vvEPA
United States
Environmental Protection
Agency
Office of Water
Office of Science and Technology
4304
EPA-822-R-03-030
December 2003
Methodology for Deriving Ambient
Water Quality Criteria for the Protection
of Human Health (2000)
Technical Support Document
Volume 2: Development of National
Bioaccumulation Factors
-------
EPA-822-R-03-030
Methodology for Deriving
Ambient Water Quality Criteria
for the Protection of
Human Health (2000)
Technical Support Document
Volume 2:
Development of
National Bioaccumulation Factors
Final
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC 20460
-------
NOTICE
The policies and procedures set forth in this document are intended solely to describe
EPA methods and guidance for developing or revising ambient water quality criteria to protect
human health, pursuant to Section 304(a) of the Clean Water Act, and to serve as guidance to
States and authorized Tribes for developing their own water quality criteria. This guidance does
not substitute for the Clean Water Act or EPA's regulations, nor is it a regulation itself. Thus, it
does not impose legally binding requirements on EPA, States, Tribes, or the regulated
community, and may not apply to a particular situation depending on the circumstances.
This document has been reviewed in accordance with U.S. EPA policy and approved for
publication. Mention of trade names or commercial products does not constitute an endorsement
or recommendation for use.
-------
FOREWORD
In 2000, the U. S. Environmental Protection Agency (EPA) published the Methodology
for Deriving Ambient Water Quality Criteria for the Protection of Human Health (2000) ("2000
Human Health Methodology"), updating and revising the existing 1980 Guidelines and
Methodology. The 2000 Human Health Methodology includes guidance on chemical risk
assessment, exposure, and bioaccumulation. The process EPA followed in developing the 2000
Human Health Methodology included gathering information from multiple stakeholders,
convening a national issues workshop, securing EPA Science Advisory Board review and public
review and comment period on the draft Human Health Methodology. A more detailed
chronology can be found in the Federal Register (65FR66444).
As part of the 2000 Human Health Methodology, EPA developed detailed procedures and
guidelines for estimating bioaccumulation factor (BAF) values for use in deriving or revising
ambient water quality criteria. This Technical Support Document Volume 2: Development of
National Bioaccumulation Factors discusses the technical basis for developing national BAFs,
the underlying assumptions and uncertainties inherent to the approach, and applying the
bioaccumulation component of the 2000 Human Health Methodology. The scientific
approaches, assumptions and science policy decisions included in this document have been peer-
reviewed as part of the comprehensive review of the 2000 Human Health Methodology, Detailed
information about this peer review process can be found an EPA's website
(www. epa. gov/waterscience).
EPA will use this technical support document to develop new ambient water quality
criteria and to revise existing recommended water quality criteria. This technical support
document will not be used alone to derive bioaccumulation factors, but rather in conjunction with
the earlier Methodology for Deriving Ambient Water Quality Criteria for the Protection of
Human Health (2000).
,
GeoHrey H.'
Director
Office of Science and Technology
in
-------
ACKNOWLEDGMENTS
Project Leader
Erik L. Winchester
Tala R. Henry
Coauthors
Lawrence P. Burkhard
Philip M. Cook
Keith G. Sappington
Erik L. Winchester
U.S. EPA Human Health
William Beckwith
JeffBigler
Sally Brough
Karen Clark
Gregory Currey
Vicki Dellarco
Charles Delos
Arnold Den
Catherine Eiden
Michael Firestone
Steven Galson
Sue Gilbertson
Joel Hansel
Wayne Jackson
William Jordan
Margaret Kelly
Sharon Lin
Roseanne Lorenzana
Gregory McCabe
Bruce Mintz
William Morrow
Jacqueline Moya
Deirdre Murphy
Joseph Nabholz
Russell Nelson
Jennifer Orme-Zavaleta
Lynn Papa
Robert Pepin
David Pfeifer
U.S. EPA Office of Science and Technology
U.S. EPA Office of Science and Technology
U.S. EPA National Health and Environmental Effects Research
Laboratory
U.S. EPA National Health and Environmental Effects Research
Laboratory
U.S. EPA National Center for Environmental Assessment
U.S. EPA Office of Science and Technology
Methodology Workgroup
U.S. EPA Region 1
U.S. EPA Office of Science and Technology
U.S. EPA Region 10
U.S. EPA Office of General Counsel
U.S. EPA Office of Waste Management
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Office of Science and Technology
U.S. EPA Region 9
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Office of Science and Technology
U.S. EPA Region 4
U.S. EPA Region 2
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Office of Children's Health Protection
U.S. EPA Office of Wetlands, Oceans, and Watersheds
U.S. EPA Region 10
U.S. EPA Region 7
U.S. EPA National Environmental Research Laboratory-RTF
U.S. EPA Office of Science and Technology
U.S. EPA National Center for Environmental Assessment
U.S. EPA Air Quality Planning and Standards
U.S. EPA Office of Prevention, Pesticides, and Toxic Substances
U.S. EPA Region 6
U.S. EPA National Health and Environmental Effects Research
Laboratory
U.S. EPA National Center for Environmental Assessment
U.S. EPA Region 5
U.S. EPA Region 5
IV
-------
Rita Schoeny
Charles Stephan
David Tomey
Fritz Wagener
Jennnifer Wigal
Jeanette Wiltse
Gary Wolinsky
Philip Woods
William Wuerthele
U.S. EPA Office of Science and Technology
U.S. EPA Office of Research and Development
U.S. EPA Region 1
U.S. EPA Region 4
U.S. EPA Office of Science and Technology
U.S. EPA Office of Science and Technology
U.S. EPA Region 9
U.S. EPA Region 9
U.S. EPA Region 8
U.S. EPA Technical Reviewers
Sidney Abel
Charlotte Bertrand
James Breithaupt
Stephanie Irene
Patricia Jennings
Ann Johnson
Michael Kravitz
Matthew Lorber
Deirdre Murphy
Lucy Shanaman
U.S
U.S
U.S
U.S
U.S
U.S
U.S
U.S
U.S
U.S
EPA Office of Pesticide Programs
EPA Office of Policy, Economics and Innovation
EPA Office of Pesticide Programs
EPA Office of Pesticide Programs
EPA Office of Pesticide Programs
EPA Office of Policy, Economics and Innovation
EPA National Center for Environmental Assessment
EPA National Center for Environmental Assessment
EPA Office of Air Quality Planning and Standards
EPA Office of Pesticide Programs
v
-------
EXTERNAL
PEER REVIEW WORKGROUP
The following professionals were part of the External Peer Review
Workgroup-Bioaccumulation Workgroup, that provided technical and scientific review regarding
the content and technical approach in the July 1998 Draft Ambient Water Quality Criteria
Derivation Methodology: Human Health. Their comments were reviewed and incorporated
where appropriate to develop this final document.
Damian Shea, Ph.D. NC State University
Brendan Hickie, Ph.D. Trent University
Robert Hale, Ph.D. Virginia Institute of Marine Science
Karen Erstfeld, Ph.D. Rutgers University
Lynn McCarty, Ph.D. LS McCarty Scientific Research and Consulting
Potential areas for conflict of interest were investigated via direct inquiry with the peer
reviewers and review of their current affiliations. No conflicts of interest were identified.
VI
-------
TABLE OF CONTENTS
Notice ii
Foreword iii
Acknowledgments iv
External Peer Review Workgroup vi
1. Introduction 1-1
1.1 Purpose 1-2
1.2 Scope 1-2
1.3 Important Bioaccumulation and Bioconcentration Concepts 1-3
2. Definitions 2-1
3. Overview of the National BAF Methodology 3-1
3.1 Summary of Four Bioaccumulation Methods 3-3
3.1.1 Field-Measured BAFs 3-3
3.1.2 BAFs Predicted from a Field-Measured BSAF 3-3
3.1.3 BAF Predicted from Laboratory-Measured BCFs 3-4
3.1.4 BAFs Predicted from Kow 3-4
3.1.5 Advantages and Limitations of BAF Methods 3-5
3.2 Framework for Deriving National BAFs 3-5
3.2.1 BAF Derivation Procedures for Inorganic and Organometallic Chemicals .... 3-7
3.2.2 BAF Derivation Procedures for Ionic Organic Chemicals 3-7
3.2.3 BAF Derivation Procedures for Nonionic Organic Chemicals 3-8
4. Background Information on Lipid Normalization, Bioavailability, and
Biomagnification 4-1
4.1 Lipid Normalization 4-1
4.1.1. Background and Theory 4-1
4.1.2 Assumptions and Limitations 4-2
4.2 Technical Basis of Freely Dissolved Normalization of Chemical Concentration
in Water 4-5
4.2.1 Background Theory and Basic Equation 4-5
4.2.2 Effects of Chemical Sorption to DOC and POC on Chemical Bioavailability . . 4-8
4.2.3 Sorptive Behavior of Nonionic Organic Chemicals with DOC and POC
(Including Plankton) in the Water Column 4-10
4.2.4 Values for the Particulate and Dissolved Organic Carbon Partition
Coefficients Kpoc and Kdoc 4-12
4.2.5 Select!on of Appropriate Kows for Partitioning (Bioavailability) Predictions .. 4-20
4.3 Importance of Sediment-Water Concentration Quotient (• socw) 4-21
vii
-------
4.4 Derivation and Use of Food Chain Multipliers 4-24
4.4.1 Derivation of FCMs Using a Food Web Model 4-25
4.4.2 Derivation of Food Chain Multipliers Using Field Data 4-40
4.4.3 Food Chain Multiplier Uncertainties 4-41
5. Calculating Baseline BAFs for Nonionic Organic Chemicals Using
the Four Methods 5-1
5.1 Method 1: Deriving Baseline BAFs from Total BAFs (BAF^s) 5-1
5.1.1 Sampling and Data Quality Considerations 5-2
5.1.2 Assumptions and Limitations 5-4
5.1.3 Validation of Method 1 5-5
5.2 Method 2: Deriving Baseline BAFs from BSAFs 5-8
5.2.1 Determination of BSAF Values 5-10
5.2.2 Relationship of Baseline BAFs to BSAFs 5-11
5.2.3 Derivation of the Baseline BAF Equation for Method 2 5-12
5.2.4 Sampling and Data Quality Considerations 5-13
5.2.5 Assumptions and Limitations 5-14
5.2.6 Validation of Method 2 5-15
5.3 Method 3: Deriving Baseline BAFs from Laboratory-Measured BCFs (BCFxs)
and FCMs 5-19
5.3.1 Sampling and Data Quality Considerations 5-19
5.3.2 Assumptions and Limitations 5-21
5.3.3 Validation of Method 3 5-22
5.4 Method 4: Baseline BAF Derived from Kow x FCM 5-23
5.4.1 Assumptions and Limitations 5-23
5.4.2 Validation of Method 4 5-24
6. Derivation of National BAFs for Nonionic Organic Chemicals 6-1
6.1 Selecting Final Baseline BAFs 6-2
6.1.1 Calculating Species-Mean Baseline BAFs for Each BAF Method 6-3
6.1.2 Calculating the Trophic Level-Mean Baseline BAF for Each BAF Method . . . 6-7
6.1.3 Selecting the Final Trophic Level-Mean Baseline BAFs 6-7
6.2 Basis for the National Default Lipid Fraction (f.) of Commonly Consumed Fish and
Shellfish 6-9
6.2.1 Variability in Lipid Content 6-9
6.2.2 Data Sources 6-10
6.2.3 Data Analysis 6-15
6.2.4 Uncertainty and Sensitivity Analysis 6-28
6.3 Basis for the National Default Values of DOC and POC 6-37
6.3.1 Data Sources 6-38
6.3.2 Data Retrieval and Screening 6-39
6.3.3 Results 6-41
6.3.4 Uncertainty/Limitations in National Default Values 6-42
Vlll
-------
7. Examples of BAF Calculations 7-1
7.1 Example 1: Calculation of a National BAF from a Field-Measured BAF (BAFj)
(Method 1) 7-1
7.1.1 Calculating a Total BAF (BAF^) 7-1
7.1.2 Calculating a Baseline BAF 7-1
7.1.3 Calculating a National BAF 7-3
7.2 Example 2: Calculation of a National BAF from Field-Measured BSAFs
(Method 2) 7-4
7.2.1 Calculating a Field-Measured BSAF 7-4
7.2.2 Determining a Sediment-Water Column Concentration Quotient (• socw) 7-5
7.2.3 Calculating a Baseline BAF 7-5
7.2.4 Calculating a National BAF 7-6
7.3 Example 3: Calculation of a National BAF for Chemical i from BCF^ x FCM
(Method 3) 7-7
7.3.1 Calculating a Laboratory-Measured ECF{ 7-7
7.3.2 Calculating a Baseline BAF 7-7
7.3.3 Calculating a National BAF 7-8
7.4 Example 4: Calculation of a National BAF for Chemical i from Kow x FCM
(Method 4) 7-10
7.4.1 Selecting aKow and FCM 7-10
7.4.2 Calculating a Baseline BAF 7-10
7.4.3 Calculating a National BAF 7-10
8. References 8-1
Appendix A A-l
Appendix B B-l
IX
-------
TABLES
Table 3-1. Strengths and Limitations of the Four BAF Methods for Deriving National
BAFs 3-5
Table 4-1. Kpoc Data 4-15
Table 4-2. Regression Equations for Dependence of Kdoc (Geometric Means)
on Kow 4-18
Table 4-3. Geometric Mean Regression Equations (log • socw = A • log Kow + B) for
Polychlorinated Biphenyls (PCBs) and Chlorinated Pesticides 4-30
Table 4-4. Average • SOCw/K0w Ratios for Three Different Ecosystems 4-33
Table 4-5. Food Web Structure for National BAF Methodology (Flint, 1986;
Gobas, 1993) 4-37
Table 4-6. Food-Chain Multipliers for Trophic Levels 2, 3, and 4 4-39
Table 4-7. Environmental Parameters and Conditions Used for Determining FCMs
for the National BAF Methodology 4-39
Table 5-1. BAF^s and Baseline BAFs Exceedance Limit Ratios for Green Bay
(All Zones Combined) 5-7
Table 5-2. Validation Statistics for Method 2: Ratio of Baseline BAFpredicted/Baseline
BAFmeasured 5-17
Table 5-3. Exceedance Levels for Ratio of Method 2-Predicted Baseline BAFs
(Geometric Mean) to BAF^s from Green Bay and Hudson River 5-17
Table 5-4. Validation Statistics for Method 4: Ratio of Baseline
BAFpredicted/Baseline BAFmeasured 5-25
Table 5-5. Summary Statistics: Differences Between Log Baseline BAFs Predicted
with Method 4 and Log Baseline BAFs Measured from Lake Ontario
(Oliver and Niimi, 1988) for Chemicals with Log Kows Exceeding 4 5-26
Table 6-1. Data Preference Hierarchy for Selecting Final Baseline BAFs for Nonionic
Organic Chemicals 6-8
Table 6-2. Categories and Mean Per Capita Consumption Rates from the USDA
CSFII 6-12
-------
Table 6-3. Lipid Content of Aquatic Organisms Used to Derive National Default
Values of Lipid Fraction (f.) 6-18
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII
Consumption Categories 6-22
Table 6-5. National Default Values of Lipid Fraction 6-28
Table 6-6. Calculation of National Default Values of Consumption-Weighted Mean
Lipid Fraction 6-29
Table 6-7. Sensitivity of National Default Values of Lipid Fraction to Different
Weighting Assumptions Among Species 6-31
Table 6-8. Sensitivity of National Default Values of Lipid Fraction to Different
Weighting Assumptions Among Trophic Levels 6-33
Table 6-9. Descriptive Statistics of Monte Carlo Simulation of National Default
Values of Lipid Fraction (10,000 Iterations) 6-37
Table 6-10. National Default Values for POC and DOC in U.S. Fresh and Estuarine
Surface Waters 6-42
Table 6-11. Descriptive Statistics from the State-Level DOC Distributions 6-46
Table 6-12. Descriptive Statistics from the Ecoregion-Level DOC Distributions 6-47
Table 6-13. Effect of DOC and POC Concentrations on the Freely Dissolved
Fraction (ffd) Relative to National Default Values of DOC and POC 6-50
XI
-------
FIGURES
Figure 1-1. A. Daily concentrations of a hypothetical nonionic organic chemical overtime
in the water column, predicted using a simple dilution model and daily flow
data for the Mississippi River at St. Paul, Minnesota 1-5
Figure 1-1. B. Daily chemical concentrations in piscivorous fish found using the kinetic
food web models of Gobas (1993) with the daily chemical concentrations in
the water column for nonionic organic chemicals with log w-octanol-water
partition coefficients (log Kows) of 2, 3, ... and 9 1-5
Figure 3-1. Framework for selection of methods for deriving national BAFs 3-2
Figure 3-2. BAF derivation for nonionic organic chemicals 3-10
Figure 4-1. Kpocs determined using the reverse-phase, sparging, ultrafiltration, and
model derived techniques 4-16
Figure 4-2. Kdocs determined using the reverse phase, equilibrium dialysis, sparging,
calculated/model derived, fluorescence quenching, solubility enhancement,
biological, and solid-phase microextraction techniques for DOCs from
different sources 4-17
Figure 4-3. Average Kdocs for individual chemicals for different DOC sources: humic
and fulvic acids, sediment porewaters, soil porewaters and groundwaters,
and surface waters 4-19
Figure 4-4. The residuals between measured log Kdocs and log Kdocs predicted using
relationship of Kdoc = 0.08 Kow. DOC sources: humic and fulvic acids
without Aldrich humic acid, sediment porewaters, soil porewaters and
groundwaters, and surface waters 4-20
Figure 4-5. The sediment-water chemical concentration quotient (• socw) for three
different chemical loading scenarios 4-23
Figure 4-6. Measured baseline BAFs for PCBs, chlorinated pesticides, and chlorinated
benzenes, chlorinated toluenes, and hexachlorobutadiene from the data of
Oliver and Niimi (1988) and BAF?s predicted using the Gobas and
Thomann models plotted against Kow for all organisms 4-27
Figure 4-7. Sediment-water column concentration quotient (• socw) for PCBs in five
different geographical zones in Green Bay, Lake Michigan 4-31
Figure 4-8. Average sediment-water column concentration quotients (• socw) for
individual PCB congeners across the five different geographical zones
in Green Bay, Lake Michigan 4-32
xn
-------
Figure 4-9. Sediment-water column concentration quotient (• socw) for PCBs at
river miles 189 and 194 4-32
Figure 4-10. The ratio of the 90th to 10th percentile baseline BAF predictions for
piscivorous fish from 100,000 Monte Carlo simulations using the Gobas
model as a function of w-octanol-water partition coefficient (Kow) 4-43
Figure 4-11. FCMs for purely pelagic and purely benthic food webs derived by
modifying the Lake Ontario food web 4-43
Figure 4-12. FCMs predicted using the Lake Ontario food web with disequilibriums
of 11.5, 23, and 46 4-43
Figure 5-1. BAF^s and Baseline BAFs for PCB congener 149 (2,2',3,4',5',6-
hexachlorobiphenyl) (±1 sd) for adult alewife for different Green
Bay zones 5-6
Figure 5-2. Box plots comparing baseline and field-measured BAFs for six PCB
congeners from Green Bay, Lake Ontario, and Hudson River ecosystems
for 13 fish species with samples segregated according to year classes and
sampling location 5-9
Figure 5-3. Relationship between measured and predicted baseline BAFs for
method 3 5-22
Figure 6-1. A schematic illustrating the aggregation of baseline BAF data by species,
trophic level, and BAF method type for nonionic organic chemicals under
Procedure 1 6-3
Figure 6-2. Frequency distribution of national default values of lipid fraction (10,000
iterations) 6-36
Figure 6-3. Reverse cumulative comparison of national default values of lipid fraction
(10,000 iterations) 6-36
Figure 6-4. Ecoregion-level DOC distributions for rivers and streams from EPA's
WATSTORE/LDC and EMAP databases 6-45
Figure 6-5. Ecoregion-level DOC distributions for rivers and streams from EPA's
WATSTORE/LDC and EMAP databases 6-46
Figure 6-6. Effect of DOC, POC, and Kow on the freely dissolved fraction (ffd) 6-50
xni
-------
1. INTRODUCTION
In 2000, the U.S. Environmental Protection Agency (EPA) published the Methodology
for Deriving Ambient Water Quality Criteria for the Protection of Human Health (USEPA,
2000a). That document (referred to here as the 2000 Human Health Methodology) presents
technical guidance and the steps that EPA will follow for deriving new and revised national
recommended ambient water quality criteria (AWQCs) for the protection of human health under
Section 304(a) of the Clean Water Act. The 2000 Human Health Methodology includes guidance
on chemical risk assessment, exposure, and bioaccumulation. To supplement the 2000 Human
Health Methodology, EPA is developing series of Technical Support Documents (TSD) on Risk
Assessment, Exposure Assessment, and Bioaccumulation. The first volume, (Volume 1: Risk
Assessment; EPA-822-B-00-005), was published with the 2000 Methodology in October 2000.
This volume (Volume 2) of the Technical Support Document (TSD) focuses on the technical
components of the 2000 Human Health Methodology that pertain to the assessment of chemical
bioaccumulation.
The 2000 Human Health Methodology incorporates a number of scientific advancements
made over the past two decades. One of these advancements is in the assessment of chemical
exposure to humans through the aquatic food web pathway. For certain chemicals, exposure via
the aquatic food web is more important than exposure from ingestion of water. Such chemicals
tend to be highly hydrophobic, to partition in aquatic environments to surficial sediments, and to
accumulate in high concentrations in fish and shellfish through the process of bioaccumulation.
One method for incorporating chemical exposure to humans through the aquatic food web
involves estimating the amount of a chemical expected to bioaccumulate in fish and shellfish that
are commonly consumed by populations in the United States. Previously, EPA primarily used
bioconcentration factors (BCFs) to estimate chemical accumulation of waterborne chemicals by
aquatic organisms. The BCF reflects contaminant exposure and accumulation by fish and
shellfish only through the water column. Over the past two decades, however, science has shown
that all the routes (e.g., food, sediment, and water) by which fish and shellfish are exposed to
highly bioaccumulative chemicals may be important in determining the chemical accumulation in
the organism's body, and that these chemicals can be transferred to humans when they consume
contaminated fish and shellfish. The EPA's approach to estimating uptake into fish and shellfish
now emphasizes the use of a bioaccumulation factors (BAFs), which account for chemical
accumulation from all potential exposure routes.
The generalized ambient water quality criterion (AWQC) formula for noncancer effects is
shown below (Equation 1-1) as an example of how the BAFs are used in the calculation of a
recommended national AWQC for the protection of human health (USEPA, 2000a). In Equation
1-1, trophic-level specific BAFs are used in the denominator, along with information on the
amount offish consumed on a daily basis (FI) for each trophic level (i), to estimate human
exposure to contaminants through the aquatic food web.
1-1
-------
AWQC = RfD • RSC BW
4
DI + S
i=2
(Equation 1-1)
where:
RfD = reference dose for noncancer effects (mg/kg/day)
RSC = relative source contribution to account for nonwater sources of exposure
BW = human body weight (kg)
DI = drinking water intake (L/day)
FI = fish intake (kg/day) at trophic level i (i = 2, 3, 4)
BAF, = bioaccumulation factor (L/kg) at trophic level i (i = 2, 3, 4)
1.1 PURPOSE
This TSD volume:
Presents the technical basis for the EPA's approach to developing national BAFs
for the different trophic levels offish and shellfish commonly consumed by
humans,
Discusses the underlying assumptions and uncertainties inherent in the approach,
and
• Provides further detail on applying the BAF component of the 2000 Human
Health Methodology.
As indicated in Equation 1-1 of Section 1, the national, trophic level-specific BAFs for a
given contaminant are used by the EPA in the derivation of AWQC for the protection of human
health. A subsequent volume (Volume 3: Development of Site-Specific Bioaccumulation Factors)
provides guidance to States and authorized Tribes for developing site-specific BAFs for the
various trophic levels when BAFs that are more representative of local conditions are preferred.
Neither of the bioaccumulation TSDs should be used alone to derive BAFs, but rather in
conjunction with the 2000 Human Health Methodology. The intended audience for both of these
documents includes the EPA scientists who are responsible for deriving water quality criteria,
State and Tribal risk assessors and stakeholders interested in the technical basis of EPA's national
BAF methodology, and other users interested in bioaccumulation issues for other applications.
1.2 SCOPE
The goal of EPA's approach for developing national BAFs is to represent the long-term
average bioaccumulation potential of a pollutant in aquatic organisms that are commonly
consumed by humans throughout the United States. National BAFs are not intended to reflect
1-2
-------
fluctuations in bioaccumulation over short periods (e.g., a few days) because human health
AWQCs are generally designed to protect humans from long-term exposures (over a lifetime) to
waterborne chemicals.
National BAFs are also intended to account for some major chemical, biological, and
ecological attributes that can affect bioaccumulation in bodies of water across the United States.
For this reason, EPA's approach includes separate procedures for deriving national BAFs
according to the type of chemical (e.g., nonionic organic, ionic organic, inorganic, and
organometallic). For the purposes of the 2000 Human Health Methodology, nonionic organic
chemicals are defined as organic compounds that do not ionize substantially in natural bodies of
water. These chemicals are also referred to as "neutral" or "nonpolar" organics in the scientific
literature. Ionic organic chemicals are considered to include those chemicals that contain
functional groups with exchangeable protons, such as hydroxyl, carboxylic, and sulfonic and
nitrogen (pyridine) groups. Ionic organic chemicals undergo ionization in water, the extent of
which depends on the pH and the pKa of the water. Ionic chemicals are considered separately
when deriving national BAFs because the behavior of the anionic or cationic species of these
chemicals in aquatic systems is much different from those of their neutral (un-ionized)
counterparts. Inorganic and organometallic chemicals include inorganic minerals, other inorganic
compounds and elements, metals, metalloids, and organometallic compounds. This TSD
document focuses primarily on the procedures for determining BAFs for nonionic organic
chemicals thatbioaccumulate. The procedures for estimating bioaccumulation of nonionic
organic chemicals are generally better developed than those for ionic chemicals. Therefore, both
the conditions under which these procedures can be applied and the limitations associated with
their application warrant further explanation.
In addition, EPA's national BAFs are derived separately for each trophic level to account
for potential biomagnification of some chemicals in aquatic food webs and broad physiological
differences among organisms that may influence bioaccumulation. As discussed in Chapter 3,
lipid contents of aquatic organisms and the amounts of organic carbon in ambient waters affect
bioaccumulation of nonionic organic chemicals in aquatic food webs. National trophic-level
specific BAFs incorporate adjustments for the lipid content of commonly consumed fish and
shellfish and for the freely dissolved fraction of the chemical in ambient water by using
nationwide averages for these two parameters. Further discussion of these parameters is provided
in Section 4.
1.3 IMPORTANT BIOACCUMULATION AND BIOCONCENTRATION CONCEPTS
Several attributes of the bioaccumulation process are important to understanding the
approach used to develop national BAFs used in setting national recommended AWQCs for the
protection of human health. First, the term bioaccumulation refers to the uptake and retention of
a chemical by an aquatic organism from all surrounding media (e.g., water, food, sediment). The
term bioconcentration refers to the uptake and retention of a chemical by an aquatic organism
from water only. For some chemicals (particularly those that are highly persistent and
hydrophobic), the magnitude of bioaccumulation by aquatic organisms can be substantially
greater than the magnitude of bioconcentration. For such chemicals, an assessment of
1-3
-------
bioconcentration alone will underestimate the extent of accumulation in aquatic biota.
Accordingly, EPA's 2000 Human Health Methodology emphasizes the consideration of chemical
bioaccumulation by aquatic organisms, whereas EPA's 1980 Methodology emphasized the
measurement of bioconcentration.
Another important aspect of the bioaccumulation process is the steady-state condition.
Specifically, bioaccumulation can be viewed simply as the result of competing rates of chemical
uptake and depuration (chemical loss) by an aquatic organism. The rates of chemical uptake and
depuration can be affected by various factors, including the properties of the chemical, the
physiology of the organism in question, water quality and other environmental conditions, the
ecological characteristics of the water body (e.g., food web structure), and the concentration and
loadings history of the chemical. When the rates of chemical uptake and depuration are equal,
tissue concentrations remain constant over time and the distribution of the chemical between the
organism and its source(s) is said to be at steady state. For constant chemical exposures and
other conditions, the steady-state concentration in the organism represents the highest
accumulation potential of the chemical in that organism under those conditions. The time needed
for a chemical to achieve steady state in the organism has been shown to vary according to the
properties of the chemical, the variability of environmental conditions, and other factors. For
example, some highly hydrophobic chemicals can require long periods (e.g., many months) to
reach steady state between environmental compartments, whereas highly hydrophilic chemicals
usually reach steady state relatively quickly (e.g., hours to days).
National recommended AWQCs for the protection of human health are typically
designed to protect humans from harmful lifetime or long-term exposures to waterborne
contaminants. Given this goal, assessing bioaccumulation that equals or approximates steady-
state accumulation is one of the principles underlying the derivation of national BAFs. For
chemicals that require relatively long periods to reach steady state in aquatic organisms, changes
in the concentration of the chemical in the water column may occur much more rapidly than
corresponding changes in concentrations in tissue. Thus, if the system departs substantially from
steady-state conditions and water concentrations are not averaged over a sufficient time period,
the ratio of the chemical concentration in tissue of organisms to that in water (i.e., the BAF) may
have little resemblance to the steady-state ratio and have little predictive value for long-term
bioaccumulation potential. Therefore, BAF measurements should be based on chemical
concentrations in the water column, averaged over a sufficient period for the chemical of interest.
In addition, the BAFs used in deriving national recommended AWQCs for the protection of
human health should be based on adequate spatial averaging of chemical concentrations in both
tissue of consumed organisms and the water column.
The concept of proper temporal averaging for the determination of BAFs is illustrated in
Figure 1-1 (taken from Burkhard, 2003). Figure 1-1A shows the daily concentrations of a
hypothetical nonionic organic chemical, using a simple dilution model and daily flow data for the
Mississippi River at St. Paul, Minnesota. These daily chemical concentrations in the river can be
transformed into daily chemical concentrations in fish by using the kinetic models of Gobas
(1993). Figure 1-1B shows the results of these transformations in piscivorous fish for chemicals
1-4
-------
with log w-octanol-water partition coefficients (Kows) ranging from 2 to 9 for a simple
hypothetical food web. Together, Figures 1-1A and 1-1B show that concentrations of nonionic
organic chemicals in fish change over time, relative to the concentration of the chemical in the
ambient water, at speeds dependent upon the hydrophobicity of the chemical, i.e., the chemical's
Kow. The response is graded in magnitude, and the rate of change decreases with increasing Kow.
For chemicals with low Kows (e.g., log Kows of 2 and 3), the speed of change is very fast, such that
concentrations of the chemical in fish mimic the trends of the chemical concentration in ambient
Jan
Apr
Jul
Oct
Jan
Apr
Jul
Day
Oct
Jan Apr Jul
Oct
Jan
Figure 1-1 (A). Daily concentrations of a hypothetical nonionic organic chemical overtime in the water column,
predicted using a simple dilution model and daily flow data for the Mississippi River at St. Paul, Minnesota.
(B) Daily chemical concentrations in piscivorous fish found using the kinetic food web models of Gobas (1993) with
the daily chemical concentrations in the water column for nonionic organic chemicals with log w-octanol-water
water. For chemicals with large Kows (e.g., log Kows of 6 and 7), concentrations of the chemical in
fish change slowly relative to those in the water, and in general, the concentrations in fish follow
the long-term trends for the chemical concentration in the water.
Clearly, BAFs based on inappropriate temporal averaging of chemical concentrations in
the water will have little predictive power; thus, BAFs should be based on concentrations in the
water column that are averaged over a sufficient period of time that is appropriate for the
chemical of interest. For this reason, a BAF was defined in the 2000 Human Health Methodology
as representing the ratio (in liters per kilogram) of the concentration of a chemical in the tissue of
an aquatic organism to its concentration in the ambient water in situations where the organism
and its food are exposed and the ratio does not change substantially over time (i.e., the ratio
reflects bioaccumulation at or near steady state). Similarly, a BCF was defined as the ratio (in
liters per kilogram) of the concentration of a chemical in the tissue of an aquatic organism to the
chemical's concentration in the ambient water, in situations where the organism is exposed
through the water only and the ratio does not change substantially over time.
1-5
-------
From the perspective of sampling for determining BAFs, chemicals with large Kows will
generally require that numerous water samples be averaged over time to establish the long-term
chemical concentrations in the water. In contrast, for chemicals with low Kows, because the
concentrations in the fish mimic those in water, the time scale for establishing the chemical
concentrations in the water shrinks to concurrent sampling of both fish and water; current
chemical concentrations in the water provide a good predictor of the chemical concentration in
the fish. Burkhard (2003) provides additional details on BAF sampling design and EPA will
provide additional information on field sampling designs for determination of BAFs in TSD
Volume 3: Development of Site-Specific Bioaccumulation Factors.
1-6
-------
2. DEFINITIONS
The following terms and their definitions are used throughout this document.
2.1 BIOACCUMULATION
Bioaccumulation. The net accumulation of a chemical by an aquatic organism as a result of
uptake from all environmental sources.
Bioaccumulation factor (BAF). The ratio (in liters per kilogram of tissue) of the concentration of
a chemical in the tissue of an aquatic organism to its concentration in water, in situations where
both the organism and its food are exposed and the ratio does not change substantially over time.
The BAF is calculated as:
t
BAP = — (Equation 2-1)
where:
C, = concentration of chemical in tissue
Cw = concentration of chemical in water
Because chemical concentrations in tissue and water can be defined in terms of chemical
partitioning to different biological or chemical phases (e.g., total concentrations in tissue or water,
concentration in lipid, concentration that is freely dissolved in water), the general equation for
BAF (Equation 2-1) is further refined below to delineate among these different phases.
Total bioaccumulation factor (BAFj). A BAF based on the total concentration of chemical in
the organism and the water. The total concentration of the chemical in tissue includes that in
either a specific tissue or a whole organism and is based on wet tissue. The total concentration of
the chemical in water includes chemical associated with particulate organic carbon, chemical
associated with dissolved organic carbon, and chemical freely dissolved in the water. A BAFx is
often referred to as a "field-measured" BAF because it is derived from analysis of tissue and
water samples collected from the field. The BAFj is expressed in liters per kilogram of lipid. The
BAFj is calculated as:
t t
BAFT = — (Equation 2-2)
where:
C, = total concentration of chemical in tissue
Cw = total concentration of chemical in water
2-1
-------
Baseline bioaccumulation factor (Baseline BAF or BAF™). For nonionic organic chemicals
(and certain ionic organic chemicals to which similar lipid and organic carbon partitioning
behavior applies1), a BAF that is based on the concentration of chemical freely dissolved in water
and the concentration of the chemical in the lipid fraction of tissue. The baseline BAF is
expressed in liters per kilogram of lipid. The baseline BAF is determined using the equation:
Baseline BAF = BAFTfd =
L
BAFTl
- 1
(Equation 2-3)
where:
BAF^ = Total BAF
ffd = fraction of the total concentration of chemical in water that is freely
dissolved
f. = fraction of tissue that is lipid
Baseline BAF can also be defined as:
Baseline BAF = BAF/d - I (Equation 2-4)
where:
BAF? = lipid-normalized and freely dissolved-based bioaccumulation factor (see
definition below)
f. = fraction of tissue that is lipid
Note: Appendix A presents the derivation of the baseline BAF and refers to it as
"BAF£d." The subscript "L" signifies concentration of the chemical specifically in
lipid, in contrast to "•," which refers to lipid normalization in which the
concentration of the chemical in total tissue is divided by the fraction of the tissue
that is lipid (f). The superscript "fd" signifies the chemical that is freely dissolved
in water rather than total chemical in water. Based on an equilibrium partitioning
assumption for the chemical's distribution in both the organism and the water,
concentrations based on the "L" and "fd" chemical expressions can be calculated
using measured or predicted values of the fraction of tissue that is lipid and
fraction of total chemical that is freely dissolved in water, respectively (see
JAs described in Section 3.2.2 and illustrated by Figure 3-1, baseline BAFs for certain ionic organic
chemicals can be derived using methods developed for nonionic organic chemicals, which rely on lipid and organic
carbon partitioning theory. In these cases, similar lipid and organic carbon partitioning behavior should be known or
inferred (i.e., based on negligible ionization) for the ionic chemical in question.
2-2
-------
Appendix A). This avoids practical limitations associated with the direct analytical
measurement of concentrations of total chemical in lipid and freely dissolved
chemical in water.
Lipid-normalized and freely dissolved-based bioaccumulation factor (BAF™). The ratio (in
liters per kilogram of lipid) of the lipid-normalized concentration of a chemical in tissue of an
organism to the concentration of the chemical freely dissolved in water, in situations where both
the organism and its food are exposed and the ratio does not change substantially over time. The
BAF? is calculated as:
fft _ i
BAFi ~ — (Equation 2-5)
where:
C. = lipid-normalized concentration of chemical in tissues
C " = concentration of chemical that is freely dissolved in water
National trophic-level specific bioaccumulation factor (National BAFTL n). A B AF based on
nationwide average lipid content for trophic level "n" and nationwide average organic carbon in
ambient waters. The national BAF(TLn) is expressed in liters per kilogram wet tissue. The national
BAF(TLn) is calculated using the equation:
National BAFTLn = [(Final Baseline BAF)TLn - (Q^ + 1] - ^ (Equation 2.6)
where:
Final Baseline BAF^n = mean baseline BAF for trophic level "n"
f-ciLn) = fraction of tissue that is lipid in aquatic organisms at trophic level "n"
ffd = fraction of the total concentration of chemical in water that is freely
dissolved
2.2 BIOCONCENTRATION
Bioconcentration. The net accumulation of a chemical by an aquatic organism as a result of
uptake directly from the ambient water, through gill membranes or other external body surfaces.
Bioconcentration factor (BCF). The ratio (in liters per kilogram of tissue) of the concentration of
a chemical in the tissue of an aquatic organism to its concentration in water, in situations where
the organism is exposed through the water only and the ratio does not change substantially over
time. The BCF is calculated as:
2-3
-------
ct
BCF = — (Equation 2-7)
where:
C, = concentration of chemical in tissue
Cw = concentration of chemical in water
Because chemical concentrations in tissue and water can be defined in terms of chemical
partitioning to different biological or chemical phases (e.g., total concentrations in tissue or water,
concentration in lipid, concentration that is freely dissolved in water), the general equation for
BCF (Equation 2-7) is further refined below to delineate among these different phases.
Total bioconcentration factor (BCFj). A BCF based on the total concentration of chemical in
the organism and the water. The total concentration of the chemical in tissue includes that in
either a specific tissue or a whole organism and is based on wet tissue. The total concentration of
the chemical in water includes chemical associated with particulate organic carbon, chemical
associated with dissolved organic carbon, and chemical freely dissolved chemical in the water. A
BCF is often referred to as a "laboratory-measured BCF" because it can be measured only in the
laboratory. The BCFj is expressed in liters per kilogram of lipid. The BCFj is calculated as:
t t
BCFT = — (Equation 2-8)
where:
C, = total concentration of chemical in tissue
Cw = total concentration of chemical in water
Baseline bioconcentration factor (Baseline BCF or BCF"). For nonionic organic chemicals
(and certain ionic organic chemicals to which similar lipid and organic carbon partitioning
behavior applies2), a BCF that is based on the concentration of chemical freely dissolved in water
and the concentration of the chemical in the lipid fraction of tissue. The baseline BCF is
expressed in liters per kilogram of lipid. The baseline BCF is determined using the equation:
2As described in Section 3.2.2 and illustrated by Figure 3-1, baseline BCFs for certain ionic organic
chemicals can be derived using methods developed for nonionic organic chemicals, which rely on lipid and organic
carbon partitioning theory. In these cases, similar lipid and organic carbon partitioning behavior should be known or
inferred (i.e., based on negligible ionization) for the ionic chemical in question.
2-4
-------
Baseline BCF = BCFLfd =
(Equation 2-9)
where:
BCF^ = Total BCF
ffd = fraction of the total concentration of chemical in water that is freely
dissolved
f. = fraction of tissue that is lipid
Baseline BCF can also be defined as:
Baseline BCF = BCF/d - I (Equation 2- 10)
where:
BCF? = lipid-normalized and freely dissolved-based bioconcentration factor (see
definition below)
f. = fraction of tissue that is lipid
Note: Appendix A presents the derivation of the baseline BCF and refers to it as
"BCF£d." The subscript "L" signifies concentration of the chemical specifically in
lipid, in contrast to "•," which refers to lipid normalization in which the
concentration of the chemical in total tissue is divided by the fraction of the tissue
that is lipid (f.). The superscript "fd" signifies the chemical that is freely dissolved
in water rather than total chemical in water. Based on an equilibrium partitioning
assumption for the chemical's distribution in both the organism and the water,
concentrations based on the "L" and "fd" chemical expressions can be calculated
using measured or predicted values of the fraction of tissue that is lipid and
fraction of total chemical that is freely dissolved in water, respectively (see
Appendix A). This avoids practical limitations associated with the direct analytical
measurement of concentrations of total chemical in lipid and freely dissolved
chemical in water.
Lipid-normalized and freely dissolved-based bioconcentration factor (BCF™). The ratio (in
liters per kilogram of lipid) of the lipid-normalized concentration of a chemical in tissue of an
organism to the concentration of the chemical freely dissolved in water, in situations where both
the organism is exposed through water only and the ratio does not change substantially over
time. The BCF? is calculated as:
2-5
-------
fd _ cl
BCFi --
(Equation 2-11)
where:
C. = lipid-normalized concentration of chemical in tissues
C " = concentration of chemical that is freely dissolved in water
2.3 ADDITIONAL TERMS
Biomagnification. The increase in concentration of a chemical in the tissue of organisms along a
series of predator-prey associations, primarily through the mechanism of dietary accumulation.
Biomagnification factor (BMF). The ratio (unitless) of the concentration of a chemical in a
predator organism at a particular trophic level to the concentration of the chemical in the tissue of
its prey organism at the next lowest trophic level for a given water body and chemical exposure.
For nonionic organic chemicals (and certain ionic organic chemicals to which similar lipid and
organic carbon partitioning behavior applies), a BMF can be calculated using lipid-normalized
concentrations of chemical in the tissue of organisms at two successive trophic levels as:
C
BMFCIL.K> = r lCIL>n> (Equation 2-12)
S PL. n- 1)
where:
BMF(TL n) = biomagnification factor for trophic level "n" (TL "n")
C. (TL, n) = lipid-normalized concentration of chemical in tissue of predator
organism at a given trophic level (TL "n")
C. (TL, n-i) = lipid-normalized concentration of chemical in tissue of prey organism
at the next lower trophic level from the predator (TL "n-1 ")
For those inorganic, organometallic, and ionic organic chemicals for which lipid and organic
carbon partitioning does not apply (see Section 5.6), a BMF can be calculated using chemical
concentrations in the tissue of organisms at two successive trophic levels as:
CIL. n- 1)
(Equation 2-13)
2-6
-------
where:
BMF(TL n) = biomagnification factor for trophic level "n" (TL "n")
Q (TL, n) = concentration of chemical in tissue of predator organism at a given
trophic level (TL "n")
Q (TL, n-i) = concentration of chemical in tissue of prey organism at the next lower
trophic level from the predator (TL "n-1")
Biota-sediment accumulation factor (BSAF). For nonionic organic chemicals (and certain ionic
organic chemicals to which similar lipid and organic carbon partitioning behavior applies), the
BSAF is the ratio (in kilograms of sediment organic carbon per kilogram of lipid) of the lipid-
normalized concentration of a chemical in tissue of an aquatic organism to its organic carbon-
normalized concentration in surface sediment, in situations where the ratio does not change
substantially over time, both the organism and its food are exposed, and the surface sediment is
representative of average surface sediment in the vicinity of the organism. The BSAF is calculated
as:
BSAF =
(Equation 2-14)
KM
where:
C. = lipid-normalized concentration of chemical in tissue
Csoc = concentration of chemical in dry sediment, normalized to sediment organic
carbon
Depuration. Loss of a chemical from an organism as a result of any active or passive process.
Equilibrium. A thermodynamic condition under which a chemical's activity, or fugacity, is
equal among all phases composing the system of interest. In systems at equilibrium, chemical
concentrations in all phases will remain unchanged over time.
Food-chain multiplier (FCM). For nonionic organic chemicals (and certain ionic organic
chemicals to which similar lipid and organic carbon partitioning behavior applies), the ratio of a
baseline BAF for an organism of a particular trophic level to the baseline BCF (usually
determined for organisms in trophic level one). For inorganic, organometallic, and certain ionic
organic chemicals to which lipid and organic carbon partitioning does not apply, a FCM can be
derived based on total (wet or dry weight) concentrations of the chemical in tissue as described in
Sections 4.4.1 and 4.4.2.
Freely dissolved concentration (C"). For nonionic organic chemicals, the concentration of the
chemical that is dissolved in ambient water, excluding the portion sorbed onto particulate or
dissolved organic carbon (POC or DOC). The freely dissolved chemical concentration is
considered to represent the most bioavailable form of an organic chemical in water and therefore
2-7
-------
is the form that best predicts bioaccumulation. The freely dissolved concentration can be
determined as:
Cwfd = C^ - ffA (Equation 2- 15)
where:
C^ = total concentration of chemical in water
ffd = fraction of the total concentration of chemical in water that is freely
dissolved
Hydrophilic. Having affinity for water; the extent to which a chemical is attracted to partitioning
into the water phase. Hydrophilic organic chemicals have a greater tendency to partition into
polar phases (e.g., water) than do hydrophobic chemicals.
Hydrophobic. Lacking affinity for water; the extent to which a chemical avoids partitioning into
the water phase. Highly hydrophobic organic chemicals have a greater tendency to partition into
nonpolar phases (e.g., lipid, organic carbon) than do hydrophilic chemicals.
