction center. The majority of the molecules are complex compounds. Some of the
molecules such as azo dyes have 8 different ionization sites.
Table D-4. Statistical Parameters of SPARC pKa Calculations
Set
Training
R2
RMS
Test
R2
RMS
Simple organic compounds
793
0.995
0.235
2000
0.995
0.274
Azo dyes compounds
50
0.991
0.550
273
0.990
0.630
IUPAC compounds1
2500
0.994
0.356
43382
0.994
0.370
1 Observed values are from many ref. such as 35-36.
2 Carbon acid pKas are not included.
Validation of the SPARC Carboxylic Acid Ester Hydrolysis Rate Constant Models
Reaction kinetics were quantitatively modeled within the chemical equilibrium
framework described previously for ionization pKa in water. It was assumed that a reaction rate
constant could be described in terms of a pseudo equilibrium constant between the reactant and
transition states. SPARC therefore follows the modeling approach described for pKa. For these
chemicals, reaction centers with known intrinsic reactivity are identified and the reaction rate
constants expressed by perturbation theory as
D-20
-------
Appendix D
Verification and Validation of the SPARC Model
log k = logfc+A,log& (°-13)
where log k is the log of the rate constant of interest; log kc is the log of the intrinsic rate
constant of the reaction center and Aplog kc denotes the perturbation of the log rate constant due
to the appended structure.
The ester hydrolysis rate constant models have been tested to the maximum extent possible as
function of temperature and solvent. The RMS deviation error for 1470 hydrolysis rate constants
in 6 solvents and at different temperature was 0.37 as shown in Figure D-14. In this test, there
were 653, 667 and 150 base, acid and general base catalyzed calculations performed as shown in
Table D-5 [14, 25], The observed-measured values are from many references such as 37-40.
y= 0.9744X-0.0892
_ro
_o
ro
O
6
<
CL
CO
-12
-12
-6
0
6
Observed (M"V1)
Figure D-14. SPARC-calculated versus observed hydrolysis rate
constants for base, acid and general base in six different solvents
and at different temperatures. The aggregate RMS was 0.37.
Table D-5. Statistical Parameters of SPARC Calculated Hydrolysis Rate Constants (M Y1)
Solvent
Base
Acid
Gbase
No RMS R2
No RMS R2
No RMS R2
Water
142
0.39 0.98
383
0.36 0.98
51
0.34 0.98
Acetone/Water
143
0.34 0.83
208
0.33 0.96
73
0.36 0.96
EthanolAVater
105
0.29 0.83
39
0.17 0.98
9
0.1 0.99
MethanolAVater
150
0.36 0.78
22
0.22 0 .95
N/A
Dioxnae/Water
90
0.47 0.75
15
0.16 0.87
17
0.47 0.67
Aceteonitrile/W ater
24
0.3 0.97
N/A
N/A
Total Molecules
654
0.37 0.96
667
0.37 0.97
150
0.39 0.97
D-21
-------
Appendix D
Verification and Validation of the SPARC Model
Validation of the SPARC Electron Affinity (EA) Models
As was the case for pKa, the SPARC computational procedure starts by locating the
potential sites within the molecule at which a particular reaction of interest could occur. In the
case of EA these reaction centers, C, are the smallest subunit(s) that could form a molecular
negative ion. Any molecular structure appended to C is viewed as a "perturber" (P). EA as
expressed in terms of the summation of the contributions of all the components, perturber(s) and
reaction center(s), in the molecule:
EA = Yj[(EA)c+5p(^EA)J
(D-14)
where the summation is over n, which is defined as the number of reaction centers in the
molecule. (EA)C is the electron affinity for the reaction center. 5P(AEA)C is a differential quantity
that describes the change in the electron affinity behavior affected by the perturber structure.
In the estimation of EA, there was no modifications to any of the pKa models or any extra
parameterization for P to calculate electron affinity from ionization pKa models other than inferring
the electronegativity and the electron affinity susceptibility of the reaction centers (C) to
electrostatic and resonance effects [4],
The EA models have been tested to the maximum extent possible on all the gas phase
electron affinity measurements reported by Kebarle, Mclver and Wentworth [4], The RMS
deviation for the 260 EA's was 0.14 e.V. and R2 was 0.98 as shown in Figure D-15. The
statistical parameters are shown in Table D-2.
4 :
3
2
T5
O
1
ra
o
0
uz
u
-1
-2
-3
jT
¦ 1 0 1
Observed
Figure D-15. SPARC-calculated versus observed electron affinity for 260
organic compounds. The RMS deviation was 0.14 e.V. and R2was 0.98.
D-22
-------
Appendix D
Verification and Validation of the SPARC Model
MONOPOLE MODELS (IONIC SPECIES)
The SPARC models were extended to ionic organic species by incorporating monopole
(charge) electrostatic interaction models to SPARC'S physical properties toolbox. These ionic
models play a major role in modeling and estimating Henry's constant for charged (ionic)
species in any solvent system. These capabilities (ionic activity) in turn allow SPARC to
calculate gas phase pKa, and non-aqueous ionization pKa and Ein chemical reduction in any
solvent system.
Validation of the SPARC Monopole Models
The SPARC monopole models have been tested on all the available data for Henry's
constant for charged molecules in water, unfortunately there was only 12 data points. However,
the SPARC Ionization pKa in water coupled with Henry's constant for charged molecules was
used to estimate 400 pKa's in the gas phase and 300 pKa's in non-aqueous solvents. Also,
SPARC electron affinity calculator coupled with Henry's constant for charged molecules was
used to estimate 352 Ei/2 chemical reduction data measurements. See Table D-2 and for more
details see reference 14.
QUALITY ASSURANCE
A quality assurance (QA) plan was developed to recalculate all the aforementioned
physical and chemical properties and compare each calculation to an originally-calculated value
stored in the SPARC databases. Every quarter, two batch files that contain more than 3000
compounds (4200 calculations) recalculate various physical and chemical properties. QA
software compares every single "new" output to the SPARC originally-calculated-value dating
back to 1993-1999. This ensures the integrity of the SPARC model as new features are added.
CONCLUSION
The strength of the SPARC chemical reactivity parameters and physical properties
calculator is the ability to estimate numerous properties for a wide range of organic compounds
within an acceptable error, especially for molecules that are difficult to measure. The SPARC
physical properties/chemical reactivity parameters calculator prediction is as reliable as most of the
experimental measurements for these properties. For simple structures, SPARC can calculate a
property of interest within a factor of 2 or even better. For complex structures where dipole-dipole
and/or H-bond interactions are strong, properties can generally be calculated within a factor of 3-4.
The true validity of the SPARC physical/chemical property models does not lie in the
models' predictive capability for pKa or solubility, but is determined by the extrapolatability of
these same models to other types of chemistry. The ability of SPARC models to be extended to
various chemical/physical properties without modification or extra parameterization to any of the
basic models provides great confidence in this powerful calculation tool.
D-23
-------
Appendix D
Verification and Validation of the SPARC Model
APPENDIX
Summary of usage of the SPARC-web version
Two months back-to-back report, which represents the usage of the SPARC calculator in
October and November 2002. November was the highest while October was the lowest usage to
date.
Summary of Activity for Report
October 2002
November 2002
Hits Entire Site (Successful) 56,875
Average Number of Hits per day on Weekdays
2,153
Average Number of Hits for the entire
Weekend 1,297
Most Active Day of the Week Thu
Least Active Day of the Week Sat
Most Active Day Ever October 24, 2002
Number of Hits on Most Active Day 4,963
Least Active Day Ever October 05, 2002
Number of Hits on Least Active Day 7
URL's of most active users
207.168.147.52 463
pi20x183.tnrcc.state.tx.us 3,986
141.189.251.7 1,720
198.137.21.14 455
57.67.16.50 327
gateway.huntingdon.com 6,823
ari es. chemi e. uni -erl angen. de 1,487
pl20x226.tnrcc.state.tx.us 67
thompson.rtp.epa.gov 413
webcache.crd.GE.COM 143
Hits Entire Site (Successful) 95,447
Average Number of Hits per day on Weekdays
4,146
Average Number of Hits for the entire
Weekend 842
Most Active Day of the Week Wed
Least Active Day of the Week Sun
Most Active Day Ever November 13, 2002
Number of Hits on Most Active Day 15,450
Least Active Day Ever November 02, 2002
Number of Hits on Least Active Day 7
URL's of most active users
141.189.251.7 1,223
gw.bas.roche.com 1,821
gateway.huntingdon.com 3,729
pl20xl83.tnrcc.state.tx.us 737
hwcgate.hc-sc.gc.ca 660
pl20x226.tnrcc.state.tx.us 379
thompson.rtp.epa.gov 563
chen.rice.edu 966
SPARC is online and can be used at http://ibmlc2.chem.uga.edu/sparc
D-24
-------
Appendix D
Verification and Validation of the SPARC Model
REFERENCES
1. S. W. Karickhoff, V. K. McDaniel, C. M. Melton, A. N. Vellino, D. E. Nute, and L. A.
Carreira., US. EPA, Athens, GA, EPA/600/M-89/017.
2. S. W. Karickhoff, V. K. McDaniel, C. M. Melton, A. N. Vellino, D. E. Nute, and L. A.
CarreiraEnviron. Toxicol. Chem. 10 1405 1991.
3. S. H. Hilal, L. A. Carreira and S. W. Karickhoff, "Theoretical and Computational
Chemistry, Quantitative Treatment of Solute/Solvent Interactions", Eds. P. Politzer and
J. S. Murray, Elsevier Publishers, chapter 9, 1994.
4. S. H. Hilal, L. A. Carreira, C. M. Melton and S. W. Karickhoff, Quant. Struct. Act. Relat,
12 389 1993.
5. S. H. Hilal, L. A. Carreira, C. M. Melton, G. L. Baughman and S. W. Karickhoff, J. Phys.
Org. Chem. 7, 122 1994.
6. S. H. Hilal, L. A. Carreira and S. W. Karickhoff, Quant. Struct. Act. Relat. 14 348 1995.
7. S. H. Hilal, L. A. Carreira, S. W. Karickhoff, M. Rizk, Y. El-Shabrawy and N. A.
Zakhari, Talanta, 43 , 607 1996.
8. S. H. Hilal, L. A. Carreira and S. W. Karickhoff, Talanta., 50 827 1999.
9. S. H. Hilal, L. A. Carreira, S. W. Karickhoff, J. Chromatogr., 269 662 1994.
10. S. H. Hilal, L. A. Carreira and S. W. Karickhoff, Accepted, Quant. Struct. Act. Relat.
11. S. H. Hilal, J.M Brewer, L. Lebioda and L.A. Carreira, Biochem. Biophys. Res. Com., 607
211 1995
12. SAB Report, Evaluation of EPA 's Research on Expert Systems to Predict the Fate and
Effects of Chemicals, November 1991.
13. Peer Review of the Research Programs of the Ecosystems Research Division. U.S. EPA,
NERL, Athens, Ga, June 1997.
14. S. H. Hilal, a companion U.S. E.P.A report, "Estimation of Physical Properties and
Chemical Reactivity Parameters from Molecular Structure using SPARC'.
15. J. E. Lemer and E.Grunwald, Rates of Equilibria of Organic Reactions, John Wiley &
Sons, New York, NY, 1965.
16. Thomas H. Lowry and Kathleen S. Richardson, Mechanism and Theory in Organic
Chemistry. 3ed ed., Harper & Row, New York, NY, 1987.
D-25
-------
Appendix D
Verification and Validation of the SPARC Model
17. L. P. Hammett, Physical Organic Chemistry, 2nd ed. McGraw Hill, New York, NY, 1970.
18. R.W. Taft, Progress in Organic Chemistry, Vol.16, John Wiley & Sons, New York, NY,
1987.
19. M. J. S. Dewar, The Molecular Orbital Theory of Organic Chemistry, McGraw Hill, New
York, NY, 1969.
20. M. J. S. Dewar and R. C. Doughetry, The PMO Theory of Organic Chemistry, Plenum Press,
New York, NY, 1975.
21. J. Dykyj, M. Repas and J. Anmd Svoboda., Vapor Pressure of Organic substances.
VEDA, Vydavatel Stvo, Slovenskej Akademie Vied, Bratislava, 1984.
22. S. W. Karickhoff and MacArthur Long, US. EPA Internal Report, April 10 1995.
23. C. R. Willke and C. Y. Lee, Ind. Eng. Chem. 47 1253 1955.
24. P. D. Neufeld, A. R. Janzen and R. A. Aziz, J. Chem. Phys. 57 1100 1972.
25. S. H. Hilal, L. A. Carreira and S. W. Karickhoff, To be Submitted.
26. R. C. Reid, J. M. Prausnitz and J. K. Sherwood, The Properties of Gases and Liquids, 3ed .,
McGraw-Hill Book Co., 1977.
27. R. R. Dreisbach Physical Properties of Chemical Compounds: Advanced in Chemistry
Series, Dow Chemical Co., ACS, Washington, D.C., (A) Volume I, 1955, (B) Volume n,
1959, (C) Volume III, 1961.
28. R. C. Wilhoit and B. J. Zwolinski,./. Phys. Chem. Ref. Data, 2 1, 1973. Supplement No. 1.
29. T. E. Jordan, The Vapor Pressure of Organic Compounds, Interscience Publisher Inc,
Philadelphia, Pennsylvania, 1954.
30. R. Weast and M. Astle, CRC Handbook of Chemistry and Physics, 79th ed., CRC Press Inc.,
West Palm Beach, 1999.
31. D. Mackay, W. Y. Shiu and K.C Ma, Illustrated Handbook ofPhysical/Chemical Properties
and Environmental Fate of Organic Chemicals, Lewis Publishers, volume I, II, III, 1993.
32. Douglas Hartley and Hamish Kidd, The Agro Chemical Handbook, Royal Society of
Chemistry, University of Nottingham, England, 1983.
33. S. R. Heller, D.W. Bigwood, P. Laster and K. Scott, The ARSPesticide Properties Database,
Maryland, U.S.A.
34. H. A. Hornsby, D. R. Wauchope and E. Albert Herner, Pesticide Properties in the
Environment, Springer, New York, NY, 1996.
D-26
-------
Appendix D
Verification and Validation of the SPARC Model
35. E. P. Serjeant and B. Dempsey, Ionization Constants of Organic Acids in Aqueous Solution,
Pergamon Press, Oxford, 1979.
36. (A) D. D. Perrin, Dissociation Constants of Organic Bases in Aqueous Solution,
Butterworth & Co, London, 1965 & Supplement 1972. (B) Supplement 1972.
37. N. B. Chapman, J. Chem. Soc., 1291 1963.
38. L. W. Deady and R. A. Shanks, Aust. J. Chem., 25, 2363 1972.
39. M. L. Bender and Robert J. Thomas, J. Am. Chem. Soc., 83 4189 1961.
40. DeLos DeTar and Carl J. Tenpas, J. Am. Chem. Soc., 7903 1976.
D-27
-------
Appendix D
Verification and Validation of the SPARC Model
this document intentionally left blank.
D-28
-------
Appendix E
Review of Validation Studies Concerning the
U.S. EPA Geochemical Speciation Model
MINTEQA2
-------
This page intentionally left blank.
-------
Appendix E
MINTEQA2 Verification and Validation
Review of Validation Studies Concerning
the U.S. EPA Geochemical Speciation Model
MINTEQA2
Prepared for the
U.S. Environmental Protection Agency
Office of Research and Development
National Exposure Research Laboratory
Ecosystems Research Division
Athens, GA
Prepared by
Allison Geoscience Consultants, Inc.
Flowery Branch, GA
August 30, 2002
E-iii
-------
Appendix E
MINTEQA2 Verification and Validation
This page intentionally left blank.
E-iv
-------
Appendix E
MINTEQA2 Verification and Validation
Table of Contents
Executive Summary v
E.l Introduction and Background 1-5
E.l.l Description of MINTEQA2 1-5
E.l.2 History of MINTEQA2 1-6
E.l.2.1 Code modifications 1-7
E.l.2.2 Database modifications 1-8
E.2 Verification and Validation 2-1
E.2.1 Verification of MINTEQA2 Calculations 2-3
E.2.1.1 Initial comparison with other computer models 2-3
E.2.1.2 Later verification efforts 2-4
E.2.2 Quality of the MINTEQA2 Thermodynamic Database 2-5
E.3 MINTEQA2 Validation Studies 3-1
E.3.1 Validation in Simple Systems (No Sorption, No Doc Complexation) 3-2
E.3.2 Validation in Complicated Systems (With Sorption or Complexation by Doc) 3-11
E.4 Conclusions 4-1
E.5 References 5-1
List of Figures
E-l MINTEQA2 Computes the Equilibrium Distribution of Metal 1-6
E-2 MINTEQA2 calculated pH (line) versus measured pH (diamonds) for initial results
cement leachate 3-3
E-3 MINTEQA2 calculated pH (line) versus measured pH (diamonds) with added impurities
in cement 3-4
E-4 Comparison of Zn2+ determined by ion exchange speciation (circles) and calculated by
MINTEQA2 (line) 3-7
E-5 MINTEQA2 predicted Pb in solution (line) versus experimentally observed
values (squares) 3-13
E-6 MINTEQA2 predicted Zn in solution (line) versus experimentally observed
values (squares) 3-14
E-7 MINTEQA2 predicted Ni in solution (line) versus experimentally observed
values (squares) 3-15
E-8 MINTEQA2 predicted Cd in solution (line) versus experimentally observed
values (squares) 3-16
E-9 MINTEQA2 predicted Cu in solution (line) versus experimentally observed
values (squares) 3-17
E-10 MINTEQA2 predicted Ba in solution (line) versus experimentally observed
values (squares) 3-18
List of Figures (continued)
E-v
-------
Appendix E
MINTEQA2 Verification and Validation
E-l 1 MINTEQA2 predicted Be in solution (line) versus experimentally observed
values (squares) 3-19
E-12 Simulated versus measured pH along the flow path 3-23
E-13 Simulated versus measured Fe concentration along the flow path 3-24
E-l4 Simulated versus measured Mn concentration along the flow path 3-25
E-l5 Simulated versus measured Cu concentration along the flow path 3-26
E-l6 Simulated versus measured Co concentration along the flow path 3-27
E-l7 Simulated versus measured Ni concentration along the flow path 3-28
E-l8 Simulated versus measured Zn concentration along the flow path 3-29
E-l9 MINTEQA2-simulated versus measured Mo04 for column experiments using
aquifer materials and water from sewage-contaminated well (F347-46) and
uncontaminated well (F347-20) 3-32
E-20 Comparison of MINTEQA2 estimated degree of DOC complexation for Cd, Ni, and Zn
versus degree of complexation in resin and batch aquifer material experiments. ... 3-36
E-21 Comparison MINTEQA2 estimated degree of DOC complexation for Cu and Pb
versus degree of complexation in resin experiments 3-38
E-22 Comparison of MINTEQA2 predictions of DOC complexation of Cd, Ni, and Zn
versus experimentally determined values (resin exchange) at various pH values. . . . 3-40
E-23 Percent of U022+ predicted as remaining in solution by MINTEQA2 (line) versus
measured (X) over range of pH 3-42
E-24 Percent of Pb predicted as remaining in solution by MINTEQA2 (line) versus
measured (X) over range of pH 3-43
E-25 Dissolved Cu and Ni concentrations simulated by QWASI model using
MINTEQA2-generated partition coefficients versus measured metal concentrations. 3-46
E-vi
-------
Appendix E
MINTEQA2 Verification and Validation
Executive Summary
This report documents and reviews available reports, papers, and studies relating to the
validation and verification of the U.S. EPA geochemical speciation model, MINTEQA2.
Evidence relating to the verification of MINTEQA2 calculations, including its initial verification
and verification efforts over the course of its history are reviewed. Quality assurance efforts
used during modifications of the MINTEQA2 thermodynamic database are discussed. There
have been hundreds of reported studies in which MINTEQA2 was applied. Only a small fraction
of these qualify as validation studies, and in many of these, validation of MINTEQA2 was not
the central focus of the work. Regardless of the overall focus of the reported study, the key
requirement for considering a particular work as a MINTEQA2 validation study is that results
calculated by MINTEQA2 are compared with some measure of reality. We have attempted to
include all studies that illuminate the question of MINTEQA2 validation, regardless of whether
pro or con. The table below summarizes the validation studies discussed in the text.
Study Citation
Basis of Validation
Metals
Unpublished MINTEQA2
workshop problem
Comparison of a pH curve from a leaching experiment
showing leachate pH versus concentration of acetic acid in
the leachant with similar curve computed by MINTEQA2.
pH
Frandsen and Gammons
(2000)
Comparison of dissolved metal concentrations predicted by
MINTEQA2 with measured values
Zn, Cu, Fe
Marani el al. (1995)
Comparison of equilibrium mineral phases predicted by
MINTEQA2 with sample mineral phases identified by X-ray
diffraction.
Pb
Fotovat and Naidu (1997)
Compares free Cu2+ and Zn2+ in solution as determined using
ion exchange procedures versus computed by MINTEQA2
Cu, Zn
Jensen et al. (1998)
Compares speciation of Fe(II) and Mn(II) in solution as
determined using ion exchange procedures versus computed
by MINTEQA2
Fe(II), Mn(II)
Yu (1996)
Comparison of solid phases predicted to precipitate by
MINTEQA2 versus solid phases identified in field samples
using X-ray diffraction and other analytical methods
Fe, Al
Palmer et al. (1998)
Comparison of Cu2+ activity measured using an ion-selective
electrode with Cu2+ activity computed by MINTEQA2
Cu
Davis et al. (1992)
Comparison of MINTEQA2-calculated metal solubilities with
measured solubilities
As, Pb
Louxetal. (1989)
Comparison of fraction of metal remaining in solution in
batch equilibrium experiments using aquifer materials versus
fraction of metal dissolved at equilibrium as calculated by
MINTEQA2 using diffuse-layer sorption model
Ba, Be, Cd, Cu,
Ni, Pb, Tl, Zn
Jenne (1994)
Solid phases predicted to exist at equilibrium by MINTEQA2
were compared with solid phases identified using analytical
methods
Ca, Fe, Mn, Al, Si
Stollenwerk (1994)
Comparison of dissolved concentrations of metals measured
in a series of wells with solution concentrations predicted
using MINTEQa2 with diffuse-layer model
Al, Fe, Mn, Ca,
Cu, Co, Ni, Zn,
pH, S04
E-vii
-------
Appendix E
MINTEQA2 Verification and Validation
Study Citation
Basis of Validation
Metals
Stollenwerk (1996)
Comparison of dissolved concentrations of metals measured
in a column experiment effluent with dissolved concentrations
predicted using MINTEQA2 with diffuse-layer model
Al, Cu, Co, Ni,
Zn
Stollenwerk (1995)
Comparison of dissolved concentration of molybdate in
column experiment effluent with dissolved concentration
predicted using MINTEQA2 with diffuse-layer model
Mo04
Doyle et al. (1994)
Comparison of dissolved concentrations of As in batch and
column tests with dissolved concentrations predicted by
MINTEQA2
As(V)
Saunders and Toran (1995)
Comparison of dissolved concentrations of metals at
monitoring wells near a disposal pond with dissolved
concentrations predicted by MINTEQA2
Co, Cd, Pb, Sr, U,
andZn
Christensen and
Christensen (1999)
Concentrations of metal-DOC complexes determined in batch
experiments using a resin exchange method were compared
with concentrations of metal-DOC complexes computed by
MINTEQA2
Cd, Ni, Zn
Christensen et al. (1999)
Concentrations of metal-DOC complexes determined in batch
experiments using a resin exchange method were compared
with concentrations of metal-DOC complexes computed by
MINTEQA2
Cu, Pb
Christensen and
Christensen (2000)
Concentrations of metal-DOC complexes determined in batch
experiments using a resin exchange method were compared
with concentrations of metal-DOC complexes computed by
MINTEQA2 over a range of pH values
Cd, Ni, Zn
Khoe and Sinclair (1991)
Comparison of dissolved metal concentrations predicted by
MINTEQA2 versus concentrations measured in neutralization
experiments
Al, Fe, Ca, Mn,
Si02, P04, Pb, U
Webster and Webster
(1995)
Comparison of dissolved As concentrations measured in batch
experiments with dissolved concentrations predicted by
MINTEQA2 using the diffuse-layer model
As(III), As(V)
Woodfine et al. (2000)
Comparison of QWASI simulation results of average lake
water dissolved metal concentrations with observed values
when MINTEQA2-predicted partition coefficients are used
Cu, Ni
Routh and Ikramuddin
(1996)
Comparison of MINTEQA2-predicted equilibrium solid
phases with solid phases observed by X-ray diffraction and
comparison of predicted water concentrations with observed
values
Pb, Zn
E-viii
-------
Appendix E
MINTEQA2 Verification and Validation
E.l Introduction and Background
The purpose of this report is to review and document available reports, papers, and
studies relating to the validation and verification of the U.S. EPA geochemical speciation model,
MINTEQA2. MINTEQA2 has been used to estimate metal partition coefficients in several
rulemaking activities by the U.S. EPA Office of Solid Waste. Given the importance of the
partition coefficient in determining the outcome of fate and transport modeling for metals, it is
anticipated that those responsible for reviewing these rules and those to be regulated by them
will be interested in the degree to which MINTEQA2 has or has not been validated. This report
reviews studies that relate to validation of MINTEQA2 to provide an assessment of its validation
status. Prior to the presentation of the validation studies, the efforts that have been made to
verify MINTEQA2 calculations are documented.
E.l.l Description of MINTEQA2
MINTEQA2 is an equilibrium geochemical speciation model maintained and distributed
by the U.S. EPA. From input data consisting of total concentrations of chemical constituents,
MINTEQA2 calculates the fraction of a contaminant metal that is dissolved, adsorbed, and
precipitated at equilibrium (see Figure 1). As input data, the total concentrations of major and
minor ions, trace metals and other chemicals are specified in terms of key species known as
components. MINTEQA2 automatically includes an extensive database of solution species and
solid phase species representing reaction products of two or more of these input components.
The model does not automatically include sorption reactions, but these can be included in the
calculations if supplied by the user. When sorption reactions are included, the dimensionless
partition coefficient can be calculated from the ratio of the sorbed metal concentration to the
dissolved metal concentration at equilibrium. The dimensionless partition coefficient is
converted to Kd with units of liters per kilogram (L/kg) by normalizing by the mass of soil (in
kilograms) with which one liter of solution is equilibrated (the phase ratio). An isotherm is
generated when the equilibrium metal distribution between sorbed and dissolved fractions is
estimated for a series of total metal concentrations.
Progress in accounting for sorption in equilibrium calculations over the past decade has
resulted in the development of coherent databases of sorption reactions for particular sorbents.
These databases include acid-base sorption reactions and reactions for major ions in aquatic
systems (Ca, Mg, S04, etc.). Including such reactions along with those representing sorption of
trace metals makes it possible to estimate sorption in systems of varying pH and composition.
Examples of coherent databases of sorption reactions include that for the hydrous ferric oxide
surface presented by Dzombak and Morel (1990) and a similar database for goethite presented
by Mathur (1995).
E-l
-------
Appendix E
MINTEQA2 Verification and Validation
Dissolved
Total
Component
Concentrations
Adsorbed
Precipitated
Figure E-1. MINTEQA2 Computes the Equilibrium
Distribution of Metal
E.1.2 History of MINTEQA2
The original version of this model (called MINTEQ) was developed in the early 1980's at
Battelle Pacific Northwest Laboratory by Felmy and coworkers (Felmy el al., 1984) as a
cooperative effort funded by the U.S. Department of Energy and the U.S. EPA. The MINTEQ
package was delivered to the U.S. EPA Environmental Research Laboratory at Athens, Georgia
(AERL) in 1985.1 The model was renamed MINTEQA1 to designate this Athens version which
was anticipated to diverge from the original as it was adapted to the special needs of the U.S.
EPA. The model was first distributed with this name from the Athens lab in 1986, but there
were very few differences between this MINTEQA1 and the original MINTEQ. The distribution
package, available for DOS-based PC's or for Digital Equipment Corporation VAX machines,
included a preprocessor program PRODEFA1 for the preparation of MINTEQA1 input files.
After more significant revisions were made in the late 1980's, the name was changed to
MINTEQA2. With further development, version numbers were used to indicate new versions,
and the model's formal name was left as MINTEQA2.
E.l.2.1 Code Modifications
The original MINTEQ developed by Felmy and coworkers was produced by combining
the mathematical structure of MINEQL (Westall el al., 1976), with the thermodynamic database
of the WATEQ3 model developed by the U.S. Geological Survey (Ball et al., 1981). The
mathematical formulation which MINEQL embodies was also used in MINTEQ. Many of the
Fortran subroutines in MINEQL were used directly or with little modification in MINTEQ,
which was also written in Fortran. After MINTEQ was delivered to the U.S. EPA, it underwent
continuous testing and development by an on-site computer specialist and geologist staff person
assigned for that purpose through 1992. Development slowed after 1992, with further
modifications occurring in response to specific U.S. EPA needs. Important revisions to the code
since its initial delivery in 1986 include:
1 This lab is now known as the National Exposure Research Laboratory, Ecosystems Research Division.
E-2
-------
Appendix E
MINTEQA2 Verification and Validation
¦ Addition of the diffuse-layer adsorption model (also known as the MIT two-layer
model) with a database for sorption onto hydrous ferric oxide (HFO) as compiled by
David Dzombak (Dzombak, 1986; later modified in Dzombak and Morel, 1990)
¦ Addition of the Gaussian distribution model for dissolved organic matter (DOM) with
an associated database of reactions (Dobbs et al., 1989; Susyeto et al., 1991; Allison,
1997)
¦ Alteration of the code to include sparse-matrix techniques to increase execution speed
(Allison, 1997)
¦ Addition of an automated procedure for producing successive model executions as
one parameter (e.g., pH, pe, total concentration of a species) is varied
¦ Addition of a procedure to write key equilibrium results to a file for input to a
spreadsheet program and subsequent plotting
¦ Provision to display the source of thermodynamic data for each species listed in the
output file. The source of the thermodynamic data was added to the database as
described in the section on database quality
¦ Continual improvements in convergence methods and correction of minor errors
encountered and reported by users worldwide
¦ Continual improvement in the PRODEFA1 (later, PRODEFA2) preprocessor
program.
The version of the model distributed publicly by the U.S. EPA Center for Exposure Assessment
Modeling (CEAM) was periodically updated as these improvements were implemented. The
succession of model names and versions numbers was: MINTEQA1 (no version numbers);
MINTEQA2, versions 2.0, 2.01, 3.0, 3.10, 3.11, 4.0, 4.01, 4.02 (current version; released 1999).
Since the delivery of MINTEQ by Battelle, the U.S. EPA has maintained and distributed
the model through the Center for Exposure Assessment Modeling (CEAM) at the EPA
Environmental Research Laboratory (Athens). CEAM has provided an orderly procedure for
maintaining and distributing MINTEQA2 and other computer codes. Users who obtain
MINTEQ A2 through CEAM are provided with a contact to report errors or suspected errors in
the model or its results. Also, prior to releasing any new version of MINTEQ A2, the revised
code is subjected to verification tests by CEAM to ensure that the modifications work as
intended and produce the desired results. These tests include compiling on multiple compilers
and test executions of a battery of equilibrium problems with known solutions. The tests were
designed to exercise all important algorithms in the code. Each successive version of the model
has been required to pass these tests before release.
E-3
-------
Appendix E
MINTEQA2 Verification and Validation
E.l.2.2 Database Modifications
The thermodynamic database contained in WATEQ3 was selected for use in the original
MINTEQ because it was felt to be among the best such databases available in the 1980's. The log
equilibrium constants for the reactions it contained had been tested for consistency (Nordstrom
and Munoz, 1986). The main shortcoming of its implementation in MINTEQ was that the
source of the thermodynamic constants (log equilibrium constants and enthalpies of reaction)
were not documented. The MINTEQA2 thermodynamic database has been updated occasionally
throughout the model's history to add new component species or to update the thermodynamic
constants of existing reaction species. MINTEQ thermodynamic data for species involving the
trace metals chromium (+2, +3, and +6 oxidation states), mercury (+1 and +2 oxidation states) ,
selenium (+4 and +6 oxidation states), and thallium (+1 and +3 oxidation states) were updated in
an EPA-funded project shortly after the model was developed (Deutsch and Krupka, 1985).
Components and reaction products with accompanying thermodynamic data were added in 1989
for antimony (+3 and +5 oxidation states) and cyanide (Sehmel, 1989). The thermodynamic
database released with version 2.0 in 1988 included the addition of 22 organic ligands and
several hundred reaction products between those ligands and trace metals. The metal-organic
complexes were added through an EPA-funded research project at Colorado School of Mines in
which carboxylic and dicarboxylic acids and other organic substances associated with landfill
leachate were identified. Nine more organic ligands and associated reactions were added in
version 2.01 (1989). An article appearing in Water Research in 1996 pointed out several errors
in the formulation of these metal-organic complexes as entered in the MINTEQ A2 database
(Serkiz et al., 1996). Those database entries were reviewed and corrected and a revised database
was prepared in 1996 (Allison, 1996). The database was modified again in 1999 for version 4.0
to add species for beryllium (Be), cobalt (Co), molybdenum (Mo), and tin (Sn); to update the
equilibrium constants for other inorganic species; and to add the source of the reaction
thermodynamic constants (log K and enthalpy of reaction) for all species for which these could
be determined (USEPA, 1998). These last two modifications to the thermodynamic database
utilized the electronic databases CRITICAL (NIST, 1997) and SC-D AT ABASE (IUPAC, 1998)
as primary data sources. The former represents the continued development of the compilation of
critically reviewed metal stability constants by Smith and Martell published in book format in
the 1970's and 80's. The sources of thermodynamic data are important in considering the validity
of the modeling results; these sources are discussed more fully in Section 2.2 below.
E-4
-------
Appendix E
MINTEQA2 Verification and Validation
E.2 Verification and Validation
In the background material above, custom has been followed in that MINTEQA2 has
been referred to as a "model." In a sense, this is incorrect. MINTEQA2 should rather be
thought of as a convenient framework that within which a particular geochemical model may be
implemented. That MINTEQA2 includes its own database of thermodynamic reactions may
obscure the fact that a geochemical model does not exist until the user specifies certain
MINTEQA2 input parameters that define the system. Defining the system includes specifying
components to represent total concentrations of major and trace ions and imposing equilibrium
conditions such a pH, Eh, solid phases present at equilibrium, equilibrium gas partial pressures,
etc. The point that MINTEQA2 is a modeling framework rather than a model is important when
one considers its validation status. To help clarify this point, the term partial validation has
sometimes been used to indicate the validation of a particular geochemical model posed within
the MINTEQA2 framework or that of another speciation "model" (Krupka et al, 1983; Jenne
andKrupka, 1985; Zachara et al., 1987; Jenne, 1994).