Lipid-normalized concentration (C). The total concentration of a chemical in a tissue or whole
organism divided by the fraction of that tissue or whole organism that is lipid. The lipid-
normalized concentration can be calculated as:
Ct
C| - — (Equation 2- 16)
where:
C, = concentration of chemical in tissue
f. = fraction of tissue that is lipid
fl-Octanol-water partition coefficient (Kow). The ratio of the concentration of a chemical in the
w-octanol phase to its concentration in the aqueous phase in an equilibrated two-phase w-octanol-
water system. For log Kow, the log of the w-octanol -water partition coefficient is a base 10
logarithm.
Sediment organic carbon-normalized concentration (Csoc). For sediments, the total
concentration of a contaminant in sediment divided by the fraction of organic carbon in
sediment. The sediment organic carbon-normalized concentration can be calculated as:
c
'sec
c— = -^- (Equation 2-17)
2-8
-------
where:
Cs = concentration of chemical in dry sediment
fsoc = fraction of dry sediment that is organic carbon
Sediment-water column concentration quotient (• socw). The ratio (in liters per kilogram of
organic carbon) of the concentration of chemical in the sediment, on an organic carbon basis, to
that in the water column, on a freely dissolved basis. Ilsocw when divided by the Kow of the
chemical provides a measure, for a given ecosystem, of the chemical's thermodynamic gradient
between the sediment and the water column. The sediment-water column concentration quotient
is calculated as:
— SCXS
~ — • iJ (Equation 2- 18)
where:
Csoc = concentration of chemical in dry sediment, normalized to sediment organic
carbon
C " = concentration of chemical that is freely dissolved in water
Steady state. A condition reached by a system when rates of chemical movement between
phases and reactions within phases are constant so that concentrations of the chemical in the
phases of the system are unchanged over time. A system at steady state is not necessarily at
equilibrium; steady-state conditions often exist when some or all of the phases of the system
have different activities or fugacities for the chemical.
Uptake. Movement of chemical from the environment into an organism as the result of any
active or passive process.
2-9
-------
3. OVERVIEW OF THE NATIONAL BAF METHODOLOGY
This section provides an overview of the methodology EPA will use for deriving national
BAFs for setting AWQCs for the protection of human health. As mentioned in Section 1,
national BAFs are intended to account for some major chemical, biological, and ecological
attributes that can affect bioaccumulation in bodies of water across the United States. Therefore,
EPA will use separate procedures for deriving national BAFs depending on the type of chemical
(i.e., nonionic organic, ionic organic, inorganic, and organometallic). In addition, to account for
other factors, such as biomagnification and broad physiological differences between trophic
levels, EPA's national BAFs are derived separately for each trophic level. The methodology
results in three national trophic level-specific BAFs for each chemical, one specific for each of
trophic levels 2, 3, and 4 (BAF2, BAF3, and BAF4).
BAFs can be measured or estimated with a variety of methods, ranging from empirically
driven approaches that rely on measurements of chemical concentrations in aquatic organisms
and their surrounding environmental media (water and sediment) to mechanistically driven
approaches that rely on food web models in combination with information about the properties
of chemicals and ecosystems to estimate bioaccumulation. The four methods that EPA will use
for deriving national BAFs are described in the following sections. For a given chemical, the
choice of which method to use for deriving a national BAF depends on several factors. These
factors include the properties of the chemical of interest, the relative strengths and limitations of
the BAF method, and the level of uncertainty associated with the bioaccumulation or
bioconcentration measurements. Because multiple evaluation steps are involved in selecting the
most appropriate BAF method(s) for a given chemical and data set, EPA has developed a
decision framework for deriving national BAFs (Figure 3-1). This framework illustrates the major
steps and decisions that will ultimately lead to calculating a national BAF. Use of this framework
leads to selection of one of six possible procedures (shown at the bottom of
Figure 3-1) for deriving national BAFs. Each procedure includes those BAF derivation methods
that are suitable for the class and properties of chemicals to which the procedure applies. The
following subsections are a prelude to the detailed discussion of the national BAF methodology
provided in Sections 4 through 7. Section 3.1 introduces each of the four methods available for
deriving national BAFs, including a discussion of their relative strengths and limitations. Section
3.2 provides additional discussion and explanation of the BAF derivation framework that applies
to all chemical types.
3-1
-------
(''DEFINE CHEMICAL\
\^ OF CONCERN J
(COLLECT & REVIEW}
\^ DATA J
CLASSIFY CHEMICAL^]
OF CONCERN
I
PROCEDURE #1
1. Field BAF
2. BSAF
3. LAB BCF*
4. K* FCM
#1
;M
i
PRC
1. F
L<
2. K
r
PROCEDURES
1 . Field BAF
2. BSAF
3. Lab BCF
I
PROCEDURE#3
1. Field BAFor
Lab BCF
I
Inorganic &
Organometallic
1
BIOMAGNIFICATION?)
^}
Yes
Figure 3-1. Framework for selection of methods for deriving national B AFs.
3-2
-------
3.1 SUMMARY OF FOUR BIO ACCUMULATION METHODS
Bioaccumulation factors used to derive national trophic level-specific BAFs can be
measured or predicted using one or more of the following four methods, depending on the type
of chemical and its properties:
1. Measured BAFs derived from data obtained from a field study (i.e., field-
measured BAFs)
2. BAFs predicted from biota-sediment accumulation factors (BSAFs) obtained from
a field study (i.e., field-measured BSAFs)
3. BAFs predicted from laboratory-measured BCFs, with or without adjustment by a
food-chain multiplier
4. BAFs predicted from a chemical's w-octanol-water partition coefficient (Kow), with
or without adjustment by a food-chain multiplier
Each of the four methods is summarized below. Details of each of the four methods are
described in Section 5.
3.1.1 Field-Measured BAFs
A BAF derived from data obtained from field-collected samples of tissue and water—
referred to here as a "field-measured BAF"—is the most direct measure of bioaccumulation. A
field-measured BAF is determined from measured chemical concentrations in an aquatic
organism and the ambient water collected from the same field location. Because the data are
collected from a natural aquatic ecosystem, a field-measured BAF reflects an organism's
exposure to a chemical through all relevant exposure routes (e.g., water, sediment, diet). A field-
measured BAF also reflects factors that influence the bioavailability and metabolism of a
chemical that might occur in the aquatic organism or its food web. Therefore, field-measured
BAFs are appropriate for all chemicals, regardless of the extent of chemical metabolism in biota.
3.1.2 BAFs Predicted from a Field-Measured BSAF
For nonionic organic chemicals (and certain ionic organic chemicals to which similar lipid
and organic carbon partitioning behavior applies), BAFs can also be predicted from BSAFs. A
BSAF is similar to a field-measured BAF in that the concentration of a chemical in biota is
measured from field-collected samples and it reflects an organism's exposure to all relevant
exposure routes. A BSAF also accounts for bioavailability and chemical metabolism that might
occur in the aquatic organism or its food web. A BSAF references the concentration of the
chemical in an organism to the concentration of chemical in sediment, but it may be converted to
a BAF when the chemical's distribution between sediments and water can be estimated. The
BSAF procedure is used only to predict a BAF for moderate to highly hydrophobic organic
chemicals.
3-3
-------
3.1.3 BAFs Predicted from Laboratory-Measured BCFs
A laboratory-measured BCF can be used to estimate a BAF for organic and inorganic
chemicals either with or without adjustment with a food-chain multiplier, depending on the
importance of nonaqueous exposure routes. However, unlike a field-measured BAF or one
predicted from a field-measured BSAF, a laboratory-measured BCF typically reflects only the
accumulation of chemical through the water exposure route. A laboratory-measured BCF may
therefore underpredict BAFs for chemicals for which accumulation from sediment or dietary
sources is important. In these cases, laboratory-measured BCF can be adjusted by a factor known
as a food-chain multiplier (FCM) to better reflect accumulation through the food web from
dietary exposures. Because a laboratory-measured BCF is determined by using the measured
concentration of a chemical in an aquatic organism and its surrounding water, a laboratory-
measured BCF often reflects metabolism of the chemical that occurs in the organism during the
BCF measurement, but not in the food web.
3.1.4 BAFs Predicted from Kow
A chemical's Kow (measured or predicted) can also be used to predict a BAF for nonionic
organic chemicals. This procedure is appropriate for nonionic organic chemicals but can also be
applied to certain ionic chemicals that have lipid and organic carbon partitioning behavior similar
to that of nonionic organics. The Kow is strongly correlated with the BCF for nonionic organic
chemicals, in particular those chemicals that are poorly metabolized by aquatic organisms. For
nonionic organic chemicals where food web exposure is important, use of the Kow alone, as an
estimate of BCF, will underpredict the BAF because the BCF accounts only for chemical
exposure from water. In such cases, the Kow is adjusted with an FCM as described for the BAF
method in Section 3.1.3.
3.1.5 Advantages and Limitations of BAF Methods
Each BAF derivation method summarized above has strengths and limitations associated
with it that will be considered and balanced when deriving national BAFs. These strengths and
limitations, as summarized in Table 3-1, form the basis for the Framework for selecting methods
for deriving national BAFs (Figure 3-1) that is described in Section 3.2. For example, use of the
field-measured BAF method is advantageous in that it applies to all chemical types, and accounts
for site-specific factors that affect bioavailability, biomagnification, and metabolism. However,
the current database of acceptable field-measured BAFs is relatively limited, in terms of both
number of sites and chemicals for which they have been derived. Furthermore, field-measured
BAFs cannot be readily determined for chemicals that are very difficult to accurately measure in
the water column (e.g., 2,3,7,8-TCDD). BAFs derived from field-measured BSAFs offer a
number of the same strengths as field-measured BAFs (e.g., they account for biomagnification,
metabolism, and site-specific factors affecting bioavailability). In addition, the BSAF method is
the only field-based method that can be used for chemicals such as 2,3,7,8-TCDD that are
difficult to measure in ambient water. In EPA's framework, however, application of the BSAF
method is currently limited to nonionic organic chemicals of moderate to high hydrophobicity.
BAFs predicted from laboratory-measured BCF can be applied to all chemical types, and data are
3-4
-------
generally more plentiful than with field-measured BAFs. However, laboratory-based BCFs by
themselves do not address chemical biomagnification in food webs unless they are adjusted with
a field- or model-derived FCM. In addition, acceptable BCFs for highly hydrophobic chemicals
(i.e., those with a log Kow > 6) appear to be very limited, often because of lack of ancillary data
that affect bioavailability (e.g., dissolved organic carbon). Finally, the model-derived BAF
derivation method (using Kow and FCMs where appropriate) offers a distinct advantage in that no
laboratory data (besides a Kow) or field data are needed to derive a BAF. However, this method is
limited to nonionic organic chemicals and is currently constrained by the lack of in vivo data on
chemical metabolism.
Table 3-1. Strengths and Limitations of the Four BAF Methods for Deriving National BAFs
BAF derivation
method
Strengths
Limitations
1. Field-
measured
BAF
2. BAF
predicted
from field-
measured
BSAF
3. BAF
predicted
from lab-
measured
BCF x FCM
4. BAF
predicted
from a Kow
FCM
Applicable to all chemical types
Incorporates chemical biomagnification
and metabolism
Reflects site-specific attributes that affect
bioavailability and dietary exposure
Incorporates chemical biomagnification
and metabolism
Reflects site-specific attributes that affect
bioavailability and dietary exposure
Useful for chemicals that are difficult to
analyze in water
Use of chemical concentrations in
sediment reduces temporal variability
Applicable to all chemical types
BCF may account for chemical
metabolism in test organisms
Large BCF database available
Standardized test methods
Readily applied with minimal input data
High-quality data currently limited to few
sites and chemicals
Representative chemical concentration in
water may be difficult to quantify
Limited to nonionic organic chemicals
with log Kow- 4
High-quality data currently limited to few
chemicals and sites
Accuracy depends on representativeness
and quality of estimate of chemical
distribution between sediment and water
Chemical metabolism, when present in
food web, generally not accounted for
High-quality data currently limited for
highly hydrophobic chemicals, in part
because of lack of ancillary data that affect
bioavailability
Limited to nonionic organic chemicals
Chemical metabolism, when present, not
accounted for
Accuracy depends on accuracy of Kow
3.2 FRAMEWORK FOR DERIVING NATIONAL BAFs
The EPA's framework for deriving national BAFs is depicted in Figure 3-1. The goal of
this framework and the BAF guidance presented in the 2000 Human Health Methodology is to
facilitate the full use of available data and methods for deriving national BAFs while prioritizing
and restricting the use of certain BAF methods based on their inherent strengths and limitations,
as summarized in Section 3.1. Use of this decision framework results in selection, based on the
class and properties of the chemical, of one of six "Procedures," each of which can be used to
3-5
-------
derive national BAFs for a chemical having the specified class and properties. Each procedure
includes one or more of the methods described in Section 3.1. Within a procedure, the number
next to each BAF method indicates its general order of preference in the hierarchy for calculating
national BAFs. For example, a field-measured BAF is generally given the highest preference for
deriving a national BAF using Procedure 1, followed by a BAF predicted from a BSAF, a BAF
predicted from a BCF x FCM, and, finally, a BAF predicted from a Kow x FCM However, the
hierarchy of methods within each procedure is not intended to be inflexible, as explained in
Section 6.1 and in the 2000 Human Health Methodology. Some situations may indicate that
greater uncertainty is likely to occur when applying a BAF derived from a "more highly
preferred" method (e.g., a field-measured BAF within Procedure 1) than with a "less preferred"
method (e.g., BAF predicted from BCF x FCM within Procedure 1), for example, when data
from the more preferred method are limited in terms of their representativeness, quantity, or
quality relative to the lower-tier method. In these situations, data from the lesser preferred, but
least uncertain, method should be used to derive the national BAFs.
The first step in the national BAF derivation framework involves precisely defining the
chemical of concern. The purpose of this step is to ensure consistency between the form(s) of
chemical used to derive national BAFs and the form(s) used as the basis of the health assessment
(e.g., the reference dose or point of departure/uncertainty factor). Although this step is usually
unambiguous for single chemicals that are stable in the environment, complications can arise
when assessing chemicals that occur as mixtures or undergo complex transformations in the
environment.
The second step of the framework consists of collecting and reviewing data on
bioaccumulation and bioconcentration. The third step involves classifying the chemical into one
of three broadly defined categories: nonionic organic, ionic organic, and
inorganic/organometallic. This step is important because some of the four BAF methods
summarized in Section 3.1 are specific to certain chemical groups (e.g., the BSAF method for
nonionic organic chemicals). For the purposes of the 2000 Human Health Methodology, nonionic
organic chemicals are defined as organic compounds that do not ionize substantially in natural
bodies of water. These chemicals are also referred to as "neutral" or "nonpolar" organics in the
scientific literature. Ionic organic chemicals are considered to include those chemicals that
contain functional groups with exchangeable protons, such as hydroxyl, carboxylic, sulfonic, and
nitrogen (pyridine) groups. Ionic organic chemicals undergo ionization in water, the extent of
which depends on the pH and the pKa of the chemical. Ionic chemicals are considered separately
when deriving national BAFs because the behavior of the anionic or cationic species of these
chemicals is much different from those of their neutral (un-ionized) counterparts. Inorganic and
organometallic chemicals include inorganic minerals, other inorganic compounds and elements,
metals, metalloids, and organometallic compounds. Additional guidance on the first three steps
of the framework is found in Section 5.3 of the 2000 Human Health Methodology.
Once the chemical is classified into one of the three chemical categories, additional
evaluation steps are necessary to determine which of the BAF procedures should be used to
derive a national BAF. These steps are summarized below for each of the three chemical
categories.
3-6
-------
3.2.1 BAF Derivation Procedures for Inorganic and Organometallic Chemicals
For inorganic and organometallic chemicals, the primary factor to be evaluated is the
likelihood that the chemical will undergo biomagnification in the food web. At present, evaluating
the biomagnification potential for this group of chemicals is almost exclusively limited to
analyzing empirical data on the importance of food web (dietary) exposure and biomagnification
in determining chemical concentrations in aquatic species. For example, available data indicate
that methylmercury biomagnifies in aquatic food webs, whereas other chemicals in this category
do not routinely biomagnify (e.g., copper, zinc, lead). If biomagnification is considered to be
likely, then field-measured BAFs are the preferred BAF method, followed by laboratory-
measured BCF adjusted with an FCM. If biomagnification is determined to be unlikely, field-
measured BAFs and laboratory-measured BCF are considered to be of equal utility for deriving
national BAFs, all other factors being equal. Additional guidance on determining national BAFs
for inorganic and organometallic chemicals is provided in Section 5.6 of the 2000 Human Health
Methodology. It should be noted that metal bioaccumulation can vary substantially across
organisms due to a number of factors, including physiological differences and variation in
mechanisms by which organisms take up, distribute, detoxify, store, and eliminate metals from
their tissues. As a result of the complexity of assessing the fate and effects of metals in the
environment, EPA has embarked on an initiative to provide additional guidance on conducting
metal assessments, including metals bioaccumulation by aquatic organisms (USEPA, 2002).
3.2.2 BAF Derivation Procedures for Ionic Organic Chemicals
For chemicals classified as ionic organic chemicals, the primary evaluation step involves
estimating the relative extent of ionization and evaluating their partitioning behavior with lipids
and organic carbon. If the relative extent of ionization that is likely to occur at pH ranges that are
typical of U.S. surface waters is negligible (see the 2000 Human Health Methodology for
guidelines on this determination), and if the un-ionized form of the ionic chemical behaves like a
nonionic organic chemical, in which lipid and organic carbon partitioning controls the behavior of
the chemical, then the chemical can be treated essentially as a nonionic chemical for the purposes
of deriving national default BAFs. If ionization is considered potentially important, or if non-lipid
and non-organic carbon mechanisms control the behavior of the chemical, then the ionic
chemical is treated in the same way as inorganic and organometallic chemicals for deriving
national BAFs. Additional guidance for deriving national BAFs for ionic organic chemicals is
provided in Section 5.5 of the 2000 Human Health Methodology.
Perfluorinated alkyl acids are an example of ionic organic chemicals. Some of these
chemicals bioconcentrate and biomagnify in food webs via non-lipid mediated mechanisms; i.e.,
lipid and organic carbon partitioning behavior observed for nonionic organic chemicals does not
apply. For the perfluorinated alkyl acids, Procedure 6 (Figure 3-1) would be used to derive
national default BAFs.
3-7
-------
3.2.3 BAF Derivation Procedures for Nonionic Organic Chemicals
Deriving national BAFs for nonionic organic chemicals is somewhat more complex than
for the other two chemical classes. First, four national BAF derivation procedures are applicable
to nonionic organic chemicals. Second, selecting the most appropriate derivation procedure
depends greatly on chemical properties, which are evaluated in two decision steps (see Figure
3-1). Finally, once the derivation procedure is selected, additional adjustments are made to the
BAFs in order to account for differences in factors that affect bioaccumulation of this group of
chemicals in aquatic organisms (e.g., lipid content in test organisms and organic carbon content
in water).
Figure 3-2 shows the national BAF derivation process for nonionic organic chemicals.
This process is divided into four steps:
Step 1. Selecting the BAF derivation procedure
Step 2. Calculating individual baseline BAFs
Step 3. Selecting final baseline BAFs
Step 4. Calculating national BAFs from the final baseline BAFs
A summary of each step follows.
Step 1: Selecting a BAF Derivation Procedure
Step 1 of the approach determines which of the four BAF procedures described in
Section 3-1 will be appropriate for deriving the national BAF for a given nonionic organic
chemical. As shown in Figure 3-1, there are two decision points. The first decision point requires
knowledge of the chemical's hydrophobicity (i.e., the Kow of the chemical). The Kow provides an
initial basis for assessing whether nonaqueous (e.g., food web, sediment) exposure and
biomagnification may be a concern for nonionic organic chemicals. Knowledge of the likely
importance of nonaqueous routes of exposure determines whether or not some methods (e.g.,
lab-measured BCF, Kow-derived BAF) require additional adjustments to account for this
exposure. Guidance for selecting the Kow for a chemical is provided in Appendix B of this TSD.
For the purposes of the 2000 Human Health Methodology, nonionic organic chemicals with log
Kow values equal to or greater than 4.0 are classified as "moderately to highly hydrophobic." For
moderately to highly hydrophobic nonionic organic chemicals, available data indicate that
exposure through the diet and other nonaqueous routes can become important in determining
chemical residues in aquatic organisms (e.g., Russell et al., 1999; Fisk et al., 1998; Oliver and
Niimi, 1983, 1988; Niimi, 1985; Swackhammer and Kites, 1988). Below a log Kow of 4, available
information indicates that nonaqueous exposure to these chemicals is not likely to be important.
The second decision point involves assessing the importance that chemical metabolism
might have in determining chemical concentrations in aquatic organisms. Assessing metabolism
is important because it affects the degree to which a chemical bioaccumulates (and biomagnifies)
in aquatic food webs. For example, some polynuclear aromatic hydrocarbons have Kow values
that would warrant initial concern for biomagnification (i.e., log Kow > 4), but chemical
3-8
-------
metabolism by higher organisms (primarily fish) often results in reduced concentrations in fish
(Endicott and Cook, 1994; Burkhard, 2000). Guidance for assessing whether a high or low rate of
metabolism is likely for a given chemical is provided in Section 5.4.2.3 of the 2000 Human Health
Methodology.
Together, the hydrophobicity and metabolism decision points lead to the selection of one
of four BAF procedures. Procedure 1 applies to chemicals with moderate to high Kows, where
(1) the influence of chemical metabolism is suspected to be minor (e.g., polychlorinated
biphenyls [PCBs], dichlorodiphenyltrichloroethane [DDT], dieldrin, etc.) or (2) there are
insufficient data on chemical metabolism to make a determination (this reflects a policy decision
to err on the side of public health protection in the absence of data). Procedure 2 applies to
moderate-to-high Kow chemicals for which the influence of chemical metabolism on
bioaccumulation is considered to be important (e.g., selected polynuclear aromatic
hydrocarbons). Within this procedure, the use of Kow-based estimates (with or without FCMs) of
BAFs is restricted because the Kow may substantially overpredict bioaccumulation for chemicals
that are metabolized. Procedure 3 applies to low-Kow chemicals for which chemical metabolism is
not considered significant. For such chemicals, no preference is given to field-measured BAFs
over laboratory-measured BCF (i.e., both methods are appropriate), since biomagnification is not
considered important for low-Kow chemicals. Procedure 4 applies to low-Kow chemicals for which
metabolism is considered to be important. In this procedure as in Procedure 2, use of Kow-
predicted BAFs is not recommended because the Kow may substantially overpredict
bioaccumulation.
Step 2: Calculating Individual Baseline BAFs
Step 2 involves calculating individual, species-specific baseline BAFs using all of the
methods available within the selected BAF derivation procedure. Calculating an individual
baseline BAF involves normalizing the field-measured BAFx (or laboratory-measured BCFj),
which are typically based on total concentrations in tissue and water by the lipid content of the
study organism and the fraction of total chemical that is freely dissolved in the ambient water.
Both the lipid content in the organism and the freely dissolved chemical concentration (as
influenced by organic carbon in water) have been shown to be important factors that influence
the bioaccumulation of nonionic organic chemicals (e.g., Mackay, 1982; Connolly and Pederson,
1988; Thomann, 1989; Suffet et al., 1994). Therefore, baseline BAFs, which are expressed on the
basis of the chemical concentration in the lipid fraction of tissue and freely dissolved in water, are
considered more amenable to being applied across different species and bodies of water than are
BAFs or BCF expressed on the basis of the total concentrations in the tissue and water. Because
bioaccumulation can be strongly influenced by the trophic position of aquatic organisms
(through either biomagnification or physiological differences), extrapolation of baseline BAFs
should not be performed between species of different trophic levels. An example of how a
baseline BAF is calculated from a field-measured BAFx is shown by Equation 3-1. Equations for
calculating baseline BAFs differ according to the BAF derivation method. Examples of baseline
BAF equations for other BAF derivation methods are provided in Sections 5.1 through 5.4 of this
TSD and in Sections 5.4.3 through 5.4.6 of the 2000 Human Health Methodology.
3-9
-------
Baseline BAF = BAFLfd =
BAF
- 1
(Equation 3-1)
where:
ffd
f.
= Total BAF
= fraction of the total concentration of chemical in water that is freely
dissolved
= fraction of the tissue that islipid
c
: CHEMICAL ~\
METABOLISM DATA^ ~~
Step 1 .
Select BAF
Procedure
BIOMAGNIFICATION \
DATA j
( LIPID CONTENT OF\
{. STUDY ORGANISM l~
Step 2.
Calculate
Individual
Baseline BAFs
I DATA PREFERENCE \
\ HIERARCHY ~
FREELY DISSOLVED\
FRACTION IN STUDY )
WATER J
Step 3.
Select Final
Baseline BAF(s)
(UNCERTAINTY IN \
BASELINE BAFs J
I LIPID CONTENT OF
{ CONSUMED
\ ORGANISMS
Step 4.
Calculate
National BAF(s)
FREELY DISSOLVED^
FRACTION IN AWQC j
WATERS
Figure 3-2. BAF derivation for nonionic organic chemicals.
3-10
-------
Step 3: Selecting Final Baseline BAFs
Step 3 of the methodology consists of selecting the final baseline BAFs from the
individual baseline BAFs by using a weight-of-evidence approach that takes into account the
uncertainty in the individual BAFs and the data preference hierarchy (i.e., field-measured BAFs
are preferred over BAFs derived using the other methods). The individual baseline BAFs should
be calculated using as many of the methods as possible under the appropriate BAF derivation
procedure. As described earlier, the data preference hierarchy discussed in Section 5.4.2 of the
2000 Human Health Methodology is not inflexible. Rather, it is intended to be a guide for
selecting the most appropriate final BAF when the uncertainty is similar between two individual
baseline BAFs calculated using different methods. Section 6. 1 of this TSD and Section 5.4.3.2 of
the 2000 Human Health Methodology provide more detailed discussions of this step.
Step 4: Calculating National BAFs
The fourth and final step in calculating national BAFs for nonionic organic chemicals
involves calculating three trophic-level specific BAFs that will be used in the equation to calculate
national recommended AWQC for the protection of human health. This step involves adjusting
final baseline BAFs to reflect the average lipid content of commonly consumed fish and shellfish
and bioavailability of the chemical in waters to which the national recommended AWQC will
apply. Converting baseline BAFs to national BAFs requires information on (1) the percent lipid
of the aquatic organisms commonly consumed in the United States and (2) the fraction of
chemical that is freely dissolved that is expected to be present in the ambient waters of interest.
Baseline BAFs are not used directly in the derivation of the national AWQC because they do not
reflect the conditions that affect chemical bioavailability in U.S. waters or chemical accumulation
due to lipid content of the fish and shellfish residing in U.S. waters. The equation for calculating a
national BAF for each trophic level is:
National BAFTLn = [(Final Baseline BAF)TLn - (Q^ + 1] - ^d (Equation 3-2)
where:
Final Baseline BAF^ n = mean baseline BAF for trophic level "n"
f-ciLn) = fraction of tissue that is lipid in aquatic organisms at trophic
level "n"
ffd = fraction of the total concentration of chemical in water that is
freely dissolved
The technical basis of Equation 3-2 is provided in Section 4. Procedures EPA will use for
determining each component of Equation 3-2 are provided in Sections 6.4 and 6.5.
3-11
-------
4. BACKGROUND INFORMATION ON LIPID NORMALIZATION,
BlOAVAILABILITY, AND BlOMAGNIFICATION
National trophic-level specific BAFs are intended to represent the long-term, average
bioaccumulation potential of a pollutant in aquatic organisms of a particular trophic level
(i.e., 2, 3, or 4) that are commonly consumed by humans throughout the United States. For
certain chemicals (e.g., nonionic organics), chemical bioavailability, biota lipid content, and
trophic transfer can affect bioaccumulation potential and ultimately the magnitude of BAFs.
Because chemical bioavailability, biota lipid content, and trophic transfer can vary across
locations and species, these factors should be accounted for in the derivation of national BAFs.
Figure 3-2 in Section 3.2.3 presents EPA's stepwise process for developing national BAFs for
nonionic organics, of which a key step is derivation of baseline BAFs.
The scientific basis for the lipid and freely dissolved fraction normalizations for nonionic
organic chemicals are presented in Sections 4.1 through 4.3. Section 4.4 presents a discussion on
how biomagnification is incorporated into the baseline BAFs in certain BAF methods.
4.1 LIPID NORMALIZATION
4.1.1 Background and Theory
The importance of lipid content in influencing the bioaccumulation of nonionic organic
chemicals in aquatic organisms is well documented. Early work by Reinert (1969) and Reinert et
al. (1972) demonstrated that nonionic organic chemicals concentrate in the lipids of organisms,
and that differences in DDT concentrations between species and size groups are reduced when
the concentrations of chemicals are normalized by lipid content. Numerous other studies have
confirmed the role of lipid content in the bioconcentration and bioaccumulation of organic
chemicals by aquatic organisms (e.g., Baron, 1990; van den Heuvel, et al., 1991; Leblanc, 1995;
Stow et al., 1997). The lipid compartment is fundamental to equilibrium partitioning theory and to
most bioaccumulation models of organic chemicals, wherein bioconcentration is described as a
chemical partitioning process between the lipid and water compartments (e.g., Mackay, 1982;
Barber et al., 1991; Gobas, 1993; Thomann, 1989; Di Toro et al., 1991). Although other
compartments are assumed to exist in aquatic organisms (e.g., interstitial water, nonlipid
biological material), partitioning to lipids becomes increasingly important as chemical
hydrophobicity increases.
Recognition of the importance of lipids when assessing and predicting bioaccumulation
of nonionic organic chemicals has led to the practice of normalizing chemical concentrations in
tissue by lipid content. Lipid normalization, which is the process of dividing the total
concentration of a chemical in tissue by the fraction of the tissue that is lipid (f), is usually
performed to account for variation in bioaccumulation between species (or individuals within a
species) that results from differences in lipid content alone. Although quantifying chemical
concentrations in lipids would be a direct measure of chemical partitioning to lipids, it is
technically difficult to do this because of the diffuse nature of lipids in tissues of aquatic
organisms. Lipid normalization has been conducted since at least the 1980s for deriving AWQC
4-1
-------
for the protection of human health (USEPA, 1980), and more recently in developing Equilibrium
Partitioning Sediment Benchmarks (USEPA, 2000b).
In the 2000 Human Health Methodology, EPA continues to recommend that BAFs be
adjusted by the fraction of tissue that is lipid in order to account for differences in
bioaccumulation that result from variation in lipid content among aquatic species (USEPA,
2000a). As depicted by Figure 3-2, BAFs are adjusted by lipid fraction (f.) in two separate steps.
In the first step, data on the fraction of tissue that is lipid (f.) and data on the freely dissolved
chemical concentration in water (C ") are used to calculate a baseline BAF from a field-measured
BAFx or a lab-measured ECF{. This step is illustrated by Equation 3-1 for field-measured BAF^s.
Here, lipid normalization is conducted to enable more precise estimates of BAFs across multiple
sites and species within a trophic level by accounting for the confounding influence of lipid
variability on BAF^s. In the second step, the final baseline BAF calculated in step 1 is converted
to a national BAF that reflects the lipid fraction of commonly consumed aquatic organisms (and
the fraction of chemical that is freely dissolved calculated for U.S. surface waters). This step is
illustrated by Equation 3-2 in Section 3.2.3.
4.1.2 Assumptions and Limitations
Although theory and empirical evidence support the concept of adjusting BAFs and
BCFs by lipid content to facilitate their extrapolation between species and sites, this practice
nevertheless involves making a series of assumptions that deserve to be explicitly stated and
evaluated. These assumptions can be stated as:
1. For a given species and exposure condition, the total concentration of a nonionic
organic chemical in the tissue of an organism at or near steady state varies in direct
proportion to the lipid content in the tissue of interest.
2. The degree of proportionality of chemical concentration with lipid content does
not depend on the amount or composition of lipids present in tissue.
As described in Section 4.1.1, the first assumption is generally supported by the empirical
evidence and underlying theory that supports many widely used bioaccumulation models. This
assumption is also supported by the findings that for organic chemicals that are not metabolized,
BCF is strongly correlated with Kow. (e.g., Veith et al., 1979b; Isnard and Lambert, 1988; de Wolf
et al., 1992). In determining Kows, w-octanol is considered to be a surrogate for lipid. Chiou (1985)
used triolein (glyceryl trioleate) as a surrogate for lipid and also found good agreement between
BCFs and triolein/water partition coefficients. Evidence of the utility of lipid normalization is
presented in Section 5.1.3, Figure 5-2, where it is shown that normalization by the fraction of
tissue that is lipid (f.) and the fraction of chemical in water that is freely dissolved (ffd)
substantially reduces variation in BAF^s.
Although the general utility of lipid normalization has been well established, this
adjustment does not account for all of the variation in BAFs that may occur. Bioaccumulation
can be affected by other factors that differ between species, such as composition of diet, growth
4-2
-------
rate, chemical metabolism, and trophic position. Ecosystem factors, such as chemical loading
history, food web structure, and bioavailability also contribute to variation in BAFs (EPA's
calculation of baseline and national BAFs address some differences in trophic position and
bioavailability). Therefore, the effectiveness of lipid normalization in reducing variability in BAFs
and BCFs is likely to be greatest when conducted between species (or individuals within a
species) that are substantially different in lipid content but have experienced similar chemical
exposure conditions. In situations where the difference in lipid content between species is
minimal, or when the aforementioned factors (e.g., loadings history, food web structure) differ
substantially between sites, the efficacy of lipid normalization may be substantially reduced or
masked. Such situations may have contributed to reports of little or questionable benefit derived
from lipid normalization in some field studies (e.g., Amrhein et al., 1999; Bergen et al., 2001).
Other procedures, such as analysis of covariance, have been proposed to improve the statistical
basis of lipid normalization (Hebert and Keenleyside, 1995). When sufficient data exist, analysis
of covariance may improve in the statistical basis for lipid-normalizing BAFs. However, the
limited data associated with typical BAF/BCF studies often restrict application of this approach
for deriving national BAFs.
The second assumption pertains to the utility of the total lipid content as a normalizing
factor for species and tissues with widely varying lipid fractions and lipid compositions. The
process of normalizing BAFs and BCFs on the basis of the total fraction of tissue that is lipid
assumes that lipids are a single, uniform compartment. In reality, total lipid content in fish
includes different lipid classes, including relatively polar phospholipids, which are common in
cell membranes, and generally nonpolar triacylglycerols, which are common in storage lipids
(Henderson and Tocher, 1987). The variation in lipid-partitioning behavior of nonionic organic
chemicals is thought to be a function of differences in polarity of lipid classes, as fewer chemicals
become associated with the more polar "membrane-bound" lipids than storage lipids (Ewald and
Larsson, 1994; van Wezel and Opperhuizen, 1995; Randall et al., 1998).
In practical terms, the potential impact that differences in lipid composition might have on
chemical partitioning and lipid normalization seems to be most relevant for very lean tissues (e.g.,
those less than l%-2% total lipids). This suggestion is based on observations that lean tissues of
some fish species contain a much greater proportion of polar phospholipids (24%-65%) than do
"fatty" tissues (1.5%-8.7%; Ewald and Larsson, 1994). Similar observations have been made
with populations of ribbed mussels, for which Bergen et al. (2001) reported significantly higher
fractions of polar lipids in leaner populations compared with fatter populations. Because of their
greater polarity with respect to lipid content, very lean tissues are likely to exhibit different
chemical/lipid-partitioning behavior than fatty tissues. Bergen et al. (2001) reported stronger
correlations between chemical concentrations and mussels with higher total (and nonpolar) lipid
content, which led to their suggestion that lipid normalization may work best above some
threshold of lipid content. However, the narrow range of lipid content evaluated in their study
(about a factor of two) and the reliance on total PCB measurements (as opposed to individual
congeners) might have limited their ability to identify meaningful trends between chemical
concentrations and lipid content.
4-3
-------
Differences in lipid composition in tissues of aquatic organisms also relate to a
complication associated with methods used to determine lipid content. Specifically, different
solvents have been used to extract lipids, which leads to different quantities (and types) of lipid
being extracted from the same tissue of aquatic organisms. In a study by Randall et al. (1991),
lipid fraction varied by nearly fourfold among four extraction methods but varied twofold or less
among two of the more common extraction methods (chloroform-methanol and acetone-
hexane). Following up on their previous work, Randall et al. (1998) report that if different
solvents are used to extract lipids and PCB congeners, differences among lipid-normalized
concentrations can vary more than fivefold, depending on the solvent combination. The relative
difference among lipid extraction methods depends not only on the polarity of the solvent but
also the lipid content of the tissue. Because lean tissues contain proportionally more polar lipids
than fatty tissues, differences in the lipid extraction efficiency for different solvents tend to be
greatest for lean tissues (de Boer, 1988; Ewald et al., 1998). This finding led these authors to
caution the use of lipid data from lean tissues that have been extracted using strictly nonpolar
solvent systems. Notably, other attributes (e.g., high temperature, pH, lipid decomposition due to
exposure to light and oxygen) can also affect lipid extractions, but these have been less studied
than has extraction solvent.
Although a variety of solvent systems that extract various lipid classes have been
proposed for use in normalizing tissue chemical concentrations by lipid content, a clear
consensus has not emerged on which method is most appropriate for all tissues, species, and
nonionic organic chemicals. Although it is desirable to have one standardized lipid extraction
method for normalizing concentrations of nonionic organic chemicals, it seems possible that no
single method would be equally appropriate for all chemical and tissue types, because different
tissues have different lipid compositions that, in turn, may alter the chemical/lipid partitioning
process. From a toxicological perspective, the science is not presently clear on which classes of
lipids (e.g., phospholipids, free fatty acids, mono-, di-, and triglycerides) are most relevant with
respect to different organic chemicals. For example, DDT has been reported to bind to relatively
polar membrane-bound lipids, which suggests membrane lipids might be relevant to DDT
toxicity (Chefurka and Gnidec, 1987). Randall et al. (1998) reported that 27% of extractable PCBs
were associated with the more polar, membrane-bound lipid pool (i.e., extractable with
chloroform/methanol), whereas 73% were associated with the neutral lipid pool (i.e., extractable
with hexane). Similarly, de Boer (1988) reported that chlorobiphenyls were associated with both
bound (membrane) and unbound (storage) lipid pools in fish. These findings further suggest that
membrane-bound lipids should not be ignored when selecting lipid extraction methods.
To promote consistency in measuring BAFs and BSAFs in field studies, EPA
recommends the continued use of the Bligh and Dyer (1959) chloroform/methanol extraction
method (or the less toxic solvent system of Kara and Radin (1978), in which hexane/isopropanol)
in combination with gravimetric measurement of lipid. The Bligh-Dyer method is recommended
because it is widely used for lipid measurements and has been well characterized in terms of the
types of lipids extracted. The Bligh-Dyer method also extracts both polar and nonpolar lipids.
Based on these and other considerations, Randall et al. (1998) also recommend the Bligh-Dyer
method as a standard technique for total lipid extraction pending more research to identify the
complex neutral chemical/lipid relationships and subsequent development of a definitive standard
4-4
-------
method. Randall et al. (1998) also recommend that if other lipid extraction methods are used,
results should be compared to results obtained using the Bligh-Dyer method to allow conversion
of the results to Bligh-Dyer equivalents. When using exiting data on lipid fraction, EPA may
consider it appropriate to exclude certain data when differences in baseline BAFs or BCFs are
substantial and are believed to be caused largely by differences in lipid extraction methods.
4.2 TECHNICAL BASIS OF FREELY DISSOLVED NORMALIZATION OF
CHEMICAL CONCENTRATION IN WATER
The 2000 Human Health Methodology for deriving trophic-level specific national BAFs
for nonionic organic chemicals uses baseline BAFs in an intermediate step. The baseline BAFs
are based on the concentration of chemical that is freely dissolved in water (C™) and the fraction
of the organism that is lipid (f). EPA uses these adjustments because they express the BAF on a
thermodynamic or fugacity basis and allow better extrapolation of BAFs from one ecosystem to
another.
By basing the baseline BAFs on C ", EPA does not ignore the chemical-associated
dissolved organic carbon (DOC) and particulate organic carbon (POC) in the water column. As
discussed in the following sections, a chemical associated with DOC and POC in the water
column is assumed to be in equilibrium with the chemical freely dissolved in the water column
(an assumption made by EPA; see Section 4.2.3). Therefore, any additions or removal of
chemical from any of the three phases (i.e., freely dissolved chemical, chemical associated with
DOC, and chemical associated with POC) will cause a re-equilibration of the chemical among the
three phases. Due to the equilibrium conditions among these three phases, the chemical
concentration in the water column expressed using any of the three phases, individually or in
combination, is indicative of the chemical concentrations in the other water column phases for a
given set of ecosystem conditions. Therefore, a BAF could be based on any combination of the
three phases and include the influences of the other water column phases.
The relationship among the freely dissolved chemical and the chemical associated with
DOC and POC, presented below, assumes equilibrium among these phases. For a given
ecosystem, DOC and POC define the partitioning of the chemical among the three phases.
National BAFs, calculated from the baseline BAFs, require both the average lipid content offish
and shellfish consumed by the U.S. population as well as average DOC and POC values for the
nation's waters. These required parameters result in expression of national BAFs on the basis of
the weight offish/shellfish tissue and total chemical concentration in the water column,
i.e., (micrograms of chemical per kilogram of wet tissue)/ (micrograms of chemical per liter of
water).
4.2.1 Background Theory and Basic Equation
Experimental evidence shows that hydrophobic organic chemicals exist in water in three
phases: (1) the freely dissolved phase, (2) sorbed to suspended solids (particulate organic
carbon), and (3) sorbed to dissolved organic matter (Hassett and Anderson, 1979; Carter and
Suffet, 1982; Landrum et al., 1984; Gschwend and Wu, 1985; McCarthy and Jimenez, 1985a;
4-5
-------
Eadie et al., 1990, 1992). The total concentration of the chemical in water is the sum of the
concentrations of the freely dissolved chemical and the sorbed chemical (Gschwend and Wu,
1985; USEPA, 1993):
Cj = Cjd + POC • C + DOC • C
'doe
where:
c;
pfd _
^poc
Cdoc =
POC =
DOC =
(Equation 4-1)
total concentration of chemical in water
concentration of chemical that is freely dissolved in water
concentration of the chemical partitioned to the particulate organic
carbon in the ambient water
concentration of the chemical sorbed to the dissolved organic carbon
in the water
concentration of parti culate organic carbon in water (kilograms of
paniculate organic carbon per liter of water)
concentration of dissolved organic carbon in water (kilograms of
dissolved organic carbon per liter of water)
The above equation can also be expressed using partitioning relationships as:
= Cwfd ' (1 + POC
DOC
(Equation 4-2)
where:
K
poc
K
poc
C /Cfd
VxpQQ / V^\v
C /Cfd
V_x ^JQQ / V_x yy
equilibrium partition coefficient of the chemical between POC phase
and the freely dissolved phase of water (liters of water per kilogram of
particulate organic carbon)
equilibrium partition coefficient of the chemical between DOC phase
and the freely dissolved phase of water (liters of water per kilogram of
dissolved organic carbon)
From Equation 4-2, the fraction of the chemical that is freely dissolved in the water can be
calculated using the following equation:
K
doc
4-6
-------
1 + POC • K^ + DOC • K^
(Equation 4-3)
Experimental investigations by Eadie et al. (1990, 1992), Landrum et al. (1984), Yin and
Hassett (1986, 1989), Chin and Gschwend (1992), and Herbert et al. (1993) have shown that Kdoc
is directly proportional to the Kow of the chemical and is less than the Kow. When measured values
of Kdoc are not available, it can be estimated using the following equation:
* 0.08
(Equation 4-4)
Experimental investigations by Eadie at al. (1990, 1992) and Dean et al. (1993) have
shown that Kpoc is approximately equal to the Kow of the chemical. When measured Kpoc values
are not available, it can be estimated using the following equation:
(Equation 4-5)
By substituting Equations 4-5 and 4-6 into Equation 4-4, the following equation is
obtained:
1
1 + POC • Kw + DOC • 0.08 • K^
(Equation 4-6)
Burkhard et al. (1997) evaluated the utility of using Equation 4-6 to derive baseline BAFs
that are applicable to multiple sites. In their study, Burkhard et al. (1997) measured BAFs for
various chlorinated butadienes, chlorinated benzenes, and hexachloroethane for three species of
forage fish and blue crab in Bayou d'Inde of the Calcasieu River system in Louisiana. Using
Equation 4-6, field-measured BAFs were converted to baseline BAFs and compared to baseline
BAFs determined for other trophic level three species in two other field studies (Pereira et al.,
4-7
-------
1988; Oliver andNiimi, 1988). One of these field studies, by Pereira et al. (1988), was conducted
in different sites within the Calcasieu River system; the other study, by Oliver and Niimi (1988),
was carried out in Lake Ontario. Burkhard et al. (1997) found no significant difference between
baseline BAFs determined in their own study and those determined by Pereira et al. (1988)
(Tukey's, • = 0.05). However, for one chemical (hexachlorobutadiene), a difference between the
two studies of about 1 order of magnitude was observed in the baseline BAFs. Burkhard et al.
(1997) further noted that their own baseline BAFs were not substantially different from those
derived for similar trophic level fish in Lake Ontario, suggesting broader applicability of properly
derived baseline BAFs.
The EPA's BAF methodology incorporates four decisions/assumptions associated with
the three-phase partitioning model for estimating the concentrations of nonionic organic
chemicals that are freely dissolved in ambient waters. These four decisions/assumptions are:
1. Sorption of the chemical to DOC and POC reduces chemical bioavailability to
aquatic organisms.
2. Chemicals in the freely dissolved phase of the water are in equilibrium with
chemical associated with the DOC and POC (including plankton) phases of the
water column.