In every application of a speciation model, it is incumbent upon the user to apply
geochemical wisdom in reviewing the completeness and quality of the thermodynamic database
and to make necessary additions or changes. The completion of the database review step is
another requirement necessary to develop a true geochemical model within the MINTEQA2
modeling framework. This requirement is especially important in selection of solid phases that
are to be allowed to precipitate. For example, a user designing a model to represent a laboratory
system must consider precipitation kinetics (or dissolution kinetics) in deciding which phases
MINTEQA2 should consider. A particular solid may indeed be the equilibrium phase if the
system is to be allowed years to equilibrate, but not if the equilibration period is one day.
Sorption modeling imposes another critical chore of database management on the user. Sorption
reactions and their associated equilibrium constants are specific to the particular sorbent type:
aluminum oxide, amorphous hydrous ferric oxide, various crystalline oxides, etc. One cannot
directly use reactions describing trace metal sorption onto hydrous ferric oxide to represent
reactions onto an aluminum oxide surface. Interchangeability of reactions is even less likely for
clay surfaces and more complicated natural sorbents such as organic matter. Another
complicating factor in including sorption reactions is that the available reactions are specific to
the sorption model within which they were derived. MINTEQA2 includes seven sorption
models. A uniform database containing acid-base, major ion, and trace metal reactions is
included for only one sorbent: hydrous ferric oxide (Dzombak and Morel, 1990). This database
is designed for use in the diffuse-layer sorption model and cannot be used directly for any other.
Thus, the geochemical model actually consists of the combination of the user's input parameters
that describe and constrain the system, the user-approved or user-modified thermodynamic
database, and the computer code that implements equilibrium in the system.
E-5
-------
Appendix E
MINTEQA2 Verification and Validation
Verification refers to tests and studies that, by design or otherwise, show that the
computations performed by the computer code representing a model are true to the intent of the
conceptual model. Verification tests determine whether the computer code, compiled into an
executable program, arrives at the intended and expected answer for a given set of input values.
Here, "input values" include total concentrations of components, the set of equilibrium reactions
and their thermodynamic constants, perhaps one or more imposed equilibrium conditions, and
the settings of various program flags and options such as ionic strength, method of estimating
activity coefficients, and system temperature. The "answer" obtained from the computer code
consists of all computed quantities including the equilibrium concentrations of all solution
species and the amounts of solid phases that have dissolved from an initially present solid or
precipitated from the solution. The computed answer also may include the ionic strength and
activity coefficients of solution species. Verification can be achieved for individual algorithms
that make up the entire computer program. For MINTEQA2, it is impossible to test all program
options and features in one program execution. The basis of judging a computer code as verified
rests on comparison of the computed answer with a standard. The standard may be the result of
a hand calculation or the result from a similar computer code that has itself been verified In
either case, it is necessary to use the same reactions and thermodynamic data when calculating
the results from the code to be verified as was used in calculating the results for the standard of
comparison.
Validation refers to tests and studies that show that the geochemical model that is
implemented by the combination of the user's input parameters, the thermodynamic database,
and the computer code provides an acceptable representation of reality or that it produces an
outcome that is an acceptable representation of reality. This definition presupposes that there
exists a measurement or group of measurements that may be taken as reality and that can be used
as the standard to which the model result is compared. For geochemical models, validation is
significantly more complicated and uncertain than verification.
There are difficulties in validating geochemical models regardless of whether the model
outcome is compared with measurements on natural field systems or lab systems that mimic the
natural environment. Natural systems are replete with complicating factors that result in
imprecise or uncertain measurements and conditions that fail to correspond to the primary tenet
of MINTEQA2-based geochemical models: that the system reflects equilibrium conditions.
Problems and issues in measuring (analytical methods, sample handling, determining redox
status), problems in incomplete knowledge of the natural environment (true nature of sorption
reactions, partial pressures of gases, rates of reaction, degree of mediation by biota), and the
high degree of variability in important chemical characteristics of natural systems all serve to
complicate comparisons of model systems with their real counterparts. In consideration of the
challenges of validating geochemical models, the U.S. EPA convened a meeting of geochemists,
soil scientists, and other groundwater professionals at the Athens, Georgia, Environmental
Research Lab in 1989. Opinions were varied among those present as to what might constitute a
validation of the model (Dr. Dave Brown, pers. comm., 2002). Those professionals who
specialized in laboratory analysis felt that comparisons of MINTEQA2 predictions with
measurements made on closely controlled laboratory systems would provide the most relevant
validations. Those who were more field-oriented felt that laboratory systems could not faithfully
represent the real systems that are of interest; they preferred a validation exercise closely tied to
field sampling of an appropriate system. The statisticians pointed out that the natural variability
E-6
-------
Appendix E
MINTEQA2 Verification and Validation
in many important MINTEQA2 input parameters would required that the model be validated at a
host of diverse sites. All of these viewpoints have merit. This diversity of opinions on how to
validate MINTEQA2 is a reflection of the difficulty of the undertaking. However, in the
fourteen years since that workshop, numerous and varied studies have been performed that,
directly or indirectly, relate to the validation of MINTEQA2. These are discussed individually
in Section 3.0.
E.2.1 Verification of MINTEQA2 Calculations
This section documents evidence relating to the verification of MINTEQA2. As
mentioned above, it is not possible to verify all parts of the MINTEQA2 computer code in a
single model execution. In part, this is due to the mutually exclusive nature of some program
options. All parts of the MINTEQA2 code have been verified. Those sections that existed at the
beginning of the model's lifetime were verified by comparison of results with those of similar
(verified) models. As a quality assurance measure, CEAM policy has required that all code
modifications and additions to MINTEQA2 be tested by a combination of compiler tests and
model execution tests before final adoption. The compiler tests required that the model be
compiled using Fortran compilers from multiple vendors and that the effect of various compiler
options on execution time and computed results be examined and accounted for. The execution
tests for MINTEQA2 consisted of a series of equilibrium problems for which the answer was
known or could be computed by hand calculations. This quality assurance requirement is a
primary basis that supports the assertion that MINTEQA2 calculations have been verified.
E.2.1.1 Initial Comparison with Other Computer Models
In a 1988 report discussing the feasibility of validating MINTEQ, Zachara and coworkers
at the Battelle PNL where MINTEQ was developed stated that all major code algorithms
including calculation of mass balance, activity coefficients, and equilibrium speciation were
verified by comparison with hand calculations during the model's initial development (Zachara
et al., 1988). Speciation results from test executions, some involving adsorption reactions, also
were found to agree with identical test calculations using WATEQ3 and MINEQL. In a study
comparing the results of several equilibrium speciation models including MINTEQ, Morrey et
al. (1985) verified that results from these models are the same when identical thermodynamic
data and program options are used.
E.2.1.2 Later Verification Efforts
As mentioned above, after its delivery to the U.S. EPA, MINTEQ was developed further
(under the names MINTEQA1 and finally MINTEQA2). Development of the code was
performed under the auspices of a centralized modeling group, the Center for Exposure
Assessment Modeling (CEAM) at the US. EPA lab at Athens, Georgia. The CEAM group
administered the development, public distribution, and user-assistance for MINTEQA2 and other
useful water quality models. CEAM established guidelines that required modifications to model
codes be tested by comparison of before and after results and by compiling the modified code
using multiple compilers. A standard battery of test problems was used in these comparison
tests.
E-7
-------
Appendix E
MINTEQA2 Verification and Validation
Additional verification tests were performed when new algorithms were added to give
MINTEQA2 capabilities. For example, the diffuse-layer (Generalized Two-Layer) sorption
model was added in 1989. This sorption model had previously been added to MICROQL
(Westall, 1979) and MINEQL by David Dzombak (Dzombak, 1986). Test problems were
presented in Dzombak (1987) along with their computer solutions (using MICROQL) and hand
calculated solutions. These formed the basis of comparison to verify the correct implementation
of the diffuse-layer model in MINTEQA2. In similar manner, the Gaussian distribution model
for computing the complexation of metals with organic matter was verified by hand calculation
when it was added to MINTEQA2 version 3.11 and further developed in version 4.0 (Allison,
1997). The correct implementation of the Gaussian DOM model was also verified by a
procedure described in (Fish etal., 1986) in which the Gaussian distribution of ligand
concentration and log K is approximated by a small set of ligands whose concentrations and
reaction log K values are scaled to conform to the Gaussian distribution. This method can be
implemented in MINTEQA2 without recourse to the formally introduced Gaussian model
option. Results from the formally implemented Gaussian DOM model were found to agreed
with the "manually implemented" Gaussian results, excepting a margin of error in the latter
inversely proportional to the number of ligands used to implement it (Allison and Perdue, 1994).
The organized public distribution of MINTEQA2 under CEAM provided a useful
clearinghouse chore for reporting suspected errors in the code or in the thermodynamic database.
Especially during the early years of its distribution, many errors (especially in the pre-processor
PRODEFA2) were discovered and corrected. The confidence that can be placed in MINTEQA2
has been enhanced by its use by the public and their feedback in reporting errors. Modifications
made to correct errors as well as to enhanced the model were alike subject to verification tests
prior to release in new versions.
Finally, verification of the MINTEQA2 code has been assured through the use of the
model in solving elementary speciation problems for use in teaching geochemistry. Several
universities have used MINTEQA2 in their courses. Dr. Willard Lindsay has discussed the use
of MINTEQA2 in teaching soil chemistry (Lindsay and Ajwa, 1995). Classroom problems
typically are made amenable to hand calculations so that the student may better appreciate the
nuances of how the problem is solved. Dr. James Drever also used MINTEQA2 to illustrate the
solution of problems in groundwater chemistry in his book The Geochemistry of Natural Waters
(Drever, 1997). Simple problems with answers easily calculated by hand were also used in
MINTEQA2 workshops sponsored by the U.S. EPA in the late 1980's and early 90's. Typically,
a simple system involving CaC03 in water is solved in a series of successively more complicated
problems (solution equilibria only, solution equilibria with constrained pH, with constrained C02
partial pressure, with solid phases). Classroom exercises illustrating correct answers using redox
and sorption reactions were also performed. Wrong answers for such exercises would quickly
become apparent as the results are used to explain the chemistry in intuitive terms in the
classroom. Also, before MINTEQA2 was used in the first such workshop, its answers were
compared with those of MINEQL for the same set of ten simple problems and were found to
agree.
E-8
-------
Appendix E
MINTEQA2 Verification and Validation
E.2.2 Quality of the MINTEQA2 Thermodynamic Database
As discussed in Section 1.2.2, the thermodynamic database used by MINTEQA2 has
been revised repeatedly to correct and update thermodynamic constants and to add new species.
Any assessment of the validity of MINTEQA2 applications must consider the quality of the
thermodynamic database that it uses. The following paragraphs describe the data sources that
have been used to update and add to the MINTEQA2 database in recent years. Steps taken to
reduce unintended errors during the process of updating the database are also described. For the
1996 and 1999 modifications, several recognized compilations of thermodynamic constants were
used as data sources. Data for reactions were also obtained from journal articles. The sources
were accorded priority according to their order in the list below. If data were not found, the next
source in the list was consulted, etc.
1. Critical Stability Constants of Metal Complexes Database (CRITICAL) published by
the National Institute of Standards and Technology (NIST Standard Reference
Database 46). Multiple versions of this database were used beginning with NIST
version 2.0 (released in late 1995), and ending with NIST version 4.0 (released in late
1997). The correction and update of the MINTEQA2 v3.11 metal-organic reactions
was completed first (see database history in Section 1.2.2). This update employed
version 2.0 of the NIST CRITICAL database. The update of the general inorganic
species began while version 3.0 of CRITICAL was the current NIST product.
Version 4.0 of CRITICAL was released during the course of updating the inorganic
species and was used to finish the project. In the final MINTEQA2 version 4.0
database, a data source reference code was added for each species. Those updated
using the CRITICAL database were indicated with the code "NIST46.2,"
"NIST46.3," or "NIST46.4" for CRITICAL versions 2.0, 3.0, or 4.0).
2. Stability Constants Database (SC-DATAB ASE) published by the International Union
of Pure and Applied Chemistry (IUPAC) and Academic Press. Two different
versions of this database were used. In the correction of thermodynamic constants for
metal-organic complexes in 1996, version 2.62 (released in 1996) was used. In the
database review for inorganic species in 1999, version 3.02 (released in early 1998)
was used. In the final MINTEQA2 version 4.0 database, the source reference code
for each species updated with SC-DATAB ASE indicates the version from which the
data were obtained ("SCD2.62" and "SCD3.02," respectively for versions 2.62 and
3.02). The reference cited in the actual MINTEQA2 database indicates the journal
article reference within SC-DATAB ASE for version 3.02 citations.
3. Nordstrom et al. (1990) presented data intended to update and document data
appearing in the U.S. Geological Survey equilibrium model WATEQ. Many of the
reactions updated in that model also appear in the MINTEQA2 database, so those
updates were incorporated. Data from this source are indicated in the final
MINTEQA2 version 4.0 database with the source reference code "Nord90."
4. Relevant data from journal articles and other compilations. Use of data from journal
articles is indicated by a code with the year followed by the first two author's initials
(surnames) and a suffixed letter to insure uniqueness (e.g., 1993 DKa). The complete
E-9
-------
Appendix E
MINTEQA2 Verification and Validation
reference is given in a bibliographic entry in a user manual supplement
accompanying MINTEQA2 version 4.0 and later versions.
5. Gibbs free energy of formation (A/7") and enthalpy of formation (AfH") values from
four different sources were used to compute the Gibbs free energy of reaction and
enthalpy of reaction. The former was used to compute the log K for the reaction. The
four data sources (in the preferred priority) were: 1) COD ATA Key Values for
Thermodynamics published by the Committee on Data for Science and Technology
(COD AT A) in 1989; 2) NIST JANAF Thermochemical Tables 1985, Standard
Reference Database 13, version 1.0. Released in electronic format in 1993, the last
update of the actual data in this database was 1985; 3) The NIST Chemical
Thermodynamics Database, Standard Reference Database 2, version 1.1. This is the
electronic form of the older National Bureau of Standards thermodynamic database.
Version 1.1 was released in 1992, but the latest revisions to the data are from 1989;
and 4) Standard Potentials in Aqueous Solution (Bard etal.. 1985). Data obtained
from these four sources are denoted by the codes "CODATA89," "NIST13.1,"
"NIST2.1.1," and "Bard85," respectively, in the source indication in the final
MINTEQA2 database.
Apart from the accuracy of information recorded and presented in CRITICAL, SC-
DATABASE, and other data sources, the process of querying a data compilation, recording the
retrieved information, reducing the data in some fashion, entering the result in the MINTEQA2
database requires attention to details and repeated checking for accuracy to insure a final product
that is free of secondary errors. The following steps were taken to minimize errors in the final
MINTEQA2 thermodynamic database:
¦ Data obtained from the CRITICAL and SC-D AT ABASE compilations of log K and
enthalpy were recorded on data entry sheets with the exact reaction as given in the
source and the pertinent ionic strength and temperature. The information recorded on
each data entry sheet was double-checked, then entered in a data storage and
manipulation program, MINCHEK, via on-screen prompts having the same format as
the data entry sheet. After entering, the displayed data was compared against the data
entry sheet for accuracy.
¦ The A/}" and AtH" values from the JANAF and the NIST Reference Database 2
electronic databases were read directly into the MINCHECK program without
intermediate transcription.
¦ After all data was collected and entered, the data reduction module in MINCHEK
was used to correct the log K values for ionic strength, temperature corrections were
computed and applied, and the data were reformulated such that all reactions were
expressed in terms of MINTEQA2 components. The latter step was accomplished by
adding reactions and their log K and AHR values as required. All data reduction steps
were computed and applied internally by MINCHEK. Gram-formula weights and
species charge were computed automatically by MINCHEK for each species from the
stoichiometry, gram-formula weight (or atomic weight), and charge of each reactant
(component).
E-10
-------
Appendix E
MINTEQA2 Verification and Validation
¦ After log K and AHR values were selected in MINCHEK for each species, the new
MINTEQA2 database was automatically written in the required format, including
reference citations for the data source.
¦ A table showing "old" and updated values of log K and AHR side-by-side for each
species was examined to find instances of large disparity. These were individually
examined to be sure the updated data were correct.
In spite of the care taken to ensure the quality of the thermodynamic database, it must be
recognized that it is still the responsibility of the MINTEQA2 user to review and edit the log K
and enthalpy values for all important reactions for the system of interest.
E-ll
-------
Appendix E
MINTEQA2 Verification and Validation
This page intentionally left blank.
E-12
-------
Appendix E
MINTEQA2 Verification and Validation
E.3 MINTEQA2 Validation Studies
The consensus at the workshop on feasibility of MINTEQA2 validation was that a single
study could not be regarded as validating MINTEQA2 regardless of its outcome. The natural
environment is too varied and complex and the sources of uncertainty and error are too great to
allow much confidence in the results of one study. Rather, it was felt that the most convincing
validation would be the accumulated weight of many varied studies that each tend to provide
some measure of confidence in the model's predictions. This section presents studies that show
instances where MINTEQA2 model predictions have (and have not) been borne out by
measurements of the corresponding real systems. These validation studies have been assembled
from those known personally to the author and from other studies discovered in a literature
search. The literature search employed the keyword "MINTEQA2" as the search term. There
have been hundreds of reported studies in which MINTEQA2 was applied. Only a small fraction
of these qualify as validation studies, and in many of these, validation of MINTEQA2 was not
the central focus of the work. Many of the studies cited below were not undertaken specifically
to provide validation support for MINTEQA2. Regardless of the overall focus of the reported
study, the key requirement for considering a particular work as a MINTEQA2 validation study is
that results calculated by MINTEQA2 are compared with some measure of reality.
We have attempted to include all studies that illuminate the question of MINTEQA2
validation, regardless of whether pro or con. However, researchers are more apt to report their
successes than their failures. Although there is little reason for most researchers to personalize
the failure of a MINTEQA2 application, researchers who find that the model result does not
agree with their experimental work may be reluctant to mention this unless they are confident
that they have used the model properly. This is especially true if the comparison is not the raison
d'etre of their work. These considerations may result in more reported studies that document the
success of MINTEQA2 application than failure.
It is convenient to group the validation studies into those that involve simple solution
chemistry or simple solid phases and those that involve sorption and/or natural organic matter.
In the presentation below of published model validation studies, information identifying the
study is followed by a brief statement of the standard of comparison or basis of the validation
that entitles the study to be regarded as bearing on validation of MINTEQA2. A list or relevant
trace metals involved in the validation and a brief description of the work are also included.
Graphical results that show the match between MINTEQA2 calculations and the standard of
comparison are presented where practical. Concerning graphical results, the plots presented
were scanned from the originals because it was impossible to obtain the original data. Thus, all
plots should be regarded as the work of the authors listed with the study.
E-13
-------
Appendix E
MINTEQA2 Verification and Validation
E.3.1 Validation in Simple Systems (No Sorption, No Doc Complexation)
Title: unpublished problem submitted for MINTEQA2 workshop
Date: 1990
Metals: pH only
Basis of validation:
Comparison of a pH curve from a leaching experiment showing leachate pH versus
concentration of acetic acid in the leachant with similar curve computed by MINTEQA2.
Description:
MINTEQA2 was used to model the behavior of a cement solidification medium for hazardous
wastes under the conditions of an acid neutralization test. This study problem was submitted for
discussion at a U.S. EPA sponsored MINTEQA2 workshop in 1990. The cement used for
solidification of hazardous waste was a dolomitic lime consisting of Ca(OH)2(s, portlandite) and
MgO(s, brucite) in 1:1 Ca to Mg molar ratio. One gram of solid was leached with 20 grams of
aqueous acid solution. The acid solution was acetic acid in water. The experiment was repeated
with successively more concentrated solutions; the acid concentration varied from 0 to 40
equivalents of acid per kg solid. MINTEQA2 was used to predict the result of the leaching
experiment. The comparison of the MINTEQA2-generated equilibrium pH curve with the
experimental data is shown in Figure 2. As seen in the figure, MINTEQA2 predicts more a bit
more buffering that was observed in the experiment. Discussion with those who performed the
experiment revealed that the cement mixture was commercial construction grade material
expected to include some inert impurities. Figure 3 shows the MINTEQA2 pH response curve
when a reasonable value of 5.7 weight% inert solids in the cement was assumed. The
MINTEQA2 pH response is almost identical with the observed response. The reader will
observe that this result tends to validate MINTEQA2, but only for the dolomitic lime system
under discussion.
14
Portlandite depleted
x
Q.
E 10
12
/\
Brucite depleted
O
'3 8
O"
LU
6
Initial
, results
10
20
Added Acid (eq/L)
30
40
Figure E-2. MINTEQA2 calculated pH (line) versus measured pH (diamonds)
for initial results cement leachate.
E-14
-------
Appendix E
MINTEQA2 Verification and Validation
Portlandite depleted
Brucite depleted
x
Q.
E
3
_Q
Wth 5.7%(wt)
impurities
0 10 20 30 40
Added Acid (eq/L)
Figure E-3. MINTEQA2 calculated pH (line) versus measured pH (diamonds)
with added impurities in cement.
Title: Heavy Metal Complexation with Aqueous Sulfide in an Anaerobic Treatment
Wetland
Authors: Frandsen, A. K., and C. H. Gammons
Source: Wetlands and Remediation: An International Conference, Battelle Press, Columbus,
OH, pp. 423-430.
Date: 2000
Metals: Zn, Cu, Fe
Basis of validation:
Comparison of dissolved total soluble metal concentrations predicted by MINTEQA2 with
measured total metal concentrations
Description:
This study compared the dissolved concentrations of Zn, Cu, and Fe measured in filtered water
samples collected from an anaerobic treatment wetland with corresponding soluble
concentrations predicted by MINTEQA2. Experimental results gave higher dissolved metal
concentrations than predicted, a circumstance which the authors attributed to the absence of
quality solution metal-sulfide complexes in the MINTEQA2 database and the consequent
prediction of precipitation of metal sulfides.
Title: Lead Precipitation in the Presence of Sulphate and Carbonate: Testing of
Thermodynamic Predictions
Authors: Marani, D., G. Macchi, and M. Pagano
Source: Water Research, 29(4): 1085-1092
Date: 1995
Metals: Pb
E-15
-------
Appendix E
MINTEQA2 Verification and Validation
Basis of validation:
Comparison of equilibrium solid mineral phases predicted by MINTEQA2 with solid mineral
phases identified by X-ray powder diffraction (XRD).
Description:
The stated objective of this study was to evaluate the capability of MINTEQA2 for predicting
realistic residual lead concentrations in aqueous solutions following an alkaline wastewater
treatment procedure. Test solutions with pH range 4 to 11 were prepared with 10 mM sulfate
concentration and variable carbonate concentration to simulate the alkaline treatment of battery
acid wastewaters. Solutions were aged for several months, then the Pb concentration in solution
was measured and precipitates were identified using XRD. Solution lead concentrations
predicted by MINTEQA2 were in reasonable agreement with measured values for pH < 7. In the
alkaline range, the MINTEQA2-predicted concentrations were significantly less than observed (a
discrepancy of two to three orders of magnitude, depending on pH). The equilibrium solid phase
predicted by MINTEQA2 in the alkaline range was Pb(OH)2(s). However, XRD analyses
showed that the actual precipitates were anglesite (PbS04), cerrusite (PbC03), and hydrocerrusite
(Pb3(C03)2(0H)2). It was found that if Pb(OH)2(s) is prohibited from precipitating in the
MINTEQA2 model runs, one or more of anglesite, cerrusite, and hydrocerrusite precipitate
depending on pH and carbonate concentration and the Pb in solution is in keeping with
observations. The authors reasoned that Pb(OH)2(s) is kinetically limited and apparently forms
upon extensive aging rather than as a direct precipitate.
Title: Ion Exchange Resin and MINTEQA2 Speciation of Zn and Cu in Alkaline Sodic and
Acid Soil Extracts
Authors: Fotovat, A. and R. Naidu
Source: Australian Journal of Soil Research, 35: 711 -726
Date: 1997
Metals: Cu, Zn
Basis of validation:
Compares speciation of Cu and Zn in solution as determined from ion exchange procedures with
speciation computed using MINTEQA2
Description:
Analytical speciation results obtained using the cation exchange resin Amberlite were
determined for copper (Cu) and zinc (Zn) in soil water for eleven soils of varied chemical
composition. The soil pH values ranged from 5.3 to 9.1. Batch experiments were conducted at
various pH and major ion concentration levels to study the procedural steps to accurately
measure free Zn2+ and Cu2+ at various total Zn and Cu concentrations. After adding stability
constants obtained from the literature (Stevenson and Fitch, 1986) for metal-DOC complexes,
the metal and ligand species in the soil extracts were calculated using MINTEQA2. The average
of the absolute differences between free Zn and Cu concentrations obtained using the exchange
resin and calculated using MINTEQA2 was 4.3 percent. The comparison of Zn2+ from the two
methods is shown in Figure 4. The authors observed that Zn was often present as free hydrated
Zn2+ in these soil extracts, with the proportion dependent primarily on pH. In contrast, Cu
occurred primarily in complexed forms in all soils. For both metals they concluded that
speciation determined by the ion exchange method and by MINTEQA2 were in close agreement.
E-16
-------
Appendix E
MINTE0A2 Verification and Validation
-6
-7
c
^ -
4. 05
£
c <2
tl-5
D> S
¦2 c
o
-9
-10
-11
y = 1.04x + 0.31 r=0.98
-11 -10 -9 -8 -7 -6
log (Zn2+) (M)
MINTEQA2
Figure E-4. Comparison of Zn2+ determined by ion exchange
speciation (circles) and calculated by MINTEQA2 (line).
(After Fotovat and Naidu, 1997)
Title: Speciation of Dissolved Iron(II) and Manganese(II) in a Groundwater Pollution
Plume
Authors: Jensen, D.L., J.K. Boddum, S. Redemann, and T.H. Christensen
Source: Environmental Science & Technology, 32(18): 2657-2664
Date: 1998
Metals: Fe(II), Mn(II)
Basis of validation:
Compares speciation of Fe(II) and Mn(II) in solution as determined from ion exchange
procedures versus speciation computed using MINTEQA2
Description:
Analytical speciation results obtained using the cation exchange resin Amberlite were
determined for Fe(II) and Mn(II) in groundwater having high concentrations of these
constituents from an anaerobic pollution plume down-gradient from a landfill. The groundwater
sample pH ranged from 5.2 to 7.1. Batch experiments on the samples and on a reference
solution were conducted such that the experiments differed only with respect to the possible
presence of complexing ligands in the actual samples (but not in the reference). The reference
experiment provided the required information about the distribution of free Fe2+ and Mn2+
between the resin and the solution under conditions comparable to those of the sample. The free
Fe2+ and Mn2+ concentrations calculated for the samples was compared with that obtained using
MINTEQA2. The original database of MINTEQA2 was changed to include the stability
constants of Fe(II) and Mn(II) carbonate and bicarbonate complexes as given by Nordstrom el
al., 1990. MINTEQA2 predicted about 20 percent less free divalent Fe and Mn than calculated
from ion exchange measurements. The authors speculated that this difference might be due to
E^7
-------
Appendix E
MINTEQA2 Verification and Validation
uncertainty in the equilibrium constants of Fe(II) and Mn(II) carbonate and bicarbonate
complexes. They observed that the constants adopted from Nordstrom et al. (1990) are at the
high end of a range reported in the literature. Upon adjusting the equilibrium constants for
FeHC03+ and MnHC03+ to lower values within their reported ranges, MINTEQA2 predicted the
same fraction of free divalent Fe and Mn as was obtained from the ion exchange experiments.
They noted that MINTEQA2 indicated negligible complexation of Mn(II) with dissolved organic
matter (DOM) and that this was consistent with their interpretation of the ion exchange results.
There was no complexation reaction for Fe(II) with DOM in the MINTEQA2 database, but the
ion exchange results indicated that this also was negligible.
Title: Precipitation of Fe and Al Compounds from the Acid Mine Waters in Dogyae Area,
Korea: A Qualitative Measure of Equilibrium Modeling Applicability and
Neutralization Capacity?
Authors: Yu, J.Y.
Source: Aquatic Geochemistry, 2: 81-105
Date: 1996
Metals: Fe, Al
Basis of validation:
Comparison of the solid mineral phases that were predicted to precipitate by MINTEQA2 with
the solid phases observed in field samples using X-ray diffraction (XRD), infared (IR), thermal
and chemical analyses.
Description:
MINTEQA2 was use to predict the assemblage of Fe and Al solid mineral phases at equilibrium
with acid mine waters mixed with stream water. Stream water contaminated with acid mine
drainage water was collected along with associate streambed precipitates. The major ion
concentration and pH of the water was characterized and used as input values for MINTEQA2.
The precipitates in the collected samples were analyzed by XRD, IR, thermal, and chemical
methods to identify the solid phases. The MINTEQA2 equilibrium calculations indicated that
ferihydrite, FeOHS04, gibbsite, and A10HS04 are at equilibrium with the samples. However,
only ferrihydrite and Al4(OH)10SO4 were identified in the experimental analyses of streambed
samples. The authors suggested that FeOHS04 and A10HS04 are kinetically inhibited and the
metastable ferrihydrite and Al4(OH)10SO4 appear instead.
Title: Toxicity to Embryo and Adult Zebrafish of Copper Complexes with Two Malonic
Acids as Models for Dissolved Organic Matter
Authors: Palmer, F.B, C.A. Butler, M. H. Timperley, and C. W. Evans
Source: Environmental Toxicology and Chemistry, 17(8): 1538-1545
Date: 1998
Metals: Cu(II)
Basis of validation:
Comparison of Cu2+ activity calculated by MINTEQA2 with Cu2+ activity measured by an ion-
selective electrode.
Description:
Copper complexes with benzylmalonic acid and //-hexadecylmalonic acid were added to the
MINTEQA2 database and the model was used to calculate copper speciation in test solutions
used in toxicity studies. The calculated free Cu2+ concentration was compared with Cu2+
measured by ion selective electrodes. Good agreement was obtained for all test solutions. In
E-18
-------
Appendix E
MINTEQA2 Verification and Validation
addition, the Cu2+ concentration calculated by MINTEQA2 was found to correlate with the
median hatching times for zebrafish embryos in the test solutions.
Title: Bioavailability of Arsenic and Lead in Soils from the Butte, Montana, Mining District
Authors: Davis, A., M. V. Ruby, and P. D. Bergstrom
Source: Environmental Science & Technology, 26(3):461-468
Date: 1992
Metals: As, Pb
Basis of validation:
Comparison of MINTEQA2 calculated metal solubilities with measured metal solubilities in an
experimental solution.
Description:
After determining that enargite (Cu3AsS4) and anglesite (PbS04) were the likely controls on
arsenic and lead solubility in a test soil, the ability of MINTEQA2 to simulate the solution
concentration of these solids in the human GI tract was compared with experimental results. The
MINTEQA2-simulated solubility of anglesite compared well with measured solubilities: 37
mg/L simulated versus 35 mg/L measured in water at pH 7; 45 mg/L simulated versus 38 mg/L
measured in a 1 mM acetate solution. Simulation of enargite solubility at conditions of the GI
tract (pH 2, Eh +200 mv) showed virtually limitless solubility, as expected. The authors
concluded that MINTEQA2 can predict the equilibrium solubility of these minerals, but that
bioavailability of As and Pb from ingested soils would likely depend on dissolution kinetics and
soil resident time in the gastric system.
E.3.2 Validation in Complicated Systems (With Sorption or Complexation by
Doc)
The first three studies listed in this section were part of an organized effort by workers at
the U.S. EPA and the U.S. Geological Survey (USGS) to validate MINTEQA2. The interest of
the former in validating MINTEQA2 resulted from plans to possibly use the model in
development of Office of Solid Waste (OSW) regulations for the disposal of metal-bearing
wastes. One conclusion from the 1988 workshop discussing the feasibility of validating
MINTEQA2 was that efforts should be made to conduct a field test in which MINTEQA2 model
predictions could be compared with field measurements (Zachara et al., 1988). It was conceded
that successful application at one or two sites could not constitute a full and complete validation
of MINTEQA2 because the range of geochemical conditions involved in the test (pH, E,„ nature
of sorbents, solid phases, metals present, etc.) would necessarily be limited. However, one or
more partial validations were viewed as important in lending credibility to the U.S. EPA's
planned application of MINTEQA2 in modeling the movement of metal pollutants from landfill
sites.
Title: Chemical Speciation and Competitive Cationic Partitioning on a Sandy Aquifer
Material
Authors: Loux, N. T., D. S. Brown, C. R. Chafin, J. D. Allison, and S. M. Hassan
Source: Chemical Speciation and Bioavailability, 1(3): 111-125
Date: 1989
Metals: Ba, Be, Cd, Cu, Ni, Pb, Tl, Zn
E-19
-------
Appendix E
MINTEQA2 Verification and Validation
Basis of validation:
Comparison of percent of metal remaining in solution in batch equilibrium experiments using
aquifer materials versus percent of metal dissolved at equilibrium as calculated in MINTEQA2
simulations in which sorption and precipitation were operative.
Description:
This study was conducted specifically to test the ability of MINTEQA2 to account for metal
attenuation processes in a sandy aquifer material in an oxidized environment. The aquifer
material and associated groundwater were obtained from a Wisconsin aquifer having a high sand
content. Batch partitioning experiments on paired aquifer/groundwater samples were conducted
by spiking the samples with a solution containing Ba, Be, Cd, Cu, Ni, Pb, Tl, Zn at approximate
concentrations of 3 mg/L. Partitioning experiments were conducted over a pH range of 4 to 9.
At the conclusion of each 48 hour equilibration period, supernatant was removed and analyzed
by Inductively Coupled Plasma spectroscopy (ICP). The percent of the originally spiked metal
that remained in solution was plotted versus pH. MINTEQA2 was used to simulate the
partitioning experiments. Both precipitation and sorption were operative in the modeling.