'i ~K =~K
•J • ^poc ^ow
4. Kdoc = 0.08-Kow
These assumptions are based on experimental evidence referenced above. Detailed
discussions of the evidence and other information supporting these assumptions are presented in
the following subsections: assumption 1 in Section 4.2.2, assumption 2 in Section 4.2.3, and
assumptions 3 and 4 in Sections 4.2.4 and 4.2.5.
4.2.2 Effects of Chemical Sorption to DOC and POC on Chemical Bioavailability
Numerous reports demonstrate the partitioning of hydrophobic nonionic organic
chemicals to POC and DOC (see Section 4.2.4). Concurrent with the research on partitioning of
hydrophobic nonionic organic chemicals to POC and DOC, research efforts have focused on
bioavailability of hydrophobic nonionic organic chemicals to fish and other aquatic organisms in
the presence of DOC and POC. The results of this research show that the concentration of
chemical that is freely dissolved in sediment porewaters and ambient surface waters is the best
measure currently available of the fraction of nonionic organic chemicals available for uptake by
aquatic organisms (Suffet et al., 1994; DiToro et al., 1991).
Reduced chemical uptake by aquatic organisms in the presence of DOC has been
extensively reported for both ambient waters and waters containing added DOC (Leversee et al.,
1983; Landrum et al., 1985; McCarthy and Jimenez, 1985b; McCarthy et al., 1985; Carlberg et al.,
1986; Black and McCarthy., 1988; Servos and Muir, 1989; Kukkonen et al., 1989). For example, it
4-8
-------
has been reported that the percentage reduction in gill uptake efficiency of benzo[a]pyrene and
2,2',5,5'-tetrachlorobiphenyl in rainbow trout is equal to the percentage reduction in freely
dissolved chemical concentration in the presence of DOC (Black and McCarthy, 1988). The
authors of this study concluded that only the chemical that was freely dissolved in the water was
available for uptake by the fish. Similarly, Landrum et al. (1985), McCarthy et al. (1985), and
Servos and Muir (1989) reported that chemical uptake rates were reduced when DOC was present
and that the concentration of chemical that is freely dissolved in the water column decreases in
proportion to the amount of DOC present in the water. These studies clearly support EPA's
assumption that chemical bioavailability of nonionic organic chemicals to aquatic organisms is
reduced in the presence of DOC and POC. Excellent reviews on the science of bioavailability are
provided by Hamelink et al. (1994) and Kukkonen (1995).
There are a few reports in the scientific literature of increases in the bioavailability of
nonionic organic chemicals to aquatic organisms in the presence of low concentrations of DOC
(see Haitzer et al., 1998). In their review, Haitzer et al. (1998) compared BCFs determined using
laboratory waters with those determined using lake waters and laboratory waters with added
DOC as the exposure media. When BCFs derived from laboratory water experiments were
smaller than those derived from the other waters, the authors concluded that increased
bioavailability had occurred. The EPA believes that some of these findings are artifacts of the
experimental design. For example, Haitzer et al. (1998) reported that bioavailability of hepta- and
octa-chlorodibenzo-p-dioxins to rainbow trout was enhanced when DOC was low. Careful
examination of the original report by Servos et al. (1989), however, reveals that the solubility
limits for the hepta- and octa-chlorodibenzo-p-dioxins were exceeded in the experiment, and
therefore any conclusions about increased bioavailability are clearly suspect. The EPA also
believes that other factors could explain reports of apparent increases in bioavailability. Verhaar
et al. (1999) and others have pointed out that in performing any experiment from which BCFs
will be derived, the organisms will introduce DOC—for example, from mucous layers, feces, and
urine—into the aqueous phase. Because the measurements made to determine BCF do not
typically measure 'bioavailable' chemical (i.e., the concentration of chemical that is freely
dissolved) but rather the total concentration of chemical that is in the exposure water, addition of
DOC by the organisms during the experiment most certainly confounds the assumption that
DOC concentrations were actually low. It is entirely possible that the concentration of chemical
that was freely dissolved in experiments using laboratory waters was substantially different from
that in experiments using lake waters and laboratory waters with added DOC. A recent report by
the authors of the 1998 review (i.e., Haitzer et al.) supports EPA's belief that the increased
bioavailability in the presence of low concentrations of DOC for BCF measurements is caused by
experimental artifacts. After very careful study of BCF measurements performed with low DOC
concentrations, Haitzer et al. (2001) concluded that"... BCF enhancements that have been
reported in the literature are more likely the result of random, experimental variations than the
result of systematic enhancement of bioconcentration."
On the basis of the information presented above, EPA has assumed that the
bioavailability of nonionic organic chemicals to aquatic organisms is reduced in the presence of
DOC and POC. EPA acknowledges that there are a few reports of increased bioavailability in the
4-9
-------
scientific literature and believes that the causes of the increased bioavailability are, in all
likelihood, experimental artifacts of the BCF measurements.
4.2.3 Sorptive Behavior of Nonionic Organic Chemicals with DOC and POC (Including
Plankton) in the Water Column
In using the three-phase partitioning model to determine the concentration of chemical
that is freely dissolved in the water column, EPA has assumed that the chemical freely dissolved
in the water column is in equilibrium with the chemical associated with DOC and POC. The basis
for this assumption is presented below.
In the development of the fluorescence quenching technique for measuring partitioning
between the freely dissolved chemical and DOC, investigators have studied the time required for
PAHs to equilibrate with DOC. Gauther et al. (1986), McCarthy and Jimenez (1985a), and
Schlautman and Morgan (1993) have reported times ranging from less than 1 minute to
approximately 10 minutes for PAHs to equilibrate with DOC. These very short equilibration
times suggest that equilibrium conditions should exist between nonionic organic chemicals and
DOC in the environment.
Kpoc data are quite limited, however, and EPA is unaware of any research efforts studying
the kinetics of partitioning of nonionic organic chemicals between the POC and the freely
dissolved phases of water. Insights into the behavior and kinetics of partitioning with POC can be
gained by examining the experimental evidence on partitioning of nonionic organic chemicals in
sediments/soils. Karickhoff et al. (1979) and Gschwend and Wu (1985) have shown that sorption
and desorption of nonionic organic chemicals to sediment and soil organic carbon are reversible.
In the 1980s, most investigators believed that time periods on the order of hours to a few days
were required for chemical to equilibrate between the freely dissolved and organic carbon phases
(Tomson and Pignatello, 1999). More recently, it had been found that attainment of steady-
state/equilibrium conditions in these systems takes substantially longer periods of time (e.g.,
upwards of 100 days), and the time period is dependent on the concentration of suspended solids
in the system (Jepsen et al., 1995).
Numerous investigations have studied the kinetics of sorption and desorption of nonionic
organic chemicals to sediment and soil organic carbon, and these studies suggest the existence of
fast and slow sorption and desorption phases (Pignatello and Xing, 1996). The desorption process
can be characterized as having a fast initial release of chemical followed by a slow, prolonged
release of the chemical. Numerous models have been developed to explain this behavior (Chen et
al., 1999). Many investigators have modeled the sorption process at the surface of the organic
carbon as a quick equilibrium process between the organic carbon surface and the chemical freely
dissolved in the water column (Schwarzenbach et al., 1993). Examples include the retarded
diffusion model of Lick and Rapaka (1996) and the radial diffusion model of Wu and Gschwend
(1986).
There is no clear consensus on a kinetic model for describing the partitioning of nonionic
organic chemicals between POC (or sediment/soil particles) and the freely dissolved phases (see
4-10
-------
Chen et al. [1999] for a listing of models). Such a model would have to account for both fast and
slow sorption/desorption processes. Given the limited amounts of Kpoc data, as well as kinetic
data for sorption/desorption processes with POC, the selection of an appropriate kinetic model is
clearly problematic. From an uncertainty standpoint, modeling nonequilibrium conditions using
equilibrium condition assumptions would cause the freely dissolved concentration of the
chemical to be too small, because the Kpoc for kinetic conditions would be less than that for
equilibrium conditions.
In some situations, the concentration of chemical that is determined to be freely dissolved
using the three-phase model might be too large. As discussed by Gustafsson et al. (1997), PAHs
partition more strongly to soot (i.e., organic carbon derived from incomplete combustion) than to
organic carbon in sediments, whereas other chemical classes, such as PCBs, do not appear to be
influenced by the soot phase. Unfortunately, the soot contents of natural waters are largely
unknown. In situations where significant amounts of soot exist, the three-phase model could be
modified to include a fourth phase consisting of soot. Gustafsson et al. (1997) describe a
methodology for estimating the partition coefficients for soot.
By definition, POC is material retained by filtering or by centrifugation. Therefore, POC
includes plankton. Because EPA assumes that the chemical associated with POC is in equilibrium
with the chemical freely dissolved in water, chemical in plankton retained by the filter must be in
equilibrium with the chemical freely dissolved in water. Under certain conditions—for example,
algal blooms—the plankton is probably not in equilibrium with the chemical freely dissolved in
water. However, because EPA's basis for deriving AWQC involves long-term average, steady-
state, or near steady-state conditions, EPA believes that it is reasonable to assume that chemicals
associated with plankton (in the POC) is on average in equilibrium with the chemicals freely
dissolved in water. It should be noted that larger plankton is not included in POC samples
because a prefiltering step is generally used to remove larger particulates before filtering or
centrifuging to separate POC; see, for example, Broman et al. (1991), who used a 100-|im
prefilter to define the upper size cutoff for POC.
The EPA, in using the three-phase model for determining the concentration of chemical
that is freely dissolved, assumes equilibrium exists between the chemical associated with the
POC and the chemical that is freely dissolved. This assumption is based on the consideration that
POC in the environment is constantly exposed to the chemical of interest, and the sorption and
desorption processes for the chemical to and from the POC are dominated by fast-phase kinetics
specifically, the quick equilibrium process occurring at the surface of the organic carbon
(Schwarzenbach et al., 1993). Because of fast-phase kinetics, short-term fluctuations in ambient
concentrations are quickly accounted for in natural waters. Furthermore, AWQCs are developed
with the assumption that conditions in ambient waters are representative of long-term averages,
which are best captured using steady-state or near steady-state conditions. EPA believes that the
three-phase partitioning model provides a reasonable approximation of these types of conditions.
Errors associated with calculation of the freely dissolved concentration are somewhat
offset by using the three-phase partitioning model twice in EPA's methodology for developing
national BAFs. First, a baseline BAF is calculated from a measured BAFx wherein the measured
4-11
-------
BAFj is lipid- normalized and corrected for bioavailability considerations using the fraction of
chemical that is freely dissolved. Second, a national EAF{ is calculated using the average lipid
content for species consumed in the United States and the fraction of chemical that is freely
dissolved in U.S. waters. EPA believes that use of the freely dissolved concentration for
converting both to and from the intermediate baseline BAF value, that is, from measured BAF-to-
baseline BAF and again from baseline BAF-to-national BAF offsets the error associated with
calculating the freely dissolved concentration.
Given the above considerations, EPA has decided to use the three-phase partitioning
model with the assumption of equilibrium conditions for the calculation of the freely dissolved
concentrations for nonionic organic chemicals for the following reasons:
1. Available data indicate complete, rapid partitioning to DOC;
2. Initial partitioning to POC is also rapid and near complete and although some
kinetic limitations on chemical partitioning with POC occur, these are not likely to
be important on the time scale applicable to human health water quality criteria;
3. No consensus exists on available kinetic models specific to POC; and
4. Use of the freely dissolved fraction twice in derivation of national BAFs offsets
model error.
4.2.4 Values for the Particulate and Dissolved Organic Carbon Partition Coefficients Kpoc
and Kdoc
In using the three-phase partitioning model for calculating the fraction of a chemical's
concentration that is freely dissolved in water (ffd), EPA will define Kpoc and Kdoc as follows:
Kpoc = Kow with 95% conference limits of a factor of 8 in either direction.
Kdoc = 0.08 • Kow with 95% confidence limits of a factor of 20 in either direction.
The basis for these relationships is presented below.
The separation of POC from DOC in water samples is operationally defined by filtering or
centrifugation. With both techniques, the operational cutoffs between POC and DOC fractions
can differ depending upon membrane selection and hardware; for example, a membrane with a
0.45-|im cutoff may be used in one study, whereas centrifugation that retains all particles with a
size of 1.0 jim or greater may be used in another study. Typically, the size cutoff between POC
and DOC fractions is 0.1-1 jim. DOC is principally composed of carbohydrates, carboxylic acids,
amino acids, hydrocarbons, hydrophilic acids, and humic and fulvic acids. POC is principally
composed of some larger humic acids, microbes, small plankton, plant litter, and ligneous matter
(Suffet et al., 1994; Thurman, 1985). The material retained by filtration or centrifugation is the
POC fraction. Organic carbon and chemical-specific analyses are performed on the POC fraction
4-12
-------
to determine POC and Cpoc, respectively. The DOC fraction is defined as the ambient water
remaining after filtration or centrifugation is performed. The DOC fraction contains both the
chemicals that are freely dissolved and the chemicals associated with the DOC. To determine the
concentration of chemical that is freely dissolved in the DOC fraction, a variety of analytical
techniques are available, for example, fluorescence quenching, purging or sparging techniques,
solid phase microextraction (SPME), equilibrium dialysis, solubility enhancement, ultrafiltration,
reverse-phase separation, size exclusion chromatography, and liquid-liquid extraction. Some of
these techniques directly measure the concentration of chemical that is freely dissolved, whereas
others physically separate the DOC-bound chemical fraction from the freely dissolved chemical
fraction.
All of the methods for measuring freely dissolved chemical in water have limitations that
can lead to uncertainties in the Kpoc and Koc. However, limitations associated with some methods
can lead to larger uncertainties than others. The methods that appear to have smaller biases are
sparging, fluorescence quenching, SPME, and possibly equilibrium dialysis. An excellent
discussion on the individual techniques (except SPME) and their limitations is presented by
Suffet et al. (1994). The reader can refer to Poerschmann et al. (1997) or Ramos et al. (1998) for
further information on the SPME method.
A review of the scientific literature reveals that Kpoc measurements are not as prevalent as
Kdoc and Koc measurements. Kocis defined as the partition coefficient of the chemical
concentration on soils or sediments (on an organic carbon basis) to the chemical concentration in
water after the removal of the solid phase. Koc is expressed as liters of water per kilogram of
organic carbon. Kpoc is defined as the partition coefficient of the chemical concentration on the
water column particles (on an organic carbon basis) to the chemical concentration freely
dissolved in the water. Kpoc is expressed as liters of water per kilogram of organic carbon.
Measured Koc values should not be assumed to be equal to Kpoc, because (1) the types of organic
carbon in the soils and sediments can be very different from those in the water column, and
(2) their denominators are different. Sediments and soils tend to be more weathered than water
column particulates, because the latter include organic matter derived from sources such as
recently deceased as well as live plankton and algae and fecal matter from aquatic organisms.
However, in some cases, the organic matter composing the Koc and Kpoc might be very similar,
because sediment resuspension and erosional inputs could be responsible for a majority of the
particulates in the water column. In all cases, the chemical concentrations in water used in
measuring Koc and Kpoc are different. The determination of Kpoc is based on the concentration of
chemical freely dissolved in the water (see Section 4.2.1 for derivation), whereas Koc is
determined by using an operational definition of "dissolved" in water. This operational definition
includes both freely dissolved chemical and chemical sorbed to DOC in the aqueous phase.
Data for Kpoc are limited for a number of reasons. First, the measurement of Kpocin field
situations is often very difficult because of the extremely low concentrations of hydrophobic
pollutants in natural waters, often 1 ppt or less on a total basis. With low concentrations, large
volumes of water must be processed in order to obtain enough of the chemical to measure. For
example, Broman et al. (1991) processed approximately 2,000 L of Baltic Sea water to obtain
measurable amounts of poly chlorinated dioxins and furans on the particulates retained by
4-13
-------
filtering and in the water passing through the filter. Second, the techniques developed for
measuring freely dissolved concentrations of chemical in natural waters are not amenable to field
sampling situations in which large volumes of water need to be processed. Third, in laboratory
studies, many investigators use sediment particles as a surrogate for naturally occurring water
column particulates, and it is somewhat tenuous to assume that sediment particles are equivalent
to water column particulates. Fourth, because of operational and analytical factors, Kocs and Kds
(Kocs expressed on the basis of dry weight rather than organic carbon on the solids,
i.e., Koc = Kd/foc) are more often reported than Kpoc values, because Kocs and Kds are much easier
to determine.
Kpoc measurements found in a search of the scientific literature are reported in Table 4-1
and plotted in Figure 4-1. These values were determined by using the reverse-phase, sparging,
and ultrafiltration method with samples primarily from Great Lakes ecosystems. An equation of
the form log Kpoc = a + b • log Kow was computed by using the geometric mean regression
technique (Ricker, 1973); this equation is:
log Kpoc = +1.19 (±2.18)+ 0.81 (±0.11)'log Kow df= 14, r = 0.84, Sxy = 0.40
The geometric mean regression technique was used because the X variable (log Kow) was
measured with error. The equation and its 95% confidence limits for any single predicted log Kpoc
are plotted in Figure 4-1. The slope of this regression line is not significantly different from 1.0 (•
= 0.05). Assuming a slope of 1.0 results in an equation of the form log Kpoc= log Kow+ B. This
equation, by rearrangement, results in B = log Kpoc- log Kow= log (Kpoc/Kow), and "B" can be
found by averaging the differences of the log Kpoc and log Kowfor the individual chemicals or by
averaging the logarithms of the ratio of the Kpocto Kow for the individual chemicals. For this data
set, an average difference (standard deviation, number of data points) of 0.023 (0.426, 16) was
obtained. Transforming the average difference to an antilog scale results in predictive relationship
of Kpoc= 1.05 • Kow with 95% confidence limits of a factor of 8 [ 10(standard deviation' * = 5%> #= 15» =
j0(0.426- 2.i3i)] in either direction from the mean predicted Kpoc.
Based on the data presented above, EPA will use the following relationship for
determining Kpoc values for use in the three-phase partitioning model:
Kpoc =1.0* Kow with 95% confidence limits for a predicted Kpoc of a factor of 8 in either
direction from the mean predicted Kpoc.
4-14
-------
o
1,
s o
—
2
M
M
•a
o
a.
o
U
Os Os Os Os
oo oo oo oo
Os Os Os Os
+-T +-T +-T +-T
OOOOOOOOOOOOOOOOOOOOOOOOOOOO
OSOSOSOSOSOSOSOSOSOSOSOSOSOS
O O O O O O
Os Os Os Os Os Os
Os Os Os Os Os Os
(L>
c/a c/a c/a c/a c/a c/a c/a c/a c/a c/a c/a
>>>>> >>>>>
Q
-"-H
CS
e
^^^^^^^^^^^^^^
oooooooooooooo
^^^^^^^^^^^^^^
mmmmmmmmmmmmmm
T3T3T3T3T3T3T3T3T3T3T3T3T3T3
'g'3'3'3'3'3'3'3'3'3'3'3'3'3
oooooooooooooo
t- t- t- t- t- t-
fNCNfNCNfN
000000 00000 0
o ON "* oo "* oo >nt-O'tO'/itst-
t- t- t- m ts t- vovo^ooooooootS
osososos
OOOOOOOO
in
OO
vot-'*t-t-'/itsoooO'*m'*fS
TtOtSV£)VOt-vOtSfSmt-OOVOO\
•G -G .e .e
3 3 PH PH
fi fi o o
222222
O O O O O O
o
i2 « « « « « « « « « « « « «
cncnc/ic/ic/ic/ic/ic/ic/ic/ic/ic/ic/ic/i
OOOOO
•§
I
.
T*H
'
ll
HH
T*H T*H T*H T*H *S *S "^ "^ O
-S-S-S'S 2 2 § § °
--g-g-SUlla
s s s a a a a ^
H H H H - rf
- a .& .& .&
*•
s s s s
1 a!
Si;
'
-§ 3
•
-g -g -g
fNCNfNCNfNCNfNCNfNCNfNCNfNCN
-------
The data used to derive the above relationship were primarily from the Great Lakes ecosystem.
Kpoc data from Butcher et al. (1998) for the Hudson River, which are not used in deriving the
predictive relationship, are also plotted in Figure 4-1 (circles) and are in good agreement with the
Kpoc data from the Great Lakes. The comparability of the Hudson River data to those obtained
from the Great Lakes ecosystem indicates that the above relationship for determining Kpoc has
broader applicability than just the Great Lakes ecosystem.
O)
O 5
.-a'"
-------
coefficients for the relationship are rather small, • 0.3, although the slopes are similar to 1. The
small dependence of Kdoc on Kow for DOC from natural waters is consistent with the findings of
Kukkonen and Oikari (1991) and Evans (1988), which suggest that factors other than the
hydrophobicity are important in the sorption or association of nonionic organic chemicals with
surface water DOC.
Figure 4-2. Kdocs determined using the reverse phase
(circle), equilibrium dialysis (open square), sparging
(plus diamond), calculated/model-derived
(downwards triangle), fluorescence quenching
(upwards triangle), solubility enhancement (open
diamond), biological (plus square), and solid phase
microextraction (open plus diamond) techniques for
DOCs from different sources. The geometric mean
regression and their 95% prediction confidence limits
are plotted.
Humic and fulvic acids
w/o Aldrich humic acid
Soil porewater and
Groundwater
4-17
-------
The Kdoc data from surface water DOC appear to have more variability than that observed
for the other DOC sources (Figure 4-2). On a diagenesis basis, DOC from the water column
might be expected to be more variable than that from sediments and humic and fulvic acids,
because surface water DOC contains detritus from recently deceased plankton and algae,
macrophytes, and so forth. Some of the variability in all five graphs in Figure 4-2 is caused by
differences among the measurement techniques. Comparisons of the reverse-phase and dialysis
techniques by Landrum et al. (1984) and Kukkonen and Pellinen (1994) suggest that differences
of an order of magnitude or more can occur between these two methods, but that typically the
dialysis technique provided values that were a factor of 2-5 higher. Biphenyl with a log Kow of
4.09 provides an example of the differences that are possible between these two techniques.
Landrum et al. (1984) performed side-by-side Kdoc measurements with biphenyl, using water
samples from Lake Erie and Huron River. The Kdoc data derived from the two techniques differed
by a factor of 3 for the Lake Erie samples and by a factor of 34 for the Huron River samples. The
variability in the Kdoc data from these surface water DOC might also be related to the time period
in which the measurements were made. Most of the measurements occurred in the 1980s, when
methods for measuring Kdoc were new and/or evolving.
Table 4-2. Regression Equations for Dependence of Kdoc (Geometric Means) on Kow
DOC Source Geometric Mean Regression Equation n r sxy
Aldrich humic acid log K^.= 0.85 (±0.03)a • log K,,w+ 0.27 (±0.20) 269 0.77 0.52
Humic and fulvic acids without Aldrich log K^ 0.88 (±0.06) • log K^-0.11 (±0.31) 230 0.29 0.65
humic acid
Sediment porewaters
Soil porewaters and groundwaters
Surface waters
All DOC including Aldrich humic acid
Naturally occurring DOC (no Aldrich
humic acid)
log Kdoc= 0.
log Kdoc= 0.
log Kdoc= 0.
log Kdoc= 0.
log Kdoc= 0.
,99 (±0.04) • log
,91 (±0.13) -log
,97 (±0.06) • log
,85 (±0.04) • log
,85 (±0.06) • log
; K^- 0.88 (±0.23)
; K^- 0.22 (±0.68)
jK^- 1.27 (±0.40)
5^- 0.11 (±0.21)
; K^- 0.25 (±0.34)
396 0.64 0.66
47 0.31 0.61
210 0.32 0.99
223 0.78 0.52
127 0.67 0.60
n = number of data points, r = correlation coefficient, sxy = standard error of estimate,a (±standard deviation).
To compare the Kdocs for all DOC sources, Kdoc values for each chemical were averaged
across analytical methods within each DOC source and replotted (Figure 4-3). Average Kdoc
values were used in part because of the unevenness in the numbers of measurements per
chemical. For example, biphenyl and benzo[a]pyrene had 4 and 49 Kdoc measurements,
respectively, for DOC from surface waters. The plot of log Kdoc versus log Kow shows
considerable consistency, and a strong dependence of Kdoc upon Kow is apparent (Figure 4-3). The
average Kdocs for the Aldrich humic acids are on average higher than those derived from natural
sources (Figure 4-3). These results are consistent with the differences in affinities for Aldrich
humic acid and naturally occurring organic carbon reported in the literature for nonpolar organic
chemicals (Suffet et al., 1994).
4-18
-------
X.
o
X.
o
v;. •• ' ^
With Aid rich humic acid _.. <>o ^
o
Without Aldrich humic acid ,.
5 6
K
Figure 4-3. Average K^s for
individual chemicals for different
DOC sources: humic and fulvic
acids (open diamond), sediment
porewaters (downwards triangle),
soil porewaters and groundwaters
(plus square), and surface waters
(upwards triangle). The geometric
mean regression and their 95%
prediction confidence limits are
plotted.
On a theoretical basis, the equation log Kdoc, Kpoc, or Koc= A • log Kow+ B has a slope of 1
when the ratio of the activity coefficients of the chemical in w-octanol to that in the organic
carbon phase is constant for chemicals with different Kows (Seth et al., 1999). The relationships of
log Koc = A • log Kow+ B derived by Seth et al. (1999), DiToro et al. (1991), and Karickhoff (1984)
had slopes of 1 and are clearly consistent with the hypothesis that the ratio of activity coefficients
is constant. Given the above theoretical basis and experimental data, a slope of 1 is assumed for
the relationship in this investigation, that is, log Kdoc = log Kow+ B. This equation, by
rearrangement, results in B = log Kdoc» log Kow = log (Kdoc/Kow), and B can be found by averaging
the differences of the log Kdoc and log Kow for the individual chemicals or by averaging the
logarithms of the ratio of the Kdoc to Kow for the individual chemicals. For the data set consisting
of naturally occurring DOC (no Aldrich humic acid), an average difference (standard deviation,
number of data points) of -1.11 (0.659, 127) was obtained. Transforming the average difference to
an antilog scale results in a predictive relationship of Kdoc = 0.08 Kow, with the 95% confidence
limits of a factor of 20 in either direction from the predicted mean Kdoc. When Aldrich humic
acids are included, an average difference of-0.966 (standard deviation, 0.578; number of data
points, 223) was obtained, resulting in a predictive relationship of Kdoc = 0.11 Kow, with 95%
confidence limits of a factor of 14 in either direction.
On the basis of the above data, EPA will use the following relationship for determining
Kdoc values for use in the three-phase partitioning model:
4-19
-------
The 95% confidence limits for a predicted log Kdoc are:
± standard deviation • t(. = 5o,
5%, df=\TJ)
= ±0.659' 1.979 = ±1.30
and transforming the 95% confidence limits to an antilog basis results in a factor of 20 in either
direction from the mean predicted Kdoc value.
In Figure 4-4, the residuals, that is, measured log Kdocs minus log Kdocs predicted by using
the relationship Kdoc = 0.08 • Kow, along with the 95% confidence limits are plotted. The
distribution of the residuals is normally distributed (• = 10%) with 64 negative and 63 positive
residuals. The residuals for the log Kdocs predicted for naturally occurring DOC have a slight
dependence on the Kow (Figure 4-4).
03
D
;g
-------
approach to selecting Kow values, which EPA developed and had peer reviewed more recently.
Based on EPA's own scientific rationales and the comments received from the peer reviewers,
EPA will select Kow values based on the guidance contained in the more recently developed
protocol. EPA's methodology for selecting Kowvalues divides the range of Kows into three groups
to reflect the differences in chemical properties and behaviors with differing hydrophobicities.
Specific details of the Kow guidance are presented in Appendix B.
4.3 IMPORTANCE OF SEDIMENT-WATER CONCENTRATION QUOTIENT (• socw)
The distribution of a chemical between surface sediments and the water column in an
ecosystem is most effectively described as the sediment-water concentration quotient (• socw),
which is further defined below. BAFs for nonionic organic chemicals are sensitive, in proportion
to hydrophobicity, to differences in the chemical's • socw in an ecosystem. • socw is either
implicitly (field-measured BAFs) or explicitly (modeled BAFs) involved in each of the four BAF
methods described in Section 5. The national BAF procedure for nonionic organic chemicals
involves setting BAFs to the extent possible on the basis of current national average values for
several key parameters, including • socw. In the case of BAF methods 1 and 2, this is done by
averaging measured BAFs from different ecosystems. For methods 3 and 4, a national value for
• socw that was selected by EPA on the basis of high-quality measurements from three different
ecosystems (Section 4.5.1) was used with the Gobas food chain model (Gobas, 1993) to
determine food chain multipliers (FCMs).
Bioaccumulation of hydrophobic nonionic organic chemicals in aquatic organisms is
dependent on a number of ecosystem conditions including food chain length (Rasmussen et al.,
1990), food web composition (Vander Zanden and Rasmussen, 1996; Burkhard, 1998), and the
chemical distribution between sediments and water (Thomann, 1992; Endicott and Cook, 1994).
The impacts of food web composition and chemical distribution between sediments and water are
interrelated because sediments and water are the primary exposure media for the benthic and
pelagic components, respectively, of the food web (Burkhard, 1998). For a benthic food web,
chemical concentrations in benthic invertebrates at the base of the benthic food web are directly
controlled by the concentrations of chemicals in the sediments. Chemical concentrations at the
base of the pelagic food web, for example, in phytoplankton and diatoms, are directly controlled
by the concentration of chemicals in the water. Therefore, differences in distribution of chemical
between sediment and water, as well as differences in benthic versus pelagic food web
composition, will affect the bioaccumulation of nonionic organic chemicals in forage and
piscivorous fish.
The distribution of chemical between the sediment and overlying water in a water body or
a zone of reference within a water body is described by the sediment-water (column)
concentration quotient (• socw), which is defined as:
(Equation 4-7)
4-21
-------
where:
Csoc = concentration of chemical in dry sediment, normalized to sediment organic
carbon
C™ = concentration of chemical that is freely dissolved in water
By expressing the concentration of chemical in sediment on an organic carbon
normalized basis and the concentration of chemical in water on a freely dissolved basis, this
quotient is a measure of the degree to which the chemical's distribution between the surface
sediment and the water column approaches or deviates from a condition of thermodynamic
equilibrium for the water body. The degree of disequilibrium (departure from equilibrium) is
proportional to the degree to which • socw/Kowfor the chemical diverges from a value of 1.0
\ socw J^-ow/-
In the aquatic environment, three factors are primarily responsible for causing • socw to
differ among ecosystems. First, concentrations of nonionic organic chemicals in the water
column and sediment are the result of well-known fate and transport processes, such as particle
sedimentation and resuspension, chemical sorption to and desorption from suspended particles
and the sediments, and ecosystem hydrodynamic properties. These processes vary among
ecosystems. Second, the chemical loading history to the ecosystem plays an important role in its
* socw For example, increasing the loading of a chemical to the water column causes an
immediate rise in the concentration of the chemical in the water, and over time, the concentration
of the chemical in the sediment will gradually increase through sedimentation processes. If the
loading of a chemical to the water column is decreased, the concentration of the chemical in the
water column drops quickly, whereas the concentration of the chemical in the sediments
decreases slowly through burial of older and more contaminated sediments by newer and less
contaminated sediments. Third, differences in organic carbon content in water column
particulates (or suspended solids) and surface sediment vary among ecosystems. The ratio of
organic carbon contents (water column to surface sediment) approximates the steady-state value
of • SOCw/Kow for the ecosystem due to diagenesis processes on the newly deposited surface
sediments.
The importance of chemical loading on • socw is illustrated in Figure 4-5 for three different
loading scenarios: (a) constant loading of a chemical to the ecosystem over time, (b) constant
loading of chemical to the ecosystem with a doubling of loading at year 50, and (c) constant
loading of chemical to the ecosystem with an 80% reduction in loading at year 50. These figures
were created by using a two-compartment mass balance model consisting of a sediment surficial
layer and the water column for a nonmetabolizable chemical with a log Kow of 6, using conditions
and parameters for a large lake ecosystem. In all three loading scenarios, the concentration of the
chemical in the water column responds quickly to the change in loading, in contrast to the
relatively slow response of the concentration of chemical in sediment. In these scenarios,
sediment and water column particulates had organic carbon contents of 3% and 15%,
respectively. In all three scenarios, • SOCW/KOW reaches a plateau of a value of 4.91, nearly equal to
the ratio between the percentage of organic carbon in suspended particulates and surface
4-22
-------
sediments. Many chemicals that have been available as sediment contaminants for many years,
such as PCBs and DDTs, which are now no longer manufactured or used, are often found to be
present in sediments in concentrations that exceed thermodynamic equilibrium with the water
column.
- Water, freely dissolved
Sediment
Elapsed Time (years)
Elapsed Time (years)
Elapsed Time (years)
Figure 4-5. The sediment-water chemical concentration quotient (• socw) for three different chemical loading
scenarios: (a) constant loading of a chemical to the ecosystem overtime, (b) a constant loading of chemical to the
ecosystem with a doubling of loading at year 50, and (c) a constant loading of chemical to the ecosystem with an 80%
reduction in loading at year 50. Simulations performed for a chemical with a log Kow of 6 using Lake Ontario
conditions and parameters.
The latter portion of scenario (c) described above (constant loading of chemical to the
ecosystem with and 80% reduction in loading at year 50) illustrates how • socw changes over time.
Differences in ecosystem parameters and conditions, such as hydraulic retention rates,
sedimentation and resuspension rates, water column and surficial sediment layer volumes, and
chemical loading rates between ecosystems, affect the specific time scales and slopes of the
changes in C ", Csoc, and • socw associated with changes in chemical loading over time.
Ecosystems at thermodynamic equilibrium, a condition that rarely exists in nature, should
theoretically have • socws equal to the chemical's Kow. Consequently, ecosystem models typically
characterize • socw by using its ratio to Kow as a measure of the degree to which the ecosystem is in
disequilibrium (Thomann et al., 1992), or, alternatively, as a measure of the fugacity ratio
(Campfens and Mackay, 1997). A • SOCW/KOW ratio of 1 is equivalent to equilibrium conditions
between the sediments and the water column. A ratio of 25, which has been typical of Lake
Ontario conditions for PCBs and DDTs since the 1970s, is a disequilibrium condition in which
the chemical is enriched in the sediments relative to the water column because of greater loadings
of the chemical to the ecosystem in the past. For ratios less than 1, the chemical is enriched in the
water column relative to the sediments; in this situation, the aquatic ecosystem is being loaded
with the chemical, but sediments have not reached steady state with the water (• socw constant).
With continued loading, sediment contamination increases until a steady-state condition is
reached (• socw constant) and the • SOCW/KOW ratio is in the 2-10 range. The lower bound of 2 arises
4-23
-------
from minimum expected differences in the organic carbon content of particulate matter in the
water column and sediments. The upper bound of 10 allows for the effects of chemical gradients
and greater relative organic carbon amounts in the water column. Green Bay, a fairly shallow and
vertically well-mixed ecosystem receiving a continuous load of PCBs from the contaminated Fox
River in Wisconsin, has a • SOCW/KOW ratio of approximately 5. This ratio indicates that the system
is close to steady state and that most or all of the disequilibrium is attributable to differences in
organic carbon in the water and sediments.
Guidelines for sampling and measurement of • socw are identical to those for sampling and
measurement of C™ under BAF method 1, as described in Sections 1.2 and 5.1, and Csoc under
BAF method 2, as described in Section 5.2. Because concentrations of bioaccumulative
chemicals in surficial sediments are relatively constant on an annual basis in most carbonaceous,
fine-sediment deposit!onal areas, determination of an appropriate average C™ in systems with
temporal fluctuations is the greatest challenge in measurement or estimation of • socw. On the
basis of monitoring reports and historical loading data, EPA expects that most persistent nonionic
organic chemicals will have • SOCW/KOW ratios in the range of 2-40. This expectation does not apply
when such chemicals have not been present in an ecosystem long enough to approach expected
steady-state concentrations in surficial sediments. In this case, • SOCw/Kow will be substantially
lower than 2, indicating low exposure potential through the benthic food web. Because the
national BAF methodology assumes that BAFs are determined for approximate long-term
average conditions, • SOCw/Kow values of less than 2 are unlikely to be relevant for persistent,
hydrophobic chemicals.
4.4 DERIVATION AND USE OF FOOD CHAIN MULTIPLIERS
FCMs are used in Procedure 1 (Figure 3-1) to estimate the dietary transfer of a chemical
up the food web for chemicals where metabolism is believed or assumed to be negligible. In
Procedure 1, FCMs are used with two of the four methods for deriving national BAFs. FCMs are
determined using a food web model and/or field data, and FCMs represent a measure of the
chemical's tendency to biomagnify in aquatic food webs. By definition, an FCM is:
cr,., _ Baseline BAF Baseline BAF
= Kl Baseline BCF (Equation 4-8)
This equation assumes that a BCF that is corrected for growth dilution, lipid normalized,
and corrected for bioavailability considerations—that is, a baseline BCF—is equal to Kow. The
scientific basis for this assumption is presented in Section 5.4.2. Because a baseline BCF is
determined by using a water-only exposure to the chemical, it represents a trophic level 1
exposure for the organisms. When organisms occupy higher trophic levels in food webs,
concentrations of certain chemicals in their tissues can exceed those that are due to water
exposure only because of dietary uptake of the chemical. The baseline BCF, when multiplied by
the FCM for the organism's trophic level, accounts for the influences of dietary uptake by the
4-24
-------
organism. Dietary uptake of the chemical generally becomes important when the chemical's
hydrophobicity exceeds a log Kowof 4 and the rate of chemical metabolism by the organism is
small. Thus, for nonionic organic chemicals, the third and fourth methods of Procedure 1 are
applicable only to chemicals with log Kows of 4 or greater.
4.4.1 Derivation of FCMs Using a Food Web Model
To derive FCMs using a food web model, a model and its input parameters must be
selected. The following subsections discuss how EPA selected a food web model for use in the
2000 Human Health Methodology. Also described are the parameters used with the model—the
food web structure, • socw (or, equivalently, C " and Csoc), and the chemical metabolism rate in the
food web. Because all food web models require the above input parameters, these input
parameters are not unique to the food web model selected by EPA.
Selection of a Food Web Model
For a food web model to provide useful predictions, it must have the following general
characteristics and qualities. First, the model must include all biotic components of the food web,
that is, plankton, benthic invertebrates, forage fish, and piscivorous fish. Second, the model must
account for chemical uptake and loss from both food and water for all organisms. Third, the
model must include chemical concentrations in sediment and the water column, because these
environmental compartments are the primary exposure media for benthic invertebrates and
phytoplankton, respectively, and these organisms reside at the base of the benthic and pelagic
food web. Fourth, because AWQCs for the protection of human health are designed for long-
term average conditions in ambient waters, steady-state solutions for predicting bioaccumulation
in the food chain model are preferred over time-variant dynamic solutions for the food chain
model (see Section 1.2). Other desirable qualities include (1) the model is easy to run by the
average user, (2) the model does not mix fate and transport models with the food chain model,
(3) the model code does not require substantial validation each time it is used, and (4) the model
is easy to parameterize.
Food chain models with the characteristics and desirable qualities summarized above
include the models of Gobas (1993) and Thomann et al. (1992). Other models are available, for
example, Levels I, II, and III fugacity models (Mackay, 1991); RIVER/FISH (Abbott et al., 1995);
AQUATOX (USEPA, 2000c,d); lannuzzi et al. (1996); Ecofate (Gobas et al., 1998); and
BASS/FGETS (Barber, 2000). The AQUATOX (USEPA, 2000c,d), lannuzzi et al. (1996), and
Ecofate (Gobas et al., 1998) models incorporate the submodels of Gobas (1993) and Thomann et
al. (1992) for modeling chemical uptake and loss. The RIVER/FISH (Abbott et al., 1995),
AQUATOX (USEPA, 2000c,d), Ecofate (Gobas et al., 1998), and BASS/FGETS (Barber, 2000)
models have extensive input data requirements and are principally designed for time and spatially
variant dynamic solutions for the food web. The time and spatially variant models include fate
and transport submodels along with the bioaccumulation submodel, and thus model predictions
include the uncertainties associated with both of the submodels. The fugacity models (Mackay,
1991) are designed for assessing the general behavior of chemical in model environments. On the
basis of the above characteristics and desirable qualities, EPA selected the models of Gobas
4-25
-------
(1993) and Thomann et al. (1992) for further consideration and evaluation for calculation of
FCMs. These two models are widely accepted in the scientific community and are being used in a
number of scientific and regulatory applications.
Burkhard (1998) performed a thorough evaluation of the Gobas (1993) and Thomann et
al. (1992) steady-state food web models for predicting chemical concentrations in aquatic food
webs. Burkhard (1998) assessed (1) the accuracy and precision of the models, (2) the sensitivity
of the predicted concentrations to changes in input parameters, and (3) the uncertainty associated
with the concentrations predicted by the models. These evaluations were performed with field
data from the Lake Ontario and its food web structure. A brief summary of this evaluation is
provided in this TSD. For further details, the reader can refer to Burkhard (1998).
Model Scope and Theoretical Basis. The Gobas and Thomann models are quite similar
in many ways. Both models include benthic and pelagic food web components, thereby
incorporating exposure of organisms to chemicals from both the sediments and the water
column. Both models contain rate equations for the estimation of steady-state conditions but also
treat some chemical distributions as equilibrium partitioning. Both models require specification of
the food web structure and the lipid contents and weights of the organisms. Both models also
incorporate the organic carbon contents of the sediment and the water column. However, the two
models also have distinct differences. The major difference pertains to the methods used to
predict chemical concentrations in benthic invertebrates and zooplankton. With the Gobas (1993)
model, concentrations are predicted by using equilibrium partitioning, whereas with the Thomann
(1992) model, concentrations are predicted by using uptake and loss rates based on respiration,
dietary consumption, and growth of the organism.
The Gobas and Thomann models do not include solubility limits or controls for the
concentration of the chemical in any compartment. Thus, for a given ratio of the chemical
concentration in sediment to that in the water column, the models will predict the same BAF and
BSAF regardless of the numerical values used for the chemical concentrations, provided the ratio
is maintained. If these models are used to predict chemical concentrations in aquatic organisms,
the actual chemical concentrations in the sediment and water column will be required.
Model Accuracy. Burkhard's (1998) evaluation using field data from Lake Ontario (Oliver
and Niimi, 1988) demonstrated that the Gobas and Thomann models have similar predictive
ability for all species for chemicals with log Kows ranging from 3 to 8 (Figure 4-6). The baseline
BAFs predicted with the Gobas model were in slightly better agreement with measured baseline
BAFs (using Lake Ontario field data) than those predicted with the Thomann model. For
chemicals with log Kows of 8 or greater, the models provided significantly different predictions.
For the Gobas model, average ratios of the predicted to the measured baseline BAFs were 1.6 for
sculpin, 1.0 for alewife, 1.4 for small smelt, 1.2 for large smelt, and 1.2 for piscivorous fish. For
the Thomann model, average ratios of the predicted to the measured baseline BAFs were 4.0 for
sculpin, 2.2 for alewife, 3.1 for small smelt, 3.0 for large smelt, and 2.5 for piscivorous fish. On
average, the Thomann model predicted slightly higher baseline BAFs than the Gobas model. For
piscivorous fish, the 10th and 90th percentile ratios (predicted/measured) were 0.4 and 5.6 for the
Thomann model and 0.3 and 2.1 for the Gobas model, respectively. Assuming a predicted
4-26
-------
concentration of 5 ppb in piscivorous fish, these ranges in baseline BAFs translate into
concentrations in fish of 2-28 ppb for the Thomann model and 1.5-10.5 ppb for the Gobas
models. These ranges are relatively narrow, varying by a factor of about 10.
Log K
Figure 4-6. Measured baseline BAFs for PCBs (•), chlorinated pesticides (•), and chlorinated benzenes, chlorinated
toluenes, and hexachlorobutadiene (•) from the data of Oliver and Niimi (1988) and BAF^s predicted using the
Gobas (- -) and Thomann (—) models plotted against Kow for all organisms (Burkhard 1998). For phytoplankton, the
predicted baseline BAFs were the same for both models and thus both lines coincide in the phytoplankton plot.
Model Sensitivity. A sensitivity analysis was performed to evaluate which input
parameters most affected the model. The sensitivity analysis of the input parameters used by the
Thomann and Gobas models revealed that • socw, Kow, lipid contents of the organisms, feeding
preferences of forage fish upon benthic invertebrates, and feeding preferences of the benthic
invertebrates (Thomann model only) were the most sensitive input parameters for the models.
Sensitivity analyses were performed using a variety of deviations to the input parameters
(i.e., ±10%, ±25%, ±50%, and ±75%). For both models, the magnitude of effect of a given change
in • socw, organism weight, and organism feeding preferences on the overall model outcome was
directly proportional to the change in each of the input parameters. Input parameters with
moderate nonproportionalities were lipid content (Thomann model) and temperature (Gobas
model). Input parameters with large nonproportionalities were lipid content (Gobas model) and
Kowfor both models. For both models, • socw, lipid content, and feeding preferences upon forage
fish showed little or no sensitivity for chemicals with log Kows of less than 4, a steep increase (or
decrease) in sensitivity for chemicals with log Kows between 4 and 6, and sensitivities of
approximately 1 (or -1) for chemicals with log Kows exceeding 6. A sensitivity of 1 means that a
4-27
-------
10% change in an input parameter results in a 10% change in the predicted baseline BAF. The
greater sensitivity of the models to the Kow input parameter is logical, because all submodels use
the hydrophobicity of the chemical to define rates of chemical uptake and loss. In general, a
+10% change in the Kow parameter results in a +10% to +20% change in the chemical
concentration in fish for log Kows up to • 7. The Thomann model was also extremely sensitive to
the feeding preference of the benthic invertebrates upon phytoplankton. For log Kows exceeding
4, the sensitivities became very large and approached values of-20 (in response to a +10%
change in the input parameter) for chemicals with log Kows of 6 or greater. The large sensitivity
for this input parameter occurs because the sediments are the predominant source of the chemical
to the benthic invertebrates but are only a minor part of the diet for the organism. Sensitivities for
other input parameters, such as organism weight, temperature (Gobas model), and other feeding
preferences, were relatively small.
Model Uncertainty. Uncertainty analyses performed with Monte Carlo simulations and
the Lake Ontario food web demonstrated that the input parameters Kow and • socw were the
dominant sources of uncertainties for the predicted baseline BAFs in piscivorous fish for both
models (Burkhard, 1998). These analyses were performed by using distributions and variances
for each input variable based on field data, and each simulation was performed with 100,000
iterations. To assess the importance of individual as well as groups of individual parameters,
simulations were performed by setting the variances for individual or groups of individual input
parameters to zero and comparing the ranges of the predicted baseline BAFs for the predictions
with the nonzeroed and zeroed variances.
For piscivorous fish, overall uncertainties in the predicted baseline BAFs ranged from a
factor of 3.3 to 5.5 in the Gobas model and from a factor of 3.3 to 8.7 in the Thomann model for
chemicals with log Kows of less than 7.6 (based on the ratio of the 10th to 90th percentile
predictions in the distribution of possible values). To provide a perspective of the differences in
predictions between the models and their uncertainties, one can assume that for piscivorous fish
the Gobas model predicts concentrations of 4 ppb (with 1.8 and 9.4 ppb for the 10th and 90th
percentile predictions) for a chemical with a log Kow of 5.0 and a concentration of 4 ppb (2.1 and
7.6 ppb) for a chemical with a log Kow of 6.6. The Thomann model would predict a concentration
of 2.6 ppb (1.2 and 4.6 ppb) for a chemical with a log Kow of 5.0 and a concentration of 16.1 ppb
(8.5 and 30.1 ppb) for a chemical with a log Kow of 6.6. In general, these differences are not large,
and from the perspective of quantifying these concentrations analytically, these differences are
almost indistinguishable.
Burkhard's (1998) evaluation of the Gobas and Thomann food web models reveals that
the models provide quite similar predictions for all organisms in the food web and that the
predictions are not significantly different for piscivorous fish. The comparison of predicted and
measured baseline BAFs based on field data from Lake Ontario suggests that the Gobas model
provides slightly more accurate predictions than the Thomann model. The sensitivities of the
input parameters are similar for both models, with the exception of benthic invertebrate feeding
preferences. The Thomann model was extremely sensitive to small changes in this input
parameter, and the Gobas model, because of its assumption of equilibrium partitioning for
benthic invertebrates, does not use this input parameter. Uncertainty analyses performed with
4-28
-------
both models indicate that Kow and • socw are the dominant sources of uncertainty in the predicted
baseline BAFs. These analyses suggested that the uncertainties associated with predictions by the
Gobas model are slightly smaller than those with the Thomann model.
Based on the evaluation of the Gobas and Thomann food web models described above,
EPA will use the Gobas model for calculating food chain multipliers (FCMs) due to the
considerations listed below:
1. The Gobas model includes both benthic and pelagic food webs, thereby
incorporating exposure of organisms to chemicals from both the sediments and
the water column.
2. The input data needed to run the model can be readily defined.
3. The baseline BAFs predicted using the model are in good agreement with field-
measured baseline BAFs for chemicals, even those with very high log Kows.
4. The Gobas model had smaller uncertainties associated with the baseline BAFs for
fish compared to the Thomann model.
5. The Gobas model is readily available via the Internet in a Windows-based format
at http://www.rem.sfu.ca/toxicology/models.htm.
6. The model predicts chemical concentrations in benthic organisms using
equilibrium partitioning theory, which is consistent with EPA's draft equilibrium
partitioning sediment benchmarks (ESBs) (USEPA, 2000b).
Because models are continually being refined, in the future EPA may consider the use of
other appropriately validated food web models for the derivation of FCMs. Any model
considered would need to have the characteristics and qualities outlined in Section 4.4. and would
have to be subjected to a validation process to address the issues of (1) accuracy and precision of
the model predictions, (2) input parameter sensitivities, and (3) uncertainties associated with the
model predictions.
Selection of the Sediment-Water Concentration Quotient (• som)
Calculations of FCMs with the Gobas food web model requires the ratio of the chemical
concentrations in the sediments (expressed on an organic carbon basis) to those in the water
column (expressed on a freely dissolved basis). Unfortunately, measured • socws are rather limited
in ecosystem type, chemical classes, and quality because of a number of factors. These include
(1) the difficulties in measuring the concentrations of hydrophobic organic chemicals in natural
waters because they occur at very low concentrations, that is, less than 1 ng/L; (2) the collection
of sediment and water samples that are not temporally and/or spatially connected; (3) collection
of bulk sediment samples rather than the uppermost 1 or 2 cm of the sediments; (4) the fact that
measurements of POC and DOC were not performed on the water samples analyzed for the
4-29
-------
hydrophobic organic chemicals; (5) the lack of determination of the sediment organic carbon
content; and (6) the fact that studies designed specifically for determining • socw are not usually
performed. In addition, combining sediment measurements from one study with water
measurements from another study can result in large biases in • socws due to differences in
analytical methodologies (e.g., different surrogates for recovery corrections, different standards).
Review of a number of different data sets, as described in Burkhard (1998), revealed three
data sets of suitable quality for which • socws could be determined. These data sets were from
Lake Ontario (Oliver and Niimi, 1988), Hudson River (USEPA, 1997; USEPA, 1998b), and Green
Bay in the Lake Michigan ecosystem (www.epa.gov/grtlakes/gbdata/). The Green Bay and
Hudson River data sets contained data for PCBs only, and the Lake Ontario data set contained
data for chlorinated pesticides, PCBs, and a few chlorinated benzenes, toluenes, and butadiene.
The data for the chlorinated benzenes, toluenes, and butadiene in the Lake Ontario data set were
not used in this analysis because these chemicals volatilize to the atmosphere relatively easily in
comparison with the higher molecular weight PCBs and chlorinated pesticides.
Figure 4-7 shows the • socws for selected PCB congeners in five different zones of Green
Bay. For the individual PCB congeners, the geometric mean regressions were performed on data
for the five different zones in the Green Bay system because both variables were measured with
error (Ricker, 1973). The slopes of the log • SOCw-log Kow regressions from the different zones were
not significantly different among the five zones (comparison of slope test, • = 5%). Therefore,
average • socws were determined for each PCB congener with data from all zones (Figure 4-8). The
geometric mean regression statistics are reported in Table 4-3 for each zone and for the average of
all zones. Examination of Figures 4-7 and 4-8 and Table 4-3 reveals that for PCBs, • socw is
strongly dependent on the Kow and slopes of slightly less than 1 were obtained. Examination of
* socwS for Lake Ontario and Hudson River reveals trends similar to those in Green Bay; a strong
dependence of • socw on Kow for the PCBs and chlorinated pesticides (Figures 4-8 and 4-9 and
Table 4-3) and slopes of 1 and slightly less than 1 were obtained.
Table 4-3. Geometric Mean Regression Equations (log • socw = A • log Kow+ B) for
Polychlorinated Biphenyls (PCBs) and Chlorinated Pesticides
Ecosystem
Green Bay (PCBs)
Zone 1
Zone 2a
Zone 3 a
Zone 3b
Zone 4
All zones, congener averages
Hudson River (PCBs)
RM189
RM194
Slope (±sd)
0.95 (±0.04)
0.92 (±0.09)
0.87 (±0.06)
0.83 (±0.06)"
0.86 (±0.08)
0.92 (±0.06)
0.87 (±0.08)
0.72 (±0.08)"
Intercept (±sd)
1.21 (±0.22)
1.13 (±0.61)
1.61 (±0.36)
1.88 (±0.36)
1.31 (±0.53)
1.20 (±0.38)
1.81 (±0.45)
3.16 (±0.42)
n
46
31
63
60
46
77
32
27
r
0.97
0.82
0.86
0.85
0.76
0.82
0.86
0.84
**y
0.17
0.34
0.37
0.33
0.46
0.43
0.13
0.16
Lake Ontario (PCBs and chlorinated
pesticides) 1.05 (±0.08) 0.83 (±0.49) 55 0.84 0.46
n = number of data points, r = correlation coefficient, sd = standard deviation, s^ = standard error of estimate," slope significantly
different from 1.0, • = 1%.
4-30
-------
O)
o
O)
o
O)
o
O)
o
O)
o
Figure 4-7. Sediment-water column concentration
quotient (• socw) for PCBs in five different
geographical zones in Green Bay, Lake Michigan.
The circled data points are the PCB congeners
numbers (logKow) 18 (5.24), 28 + 31 (5.67), 52
(5.84), 101 (6.38), 118 (6.74), 149 (6.67), 174
(7.11), and 180 (7.36). The geometric mean
regression and their 95% confidence limits are
plotted.
4-31
-------
O)
o
10
9
8
7
6
5
4
All Zones,
Congener Averages
Figure 4-8. Average sediment-water column
concentration quotients (• socw) for individual PCB
congeners across the five different geographical
zones in Green Bay, Lake Michigan. The circled
data points are the PCB congeners numbers (log
Kow) 18 (5.24), 28 + 31 (5.67), 52 (5.84), 101
(6.38), 118 (6.74), 149 (6.67), 174 (7.11), and
180 (7.36). The geometric mean regression and
their 95% confidence limits are plotted.
K
In the Green Bay ecosystem, chemical concentrations in both sediments and the water column
decrease with increasing zone number. Zone 1 is at the mouth of the Fox River, the source of
PCBs to the bay, and zone 4 connects the bay to Lake Michigan. Zone 1, the region of highest
chemical concentrations, has much less variability in the measured • socws and the largest slope
for the log • SOCw-log Kow relationship among all sampling zones in Green Bay. Comparison of the
variability existing in zones 1 through 4, as illustrated by the 95% confidence intervals in Figure
4-7, suggests that variability increases with increasing distance from the source of the PCBs
(Table 4-4), and this trend parallels the concentration gradient in Green Bay. The tightness,
consistency, and slope of the • SOCwS-Kow relationship observed in zone 1 data might be more
illustrative of the underlying • SOCwS-Kow relationship than those of the other zones because of
lower uncertainties associated with the analytical measurements.
Hudson River Mile 189
Hudson River Mile 194
Figure 4-9. Sediment-water column concentration
quotient (• socw) for PCBs at river miles 189 and
194. The circled data points are the PCB
congeners numbers (logKow) 18 (5.24), 28 + 31
(5.67), 52 (5.84), 101 (6.38), 118 (6.74), 149
(6.67), 174 (7.11), and 180 (7.36). The geometric
mean regression and their 95% confidence limits
are plotted.
4-32
-------
Table 4-4. Average • S0cw/Kow Ratios for Three Different Ecosystems
Ecosystem
Green Bay (PCBs)
Zone 1
Zone 2a
Zone 3 a
Zone 3b
Zone 4
All zones, congener averages
Hudson River (PCBs)
RM189
RM194
Average Ratio (±sd)
9. 15 (±4.97)
6.35 (±6.73)
10.3 (±13.3)
9.48 (±10.6)
4.49 (±6.68)
7.21 (±6.68)
14.3 (±8.98)
48.4 (±47.6)
5%
4.34
1.24
1.27
1.68
0.60
1.01
6.03
18.9
Percentile
10% 90%
5.55
1.37
1.88
2.00
0.75
1.76
7.36
22.6
13.8
13.1
21.7
20.1
6.95
13.3
23.4
69.5
95%
17.3
21.0
25.6
29.9
8.10
16.5
34.7
83.6
RM189
RM194
Lake Ontario (PCBs and chlorinated
pesticides)
Overall average • SOCW/K<,W=
14.3 (±8.98)
48.4 (±47.6)
23 .4 (±25.1)
23.3 (±18.0)
6.03
18.9
2.96
7.36
22.6
3.57
23.4
69.5
52.6
34.7
83.6
82.4
sd = standard deviation
From a theoretical standpoint, log • SOCw-log Kow relationships will have a slope of 1 if the
ecosystem is at equilibrium. In addition, EPA believes that ecosystems at steady state or with
conditions that approximate the longer term average conditions will also have slopes nearly equal
to 1. A number of factors could cause the slope to be less than 1; these include volatilization
losses (volatilization rates decrease with increasing molecular weight), sorption/desorption
hysteresis (desorption rates decrease with increasing molecular weight), inaccuracies in the
calculation of the concentration of chemical that is freely dissolved in the water column (the
denominator in the • socw term), and measurement error in determining the concentrations of
chemical in the sediments and/or water column. The log • SOCw-log Kow relationships for the
Hudson River, Lake Ontario, and Green Bay ecosystems have slopes that are 1 or slightly less
than 1 for PCBs and chlorinated pesticides (Table 4-3). The smallest slopes were observed with
the Hudson River ecosystem data. The Hudson River ecosystem is much more dynamic and
possibly further from steady-state conditions than are the Lake Ontario and Green Bay
ecosystems, because of changing flows overtime and recent changes in PCB loadings. Given the
similarity in slopes among all three ecosystems, the conditions in the Hudson River do not appear
to be greatly different from those in the other two ecosystems.
Given that the slopes for the log • SOCw-log Kow relationships in Green Bay, the Hudson
River, and Lake Ontario are close to 1, and the fact that ecosystems tend to move toward the
theoretical slope of 1 over time, EPA assumes a slope one for this relationship. This causes the
log • socw = A • log Kow + B relationship to become log • socw = log Kow + B. Rearrangement of this
equation gives B = log Kdoc - log Kow= log (Kdoc/Kow), and B can be found by averaging the
differences of the log Kdoc and log Kow for the individual chemicals or by averaging the logarithms
of the ratios of the Kdoc to the Kow for the individual chemicals. This averaging was performed for
the three ecosystems (Table 4-4), yielding average • SOCw/Kow ratios of 7.21 for Green Bay, 14.3
and 48.4 for Hudson River, and 23.4 for Lake Ontario. The large differences in average • SOCw/Kow
ratios between the two Hudson River sampling stations suggest distinctly different behaviors in
the two sampling stations, and, therefore, an overall ratio was not computed for the Hudson
4-33
-------
River. An average • SOCw/Kow ratio was computed for the three ecosystems by using the average
values for Green Bay and Lake Ontario with the two ratios from the Hudson River. An average
* Socw/KoWratio for the three ecosystems of 23.3 with a standard deviation of 18.0 was obtained
(Table 4-4).
The EPA believes that the differences in average • SOCW/KOW ratios among the three
ecosystems evaluated here illustrate the range of variability that occurs among ecosystems across
the nation. Because • socws are a function of both current and past chemical loadings to the
ecosystems, • SOCW/KOW ratios both larger and smaller than those observed exist in the nation. For
highly contaminated sites, for example, Superfund sites with large concentrations of chemicals in
the sediments, • SOCW/KOW ratios could become very large. For new chemicals that are just being
introduced or discharged into the environment, • SOCW/KOW ratios will be small because very little of
the chemicals is present in the sediment. Degradation processes such as hydrolysis, photolysis,
and metabolism can also strongly influence the • SOCW/KOW ratio, depending on where these
processes occur (i.e., the sediment and/or the water column).
Because the degradation rates for the PCBs and chlorinated pesticides in the environment
are extremely slow, the average • SOCW/KOW ratio of 23.3 for the three ecosystems is representative
of chemicals that are very slowly degraded (or have long half-lives in the environment).
Chemicals with higher degradation rates will, in all likelihood, have • SOCw/Kow ratios that are
different from those for the PCBs and chlorinated pesticides, and EPA believes that the
* socw/Kow ratios will be smaller for such chemicals, on average, than those for the PCBs and
chlorinated pesticides.
On the basis of the data and information presented above, EPA will use the average value
of 23 for the • SOCw/Kow ratio for deriving FCMs with the Gobas model.
Selection of a Food Web Structure
To determine FCMs with the Gobas model, a food web structure is needed. The
information necessary to construct a food web includes the diet of the individual organisms
composing the food web and their weights and lipid contents. The sensitivity analysis performed
with the Gobas model indicated that the model predictions were relatively insensitive to organism
weights (largest sensitivity, < 0.1) and feeding preferences of piscivorous fish (largest sensitivity,
< -0.3) for all Kows. The predictions were more sensitive for • socw, feeding preferences of forage
fish upon benthic invertebrates, and lipid contents for chemicals with higher log Kows (Burkhard,
1998). The more sensitive input parameters attain sensitivities of approximately 1 or-1 at log
Kows of 6 (• socw), 5 (feeding preferences of forage fish upon benthic invertebrates), and 7 (lipid
content). The most sensitive input parameter was the feeding preferences of forage fish, that is,
the percentage of zooplankton (pelagic component) and benthic invertebrates (benthic
component) in their diet. The benthic/pelagic composition of the food web is, EPA believes, the
most important characteristic for defining the structure of the food web for piscivorous fish
because transfer of chemicals from the sediment to piscivorous fish occurs almost exclusively via
their diet.
4-34
-------
The uncertainty analysis performed with the Gobas model revealed that the • socw and Kow
input parameters were the dominant sources of uncertainty associated with the model prediction.
The higher sensitivities associated with the • socw and feeding preferences of forage fish are related
to higher sources of uncertainty associated with the • socw input parameter. As concluded in the
previous section, due to past chemical loadings and ecosystem interactions, many sediments
across the United States are currently enriched relative to the water column
(i.e., • SOCw/Kow= 23). This thermodynamic difference results in substantially different
concentrations of the chemical in benthic organisms and pelagic organisms (zooplankton). In the
Gobas model, the difference in lipid-normalized chemical concentrations between benthic and
pelagic organisms is a factor of 23, precisely the disequilibrium between the concentrations of
chemical in sediment and the water column. Consequently, small changes in the benthic portion
of the diet of forage fish will result in very different amounts of the chemical in the diets of both
forage fish and their predators. This large concentration difference is responsible for the
overwhelming importance of the benthic/pelagic composition in defining the food web. However,
in ecosystems where the disequilibrium (• SOCw/Kow) is small (or approaches equilibrium
conditions), the differences in lipid-normalized chemical concentrations between the benthic and
pelagic organisms will be much smaller. At equilibrium conditions (• SOCw/Kow =1), the lipid-
normalized chemical concentrations in the benthic and pelagic organisms are equal and differ by
the ratio of the fraction lipid of the organisms (f.), on a wet-weight basis. Therefore, for
ecosystems at or near equilibrium conditions, the benthic/pelagic composition of the food web is
much less important, because there are small differences between the chemical concentrations in
the benthic and pelagic organisms.
Food webs differ widely in their benthic/pelagic compositions among ecosystems, among
individual species, and among different age classes of species within an ecosystem. Of all the
ecosystem types, the purely pelagic food webs might be the least common for piscivorous fish.
However, purely pelagic food webs have been found in remote Ontario lakes for lake trout
(Rasmussen et al., 1990) and in Adirondack lakes for brook trout and yellow perch (Havens,
1992). Purely benthic food webs are more common than purely pelagic food webs, but are still
rather limited in nature. Some examples of purely benthic food webs can be found in tidal and
estuarine ecosystems, such as the food webs for flounder in New Bedford harbor (Connolly,
1991) and striped bass in the tidal Passaic River (lannuzzi et al., 1996). Mixed food webs are
common in all ecosystems and, EPA believes, far outnumber the purely pelagic and benthic food
webs. There are numerous examples of mixed benthic/pelagic food webs, such as the food webs
for lake trout in the Great Lakes (Flint, 1986; Morrison et al., 1997), lobster in the New Bedford
harbor (Connolly, 1991), whitefish and rainbow trout in the Fraser River (Gobas et al., 1998),
white perch in the Chesapeake Bay (Baird and Ulanowicz, 1989), and perch, bass, and crappie in
Little Rock Lake (Martinez, 1991). Purely pelagic and/or benthic species can exist in ecosystems
containing species with a mixed benthic/pelagic food web, for example, flounder and lobster in
New Bedford harbor (Connolly, 1991).
Attributes of a common aquatic food web might include multiple trophic levels, the
presence of forage and piscivorous fishes, a mixed benthic-pelagic structure, and benthic
invertebrates as important components. From the perspective of the fish and shellfish consumed
on average by the U.S. population, EPA believes that the common food web just described
4-35
-------
provides a reasonable representation of the potential exposures of humans to various segments of
the food web, for example, benthic filter feeders/detritivores such as shrimp and clams and top
predators such as trout and salmon. For birds and wildlife, such a food web might not be
reasonable because their diets, in all likelihood, are different from those of the U.S. human
population.
In selecting a food web structure for determining FCMs used in the third method of
Procedure 1 of the BAF methodology, EPA considered a number of approaches for deriving or
selecting a food web. These approaches included (1) developing a hypothetical food web
structure consistent with the desirable characteristics described above; (2) developing food web
structures for different ecosystem types and then averaging data to derive the typical food web
for the nation; (3) using fish consumption survey data, developing food web structures for
different species consumed by the U.S. population, and then averaging to derive the typical food
web for the nation; and (4) simply selecting an existing food web with the desirable
characteristics described above. In selecting an average or typical food web structure for the
nation, all of these approaches are somewhat problematic because of the large differences in food
webs across the country. For this reason, EPA strongly encourages States and Tribes to make
site-specific modifications to EPA's national BAFs (USEPA, 2000a).
The use of a purely benthic food web structure for the national BAF methodology with
the • SOCw/Kow value of 23 will result in the largest FCMs for fish. In contrast, the use of a purely
pelagic food web structure for the national BAF methodology with the • SOCW/KOW value of 23 will
result in the smallest FCMs for fish when the Gobas model is used. Because the goal of EPA's
national BAF methodology is to represent the long-term, average bioaccumulation potential of
pollutants in aquatic organisms that are commonly consumed by humans throughout the United
States, neither the purely benthic nor the purely pelagic food web structure represents average
conditions. Rather, the purely benthic and the purely pelagic food web structures are the
extremes in food web structure resulting in the largest and smallest FCMs, respectively.
On the basis of the above information and discussion, EPA will use the mixed food web
structure from the Lake Ontario ecosystem as the representative food web for determining FCMs
for the national methodology (Table 4-5) (Flint, 1986; Gobas, 1993). This selection is based on
the following considerations:
1. The Lake Ontario food web possesses the characteristics of the average or typical
food web described above, that is, four trophic levels and a benthic/pelagic
composition ration of 55:45 for the piscivorous fish (Table 4-5).
2. The Lake Ontario food web structure is not overly complex but does include
multiple forage fish with differing diets that are consumed by piscivorous fish (Table
4-5).
3. Comparisons of measured baseline BAFs and baseline BAFs predicted by using
FCMs based on the Lake Ontario data demonstrated good agreement for other
ecosystems, such as Green Bay, Hudson River, and Bayou d'Inde.
4-36
-------
4. None of the other approaches considered by EPA for deriving an average or typical
food web for the nation would have substantially lower uncertainties than those
associated with using the Lake Ontario food web.
5. A detailed investigation of the sensitivities and uncertainties for this specific food
web structure with the Gobas model can be performed (Burkhard, 1998), whereas
use of the other possible approaches described above for selecting food web
structure is not amenable to such analysis.
6. This selected food web does not represent either extreme in benthic/pelagic
composition and thus is consistent with EPA's goal for the national methodology of
representing the long-term average bioaccumulation potential of pollutants in
aquatic food webs.
Table 4-5. Food Web Structure for National BAF Methodology (Flint, 1986; Gobas, 1993)
Species Trophic Lipid Weight Diet
Trophic Lipid Weight
Level Content
Phytoplankton
Zooplankton (mysids [Mysis relicta])
Benthic Invertebrates (Diporeia)
Sculpin (Coitus cognatus)
Alewife (Alosa pseudoharengus)
Smelt (Osmerus mordax)
1
2
2
3
3
3-4
0.5%
5.0%
3.0%
8.0%
7.0%
4.0%
100 mg
12 mg
5.4 g
32 g
16 g
1 8% zooplankton, 82% Diporeia
60% zooplankton, 40% Diporeia
54% zooplankton, 21% Diporeia, 25%
Salmonids (Salvelinus namaycush,
Oncorhynchus mykiss, Oncorhynchus
velinus namaycush
11%
sculpin
2,410 g 10% sculpin, 50% alewife, 40% smelt
Calculation of Food Chain Multipliers
One additional input parameter is necessary before FCMs can be determined with the
Gobas food web model. This parameter, the rate of metabolism in forage and piscivorous fish, is
difficult to define because of the general lack of data on metabolism rate constants for individual
compounds. Procedure 1 of EPA's BAF methodology (see Section 3.1, Figure 3-1) assumes that
the rates of metabolism for the chemicals of interest are low. Consequently, EPA assumes no
metabolism; that is, metabolism rates are set equal to zero in the model when FCMs are
calculated for methods 3 and 4 in Procedure 1.
Inputs to the Gobas model (MS-DOS version) include concentrations of chemicals in the
sediment (expressed on a wet-weight basis) and in the water column (expressed on a total basis).
Because the Gobas model does not have solubility limits or controls for the concentration of
chemical in any compartment (i.e., sediment, water, and biota), the chemical concentration in the
water used with the model is arbitrary for determining the BAFs. In other words, the BAF
obtained by using a concentration of chemical of 1 ng/L will be equal to that obtained using a
4-37
-------
concentration of chemical of 150 |ig/L for a specified Kow. Thus, in deriving the FCMs, 1 ng/L
(concentration of chemical freely dissolved in the water column, C™) is used and the
corresponding chemical concentration in the sediment is calculated by using • SOCW/KOW = 23
relationship, or Cs = 23 • Kow • (1 ng/L) • foc.
In applying the Gobas model, EPA does not use Gobas's method of accounting for
bioavailability. Gobas's method for determining the freely dissolved (bioavailable) concentration
of the chemical in water makes no distinction between POC and DOC phases but rather treats
these two phases as one. In Section 4.2 of this document, the procedure used in the 2000 Human
Health Methodology for determining the concentration of chemical that is freely dissolved in the
ambient water, C™, is presented. To avoid using Gobas's method of accounting for
bioavailability, EPA set the concentration of the DOC in the model to an extremely small
number, 1.0 x 10'30 kilograms per liter. The Gobas model takes the total concentration of the
chemical in the water that is input to the model and, before doing any predictions, performs a
bioavailability correction by calculating the C™. The C™ is then used in all subsequent
calculations by the model. By setting the concentration of the DOC to 1.0 x 10"30 kilograms per
liter, the total concentration of the chemical put into the model becomes essentially equal to the
C™, because the bioavailability correction with the method of Gobas is extremely small.
For each value of Kow input to the Gobas model, predicted baseline BAFs are reported by
the model for each organism in the food web. FCMs are calculated from the predicted BAFs
using the following equation:
„„., _ Baseline BAF
K (Equation 4-9)
where:
Baseline BAF = BAF that is based on the concentration of chemical freely dissolved
in water (C™) and the concentration of chemical in the lipid fraction
of tissue
Kow = w-octanol-water partition coefficient
Using Equation 4-10, FCMs were calculated for each organism in the Lake Ontario food
web with the reported BAFs (Oliver and Niimi, 1988). Table 4-6 lists the FCMs for trophic level 2
(zooplankton), trophic level 3 (forage fish), and trophic level 4 (piscivorous fish). The FCMs for
the forage fish, trophic level 3, were determined by taking the geometric mean of the FCMs for
sculpin and alewife. The FCMs for the smelt were not used in determining the mean FCMs for
the forage fish because the diet of this organism includes small sculpin. This diet causes smelt to
be at a trophic level slightly higher than 3 but less than trophic level 4. In contrast, the diets of the
sculpin and alewife were solely trophic level 2 organisms (i.e., zooplankton andDiporeia sp.)
FCMs were determined with the Gobas model, the food web structure in Table 4-5,
* socw/Kow= 23, and the environmental parameters and conditions listed in Table 4-7. The resulting
FCMs, used for the national BAF methodology, are shown in Table 4-6.
4-38
-------
Table 4-6. Food-Chain Multipliers for Trophic Levels 2, 3, and 4 (Mixed Pelagic and
Benthic Food Web Structure and • SOcw/KoW= 23)
Log
KOW
4.0
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
5.0
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
6.0
6.1
6.2
6.3
6.4
6.5
Trophic Level
2
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
Trophic Level
3a
1.23
1.29
1.36
1.45
1.56
1.70
1.87
2.08
2.33
2.64
3.00
3.43
3.93
4.50
5.14
5.85
6.60
7.40
8.21
9.01
9.79
10.5
11.2
11.7
12.2
12.6
Trophic Level
4
1.07
1.09
1.13
1.17
1.23
1.32
1.44
1.60
1.82
2.12
2.51
3.02
3.68
4.49
5.48
6.65
8.01
9.54
11.2
13.0
14.9
16.7
18.5
20.1
21.6
22.8
Log
KOW
6.6
6.7
6.8
6.9
7.0
7.1
7.2
7.3
7.4
7.5
7.6
7.7
7.8
7.9
8.0
8.1
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
9.0
Trophic Level
2
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
Trophic Level
3a
12.9
13.2
13.3
13.3
13.2
13.1
12.8
12.5
12.0
11.5
10.8
10.1
9.31
8.46
7.60
6.73
5.88
5.07
4.33
3.65
3.05
2.52
2.08
1.70
1.38
Trophic Level
4
23.8
24.4
24.7
24.7
24.3
23.6
22.5
21.2
19.5
17.6
15.5
13.3
11.2
9.11
7.23
5.58
4.19
3.07
2.20
1.54
1.06
0.721
0.483
0.320
0.210
1 The FCMs for trophic level 3 are the geometric mean of the FCMs for sculpin and alewife.
Table 4-7. Environmental Parameters and Conditions Used for Determining FCMs for the
National BAF Methodology
Mean water temperature: 8°C
Organic carbon content of the sediment: 2.7%
Dissolved organic carbon content of the water column: 1 .OE-30 mg/L
Density of lipids: 0.9 kg/L
Density of organic carbon: 0.9 kg/L
Metabolic transformation rate constants (all organisms): 0.0 d"1
socw ^"TJW
4-39
-------
4.4.2 Derivation of Food Chain Multipliers Using Field Data
In addition to model-derived estimates of FCMs, field data can also be used to derive
FCMs for nonionic organic chemicals. Compared with the model-based FCMs described
previously, field-derived FCMs account for any metabolism of the pollutant of concern by the
aquatic organisms used to calculate the FCM.
Field-derived FCMs should be calculated with lipid-normalized concentrations of the
nonionic organic chemical in appropriate predator and prey species, using the following
equations:
= BMF^ (Equation 4- 10)
FCM ^3 = (BMF^) • (BMF ^ (Equation 4-11)
FCM ^4 = (BMF TU) • (BMF ^3) • (BMF ^2) (Equation 4-12)
where:
FCM = food chain multiplier for designated trophic level (TL2, TL3, or TL4).
The basic difference between FCMs and BMFs is that FCMs relate back to trophic level 1
(or trophic level 2, as assumed by the Gobas model [1993]), whereas BMFs always relate back to
the next lowest trophic level. For nonionic organic chemicals, BMFs can be calculated from lipid-
normalized concentrations of chemical in tissues of biota at a site according to the following
equations:
BMFTL2 = (C.jTL2)/(CvTL1) (Equation 4-13)
BMF^ = (c'.tTL3)/(c'.!TL2) (Equation 4-14)
BMF™ = (C/TL4)/(C.,TL3) (Equation 4-1 5)
where:
C. = lipid-normalized concentration of chemical in tissue or whole organism
at a specified trophic level (TL2, TL3, or TL4).
In addition to the acceptability guidelines pertaining to field-measured BAFs, the
following procedural and quality assurance guidelines apply to field-measured FCMs.
1 . Information should be available to identify the appropriate trophic levels for the
aquatic organisms and appropriate predator-prey relationships for the site from
which FCMs are being determined. Information about trophic status is most
accurate when obtained from the site(s) of interest, because predator-prey
relationships for some species can vary widely over space and time. When a
predator species consumes multiple prey species at a particular trophic level,
chemical concentrations in prey species should be appropriately weighted (if the
data are available) when used to calculate field-based FCMs. General information
4-40
-------
on determining trophic levels of aquatic organisms can be found in USEPA
(2000 e-g).
2. The aquatic organisms sampled from each trophic level should reflect the most
important exposure pathways leading to human exposure via consumption of
aquatic organisms. For higher trophic levels (e.g., 3 and 4), aquatic species used to
calculate FCMs should be those that are commonly consumed by humans. The
species sampled should also reflect size and age ranges that are typical of human
consumption patterns.
3. The study from which the FCMs are derived should contain enough supporting
information to determine that tissue samples were collected and analyzed
according to appropriate, sensitive, accurate, and precise methods.
4. The percent of tissue that is lipid should be either measured or reliably estimated
for the tissue(s) used to determine the FCM.
5. The chemical concentrations in the tissues/organisms used to calculate FCMs
should reflect long-term average exposures of the target species to the chemical of
interest; longer averaging periods are generally necessary for chemicals with
greater hydrophobicity.
4.4.3 Food Chain Multiplier Uncertainties
Uncertainties associated with predictions from the Gobas model have been assessed by
Burkhard (1998) as described in Section 4.4.1. Monte Carlo analyses were conducted by varying
all input parameters except Kow and • socw. The results of these analyses are shown in Figure 4-10.
Because the FCMs were calculated from baseline BAFs predicted with the Gobas model, the
uncertainties shown in Figure 4-10 are directly applicable to the FCMs. These results suggest that
the FCMs have fairly low uncertainties, because, in their calculation, • socw was fixed and the Kows
were assumed to have no error. For example, for a log Kow of 6.5, the ratio from the Monte Carlo
analysis was 1.74 for the 90th to 10th percentile predicted baseline BAFs (Figure 4-10). For trophic
level 4 fish, the FCM is 22.8 and the 10th and 90th percentile FCMs would be 17.3 and 30.1,
respectively.
Application of the FCMs, calculated with the assumed food web (Table 4-5) and
disequilibrium (• SOCW/KOW) of 23, to ecosystems and/or organisms with vastly different food webs
and/or disequilibriums can cause substantial biases in the baseline BAFs predicted for use in
methods 3 and 4 of Procedure 1. Although the degree and magnitude of error will vary among
sites, some general statements can be made about the direction and relative uncertainty
associated with these biases. Food webs that are more pelagic-based will tend to have smaller
FCMs, whereas food webs that are more benthic-based will tend to have larger FCMs. In Figure
4-11, FCMs for purely pelagic and purely benthic food webs, created by modifying the assumed
Lake Ontario food web and rerunning the Gobas model, are shown along with the FCMs from
4-41
-------
4-11, FCMs for purely pelagic and purely benthic food webs, created by modifying the assumed
Lake Ontario food web and rerunning the Gobas model, are shown along with the FCMs from
Table 4-5. These two modified food webs represent the extremes in benthic/pelagic composition.
Decreasing the disequilibrium (• SOCW/KOW) will cause the FCMs to become smaller, whereas
increasing the disequilibrium (• SOCW/KOW) will cause the FCMs to become larger (Figure 4-12). The
FCMs for the 2000 Human Health Methodology were derived assuming no metabolism of the
chemical in the food web. If metabolism does exist within the food web, the FCMs will be
smaller than those calculated without metabolism.
FCMs derived from field measurements (see Section 4.4.2) do not have the above biases
because the measurements incorporate the conditions existing at the field site where the
measurements were performed. This includes the existing disequilibrium, chemical metabolism,
and influences due to the structure of the food web (i.e., predator-prey relationships and benthic-
pelagic components).
4-42
-------
Figure 4-10. The ratio of the 90th to 10th
percentile baseline BAF predictions for
piscivorous fish from 100,000 Monte Carlo
simulations using the Gobas model as a
function of n-octanol-water partition
coefficient (Kow) (Burkhard 1998). The ratio
when all parameters except Kow ()))) and
except • socw and Kow ()••)) are varied.
Log KO
100
01
10
ro
.n
O
"5 0.1
0.01
Figure 4-11. FCMs for purely pelagic (••••)
and purely benthic (• • •) food webs derived by
modifying the Lake Ontario food web ()))).
5 6
Log KOW
100
in
-------
5. CALCULATING BASELINE BAFs FOR NONIONIC ORGANIC CHEMICALS USING
THE FOUR METHODS
This section presents each of the four BAF methods as they are applied to nonionic
organic chemicals under Procedure 1. Application of the four BAF methods under Procedure 1 is
generally more complex than for Procedures 2-4, and thus, more detailed discussions are
warranted on how to appropriately apply them. Nonetheless, the same general data quality
considerations, assumptions, strengths, and limitations that apply to the BAF methods under
Procedure 1 are generally relevant to Procedures 2-4, even though each method is not applied in
the same manner (or may not be used at all) under these other three procedures. The equations
for each BAF method under Procedure 1 are shown in this section, and the ability of each
method to predict BAFs is discussed, as are assumptions and limitations inherent to each
method.
5.1 METHOD 1: DERIVING BASELINE BAFs FROM TOTAL BAFs (BAFjs)
As has been noted, BAFs derived from data from samples collected in the field EAF{ are
the first preference in EPA's BAF hierarchy for deriving individual baseline BAFs. In Section 2,
the term "total BAF," denoted EAF{, was introduced to refer to "field-measured" BAFs. The
BAFj is defined as:
t _ Ct
BAFT - — (Equation 5-1)
where:
C, = total concentration of the chemical in tissue
Cw = total concentration of chemical in water
The BAFj shown in Equation 5-1 is calculated on the basis of the total concentration of chemical
in the appropriate wet tissue of the aquatic organism sampled and the total concentration of the
chemical in the ambient water at the sampling site.
A baseline BAF is calculated from a BAFj as shown in Equation 5-2 by using information
on the lipid fraction (f.) of the tissue of concern for the study organism and the fraction of the
total chemical that is freely dissolved in the study water (ffd) Appendix A provides more detailed
information on derivation of the baseline BAF equation.
Baseline BAF =
BAF^
- 1
(Equation 5-2)
5-1
-------
where:
BAF = Total BAF (BAF^ = C,/CW)
ffd = fraction of the total concentration of chemical in water that is freely dissolved
f. = fraction of tissue that is lipid
In calculating BAFs using method 1, EPA will use appropriate EAF{ data obtained from
the open literature (e.g., peer-reviewed journals, government reports, professional society
proceedings) when sufficient information is provided to indicate the quality and usability of data.
In general, the bioaccumulation data used should make it possible to calculate reliable BAF^s and
to make some assessment of the overall uncertainty in the EAF{ value.
5.1.1 Sampling and Data Quality Considerations
The data used to calculate a BAFx should be thoroughly reviewed to assess the quality of
the data and the overall uncertainty in the BAF value. The following general criteria apply in
determining the acceptability of BAF^s. Because no guidance can address all of the variation in
experimental designs and data found in the literature, best professional judgment will be
necessary to supplement these data quality guidelines in selecting the best available information
and using it appropriately.
1. Aquatic organisms used to calculate a field-measured EAF{ should generally be
representative of those aquatic organisms commonly consumed by the general
population in the United States. An aquatic organism that is not commonly
consumed by the general U.S. population can be used to calculate a field-
measured BAFx provided that the organism is considered to be a reasonable
surrogate for a commonly consumed organism. Information on the ecology,
physiology, and biology of the organism should be reviewed when assessing
whether an organism is a reasonable surrogate for a commonly consumed
organism.
2. The trophic level of the study organism should be determined by taking into
account its life stage, diet, and the food web structure at the study location.
Information from the study site (or similar sites) is preferred when evaluating
trophic status of an organism. If such information is lacking, general information
for assessing trophic status of aquatic organisms can be found in USEPA
(2000e-g).
3. In some cases, assessments of size, age, and reproductive status of the organisms
might be useful in assigning appropriate trophic levels for the study organisms.
Additionally, accumulation of chemical can vary as a result of other factors such
as different growth rates and pre-spawning versus post-spawning organisms.
Thus, the above ancillary information might be useful in deciding whether the
study organisms are appropriate representatives for field sites.
5-2
-------
4. The percent lipid of the tissue used to determine the EAF{ needs to be known,
either as measured in the field study or reliably estimated. This parameter is
necessary to permit lipid normalization of the concentration of chemical in tissue
when deriving baseline BAFs.
5. The study from which the EAF{ is derived should contain sufficient supporting
information from which to confirm that tissue and water samples were collected
and analyzed according to appropriate, sensitive, accurate, and precise analytical
methods.
6. The site of the field study should not be so unusual that the EAF{ cannot be
reasonably extrapolated to other locations where the national BAF and resulting
AWQC will apply.
7. The water concentration(s) used to derive the EAF{ should reflect the average
exposure experienced by the study organism(s). The extent of spatial and
temporal averaging that is necessary for the water samples is a function of the
variability in chemical concentration in the ecosystem. In general, greater temporal
and spatial averaging of chemical concentrations in water will be necessary with
increasing Kow. More water samples over time and space (i.e., more averaging) will
be necessary for chemicals with higher Kows and higher variability in chemical
concentrations than in ecosystems with lower variabilities in chemical
concentrations. For chemicals with higher Kows, BAF^s determined with
composite water samples over time will generally be more accurate than those
measured by individual "grab samples." For chemicals with lower Kows, BAFj s
determined with composite water samples over time will, in general, be more
accurate than those measured with individual "grab samples."
8. The home range of the organisms that are collected for determining BAFx s should
be determined or assessed such that the appropriateness of the spatial sampling
design can be evaluated within the context of the organism's mobility. For more
mobile organisms, greater spatial averaging will generally be necessary.
9. The concentrations of POC and DOC in the study water should be measured or
reliably estimated so that baseline BAFs can be derived.
10. The field study should not be conducted in an ecosystem that has recently
experienced a major change or disruption in chemical loadings or flows (for
example, a 100-year flood or the removal of a major chemical source) because it
takes time for the ecosystem to return to long-term average or steady-state
conditions. The response times depend on a number of factors, including the
nature of the disruption; the chemical's loading to and from the ecosystem; the
hydrodynamics and solids transport of the ecosystem; fish-specific parameters,
such as growth rates, chemical uptake, and depuration rates; and the Kow of the
chemical. For chemicals with higher Kows, response times might range from
5-3
-------
months to years, whereas for those with lower Kows, the response times would be
shorter, possibly less than a year.
The EPA is presently developing guidance for designing and conducting field studies for
determining field-measured BAF^s and for determining minimum data quality and quantity
requirements. This guidance will provide detailed information on how to design field sampling
studies that will yield BAF^s that are representative of the long-term average conditions in an
ecosystem and have low bias and good accuracy.
5.1.2 Assumptions and Limitations
Several assumptions and limitations are inherent in the use of BAF^s for deriving national
BAFs for nonionic organic chemicals. First, it is assumed that properly derived BAF^s can
provide a reasonable estimate of the bioaccumulation that would occur under the long-term
conditions that exist in the ecosystem. This assumption is important because human health
AWQCs are generally intended to protect humans from long-term (chronic) exposure to
chemical concentrations in water and fish. To address this assumption, concentrations of
chemicalin water and tissue must be averaged over appropriate temporal and spacial scales so
that a steady-state or long-term BAF can be reasonably approximated. Complications can arise in
situations where variability in chemical concentrations in water is high relative to concentrations
of chemical in tissue (as is usually the case with highly hydrophobic chemicals), when rapid
changes occur in chemical loadings to the ecosystem, and when organisms move between areas
in which they experience greatly differing chemical exposures. As discussed in Section 5.1.1,
achieving the most appropriate temporal and spatial averaging for determining BAF^s can be
specific to the chemical, species, and study site. In this regard, adherence to the aforementioned
sampling and data quality guidelines with respect to temporal and spatial averaging is the best
way to ensure that B AF^s reflect the bioaccumulation that would be expected at or near steady
state.
The second major assumption associated with the use of BAF^s for nonionic chemicals is
that by adjusting the EAF{ for the organism's lipid content (f.) and the chemical concentration
that is freely dissolved (ffd), it is possible to make reasonable predictions of bioaccumulation
across different species (within a trophic level) and sites. In reality, other factors influence
bioaccumulation. These factors include differences in chemical loadings histories (i.e., sediment-
water disequilibrium); food web structure; organism health and physiology; water quality factors
such as temperature; and food quality—all of which may vary across ecosystems.
Burkhard et al. (2003a) have evaluated the effectiveness of adjusting BAF^s by lipid
content and freely dissolved chemical concentration for increasing the reliability of extrapolating
BAFs across ecosystems and species. The results of these comparisons, which are discussed
further in Section 5.1.3, suggest that adjusting by f. and ffd reduces much of the variability in
BAFxS. Furthermore, this analysis suggests that BAFs can be extrapolated among species within
a trophic level and across ecosystems with reasonable accuracy. Nevertheless, some variation in
occurs from "other factors," such as those mentioned above.
5-4
-------
A third assumption involved in the use of BAF^s for deriving national BAFs is that, within
reasonable limits, BAF^s are independent of exposure concentration (i.e., BAF^s do not vary as a
function of exposure concentrations). This assumption is made when applying BAF^s derived
from one set of chemical exposure concentrations to another set of concentrations (e.g., from
higher to lower chemical concentrations or vice versa). For nonionic chemicals, this assumption
is consistent with the mechanism of chemical uptake (i.e., passive diffusion across cell
membranes) and is widely supported by reports in the literature. However, it is theoretically
possible for BAFxS to become dependent on exposure concentrations if these concentrations are
so high that they affect an organism's health and, subsequently, its rate of chemical uptake,
elimination, or metabolism. Although this is an issue typically associated with BCF studies in
which exposure concentration is controlled by the investigator, this issue will be considered
during data review in order to avoid using BAF^s from organisms that show overt signs of
toxicity. Although deviation from the concentration-independency assumption may be possible
under some circumstances, EPA is not aware of data that demonstrate the extent to which this
assumption might be violated under environmentally realistic exposure conditions. Furthermore,
it would probably be difficult to measure the extent of deviation from the concentration-
independency assumption, given the presence of other contributors, particularly differences in
bioaccumulation over space and time, to the overall variability on BAF^s.