Sorption was included in the modeling by invoking the diffuse-layer (MIT Two-Layer) model
with the database of sorption reactions for amorphous iron oxyhydroxide as given by Dzombak
(1986). The amorphous iron content used in the MINTEQA2 model runs was as measured in the
aquifer samples by the extraction method of Jenne and Crecilius (1988). Iron oxide sorbent site
densities for two site types and specific surfaces area were as recommended by Dzombak (1986;
0.2 and 0.005 moles of sites per mole Fe and 600 m2/g). A key point in this modeling was that
the database of sorption reactions and related parameters for the amorphous ferric oxyhydroxide
sorbent was used as given by Dzombak without alteration or adjustment. The sorption modeling
parameters were not fitted to the specific systems modeled. The simulated percent of metal
remaining in solution at equilibrium was compared with the measured value at each pH. Results
were presented graphically (see Figures 5-11 below). In interpreting the model results, Loux and
coworkers noted that sorption played a more important role in describing the pH-dependent
behavior of Ni, Pb, and Zn than for other metals; the equilibrium solution concentration of Cd
was better described by precipitation as a cadmium carbonate phase; pH-dependent behavior of
Ba, Be, and Cu were poorly described by the model. The authors speculated that complexation
of Cu with carbonato or organic complexes not represented in the model simulations may
explain the poor match between model and experimental result for this metal. They suggested
that Ba results could be explained by including non-specific adsorption terms or by invoking a
pH-sensitive solid- solution solubility control. Other key conclusions by the authors were that
"the agreement between model and experimental results was sufficient (for a number of
elements) to move the concept of partitioning behavior in complex environmental systems from
the realm of purely sediment-specific fitting to a more fundamental modeling basis," that the
database of thermodynamic constants for Be should be reviewed for completeness and accuracy,
and that removal of Tl from solution is relatively pH insensitive.
E-20
-------
Appendix E
MINTE0A2 Verification and Validation
100
Pb(OH) 2 pptn. significant only at pH > 9.5
Figure E-5. MINTEQA2 predicted Pb in solution (line) versus
experimentally observed values (squares).
(After Loux etal., 1989)
Zn(OH) 2 PPtn significant at pH > 9
Figure E-6. MINTEQA2 predicted Zn in solution (line) versus
experimentally observed values (squares).
(After Loux etal., 1989)
E-21
-------
Appendix E
MINTE0A2 Verification and Validation
Ni(OH) 2 pptn significant only at pH > 8.5 B
Figure E-7. MINTEQA2 predicted Ni in solution (line) versus experimentally
observed values (squares).
(After Loux eta!., 1989)
c.
o
is
o
CO
g.
cn
g
"c
n3
£
0)
cr
"D
o
100
90
80
70
60
50
40
30
20
10
0
V
V
¦
¦ ¦ ¦ " * ¦ ¦
V
V
¦
\
I
¦
¦
Adsorption + CdCO a pptn.
¦
\
(pK sp= 12.0)
¦
\
\
I ¦
10
PH
Figure E-8. MINTEQA2 predicted Cd in solution (line) versus experimentally
observed values (squares).
(After Loux etal., 1989)
E-22
-------
Appendix E
MINTE0A2 Verification and Validation
Cu(OH)2 pptn. significant only at pH > 8
Figure E-9. MINTEQA2 predicted Cu in solution (line) versus experimentally
observed values (squares).
(After Loux etal., 1989)
100
x^
¦ - — ¦ - ¦ _ ^
90
v<
* \
Adsorption + BaSO 4pptn.
\ \
log(K«i = -9.04
80 -
m \ \
¦ ¦ #
(Fruchter et. al.. 1988>
70 -
¦ _ i
7 ¦ Nv
Adsorption * BaSO 4pptn.
60 -
\
log(K = -9.42
50 -
\
\
M
40 -
\
30 -
¦
¦
20 ^
/
Adsorption * BaSO 4pptn.
S ' 1
10 -
log(K = -9.97
(crystalline)
o -
10
PH
Figure E-10. MINTEQA2 predicted Ba in solution (line) versus experimentally
observed values (squares).
(After Loux etal., 1989)
E-23
-------
Appendix E
MINTEQA2 Verification and Validation
c
o
o
CO
O)
c
*c
'cO
E
<15
GC
<8
100
90
80
70
60
50 -
40
30
20
10
0
Primarily Be(OH) 2pptn.
Adsorption (no OH "complexation Of pptn.)
Saturation of adscrptive sites
B ¦
./ i/nrrrn a
PH
10
Figure E-11. MINTEQA2 predicted Be in solution (line) versus experimentally
observed values (squares).
(After Loux eta!., 1989)
The following two studies concern the same study site: the Pinal Creek area of the
copper mining district near Globe, Arizona. The first study was commissioned by the U.S. EPA
and performed by Battelle Pacific Northwest Laboratories. It was originally suggested in the
model validation feasibility study previously mentioned (Zachara et al, 1988). The second study
at the same site was a collaborative effort by the U.S. Geological Survey (USGS) and the U.S.
EPA. It was published both as a journal article (Applied Geochemistry) and as part of a USGS
Water-Supply Paper. Both publications are cited below.
Title: Geochemical Model "Validation": Reliability of Solubility Equilibria Calculated with
Field Data from an Acidic Metal-rich Plume Near Globe, Arizona
Authors: Jenne, E.A.
Source: Unpublished report prepared for the U. S. EPA, Office of Research and Development,
Environmental Research Laboratory, Athens, GA by Battelle Pacific Northwest
Laboratories, Richland, WA
Date: 1994
Metals: Ca, Fe, Mn, Al, Si
Basis of validation:
The solid phases predicted to exist by MINTEQA2 were compared with solid phases identified
analytically in field samples (e.g., by scanning electron microscopy (SEM), x-ray diffraction
(XRD), or other means).
Description:
MINTEQA2 was used to predict the assemblage of equilibrium solid phases in samples obtained
from wells along a transect in a contaminated aquifer. For each sample, MINTEQA2 predicted
the saturation index (SI) for each possible solid phase. An SI value of zero for a solid phase
indicates equilibrium between that phase and the corresponding solution. The study site
included the area downgradient from tailings piles in the Globe, Arizona copper mining district.
E-24
-------
Appendix E
MINTEQA2 Verification and Validation
These tailings piles contain sulfide-bearing minerals, the oxidation of which has resulted in
infiltration of acidic, sulfate-rich water high in Cu and other metals. The area includes Pinal
Creek and a tributary, Miami Wash. This area in underlain by an unconsolidated alluvial aquifer
with a high hydraulic conductivity (200 to 300 m/day). This valley aquifer is bounded by
cemented formations of much lower conductivity. The pH of the uncontaminated aquifer ranges
from 6.4 to 8.4. A zone of neutralization has developed due to intrusion of the acidic water.
This zone has moved downgradient at about 0.5 m/yr. Over thirty observation wells that
penetrate the alluvial aquifer and transect the zone of neutralization are located in the valley
floor. In this study, samples from these wells, both up- and down-gradient from the
neutralization front, were used to represent the solution chemistry in MINTEQA2. In each
model run., the total concentrations of solutes, including metals, was as measured in the filtered
(0.45 |im) well water samples. The pH was constrained at the measured value of each sample.
The solid phases predicted to exist by MINTEQA2 (evidenced by SI approximately zero) were
compared with the solid phases identified analytically (e.g., by scanning electron microscopy
(SEM), x-ray diffraction (XRD), or other means). The MINTEQA2 predictions regarding the
existence of calcite and gypsum conformed reasonably with the observed occurrence of these
minerals. MINTEQA2 predictions indicated that amorphous iron hydroxide was oversaturated
in most samples; Fe(III) oxide coatings were identified by SEM. The authors suggested that
their method of quantifying the total Fe(III) for input in MINTEQA2 may have contributed to the
calculation of Fe(III) oxide over saturation. (They assumed that total measured Fe represented
Fe(II) and they calculated Fe(III) using the measured Eh value.) They also noted that a broad
range of solubility values for amorphous ferric oxide (corresponding to a range in log K) is
reported in the literature. Results also suggested that the MINTEQA2 model predictions for
rhodochrosite (MnC03) may correspond to observations. Results for A1 and Si were
inconclusive due to uncertainty in the actual controlling solid phases at the site.
Title: Geochemical Interactions Between Constituents in Acidic Groundwater and
Alluvium in an Aquifer near Globe, Arizona
Authors: Stollenwerk, K.G.
Source: Applied Geochemistry, 9:353-369
Date: 1994
Metals: Al, Fe, Mn, Ca, Cu, Co, Ni, Zn; also pH and S04
Basis of validation:
Comparison of dissolved concentrations of constituents measured in a series of wells with
solution concentrations predicted using MINTEQA2 with sorption and precipitation operative.
Description:
MINTEQA2 was used to simulate geochemical reactions between acidic mine drainage and
alluvial aquifer material. The study area was the Pinal Creek-Miami Wash area of the copper
mining district near Globe, Arizona. (The same study area as in Jenne,1994). At this site, acidic
water from tailings piles contains high sulfate concentrations from oxidation of sulfide-bearing
minerals. Higher than normal concentrations of trace metals also occur in the leachate. These
constituents move down-gradient through an unconsolidated alluvial aquifer that underlies
Miami Wash and Pinal Creek. In this study, a conceptual model of geochemical reactions was
first evaluated in a laboratory column experiment using aquifer materials. The objective of this
column calibration step was to identify the minimum set of geochemical reactions that would
explain the concentration of constituents in the column effluent. MINTEQA2 was used to
simulate those reactions and predict breakthrough curves for various constituents in the column.
E-25
-------
Appendix E
MINTEQA2 Verification and Validation
The grains of alluvial material in the column were visibly coated with iron oxides— the hydrous
ferric oxide (HFO) database of Dzombak and Morel (1990) was used in the diffuse-layer
sorption model in MINTEQA2 to estimate trace metal sorption. Aluminum and manganese
oxide sorbents were also known to be present in the alluvial material, but they were not
accounted for in the model. The specific surface area and sorption site density were set at values
recommended Dzombak and Morel for use with their database (600 m2/g and 0.2 moles of sites
per mole Fe; Dzombak and Morel, 1990). The amount of HFO used was the sum of an amount
determined by chemical extraction of the original alluvium and an amount estimated to
precipitate as the acidic leachate moved through the alluvium in the column. The equilibrium
constants for H, Cu, Ni, and Zn for sorption onto HFO were as specified in Dzombak and Morel
(1990). Published equilibrium constants for Co and Mn sorption were adjusted to fit the extent
of sorption in the column experiment. After successfully simulating breakthrough curves from
the column experiment, MINTEQA2 was used to simulate measured changes in concentration of
aqueous constituents in wells disposed along a flow path in the alluvial valley aquifer along
Miami Wash and Pinal Creek. The flow path chosen for simulation connected the most
contaminated well in each of six observation well nests and two surface water observations so
that a continuous progression from most contaminated to least contaminated water was
represented. In applying the column-calibrated geochemical model (embodied in MINTEQA2)
to the field, it was necessary to account for dilution of constituents in the acidic plume by
groundwater from other sources. A chloride tracer experiment was conducted to assess the
extent of dilution in the study area. PHREEQE was used to account for dilution of the
groundwater along the flowpath. The resulting mixed water chemistry was used as input to the
geochemical model represented in MINTEQA2. Results of the simulations were plotted as
constituent concentration versus distance with observations from the wells included for
comparison. Figures 12-18 present these comparisons for pH, Fe, Mn, Cu, Co, Ni, and Zn. In
Figure 12, the simulated change in pH along the flow path is shown. The reactions in the
geochemical model that accounted for the pH behavior were initial buffering by calcite and
dolomite until these minerals were depleted. After depletion of carbonates, the pH was
controlled by adsorption of H+ on oxide minerals and by reactions with aluminum (precipitation
of amorphous Al(OH)3 and A10HS04). Figures 13 and 14 show the simulation of Fe and Mn
concentration with distance compared with observed concentrations. In the model, oxidation of
Fe(II) to Fe(III) and subsequent precipitation was the controlling factor in Fe chemistry. The
Fe(II) was oxidized by Mn02 dissolution and reduction of Mn. The authors report that others
have provided evidence that the dissolution of Mn02 does occur along this flow path (Ficklin el
al., 1991). The total amount of Mn dissolved in the column experiment was less that the amount
predicted to dissolve in the oxidation of Fe(II). The authors discussed several explanations of
this difference. They chose to eliminate some of the Mn(II) predicted by MINTEQA2 because
the column experiment data indicated that approximately 65 percent of the Mn reduced by Fe(II)
remained in the solid phase (possibly reprecipitating as Mn(Fe02)2). Results for Cu, Co, Ni, and
Zn are shown in Figures 15-18. The controlling reactions for these constituents were the
sorption reactions onto ferric oxide. There was little sorption of Co, Ni, and Zn predicted along
the low pH region of the flow path (first 10 km). The simulated concentrations along this
portion of the transect were explained by dilution alone. Sorption of these metals became more
important down-gradient from 10 km where the pH increased. Sorption of Cu was important in
controlling Cu solution concentrations along the entire flow path and the combination of dilution
and adsorption accurately simulated Cu concentrations in the groundwater.
E-26
-------
Appendix E
MINTE0A2 Verification and Validation
1 1 1
A
~—~
I ' I
MEASURED
SIMULATED
I 1
I 1 1
1 1 1 1
1 1 1
09498400
452
- 09498380
RANGE IN MEASJRED
CONCENTRATION
AT SITE— NumDer is
site identifier. Three-
digit number rs well.
Number in italics is
streamflow-gaging
station
09498380
-
ffl 501
—
-
402
452 J
-
K
m
©
I ,
H01
' 1
i_
i , i ,
1 .
1 , 1
, I , I
. 1 . .
-2 0 S 4 6 8 10 12 14 16 18
DISTANCE, IN KILOMETERS
Figure E-12. Simulated versus measured pH along the flow path.
(Stollenwerk, 1994 and 1996)
60
A MEASURED
~—~ SIMULATED
051
MEASURED pH
RANGE IN MEASURED
CONCENTRATION
AT SITE — Number is
site identifier. Three-
digit number Is well.
Number in italics is
streamflow-gaging
452
09498380
rt 40
101
t- 20
304
402
452
09498400
501 09498380
-2
DISTANCE, IN KILOMETERS
Figure E-13. Simulated versus measured Fe concentration along the flow path.
(Stollenwerk, 1994 and 1996)
E-27
-------
Appendix E
MINTE0A2 Verification and Validation
tC
£
O
5
O
o
09498380
1 I " I
A MEASURED
O ~ SIMULATED
MEASURED pH
452
09498380
RANGE IN MEASURED
CONCENTRATION
AT SITE— Number is
site identifier. Three-
digit number is well.
Number in italics is
streamflow-gaging
station
09498400
¦ I
6 8 10
DISTANCE, IN KILOMETERS
X
Q.
Figure E-14. Simulated versus measured Mn concentration along
the flow path. (Stollenwerk, 1994 and 1996)
o
Q
o
A MEASURED
~ ~ SIMULATED
452
09498380
MEASURED pH
RANGE IN MEASURED
CONCENTRATION
AT SITE— Nurrber is
site dentifier. Three-
digit number is well.
Number in italics is
streamflow-gaging
station
6 8 10
DISTANCE, IN KILOMETERS
09498400
J ^
Figure E-15. Simulated versus measured Cu concentration along
the flow path. (Stollenwerk, 1994 and 1996)
E-28
-------
Appendix E
MINTE0A2 Verification and Validation
—p—i 1 > r
A MEASURED
O ~ SIMULATED
t 10
452
09498380
MEASURED pH
RANGE IN MEASURED
CONCENTRATION
AT SITE— Number is
ste Identifier. Three-
digit number is well
Number In italics is
streamflow-gaging
station
~~ 8
ii 402
09498400
a
09498380
— 4
— 2
6 8 10
DISTANCE, IN KILOMETERS
16 18
Figure E-16, Simulated versus measured Co concentration along
the flow path. (Stollenwerk, 1994 and 1996)
—I—'—I—r
A MEASURED
O ~ SIMULATED
MEASURED pH
1—1—i—1—r
0.06 f— 0511
452
09498380
RANGE IN MEASURED
CONCENTRATION
AT SITE— Number is
site identifier. Three-
digit number Is well.
Number
-------
Appendix E
MINTE0A2 Verification and Validation
D
0.4 | , j r- , - l 1 T 1 1 1 1 1 1 1" I I 1 I T
A MEASURED
~ ~ SIMULATED
RANGE IN MEASURED
CONCENTRATION
AT SITE—Number is
site Identifier. Three-
digit number is well.
Number in italics is
streamflow-gaging
station
452
09498380
09498400
DISTANCE. IN KILOMETERS
Figure E-18. Simulated versus measured Zn concentration along
the flow path. (Stollenwerk, 1994 and 1996)
Title: Simulations of Reactions Affecting Transport of Constituents in the Acidic Plume,
Pinal Creek Basin, Arizona
Authors: Stollenwerk, K.G.
Source: U.S. Geological Survey Water-Supply Paper 2466, pp. 21-49
Date: 1996
Metals: Al, Fe, Mn, Ca, Cu, Co, Ni, Zn; also pH and S04
Basis of validation: Comparison of dissolved concentrations of constituents in the effluent from
a column experiment with dissolved concentrations predicted by MINTEQA2 with sorption
operative.
Description: This paper reports on the same study as the preceding entry (Stollenwerk, 1994).
It also includes a new section not present in the 1994 paper. This new section discusses the
application of the same geochemical model (implemented in MINTEQA2) to transport of
contaminants in an aquifer that underlies the originally studied alluvium. This underlying basin-
fill area is characterized by a much lower hydraulic conductivity than the alluvium and by a
higher carbonate content. Some acidic leachate has penetrated this aquifer. The author applied
the same model as used in the valley aquifer to explain contaminant concentrations in this
aquifer. The pH-dependent sorption of Co, Cu, Ni and Zn and the precipitation of Al were
simulated. The basin-fill aquifer data (obtained from a column experiment using basin-fill
aquifer materials) were simulated using MINTEQA2 with the diffuse-layer sorption model with
parameters as described in Stollenwerk, 1994. The model predicted concentrations of Co, Ni,
and Zn that matched the experimental data reasonably well. (The match was described
qualitatively only in this paper.) Cu and Al were not detected in the effluent from the column
and this was consistent with MINTEQA2 predictions that virtually all Cu would be sorbed and
that Al would precipitate as amorphous Al(OH)3.
E-30
-------
Appendix E
MINTE0A2 Verification and Validation
Title: Modeling the Effects of Variable Groundwater Chemistry on the Adsorption of
Molybdate
Authors: Stollenwerk, K.G.
Source: Water Resources Research, 31(2):347-357
Date: 1995
Metals: Mo
Basis of validation: Comparison of dissolved concentration of molybdate in the effluent from a
column experiment with dissolved concentrations predicted by MINTEQA2 with sorption
operative.
Description: MINTEQA2 was used to estimate adsorption of molybdate (Mo04) in column
experiments using sediment and groundwater collected from a shallow alluvial aquifer near Cape
Cod, MA. Two column experiments were performed, one using groundwater from a sewage
contaminated well (F347-46) and one using groundwater from a well not contaminated with
sewage (F347-20). The modeling of the column experiments employed the diffuse-layer model
(Dzombak and Morel, 1990). Surface acidity constants for H+ sorption and adsorption reactions
for S04 and P04 were included to account for change in Mo04 adsorption with pH and with
competition for sorption sites due to changing concentrations of these ions. The acidity
constants and other equilibrium constants were determined in separate batch experiments. The
surface area and surface site density were measured for the alluvial aquifer material. The match
between predicted and measured Mo04 adsorption for both groundwaters for various pH values
Figure E-19. MINTEQA2-simulated versus measured
Mo04 for column experiments using aquifer materials
and water from sewage-contaminated well (F347-46) and
uncontaminated well (F347-20). (Stollenwerk, 1995)
is shown in Figure 19.
E-31
-------
Appendix E
MINTEQA2 Verification and Validation
Title: Predicting the Environmental Stability of Treated Copper Smelter Flue Dust
Authors: Doyle, T.A., A. Davis, and D.D. Runnels
Source: Applied Geochemistry, 9:337-350
Date: 1994
Metals: As(V)
Basis of validation: Comparison of dissolved concentrations of As in batch and column tests
with dissolved concentrations predicted by MINTEQA2
Description: The leachability of As, Cd, Cu, Fe, and Pb from treated copper smelter flue dust
was investigated by batch and column tests. Electron microprobe spectroscopy of the treated
flue dust identified scorodite (FeAs04.2H20) as the primary As mineral. Modeling of the system
using MINTEQA2 also predicted that scorodite controlled the leaching of As from the column.
The MINTEQA2 predicted leachate concentration of As (1270 |ig/L) compared well with the
measured value (1330 |ig/L) from the continuously eluted column test. The triple-layer model in
MINTEQA2 was used to model sorption of As(V). Including sorption in MINTEQA2
simulations of batch tests resulted in overestimating As removal. However, results for a
continuously eluted column test showed good agreement between dissolved As concentration
estimated using MINTEQA2 with the triple-layer model for sorption (1120 |ig/L) and actual
measured As in the column leachate (1330 |ig/L). The triple-layer model in MINTEQA2 was
used with reactions and parameters from Davis and Leckie (1978, 1980). The equilibrium
constant for formation of scorodite used in the simulation was calculated from data provided in
Robins (1990).
Title: Modeling of radionuclide and heavy metal sorption around low- and high-pH waste
disposal sites at Oak ridge, Tennessee.
Authors: Saunders, J.A. and L.E. Toran
Source: Applied Geochemistry, 10:673-684
Date: 1995
Metals: Co, Cd, Pb, Sr, U, and Zn
Basis of validation: Comparison of dissolved concentrations of metals at monitoring wells near
a disposal pond with dissolved concentrations of these metals predicted by MINTEQA2.
Description: MINTEQA2 was used to predict the mobility of radioisotopes at low-level
radioactive waste disposal sites. Mineral precipitation and metal sorption reactions were
included in the simulations. The simulations included estimating the neutralization of both low-
and high-pH waste leachates due to interaction with soils. Complexation reactions with EDTA
were also included. The sorption model used was the diffuse-layer model with sorption
reactions for iron oxyhydroxide with associated sorption site densities and surface area as
specified in Dzombak and Morel (1990). The authors compared MINTEQA2-computed
dissolved Co, Cd, Pb, Sr, U, and Zn at various pH values with observations in near-source
monitoring wells. Although they acknowledged several complicating factors not included in
their modeling methodology (presence of additional soil sorbents besides iron oxyhydroxide,
kinetics of mineral dissolution and precipitation, effects of flow along fractures, dilution by
groundwater), they concluded that "predictions about pH-dependent mineral precipitation and
metal sorption reactions from the modeling generally match field observations..." for both the
acidic and high-pH disposal sites. (Figures showing model versus measured concentrations for
Cd and Zn are included in the original paper, but their quality is too poor to allow reproduction.)
E-32
-------
Appendix E
MINTEQA2 Verification and Validation
Title: Complexation of Cd, Ni, and Zn by DOC in Polluted Groundwater: A Comparison of
Approaches Using Resin Exchange, Aquifer Material Sorption, and Computer
Speciation Models (WHAM and MINTEQA2)
Authors: Christensen, J.B. and T.H Christensen
Source: Environmental Science & Technology, 33:3857-3863
Date: 1999
Metals: Cd, Ni, and Zn
Basis of validation: Concentrations of metal-DOC complexes determined in batch experiments
using a resin exchange method were compared with concentrations of metal-DOC complexes
computed by MINTEQA2
Description: This study compared the extent of metal-DOC complexation measured in a resin
exchange method, measured in batch sorption experiments using actual aquifer materials, and
estimated using the speciation models WHAM and MINTEQA2. Groundwater samples were
obtained from the DOC-rich leachate plume downgradient from a landfill. A synthetic leachate
solution was prepared so as to be similar to each of these leachate samples except that DOC was
not included. The with- and without-DOC solutions were used in the resin exchange method and
in batch experiments with aquifer materials to estimate DOC complexation of Cd, Ni, and Zn.
DOC complexation of these metals was also estimated by simulating the leachate solutions in
WHAM (Tipping, 1994) and MINTEQA2. The authors reported that the resin exchange method
and the batch equilibration with aquifer materials provided similar measures of DOC
complexation of these metals. The ability of the models to predict the complexation of these
metals with DOC varied among the metals and the different samples. Concerning the predictions
by MINTEQA2, excellent agreement with the experimental methods was noted for complexation
of Cd. Agreement of MINTEQA2 results with experimental methods for Zn were described as
"fair." For Ni, MINTEQA2 results were described as in "fair" agreement with experimental
results for one sample, but for a second sample, MINTEQA2 significantly underestimated Ni
complexation with DOC. Several of these comparisons are shown in Figure 20. It was also
noted that the shape of the curves produced by increasing DOC concentration in the experiments
was generally well matched in the MINTEQA2 results. The authors concluded that in leachate
polluted groundwater, the MINTEQA2 model gives a useful first approximation of Cd, Ni, and
Zn complexation by DOC.
E-33
-------
Appendix E
MINTE0A2 Verification and Validation
(M+MDOC)/M
LI Cadmium
8-
6-
5-
4-
3-
2-•
200
250
150
0
50
too
(Or
8 - -
6-
5-
4-
L1 Nickel
3--
2
200
250
50
100
150
0
DOC (mgCL'1)
~
Resin experiments
~
Aquifer material*
o
Aquifer material11
f
WHAM default proton parameters
—
WHAM specific proton parameters
M1NTEQA2 default proton parameters
(M+MDOC)/M
L2 Cadmium
4 ¦ ¦
3-
200
250
100
150
50
8
6-
5-
4-
3-
2-
100
150
200
250
0
50
1001-
L2 Zinc
50
20
10::
5 ¦
2
250
50
100
150
200
0
DOC (mg C L"')
Figure E-20. Comparison of MINTEQA2 estimated degree of DOC complexation for Cd, Ni,
and Zn versus degree of complexation in resin and batch aquifer material experiments.
(Christensen and Christensen, 1999)
Title: Complexation of Cu and Pb by DOC in polluted groundwater: A comparison of
experimental data and predictions by computer speciation models (WHAM and
MINTEQA2)
Authors: Christensen, J.B., J.J. Botma, and T.H Christensen
Source: Water Research, 33(15):3231-323 8
Date: 1999
Metals: Cu, Pb
Basis of validation: Concentrations of metal-DOC complexes determined in batch experiments
using a resin exchange method were compared with concentrations of metal-DOC complexes
computed by MINTEQA2
Description: This study compared the extent of metal-DOC complexation measured in a resin
exchange method and estimated using the speciation models WHAM and MINTEQA2.
Groundwater samples were obtained from the DOC-rich leachate plume downgradient from a
landfill. A synthetic solution was prepared so as to be similar to each of these leachate samples
E-34
-------
Appendix E
MINTE0A2 Verification and Validation
except that DOC was not included. The with- and without-DOC solutions were used in the resin
exchange method to estimate DOC complexation of Cu and Pb. DOC complexation of these
metals was also estimated by simulating the leachate solutions in WHAM (Tipping et al., 1994)
and MINTEQA2. The ability of the models to predict the complexation of these metals with
DOC varied among different samples. Concerning the predictions by MINTEQA2, agreement
between experimental and observed complexation was generally good (see Figure 21). For one
sample MINTEQA2 underestimated the degree of Cu complexation by an amount corresponding
to a maximum of a factor of two in the free Cu2+ concentration. For one sample MINTEQA2
overestimated DOC binding with Pb, again by an amount corresponding to a factor of two in the
free Pb2+ concentration.
Copper
degree of complexation
1.0
}.9
).e
).7
>.6
>.5
>.4
).3
),2
>.1
0
0 50 100 150
degree of complexation
0
>.9
1.8
1.7
1.6
1.5
1.4
1.3
(.2
1.1
0
0 50 100 150
DOC concentration (mg C/l)
Degree of complexation
1.0
09/ Copper
! ~ co
0.8 ¦
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Lead
S2
0 50 100 150
Degree of complexation
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
50 100 150
DOC concentration (mg C/l)
~ Resin experiments WHAM MINTEQA2
Figure E-21. Comparison MINTEQA2 estimated degree of DOC complexation for Cu and Pb
versus degree of complexation in resin experiments.
(Christensen etal., 1999)
E-35
-------
Appendix E
MINTEQA2 Verification and Validation
Title: The Effect of pH on the Complexation of Cd, Ni, and Zn by Dissolved Organic
Carbon from Leachate-polluted Groundwater
Authors: Christensen, J.B. and T.H Christensen
Source: Water Research, 34:3743-3754
Date: 2000
Metals: Cd, Ni, and Zn
Basis of validation: Concentrations of metal-DOC complexes determined in batch experiments
using a resin exchange method were compared with concentrations of metal-DOC complexes
computed by MINTEQA2 over a range of pH values.
Description: This study compared the extent of metal-DOC complexation measured in a resin
exchange method and estimated using the speciation models WHAM and MINTEQA2 over a
range of DOC concentrations and pH values. Groundwater samples were obtained from the
DOC-rich leachate plume downgradient from a landfill. A synthetic solution was prepared so as
to be similar to each of these leachate samples except that DOC was not included. The with- and
without-DOC solutions were used in the resin exchange method to estimate DOC complexation
of Cd, Ni, and Zn at various combinations of DOC concentration and pH . DOC complexation
of these metals was also estimated by simulating the leachate solutions in WHAM (Tipping et
al., 1994) and MINTEQA2. Concerning the predictions by MINTEQA2, a very poor match with
experimental results was obtained for all three metals because the MINTEQA2 result did not
show appropriate pH response (see Figure 22). The degree of DOC complexation predicted by
MINTEQA2 did not change appreciably as the pH was changed. The authors suggested that
MINTEQA2 predictions of pH dependent complexation by DOC could be improved
considerably by including a second site type (phenolic) in the MINTEQA2 representation of
DOC. The importance of the phenolic site relative to MINTEQA2 predictions has been
discussed previously by Allison and Perdue (1994).
E-36
-------
Appendix E
MINTE0A2 Verification and Validation
(M+MDOC)/M
(M+MDOC)/M
S2 Cadmium
S1 Cadmium
52 Nickel
S1 Nickel
S1 Zinc
S2 Zinc
100
DOC (rng C/L)
50 100 150
DOC (mg C/L)
200
200
200
~ pH 5 O pH 5.5 • pH 6 O pH 6.5 ¦ pH 7 ~ pH 7.5 ~ pH 8
— MINTEQA2 pH 5 MINTEQA2 pH 6 M1NTEQA2 pH 7 MINTEQA2 pH 8
Figure E-22. Comparison of MINTEQA2 predictions of DOC complexation of Cd, Ni, and Zn
versus experimentally determined values (resin exchange) at various pH values.
(Christensen and Christensen, 2000)
E-37
-------
Appendix E
MINTE0A2 Verification and Validation
Title: Chemical Modeling of the Neutralisation Process for Acid Uranium Mill Tailings
Authors: Khoe, G.H. and G. Sinclair
Source: Proceedings, Hydrometallurgy and Aqueous Processing Symposium, 1991 Annual
Meeting of the Metallurgical Society of AIME, New Orleans, LA, EP Congress 91
Date: 1991
Metals: Al, Fe, Ca, Mn, Si, P04, Pb, U
Basis of validation: Comparison of dissolved metal concentrations predicted by MINTEQA2
versus concentrations measured in neutralization experiments.
Description: MINTEQA2 was used to model the neutralization of uranium mill tailings. The
adsorption of U022+ via the triple-layer model was included. The acid tailings solution was
represented in MINTEQA2 and successive additions of the liming agent were simulated by
additions of portlandite. Comparison of the predicted pH and experimental pH measurements
were similar. The MINTEQA2-predicted Al, Fe, Ca, Si, and P04 solution concentrations versus
pH were in good agreement with measured values. The agreement between dissolved Mn
concentration versus pH predicted by MINTEQA2 and experimental values was not as good.
The authors conducted other experiments which they interpreted as suggesting C02 equilibria as
the reason for this discrepancy. (The experimental solution did not have time to fully equilibrate
with the atmosphere.) This study also included sorption modeling for Pb and U in the
neutralized tailings using the triple-layer model in MINTEQA2. This modeling employed
sorption reactions and equilibrium constants published by Davis and Leckie (1978) and Payne
and Waite (1990). Comparison of model predictions and experimental results are shown in
Figures 23 and 24 for U and Pb.
100
with adsorption
without
adsorption
c
o 60
e
S
g
0> 40
Q.
* experimental
data
Figure E-23. Percent of U022+ predicted as remaining in
solution by MINTEQA2 (line) versus measured (X) over
range of pH. Circles and triangles denoted model
simulations with and without including sorption.
(Khoe and Sinclair, 1991)
E-38
-------
Appendix E
MINTE0A2 Verification and Validation
Title: Arsenic Adsorption from Geothermal Water
Authors: Webster, J.G . and K.S. Webster
Source: Sixteenth AnnualPNOC Geothermal Workshop Proceedings, Manila, Philippines, pp.
Date: 1995
Metals: As(III) and As(V)
Basis of validation: Comparison of solution As concentrations measured in batch experiments
with dissolved concentrations predicted by MINTEQA2 using the diffuse-layer sorption model
Description: MINTEQA2 was used to predict the sorption of arsenic (As) in batch experiments
using As contaminated geothermal bore water and a prepared, freshly precipitated hydrous ferric
oxide (HFO) sorbent. The major ion concentrations of the water samples were characterized and
entered in MINTEQA2. Experiments were carried out over a range of pH values (3 to 11).
Experiments were also conducted using a 0.1 M NaN03 solution rather than the geothermal bore
water. The MINTEQA2 modeling employed the diffuse-layer sorption model and associated
database for HFO (Dzombak and Morel, 1990). MINTEQA2 predictions were consistent with
experimental observations for the 0.1 M NaN03 solution, especially for pH < 9. However, for
both As(III) and As(V), MINTEQA2 consistently predicted less As remaining in solution (i.e.,
more As sorption) than was observed in the bore water experiments. For pH <9, observed
As(III) sorption was < 50 percent of the value predicted by MINTEQA2 and observed As(V)
sorption ranged from 33 to 95 percent of the predicted value. The authors were unable to explain
the discrepancy between the model and experimental results at the time of publication of this
paper. However, in a follow-up study they concluded that very high silica concentration in the
geothermal bore water samples resulted in sorption of Si02 and reduced the sorption of As in the
bore water experiments (J. Webster, pers. comm, 1996). Silica adsorption had not been included
in the original MINTEQA2 modeling. They repeated the MINTEQA2 modeling of the bore
water with sorption reactions for silica onto HFO (equilibrium constants estimated) and obtained
a much closer match with experimental results.