Currently, the greatest limitation in using BAF^s to derive national BAFs is the paucity of
high-quality field data. The primary deficiencies that limit the use of available BAFj data include
the lack of proper spatial and temporal averaging, insufficient ancillary data (e.g., DOC, POC,
lipid content of organisms), and the lack of samples co-located in space and time. These
deficiencies often reflect limitations on available resources, but they also reflect study designs
that are inconsistent with the goals of a BAF study. These data gaps are expected to be filled as
additional field-measured data are generated to meet demands for site-specific BAFs and as
future guidance is developed for properly designing field BAF studies.
5.1.3 Validation of Method 1
As has been mentioned, use of baseline BAFs for nonionic organic chemicals allows
BAFxS to be extrapolated among species and across locations and improves the accuracy of this
extrapolation. To validate this approach, Burkhard et al. (2003a) EPA made two different
evaluations. In the first evaluation, BAF^s for PCBs and several aquatic species from Green Bay,
Lake Michigan, were compared among different geographical areas of the Bay (zones) and across
the entire Bay. In the second evaluation, BAF^s and baseline BAFs for six PCB congeners were
compared among species and across ecosystems.
In the Green Bay evaluation, BAF^s and baseline BAFs for the PCB congeners 18, 52,
149, and 180 in adult alewife, age 4 walleye, and age 10 carp were compared. These species were
selected because they were those species most frequently sampled across the different zones. The
PCB congeners used in the evaluation are major components of the PCB mixture present in
Green Bay, and uncertainties associated with their measurement are low. In addition, these
congeners had hydrophobicities that spanned a wide range: log Kows are 5.24 for PCB 18, 5.84 for
PCB 52, 6.67 for PCB 149 and 7.36 for PCB 180. BAF^s and baseline BAFs were calculated for
5-5
-------
six different zones. Comparing across the zones, the baseline BAFs varied less than the BAF^s;
that is, the baseline BAFs were more constant across zones than were the BAF^s (Figure 5-1;
Burkhard et al., 2003a). The BAF^s increased from zone 1 to 4 (Figure 5-1; Burkhard et al.,
2003a), and the differences are more pronounced for the more hydrophobic PCBs 149 and 180,
which is consistent with equilibrium partitioning theory. The observed trend of increasing BAF^s
across zones is due to increasing bioavailability of dissolved PCBs, which is caused by decreasing
POC and DOC across zones. This trend appeared to disappear with the adjustment of the BAF^s
to baseline BAFs (Figure 5-1; Burkhard et al., 2003a) because, as presented in
Section 4, the baseline BAF adjustment of chemical concentration to that which is freely
dissolved accounts for differences in POC and DOC.
8-
7-
6-
5-
2a
2b
3a
3b
Zone
Figure 5-1. BAF^s (•) and Baseline BAFs (•) for PCB congener 149
(2,2%3,4%5%6-hexachlorobiphenyl) (±1 sd) for adult alewife for different Green Bay zones.
To further evaluate the relative variances associated with BAF^s and baseline BAFs, bay-
wide BAFs were compared. Bay-wide BAF^s and baseline BAFs were calculated using a sample-
size weighted average of the BAFs for each of the geographical zones. The variances of the bay-
wide BAFxS and baseline BAFs were calculated as described in detail in Burkhard et al., 2003a.
The results of these calculations are summarized, by species, using the ratio of 90th to 10th and 95th
to 5th percentile exceedance limits in Table 5-1 and Burkhard et al., 2003a. Overall, the baseline
BAFs had smaller ratios than the BAF^s and the adjustment/conversion of BAF^s to baseline
BAFs resulted in an approximately twofold decrease in variability (Burkhard et al., 2003a).
5-6
-------
Table 5-1. BAF^s and Baseline BAFs Exceedance Limit Ratios for Green Bay (All Zones
Combined)
^,,T. 90th to 10th Percentile Exceedance Limit Ratio
Congener
Adult alewife
18
52
149
180
Age 4 walleye
18
52
149
180
Age 10 carp
18
52
149
180
BAF'Ts
4.98
5.48
3.33
4.08
3.57
4.04
3.11
3.96
4.87
6.75
5.96
7.09
Baseline BAFs
3.11
2.85
1.88
2.20
3.50
2.74
2.12
2.12
4.23
3.49
1.87
2.17
95th to 5th Percentile
BAF'Ts
7.86
8.90
4.70
6.10
5.14
6.01
4.30
5.87
7.65
11.6
9.91
12.4
Exceedance Limit Ratio
Baseline BAFs
4.30
3.84
2.26
2.76
5.00
3.65
2.62
2.63
6.39
4.99
2.24
2.71
To assess across-ecosystem variability, Burkhard et al. (2000a) assembled BAF^s and
baseline BAFs for six PCB congeners—PCBs 22, 52, 85, 118, 146, and 149—from the Green
Bay, Lake Ontario, and Hudson River ecosystems for 13 fish species (Figure 5-2). When
possible, age class BAF^s were assembled, and trophic levels for the different species were
assigned with nominal/rounded trophic levels. These assignments caused species with slightly
lower trophic level positions (e.g., adult gizzard shad, with an average trophic level of 2.5) to be
lumped with species with slightly higher trophic levels (e.g., adult alewife, with an average trophic
level of 3.5) within the nominal trophic levels shown in Figure 5-2. As shown in Figure 5-2, the
baseline BAFs had substantially lower variability than the BAF^s for trophic level 3 and 4 fish.
The coefficients of variation (in arithmetic space) for the trophic level 3 baseline BAFs were 85%
for PCB 22, 73% for PCB 52, 70% for PCB 85, 61% for PCB 118, 92% for PCB 146, and 59% for
PCB 149. For the BAF^s, these values were 116% for PCB 22, 97% for PCB 52, 104% for PCB
85, 104% for PCB 118, 615% for PCB 146, and 68% for PCB 149 (Burkhard et al., 2003a). Similar
differences in the coefficients of variation were found for trophic level 4 fish (Burkhard et al.,
2003a). On average, the 75th/25th and 90th/10th percentile ranges in baseline BAFs were • 2x and
• 5x smaller than the ranges for BAF^s. These results suggest that the corrections for tissue or
organism lipid content (f.) and the fraction of chemical concentrations that is freely dissolved in
water (ffd) reduce variability when BAFs are extrapolated among species of similar trophic levels
and across ecosystems. The variability (that is, the remaining spread or range in the baseline
BAFs for each trophic level) that was not due to differences in lipid content and freely dissolved
concentration of chemical is shown in Figure 5-2. Sources of the underlying variability could
include differences in nominal versus actual trophic level assignments for the individual species,
differences in disequilibrium of the ecosystem, and differences in age, size, growth rate, and/or
reproductive status of the individual organisms.
5-7
-------
5.2 METHOD 2: DERIVING BASELINE BAFS FROM BSAFs
When acceptable BAF^s are not available for a nonionic organic chemical with a log Kow
of • 4, EPA recommends the use of BSAFs to predict the baseline BAF as the second method in
the BAF data preference hierarchy under Procedure 1 or 2. Although BSAFs may be used for
measuring and predicting bioaccumulation directly from concentrations of chemicals in surface
sediment, they also can be used to estimate baseline BAFs (USEPA, 1995b; Cook and Burkhard,
1998). Because BSAFs based on field data incorporate the effects of metabolism,
biomagnification, growth, and other factors, baseline BAFs estimated from BSAFs will also
account for all these factors. The BSAF approach is particularly beneficial for developing
AWQCs for chemicals that are detectable in fish and shellfish tissues but are difficult to detect
and measure in ambient water. The BSAF method is also beneficial for measuring the degree to
which bioaccumulation is reduced for chemicals, such as polychlorinated dibenzo-p-dioxins,
dibenzofurans, certain biphenyl congeners, and polycyclic aromatic hydrocarbons, through
metabolism in food webs or the species of concern.
Prediction of a baseline BAF from a BSAF requires data for one or more reference
chemicals for which concentrations in ambient water, as well as sediment, can be measured,
preferably from a common sediment-water-biota data set. This method, in effect, translates
relative differences between measured BSAFs for two chemicals into relative differences in
baseline BAFs for the chemicals when the baseline BAF for one chemical cannot be measured.
Relative differences in bioaccumulation can be accurately measured when each chemical's
concentrations are analyzed from the same or equivalent environmental samples collected from a
site suitable for this purpose. BSAFs must be measured for the chemical of interest in order to
provide the basic measure of the chemical's bioaccumulation potential. Specifically, this method
uses measured sediment-water concentration quotients (• socws) for reference chemicals to
estimate values of C " that cannot be measured for the chemical of interest. Each chemical's Kow
must also be acquired, because the ratio of • socw to Kow provides the basis for relating reference
chemicals to the chemicals of interest. The following sections describe more completely the
determination of BSAF values; the relationship of baseline BAFs to BSAFs; the derivation of the
BSAF method equation; sampling and data quality considerations; assumptions and limitations
associated with the method; and the validation of this method for estimating baseline BAFs with
data from Lake Ontario and other ecosystems.
5-8
-------
2,3,4'-Trichlorobiphenyl
(PCB22) log Kow = 5.58
< 7
° 6
2,2',5,5'-tetrachlorobiphenyl
(PCB52) log Kow = 5.84
< 7
° 6
TL 3 B TL 3 F TL 4 B TL 4 F
TL 3 B TL 3 F TL 4 B TL 4 F
2,2',3,4,4'-pentachlorobiphenyl
(PCB85) log K =6.30
< 7
° 6
2,2',3,4',5',6-hexachlorobiphenyl
(PCB149) log K =6.67
< 7
° 6
TL 3 B TL 3 F TL 4 B TL 4 F
TL 3 B TL 3 F TL 4 B TL 4 F
2,3',4,4',5-pentachlorobiphenyl
(PCB118) log K,,w= 6.74
< 7
° 6
2,2',3,4',5,5'-hexachlorobiphenyl
(PCB146) log K,,w= 6.89
< 7
° 6
TL 3 B TL 3 F TL 4 B TL 4 F
TL 3 B TL 3 F TL 4 B TL 4 F
Figure 5-2. Box plots comparing baseline (TL 3 or 4 B) and field-measured (TL 3 or 4 F) BAFs for six PCB congeners obtained
from Green Bay, Lake Ontario, and Hudson River ecosystems for 13 fish species with samples segregated according to year
classes and sampling location, e.g., 4-year-old walleye from zone 4 in Green Bay and adult perch from RM 194 in the Hudson
River. For box plots, the median is the line inside the box, the 25th and 75th percentiles are the ends of the box, the 10th and 90th
percentiles are the T-lines, and outliers, points beyond the 10th and 90th percentiles, are the dots (•).
5-9
-------
5.2.1 Determination of BSAF Values
As shown in Equation 5-3, the BSAF is determined by relating the lipid-normalized
concentration of a chemical in a tissue or organism to the organic carbon-normalized
concentration of the chemical in surface sediment. A BSAF is expressed in grams of organic
carbon in sediment per gram lipid in tissue.
c
BSAF = —L (Equation 5-3)
where:
C. = lipid-normalized concentration of chemical in tissues
Csoc = concentration of chemical in dry sediment, normalized to sediment organic
carbon
The lipid-normalized concentration of a chemical in an organism (C.) is determined by:
„ _ Ct
S - — (Equation 5-4)
where:
C, = concentration of chemical in tissue
f. = fraction of the tissue that is lipid
The sediment organic carbon-normalized concentration (Csoc) is determined by:
Cs
-f— (Equation 5-5)
eoc
where:
Cs = concentration of chemical in dry sediment
fsoc = fraction of dry sediment that is organic carbon
The appropriate use of BSAFs for calculation of baseline BAFs does not require the
existence of steady state between the chemical mass loading and concentrations in sediments.
However, BSAFs are most useful when measured under conditions in which chemical
concentrations in water are linked to slowly changing concentrations in sediment. BSAFs
5-10
-------
measured when concentrations in water are rapidly changing, either through onset of
contamination or the abrupt cessation of loading to the water, are likely to be unreliable without
additional modeling to extrapolate the values to longer term or steady-state conditions.
BSAFs are rarely measured for ecosystems at thermodynamic equilibrium, so a BSAF
inherently includes a measure of the "disequilibrium" associated with the distribution of a
chemical in the ecosystem. The deviation of a BSAF from the expected equilibrium value of
approximately 1-2 is determined by the net effect of all factors that contribute to the
disequilibrium between sediment and aquatic organisms. A value greater than 1-2 can occur
through biomagnification or when surface sediment has not reached steady state with water. A
value of less than 1-2 can occur from diagenesis of organic carbon in sediments, kinetic
limitations for chemical transfer from sediment to water or water to the food web, and biological
processes (such as growth or metabolism/biotransformation of the chemical in biota or its food
web).
5.2.2 Relationship of Baseline BAFs to BSAFs
Both BSAFs and baseline BAFs can provide good measures of the relative
bioaccumulation potential of hydrophobic organic chemicals if based on accurate measurements
of concentrations in appropriate samples of biota, sediment, and water. When calculated from a
common organism-sediment-water sample set, chemical-specific differences in BSAFs or
baseline BAFs reflect the net effect of biomagnification, metabolism, bioenergetics, and
bioavailability factors on each chemical's bioaccumulation. The purpose of method 2 is to
convert the bioaccumulation information contained in a measured BSAF to the corresponding
baseline BAF value for a chemical. The relationship between a measured BSAF for a chemical
and its baseline BAF depends strictly on the value of the chemical's sediment-water
concentration quotient, (• socw). Method 2 uses measurements of* socw for reference chemicals (r),
(* socwXs, to determine the value of (• socw) for a chemical of interest, i, which is unmeasurable.
(* socwXs are determined by:
= spc'r
(Equation 5-6)
where:
(Csoc)r = concentration of a reference chemical in dry sediment, normalized to
sediment organic carbon
(C")r = concentration of the reference chemical that is freely dissolved in water
From the definitions of BAF? (Equation 2-5), BSAF (Equation 2-14), and • socw, the
sediment-water column concentration quotient (Equation 2-18), the relationship between • S(
BAF? and BSAF may be derived for chemical i:
5-11
-------
J CBSAF,fd \
= (BSAF). (Equation 5-7)
Equation 5-7 can be rearranged to give (BAF?); = (BSAF); • (• socw);. Then (BSAF); •
(* socw); can be substituted for (BAF?); in equation 2-4 to express the baseline BAF for chemical i
as:
(Baseline BAfy = (BSAF); (17^ - = (Equation 5-8)
Equation 5-8 reveals that a baseline BAF could be directly estimated from a BSAF if a
reasonably certain estimate of • socw is available for the chemical. Since ecosystems are often not
under steady-state chemical loading conditions, uncertainty is expected to be less when (• socw); is
based on measurements for chemicals with similar Kows.
5.2.3 Derivation of the Baseline BAF Equation for Method 2
In many cases, the fugacity ratios between sediments and water (• SOCW/KOW) for both
reference chemicals and the chemical of interest are arguably similar. In fact, this similarity
provides a useful criterion for the selection of reference chemicals. In cases where evidence exists
for a significant difference, the explicit difference may be represented by Dj/r:
D . . „ , ,
" ' OL-X (Equa,,on5-
Thus,
.
(Equation 5-10)
By substituting Equation 5-10 into Equation 5-8, the method 2 equation (5-11) is
obtained. For each aquatic species for which a field-measured BSAF for a chemical of interest, i,
5-12
-------
is available, a baseline BAF may be calculated using the following equation with an appropriate
value of (• Socw/Kow)r:
(Basdine BAF)i = (BSAF)i f | - (Equation 5-11)
where:
(BSAF); = biota-sediment accumulation factor for chemical of interest "i"
(* socwX = sediment-water column concentration quotient of reference chemicaF'r"
(Kow); = w-octanol-water partition coefficient for chemical of interest "i"
(Kow)r = w-octanol-water partition coefficient for reference chemical "r"
Dj/r = ratio between • SOCW/KOW for chemicals "i" and "r"
(normally chosen so that Dj/r = 1)
5.2.4 Sampling and Data Quality Considerations
Reference chemicals with • SOCW/KOW similar to that of the chemical of interest are preferred
for method 2 and often are available. Theoretically, the difference between sediment-to-water
fugacity ratios for two chemicals, "i" and "r" (Dj/r), can be used when reliable reference chemicals
that meet the fugacity equivalence condition are not available. Nonionic organic chemicals with
concentrations in water at approximate steady state with respect to concentrations in surface
sediments should have similar, if not equal, values of • SOCW/KOW that are related to the fraction of
organic carbon in suspended solids when compared with the fraction of organic carbon in the
surface sediments. When steady-state conditions are not present, as is often the case, • SOCW/KOW
values for related chemicals may be similar. Similarity of* SOCW/KOW for two chemicals can be
indicated on the basis of similarities in molecular structure, which lead to similar physical
chemical behavior in water (persistence, volatilization), similar mass loading histories, and similar
concentration profiles in sediment cores. In many cases, PCBs serve as effective reference
chemicals.
The following sampling and data quality considerations should be met when field-
measured BSAFs are used to predict BAFs:
1. The reference chemicals and the chemical of interest should have similar
physicochemical properties, as well as persistence in water and sediment.
2. When possible, (• socw)r data for several reference chemicals with similar Kows
should be obtained from the same water and sediment samples to ensure that
predictions are more robust than those that would be obtained with only one
reference chemical.
5-13
-------
3. Data for several reference chemicals and the chemical of interest should come
from a common organism-water-sediment data set for a particular site. (Csoc)r and
(Csoc)i should be measured from the same sediment samples, because this
eliminates uncertainty attributable to spatial heterogeneity of Csoc.
4. The Kow value for the target and reference chemicals should be selected as
described in Section 4.5 of this TSD.
5. Whenever possible, the loadings history of the reference chemicals and the
chemical of interest should be similar, such that their sediment-water
disequilibrium ratios (• SOCw/Kow) would not be expected to be substantially
different (Dj/r ~ 1).
6. Samples of surface sediments (0-1 cm is ideal) should be collected from locations
in which carbonaceous sediment, containing the reference chemicals and the
chemical of interest, is regularly deposited and is representative of average surface
sediment in the vicinity of the organism.
7. All sampling and data quality considerations described in Section 5.1.1 for
determining BAF^s should also be met.
5.2.5 Assumptions and Limitations
Although EPA is currently restricting the application of this method for baseline BAF
derivation to nonionic organic chemicals with a log Kow of • 4, this restriction primarily reflects
lack of validation of this method as applied to chemicals with a log Kow of < 4. In addition, the
need for this method is greater for chemicals with higher log Kows because of the difficulties
associated with detecting and measuring such chemicals in ambient water. Future development
and evaluation of this method may lead to its application to a broader range of chemicals.
The primary assumptions and limitations discussed in Section 5.1.1 for method 1 also
apply to method 2. The primary limitation associated with method 2 for calculating and applying
the baseline BAF—namely, variability of C™—is common to all methods and models for
predicting and measuring BAFs. In deriving Equation 5-10, the assumption is made that • socw
values are chosen from a common sediment data set (i.e., both BSAF and • socw are based on the
same value for Csoc). In the event that this cannot be done, the relative percent error in the
baseline BAF associated with the Csoc inequality will equal 100 times the difference in Csocs used
for the BSAF and • socw, divided by the Csoc used for the BSAF measurement.
Although EPA recommends that Csoc values represent spatially averaged surface sediment
contamination levels in the region affecting the organism's exposure, method 2 should be
accurate even when the Csoc value used for the BSAF and • socw does not well represent spatially
averaged conditions. This is because the Csoc need only reflect the relative level of contamination
of sediments over time. The magnitude of errors associated with fluctuations in C " will be the
5-14
-------
same for method 2 as for method 1. Temporal changes in C " are responsible for most deviations
from steady state between biota, water, and sediments.
Inaccuracies associated with the use of • SOCW/KOW from reference chemicals to estimate
C"s for chemicals of interest under method 2 have a linear impact on the accuracy of baseline
BAFs. For example, if* SOCW/KOW is 10 but the estimate used is 20, the calculated baseline BAF
will be greater than the true value by a factor of 2. The measurements of • SOCw/Kow to date indicate
an expected range of 5-40 for most contamination scenarios. If the data quality considerations
for choosing • SOCW/KOW for the chemical of interest are followed, the magnitude of the errors
associated with the choice of • SOCW/KOW should be no greater than twofold.
The strength of method 2 is that it utilizes measurements of relative (not absolute)
differences in bioaccumulation between chemicals with structural similarity. When properly
sampled, sediments provide time-stable measures of concentrations of persistent bioaccumulative
chemicals in aquatic systems. Method 2 is currently the only viable method for estimating
baseline BAFs for nonionic organic chemicals with (1) a log Kow of • 4, (2) concentrations in
water that are often undetectable, and (3) significant rates of chemical metabolism by organisms.
Important examples of chemicals with these characteristics are polychlorinated dibenzo-p-
dioxins, polychlorinated dibenzofurans, and non-orthochlorinated biphenyls.
5.2.6 Validation of Method 2
For method 2, validation efforts were conducted with data collected from three aquatic
ecosystems in the United States: Lake Ontario; Green Bay/Fox River, Wisconsin; and the
Hudson River, New York. EPA previously published information on validation of the method 2
approach by using data on PCBs, chlorinated benzenes, pesticides, and 2,3,7,8-
tetrachlorodibenzo-p-dioxin (TCDD) collected from Lake Ontario and the mid-bay region of
Green Bay (USEPA, 1995c). Baseline BAFs for PCBs, chlorinated benzenes, and some pesticides
were predicted from BSAFs for Lake Ontario salmonids and compared with measured baseline
BAFs from the same system. The baseline BAFs predicted from BSAFs were within a factor of 4
of the measured baseline BAFs. Furthermore, when predicted baseline BAFs for TCDD and
PCBs from Green Bay salmonids and Lake Ontario brown trout were compared, the baseline
BAFs predicted from BSAFs were generally within a factor of 2 of the measured baseline BAFs.
Although there were a few outliers in the observed trends, the results of this validation effort
showed method 2 generally works well, not only for predicting baseline BAFs with data from the
same ecosystem (Lake Ontario), but also for predicting baseline BAFs between systems (Green
Bay vs. Lake Ontario).
For this TSD, Burkhard et al. (2003a) extended the previous validations for method 2 by
comparing results of field-measured baseline BAFs with baseline BAFs predicted from BSAFs
using additional PCB data collected from Green Bay/Fox River and the Hudson River. The data
sets for this latest validation effort were selected from the 1989-1990 Green Bay Mass Balance
Study (http://www.epa.gov/grtlakes/gbdata) and the Hudson River PCBs Reassessment Remedial
Investigation/Feasibility Study (USEPA, 1998). The former study included data from the lower
Fox River and the inner, middle, and outer zones of Green Bay. The Hudson River data were
5-15
-------
collected over several years by a number of Federal and State agencies and private groups and
were assembled into a single database (USEPA, 1998) from which data were selected for this
analysis. The reference PCB congeners used in this validation effort included those used in the
previous validations (PCB 52, 105, 118) (USEPA, 1995b) as well as PCBs 18, 28, 149, 174, and
180. This validation was performed using the geometric mean of the baseline BAFs predicted by
using as many as possible of the eight reference PCB congeners listed above. As noted
previously (see Section 5.4.2), EPA recommends that several reference chemicals be used with
method 2 and that Kows be matched as closely as possible because slightly smaller predictive
errors were observed in the validation study when the chemicals of interest and the reference
chemicals had more closely matched Kows (Burkhard et al., 2003a). The recent validation effort by
Burkhard et al. (2003a) also included baseline BAFs for several fish species in addition to
salmonids (e.g., carp, walleye, shad, alewife, yellow perch, white perch, pumpkinseed, red-
breasted sunfish, and largemouth bass), some of which spanned several age classes.
A summary of the validation exercise is presented here and a detailed discussion is
provided by Burkhard et al. (2003a). Baseline BAFs predicted with method 2 were plotted against
field-measured baseline BAFs. The geometric mean baseline BAF predicted from BSAFs are
plotted; these differ from the individual reference chemical predictions only in terms of vertical
displacement due to the differences in congener-specific • socw. The ratio of predicted-to-
measured congener-specific baseline BAFs (BAFpredicted/BAFmeasured) was used to evaluate the
agreement between method 2-derived baseline BAFs and field-measured baseline BAFs. Table
5-2 presents zone (Green Bay data) and location-specific (Hudson River data) statistics for the
BAFpredicted/BAFmeasured ratio. Table 5-2 also presents the percentage of BAFpredicted/BAFmeasured ratios
that fall within specified ranges of the distribution. In general, the agreement between method 2-
predicted baseline BAF and field-measured baseline BAF values is very good, with a majority of
predicted BAF values falling within a factor of 2 of the field-measured BAF values. In addition,
>90% of method 2-predicted BAFs (95% from Green Bay and 91% from Hudson River) are
within a factor of 5 of the field-measured baseline BAFs. Table 5-3 presents exceedance levels
(i.e., certain points within the data distribution) for the ratio of predicted to measured congener-
specific baseline BAFs (BAFpredicted/BAFmeasured) for each fish species and ecosystem zone/location.
For most zones in Green Bay, the 95% exceedance levels (i.e., 95% of the BAFpredicted/BAFmeasured
values) fall within the range of 0.2 (one-fifth of the predicted baseline BAF) to 5.0 (five times the
predicted baseline BAF). Results for the Hudson River indicated generally similar agreement
between method 2-predicted baseline BAFs and field-measured baseline BAFs, although
exceedance levels are noticeably wider at river mile (RM) 169. Method 2 also appears to
overpredict BAFs at RM 122 and 114, the only locations where bias in this method was found.
Overall, these analyses support the use of method 2 to estimate baseline BAFs from field-
measured BSAFs.
5-16
-------
Table 5-2. Validation Statistics for Method 2: Ratio of Baseline BAFpredicte(/Baseline
measured
Location
Green Bay
Zone 1
Zone 2a
Zone 3 a
Zone 3b
Zone 4
All zones
Hudson River
RM194
RM189
RM169
RM144
RM122
RM114
All stations
95%
0.39
0.25
0.21
0.16
0.31
0.22
0.46
0.33
0.11
0.67
0.70
1.20
0.13
Method 2:
Mean
0.88
1.27
1.25
1.08
3.33
1.53
1.12
1.00
2.01
1.19
2.43
3.86
1.50
Exceedance Levels and Comparison Statistics
Median
0.88
0.89
0.73
0.69
1.07
0.81
0.99
1.03
0.59
0.97
2.16
3.78
1.10
5%
1.66
3.47
3.78
2.71
3.79
3.29
2.12
1.55
9.91
2.14
4.81
6.91
4.42
% within 2x
87.6
69.8
51.5
53
31.9
55.7
81.9
87.5
19.0
92.3
45.8
16.7
64.9
% within 5x
100
92.8
94.1
91.4
97.4
94.5
95.2
100
68.3
100
95.8
83.3
90.7
Includes number of species and range in "n" across sites
RM = river mile.
Table 5-3. Exceedance Levels for Ratio of Method 2-Predicted Baseline BAFs
(Geometric Mean) to BAF^s from Green Bay and Hudson River
Location
Green Bay Zone 1
Adult alewife
Age 1 carp
Age 1 walleye
Age 3 walleye
Age 4 walleye
Green Bay Zone 2a
YOY alewife
Adult alewife
Age 2 carp
Age 8 carp
YOY shad
YOY smelt
Adult smelt
Age 3 walleye
Age 4 walleye
Green Bay Zone 3a
YOY alewife
Adult alewife
Age 1 carp
5%
0.34
0.4
0.4
0.4
0.4
0.27
0.29
0.33
0.27
0.2
0.27
0.27
0.27
0.27
0.22
0.22
0.24
10%
0.45
0.48
0.5
0.49
0.49
0.37
0.42
1.42
0.37
0.3
0.36
0.37
0.37
0.37
0.3
0.31
0.28
Percentile
90%
1.27
1.21
1.26
1.26
1.2
2.88
2.89
2.89
2.88
2.88
2.89
2.88
2.88
2.88
2.62
2.45
2.44
95%
1.39
1.4
1.61
1.63
1.4
3.22
3.28
3.31
3.22
3.19
3.25
3.22
3.22
3.22
3.87
3.25
2.94
5-17
-------
Location
YOY smelt
Adult smelt
Age 2 brown trout
Age 3 brown trout
Age 4 walleye
Green Bay Zone 3b
Adult alewife
Age 8 carp
Age 10 carp
YOY smelt
Adult smelt
Age 2 brown trout
Age 3 brown trout
Age 3 walleye
Age 4 walleye
Green Bay Zone 4
Adult alewife
Age 10 carp
YOY smelt
Adult smelt
Age 2 brown trout
Age 3 brown trout
Age 4 walleye
Age 5 walleye
Hudson River RM 194
Carp
Yellow perch
Red-breasted sunfish
Largemouth bass
Hudson River RM 189
Yellow perch
Pumpkinseed
Red-breasted sunfish
Largemouth bass
Hudson River RM 169
Yellow perch
Pumpkinseed
Red-breasted sunfish
Largemouth bass
Hudson River RM 144
Yellow perch
Hudson River RM 122
Yellow perch
White perch
Hudson River RM 114
White perch
5%
0.2
0.2
0.2
0.2
0.2
0.15
0.15
0.17
0.15
0.15
0.15
0.15
0.17
0.15
0.31
0.33
0.31
0.29
0.31
0.32
0.32
0.32
0.46
0.51
0.46
0.46
0.32
0.33
0.33
0.33
0.11
0.11
0.11
0.11
0.67
0.86
0.64
1.2
10%
0.27
0.26
0.26
0.26
0.26
0.26
0.23
0.27
0.23
0.23
0.23
0.23
0.27
0.23
0.33
0.34
0.33
0.33
0.33
0.34
0.34
0.34
0.53
0.66
0.54
0.54
0.46
0.48
0.5
0.5
0.12
0.12
0.12
0.12
0.7
1.05
0.77
1.43
Percentile
90%
2.74
2.53
2.45
2.59
2.44
2.27
2.18
2.25
2.25
2.25
2.25
2.25
2.26
2.24
2.92
2.91
2.92
3
2.91
2.96
2.91
2.91
1.78
1.85
1.77
1.77
1.53
1.52
1.52
1.52
4.79
4.97
4.79
4.79
1.82
4.41
3.29
4.99
95%
3.93
3.82
3.21
3.85
3.17
2.66
2.67
2.65
2.65
2.65
2.65
2.65
2.66
2.65
3.79
3.37
3.44
3.81
3.29
3.8
3.34
3.32
2.02
2.06
2.01
2.01
1.78
1.75
1.73
1.73
6.85
7.09
6.85
6.85
2.14
5.37
4.01
6.91
RM = river mile; YOY = young of year.
5-18
-------
5.3 METHOD 3: DERIVING BASELINE BAFs FROM LABORATORY-MEASURED
BCFs (BCF js) AND FCMs
Method 3 in Procedure 1 is appropriate for nonionic organic chemicals that have
moderate-to-high hydrophobicity (log Kow • 4) and low potential for being metabolized. For
method 3, a laboratory-measured BCF (BCFj) and FCM are used to predict a baseline BAF. The
BCFj must be used in conjunction with an FCM because nonaqueous routes of exposure and
subsequent biomagnification are of concern for the types of chemicals to which Procedure 1
applies. Although a ECF{ accounts for chemical metabolism that occurs in the organism used to
calculate the BCFj, it does not account for metabolism that may occur in other organisms of the
aquatic food web. Method 3 uses the following baseline BAF and BCF equations:
Baseline BAP = FCM
(Equation 5-12)
where:
= Total BCF (BCF^ = C,/CW)
ffd = fraction of the total concentration of chemical in water that is freely
dissolved
f. = fraction of the tissue that islipid
FCM = the food-chain multiplier for the appropriate trophic level, obtained by
linear interpolation (Table 4-6) or from appropriate field data
The technical basis for Equation 5-12 is provided in Appendix A. Presented below are
detailed discussions and information on selecting appropriate BCFxS and FCMs and the
derivation of FCMs using food web models and field data.
5.3.1 Sampling and Data Quality Considerations
The BCFj should be calculated by using information on the total concentration of the
chemical in the tissue of the organism and the total concentration of the chemical in the
laboratory test water. The data used to calculate a ECF{ should be thoroughly reviewed to assess
the quality of the data and the overall uncertainty in the BCF value. The following general criteria
apply in determining the acceptability of BCF^. Because no guidance can address all of the
variation in experimental designs and data found in the literature, best professional judgment will
be necessary to supplement these data quality guidelines in selecting the best available
information and using it appropriately.
1. Aquatic organisms used to calculate a ECF{ should be representative of those
aquatic organisms commonly consumed in the United States. An aquatic
organism that is not commonly consumed in the United States can be used to
calculate an acceptable ECF{ provided that the organism is considered to be a
reasonable surrogate for a commonly consumed organism. Information on the
5-19
-------
ecology, physiology, and biology of the organism should be reviewed when
assessing whether an organism is a reasonable surrogate.
2. The test organism should not be diseased, unhealthy, or adversely affected by the
concentration of the chemical, because these conditions may alter accumulation of
chemicals.
3. The test organisms should be exposed to the chemical under flow-through or
renewal conditions.
4. The concentrations of the chemical in the laboratory test water must not exceed
the solubility of the chemical in water. Micelles, which indicate the chemical is not
dissolved, should not be present in the exposure water. Older ECF{ measurements
for highly hydrophobic chemicals, such as those with a log Kow of > • 6 are often
unreliable because solubility limits were exceeded or the chemical was present in
the exposure water in the form of micelles.
5. The total concentration of the chemical in the water should be measured and
should be relatively constant during the exposure period.
6. The concentrations of POC and DOC in the study water should be measured or
reliably estimated.
7. The percent of the tissue or organism that is lipid (i.e., fraction lipid, f. ) must be
measured or reliably estimated to permit lipid normalization.
8. The calculation of the ECF{ should appropriately address growth dilution, which
can be particularly important for poorly depurated chemicals.
9. Other aspects of the methodology used should be similar to those described by
the American Society of Testing and Materials (ASTM, 1990) and U.S. EPA
Ecological Effects Test Guidelines (USEPA, 1996).
10. If BCFj consistently increases or decreases as the concentration of the chemical
increases in the test solutions (and this variation is not due to changes in lipid
fraction of organisms, freely dissolved fraction of chemical in water, or changes in
health of the organisms), the BCF determined at the concentration of chemical
that is closest to the expected AWQC concentration should be used in deriving the
AWQC.
11. BCFj may be based on measurement of radioactivity only when the ECF{ is
intended to include metabolites, when there is confidence that there is no
interference due to metabolites of the parent chemical, or when studies are
conducted to determine the extent of metabolism, thus allowing for a proper
correction.
5-20
-------
12. All considerations described in Section 4.4.2 for determining FCMs should also be
met.
5.3.2 Assumptions and Limitations
In using method 3, EPA will assume that (1) a high-quality ECF{ is a better measure of
the bioconcentration potential of a chemical than simply assuming that the baseline BCF is equal
to Kow, an assumption used with method 4, (2) the measured BCFj and the baseline BAF
predicted with method 3 are independent of chemical concentration in the water, and (3) FCMs
account for biomagnification processes caused by the consumption of contaminated food in
aquatic food webs. Assumptions, limitations, and uncertainties associated with FCMs and
concentration independence of ECF{ and EAF{ are discussed in Sections 4.4.1 and 5.1.4,
respectively.
BCFj for chemicals that are metabolized by the test organisms incorporate the effects of
the metabolism on the concentration of chemical that is accumulated in the organism. However,
if induction of metabolic systems is required or co-occurring contaminants (i.e., that exist in the
environment) are required for the metabolism to take place, then the effect of metabolism may
not be captured in the BCFj measurement. Therefore, the range of effects of metabolism on
BCFj will be chemical specific. Nevertheless, EPA believes that high-quality BCFxS provide a
better measure of bioconcentration potential for chemicals than simply assuming the baseline
BCF is equal to the chemical's Kow because of the potential of the ECF{ to include the effects of
metabolic processes. Furthermore, BCFxS can be obtained for specific species of interest. This
specificity may reduce uncertainties associated with extrapolating bioaccumulation factors
among species with known or suspected differences in metabolic pathways or capacity.
For method 3, baseline BAFs are calculated with the FCM and the BCF^. As discussed in
Section 4.4, FCMs that will be used by EPA in deriving national BAFs are derived using the
Gobas food web model with a number of assumptions and input parameters, namely, no
metabolism, an assumed food web structure, and • SOCw/Kow= 23. Limitations and uncertainties
associated with the FCMs have been discussed previously (see Section 4.4.1), and these
limitations and uncertainties are incorporated into the baseline BAFs derived using method 3.
The baseline BAFs derived with method 3 for chemicals that are metabolized will not
include the effects of all metabolism processes because of the assumption of no metabolism used
in deriving the FCMs. However, the method will incorporate those metabolism processes or
effects that are captured in the BCF measurement. Baseline BAFs predicted from measured BCF
for chemicals that are metabolized will be smaller than those predicted from measured BCFs for
chemicals of equal hydrophobicity but which are not metabolized.
A major limitation associated with method 3 is the current lack of high-quality measured
BCF data for highly hydrophobic chemicals in any organism class. This lack of data is due
principally to the difficulties associated with performing BCF measurements for highly insoluble
(hydrophobic) chemicals. Conditions appropriate for performing these measurements are
described in Section 5.3.1. When evaluating literature BCF data, one often finds measurements
5-21
-------
performed with (1) conditions that do not meet current standards, for example, a solvent carrier
such as acetone is used to introduce the chemical into the aqueous phase or chemical solubilities
in water are exceeded, and (2) poor and/or incomplete reporting of measurement conditions and
parameters, for example, no lipid data, no POC and DOC data, and/or an inability to determine
whether steady-state conditions were obtained in the experiment. In addition, some BCFs were
measured with chemical mixtures, such as Aroclors, and resolving the effects of co-occurring
chemicals on micelle formation is often intractable. As BCF data become available for highly
hydrophobic chemicals in the future, the impact of this limitation will lessen.
5.3.3 Validation of Method 3
To date, EPA has performed only a limited number of evaluations of method 3 because of
a lack of BCF data of the appropriate quality, as described in Section 5.1.2. and in Section 5.3.2.
For example, EPA invested considerable effort in examining the scientific literature for measured
BCFs for PCB congeners and was not able to find BCFs of appropriate quality.
Burkhard et al. (1997) evaluated method 3 by using field data for chlorinated benzenes,
butadienes, and hexachloroethane from Bayou d'Inde, Lake Charles, Louisiana. The results of
this evaluation showed that field-measured baseline BAFs were within a factor of 3 for 88% and a
factor of 5 for 94% of the baseline BAFs predicted using method 3 (n = 32) (Figure 5-3). The
median of the ratios of the field-measured baseline BAFs to predicted baseline BAFs was 1.03,
and approximately one-half of the predicted baseline BAFs were less than the measured baseline
BAFs (53%, n = 32). The chemicals whose field-measured baseline BAFs were in least agreement
with the predicted baseline BAFs were hexachloroethane, Z-pentachlorobutadiene, and
hexachlorobutadiene for Callinectes sapidus (blue crab). Metabolism of these chemicals by C.
sapidus is suggested as the cause of the poor agreement between the field-measured BAFs and
the baseline BAFs predicted using method 3 (Burkhard et al., 1997).
Figure 5-3. Relationship between measured and predicted
baseline BAFs for method 3. The dotted and dashed lines
represent a factor of 3 and 5 difference between the
measured and predicted baseline BAFs, respectively.
Baseline BAFs measured using Callinectes sapidus (' ),
Micropoganias undulatus (' ), Fundulus heteroclitus (• ),
and Brevoortia patronus (• ).
2345
Predicted Log BasdheBAF
5-22
-------
5.4 METHOD 4: BASELINE BAF DERIVED FROM Kow x FCM
Method 4 in the tiered hierarchy of Procedure 1 consists of using Kow and an appropriate
FCM for estimating the baseline BAF. In Procedure 1, this method is used only for nonionic,
moderate to highly hydrophobic chemicals whose metabolism is considered negligible or is
unknown. In this method, the Kow is assumed to be equal to the baseline BCF, and thus the
organic carbon and lipid normalization procedures are not needed. To account for
biomagnification in method 4, the Kow value is multiplied by an appropriate FCM. Method 4 uses
the following baseline BAF equation:
Baseline BAF = Kow x FCM (Equation 5-13)
where:
FCM = food-chain multiplier for the appropriate trophic level, obtained by linear
interpolation (Table 4-7) or from appropriate field data
Kow = w-octanol-water partition coefficient
Detailed information on selection of appropriate FCMs and Kow values can be found in
Section 4.4 and Appendix B, respectively.
5.4.1 Assumptions and Limitations
A number of assumptions are associated with baseline BAFs predicted with method 4.
First, it is assumed that the Kow is equal to the chemical's baseline BCF for a non-metabolized
chemical. Second, it is assumed that there is no metabolism of the chemical in the food web.
Third, the assumptions incorporated into the FCMs—namely, • SOCW/KOW= 23, mixed benthic and
pelagic food web, and adequacy of the Gobas model—are directly incorporated into the
predictions made with method 4. Discussion of these assumptions and limitations is presented
below.
Method 4 assumes that the Kow is equal to the chemical's baseline BCF. Use of the Kow in
place of the baseline BCF is supported by equilibrium partitioning theory. This theory assumes
that (1) the bioconcentration process can be viewed as a partitioning of a chemical between the
lipid of aquatic organisms and water and the Kow is a useful surrogate for this partitioning process,
and (2) a linear relationship exists between the Kow and the BCF. Mackay (1982) demonstrated
the usefulness of Kow as a surrogate for this partitioning process by presenting a thermodynamic
basis for the partitioning process for bioconcentration. In theory, it follows that the baseline BCF
(i.e., BCF based on the concentration of chemical in lipid of organisms and freely dissolved in
water) should be similar, if not equal, to the Kow for organic chemicals. Numerous investigations
have provided empirical data to support this theory. As summarized by Isnard and Lambert
(1988), numerous studies have demonstrated a linear relationship between the log Kow for organic
chemicals and the log BCF measured for fish and other aquatic organisms exposed to those
chemicals. In addition, when the regression equations are constructed with BCFs reported on a
lipid-normalized basis, the slopes and intercepts are not significantly different from 1 and 0,
5-23
-------
respectively. For example, de Wolf et al. (1992) adjusted a relationship reported by Mackay
(1982) to a 100% lipid basis (lipid-normalized basis) and obtained the following relationship:
logBCF = 1.00 log K^H- 0.08 (Equation 5-14)
For chemicals with large log Kows (>6.0), reported BCFs are often not equal to the Kow
even for nonmetabolized chemicals, because the measurements were not performed and/or
reported with appropriate experimental conditions. BCFs for nonmetabolized chemicals are equal
to the Kow when the BCFs are reported on a lipid-normalized basis; determined using the
concentration of the chemical that is freely dissolved in the exposure water; corrected for growth
dilution; determined under steady-state conditions or from accurate measurements of the
chemical's uptake (kj) and elimination (k2) rate constants; and determined with no solvent
carriers in the exposure. If, when reviewing the literature for BCFs, EPA cannot verify that
measured BCFs are measured under the appropriate conditions, as described in Section 5.3.1,
EPA will use the Kow as an approximation of the baseline BCF.
As discussed in Section 3.1.4, method 4 is used only for nonionic organic chemicals with
log Kows greater than or equal to 4 and low rates of metabolism (Figure 3-1). The restriction of the
use of method 4 to only non-metabolized chemicals is based on the fact that the assumption that
Kow equals BCF is valid only for non-metabolizable chemicals. When a chemical is metabolized
by an organism during the measurement of BCF, the measured BCF will be smaller than the Kow.
In addition, as discussed in Section 4.4, FCMs are also calculated using the assumption that no
metabolism of the chemical takes place in the food web. If method 4 is used when metabolism of
the chemical occurs in the food web, predicted BAFs will be larger than field-measured BAFs.
For detailed information on the assumptions incorporated into the FCMs, refer to Section 4.4.
5.4.2 Validation of Method 4
As noted in Section 5.2.6, Burkhard et al. (2003a) have performed exercises to validate the
predictive power of methods 2 and 4. The validation exercises were performed by using data
collected from Lake Ontario, Green Bay/Fox River, the Hudson River, and Bayou d'Inde,
Louisiana. With these data sets, baseline BAFs predicted using method 4 were plotted against
field-measured baseline BAF values. The agreement between baseline BAFs predicted using
method 4 and field-measured baseline BAF values is generally good for Green Bay, although not
as good as was seen for method 2 (see Section 5.2.5 and Burkhard et al., 2003a). In Green Bay,
59% of the baseline BAFs predicted using method 4 are within a factor of 2 and 93% are within a
factor of 5 of the measured baseline BAFs (Table 5-4). The validation exercises using the Green
Bay/Fox River and Hudson River data are described in detail in Burkhard et al. (2003a).
5-24
-------
Table 5-4. Validation Statistics for Method 4: Ratio of Baseline BAFpredicted/Baseline
RA17
JJ-^VJ measured
Location
Green Bay
Zone 1
Zone 2a
Zone 2b
Zone 3 a
Zone 3b
Zone 4
All zones
Hudson River
RM194
RM189
RM169
RM144
RM122
RM114
All stations
95%
0.32
0.17
0.23
0.33
0.23
0.15
0.21
0.06
0.12
0.10
0.42
0.40
0.40
0.08
Method 4:
Mean
1.17
1.17
1.18
1.58
1.35
1.43
1.30
0.16
0.26
0.95
0.72
0.70
0.78
0.5
Exceedance
Median
0.89
0.74
0.83
1.05
0.90
0.61
0.84
0.11
0.20
0.41
0.67
0.67
0.73
0.24
Levels and Comparison
Statistics
5% % within 2x
2.75
3.40
3.01
4.71
4.15
5.28
3.90
0.38
0.55
1.89
1.14
1.27
1.29
1.07
69.8
54.6
61.0
64.0
60.5
40.5
58.6
3.6
9.0
35.3
76.5
80.0
76.9
26.3
% within 5x
98.1
91
94.9
94.7
94
82.2
92.7
25.3
55
76.5
100
100
100
60.7
RM = river mile.