Title: Simulating the Response of Metal Contaminated Lakes to Reductions in Atmospheric
35-42
100
80
wtth adsorption
* experimental data
2
3
4
5
6
PH
Figure E-24. Percent of Pb predicted as
remaining in solution by MINTEQA2 (line)
versus measured (X) over range of pH.
Circles and triangles denoted model
simulations with and without including
sorption. (Khoe and Sinclair, 1991)
E-39
-------
Appendix E
MINTEQA2 Verification and Validation
Loading Using a Modified QWASI Model
Authors: Woodfine, D.G., R. Seth, D. Mackay, and M, Havas
Source: Chemosphere, 41:1377-1388
Date: 2000
Metals: Cu, Ni
Basis of validation: Comparison of QWASI simulation results of average lake water dissolved
metal concentrations with observed values when MINTEQA2-predicted partition coefficients are
used.
Description: MINTEQA2 was used to simulate the effect of pH, solution chemistry, particulate
(sorbent) concentrations and metals loading on metal partition coefficients. The simulated
coefficients were used in the QWASI (Quantitative Water Air Sediment Interaction) transport
model. This model accounts for metals loading of surface water bodies via atmospheric
deposition of metal-laden dust particles originating at smelter sites. Previous versions of the
QWASI model had employed constant partition coefficients obtained from the literature or
experimentally. The study described here included the use of MINTEQA2 for metal speciation
and to calculate partition coefficients for association of metals with suspended particles in two
water bodies (Alice Lake and Baby Lake). The lake suspended matter was assumed to be
dominated by hydrous ferric oxide (HFO) and the database of sorption reactions for HFO from
Dzombak and Morel (1990) was used in MINTEQA2 to quantify sorption. The surface site
density for the suspended particles was as determined by Hamilton-Taylor et al. (1997) for lake
suspended particles (0.8 mmol sites per g). Estimated copper (Cu) and nickel (Ni) deposition
rates over a twenty year period beginning in 1972 were used to parameterize the QWASI model.
The QWASI simulations produced annual estimated lake concentrations for Cu and Ni using the
MINTEQA2-predicted partition coefficients for both lakes. Results are plotted versus measured
dissolved metal concentrations in these lakes (Figure 25). Previous studies for metals using the
QWASI model employed constant partition coefficients from experiments or from the literature.
In this first use of a metal speciation model to predict metal partition coefficients for variable
chemical conditions and metal loading within QWASI, the authors concluded that the model
simulations were reasonable given uncertainties in the input data.
E-40
-------
Appendix E
MINTE0A2 Verification and Validation
BABY LAKE
3.5
cr
E,
d
C
o
O
B
35
TO
* Measured
Simnlatprl
Nickel
¦ 1972 1977 1962 1987 1992
Year
Measured
Simulated
E 05
Copper
1982
1987 1992
ALICE LAKE
Measured
.Simulated
Nickel
1977
1982
Year
1987 1992
Measured
Simulated
Copper
B 0,4
1972 1977 1982 1987 1992
Year
Figure E-25. Dissolved Cu and Ni concentrations simulated by QWASI model using
MINTEQA2-generated partition coefficients versus measured metal concentrations.
(Woodfine et a!., 2000)
E-41
-------
Appendix E
MINTEQA2 Verification and Validation
Title: Trace Element Geochemistry of Onion Creek near Van Stone Lead-Zinc Mine
(Washington, USA) - Chemical Analysis and Geochemical Modeling
Authors: Routh, J. and M. Ikramuddin
Source: Chemical Geology, 133:211-224
Date: 1996
Metals: Pb, Zn
Basis of validation: Comparison of MINTEQA2-predicted equilibrium solid phases with solid
phases observed by X-ray diffraction and predicted water concentrations with observed
concentrations.
Description: MINTEQA2 was used to predict the concentration of trace elements in surface
waters of Onion Creek. Onion Creek (Washington) receives surface runoff and seepage from
tailings ponds associated with lead-zinc mining activities. The MINTEQA2 modeling included
precipitation reactions for major ions and trace elements and sorption reactions for Pb and Zn.
An experimentally determined amount of ferrihydrite was used to represent the sorbent in
MINTEQA2; its surface parameters were as described by Dzombak and Morel (1990) for
hydrous ferric oxide (surface area 600 m2/g and site density 0.2 moles of sites per mole Fe).
Sorption reactions also were as specified by Dzombak and Morel (1990). The pH of Onion
Creek water was 7.9. X-ray diffraction analyses of sediment samples confirmed the presence of
dolomite, calcite, A1 -oxide, Fe-Mn-oxide, and zinc sulfide as predicted by MINTEQA2. The
model also predicted low dissolved concentrations of Pb and Zn due to sorption onto
ferrihydrite. This finding was supported by high partition coefficients computed for these
elements in Onion Creek waters (> 104 L/kg).
E-42
-------
Appendix E
MINTEQA2 Verification and Validation
E.4 Conclusions
The MINTEQA2 validation studies reviewed present a broad spectrum of natural
conditions, analytical methods, and metals of interest. The studies also cover most of the
important MINTEQA2 sub-models that are important to the use of MINTEQA2 to support
rulemaking activities by the U.S. EPA Office of Solid Waste. These activities include the
Hazardous Waste Identification Rule (circa 1995-96) and the 3MRA of 1999. In modeling
support for both of these activities, MINTEQA2 modeling was performed to estimate metal
partition coefficients for use in groundwater transport modeling (US EPA, 1996 and 1999). This
modeling involved precipitation of major ions (Al, Ca, Fe, etc.) and sorption of trace metals.
The sorption modeling was included by means of the diffuse-layer model with the database of
sorption reactions provided by Dzombak (1986) and Dzombak and Morel (1990). The Gaussian
DOM model was used to estimate metal binding with dissolved organic matter. The modeling
was performed over a broad range of pH, HFO sorbent, and metal concentrations.
The pH range covered among all studies cited is 4 to 12. Some studies were performed
using laboratory systems exclusively (e.g., freshly precipitated pure phase HFO as the sorbent)
and some involved natural aquifer materials and soils. The analytical methods used included
ICP spectroscopy, X-ray diffraction, scanning electron microscopy, speciation using ion
exchange resins, and ion-selective electrodes. The metals for which validating comparisons
were attempted included Al, Ca, Fe, Mn, P04, Si02, S04, Cd, As, Ba, Be, Cu, Cd, Co, Mo, Ni,
Pb, Sr, Tl, U, and Zn (and pH). The studies presented results for MINTEQA2 modeling in
which the main process removing metals from solution was assumed to be precipitation as well
as studies in which sorption was assumed to be the operative process in reducing dissolved metal
concentrations. Most of the studies that focused on sorption employed the diffuse-layer model
with the database provided by Dzombak and Morel (1990). The Gaussian DOM model was also
included in some modeling studies.
Most of the studies were not undertaken for the specific purpose of validating
MINTEQA2, although a few were (e.g., Loux et al., 1989, Stollenwerk 1994, 1996). These may
perhaps carry more weight in assessing the overall validation status of MINTEQA2 in that both
included considerable connection with natural field materials and both place emphasis on the
role of sorption. As previously discussed, the 1988 workshop on MINTEQA2 validation
presented a broad range of opinions as to what might constitute a validation of this model. The
consensus was that a broad range of studies that show some level of correspondence between the
model's predictions and actual measurements might be the best that could be hope for. Here, we
leave the reader to review the studies cited and form his or her own opinion regarding the
validation status of MINTEQA2.
E-43
-------
Appendix E
MINTEQA2 Verification and Validation
This page intentionally left blank.
E-44
-------
Appendix E
MINTEQA2 Verification and Validation
E.5 References
Allison, J.D.,1997. An Implementation of the Competitive Gaussian Model for Metal-humic
Binding in a General Speciation Model, Ph.D. Dissertation, Georgia Institute of
Technology, Atlanta, GA.
Allison, J.D., 1996. Update and Correction of Metal-Organic Complexes for the MINTEQA2
Thermodynamic Database, prepared for U.S. Environmental Protection Agency, National
Exposure Research Lab, Ecosystems Research Division, Athens, GA.
Allison, J.D. and E.M. Perdue, 1994. Modeling metal-humic interactions with MINTEQA2, in
Humic Substances in the Global Environment and Implications on Human Health, edited
by N. Senesi and T.M. Miano, Elsevier Science B.V.
Ball, J.W., D.K. Nordstrom, and E.A. Jenne, 1981. Additional and Revised Thermochemical
Data and Computer Code for WATEQ2- A computerized Chemical Model for Trace and
Major Element Speciation and Mineral Equilibrium of Natural Waters, U.S. Geological
Survey, Water Resources Investigations, WR1 78-116.
Bard, A. J., R. Parsons, and J.Jordan (eds.), 1985. Standard Potentials in Aqueous Solution,
International Union of Pure and Applied Chemistry, Marcel Dekker, Inc., New York,
834p.
Christensen, J.B. and T.H Christensen, 1999. Complexation of by DOC in polluted groundwater:
A comparison of approaches using resin exchange, aquifer material sorption, and
computer speciation models (WHAM and MINTEQA2), Environmental Science &
Technology, 33:3857-3863.
Christensen, J.B., J.J. Botma, and T.H Christensen, 1999. Complexation of Cu and Pb by DOC
in polluted groundwater: A comparison of experimental data and predictions by
computer speciation models (WHAM and MINTEQA2), Water Research, 33(15):3231-
3238.
Christensen, J.B. and T.H Christensen, 2000. The effect of pH on the complexation of Cd, Ni,
and Zn by dissolved organic carbon from leachate-polluted groundwater, Water
Research, 34:3743-3754.
Davis, A., M. V. Ruby, and P. D. Bergstrom, 1992. Bioavailability of arsenic and lead in soils
from the Butte, Montana, mining district, Environmental Science & Technology,
26(3):461-468.
E-45
-------
Appendix E
MINTEQA2 Verification and Validation
Davis, J. A. and J.O. Leckie, 1978. Surface ionization and complexation at the oxide/water
interface II. Surface properties of amorphous iron hydroxide and adsorption of metal
ions, Journal of Colloidal and Interface Science, 67:90-107.
Davis, J. A. and J.O. Leckie, 1980. Surface ionization and complexation at the oxide/water
interface 3. Adsorption of anions, Journal of Colloidal and Interface Science, 74:32-43.
Deutsch, W.J. and K.M. Krupka, 1985. MINTEQ Geochemical Code: Compilation of
Thermodynamic Database for the Aqueous Species, Gases, and Solids Containing
Chromium Mercury, Selenium, and Thallium. Unpublished report prepared by Battelle
Pacific Northwest Laboratory for the U.S. Environmental Protection Agency, Athens,
GA.
Dobbs, J.C., W. Susetyo, L.A. Carreira, and L.V. Azarraga, 1989. Competitive binding of
protons and metal ions in humic substances by lanthanide ion probe spectroscopy,
Analytical Chemistry, 61:1519-1524.
Doyle, T.A., A. Davis, and D.D. Runnels, 1994. Predicting the environmental stability of treated
copper smelter flue dust, Applied Geochemistry, 9:337-350.
Drever, J.I., 1997. The Geochemistry of Natural Waters- Surface and Groundwater
Environments (3rd ed.), Prentice-Hall, New Jersey.
Dzombak, D.A. and F.M.M. Morel, 1990. Surface Complexation Modeling: Hydrous Ferric
Oxide. John Wiley & Sons, New York, 393p.
Dzombak, D.A., 1986. Toward a Uniform Model for the Sorption of Inorganic Ions on Hydrous
Oxides, Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge, MA.
Dzombak, D.A. 1987. A Practical Guide for Use of the Generalized Two Layer Sorption Model,
prepared for the U.S. Environmental Protection Agency, Athens, GA.
Felmy, A.R., D.C. Girvin, and E.A. Jenne, 1984. MINTEQ - A Computer Program for
Computing Aqueous Equilibria, prepared by Battelle Pacific Northwest Laboratories for
the U.S. Environmental Protection Agency, Athens, GA.
Ficklin, W.H., A.H. Love, and C.S.E. Papp, 1991. Solid phase variations in an aquifer as the
solution changes, Globe, Arizona. U.S. Geological Survey Water Resource
Investigations Report 91-4034, pp.475-480.
Fish, W., D.A. Dzombak, and F.M.M. Morel, 1986. Metal-humate interactions. 2. Application
and comparison of models, Environmental Science & Technology, 20: 676-683.
Fotovat, A. and R. Naidu,1997. Ion exchange resin and MINTEQa2 speciation of Zn and Cu in
alkaline sodic and acid soil extracts, Australian Journal of Soil Research, 35: 711-726.
E-46
-------
Appendix E
MINTEQA2 Verification and Validation
Frandsen, A. K., and C. H. Gammons, 2000. Heavy metal complexation with aqueous sulfide in
an anaerobic treatment wetland, in Wetlands and Remediation: An International
Conference, Battelle Press, Columbus, OH, pp. 423-430.
Hamilton-Taylor, J., L. Giusti, W. Davidson, W. Tych, and C.N. Hewitt, 1997. Sorption of
trace metals (Cu, Pb, Zn) by suspended lake particles in artificial (0.005 M NaN03) and
natural (Esthwaite Water) freshwaters, Colloids Surfaces A: Physicochem. Engrg.
Aspects, 120:205-219.
IUPAC, 1998. Stability Constants Database SC-DATABASEfor Windows v2.62 and 3.02,
International Union of Pure and Applied Chemistry (IUPAC) and Academic Software,
United Kingdom..
Jenne, E.A., 1994. Geochemical Model "Validation": Reliability of Solubility Equilibria
Calculated with Field Data from and Acidic Metal-rich Plume near Globe, Arizona,
prepared by Battelle Pacific Northwest Laboratories for the U.S. Environmental
Protection Agency, Athens, GA.
Jenne, E. A. and E. A. Crecilius, 1988. Determination of sorbedmetals, amorphic Fe, oxidixMn,
and reactive particulate organic carbon in sediments and soils. Presented at the 3rd
International Conference on Environmental Contamination, Venice, Italy, Sept 26-29,
1988.
Jenne, E.A. and K.M. Krupka, 1985. Validation of geochemical models, in Workshop on
Fundamental Geochemistry Needs for Nuclear Waste Isolation, B.R. Erdal (ed.), pp. 143-
152.
Jensen, D.L., J.K. Boddum, S. Redemann, and T.H. Christensen, 1998. Speciation of dissolved
iron(II) and manganese(II) in a groundwater pollution plume, Environmental Science &
Technology, 32(18): 2657-2664.
Khoe, G.H. and G. Sinclair, 1991. Chemical modeling of the neutralisation process for acid
uranium mill tailings, Proceedings, Hydrometallurgy and Aqueous Processing
Symposium, 1991 Annual Meeting of the Metallurgical Society of AIME, New Orleans,
LA, EP Congress 91.
Krupka, K.M., E.A. Jenne, and W.J. Deutsch, 1983. Validation of the WATEQ Geochemical
Model for Uranium, PNL-4333, Battelle Pacific Northwest Laboratories, Richland, WA.
Lindsay, W. L. and H.A. Ajwa, 1995. Use of MINTEQA2 for Teaching Soil Chemistry, in
Chemical Equilibrium and Reaction Models, edited by R.H. Loeppert, A.P. Schwab, and
S. Goldberg, Soil Science Society of America, Special Publication 42, Madison,
Wisconsin, pp.241-252.
Loux, N.T., D. S. Brown, C.R. Chapin, J.D. Allison, and S.M. Hassan, 1989. Chemical
speciation and competitive cationic partitioning on a sandy aquifer material, Chemical
Speciation and Bioavailability, 1:111-125.
E-47
-------
Appendix E
MINTEQA2 Verification and Validation
Marani, D., G. Macchi, and M. Pagano,1995. Lead precipitation in the presence of sulphate and
carbonate: Testing of thermodynamic predictions, Water Research, 29(4): 1085-1092
Mathur, S. S., 1995. Development of a Database for Ion Sorption on Goethite Using Surface
ComplexationModeling, Master's Thesis, Department of Civil and Environmental
Engineering, Carnegie Mellon University, Pittsburgh, PA.
Morrey, J.R., C.T. Kincaid, and C.J. Hostetler, 1985. Geochemical Models for Solute Migration,
Volume 3: Evaluation of Selected Computer Codes for Modeling Aqueous Solutions and
Solute Migration in Soils and Geologic Media, EA-3417, Battelle Pacific Northwest
Laboratories, Richland, WA.
NIST, 1997. Critical Stability Constants of Metal Complexes Database (CRITICAL, version 2.0
and 4.0), Standard Reference Database 46, National Institute of Standards and
Technology, Washington, DC.
Nordstrom, D.K. and J.L. Munoz, 1986. Geochemical Thermodynamics, Blackwell Scientific,
Palo Alto, CA, pp. 365-391.
Nordstrom, D.K., L.N. Plummer, D. Langmuir, E. Busenberg, H. May, B,L. Jones, and D.L.
Parkhurst (1990) Revised chemical equilibrium data for major water-mineral reactions
and their limitations. In: Chemical Modeling in Aqueous Systems II (D.C. Melchior and
R.L. Bassett, eds.), American Chemical Society Symposium Series 416, Chapter 31 pp.
398-413.
Palmer, F.B, C.A. Butler, M. H. Timperley, and C. W. Evans, 1998. Toxicity to embryo and
adult zebrafish of copper complexes with two malonic acids as models for dissolved
organic matter, Environmental Toxicology and Chemistry, 17(8): 1538-1545.