The accuracy of baseline BAFs predicted with method 4 in the Hudson River varied
among sites. Generally, the predicted baseline BAFs are biased low; this is evident in Table 5-4,
where the mean and median predicted/measured ratios are less than 1 for all locations. At three of
the six stations in the Hudson River (river miles 114, 122, and 144), there was good agreement
between predicted and measured baseline BAFs (>75% within a factor of 2 and 100% within a
factor of 5; Table 5-4). However, for river mile 169, agreement was not as good (35% within a
factor of 2; 76% within a factor of 5). Finally, at two sites (river miles 189 and 194), there was
substantial underprediction of measured baseline BAFs with method 4. On the other hand, for
the Hudson River data set, the variability associated with baseline BAFs predicted using method
4 was generally smaller than that associated with method 2 (see Section 5.2.5 and Burkhard et al.,
2003a).
Factors that might be involved with the underprediction of the baseline BAFs for river
miles 169, 189, and 194 using method 4 include (1) the use of FCMs (Table 4-7) derived using
conditions and parameters for the nation instead of for the Hudson River, (2) the use of field
samples that were not temporally and/or spatially coordinated and/or representative of the
ecosystem, and (3) the sampling of an ecosystem with rapidly changing conditions in recent
5-25
-------
Table 5-5. Summary Statistics: Differences Between Log Baseline BAFs Predicted with
Method 4 and Log Baseline BAFs Measured from Lake Ontario (Oliver and Niimi, 1988)
for Chemicals with Log Kows Exceeding 4
Statistic
Average
Standard deviation
Count
Median
Within 2x
Within 5x
Negative residual
Positive residual
Sculpin
•0.01
0.35
51
•0.02
63%
94%
53%
47%
Alewife
•0.04
0.36
49
•0.06
59%
94%
53%
47%
Organism
Small
smelt
0.09
0.37
46
0.14
61%
94%
59%
41%
Large
smelt
•0.28
0.35
47
•0.30
47%
92%
72%
28%
Piscivorous fish
•0.08
0.36
57
•0.08
58%
96%
56%
44%
history due to the collapse of the Allen Mill gate in 1991. Burkhard et al. (2003 a) evaluated the
influences of site-specific • SOCw/Kow values, the most important site-specific parameter in
calculating FCMs (Burkhard, 1998), by deriving site-specific FCMs for river miles 169, 189, and
194 using • SOCW/KOW values of 40, 13, and 40, respectively, in comparison to the national value of
23 (see Section 4.4.1). The predictions using method 4 with the site-specific FCMs were still too
small. Increasing the • SOCw/Kow values for river miles!89 and 194 to 80 or 120 resulted in
agreement between predicted and measured baseline BAFs comparable to that observed in Green
Bay for method 4 (Table 5-4). The better agreement with much larger • socw values is consistent
with the possibility that the samples were not representative of the ecosystem. However, the
collapse of the Allen Mill gate (see below) 2 years prior to the collection of the field samples in
1993 introduces substantial and unresolvable uncertainty into the existing conditions in the river.
The gate collapse introduced elevated and variable PCB concentrations (by orders of
magnitude) into the river for approximately 15 months (QEA, 1999), and average concentrations
in water declined substantially from 1992 to 1993 for all congeners. Since the Allen Mill gate is
upstream of all the river mile locations used in this study, the most dramatic effects of this
episodic loading would be observed at the upstream locations, with a lessening of its effect with
increasing distance downstream. Better agreement with increasing distance downstream was
observed and is consistent with the expected lessening of the effect with increasing distance
downstream. Coincident with the gate failure, the lipid content of large mouth bass (and other
fishes as well) declined to extremely low levels of 0.25% in 1991 and then recovered in 1992 and
1993, to 1.0% and 1.5%, respectively. Although the causative factors are unclear for the rapid
decline in lipid contents, it appears that the food web was significantly disrupted. The overall
importance of the gate collapse and rapidly declining chemical concentrations between 1992 and
1993 and unusual lipid contents in fish for river miles 189 and 194 upon the bias observed with
method 4 at river miles 189 and 194 cannot be fully assessed with available data. At best, one can
conclude that river miles 189 and 194 were strongly influenced by the gate collapse and that
conditions in the river were not stable (e.g., sediment-water column chemical relationships,
chemical concentrations in the water, and fish lipid contents were rapidly changing). In the
derivation of FCMs for method 4, steady-state conditions in the food web are used, and chemical
concentrations in the sediment and water are constant over time. The conditions present in the
5-26
-------
river at river miles 189 and 194 clearly violate the latter assumption of relatively stable chemical
concentrations in sediment and water over time. Some or all of the above factors including
sample representativeness might be responsible for the observed constant biases at river miles
169, 189, and 194. These biases have also been observed using a complex, time-dependent food
web bioaccumulation model, calibrated to site-specific data for food web dynamics, species-
specific bioenergetics, and chemical uptake and elimination data in the same time period at the
same Hudson River locations as method 4 (QEA, 1999). These results further suggest unusual
conditions in the river.
Burkhard et al. (1997) evaluated the predictiveness of method 4 against field-measured
baseline BAFs for trophic level 3 fish sampled from the Bayou d'Inde for selected chlorinated
benzenes, chlorinated butadienes, and hexachloroethane. Bayou d'Inde is a lowland channel that
meanders through a brackish-freshwater marsh that is influenced by tide. This ecosystem is very
different from either the Great Lakes or the Hudson River and provides a useful demonstration of
the applicability of method 4 across different ecosystems. Because this evaluation of method 4
was conducted before the development of the final National BAF Methodology, it was performed
with FCMs and default values for POC and DOC that are marginally different from those that are
used in the National BAF Methodology (USEPA, 2000a). Burkhard et al. (1997) found good
agreement between the predicted and measured baseline BAFs for both the fish and invertebrates
sampled. Overall, approximately 90% of the predicted baseline BAFs were within a factor of 5 of
the measured baseline BAFs, and the median ratio of the predicted baseline BAFs to the
measured baseline BAFs was 1.64. As was observed in evaluating method 2, the baseline BAFs
predicted with method 4 for hexachloroethane, Z-l,l,2,3,4-pentachlorobuta-l,3-diene, and
hexachlorobuta-l,3-diene for blue crabs (C. sapidus) were smaller than the measured baseline
BAFs. Higher Phase II metabolism activities in this species for this class of chlorinated chemicals
is suggested to contribute to the difference between the predicted and measured BAFs.
The EPA also compared the baseline BAFs predicted with method 4 to measured BAFs
for the Lake Ontario ecosystem (Table 5-5). The average differences between measured and
predicted baseline BAFs were small for both forage and piscivorous fish, and more than 90% of
the baseline BAFs predicted with method 4 were within a factor of 5 of the measured BAFs. The
residuals (predicted minus measured) were evenly distributed, except for the large smelt. The
trophic level for the large smelt is estimated to be • 3.5, owing to its consumption of smaller
forage fish, and consequently, it was anticipated that the predicted baseline BAFs with trophic
level 3 FCMs would be slightly lower than the measured BAFs for this species.
As summarized above, the predictive accuracy of method 4 has been evaluated with field
data from four different ecosystems. For the Lake Ontario, Green Bay/Fox River, and Bayou
d'Inde ecosystems, baseline BAFs predicted with method 4 were in excellent agreement with the
measured BAFs: More than 90% of the predicted baseline BAFs were within a factor of 5 of the
measured baseline BAFs. In the Hudson River, for three of the sampling stations, baseline BAFs
predicted with method 4 were in excellent agreement with measured BAFs: 100% of the
predictions were within a factor of 5 of the measured baseline BAFs. For the other three sampling
stations in the Hudson River, baseline BAFs predicted with method 4 were much smaller than the
5-27
-------
measured BAFs, but the predictions were consistent with those based on the complex site-
specific time-dependent food web bioaccumulation model.
Overall, EPA believes that method 4 provides excellent predictions for ecosystems that
have not recently experienced a major change or disruption in chemical loadings or flows. Of all
the ecosystems examined, the extreme temporal dynamics observed for several important factors
(e.g., fish lipid content, food web structure, exposure concentrations) in the Hudson River makes
this site a severe test of all the BAF methodologies. In fact, the Hudson River data set may
arguably fail to meet the sampling and data quality considerations specified in Section 5.1.2 for
deriving baseline BAFs from field data. Nonetheless, EPA believes that the application of the
BAF methods to this location was a useful exercise and illustrates that useful predictions are
possible using method 4 in ecosystems with extreme temporal dynamics.
5-28
-------
6. DERIVATION OF NATIONAL BAFs FOR NONIONIC ORGANIC CHEMICALS
This section describes the process and technical documentation for determining national
BAFs for nonionic organic chemicals once all appropriate individual baseline BAFs have been
determined by using the methods described in Section 5. The distinction between national and
baseline BAFs is important and is illustrated by Figure 3-2 in Section 3. Specifically, baseline
BAFs are BAFs that have been adjusted to account for the lipid content of the tissues and the
amount of freely dissolved chemical in water. As explained in Section 4, these two factors are
important in affecting the bioaccumulation of nonionic organic chemicals. However, baseline
BAFs are not directly used to determine national human health AWQC, because they do not
reflect the lipid content of target aquatic organisms and the fraction of chemical that is freely
dissolved in water for the sites to which the AWQC applies. In effect, baseline BAFs are
"normalized" by lipid fraction and are based on the freely dissolved chemical (i.e., expressed on a
100% lipid and 100% freely dissolved basis). Furthermore, baseline BAFs need to be converted
to BAFs, expressed as total concentrations in tissue and water, to be compatible with national
human health AWQCs, which are based on the total concentration of a chemical in water.
To calculate national BAFs from baseline BAFs, two additional steps must be taken. First,
a final baseline BAF must be determined for each trophic level from all appropriate individual
baseline BAFs calculated by the methods described in Section 5. Guidance for determining a final
baseline BAF is provided in Section 6.1. After a final baseline BAF has been selected for each
trophic level, national BAFs are calculated for each trophic level using information on lipid
fraction of consumed aquatic organisms and the fraction of chemical that is freely dissolved in
water at the sites to which the AWQC applies. The calculation of a national BAF from a final
baseline BAF is shown in Equation 6-1.
National RAF^ = [(Final Baedine RAF^ ' (Q^ + 1] ' (fa> (Equation 6-1)
where:
Final Baseline BAF = mean baseline BAF for trophic level "n"
f.(TLn) = fraction of tissue that is lipid in aquatic organisms at trophic
level "n"
ffd = fraction of the total concentration of chemical in water that is
freely dissolved
For deriving national BAFs, EPA uses national default values of lipid fraction (f.) that are
specific to each trophic level. The national default values of lipid fraction are:
Trophic Level 2: 0.019
Trophic Level 3: 0.026
Trophic Level 4: 0.030
6-1
-------
These national default values reflect consumption-weighted mean values of the lipid fraction of
aquatic organisms that are commonly consumed throughout the United States. The technical
basis of EPA's national default values for lipid fraction is presented in Section 6.2.
The same equation used to estimate the fraction of a chemical that is freely dissolved (ffd)
for deriving baseline BAFs (Equation 4-6) is used here to estimate the chemical's freely dissolved
fraction for deriving national BAFs. For deriving national BAFs, EPA uses national default values
DOC and POC for estimating a representative fraction of chemical that is freely dissolved in U.S.
surface waters. The national default values of DOC and POC are:
DOC: 2.9mg/L
POC: 0.5 mg/L
These national default values reflect central tendency estimates of DOC and POC for bodies of
water distributed throughout the United States. The technical basis of EPA's national default
values for POC and DOC is presented in Section 6.3.
Once the national BAFs are determined for each trophic level, they are used in Equation
1-1, 1-2, or 1-3 in the 2000 Human Health Methodology to derive national human health
AWQCs. As discussed earlier in this document, both the fraction of tissue that is lipid (f.) and the
fraction of chemical that is freely dissolved (ffd) are two parameters that may be adjusted by
States and Tribes to reflect local or regional conditions. Details on adjusting national BAFs to
reflect local or region-specific values for lipid fraction of consumed aquatic organisms and the
fraction of chemical that is freely dissolved at the site(s) of interest are provided in a subsequent
volume of this TSD (Volume 3: Development of Site-Specific Bioaccumulation Factors).
6.1 SELECTING FINAL BASELINE BAFs
Once individual baseline BAFs have been determined by using the appropriate BAF
methods within the applicable BAF derivation procedure, the next step in deriving national BAFs
for nonionic organic chemicals consists of selecting the final baseline BAF for each trophic level
(Section 3, Figure 3-2). As shown by Equation 6-1, final baseline BAFs are used to derive
national BAFs by adjusting for the organic carbon content expected in representative U.S. surface
waters and the lipid content of commonly consumed aquatic organisms. Determination of the
final baseline BAF for each trophic level from individual baseline BAFs essentially involves a
series of data aggregation steps. First, for each BAF method and trophic level, the mean of the
corresponding individual baseline BAFs is calculated for each species to produce a set of
"species-mean baseline BAFs." Next, for each BAF method and trophic level, the mean of the
corresponding species-mean baseline BAFs is calculated to produce a set of "trophic level-mean
baseline BAFs." Finally, a single "final baseline BAF" is selected or derived for each trophic level
from the available set of trophic level-mean baseline BAFs. Although simple in concept, the
process for calculating final baseline BAFs involves the use of best professional judgment in
combination with other considerations, including the data preference hierarchy, the relative
uncertainty among BAF estimates, and the weight of evidence among BAF methods. A summary
of the steps involved in determining final baseline BAFs is provided in Section 5.4 of the 2000
6-2
-------
Human Health Methodology. Additional guidance for determining final baseline BAFs is
described below.
6.1.1 Calculating Species-Mean Baseline BAFs for Each BAF Method
For each trophic level and BAF method combination, species-mean baseline BAFs are
calculated as the geometric mean of acceptable baseline BAFs. An illustration of this step is
provided in Figure 6-1 for the four BAF methods of Procedure 1. Procedure 1 is the derivation
procedure that applies to nonionic organic chemicals with moderate-to-high hydrophobicity and
negligible or unknown metabolism rates. All four BAF methods can be used in Procedure 1. Each
unique species-trophic level-BAF method combination "sub-cube" in Figure 6-1 may consist of
multiple baseline BAFs (illustrated on the right face of the entire cube in Figure 6-1 for BAF
methods 1, 2, and 3 under trophic level 4). Species delineations are lacking for method 4 (Kow x
FCM) because this method produces a single baseline BAF for each trophic level. This is
illustrated by the single "baseline BAF column" for each trophic level of method 4 rather than a
sub-cube for each species. For illustration purposes, Figure 6-1 implies that, for each of the
applicable BAF methods, one or more baseline BAFs are available for each sub-cube in trophic
level 4. In practice, acceptable baseline BAFs may not be available for a BAF method or may be
available for only one or two trophic levels for a BAF method.
/ / /
/ / /
/ / /
/ / /
Meaa BAF
(Method l,TL2,spl)
Meaa BAF
(Method l,TL2,sp2)
Meaa BAF
(Method l,TL2,sp3)
Meaa BAF
(Method l,TL3,spl)
Meaa BAF
(Method l,TL3,sp2)
MeaaBAF
(Method l,TL3,sp3)
MeaaBAF
(Method l,TL4,spl)
MeaaBAF
(Method l,TL4,sp2)
MeaaBAF
(Method l,TL4,sp3)
I
t
t
/
$
/ >
^
/
#
'*
/
*
/(
V
1
^
/r
\
£
'\
b
^
N
Trophic Level
Figure 6-1. A schematic illustrating the aggregation of baseline BAF data by species, trophic level, and
BAF method type for nonionic organic chemicals under Procedure 1.
6-3
-------
When species-mean baseline BAFs are being calculated, individual baseline BAFs should
be reviewed carefully to assess their quality, variability, and overall uncertainty. This evaluation
will support decisions about whether to exclude certain baseline BAFs from the calculation of
national BAFs. This evaluation will usually be qualitative, because the availability of
bioaccumulation data is currently limited for many chemicals. Highly uncertain baseline BAFs
should not be used. Large differences in individual baseline BAFs for a given species (e.g., greater
than a factor of 10) should be investigated further. Some or all of the baseline BAFs for a given
species might not be used. Although all of the procedural and quality assurance guidelines
described in Section 5 apply for evaluating the quality, variability, and uncertainty of baseline
BAFs, several issues of special concern are discussed below. These issues include:
1. Temporal and spatial averaging of chemical concentrations
2. Spatial and temporal connectivity of samples
3. Chemical loadings history and steady state
4. Differences in food web structure
5. Reliability of lipid and organic carbon measurements
Temporal and Spatial Averaging of Chemical Concentrations
The extent of temporal and spatial averaging of chemical concentrations that are used to
calculate each baseline BAF should be explicitly evaluated as part of determining its overall
reliability (or uncertainty). Sufficient temporal and spatial averaging of chemical concentrations is
critical for accurately determining baseline BAFs from BAF^s and BSAFs (i.e., methods 1 and 2
of Procedure 1). As discussed previously, the extent of averaging that is considered ideal varies
according to the chemical properties (e.g., hydrophobicity) and the variability of chemical
concentrations in compartments of the ecosystem (e.g., water, sediment, tissue). Greater spatial
and temporal averaging will be needed for highly hydrophobic chemicals in ecosystems showing
high variability in chemical concentrations than for chemicals of lower hydrophobicity or those
from ecosystems with low variability in chemical concentrations (see Section 1.2). Spatial
sampling of chemical concentrations should in all likelihood span the immediate/local home
range of the aquatic species. As illustrated by Figure 1-1, variability of chemical concentrations in
water does not necessarily equate to variability of chemical concentrations in sediment and biota,
particularly for highly hydrophobic chemicals where variability of chemical concentrations in
sediments and biota tends to be dampened compared to that in water. Therefore, the extent of
temporal averaging of chemical concentrations required to achieve accurate estimates of baseline
BAFs that reflect steady-state conditions may vary depending on the environmental
compartment. For baseline BAFs derived from BAF methods 1 and 2, which incorporate
measurements of chemical concentrations in ambient water, variability in chemical
concentrations in water may be especially important.
In situations where variability in chemical concentrations is not well characterized, some
inference may be made from the overall hydrodynamics and chemical loading patterns in an
ecosystem. For example, estuaries and rivers generally display greater hydrologic fluctuations
than do large lake systems and, all else being equal, would generally be expected to have greater
variability in chemical concentrations in water and sediments, and thus potentially in biota.
6-4
-------
Chemical loadings that are highly variable over time and/or are spatially complex (e.g., multiple
point sources at different locations) will also contribute to greater variability in chemical
concentrations in an ecosystem compared with more constant and uniform loading situations
(e.g., continuous releases of similar magnitude over wide geographic areas, such as PCB releases
from sediments in Lake Ontario). For moderately to highly hydrophobic chemicals with low rates
of metabolism, less confidence (greater uncertainty) should usually be assigned to baseline BAFs
derived from studies with minimal or no spatial or temporal averaging than with those from
studies with greater averaging, unless overall variability in concentrations is very small.
Moderately to highly hydrophobic chemicals that lack persistence in the environment may
require less spatial or temporal averaging to the extent that their distributions are known to be
limited with respect to more persistent chemicals of similar hydrophobicity. Concentrations in
water and tissue of chemicals of low hydrophobicity (i.e., log Kow • 3) tend to parallel one another
temporally because of rapid uptake and elimination kinetics. Thus, less temporal and spatial
averaging of contemporaneous concentrations in water and tissue is required to produce reliable
BAFs for chemicals of low hydrophobicity.
Temporal and Spatial Connectivity of Samples
The connectivity of samples in space and time should be evaluated for assessing the
reliability of baseline BAFs for accurately representing steady-state bioaccumulation. Depending
on chemical and ecosystem properties, BAF^s that are derived from tissue samples that are
widely separated in space or time from water samples can be highly uncertain. Special attention
should be paid to situations in which geographic gradients in concentrations are known or
suspected, because geographic asynchrony in tissue and water samples can lead to erroneous or
biased estimates of a EAF{ under these circumstances. For example, if water samples are
collected from an area of high concentrations in an exposure gradient (e.g., near a discharge) and
fish are collected from a "down gradient" area (i.e., where exposure concentrations in water are
expected to be substantially lower), BAF^s can be underestimated because of overestimated
chemical concentrations in water relative to those in tissue. Even when water, sediment, and
biological samples are co-located in space, the mobility of organisms such as fish can be
problematic when strong gradients in chemical concentrations exist. For highly hydrophobic
chemicals whose metabolism is not important, modeling results suggest that fish tissue should be
sampled toward the end of the water sampling period in order to account for the time lag
associated with the slow accumulation kinetics of these compounds. For chemicals of low
hydrophobicity, rapid uptake and elimination kinetics indicates that tissue and water samples
should be closely connected temporally and spatially to produce reliable estimates of the EAF{
(e.g., sampling fish and water at the same time and location).
Chemical Loadings History and Steady State
As discussed in Section 4.3, the history of chemical loadings to an ecosystem has a direct
bearing on the extent of the disequilibrium between chemical concentrations in water and
sediment (• SOCw/Kow). This in turn can affect the magnitude of the BAFx for nonionic organic
chemicals (Burkhard et al., 2003b). Furthermore, rapid changes in chemical loadings to an
ecosystem can lead to biased estimates of long-term BAF^s when sufficient time is not allowed
6-5
-------
for tissue, water, and sediment concentrations to approach steady state, or when concentrations
are not averaged properly over space and time. A longer time is generally necessary for highly
hydrophobic chemicals in fish to reach steady state with respect to water than for chemicals with
low hydrophobicity. For ECF{ measurements, evaluation of steady state is particularly important
when the steady-state method is used, because highly hydrophobic chemicals may require
substantially longer than the 28-day test duration typical of many bioconcentration tests to reach
or approach steady state. Therefore, when variability and uncertainty in baseline BAFs are being
evaluated for a given species, special attention should be paid to differences in chemical loadings
history and the likelihood that BAF^s (or BCFxs) reflect steady-state conditions as possible
explanatory factors. For highly hydrophobic chemicals (e.g., log Kow of approximately 5 or
greater), baseline BAFs may be highly biased when derived from field measurements in
ecosystems where there has been a recent and substantial change in chemical loadings or
chemical concentrations. Values of BAFx that are known or suspected of being substantially
biased with respect to representing steady-state bioaccumulation conditions should not be used
in calculating species-mean baseline BAFs.
Food Web Structure
Another factor to evaluate when comparing baseline BAFs derived from BAF^s or BSAFs
is the difference in food web structure that may exist across sites (or even season) for the same
species. Although there is no one "right" food web structure from which to judge the
acceptability of a baseline BAF, differences in food web structure may help to explain some of
the variation observed in baseline BAFs for a given species across sites. Model predictions and
field observations indicate that food web structure can affect the magnitude of bioaccumulation,
particularly for highly hydrophobic organic chemicals. While the trophic position of a given
organism (and, by extension, the magnitude of its dietary exposure) can vary as a function of its
age, size, and reproductive status, variations in the availability of and competition for prey items
can directly influence dietary exposures. In some cases, individual organisms of the same species
that differ in size or age are classified into separate trophic levels because of size or age-related
differences in feeding preference and diet. Finally, for highly hydrophobic chemicals, for which
significant disequilibrium exists between water and sediment concentrations, models indicate that
a species with a benthic-driven diet tends to accumulate higher concentrations than does the
same species with a pelagic-driven diet (see Section 4.4 and Burkhard et al., 2003b).
Organic Carbon and Lipid Measurements
The reliability of organic carbon (DOC, POC) and lipid measurements is also important to
review when evaluating uncertainty in baseline BAFs. Both are used directly in deriving baseline
BAFs from field or laboratory data. Concentrations of DOC and POC in a body of water are
expected to vary overtime as a function of precipitation events, season, hydrodynamics, and
numerous other attributes of a watershed. Thus, sufficient sampling of DOC and POC
concentrations over space and time is needed to achieve representative estimates of baseline
BAFs, the extent of which will vary according to the variability in the particular ecosystem and
the hydrophobicity of the chemical in question. Samples for the analysis of DOC and POC
should be collected simultaneous with water samples collected for the analysis of the chemical of
6-6
-------
interest. For highly hydrophobic chemicals where greater temporal and spatial averaging of
chemical concentrations is needed to determine a EAF{ that is representative of long-term
conditions, DOC and POC concentrations will need to be similarly averaged when used to
calculate the corresponding baseline BAF for the study site. This is especially important for
highly hydrophobic chemicals because the impact of DOC and POC on calculation of the
baseline BAFs is greatest with these chemicals (see Section 6.3).
In estimating lipid fraction for use in deriving a baseline BAF, care should be taken to
review the differences in the extraction method used to measure the lipid content of a given
species across studies. As discussed in Section 4.1, differences in the polarity of solvents used to
extract lipids from tissue can result in the extraction of different amounts of lipid. This can lead to
variation in lipid-normalized concentrations and, consequently, in baseline BAFs because of the
solvent system used. Of particular concern are differences in the solvent extraction efficiencies of
lipid and chemicals in extremely lean tissues (e.g.,
-------
3. The weight of evidence suggested by BAFs determined by different BAF
derivation methods.
The data preference hierarchy for each BAF derivation procedure is summarized in Figure
3-1 and further detailed in Table 6-1. It is based on the relative strengths and limitations of each
BAF method and reflects the general preference of field-measured data over laboratory- or
model-based estimates of bioaccumulation. Importantly, this hierarchy is intended for use as a
guide for selecting the final baseline BAF rather than as a steadfast rule. Departures from this data
preference hierarchy are entirely appropriate when considerations of uncertainty and weight of
evidence indicate that a lower tier method would be preferred over a higher tier method.
In general, when trophic level-mean baseline BAFs are available for more than one BAF
method within a given trophic level, the final trophic level-mean baseline BAF should be selected
from the most preferred BAF method, as defined by the data preference hierarchy for the
applicable derivation procedure (Figure 3-1; Table 6-1). If uncertainty in a trophic level-mean
baseline BAF based on a higher tier (more preferred) method is judged to be substantially greater
than one from a lower tier method, and the weight of evidence from the various methods
suggests that a BAF value from a lower tier method is likely to be more accurate, then the final
baseline BAF for that trophic level should be selected from the lower tier method.
Table 6-1. Data Preference Hierarchy for Selecting Final Baseline BAFs for Nonionic
Organic Chemicals
BAF
Derivation
Procedure
1
2
3
Applicability
K,,w • 4, metabolism
negligible or
unknown
Kow • 4, metabolism
significant
Kow <4, metabolism
negligible or
unknown
1.
2.
3.
4.
1.
2.
3.
1.
2.
Data Preference Hierarchy
Baseline BAF from an acceptable BAF); (method 1)
Baseline BAF predicted from an acceptable BSAF (method 2)
Baseline BAF predicted from an acceptable BCF); and FCM (method 3)
Baseline BAF predicted from an acceptable K,,w and FCM (method 4)
Baseline BAF from an acceptable BAF ); (method 1)
Baseline BAF from an acceptable BSAF (method 2)
Baseline BAF from an acceptable BCF); (method 3)
Baseline BAF from an acceptable BAF);. (method 1) or BCF); (method 3)
Baseline BAF predicted from an acceptable Kow value (method 4).
4 K,,w <4, metabolism 1. Baseline BAF from an acceptable BAF); (method 1) or BCF); (method 3)
significant
When the weight of evidence among the various BAF methods is being considered,
greater confidence in a final baseline BAF is generally assumed when baseline BAFs are in
agreement across a greater number of methods within a given trophic level. However, lack of
agreement among baseline BAFs derived from different methods does not necessarily indicate
less confidence, if such disagreements can be adequately explained. For example, if the chemical
of concern is metabolized by aquatic organisms represented by a baseline BAF value, one would
6-8
-------
expect disagreement between a baseline BAF derived from a EAF{ (the highest priority data) and
a baseline BAF predicted from a Kow and model-derived FCM. In addition, consideration should
also be given to the quantity and diversity of bioaccumulation measurements that underlie the
calculation of a trophic level-mean baseline BAF. This is particularly relevant because national
BAFs are intended to reflect central tendency estimates of bioaccumulation expected across U.S.
surface waters. In some cases, the uncertainty associated with very limited BAF data from a
"more preferred" method may be offset by the greater quantity and diversity of data that are
available from an otherwise "less preferred" method for a given data preference hierarchy.
6.2 BASIS FOR THE NATIONAL DEFAULT LIPID FRACTION (f.) OF
COMMONLY CONSUMED FISH AND SHELLFISH
This section provides the technical basis of EPA's recommended national default values
of lipid fraction (f.) that are used to derive national BAFs for nonionic organic chemicals (0.019
for trophic level 2 organisms, 0.026 for trophic level 3 organisms, and 0.030 for trophic level 4
organisms). As indicated by Equation 6-1 and Figure 2-2, the lipid fraction of commonly
consumed aquatic species is needed to adjust a baseline BAF (which reflects partitioning to 100%
lipids) to a BAF that reflects the lipid fraction of aquatic organisms commonly eaten by U.S.
consumers. Information on lipid content is used to adjust BAFs for nonionic chemicals because it
has been shown to influence the magnitude of bioaccumulation in aquatic organisms (Mackay,
1982; Connolly andPederson, 1988; Thomann, 1989). Therefore, lipid content in consumed
aquatic organisms is considered to be an important factor for characterizing potential human
exposure to nonionic organic chemicals.
Although EPA uses national default values of lipid fraction to derive national human
health AWQC, EPA encourages States and authorized Tribes to use local or regional data on the
lipid content and consumption rates of consumed aquatic species when adopting criteria into
their own water quality standards. The use of such locally or regionally derived data is
encouraged over national-scale data because local or regional consumption patterns offish and
shellfish (and thus the amount of lipid consumed from aquatic organisms) can differ from
national consumption patterns. Additional guidance on developing local or region-specific values
of lipid fraction, including a database of lipid fraction for many commonly consumed aquatic
organisms, is found in a subsequent volume of this TSD (Volume 3: Development of Site-
Specific Bioaccumulation Factors). Nevertheless, EPA recognizes that there will be situations
when such local or regional data are not available or are inadequate for deriving representative
values of lipid fraction for setting State or Tribal water quality standards. In these cases, EPA
recommends the use of its national default values of lipid fraction for deriving BAFs and resulting
water quality criteria.
6.2.1 Variability in Lipid Content
One issue associated with setting national default values of lipid fraction in consumed
aquatic organisms is how to address intraspecies and interspecies variability in lipid content. For
example, the mean percent lipid in fillets of lake trout, Salvelinus namaycush, a notoriously
"fatty" species, is estimated to be about 12%. This value is about 18 times the mean percent lipid
6-9
-------
found in fillets of northern pike, Esox lucius (0.7%). Wide variation in lipid content can also
occur within a species. Based on data presented later in this section, the coefficient of variation of
percent lipid can approach or, in some cases, exceed 100% within a species, even when data are
limited to specific tissue types.
A number of factors can lead to variability in lipid content of aquatic organisms. Many of
these factors are fundamentally related to differences in physiology, metabolism, organism health
or condition, and feeding ecology among and within species. These factors and, consequently,
the lipid content in a particular tissue can vary as a function of season, temperature, reproductive
status, migratory patterns, sampling location (both within and across water bodies), age, size, life
stage, the availability of prey, and other factors. In addition, the distribution of lipids in a
particular aquatic organism is not uniform across all tissue types, thus resulting in differences in
lipid fraction depending on the tissue sampled (e.g., fillet, whole body, muscle). Finally,
differences among analytical methods used to extract and measure lipids and associated
analytical error can contribute to variability in reported values of lipid fraction (see Section 6.2.2).
For the purposes of deriving national default values of lipid fraction, EPA has addressed
the issue of variability in lipid content in several ways. First, only data for the most commonly
consumed aquatic species in the United States were considered. The identity of these species was
determined by the U.S. Department of Agriculture's Continuing Survey of Food Intake by
Individuals (CSFII) (USDA, 1998) for 1994 through 1996 (the most recent data at the time of this
analysis). Second, data were limited to the tissues that are most commonly consumed within a
species. Third, when size information was available, data were further limited to sizes of aquatic
species that are typically eaten by U.S. consumers. Finally, national default values of lipid
fraction were determined by weighting individual mean values of lipid fraction for each species
(or group of species) by the appropriate consumption rates determined for the U.S. population. In
this manner, EPA's national default values of lipid fraction better reflect national consumption
patterns of aquatic organisms in comparison to simply weighting the lipid fraction for each
species equally.
The following sections present the data sources, analysis, assumptions, and uncertainty
associated with the derivation of national default lipid values.
6.2.2 Data Sources
The national default values of lipid fraction (f.) were derived by using three types of
nationally aggregated data:
1. National per capita consumption rates of aquatic organisms
2. Lipid fraction in consumed aquatic organisms
3. Trophic status of consumed aquatic organisms
A summary and description of this information are provided below.
6-10
-------
National Mean Per Capita Fish Consumption Rates
Information on the types and quantity of aquatic organisms consumed in the United
States was obtained from the CSFII (USDA, 1998). This national survey provides daily mean per
capita estimates offish consumption for the U.S. population for categories of estuarine,
freshwater, and marine fish and shellfish (among other dietary categories). Although other
regional or local surveys were available, the CSFII was selected because it provided consumption
information on a national basis and contained the most recent data available. Furthermore,
information from the same USDA survey was used to derive national default values offish
consumption rates for calculating national human health AWQCs. Reliance on the same fish
consumption survey ensures consistency between the derivation of national default values for
lipid fraction and national default values offish consumption rates.
Table 6-2 shows the habitat classification, CSFII consumption categories, estimated mean
per capita consumption rates, and fraction of total estuarine and freshwater consumption
represented by each category in the CSFII. Mean per capita consumption rates were estimated for
individuals aged 18 years and older for categories of freshwater and estuarine aquatic organisms.
This same category of consumers was used to calculate the national default value offish
consumption for the general adult population and sport anglers (i.e., a 90th percentile total intake
rate of 17.5 g per person per day, as described in the 2000 Human Health Methodology).
However, sufficient data were not available in the CSFII survey to adequately describe the pattern
offish consumption (as opposed to the total fish and shellfish intake rate) for individuals
constituting the 90th percentile. Therefore, the consumption pattern represented by the mean per
capita consumption rates was chosen for this analysis. In calculating the estimated mean
consumption rates, two CSFII consumption categories were classified as being unknown by the
survey (i.e., "unknown fish" and "unknown seafood"). By using the same approach adopted for
deriving the national default value for fish consumption, 39% of these two "unknown" fish
consumption categories were apportioned to the freshwater and estuarine consumption
categories on a consumption rate-weighted basis. Further details of the mean per capita
consumption rates from the CSFII survey data and assignment of habitat designations for fish
and shellfish can be found in the Exposure Assessment volume of this Technical Support
Document.
It is apparent from Table 6-2 that estimated mean per capita consumption rates vary
widely across categories and that consumption in a relatively few categories dominates the overall
consumption pattern of freshwater and estuarine organisms. For example, the consumption rate
in the "shrimp category" constitutes about 35% of the total estimated mean per capita
consumption rate of freshwater and estuarine organisms. Similarly, the consumption rates
corresponding to the flounder, catfish (fresh and estuarine), and flatfish categories constitute
approximately 30% of the total consumption rate of freshwater and estuarine organisms. To
account for the disproportional nature of the national per capita consumption rates revealed by
the CSFII survey, the derivation of national default values of lipid fraction were determined on a
consumption-weighted basis, as described later in this section.
6-11
-------
Table 6-2. Categories and Mean Per Capita Consumption Rates from the USDA CSFII
Habitat
USDA CSFII Consumption
Category
Estimated Mean Consumption
Rate (g/person/day)
Proportion of Total Freshwater and
Estuarine Consumption Rate (%)
Estuarine
Freshwater
Total
Source: USDA
Notes: (l)Esti
Shrimp
Flounder
Catfish (estuarine)
Flatfish (estuarine)
Crab (estuarine)
Perch (estuarine)
Croaker
Oyster
Herring
Trout, mixed spp. (estuarine)
Salmon (estuarine)
Anchovy
Rockfish
Mullet
Clam (estuarine)
Smelts (estuarine)
Eel
Scallop (estuarine)
Smelts, rainbow (estuarine)
Sturgeon (estuarine)
Catfish (freshwater)
Trout (rainbow)
Perch (freshwater)
Carp
Trout, mixed spp. (freshwater)
Pike
Whitefish (freshwater)
Crayfish
Snails (freshwater)
Cisco
Salmon (freshwater)
Smelts, rainbow (freshwater)
Sturgeon (freshwater)
combined 1994-1996 Continuing Survey
mates were based on 2-day averages and i
2.65492
0.73482
0.60335
0.45173
0.42111
0.22331
0.17792
0.17485
0.16428
0.15305
0.05915
0.05815
0.05428
0.04512
0.01732
0.00880
0.00466
0.00140
0.00076
0.00018
0.60335
0.25361
0.22331
0.19071
0.15305
0.04021
0.01309
0.01028
0.00207
0.00179
0.00118
0.00076
0.00018
7.50273
of Food Intakes by Individuals (CSFII) (USDA,
are projected from a sample of 9,596 individuals
35.4
9.8
8.0
6.0
5.6
3.0
2.4
2.3
2.2
2.0
0.8
0.8
0.7
0.6
0.2
0.1
0.06
0.02
0.01
0.002
8.0
3.4
3.0
2.5
2.0
0.5
0.2
0.1
0.03
0.02
0.02
0.01
0.002
100.0
1998).
18 years of age and
older in the U.S. population of 190,931,846 individuals 18 years of age or older, using 3-year combined survey weights.
Weights are for uncooked fish and shellfish. (2) The fish component of foods containing fish was calculated by using
data from the recipe file of USDA's Nutrient Database for individual food intake surveys. (3) Values reflect
apportionment of 39% of the consumption rate of "unknown fish" and "unknown seafood" categories to freshwater and
estuarine categories. (4) The number of digits does not imply their statistical significance. See Section 6.2.2 and the
Exposure Assessment volume of this Technical Support Document for additional information about the CSFII and
habitat classification.
6-12
-------
Lipid Content of Consumed Aquatic Species
In reviewing the available literature, EPA was unable to find a single comprehensive
database containing information on the lipid content of consumed aquatic organisms (fresh and
estuarine). As a result, information on the lipid fraction of aquatic organisms was obtained from a
variety of primary and secondary sources. The following major sources of lipid data were used in
the derivation of national default values of lipid fraction:
EPA's National Sediment Quality Survey database (USEPA, 2001 a)
EPA's National Study of Chemical Residues in Fish (USEPA, 1992a)
EPA's Green Bay Mass Balance Study (USEPA, 1992b, 1995c),
U. S. Department of Agriculture's (USD A) Nutrient Data Bank (Exler, 1987)
• A review from National Marine Fisheries Service of the National Oceanic and
Atmospheric Administration (NOAA) (Sidwell, 1981)
Two California databases (California Toxic Substances Monitoring Program and
Bay Protection and Toxic Cleanup Program)
When insufficient data were available from the above sources for certain species, targeted
literature searches were conducted and data from primary literature were used. Importantly, care
was taken to avoid duplication of data among the various data sources. For example, examination
of the National Sediment Quality Survey lipid data and data from various studies conducted on
the Hudson River revealed significant overlap and apparent duplication. Similarly, multiple
literature reviews (e.g., USDA and NOAA reports) were not used for a given species unless it was
clear that the primary literature upon which each was based was unique. Each of these data
sources is discussed in more detail below.
National Sediment Quality Survey. Data on lipid content were extracted from a
prerelease version of EPA's National Sediment Quality Survey (NSQS) database from 1980
through 1998, the last year for which data were made available in this version of the database. A
description of the NSQS database can be found in USEPA (200 la). The primary source of data
contained in the NSQS was EPA's STORET (Storage and Retrieval of U.S. Waterways
Parametric Data) database, recently renamed the Legacy Data Center database. STORET is a
waterway-related monitoring database that contains data from many Federal and State
government agencies. Geographically, the lipid data extracted from the NSQS database were
mostly limited to organisms in freshwater habitats. More than 47,000 records of lipid content
were extracted, representing more than 200 species and taxonomic groupings. Despite a large
number of species represented in the database, the quantity of lipid data was not evenly
distributed among species. For example, 10 species (mostly catfish, bass, perch, and salmonids)
represented 60% of the total number of records. Data on lipid content (e.g., percent lipid, tissue
type), organism attributes (e.g., common name, scientific name, and, where available, age,
weight, length, and sex), and sampling station (e.g., latitude, longitude, sampling date,
investigator names) were extracted and combined in an MS ACCESS® database. Information on
the method of lipid analysis was not reported in the NSQS database.
6-13
-------
National Study of Chemical Residues in Fish (NSCRF). Data on the lipid content of
aquatic organisms were extracted from EPA's National Study of Chemical Residues in Fish
(USEPA). This study represents a one-time screening investigation to determine the prevalence of
selected bioaccumulative pollutants in fish. Samples were collected from 388 locations in the
United States that included targeted sites near point and non-point pollution sources, background
sites with minimal expected pollution sources, and a few sites corresponding to the U.S.
Geological Survey's National Stream Quality Accounting Network (NASQAN) to obtain
nationwide coverage. Species represented in the study generally included a bottom-feeder and a
game-fish species. Carp was the most common species sampled, followed by largemouth bass,
white sucker, channel catfish, and smallmouth bass. Three to five fish collected from one location
were used for each composite sample. For each composite sample, two measurements of the
lipid content were obtained, one from the test for dioxins/furans and one from the test for other
xenobiotics. The average of the two lipid values was used to represent each sample data point in
the lipid database. Location and sampling date information was available, as were the common
name of the species collected and the tissue type sampled (whole body, fillet). Lipids were
measured gravimetrically after extraction with hexane and methylene chloride (1:1).
U.S. Department of Agriculture (USDA). Information on the composition of finfish and
shellfish products (including lipid content) was summarized by Exler (1987), using data from
USDA's Nutrient Data Bank. These data were particularly useful because they included estuarine
species that were poorly represented by the other databases. Sources of the data summarized by
Exler (1987) included journal articles, technical reports, and other scientific and technical
literature (published and unpublished). Raw data on lipid content were not reported by Exler;
only summary statistics (mean, standard error, and number of samples) were reported. Because it
was not possible to combine the summary data from Exler (1987) with individual records from
other sources and still maintain statistical reliability, the Exler (1987) data were used on an
exclusive basis for a given species. Tissue types were restricted to the "edible portion" of species.
Although data were available on various lipid fractions (e.g., saturated, monosaturated, and
polyunsaturated fatty acids; cholesterol), only total lipid data were used in raw (unprocessed)
samples. Information on the method of lipid analysis used was not reported.
National Marine Fisheries Service. Data on the lipid content of estuarine species from
the National Marine Fisheries Service of NOAA were available from a review by Sidwell (1981).
This review consists of compilations of data from primary literature sources. Information on the
specific location and number of individuals per value was not available in these reviews. Data
were restricted to species collected in North America, when information was available to make
this distinction. Information was available on species' common and Latin names, tissue type, and
method of preparation (e.g., raw, cooked). Only samples that were indicated as being fresh or raw
(or for which no preparation information was available) were used in the analysis of lipid data.
Other information, such as the number of individuals in a sample and their age, weight, and sex,
was not available. Importantly, the Sidwell (1981) data were not used in combination with the
USDA data for a given species, because both represent reviews of the primary literature and may
contain redundant data. A later review was also available from the National Marine
Fisheries Service (Kryznowek and Murphy, 1987); however, this was not used because its data
were presented in an aggregated format.
6-14
-------
California Toxic Substances Monitoring Program. Lipid data were obtained from the
Toxic Substances Monitoring Program (TSMP) database sponsored by the California
Environmental Protection Agency, California State Water Resources Control Board, via
downloading from their Internet site (www.swrcb.ca.govl The TSMP was initiated in 1976 to
provide a uniform statewide approach to monitor for toxic substances in freshwater and, to a
limited extent, in estuarine and marine waters, through the analysis offish tissue and other
aquatic life. Samples are collected annually and composite samples of six fish are collected when
possible. The database provides information on age, sample collection date, location, number of
organisms per sample, weight, and length. Most samples for the species of interest are fillet
samples, although some whole-organism samples are also present. Lipids were measured
gravimetrically with petroleum ether solvent. A complete description of the TSMP database is
found in Rasmussen (1998).
Bay Protection and Toxic Cleanup Program. Relevant information on lipid content was
extracted from two small databases available from the Bay Protection and Toxic Cleanup
Program (BPTCP) (also at www.swrcb.ca.govY One database contained samples from San
Francisco Bay, and the second included samples from San Francisco Bay and other California
locations. Lipids were measured gravimetrically with methylene chloride as the extraction
solvent.
Green Bay Mass Balance Study. The 1989-1990 Green Bay Mass Balance Study
(GBMB) was conducted to generate a comprehensive data set for modeling the fate, transport,
and bioaccumulation of several toxic chemicals, including polychlorinated biphenyls, in Green
Bay, Lake Michigan. Six species were represented in the database for lipid content (walleye,
brown trout, carp, alewife, rainbow smelt, and gizzard shad). All data are from composite
samples of whole-body tissue. Included in the database is information on collection date, zone of
bay in which the sample was collected, age, and number offish in the composite. Data were
retrieved from the Green Bay Relational Database, which is maintained by EPA's Office of
Research and Development. Lipids were measured gravimetrically after extraction with
hexane/methylene chloride (1:1).
6.2.3 Data Analysis
The following steps were taken in the calculation of national default values of lipid
fraction:
Removing suspect data (e.g., duplicate records, extreme values)
Classifying species into CSFII consumption categories
Excluding data for tissue types and size ranges not typically consumed
Calculating mean lipid fraction for CSFII species
Assigning trophic levels to species and CSFII consumption categories
Calculating consumption-weighted values of lipid fraction within each trophic
level
6-15
-------
Each of these steps is further described below.