Payne, T.E. and T.D. Waite 1990, Surface complexation modeling of uranium sorption data
obtained by Isotope exchange techniques, Radiochim. Acta,
Robins, R.G., 1990. The stability and solubility of ferric arsenate: an update. Paper presented at
th q Metallurgical Society Annual Meeting, Feb. 18-22, 1990, Anaheim, CA.
Routh, J. and M. Ikramuddin, 1996. Trace element geochemistry of Onion Creek near Van Stone
lead-zinc min (Washington, USA) - chemical analysis and geochemical modeling,
Chemical Geology, 133:211-224.
Saunders, J. A. and L.E. Toran, 1995. Modeling of radionuclide and heavy metal sorption around
low- and high-pH waste disposal sites at Oak ridge, Tennessee, Applied Geochemistry,
10:673-684.
Sehmel, G.A., 1989. Cyanide and Antimony Thermodynamic Database for the Aqueous Species
and Solids for the EPA-MINTEQ Geochemical Code, prepared by Battelle Pacific
Northwest Laboratory for the U.S. Environmental Protection Agency, Athens, Georgia.
E-48
-------
Appendix E
MINTEQA2 Verification and Validation
Serkiz, S.M., J.D. Allison, E.M. Perdue, H.E. Allen, and D.S. Brown, 1996. Correcting errors in
the thermodynamic database for the equilibrium speciation model MINTEQA2. Water
Research, 3 0(8): 193 0-193 3.
Stevenson, F.J. and A. Fitch, 1986. Chemistry of complexation of metal ions with soil solution
organics, in Interactions of Soil Minerals with Natural Organics and Microbes, P.M.
Huang and M. Schnitzer (eds.), Soil Science Society of America.
Stollenwerk, K.G., 1994. Geochemical interactions between constituents in acidic groundwater
and alluvium in an aquifer near Globe, Arizona, Applied Geochemistry, 9:353-369.
Stollenwerk, K.G., 1995. Modeling the effects of variable groundwater chemistry on the
adsorption of molybdate, Water Resources Research, 31(2):347-357.
Stollenwerk, K.G., 1996. Simulations of reactions affecting transport of constituents in the acidic
plume, Pinal Creek basin, Arizona, U.S. Geological Survey Water-Supply Paper 2466,
pp. 21-49.
Susetyo, W., L.A. Carreira, L.V. Azarraga, and D.M. Grimm, 1991. Fluorescence techniques for
metal-humic interactions. Fresenius Jour. Analytical Chemistry, 339:624-635.
Tipping, E. ,1994. WHAM— A chemical equilibrium model and computer code for waters,
sediments, and soils incorporating a discrete site/electrostatic model of ion-binding by
humic substances, Computers & Geosciences, 20: 973-1023.
USEPA, 1996. Background Document for Metals; EPA Composite Model for Leachate
Migration with Transformation Products (EPACMTP), Office of Solid Waste, U.S.
Environmental Protection Agency, Washington, DC.
USEPA, 1998. MINTEQA2/PRODEFA2, A Geochemical Assessment Model for Environmental
Systems: User Manual Supplement for Version 4.0, prepared by HydroGeoLogic, Inc.
and Allison Geoscience Consultants, Inc. for the U.S. Environmental Protection Agency,
Athens, GA.
USEPA, 1999. Changes in the MINTEQA2 Modeling Procedure for Estimating Metal Partition
Coefficients in Groundwater, prepared by HydroGeoLogic, Inc. for the U.S.
Environmental Protection Agency, Washington, DC.
Webster, J.G . and K.S. Webster, 1995. Arsenic adsorption from geothermal water, Sixteenth
AnnualPNOC Geothermal Workshop Proceedings, Manila, Philippines, pp. 35-42.
Westall, J.C., J.L. Zachary, and F.M.M. Morel, 1976. MINEQL: A Computer Program for the
Calculation of Chemical Equiloibrium Composition of Aqueous Systems, Technical Note
18, Ralph M. Parsons Laboratory, Dept. of Civil Engineering, Massachusetts Institute of
Technology, Cambridge, MA.
E-49
-------
Appendix E
MINTEQA2 Verification and Validation
Westall, J.C., 1979. MICROQL. II. Computation of Adsorption Equilibria in BASIC, Swiss
Federal Institute of Technology, EAWAG, Deubendorf, Switzerland.
Woodfine, D.G., R. Seth, D. Mackay, and M, Havas, 2000. Simulating the response of metal
contaminated lakes to reductions in atmospheric loading using a modified QWASI
model, Chemosphere, 41:1377-1388.
Yu, J.Y.,1996. Precipitation of Fe and A1 compounds from the acid mine waters in Dogyae area,
Korea: A qualitative measure of equilibrium modeling applicability and neutralization
capacity?, Aquatic Geochemistry, 2: 81-105
Zachara, J.M., D.C. Girvin, R.L. Schmidt, and C.T. Resch, 1987. Chromate adsorption on
amorphous iron oxyhydroxide in the presence of major groundwater ions, Environental
Science & Technology, 21(6):589-594.
Zachara, J.M., R.L. Schmidt, and E.A. Jenne, 1988. Feasibility of Field Testing the MINTEQ
Geochemical Code, prepared for the U.S. Environmental Protection Agency by Battelle
Pacific Northwest Laboratories, Richland, WA.
E-50
-------
Appendix F
Professional Publications and Presentations:
Multimedia, Multipathway, and Multireceptor Risk
Assessment (3MRA) Modeling System
-------
This page intentionally left blank.
-------
Appendix F
Professional Publications and Presentations
Appendix F
Professional Publications and Presentations:
Multimedia, Multipathway, and Multireceptor Risk Assessment (3MRA)
Modeling System
Investigating uncertainty and sensitivity in integrated, multimedia environmental
model: Tools for FRAMES-3MRA. J.E. Babendreier, US Environmental protection Agency,
Athens, GA and K.J. Castleton, Battelle Pacific Lab, Richland, WA, in publication, Journal of
Environmental Modeling and Software, 2003.
The 3MRA Risk Assessment Framework. A flexible approach for performing
multimedia, multipathway, and multireceptor risk assessments under uncertainty. C.M. Marin,
Ambiotec Group, Inc.; V. Guvanasen, HGL, Inc, Herndon, VA; Appearing as the lead paper in:
International Journal of Human and Ecological Risk Assessment, December, 2003.
An overview of USEPA's integrated multimedia, multipathway, and multireceptor
exposure and risk assessment tool. The 3MRA Model. B. Johnson, S. Kroner, D. Cozzie,
and Z. Saleem, US Environmental Protection Agency, Washington, D.C. Published in
Brownfield Sites: Assessment, Rehabilitation and Development; edited by C.A. Brebbia,
D. Almorza, H. Klapperich. Wessex Institute of Technology, Southampton, UK, 2002.
The Role of Uncertainty of Vadose Zone Flow and Transport in Multimedia,
Multireceptor, and Multipathway Risk Assessment. Varut Guvanasen, Carlos Marin,
Theodore P. Lillys, and Zubair A. Saleem, American Geophysical Union, Annual Meeting,
Washington, DC, May 30, 2002 .
Society for Risk Analysis (SRA): Symposium on Multimedia, Multipathway, and
Multireceptor Risk Assessment for Identification of Hazardous Wastes, Atlanta, GA,
December 5- 8, 1999.
Session 1
Johnson, W. Barnes, US EPA, 401 M. Street, SW, Washington, DC 20460. AN OVERVIEW
OF EPA'S RISK ASSESSMENT FOR IDENTIFICATION OF HAZARDOUS
WASTES.
F-3
-------
Appendix F
Professional Publications and Presentations
Saleem. Z.A.. US EPA, 401 M Street, Washington, D.C. 20460; Marin, C.M., Ambiotec,
Harlingen, TX; and Guvanasen, V., HydroGeoLogic, Herndon, VA. AN OVERVIEW
OF THE MULTIMEDIA, MULTIPATHWAY, AND MULTIRECEPTOR RISK
ASSESSMENT (3MRA) FRAMEWORK FOR HWIR.
Laniak. G. F.. US EPA, Athens, GA 30613; Castleton K. J., and Whelan, G., Pacific Northwest
National Laboratory, Richland, WA 99352. AN OVERVIEW OF A NATIONAL
MULTIMEDIA, MULTIPATHWAY AND MULTIRECEPTOR RISK
ASSESSMENT TECHNOLOGY DEVELOPMENT.
Little. KW. and Coburn, JB, Research Triangle Institute, PO Box 12194, Research Triangle Park,
NC 27709; Labieniec, P, Labieniec Consulting, 10701 SpringRun Road, Chesterfield, VA
23832; Saleem, Z. and Cozzie, D. US Environmental Protection Agency, Office of Solid
Waste, 401 M Street, SW (MD-5307), Washington, DC 20460; Guavanasen, D,
HydroGeoLogic, Inc., 1155 Herndon Parkway, Suite 900, Herndon, VA 20170.
SIMULATING MULTIMEDIA RELEASES FROM WASTE MANAGEMENT
UNITS FOR HWIR.
Session 2
Saleem. Z.A.. US EPA, 401 M Street, SW, Washington, DC 20460; Ambrose, R.B., US EPA,
Athens, Georgia 30613; Schwede, D.B., National Oceanic and Atmospheric
Administration, RTP, NC, Little, K.L., Research Triangle Institute, RTP, NC 27709;
Guvanasen, D., and Lillys, T.P., HydroGeoLogic, Inc., Herndon, VA 20170.
SIMULATING INTEGRATED MULTIMEDIA CHEMICAL FATE AND
TRANSPORT FOR NATIONAL RISK ASSESSMENTS.
Pierson. TK. Little, KW, Lutes, AC, Research Triangle Institute, PO Box 12194, Research
Triangle Park, NC 27709; Kroner, SM, US Environmental Protection Agency, Office of
Solid Waste, 401 M Street, SW (5307W), Washington, DC 20460; Tohmaz, A, 109
Norcross Place, Cary, NC 27513. HUMAN EXPOSURE AND POPULATION RISK
MODELS FOR IMPLEMENTING 3MRA.
Beaulieu. SM. McLean, JS, Conrad, GT, Research Triangle Institute, PO Box 12194, Research
Triangle Park, NC 27709; and Cozzie, DA, US Environmental Protection Agency, 401 M
Street, SW (MD-5307), Washington, DC 20460. PROPOSED METHODOLOGY TO
ASSESS ECOLOGICAL EXPOSURE AND RISK FOR 3MRA.
Truesdale. RS. Conrad, GT, Research Triangle Institute, PO Box 12194, Research Triangle Park,
NC 27709; and Kroner, SM, US Environmental Protection Agency, Office of Solid
Waste, 401 M Street, SW (5307W), Washington, DC 20460. STRATEGIES FOR
F-4
-------
Appendix F
Professional Publications and Presentations
NATIONAL SCALE DATA COLLECTION: EXTANT SOURCES AND SPATIAL
(GIS) DATA PROCESSING.
Society for Environmental Toxicology and Chemistry (SETAC): Interactive Poster Session
on Modeling for National Risk Assessment. Philadelphia, PA. 14 -18, November, 1999.
Simulation of Surface Impoundments in National Risk Assessments. Guvanasen, V.,
Hy droGeologic, Inc., Herndon, VA; Coburn, J.B. Research Triangle Institute, RTP, N.C.; and
Saleem, Z., U.S. Environmental Protection Agency, Washington, D.C.
Simulating Emissions of Hazardous Constituents from Non-wastewater Waste
Management Units. Labieniec, P.A.*, Labieniec Consulting Services, Chesterfield, VA., Little,
K.W., Lawless, P.A., Research Triangle Institute, RTP, NC.
Simulating Atmospheric Exposure in a National Risk Assessment Using an Innovative
Meteorological Sampling Scheme. Schwede, D.B., National Oceanic and Atmospheric
Administration, Research Triangle Park, N.C., Brode, R.W. and Jindal, M., Pacific
Application of Exams as the Surface Water Module in the HWIR Multimedia Risk
Assessment System. Ambrose, Robert B., Jr., P.E.*, U.S. EPA, Athens, GA and Burns,
Lawrence A., Ph.D., U.S. EPA, Athens, GA
Simulating Dynamic Response of Regional Watersheds to Emissions from Waste
Management Units. Little, K.W.*, Research Triangle Institute, RTP, NC, Labieniec, P.A.,
Labieniec
Version 4.0 of U.S. EPA's Geochemical Speciation Model MINTEQA2 for Use in National
Risk Assessments. Allison, J.D., Allison, T.L., Brown, D.S.*, and Ambrose, R.B.,
HydroGeoLogic, Herndon, VA; Allison Geoscience Consultants, Flowery Branch, GA; and U.S.
EPA, Athens, GA.
Effects of Heterogeneity of porous Media on Fate and Transport Modeling for National
Risk Assessment. Guvanasen, V., Kim, J., Hy droGeoLogic, Inc., Herndon, VA, Saleem, Z.A.,
U.S.EPA, Washington, D.C., Schmelling. S., U.S.EPA, Ada, OK, Lee, S., Chen, J-S., Dynamac
Corp., Ada. OK.
Proposed Approach for Development of Ecological Benchmarks for a National-Scale Risk
Assessment. Beaulieu, S.M.*, Research Triangle Institute, RTP, NC, Spencer, M., University
of Haifa, Haifa, Israel, Harmon, A. B., Research Triangle Institute, RTP, NC, and Cozzie, D.A.,
U.S. EPA, Washington, DC.
F-5
-------
Appendix F
Professional Publications and Presentations
Application of GIS Techniques and Other Data Sources for National Risk Assessments:
Part 2. "Result (Interactive Demonstration)". Conrad, G. T*., Truesdale, R. S., Andrews,
L.S., Rickman, E. E., Andrews. M., Research Triangle Institute, RTP, NC.
Society for Environmental Toxicology and Chemistry (SETAC). 20th Annual meeting,
Philadelphia, PA. 14 - 18, 1999. Platform Sessions on Multimedia, Multipathway, and
Multireceptor Risk Assessment:
Session 1
A Framework For a National Multimedia, Multipathway And Multireceptor Risk
Assessment (3MRA) For Identification of Hazardous Wastes. Marin, C M ., Ambiotec,
Harlingen, TX; Guvanasen, V., HydroGeoLogic, Herndon, VA; and Saleem, Z.A., USEPA,
Washington, D.C.
Successful Design and Implementation of a National Multimedia, Multipathway and
Multireceptor Risk Assessment (3MRA) Software System for the Identification of
Hazardous Wastes. Karl J. Castleton, Pacific Northwest National Laboratory, Richland, WA
99352; Gerard F. Laniak, US Environmental Protection Agency, Ecosystems Research Division,
Athens, GA 30605 and Gene Whelan, Ph.D., Pacific Northwest National Laboratory, Richland,
WA 99352.
Application of GIS Techniques and Other Spatial Data Sources for National Risk
Assessments: Part 1, Data Sources and methodology. Robert S. Truesdale, Gerald T. Conrad,
E.E. Rickman, and Linda S. Andrews. Research Triangle Institute, Research Triangle Park, N.C.
20771.
A Pro posed Approach for Conducting National-Scale Ecological Risk Assessments for the
Identification of Hazardous Wastes. Stephen M. Beaulieu, Joanie McLean and Gerald T.
Conrad, Research Triangle Institute, Research Triangle Park, N.C.; and David A. Cozzie, US
Environmental Protection Agency, Washington, DC 20460.
Session 2
SPARC Generated Chemical Properties Database for Use in National Risk Assessments.
J. MacArthur Long Ph. D., Samuel W. Karickhoff, Ph. D., Eric J. Weber, Ph.D., U.S.
Environmental Protection Agency, National Exposure Research Laboratory, Athens, GA 30605.
Development of Human Exposure Factor Distributions for Use in a National Risk
Assessment. L. E. Meyers, Ph.D., Research Triangle Institute, Research Triangle Park, N.C.
F-6
-------
Appendix F
Professional Publications and Presentations
20771, J. Moya, U.S. Environmental Protection Agency, Washington, D.C., and R. W.
Whitmore, Research Triangle Institute, Research Triangle Park, N.C. 20771.
Proposed Methodology Used to Estimate Contaminant Transfer Through Aquatic Food
Webs for a National Risk Assessment. S. M. Beaulieu, K.W. Little, A.B. Harmon, and T.K.
Pierson, Research Triangle Institute, N.C. 20771 and S. M. Kroner, U.S. Environmental
Protection Agency, Washington, DC 20460.
Variability and Uncertainty Analyses for National Probabilistic Risk Assessments
C.M. Marin, AMBIOTEC, Inc., V. Guvanasen, HydroGeoLogic, Inc. and Z. A. Saleem, U.S.
Environmental Protection Agency, Washington, DC 20460.
OTHER PRESENTATIONS TO PUBLIC FOR INFORMATION EXCHANGE
Hazardous Waste Identification Rule; Identification and Listing of Hazardous
Wastes. Application of the Multimedia, Multipathway, and Multireceptor Risk Assessment
(3MRA) Model to 36 chemicals for national regulations. Federal Register Notice, Vol 65, No.
135, pp. 44491-444506, July 18, 2000.
National RCRA Program Meeting, Hyatt Regency Capitol Hill Hilton, Washington, DC.
Multimedia, Multipathway, and Multireceptor Risk Assessment (3MRA) Workshop. August
17, 2000.
F-7
------- |