Removing Suspect Data
Data from the NSQS database were screened to ensure that all records in the database
included, at a minimum, fields for common name, scientific name, species code, lipid content,
tissue type, sample data, and sample location. Most NSQS records contained additional
information beyond this subset. Records for common name, scientific name, and species code
were cross-checked for consistency, and erroneous entries were corrected or removed if they
were ambiguous. Furthermore, data were removed from all database sources that were collected
before 1980. Extreme values of lipid content (e.g., zero values and those above 35% for finfish
other than lake trout) were removed to minimize the impact of suspected outliers on the analysis.
For lake trout, values of up to 45% were tolerated because fillets of this species are notoriously
high in fat content, which on rare occasions can approach this value. The trimming of the high
extreme values resulted in the removal of very few records (i.e., fewer than 30 of more than
47,000). Finally, records from multiple sources were compared to identify and remove duplicate
records.
Classifying Species into CSFII Consumption Categories
The next step in calculating the national default values of lipid fraction involved assigning
species to the CSFII consumption categories shown in Table 6-2. This step was conducted to
maintain consistency in the data used to determine national default consumption rates offish and
shellfish and lipid fraction. In most cases, information was not available from the CSFII to
identify exactly which species were represented by the consumption rates listed in Table 6-2.
Therefore, assignment of a species to a CSFII category was based on several factors, including (1)
its taxonomic and publicly perceived linkage to a CSFII category, (2) its likelihood of being
caught (either recreationally or commercially) and consumed in the United States, and (3) its
likelihood of inhabiting either fresh or estuarine waters for at least some portion of its life cycle.
Information from numerous published sources was used to help determine whether a species met
these criteria. Because several of the CSFII species categories were broad in terms of the types of
species that could be included, some species were assigned to multiple CSFII categories. For
example, flounder species fit into both estuarine flatfish and flounder categories. In such cases,
appropriate records were included in both CSFII categories. Data for species that could not be
unambiguously assigned to a CSFII consumption category were omitted from the analysis.
Notably, this resulted in the exclusion of lipid data for some species that are commonly
consumed in the United States but were not associated with a CSFII category (e.g., largemouth
bass and walleye).
Screening by Tissue Type and Size Ranges
As discussed previously, lipid content can vary widely by the type of tissue in which it is
measured. To derive national default values of lipid fraction and national BAFs that are
representative of human exposure potential, lipid data were screened according to the types of
tissues most commonly consumed by the U.S. population. Because lipid data originated from a
6-16
-------
variety of sources that differed in nomenclature used to classify the type of tissue, a variety of
commonly consumed tissues were accepted, depending on the species and the availability of
data. The following types of tissues reported by various data sources were included in the
derivation of the national default values of lipid fraction:
• Finfish (except anchovy and smelt): standard fillet, fillet (with skin, without skin,
skin unspecified), edible portion
• Anchovy, smelt: whole body
Clam: whole/raw, adductor muscle
Crab: edible portion, muscle, standard fillet, muscle and hepatopancreas
Crayfish, oyster, scallop, shrimp: edible portion
• Snails: whole/raw
Lipid Content of Species in CSFII Categories
On the basis of lipid data from the aforementioned sources, the average percent lipid was
calculated for each of the species in the CSFII consumption categories (Table 6-3). Next, the
average percent lipid of all species in each CSFII category was determined as the average of the
corresponding individual species-mean lipid values. Average values of lipid content were
determined because national B AFs are designed to reflect central-tendency estimates. Ideally, if
sufficient national consumption data were available at the species level, the overall average lipid
value for each CSFII category would be determined on a consumption-weighted basis. However,
sufficient consumption data were not available below the CSFII category level at a national scale.
Therefore, equal weights were assigned to each species' mean lipid value. For example, lipid data
were available for several species of trout (e.g., rainbow trout, brown trout, and others), whereas
consumption rates were available from the CSFII only for trout as a group. Thus, mean percent
lipid values for all trout species were averaged and combined with the consumption rate for trout
from the CSFII.
Trophic Level Assignments to Species and CSFII Consumption Categories
National fish and shellfish consumption data from the CSFII (see Table 6-2) indicate that,
on average, individuals consume aquatic organisms from a variety of trophic levels (e.g., oysters
and clams in trophic level 2, whitefish and herring in trophic level 3, perch and trout in trophic
level 4). Because trophic position (in particular, dietary composition) can affect the extent of
bioaccumulation in aquatic organisms, national BAFs are derived separately for each trophic
level. Similarly, because lipid content can vary by species, national default values of lipid fraction
are derived separately for each trophic level. To estimate trophic level-specific values of lipid
fraction, a trophic level designation must be assigned to each of the CSFII consumption rate
categories shown in Table 6-2. This same trophic level assignment is used to discern the fraction
of the national default fish consumption rate (17.5 g/d) that occurs at each trophic level (see the
2000 Human Health Methodology).
6-17
-------
Table 6-3. Lipid Content of Aquatic Organisms Used to Derive National Default Values of Lipid Fraction (f.)
CSFH Consumption
Category (Habitat)"
Anchovy (estuarine)
Carp (freshwater)
Catfish (freshwater)
Catfish (estuarine)
Cisco (freshwater)
Clam (estuarine)
Crab (estuarine)
Crayfish (freshwater)
Croaker (estuarine)
Eel (estuarine)
Flatfish (estuarine)
Flounder (estuarine)
Herring (estuarine)
Mullet (freshwater)
Oyster (estuarine)
Common Name
Striped anchovy
European anchovy
Northern anchovy
Common carp
White catfish
Black bullhead
Yellow bullhead
Brown bullhead
Channel catfish
White catfish
Brown bullhead
Channel catfish
Cisco
Hard shell clam
Soft shell clam
Venus clam (Littleneck
Japanese)
Venus clam (Shortneck)
Venus clam (Asari)
Venus clam
Venus clam (hard)
Blue crab
Dungeness crab
Queen crab
Crayfish (mixed sp.)
White croaker
Atlantic croaker
Yellowfin croaker
Eel, mixed species
Sole and flounder
Sole and flounder
Blueback herring
Atlantic herring
Pacific herring
Striped mullet
Pacific oyster
Scientific Name
Anchoa hepsetus
Engraulis encrasicholus
Engraulis mordax
Cyprinus carpio
Ameiurus catus
Ameiurus melas
Ameiurus natalis
Ameiurus nebulosus
Ictalurus punctatus
Ameiurus catus
Ameiurus nebulosus
Ictalurus punctatus
Coregonus Artedii
Mercenaria mercenaria
Mya arenaria
Tapes (venerupis) decussatus
Tapes japon ica
Tapes philippinarum
Venus gallina
Venus lusoria
Callinectes sapidus
Cancer magister
Chionoectes opilio
Astacus and Orconectes
Genyonemus lineatus
Micropogonias undulatus
Umbrina roncador
Anguilla spp.
Bothidae and Pleuronectidae
Bothidae and Pleuronectidae
Alosa aestivalis
Clupea harengus
Clupea pallasi
Mugil cephalus
Crassostrea gigas
Species Mean
Lipid Content (%)
2.8
4.8
10.7
5.4
4.3
1.1
1.4
2.6
5.3
4.3
2.6
5.3
1.9
0.7
1.2
1.2
1.8
2.6
0.9
0.6
1.3
1.0
1.2
1.1
4.2
3.2
1.8
11.7
1.2
1.2
7.2
9.0
13.9
3.8
2.3
CV
NR
0.34
NR
0.86
0.58
0.70
0.99
0.72
0.71
0.58
0.72
0.71
0.65
NR
NR
NR
NR
NR
NR
1.19
0.26
0.30
NR
0.88
0.47
0.70
0.28
0.80
0.80
0.45
0.51
0.39
0.62
0.33
No. Obs.
23
26
16
2,792
204
113
95
988
1,427
204
988
1,427
69
47
3
15
3
3
29
5
101
24
6
5
37
8
3
14
596
596
92
2,524
128
43
13
Data
Source0
1
2
1
3
3,4,5
3,4,5
3,5
3,4,5
3,4,5
3,4,5
3,4,5
3,4,5
2
1,6
1
1
1
1
1
1
3
2
2
2
4,5,6,7
2
5
2
2
2
3
2
2
2
2
CSFII Mean
Lipid (%)
6.1
5.4
2.9
4.0
1.9
1.3
1.1
1.1
3.0
11.7
1.2
1.2
10.0
3.8
2.4
-------
CSFH Consumption
Category (Habitat)"
Perch (estuarine and
freshwater)
Pike (freshwater)
Rockfish (estuarine)
Salmon
(estuarine & freshwater)
Scallop (estuarine)
Shrimp (estuarine)
Smelt, rainbow (estuarine
& freshwater)
Snails (freshwater)
Sturgeon
(estuarine & freshwater)
Trout, mixed spp.
(freshwater)
Trout, mixed spp.
(estuarine)
Trout, (freshwater)"
Whitefish
Common Name
Eastern oyster
White perch
Yellow perch
Northern pike
Muskellunge
Chain pickerel
Striped bass
Rockfish
Pink salmon
Chum salmon
Coho salmon
Sockeye salmon
Chinook salmon
Atlantic salmon
Scallop, mixed species
Shrimp, mixed species
Rainbow smelt
Snails, mixed species
Lake sturgeon
White sturgeon
Rainbow trout
Cutthroat trout
Brown trout
Brook trout
Lake trout
Rainbow trout
Cutthroat trout
Rainbow trout
Whitefish, mixed SDD.
Scientific Name
Crassostrea virginica
Morone americana
Percaflavescens
Esox lucius
Esox Masquinongy
Esox niger
Morone saxatilis
Sebastes spp.
Oncorhynchus gorbuscha
Oncorhynchus keta
Oncorhynchus kistuch
Oncorhynchus nerka
Oncorhynchus tshawytscha
Salmo salar
Pectinidae
Panaeidae and Pandalidae
Osmerus mordax mordax
Vivaparadidae, Helixidaed
Acipenser fulvescens
Acipenser transmontanus
Oncorhynchus mykiss
Salmo clarki clarki
Salmo trutta
Salvelinus fontinalis
Salvelinus namaycush
Oncorhynchus mykiss
Salmo clarki clarki
Oncorhynchus mykiss
Corezonus spp.
Species Mean
Lipid Content (%)
2.5
3.5
1.0
0.6
1.1
0.4
5.3
1.6
3.5
3.8
2.9
8.6
3.4
6.3
0.8
1.7
4.1
1.4
9.4
1.3
5.1
1.2
7.4
4.0
12.3
5.1
1.2
5.1
5.9
CV
0.56
0.72
0.79
1.01
0.87
0.74
0.59
NR
0.49
0.62
0.75
0.32
0.93
0.74
0.35
0.39
0.46
0.75
0.63
0.67
0.67
0.79
0.73
0.56
0.62
0.67
0.79
0.67
0.64
No. Obs.
193
682
841
904
35
72
7,657
81
144
13
617
48
873
7
114
100
130
11
51
7
556
15
615
96
910
556
15
556
68
Data
Source0
2
3,4
3,5
3,4
3
3,4
3,4,5,7
2
2
2
3
2
3
2
2
2
3,8
1
3
4,5,7
3,4,5
3
3,4,5
3,4,5
3,4,5
3,4,5
3
3,4,5
2
CSFII Mean
Lipid (%)
2.3
0.7
3.5
4.7
0.8
1.7
4.1
1.4
5.4
6.0
3.2
5.1
5.9
1 Habitat designation (freshwater, estuarine) assigned to the CSFII consumption categories. See the Exposure Assessment volume of this Technical Support Document for details.
b Coefficient of variation.
c Data sources: 1 = Sidwell (1981), 2 = Exler (1987), 3 = NSI (USEPA, 200la), 4 = USEPA (1992a), 5 = CATSMP, 6 = primary literature, 7 = BPTCP, 8 = GBMB. See Section
6.2.2 for a description of data sources.
d In addition to these two families, specific genera represented include Ampullaria, Vivaparus, Achatina, Murex, Thais, Nassa, and Aporrhais.
e Information from the CSFII survey indicates that rainbow trout is appropriate for the "trout, freshwater" category.
-------
The EPA recognizes that the dietary composition of an aquatic species (and, henceforth,
its trophic position) can vary as a function of size, age, life history, season, and food web
structure of the water body. As in the discussion presented at the beginning of Section 6.2 for
deriving values of lipid fraction in general, States and Tribes are encouraged to use local or
regional information to estimate the trophic position of consumed aquatic organisms for setting
local or regional criteria. The use of such local or regional information is encouraged because
factors that can affect trophic position often vary on a local or regional basis.
In the case of deriving national AWQC, and in situations where sufficient local or regional
data are not available, an assessment of the trophic position of consumed aquatic organisms is
necessary. Estimating the trophic level position of aquatic species requires information on the
identity, size, age, and diets of the individual aquatic species consumed. As previously discussed,
very limited data were available to further delineate the identity and size of species consumed
within each of the CSFII categories in Table 6-2. For most of the CSFII categories, this lack of
information was not viewed as problematic, because rather unambiguous assignments of trophic
position could be made to these categories (e.g., all oysters are considered to be trophic level 2).
However, for other CSFII categories, assignment of trophic position required some assumptions
to be made, which introduces greater uncertainty. To assist in estimating the trophic position of
species represented by the CSFII consumption survey, EPA relied on information summarized in
a report entitled Trophic Level and Exposure Analysis for Selected Piscivorous Birds and
Mammals (USEPA, 2000e-g). Although focused on piscivorous birds and mammals, this report
contains information on dietary composition and trophic status for numerous species in the
aquatic food web by virtue of the fact that the aquatic food web serves as the dietary basis for
piscivorous wildlife. The following procedures were used in assigning trophic position to the
CSFII consumption categories.
1. Species Trophic Level Assignments. Species trophic level assignments were performed
as follows:
a. For game fish that correspond to the CSFII categories, data were used for edible
size ranges (about 20 cm [8 inches] or larger).
b. For species where multiple size ranges were available, preference was given to the
larger specimens in determining the species trophic level.
c. Trophic level 2 was assigned to a species if appropriate trophic level data ranged
between 1.6 and 2.4; trophic level 3 if trophic level data ranged from 2.5 to 3.4;
and trophic level 4 if trophic level data were 3.5 or higher. This is consistent with
the approach taken in the Great Lakes Water Quality Initiative guidance (USEPA,
1995b).
2. CSFII Consumption Category Trophic Level Assignments. Once trophic levels were
assigned to each species that could reasonably correspond to a CSFII consumption
category, this information was used to assign a trophic level to each CSFII consumption
category, as follows:
6-20
-------
a. In situations where a CSFII category was represented by the vast majority of
species within a single trophic level, that trophic level was assigned to the CSFII
category (e.g., trout, pike, smelt).
b. For some CSFII consumption categories, the trophic status of representative
species spanned two trophic levels. In this case, consumption rate for that
category was evenly divided between the two trophic levels (e.g., for flounder,
50% to trophic level 3 and 50% to trophic level 4). This situation occurred for
opportunistic species such as flounder and flatfish, catfish, croaker (all evenly
divided between trophic levels 3 and 4), and shrimp (divided between trophic
levels 2 and 3).
3. The results of the trophic level assignments are shown in Table 6-4.
Calculation of Consumption-Weighted Lipid Content, by Trophic Level. The national
consumption-weighted mean lipid fraction for each trophic level (i.e., the national default values
of lipid fraction) was calculated according to the following equation.
CRi .,
(Equation 6-2)
where:
f. = national consumption-weighted mean lipid fraction of consumed aquatic
organisms at a given trophic level
CR; = mean per capita consumption rate of species in CSFII consumption category
"i" at the same trophic level
CR,0, = mean per capita consumption rate of species in all CSFII consumption
categories at the same trophic level
f i = average lipid fraction of species in CSFII consumption category "i"
Using Equation 6-2, EPA's national default values of lipid fraction were calculated for
trophic levels 2, 3, and 4 in Table 6-5.
These values were calculated with consumption rate data (CR; and CR,ot) that originated
from the USDA CSFII in Table 6-2, average values of lipid fraction for aquatic species described
in Table 6-3, and trophic level designations of each CSFII consumption category described in
Table 6-4. The calculation of the national default values of lipid fraction is illustrated in
Table 6-6.
6-21
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories
CSFII
Consumption Common Scientific
Category Name Name
Anchovy Bay anchovy Anchoa mitchilli
Northern anchovy Engraulis mordax
Carp Common carp Cyprinus carpio
Trophic Trophic
Level (a) Level (a)
Size (Mean) (Range)
Adult 2.8 —
— 2.3 2.1-2.5
— — 2.2-3.1
10-23 cm 3 2.8-3.1
> 23 cm 2.4 2.2-2.6
Notes (a)
9% Microinvertebrates, 58%
zooplankton, 33% organic detritus.
Feeds primarily on phytoplankton and
some zooplankton.
Young feed on zooplankton, small carp
feed on benthic invertebrates and
detritus, larger carp become more
herbivorous.
Up to 23 cm, feed primarily on benthic
invertebrates.
Larger carp (approx >23 cm) feed
primarily on plants and detritus
(60%-70%), benthic invertebrates
(15-35%), and some zooplankton
Species
Assigned
Trophic
Level (b)
3
2
3
CSFII
Assigned
Trophic
Level (c)
2(50%)
3(50%)
3
Catfish
Black bullhead Ameiurus me las — 3 2.9-3.2
Blue catfish Ictalurus furcatus — 3 —
Brown bullhead Ameiurus nebulosus — — 2.7-3.3
>10 cm 3.0 2.7-3.2
Channel catfish Ictaluruspunctatus 36-54 cm — 2.8^1
5-30 cm 3.1
30-35 cm 3.3 3-3.5
Seem to consume zooplankton and
benthic invertebrates throughout life.
Individuals >15 cm may consume some
small fish, but also plant materials.
Assumption.
Diet changes with size.
Those >10 cm feed on 20%-30%
plants and 70%-100% benthic
invertebrates (burrowing mayfly, scud,
chironomid types). Some consume
small fish as well.
Changes with age; can grow up to • 50
cm. Three studies indicate it consumes
plants; one other did not.
5-30 cm; consumes largely benthic
invertebrates (60%-80%), detritus
(10%-15%), and zooplankton
30-35 cm; consumes fish (32%),
benthic invertebrates (40%),
zooplankton (12%), and detritus (15%).
Some populations consume up to 25%
3 (50%)
4 (50%)
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories (continued)
CSFII
Consumption
Category
Catfish
Cisco
Clam
Crab
Common
Name
Channel catfish
Flat bullhead
Flathead catfish
Yellow bullhead
Cisco
Clam (general)
Hardshell clam
Softshell clam
Manila clam
Geoduck
Amethyst gemclam
Atlantic rangia
Baltic macoma
Dwarf surf clam
Blue crab
Rock crab
Dungeness crab
Stone crab
Scientific
Name Size
Ictalurus punctatus 35-45 cm
>45 cm
Ictalurus platycephalus —
Pylodictis olivaris —
Ictalurus natalis 30^16 cm
Coregonus artedii 20-30 cm
— —
Mercenaria spp. —
Mya arenaria —
Venerupis japonica —
Panopea abrupta —
— —
Rangia cuneata —
Macoma balthica —
Mulina lateralis —
Callinectes sapidus —
Cancer spp. —
Cancer magister —
Menippe spp. —
Trophic
Level (a)
(Mean)
3.8
4
3.2
3.8
2.6
3.0
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
3.2
3.3
3.5-3.7
3.4
Trophic
Level (a)
(Range)
3.5-3.9
4.0^.2
3-4
—
—
3.0-3.1
2.1-2.4
2.1-2.3
2.1-2.3
2.1-2.3
2.1-2.3
2.1-2.3
2.1-2.3
2.1-2.3
2.1-2.3
3.0-3.4
3.1-3.5
3.3^.1
3.2-3.7
Notes (a)
35^15 cm; consumes fish (67%),
benthic invertebrates (25%), and
detritus (8%). Some populations
consume up to 25% algae.
>45 cm; consumes fish (100%).
Can grow to large sizes; feeds on
mollusks (primarily clams), bryozoans,
and worms.
Diet consists primarily offish with
some crayfish, and mollusks.
Scavenger; often consumes minnows,
crayfish, insect larvae, worms, and
algae.
Primarily a plankton feeder.
Occasionally eats eggs, small fish.
Filter feeder on plankton, detritus;
includes zooplankton.
Selective filter feeder, consumes
primarily planktonic microalgae.
Nonselective filter feeder.
Nonselective filter feeder.
Filter feeder.
Filter feeder.
Filter feeder.
Filter feeder.
Filter feeder.
Feeds primarily on bivalves, organic
debris fish, crustaceans, plants, and
worms.
Feeds on herbivorous snails,
amphipods, shrimp, polychaetes, and
sea urchins.
Carnivorous; shrimp are preferred prey.
Carnivorous, feeding primarily on
mollusks.
Species
Assigned
Trophic
Level (b)
4
4
3
3
2
2
2
2
2
2
2
2
2
3
3
4
3
CSFII
Assigned
Trophic
Level (c)
3
2
3
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories (continued)
CSFII
Consumption
Category
Crayfish
Croaker
Eel
Flatfish and
flounder
Herring
Mullet
Oyster
Common
Name
Crayfish
White (Pacific)
croaker
Atlantic croaker
American eel
Gulf flounder
California halibut
Summer flounder
Southern flounder
Starry flounder
Winter flounder
Atlantic herring
Pacific herring
Mullet
American oyster
Pacific oyster
Scientific
Name Size
Astacidae —
Genyonemus lineatus —
Micropogonias —
undulatus
Anguilla rostrata —
Paralichthys albigutta —
Paralichthys >56 cm
californicus
Paralichthys dentatus —
Paralichthys —
lethostigma
Platichthys stellatus —
Pseudopleuronectes —
americanus
Clupea harengus —
Clupea pallasi —
Mugil spp. —
Crassostrea virginica —
Crassostrea gigas —
Trophic
Level (a)
(Mean)
2.4
3.4
3.7-4.0
3.9
3.5-3.9
3.5
3.5-3.9
3.5-3.9
3.2-3.6
3.0-3.6
3.2
3.2
2.1
2.2
2.2
Trophic
Level (a)
(Range)
2.0-2.7
3.2-3.7
3.3^.7
3.7^.3
3.3^.1
3.3^.0
3.3^.1
3.3^.1
2.7-3.8
2.7-3.8
3.1-3.3
3.1-3.3
2.0-2.3
2.1-2.3
2.1-2.3
Species
Assigned
Trophic
Notes (a) Level (b)
Primarily herbivorous, omnivorous; 2
animal food minor part of diet if
vegetation is available.
Opportunistic bottom feeder on small 3
fish, squid, shrimp, polychaetes, crabs,
and clams.
Opportunistic bottom feeder; feeds 4
mostly on polychaetes, copepods,
mysids, and small clams.
Feeds on small fish, young alewives, 4
salmon and trout fry, shrimp, and
crabs.
Paralichthys genus are primarily 4
piscivorous as adults but also eat
polychaetes, Crustacea, echinoderms,
and mollusks.
Feeds on anchovies and other small
fish.
See Gulf flounder. 4
See Gulf flounder. 4
Feeds primarily on small Crustacea, 3
polychaetes, bivalves, and
echinoderms. Few fish.
Feeds primarily on benthic polychaetes, 3
amphipods, coelenterates, shrimp, plant
material, and detritus.
Feeds primarily on copepods and krill. 3
Feeds primarily on copepods and krill. 3
Bottom feeding herbivore/detritivore; 2
consumes some benthic animals.
Filter feeder on phytoplankton, detritus, 2
and bacteria.
Filter feeder on phytoplankton, detritus, 2
and bacteria.
CSFII
Assigned
Trophic
Level (c)
2
3 (50%)
4 (50%)
4
3 (50%)
4 (50%)
3
2
2
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories (continued)
CSFII
Consumption
Category
Perch
Pike
Rockfish
Salmon
Scallop
Shrimp
Common
Name
Silver perch
White perch
Yellow perch
Northern pike
Pickerel
(redfin & grass)
Striped bass
Pink salmon
Coho salmon
Sockeye salmon
Chinook salmon
Bay scallop
Sea scallop
Northern shrimp
Brown shrimp
Scientific
Name Size
Bairdiella chrysoura > 8 cm
Morons americana adult
Percaflavescens 20-30 cm
Esoxlucius >10 cm
Esox americanus larger
specimens
Morons saxatilis >10 cm
Oncorhynchus —
gorbuscha
Oncorhynchus kistuch 45-60 cm
Oncorhynchus nerka —
Oncorhynchus young
tshawytscha
adult
Argopecten irradians —
Placopecten —
magellanicus
Pandalus borealis —
Pandalus aztecus —
Trophic
Level (a)
(Mean)
3.6
3.6
3.4
4
4
3.9
3.8
4
4
3
4
2.2
2.2
2.7
2.7
Trophic
Level (a)
(Range)
3.3^.2
3.3^.2
3.1-3.8
—
—
—
—
4.0^.5
—
—
—
2.1-2.3
2.1-2.4
2.3-2.9
2.3-3.0
Notes (a)
Feeds on polychaetes, shrimp, and
mollusks; becomes more piscivorous
with age.
Benthic predator, becoming
increasingly piscivorous with age.
20-30 cm; consumes 10%
zooplankton, 50% benthic
invertebrates, 34% fish (some
populations); nearly 100% fish in other
populations.
> 10 cm; diet primarily all fish.
Larger specimens consume small fish.
> 10 cm consumes 85%-97% fish.
Feeds at sea; consumes krill,
amphipods, squid, and copepods.
Adults feed primarily on alewife and
smelt. In Lake Michigan, this could
result in a higher trophic level where
alewife feed on Mysis.
Assume trophic level 4 for large
specimens.
Young in fresh water feed on terrestrial
insects taken at water surface.
Marine adults consume primarily fish,
some amphipods, and other inverts.
Smaller feeder than ocean scallop.
Filter feeder, consuming primarily algae
but also zooplankton, bacteria, and
detritus.
Large; feeds on polychaetes,
copepods, benthic and planktonic
microorganisms, and algae.
Omnivore/predator, consuming
polychaetes, amphipods, detritus, and
algae.
Species
Assigned
Trophic
Level (b)
4
4
3
4
4
4
4
4
4
3
4
2
2
3
3
CSFII
Assigned
Trophic
Level (c)
4
4
4
4
2
2 (50%)
3 (50%)
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories (continued)
CSFII
Consumption
Category
Shrimp
Common
Name
Pink shrimp
Scientific
Name Size
Panaeus duorarum —
Trophic
Level (a)
(Mean)
2.4
Trophic
Level (a)
(Range)
2.1-2.9
Notes (a)
Benthic omnivore, consuming
Species
Assigned
Trophic
Level (b)
2
CSFII
Assigned
Trophic
Level (c)
Smelt
Smelt
Snail
Sturgeon
Trout
White shrimp
Jacksmelt
Surf smelt
Rainbow smelt
Eulachon
Snail
Green sturgeon
White sturgeon
Lake sturgeon
Brook trout
Cutthroat trout
Penaeus setiferus —
Atherinopsis adult
californiensis
Hypomesus pretiosus —
Osmerus mordax
Thaleichthys pacificus adult
Acipenser medirostris > 1 m
Acipenser > 1.2 m
transmontanus
Acipenser rubicundus —
Salvelinus fontinalis 10^0 cm
Salmo clarki
<40 cm
polychaetes, other crustaceans,
mollusks, algae, and vascular plant
detritus.
2.3 2.1-2.7 Omnivore, but consuming more plant 2
material than do other shrimp.
2.5 2.2-2.8 Omnivorous; feeds on algae, diatoms, 3
detritus, and small crustaceans.
3.1 3.0-3.2 Consumes primarily amphipods, 3
euphausiids, copepods, other
zooplankton.
3.4 3.2-3.8 Feeds on krill, amphipods, polychaetes, 3
plant debris, small fish, including
herring cunner, anchovy, and
silversides.
Feeds primarily on zooplankton, 3
including krill and copepods.
Most species are strictly herbivorous, 2
grazing on periphyton or other plant
materials.
Benthic carnivore, feeding on 4
invertebrates and small fish.
Adults are benthic carnivores, feeding 4
on invertebrates, including shrimp and
bivalves. Larger juveniles and adults
feed on fish, including eulachon,
anchovies, minnows, and suckers.
— 3-4 Can grow up to 100 pounds, averages 4
about 40-50 pounds for adults;
primarily a bottom feeder, reportedly
feeding on small gastropods,
crustaceans, insect larvae, and small
fishes.
3.2 — 10^0 cm; at most, 7%-8% fish in diet; 3
remainder primarily benthic
invertebrates but also some
zooplankton in some populations.
3 — < 40 cm; consumes invertebrates. 4
3.1
2
3.7
3.7
3.0-3.2
—
3.5^.0
3.5^.0
-------
Table 6-4. Trophic Level Assignment of Aquatic Species Corresponding to CSFII Consumption Categories (continued)
CSFII
Consumption Common Scientific
Category Name Name
Size
>40 cm
old adults
Trophic
Level (a)
(Mean)
3.2
4
Trophic
Level (a)
(Range)
—
Notes (a)
> 40 cm; becomes piscivorous.
Assumption for oldest specimens.
Species
Assigned
Trophic
Level (b)
CSFII
Assigned
Trophic
Level (c)
Dolly Varden trout Salvelinus malma
Dolly Varden trout Salvelinus malma
10-30 cm
30^0 cm
>40 cm
Trout
Lake trout Salvelinus namaycush 20-30 cm
30^0 cm
>40 cm
Rainbow trout Oncorhynchus mykiss < 30 cm
30-50 cm
Whitefish
Round whitefish Prosopium
cylindraceum
> 50 cm
20-30 cm
Mountain whitefish Prosopium williamsoni > 30 cm
— 3-4 Trophic level changes with age.
3 — 10-30 cm; diet 100% benthic
invertebrates.
3.75 — 30^10 cm; diet 75% fish, 17% benthic
invertebrates.
4 — > 40 cm; diet consists of 100% fish.
3.7 3.5-4.0 20-30 cm; feeds primarily on small fish
(70%) and benthic invertebrates (30%).
3.9 3.7-4.1 30^10 cm; feeds primarily on fish
(90%) and some benthic invertebrates
(10%).
4.2 4.0-4.5 > 40 cm; feeds entirely on fish; in Lake
Michigan, feed on alewives, which feed
on Mysis, which feed on zooplankton.
3 — < 30 cm; diet completely of benthic
invertebrates or both invertebrates and
zooplankton.
3.6 — 30-50 cm; diet 35-90% fish, 25-75%
benthic invertebrates, zooplankton,
terrestrial insects.
4 — > 50 cm; diet of 100% fish.
3 — Young probably feed on zooplankton;
adults feed primarily on benthic
invertebrates.
3.5 — Greater than 30 cm; consumes large
invertebrates and small fish.
Unless otherwise specified, information on trophic status was obtained from USEPA (2000e-g). Game fish data were limited to specimens considered to be representative of the
edible size range (i.e., 20 cm or larger).
In determining species trophic level assignments, preference was given to data on larger specimens. Trophic level 4 was assigned to a species with data indicating trophic level 3.5
or higher; trophic level 3 was assigned to a species with data indicating trophic level 2.5-3.4; trophic level 2 was assigned for trophic level 1.5-2 A.
In determining CSFII category trophic level assignments, best professional judgment was used. For example, the CSFII category for catfish includes four species that are assigned
to trophic level 3 and three species assigned to trophic level 4. Thus, it is assumed that half (50%) of consumption in the catfish CSFII category is from TL3 and half from TL4.
Except for shrimp, all other CSFII categories included species that either were exclusively or predominately one trophic level (e.g., trout, estuarine flatfish, smelt).
-------
Table 6-5. National Default Values of Lipid Fraction
Trophic Level National Default Value
(percentage)
2 1.9%
3 2.6%
4 3.0%
6.2.4 Uncertainty and Sensitivity Analysis
This section discusses the uncertainty and sensitivity associated with the calculation of
national default values of lipid fraction described in the previous section. The objective here is to
identify the major sources of uncertainty in the present analysis and, where possible, provide
some insight into their potential magnitude and direction of impact on the national default values
of lipid fraction. In this way, the overall confidence in the default lipid values can be assessed (at
least qualitatively) and steps to reduce this uncertainty can be identified. Although ideally one
would attempt to address each source of uncertainty quantitatively, available data and resources
did not permit complete quantitative analysis of uncertainty. A quantitative analysis of selected
sources of uncertainty is provided at the end of this section.
Qualitative Analysis
Applicability of Fish Consumption Rate Data. An integral part of EPA's calculation of
national default values of lipid fraction involved the estimation of the type and quantity offish
and shellfish consumed by the U.S. population. As described previously, data on fish and
shellfish consumption were obtained from USDA's CSFII for the years 1994-1996 (USDA,
1998). A number of uncertainties are associated with the use of the CSFII consumption data,
some of which are described in more detail in the Exposure Assessment volume of this Technical
Support Document. First and foremost, the mean per capita rates offish and shellfish
consumption derived from the CSFII are national in scope. As a result, the national pattern offish
and shellfish consumption developed from the CSFII may differ from the consumption patterns
represented by various human subpopulations, particularly on local or regional scales. Although
the magnitude of this uncertainty has not been quantified here, its impact on the national default
values of lipid fraction is believed to be bidirectional (i.e., resulting in an overestimation or
underestimation of lipid fraction applicable to local or regional scenarios). Because the CSFII
consumption rates are weighted toward leaner aquatic organisms (e.g., shrimp, flounder, flatfish),
it is conceivable that they may lead to a greater tendency in the national default values of lipid
fraction to underestimate lipid fraction associated with some local and regional consumption
patterns compared with overestimating lipid fraction. The magnitude of uncertainty in applying
the CSFII consumption rates to local or regional situations will depend on the extent to which
local consumption patterns differ from the pattern represented by the CSFII.
6-28
-------
Table 6-6. Calculation of National Default Values of Consumption-Weighted Mean Lipid Fraction
Habitat
Estuarine
Estuarine
Freshwater
Estuarine
Estuarine
Estuarine
Estuarine
Freshwater
Estuarine
Freshwater
Estuarine
Freshwater
Freshwater
Estuarine
Estuarine
Estuarine
Estuarine
Estuarine
Estuarine
Estuarine
Estuarine
Freshwater
Freshwater
Estuarine
Freshwater
Estuarine
Estuarine
Estuarine
Estuarine
Estuarine
Freshwater
Freshwater
Estuarine
CSFII
Consumption
Category
Anchovy
Clam
Crayfish
Mullet
Oyster
Scallop
Shrimp
Snails
Anchovy
Carp
Catfish
Catfish
Cisco
Crab
Croaker
Flatfish
Flounder
Herring
Shrimp
Smelts
Smelts, rainbow
Smelts, rainbow
Whitefish
Catfish
Catfish
Croaker
Eel
Flatfish
Flounder
Perch
Perch
Pike
Rockfish
Assigned Mean
Trophic Trophic Level Percent
Level Weighting Factor Lipid
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
0.5
1
1
1
1
1
0.5
1
0.5
1
0.5
0.5
1
1
0.5
0.5
0.5
1
0.5
1
1
1
1
0.5
0.5
0.5
1
0.5
0.5
1
1
1
1
6.1
1.3
1.1
3.8
2.4
0.8
1.7
1.4
6.1
5.4
4.0
2.9
1.9
1.1
3.0
1.2
1.2
10.0
1.7
4.1
4.1
4.1
5.9
4.0
2.9
3.0
11.7
1.2
1.2
2.3
2.3
0.7
3.5
Mean
Consumption
Rate
(g/person/d)
0.05815
0.01732
0.01028
0.04512
0.17485
0.00140
2.65492
0.00207
0.05815
0.19071
0.60335
0.60335
0.00179
0.42111
0.17792
0.45173
0.73482
0.16428
2.65492
0.00880
0.00076
0.00076
0.01309
0.60335
0.60335
0.17792
0.00466
0.45173
0.73482
0.22331
0.22331
0.04021
0.05428
Trophic Level
Weighted
Consumption Rate
(g/person/day )
0.02907
0.01732
0.01028
0.04512
0.17485
0.00140
1.32746
0.00207
0.02907
0.19071
0.30168
0.30168
0.00179
0.42111
0.08896
0.22586
0.36741
0.16428
1.32746
0.00880
0.00076
0.00076
0.01309
0.30168
0.30168
0.08896
0.00466
0.22586
0.36741
0.22331
0.22331
0.04021
0.05428
CSFII Category
Weights
0.01809
0.01077
0.00640
0.02807
0.10877
0.00087
0.82575
0.00129
0.00844
0.05538
0.08761
0.08761
0.00052
0.12229
0.02584
0.06559
0.10670
0.04771
0.38551
0.00256
0.00022
0.00022
0.00380
0.12305
0.12305
0.03629
0.00190
0.09212
0.14986
0.09108
0.09108
0.01640
0.02214
Consumption
Weighted Percent
Lipid Values
0.11057
0.01380
0.00678
0.10638
0.25941
0.00066
1.42855
0.00181
0.05162
0.29671
0.35292
0.25541
0.00099
0.13928
0.07865
0.07806
0.12697
0.47898
0.66693
0.01048
0.00090
0.00090
0.02228
0.49566
0.35871
0.11046
0.02218
0.10963
0.17833
0.20599
0.20599
0.01190
0.07638
-------
Table 6-6. Calculation of National Default Values of Consumption-Weighted Mean Lipid Fraction (continued)
Habitat
Estuarine
Freshwater
Estuarine
Freshwater
Freshwater
Estuarine
Freshwater
CSFII
Consumption
Category
Salmon
Salmon
Sturgeon
Sturgeon
Trout (rainbow)
Trout, mixed sp.
Trout, mixed sp.
Assigned
Trophic
Level
4
4
4
4
4
4
4
Mean
Trophic Level Percent
Weighting Factor Lipid
1 4.7
1 4.7
1 5.3
1 5.3
1 5.1
1 3.2
1 6.0
Mean
Consumption
Rate
(g/person/d)
0.05915
0.00118
0.00018
0.00018
0.25361
0.15305
0.15305
Trophic Level
2
3
4
Total
Trophic Level
Weighted
Consumption Rate
(g/person/day )
0.05915
0.00118
0.00018
0.00018
0.25361
0.15305
0.15305
Consumption Rate
fg/nerson/dav)
1.60757
3.44341
2.45175
7.50273
CSFII Category
Weights
0.02412
0.00048
0.00007
0.00007
0.10344
0.06242
0.06242
Sum of Weights
1.00000
1.00000
1.00000
Consumption
Weighted Percent
Lipid Values
0.11420
0.00227
0.00039
0.00039
0.53028
0.19788
0.37425
Consumption-
Weighted Mean
Percent Linid
1.9
2.6
3.0
-------
Of particular importance will be the extent to which the lipid contents of locally consumed
aquatic organisms differ from those corresponding to species that drive EPA's calculation of the
national default lipid values (e.g., shrimp, flounder, flatfish, catfish). Interestingly, an analogous
derivation of consumption-weighted values of lipid fraction specific for the Great Lakes region
resulted in estimates that were similar to the national default values (i.e., 1.8% for trophic level 3,
3.0% for trophic level 4), despite considerable differences in consumption patterns (USEPA,
1995b). To address uncertainty associated with potential differences in consumption patterns and
associated lipid fraction, EPA recommends that States and Tribes use local or regional
information on fish and shellfish consumption to calculate values for lipid fraction whenever
possible.
Specificity of Fish Consumption Rate Data. Another attribute associated with the
CSFII-derived consumption rates that leads to uncertainty in the national default values of lipid
fraction is the lack of specificity in the consumption rate categories (trout, flatfish, catfish).
Specifically, consumption rate information was not available at the species level (e.g., lake trout,
brook trout). Therefore, within a CSFII consumption category, equal weighting of lipid fraction
among species was assumed. Obviously, to the extent that this assumption is violated,
uncertainty will be introduced into the default values of lipid fraction.
To provide some insight into the effect of violating EPA's assumption of equal weighting
among species lipid values on the national default values of lipid fraction, the default values of
lipid fraction were recalculated assuming 100% weighting to the species with the highest mean
lipid value within a CSFII consumption category, and again assuming 100% weighting to the
species with the lowest mean lipid value. Values of lipid fraction calculated with the species with
the lowest and highest mean lipid values are provided in Table 6-7, along with EPA's national
default values of lipid fraction (calculated with the average of species mean lipid values).
It is apparent from this exercise that substantially different assumptions about the
weighting of species mean lipid values within a CSFII consumption category have relatively little
impact on the national default values of lipid fraction (i.e., <50% increase or decrease). However,
this analysis was constrained by the limited availability of lipid data for multiple species within a
CSFII consumption category. Specifically, of the consumption categories
Table 6-7. Sensitivity of National Default Values of Lipid Fraction to Different Weighting
Assumptions Among Species
Trophic Level Calculated Using Lowest Calculated Using Average of Calculated Using Highest
Species Mean Lipid Values Species Mean Lipid Values" Species Mean Lipid Values
2 1.9 1.9 2.0
3 2.1 2.6 3.1
4 1.8 3.0 4.4
1 Weighting assumption used for calculating the national default values of lipid fraction.
6-31
-------
constituting the bulk of the total consumption rate (shrimp, catfish, flounder, flatfish), only the
catfish category was represented by multiple species. Therefore, the impact of different
assumptions about the weighting of species mean lipid values is likely to be underestimated by
the present analysis.
Model Uncertainty, Measurement Error, and Variability in CSFII-Derived Fish
Consumption Rates. Other sources of uncertainty in EPA's national default values of lipid
fraction include model uncertainty, measurement error, and variability associated with the
estimates of mean fish consumption rates from the CSFII study. Even though data and resource
limitations prevented EPA from assessing the magnitude and direction of these sources of
uncertainty, it is still considered instructive to discuss their overall characteristics. The term model
uncertainty is used here to represent uncertainty originating from the design of the CSFII study
and its application to AWQC derivation. Specifically, the fish consumption rates from the CSFII
study were based on a 2-day dietary recall from a stratified random sample of the U.S.
population. In many situations, AWQCs are derived to protect against adverse effects from long-
term (chronic) exposures to chemicals from sources including fish consumption. Under these
AWQC applications, so-called model uncertainty is introduced in the estimated fish consumption
rates to the extent that daily mean per capita consumption rates estimated over a 2-day period
deviate from the "true" daily consumption rates over the long term.
The term measurement error refers to the error associated with recalling from memory
the type and quantity offish and shellfish actually consumed. Interestingly, although the previous
discussion of model uncertainty might lead one to favor survey designs with longer recall periods
(e.g., weekly, monthly), measurement error can increase substantially for longer survey recall
periods.
Finally, one can expect the variability associated with the estimated mean per capita
consumption rates to affect the derivation of national default values of lipid fraction. This
variability would reflect variation in the amount and types offish and shellfish actually consumed
across individuals, in addition to differences in the ability of individuals to recall what they ate in
the past (measurement error). To assess the effect of this source of variability on the default
consumption rates, one would need some estimate of the variance of mean per capita
consumption rates for each CSFII category. However, limitations in the CSFII study prevent
accurate estimates of this variance at the CSFII category level.
For a more detailed discussion of uncertainty associated with the use of data from the
CSFII study, see the Exposure Assessment volume of the Technical Support Document.
Uncertainty in Trophic Level Classification. As illustrated by Table 6-4, variation exists
in the trophic position of commonly consumed aquatic organisms. Sources of this variability can
be attributed to numerous factors, including the size and life stage of the organism, the season,
the organism's life history (e.g., migratory behavior), and spatial heterogeneity in the food web
structure. To calculate national default values of lipid fraction, EPA relied on a synthesis of data
on the trophic position of aquatic organisms (USEPA, 2000e-g). Data from these syntheses of
trophic positions were ultimately rounded to nominal values (e.g., 1, 2, 3, 4), when in reality a
6-32
-------
continuum exists for an organism's trophic position (e.g., 1.3, 2.6, 3.7). To the extent that the
trophic position of consumed aquatic organisms at various locations differs from EPA's
assessment of trophic position, uncertainty will be introduced in derivation of the national default
lipid values. For some organisms (e.g., clams, oysters, scallops), the variability in trophic position
appears small (Table 6-3) and the likely impact on the default values of lipid fraction is expected
to be minimal. For other groups of organisms, (e.g., anchovy, catfish, croaker, flatfish, flounder,
shrimp), wider variation exists in trophic position within and across species, in part because of
their opportunistic feeding style. For these CSFII categories, EPA weighted the consumption
rates equally between multiple trophic levels (Table 6-6). By chance, these more "opportunistic"
species make up a large fraction of the total rate of consumption of freshwater and estuarine
species. To assess the sensitivity of EPA's national default values of lipid fraction to the
assumption of equal weighting of consumption rates across trophic levels for selected species
(anchovy, catfish, croaker, flatfish, flounder, and shrimp categories), national default values of
lipid fraction were recalculated by assuming that 100% of the consumption occurred in the lower
trophic level, and again assuming that 100% of the consumption occurred in the higher trophic
level. For example, calculations were performed assuming that all of the consumption of catfish
species occurred at trophic level 3 and again at trophic level 4. Results from this sensitivity
analysis are shown in Table 6-8.
It is apparent from Table 6-8 that the national default values of lipid fraction are relatively
insensitive to assumptions made about the trophic position of those species for which the trophic
position is particularly variable (anchovy, catfish, croaker, flatfish, flounder, and shrimp).
Table 6-8. Sensitivity of National Default Values of Lipid Fraction to Different Weighting
Assumptions Among Trophic Levels"
Trophic Level Calculated Assuming 100% Calculated Assuming 50% Calculated Assuming 100%
Consumption at the Lower Consumption at Lower and Consumption at the Higher
Trophic Level Higher Trophic Level1" Trophic Level
2 2.5 1.9 1.9
3 2.3 2.6 2.8
4 2.8 3.0 3.7
1 Different weighting assumptions were made for anchovy, catfish, croaker, flatfish, flounder, and shrimp CSFII consumption
categories.
b Weighting assumption chosen for calculating the national default values of lipid fraction.
Limitations in the Lipid Data. A number of limitations in the lipid data contribute to
uncertainty in the national default values of lipid fraction. First, the lipid data used to calculate the
national default values of lipid fraction were originally generated for a variety of purposes and by
a variety of methods. These data almost certainly do not represent a random sampling of aquatic
organisms that is properly stratified over potentially important variables such as age, tissue type,
and season. As a result, the data set may contain hidden biases that are difficult to assess without
a comparison with a truly random, stratified sample. For example, some species data sets may be
overrepresented by one or more tissue types, where multiple tissue types are being consumed
6-33
-------
(e.g., fillet with and without skin). Other species data sets might contain biases that result from
overrepresentation of certain lipid measurement methods. Second, the sample size is small for a
number of species (i.e., <10 for certain species of clam, crayfish, croaker, sturgeon, and salmon).
In general, the lower sample sizes for these species result in lower confidence in the species mean
values of lipid fraction. Some commonly consumed species may not be represented at all in a
given CSFII consumption category because lipid data were simply unavailable. Third, the number
of species represented in certain CSFII consumption categories is small or unknown. The latter
situation occurred for shrimp, flounder, flatfish, crayfish, eel, and whitefish categories, where
lipid data were available only in an aggregated form (e.g., at the family level for shrimp). Finally,
the quality of some of the lipid data (in particular, the STORET-based NSQS data) was not
documented and could not be verified directly.
Quantitative Analysis
Variability in Lipid Data. As shown in Table 6-3, estimates of variation around the
species mean value of lipid content are available for most species. Multiple sources of variation
are believed to contribute to the observed variation in species lipid content. Within a species,
these sources include measurement error; differences in lipid extraction and quantification
methods; inclusion of data from different tissue types, ages, and sizes of organisms; and different
dietary habits among individuals within a species, to name a few.
To assess how these and other sources of variability in lipid content affect the uncertainty
in EPA's calculation of national default values of lipid fraction, a probabilistic-based uncertainty
analysis was conducted, using the estimated variance and mean values shown in Table 6-2. This
analysis relied on several assumptions:
1. Mean and coefficient of variations in species lipid content were defined from the
data summarized in Table 6-3.
2. The values of lipid content for each species were assumed to be log-normally
distributed. This assumption was consistent with the positive skewness (and non-
negative nature) of percentage data and was supported by visual inspection of
frequency distributions from selected lipid data sets.
3. In a few situations (e.g., striped anchovy, northern anchovy, rockfish [Sebastes
spp.]), estimates of variance around the mean value were not available. In these
cases, the coefficient of variation was assumed to be equal to that calculated from
another species in the same CSFII consumption category.
4. For two CSFII consumption categories (clam, crayfish), no information on
variance was available from any species. As a result, no variance was assumed
around the mean values.
5. Trophic level designations and mean per capita consumption were held constant.
6-34
-------
6. All input distributions were assumed to be independent (i.e., no correlation among
distributions).
With the aforementioned information and supporting assumptions, a probabilistic
uncertainty analysis was run, using a Monte Carlo simulation technique for calculating the
national default values of lipid fraction. The Monte Carlo simulation used Crystal Ball® version
4.0 software (Decisioneering, Inc., Denver, Colorado) in combination with a Microsoft Excel® 97
spreadsheet application. For each iteration of the simulation, a consumption-weighted average
value of lipid fraction (i.e., a default value of lipid fraction) was calculated for each trophic level,
using a randomly selected value of lipid fraction from each of the species input distributions. The
simulation was run for 10,000 iterations, thus producing a distribution of default values of lipid
fraction for each trophic level. Repeated simulations indicated that 10,000 iterations produced
highly stable estimates of the mean and extreme percentiles of the default lipid values.
Figures 6-2 and 6-3 show the results of the Monte Carlo simulations. The x-axis of Figure
6-2 refers to the default value of lipid fraction (expressed as a percentage), and the y-axis displays
the probability. It is apparent that the default values of lipid fraction comprise three distinct but
somewhat overlapping distributions. For clarity, Figure 6-2 displays the same information in the
form of a reverse cumulative frequency. The y-axis displays the frequency by which a given
value of the default value of lipid fraction (x-axis) is exceeded in the data set. From Figure 6-3,
one can estimate the likelihood that the default lipid fraction would exceed a particular value.
Relevant descriptive statistics from the output distributions of default values of lipid
fraction are shown in Table 6-9. As expected, mean values produced from the Monte Carlo
simulations were identical (to two significant digits) to the national default values (Table 6-4)
calculated by using only the species mean values as inputs. Regarding variation surrounding the
mean values, it is often useful to evaluate the range between the 5th and 95th percentiles of the
distribution. From this measure, it is evident that the variability around the mean values of lipid
fraction is relatively small (a factor of • 2.5). The most sensitive input distributions to the
calculated default values of lipid fraction are shrimp (for trophic level 2), shrimp and common
carp (for trophic level 3), and rainbow trout (for trophic level 4). Each of these input distributions
contributed approximately 25% or more to the variance in the calculated default lipid values.
Finally, EPA acknowledges that there is similarity among the national default value lipid
fractions across trophic levels (i.e., 1.9, 2.6, 3.0) and that the uncertainty bounds somewhat
overlap. This degree of similarity might support the notion of calculating a single national default
value of lipid fraction rather than maintaining distinctions among trophic levels. Although EPA
considered this option, it was ultimately rejected in favor of maintaining separate national default
values of lipid fraction at each trophic level for various reasons. First, maintaining trophic level
specificity in lipid fraction is consistent with EPA's derivation of national BAFs, which are
calculated separately for trophic levels 2, 3, and 4. As explained in the 2000 Human Health
Methodology (USEPA, 2000a), trophic level-specific BAFs are derived to account for factors that
can affect bioaccumulation in aquatic organisms occupying different trophic positions in aquatic
food webs (e.g., biomagnification and broad physiological differences such as clams versus fish).
In addition to the improved technical accuracy associated with applying trophic level-specific
6-35
-------
values of lipid fraction, maintaining a separate distinction across trophic levels also provides
flexibility to States and Tribes for adjusting EPA's national default lipid values. Specifically,
adjustments can be made to the calculation of one trophic level-specific value of lipid fraction
without affecting those determined for the other trophic levels. For example, a State or Tribe may
wish to add or subtract lipid data for various top predator species (trophic level 4) without
changing the values of lipid fraction for other trophic levels. Thus, EPA believes that its use of
trophic level-specific values of lipid fraction not only achieves greater technical accuracy than
does a single estimate, but also affords greater flexibility to States and Tribes in making desired
adjustments to EPA's national default values of lipid fraction.
Overlay Chart
n
J3
O
.104 -I
.078 -•
.052 -•
.026 -•
.000
TL2-cons. wt. lipid
TL3-cons. wt. lipid
TL4 Cons. Wt. Lipid
0.50
1.75
3.00
4.25
5.50
Figure 6-2. Frequency distribution of national default values of lipid fraction (10,000 iterations).
Overlay Chart
Reverse Cumulative Comparison
n
j=i
o
1 .000 -
.750 -•
.500--
.250--
.000
TL2-cons. wt. lipid
TL3-cons. wt. lipid
TL4 Cons. Wt. Lipid
0.50
1.75
3.00
4.25
5.50
Figure 6-3. Reverse cumulative comparison of national default values of lipid fraction (10,000 iterations).
6-36
-------
Table 6-9. Descriptive Statistics of Monte Carlo Simulation of National Default Values of
Lipid Fraction (10,000 Iterations)
Statistic/Percentile
Mean
Median
CV
5th
95th
95th/5th
Minimum
Maximum
Trophic Level 2
1.9
1.8
0.30
1.2
3.0
2.5
0.7
6.1
Trophic Level 3
2.6
2.5
0.17
2.0
3.4
1.7
1.4
5.9
Trophic Level 4
3.0
2.9
0.17
2.3
3.9
1.7
1.6
6.5
6.3 BASIS FOR THE NATIONAL DEFAULT VALUES OF DOC AND POC
This section provides the technical basis of EPA's calculation of national default values of
DOC and POC concentrations in U.S. fresh and estuarine surface waters. As summarized in the
National Human Health AWQC Methodology (USEPA, 2000a), EPA's national default values of
DOC (2.9 mg/L) and POC (0.5 mg/L) are used in calculating national BAFs for nonionic organic
chemicals. Information on DOC and POC is necessary to adjust a baseline BAF, which reflects
the concentration in the lipid fraction of tissue and the freely dissolved concentration in water, to
a national BAF expressed in terms of total chemical in water and tissue (Section 3, Figure 3-2).
The national BAF incorporates values that are reflective of the lipid content of the fish and
shellfish consumed by the U.S. population and the effects of chemical binding/associating with
DOC and POC in representative U.S. surface waters. For deriving national human health
AWQCs, EPA uses national default values of DOC and POC that are representative of U.S.
surface waters for calculating the freely dissolved fraction of a nonionic organic chemical (see
Equation 4-4).
Although EPA uses national default values of DOC and POC to derive national human
health AWQC for nonionic organic chemicals, EPA encourages States and authorized Tribes to
use local or regional data on the organic carbon content of applicable waters when adopting
criteria into their own water quality standards. The EPA encourages the use of appropriately
derived locally or regionally derived values of DOC or POC over nationally derived values
because local or regional conditions that affect DOC and POC concentrations can differ
substantially from those represented by nationally derived values. Additional guidance on
developing local or region-specific values of DOC and POC is found in a subsequent volume of
this TSD (Volume 3: Development of Site-Specific Bioaccumulation Factors). Nevertheless, EPA
recognizes that there will be situations when such local or regional data are not available or are
inadequate for deriving local or regional values of DOC and POC. In these cases, EPA
6-37
-------
recommends that States and Tribes use EPA's national default values of DOC and POC when
deriving BAFs for use in establishing State or Tribal water quality criteria and standards.
The following sections present the data sources, analysis, and uncertainty associated with
EPA's derivation of national default values of DOC and POC.
6.3.1 Data Sources
Data on the concentrations of DOC and POC in U.S. surface waters were obtained from
two databases:
1. The U.S. Geological Survey's (USGS) WATSTORE database
2. EPA's historical STORET database (recently renamed the Legacy Data Center
[LDC] database)
Although EPA's historical STORET database (henceforth called "LDC" for consistency
with current EPA nomenclature) contains data from the USGS WATSTORE database, queries
indicated that the USGS data contained in LDC were at least 2 years out of date. Therefore, non-
USGS data were retrieved from the LDC database and USGS data were retrieved from the
WATSTORE database in order to obtain a comprehensive data retrieval without duplicating
records. Each database is described further below.
WATSTORE Database
The USGS developed the WATSTORE (National Water Data Storage and Retrieval
System) for the storage and retrieval of water data collected through its activities. The
WATSTORE database was established in 1972 to provide an effective and efficient means for
processing and maintaining water data collected through USGS activities and to facilitate release
of the data to the public. The system resides on the central computer facilities of the USGS at its
National Center in Reston, Virginia, and consists of related files and databases. The Water
Quality File was searched for retrieval of DOC and POC data. This file contains approximately 2
million analyses of water samples that describe the chemical, physical, biological, and
radiological characteristics of both surface and groundwater. The method for analysis of POC
followed Standard Methods #5310D—"Wet Oxidation Method for Total Organic Carbon"
(APHA, 1995), with two modifications. First, silver filters were used instead of glass fiber filters.
Second, a sonification step was added in 1997 to facilitate complete oxidation of organic carbon
(USGS, 1997; Burkhardt et al., 1999). For analysis of DOC, the wet oxidation method (Standard
Methods #5310D) was also used on filtered samples until approximately 1983. After 1983, the
persulfate-ultraviolet oxidation method was used (StandardMethods #5310Cy) on filtered
samples, which includes UV radiation with a reduced heating/digestion step (APHA, 1995;
Kammer, 2000).
6-38
-------
LDC Database
EPA's LDC database is a waterway-related monitoring database that contains data
collected by Federal, State, and local agencies; Indian Tribes; volunteer groups; academics; and
others. All 50 States, territories, and jurisdictions of the United States, along with portions of
Canada and Mexico, are represented in the database. The database used in this analysis was
called the historical, or "old," STORET database but recently was renamed the Legacy Data
Center (LDC). The LDC contains historical water quality data dating back to the early part of the
20th century and collected up to the end of 1998. The database contains raw biological, chemical,
and physical data on surface water and groundwater. Each sampling result is accompanied by
information on where the sample was taken (latitude, longitude, State, county, Hydrologic Unit
Code, and a brief site identification), when the sample was gathered, the medium sampled (e.g.,
water, sediment, fish tissue), and the name of the organization that sponsored the monitoring.
Information on the analytical methods used to quantify DOC and POC was not available in the
LDC database. Although a newer version of STORET was initiated in 1999 that contained this
and other QA/QC information, sufficient data were not available from this database at the time of
this retrieval (i.e., the new STORET database contains data collected beginning in 1999, along
with selected older data that have been properly documented and migrated from the LDC).
6.3.2 Data Retrieval and Screening
Data retrievals from the LDC and WATSTORE databases were conducted in January
2000 and combined into a single relational database. Originally, approximately 800,000 records
containing data on POC, DOC, or total organic carbon (TOC) were retrieved for the period
beginning in 1970 through the latest year data were available (1999 for WATSTORE; 1998 for
LDC). This retrieval was limited to samples taken from ambient surface waters (i.e., samples from
wells, springs, effluents, and other nonambient sources were excluded). Additionally, this initial
retrieval included multiple types of POC and DOC measurements to ensure that the initial data
retrieval would be sufficiently comprehensive.
Once these data were retrieved, the two data retrievals were combined into a single
database. Numerous steps were then taken to process and screen the DOC and POC data so that
only the most appropriate data would be retained for calculating the national default values.
These processing and screening steps are outlined below.
1. Organic Carbon Parameters. The following parameter codes were retained in the
database: 00680 (Carbon, Total Organic); 00681 (Carbon, Dissolved Organic);
00684 (Carbon, Dissolved Organic-Whatman GF/F); 00689 (Carbon, Suspended
Organic); 80102 (Carbon, Organic Particulate). All units were expressed in
milligrams per liter as C (Carbon).
2. Uncertain Values. Values that were coded in such a way as to suggest uncertainty
in the measurement were deleted from the database. For example, values coded as
"estimated value," "analyte detected in blank and sample," "sample held beyond
6-39
-------
normal holding time," "actual value is known to be greater reported value," and
similar such indicators were deleted.
3. Water Body and Station Types. The database was further restricted to the
following water body types: estuaries, lakes, reservoirs, and streams (including
rivers). This step excluded other "ambient" surface water types, such as oceans,
freshwater wetlands, and canals. Station types were restricted to those coded as
"ambient" only. This step excluded so-called specialty stations (i.e., those stations
designated for special purposes such as storm water runoff and biological and
sediment monitoring).
4. Sampling Period. Although the initial retrieval contained data dating back to 1970,
the time period for the final database was restricted to 1980 through 1999. Pre-
1980 data were eliminated because of the greater uncertainty in using these data to
represent present-day conditions that can affect organic carbon concentrations in
surface waters (e.g., secondary treatment of effluents).
5. Detection Levels. Some values for POC and DOC were reported to be below
analytical detection levels. In this situation, the value was assumed to be half of
the reported detection level. Values with "high" detection levels (i.e., >1.0 mg/L
for DOC and >0.2 mg/L for POC) were deleted from the database because of the
greater uncertainty involved in estimating definitive values of DOC and POC in
these situations.
6. Calculated Values. It was clear from reviewing the data that a substantial portion
of samples contained values of DOC and TOC, but not POC. It is apparently not
uncommon to determine the POC concentration by subtracting the DOC
concentration from the TOC concentration determined from a given sample. In
these situations, the parameter of interest (POC or DOC) was calculated by the
difference from the other two measurements (i.e., POC = TOC - DOC; DOC =
TOC - POC). This calculation was performed using data only from the same
sample to avoid introducing error into calculated POC values. The end result was
that about 40% of the total number of POC values in the database were
determined by difference. The opposite condition (i.e., TOC and POC, but no
DOC value) occurred rarely and resulted in only 0.4% of the total DOC samples
being determined by difference.
7. Extreme Values. As a final quality control step, DOC and POC values at the
extreme high end of the cumulative frequency distributions were reviewed for
consistency with extreme values reported in natural surface waters of the United
States. A small fraction of the DOC and POC concentrations in the LDC database
exceeded concentrations considered to represent upper limits of DOC and POC
concentrations reported in U.S. water bodies (i.e., 0.2% exceeded 60 mg/L for
DOC and 0.6% exceeded 30 mg/L for POC). These extreme values were based on
a review of organic carbon data by Thurman (1985), who reported extreme values
6-40
-------
of DOC concentrations of as high as 50 mg/L in distrophic lakes and 60 mg/L in
tributaries draining wetland systems. Concentrations of POC between 1 and 30
mg/L encompass 99% of the world's river systems and are at the upper range of
POC concentrations reported for U.S. rivers (as reviewed by Thurman, 1985).
The evaluation of extreme values revealed some "negative" values of POC (i.e.,
about 7% of the total number of POC values). These values occurred almost
entirely as an artifact of calculating POC values by difference (see item 6 above)
and the impact of measurement error on this process. For example, if both TOC
and DOC were near analytical detection limits or were otherwise very similar in
magnitude, it would not be surprising for reported values of DOC to be, on
occasion, slightly higher than those for TOC as a result of measurement error. The
vast majority of the negative values were relatively close to zero (i.e., between -1
mg/L and 0 mg/L).
To address concerns about the impact of extreme values on the calculation of
national default values for DOC and POC, the extreme ends of the respective
distributions were truncated. Specifically, values of DOC above 60 mg/L and of
POC above 30 mg/L were omitted from the database. These values represent the
99.8th and 99.4th cumulative percentiles of the respective DOC and POC
distributions. To avoid introducing bias into the median values of DOC and POC
by truncating one side of the distribution, DOC and POC values below the lower
0.2% and 0.6%, respectively, were also omitted. This truncation of the lower tail of
the POC distribution bounds had the impact of eliminating some, but not all, of
the "negative" POC values.
6.3.3 Results
Using the screened databases described previously, national default values of DOC and
POC were calculated to be 2.9 mg/L and 0.5 mg/L, respectively. These values represent median
(50th percentile) values from approximately 110,000 measurements of DOC and 86,000
measurements of POC in U.S. fresh and estuarine surface waters. All 50 States are represented in
the database. The EPA selected median values of DOC and POC for the national default values
for consistency with the goal of national BAFs (i.e., as central-tendency estimates).
Table 6-10 shows descriptive statistics surrounding the median values for DOC and POC,
in addition to values for specific water body types. It is evident from Table 6-10 that variation in
DOC and POC concentrations is relatively large. For example, the coefficient of variations
around the means are all above 100% and approach or equal 200% in some cases. Ratios of the
95th to the 5th percentiles range from a factor of 5 to 30, depending on water body type and
parameter. This variation is not unexpected, given the high degree of temporal and spatial
heterogeneity represented in the database. It is also apparent that the type of water body (lake,
stream, estuary) has some impact on the DOC and POC distributions. For example, median
values of DOC and POC from samples designated as "stream/river" are nearly twice those
designated as "lakes." This difference is probably related to the differing hydrologic,
6-41
-------
biogeochemical, and watershed characteristics of streams and lakes. Given the relatively high
degree of variation that is evident in DOC and POC concentrations in surface waters across the
United States, EPA recommends that States and Tribes consider deriving appropriate values of
DOC and POC by using local or regional data when sufficient data are available. However, for
deriving national AWQC, and when States and Tribes lack sufficient local or regional data, EPA
recommends the use of its national default values of DOC and POC.
Table 6-10. National Default Values for POC and DOC in U.S. Fresh and Estuarine Surface
Waters
DOC (mg/L)
Statistic
Median
Mean
Std.
CV
n
5th
10th
25th
75th
90th
95th
95th/5th
All Types
2.9
4.6
5.1
111%
111,059
0.8
1.2
2.0
5.4
9.7
14
17.5
Stream/
River
3.8
5.6
5.9
105%
69,589
0.7
1.0
2.1
6.9
11.6
16.5
23.6
Lake/
Reservoir
2.1
2.9
3.0
103%
25,704
1.0
1.4
1.8
2.6
5.0
7.8
7.8
Estuary
2.7
3.4
2.6
76%
15,766
1.7
2.0
2.3
3.2
5.0
9
5.3
All Types
0.5
1.0
2.0
200%
86,540
Oa
0
0.2
1.1
2.3
3.9
—
POC
Stream/
River
0.6
1.3
2.5
192%
48,238
0"
0"
0.2
1.4
3.1
5
—
(mg/L)
Lake/
Reservoir
0.3
0.5
1.0
200%
23,483
0.08
0.1
0.2
0.5
0.8
1.3
16.3
Estuary
0.9
1.2
1.8
150%
14,819
0.1
0.3
0.5
1.4
2.2
3
30.0
1 Values calculated to be less than zero because of measurement error; see Section 6.3.2 for explanation.
Source: U.S. EPA LDC and USGS WATSTORE databases. Data retrieval: January 2000; see Sections 6.3.1 and 6.3.2 for
description.
6.3.4 Uncertainty/Limitations in National Default Values
This section describes sources of uncertainty associated with EPA's derivation of national
default values of DOC and POC for establishing national human health AWQCs. This discussion
of uncertainty is neither exhaustive nor entirely quantitative. Rather, it focuses on sources of
uncertainty that are likely to have the greatest impact on the derivation and application of national
default values of DOC and POC. Sources of uncertainty characterized below are grouped into the
following categories: (1) sampling bias, (2) measurement error, and (3) natural variability in DOC
and POC concentrations.
6-42
-------
Sampling Bias
The national default values of DOC and POC are intended to represent central tendency
estimates of DOC and POC concentrations in U.S. surface waters. Ideally, the data used to
generate these values should originate from a random sampling of U.S. surface waters and should
be appropriately stratified and weighted by spatial and temporal factors that would be expected to
influence organic carbon concentrations in aquatic ecosystems (e.g., water body type, hydrologic
and watershed characteristics, ecoregion, season). However, this type of database was not
available on a national scale. Therefore, EPA relied on data from USGS's WATSTORE and
EPA's LDC databases to calculate its national default DOC and POC values. The strengths of
these databases include their large number of records (e.g., >110,000 DOC values and >86,000
POC values), a representation of DOC and POC values for all 50 States, and the reasonably long
period over which data were collected (1980-1999 for this analysis).
An important limitation of the WATSTORE and LDC databases is the fact that they do
not reflect a random sampling of U.S. surface waters (i.e., they may contain biases because of
sampling design). For example, about half of the DOC and POC values in the databases were
sampled in Maryland, New York, Ohio, Florida, and Delaware. Thus, some States are
disproportionally represented, even when one considers the relative area of surface water area
likely to be contained within each State. In addition, organic carbon data from these databases
were not weighted or aggregated in any way before national default (median) values were
calculated. Given these potential biases in the underlying data, it is important to address the
obvious question: How well do EPA's national default values of DOC and POC represent
average conditions across the United States?
To address the question of sampling bias and its impact on the representativeness of
EPA's national default DOC and POC values, two types of comparisons were made with the
WATSTORE/LDC data. First, the national default values were compared with central-tendency
estimates of DOC and POC obtained from independent reviews of the relevant scientific
literature. This was done to provide a qualitative assessment of the comparability of national
default values to "expected" values based on literature accounts. The second comparison was
more quantitative in design and involved contrasting geographically distinct subsets of the
WATSTORE/LDC databases with geographically similar subsets of data produced by EPA's
Environmental Monitoring and Assessment Program (EMAP). Data contained in the EMAP
databases are sampled by using a stratified, random sampling design that minimizes the effect of
biases in sampling design on resulting statistical distributions of the data. Each of these
comparisons is described below.
Comparisons with Literature Data, Thurman (1985) reviewed the literature on DOC and
POC concentrations in surface waters throughout the world. The concentrations of DOC and
POC were found to vary in surface waters as a function of water body type, trophic status (lakes),
climate, watershed size and vegetation, and season of the year. Specifically, Thurman (1985)
reported that mean values of DOC in some pristine streams range from 1 to 3 mg/L and those in
rivers and lakes typically range from 2 to 10 mg/L. Ranges of DOC concentrations in estuaries are
reported to be highest at the limit of tidal rise (i.e., essentially equivalent to DOC in rivers) and
6-43
-------
lowest as dilution with seawater becomes complete (i.e., approximating 1 mg/L on average). For
swamps, marshes, and bogs, DOC concentrations are reported to range from 10 to 30 mg/L.
Concentrations of POC in lakes reportedly range from about 0.1 to 1.0 mg/L, whereas those in
small streams range from 0.1 to 0.3 mg/L, or about 10% of the DOC. Finally, POC concentrations
in rivers are reported to be about one-half the concentration of DOC (i.e., 1-5 mg/L), although
POC may equal DOC in the largest rivers during times of high discharge. Although the ranges of
DOC and POC concentrations reported by Thurman (1985) include surface waters found beyond
the United States, they also appear to be representative of U.S. surface waters, based on data
summarized by Thurman (1985) that were specific to the United States.
Despite the aforementioned limitations in EPA's DOC and POC databases with respect to
potential sampling bias, EPA's national default values of DOC (2.9 mg/L) and POC (0.5 mg/L)
compare favorably with the ranges of DOC and POC concentrations summarized by Thurman
(1985). This comparison suggests that EPA's national default values of DOC and POC are not
unreasonable in terms of representing typical organic carbon concentrations found in U.S. rivers,
streams, lakes, and estuaries. With respect to wetland areas (marshes, swamps, bogs), it is likely
that the national default values may significantly underestimate DOC and POC concentrations in
these systems, owing to their poor representation in the DOC and POC databases. The impact of
this underestimation on national BAFs will vary as a function of the Kow of the chemical (see
"Natural Variability in DOC and POC Concentrations," below). For some highly hydrophobic
organic chemicals, this underestimation may result in a conservative estimate of the AWQC for
these systems (i.e., a lower AWQC than what might be necessary) because of a likely
overestimation of the bioavailable fraction and national BAF.
Comparisons with EMAPData. Data generated by EPA's EMAP program are based on
a stratified, random sampling strategy that is specifically designed to minimize the influence of
sampling bias on the data and to enable statistically based extrapolations across geographic
regions (Herlihy et al., 2000). Currently, however, the EMAP databases contain DOC
measurements (but not POC measurements) and are limited to smaller geographic scales and
specific water body types. Thus, DOC data from EMAP's 1997-1998 sampling of mid-Atlantic
streams and rivers (http://www.epa.gov/emap/html/datal/surfwatr/data) were compared with
similar geographic subsets from the WATSTORE/LDC database. The mid-Atlantic EMAP
database was chosen because sufficient data were available on DOC in rivers and streams to
make meaningful comparisons at the State and ecoregion levels. Similarly, the mid-Atlantic
region is also well represented in the WATSTORE/LDC database.
Figure 6-4 shows the cumulative frequency distributions of DOC contained in the EMAP
mid-Atlantic database (top panel) and the WATSTORE/LDC database (bottom panel) for rivers
and streams in Pennsylvania, Virginia, and West Virginia. Similar comparisons are made for four
mid-Atlantic ecoregions (Piedmont, Ridge and Valley, Central Appalachians, Western Allegheny
Plateau; Figure 6-5). Descriptive statistics are provided in Tables 6-11 and 6-12. From both sets of
comparisons, it is apparent that the agreement between the WATSTORE/LDC and EMAP data is
best at the middle to lower tails of the distributions and poorest at the higher end of the
distributions. At the lower tails of the distributions (e.g., 10th, 25th percentiles), the
WATSTORE/LDC DOC data are generally within 30% of the EMAP data (ecoregion 70 being
6-44
-------
the only exception). The median DOC values of the WATSTORE/LDC data show a slightly
higher bias compared with median values from the EMAP data but are usually within a factor of
1.5 (ecoregions 47 and 70 are about a factor of 2 greater). For a majority of comparisons made at
the 75th and 90th percentiles (i.e., 11 of 14), the WATSTORE/LDC DOC values are approximately
a factor of 2 greater than the corresponding percentiles from the EMAP data. This result is
expected, given the greater focus of the WATSTORE/LDC sampling sites on larger river and
stream systems and on areas receiving proportionately greater human influence compared with
the EMAP sampling sites.
8. 1
o
£ 0.8
EMAP Data
J/j~ *- ~
DOC (mg/L)
•PA-
-VA-
• WV
WATSTORE & LDC Data
DOC (mg/L|
Figure 6-4. Ecoregion-level DOC distributions for rivers and streams from EPA's WATSTORE/LDC and EMAP
databases.
Source'. EMAP data were taken from EPA's Environmental Monitoring and Assessment Program, Mid-Atlantic
Integrated Assessment, 1997-98 (http://www.epa.gov/emap/html/datal/surfwatr/data).
USGS WATSTORE and EPA LDC retrievals are explained in Section 6.3.2.
6-45
-------
EMAP Data
DOC (mg/L|
-Eco Reg. 45 •
-Eco Reg. 67 •
• Eco Reg. 69 •
- Eco Reg. 70
WATSTORE & LDC Data
-45-
-67 •
•69-
-70
Figure 6-5. Ecoregion-level DOC distributions for rivers and streams from EPA's WATSTORE/LDC and EMAP
databases.
Source: EMAP data were taken from EPA's Environmental Monitoring and Assessment Program, Mid-Atlantic
Integrated Assessment, 1997-98 (http://www.epa.gov/emap/html/datal/surfwatr/data). Ecoregion45 = Piedmont; 67
= Ridge & Valley; 69 = Central Appalachians; 70 = Western Allegheny Plateau.
USGS WATSTORE and EPA LDC retrievals are explained in Section 6.3.2.
Table 6-11. Descriptive Statistics from the State-Level DOC Distributions
Statistic
n
Mean
10th percentile
25th percentile
50th percentile
75th percentile
90th percentile
PA
89
1.7
0.8
1.0
1.5
2.1
2.5
EMAP (1997-1998)"
VA
80
2.0
0.8
1.0
1.5
1.9
3.2
WV
59
2.3
1.0
1.5
1.8
2.2
3.7
WATSTORE
PA
1,359
3.8
0.9
1.3
2.2
4.6
8.9
&LDC
VA
634
3.1
0.7
1.0
1.8
3.7
6.5
(1980-1999)
WV
682
2.0
0.7
1.2
1.7
2.5
3.5
6-46
-------
Table 6-12. Descriptive Statistics from the Ecoregion-Level DOC Distributions
Statistic
n
Mean
10th percentile
25th percentile
50th percentile
75th percentile
90th percentile
45
38
2.6
1.1
1.6
1.8
3.1
4.0
EMAP MAIA (1997-1998)"
Eco region
67 69
64 36
1.7 1.7
0.8 0.7
1.1 1.0
1.3 1.5
2.0 1.8
2.7 2.1
70
43
2.1
<0.7
1.4
1.8
2.1
2.7
WATSTORE & LDC (1980-1999)
Eco region
45
309
4.4
1.0
1.7
3.4
5.9
9.3
67 69
733 864
2.2 1.9
0.6 0.7
1.0 1.1
1.7 1.6
2.8 2.3
4.5 3.3
70
1,795
4.0
1.6
2.7
4.1
5.0
6.1
1 Length-weighted statistics for EMAP data. Ecoregion 45 = Piedmont; 67 = Ridge & Valley;
69 = Central Appalachians; 70 = Western Allegheny Plateau.
Source: EMAP data were taken from EPA's Environmental Monitoring and Assessment Program, Mid-Atlantic Integrated
Assessment, 1997-98 (http://www.epa.gov/emap/html/datal/surfwafr/data/).
USGS WATSTORE and EPA LDC retrievals are explained in Section 6.3.2.
The previous comparisons of DOC concentrations from the mid-Atlantic EMAP and
WATSTORE/LDC databases are clearly limited with respect to evaluating the impact of possible
sampling bias on EPA's national default values of DOC and POC (i.e., comparisons are restricted
to the mid-Atlantic region and no comparisons could be made for POC). Despite these
limitations, this analysis indicates that, at least for the three States that are well represented in the
WATSTORE/LDC database, the degree of sampling bias at median values is not overly
exaggerated. Best agreement between the two databases occurred at percentiles at or below the
median values of the distributions. Assuming the EMAP data represent unbiased results, a
noticeable and somewhat expected bias appears in the WATSTORE/LDC data, primarily at the
higher percentiles. Results are mixed at the ecoregion level; two of the four DOC distributions
compare favorably between the two databases (defined here as percentile values within a factor
of 2). The greater discrepancy between DOC concentrations in ecoregions 45 and 70 appears to
be related to the disproportionate influence of several stations from which large numbers of
measurements were taken relative to the other stations.
Measurement Error. Other sources of uncertainty in EPA's national default values of
DOC and POC concentrations include error associated with measuring DOC and POC
concentrations. Measurement error refers to error associated with quantifying the particular
variable of interest (e.g., DOC) and includes error associated with sample collection and handling
and analytical techniques. Measurement error varies by analytical method, laboratory, and, to
some extent, each batch of samples analyzed. For the LDC data, the analytical methods used to
determine DOC and POC concentrations were not reported in the database. For analytical
methods underlying the WATSTORE data, estimates of accuracy (percent recovery) and
6-47
-------
precision (relative standard deviation) are available for the analysis of TOC and POC. The mean
percent recovery and relative standard deviation associated with TOC measurements with the wet
oxidation method (StandardMethods #5310D) are reported as 103 ± 3.4% (APHA, 1995).
Similarly, for TOC measurements using persulfate-ultraviolet oxidation (StandardMethods
#53 IOC), the reported mean percent recovery and relative standard deviation of measurements in
two matrices are 93 ± 7% and 106 ± 6% (APHA, 1995). Finally, the reported mean percent
recovery and standard deviation for POC measurements with the wet oxidation method with
silver filtration and sonification are 97 ± 11% (USGS, 1997; Burkhardt et al., 1999). Relative to
other sources of uncertainty in national default DOC and POC values, error associated with
analytical methods appears to be small, at least where it has been quantified.
Natural Variability in DOC and POC Concentrations. As one would expect, there is
substantial variability in the median values of DOC and POC concentrations in U.S. surface
waters (Table 6-10). Specifically, the range of 95th to 5th percentile estimates approximates or
exceeds a factor of 20 in several types of surface waters. Although measurement error is reflected
in this variability, the bulk of variability is believed to result from naturally occurring conditions
and processes that contribute to spatial and temporal variability in the delivery and
biogeochemical cycling of organic carbon in surface waters. Some of these factors include
climatology (e.g., arid, arctic, alpine, and tropical zonal differences) and trophic status (e.g.,
oligotrophic, mesotrophic, and distrophic lakes), discharge volume and source (for streams and
rivers), watershed size and landscape characteristics, season, and the extent of tidal influence (for
estuaries). To address uncertainty in BAFs resulting from this natural variability in DOC and POC
concentrations, EPA encourages States and authorized Tribes to use appropriate local or regional
data on the organic carbon content of applicable waters when adopting criteria into their own
water quality standards. Nevertheless, EPA recognizes that appropriate local or regional data will
not always be available in sufficient quantity or quality. Therefore, it is appropriate to explore the
degree to which variability in DOC and POC concentrations has an impact on national BAFs.
Figure 6-6 illustrates the effect of varying concentrations of DOC and POC on the freely
dissolved fraction for nonionic organic chemicals with various Kows. The freely dissolved fraction
was calculated according to Equation 5-12 of the 2000 Human Health Methodology (USEPA,
2000a) and has a 1:1 impact on the resulting national BAF (see Equation 5-28 of the 2000 Human
Health Methodology).
From an examination of Figure 6-6, several observations can be made regarding how
variability in organic carbon concentrations is predicted to affect the freely dissolved fraction (and
subsequently the national BAF) for nonionic organic chemicals. First, the effect of DOC and
POC concentrations on the freely dissolved fraction is highly dependent on Kow. For nonionic
organic chemicals with log Kow values of about 4 or less, changes in DOC and POC
concentrations within the 5th to 95thpercentiles have very little impact on the freely dissolved
fraction. Further analysis (not shown here) indicates that this insensitivity holds true for values of
DOC and POC far exceeding the 5th and 95th percentiles. Thus, uncertainty in the DOC or POC
concentrations has very little impact on the resulting national BAFs for low Kow chemicals.
6-48
-------
A second observation is that, for nonionic organic chemicals with higher hydrophobicity
(e.g., log Kow >4), the impact of DOC and POC on the freely dissolved fraction increases as Kow
increases. Although the absolute range in the freely dissolved fraction corresponding to the 5th
and 95th percentiles of DOC and POC peaks at a log Kow of about 6 (i.e., from 0.89 at the 5th
percentile to 0.17 at the 95th percentile), the relative difference in freely dissolved fraction (as
measured by the ratio of freely dissolved fraction at various DOC and POC percentiles) increases
with Kow. Because the freely dissolved fraction is used in a multiplication step to calculate the
national BAF, the relative differences in freely dissolved fraction are more meaningful for
interpreting the impact of variability of organic carbon concentrations on the national BAF.
Table 6-13 illustrates the effect of organic carbon on the differences in freely dissolved
fraction relative to that calculated with national default values of DOC and POC. Here, relative
differences are expressed as ratios of the freely dissolved fraction calculated at various percentiles
of DOC and POC (from Table 6-12) to that calculated at the national default values of DOC and
POC. For chemicals with a log Kow of 5.0, the relative impact of DOC and POC within these
percentiles is still rather minor (i.e., a 10% increase at the 5th percentile DOC and POC values
versus a 30% decrease at the 95th percentile). For chemicals with a log Kow value of 6.0, the impact
of organic carbon is more substantial, resulting in a 50% increase in the freely dissolved fraction
at the 5th percentile DOC and POC values. The freely dissolved fraction associated with the 95th
percentile DOC and POC values drops to 30% of the fraction calculated with the national default
DOC and POC values. The effect of lower DOC and POC concentrations on the freely dissolved
fraction is still somewhat muted compared with higher concentrations, in part because the freely
dissolved fraction calculated with the national default values of DOC and POC is still relatively
high (0.93 at Kow 5.0 and 0.58 at Kow 6.0), and it cannot increase beyond 1.0. The greatest impact
of organic carbon on the freely dissolved fraction is seen at the highest Kow (8.0), where the freely
dissolved fractions calculated at the 5th and 95th percentiles are similar in magnitude to the
changes in DOC and POC values (i.e., a fivefold increase in DOC concentration and an eightfold
increase in POC concentration results in an approximately sevenfold decrease in the freely
dissolved fraction).
A final observation is that for highly hydrophobic chemicals, the freely dissolved fraction
is most sensitive to changes in POC relative to DOC. This fact is clear from examination of
Equation 4-4 and relates to the higher partition coefficient for organic chemicals to POC
(Kpoc = 1.0 • Kow L/kg) as compared with that for DOC (Kdoc = 0.08 • Kow L/kg). Therefore, in
terms of reducing overall uncertainty associated with the application of national default values of
DOC and POC, resources should be directed toward site- or region-specific organic carbon
measurements for chemicals with higher hydrophobicity (e.g., about log Kow of 5 and above).
Although both DOC and POC measurements are needed, results indicate that particular attention
should be paid to quantifying representative measurements of POC.
6-49
-------
0.0
1 .OE+03
1 .OE+04
1 .OE+05
1 .OE+06
1 .OE+07
1 .OE+08
Kow
•- Nat'l Default (50th) - - - 5th (DOC & POC*) - - -A - - 10th (DOC & POC*)
•• - - 25th (DOC & POC) •••£>-- 75th (DOC & POC) - - -A - - 90th (DOC & POC)
4--95th (DOC & POC)
Figure 6-6. Effect of DOC, POC, and Kow on the freely dissolved fraction (ffd). The solid line (• ) is the freely
dissolved fraction that corresponds to EPA's national default values of DOC (2.9 mg/L) and POC (0.5 mg/L). The
dashed lines reflect various percentiles from the distributions of DOC and POC concentrations used to derive the
national default values (e.g., 5th, 10th, 25th, 75th, 90th, 95th percentiles from Table 6-9).
* Estimated value based on statistical parameters from the POC distribution (see Table 6-9), assuming data were log-normally
distributed.
Table 6-13. Effect of DOC and POC Concentrations on the Freely Dissolved Fraction (ffd)
Relative to National Default Values of DOC and POC
Percentile
50th (National
Default)
5th
10th
25th
75th
90th
95th
DOC
(mg/L)
2.9
0.8
1.2
2
5.4
9.7
14
POC
(mg/L)
0.5
0.06"
0.09"
0.2
1.1
2.3
3.9
Fraction
5.0
0.93
0.99 [1.1]
0.98 [1.1]
0.97 [1.0]
0.87 [0.9]
0.77 [0.8]
0.67 [0.7]
Freely Dissolved
6.0
0.58
0.89 [1.5]
0.84 [1.5]
0.74 [1.3]
0.40 [0.7]
0.25 [0.4]
0.17 [0.3]
(frd) and [Ratio to National
LogKoW
7.0
0.12
0.44 [3.7]
0.35 [2.9]
0.22 [1.8]
0.06 [0.5]
0.03 [0.25]
0.02 [0.16]
Default]
8.0
0.014
0.08 [5.5]
0.05 [3.8]
0.03 [2.0]
0.006 [0.5]
0.003 [0.24]
0.002 [0.15]
1 Estimated value based on statistical parameters from the POC distribution (see text).
6-50
-------
7. EXAMPLES OF BAF CALCULATIONS
The examples presented in this section illustrate how national BAFs used in calculation of
human health AWQC are developed for a chemical of interest (chemical i) with the four BAF
methods under Procedure 1. The general process illustrated here is also applicable to chemicals
under Procedures 2-4. The equations used in the examples given here are presented in Sections 4
and 5 and the terms used in the equations are defined in Section 2. For reference, the equation
numbers provided here refer to the section where the equation was initially presented or derived.
7.1 EXAMPLE 1: CALCULATION OF A NATIONAL BAF FROM A FIELD-
MEASURED BAF (BAF j) (METHOD 1)
This example illustrates the development of a national trophic level 4 BAF using
method 1 for a hydrophobic nonionic chemical (chemical i). Calculating national BAFs using
method 1 requires the use of a EAF{ (also commonly referred to as a "field-measured" BAF).
Determination of a BAFj requires information on the total concentration of chemical i in fish
tissue and the total concentration of chemical i in the ambient water.
7.1.1 Calculating a Total BAF (BAFj)
In this example, data are available from Lake John Doe (a hypothetical lake) on the total
concentration of chemical i in lake trout (100 ng/kg) and in the water column (1.6xl08 |ig/L). A
review of the dietary preferences of the larger sizes of lake trout that are commonly consumed by
the general U.S. population confirms that these organisms belong to trophic level 4 (USEPA,
2000e-g). Data obtained from field studies indicate that the mean concentration of the chemical in
the water column reflects adequate temporal and spatial averaging, based on the Kow of this
chemical, and is representative of the average exposure of chemical i to the target fish. The EAF{
calculated for chemical i is 6.2x 10s L/kg, as shown below.
BAFTl = '"Wig = 6.2xi05 L/kg wet
(See Equation 2-1)
7.1.2 Calculating a Baseline BAF
The BAFj is converted to a baseline BAF for a specific trophic level by incorporating site-
specific information on the fraction of the chemical that is freely dissolved in the ambient water
(ffd) and the fraction of tissue or aquatic organism sampled that is lipid (f.). The equation for
calculating a baseline BAF from EAF{ is:
7-1
-------
Baseline BAF =
BAFTl
- 1
(See Equation 5-2)
Determining the fraction of chemical i that is freely dissolved (ffd) in the ambient water
requires information on the POC and DOC in the ambient water where the samples were
collected and the Kow of chemical i. For this example, the median POC concentration from Lake
John Doe is 0.6 mg/L (6.0x 10'7 kg/L) and the median DOC concentration is 8.0 mg/L (8.Ox 10'6
kg/L). It is important that the POC and DOC concentrations used in calculating the freely
dissolved fraction for baseline BAFs be determined from the water body used in the BAF study.
It is not appropriate to use national default POC and DOC concentrations to derive baseline
BAFs from BAF^s. The Kow for chemical i is 1.0* 105, or a log Kow of 5. Based on these data, the
fraction of chemical i that is freely dissolved in water is 0.89, calculated as shown below:
fM = i = 0.89
[1 + (6.0X10'7 kg/L • IxlO5 L/kg) + (8.0xlO-6 kg/L • 0.08 • IxlO5 L/kg>]
(See Equation 4-6)
The mean f. of the fish species sampled in Lake John Doe is 0.08 (8%). Using this f. and
the BAFx and ffd calculated above, a baseline BAF for lake trout of 8.7* 106 L/kg of lipid is
calculated as follows:
Baseline BAF =
6.2xlOs
- 1 • — = 8.7xl06 L/kg of lipid
0.89 0.08
(See Equation 5-2)
For the purposes of this example, it is assumed that only one acceptable BAF value is
available for trophic level 4 organisms. Thus, the baseline BAF for trophic level 4 is the baseline
BAF for lake trout. If acceptable BAF^s are available for additional trophic level 4 organisms,
baseline BAFs are calculated for each of the trophic level 4 species for which there is acceptable
data and then the baseline BAF for trophic level 4 is calculated as the geometric mean of these
baseline BAFs. Recall that in EPA's BAF methodology, BAFs are trophic-level specific. Hence,
this calculation would be carried out similarly for each trophic level.
7-2
-------
7.1.3 Calculating a National BAF
After deriving all acceptable baseline BAFs for chemical i and selecting a final baseline
BAF for trophic level 4, the next step is to calculate a national BAF for this trophic level that will
be used to derive the AWQC. In this example, it is assumed that the baseline BAF calculated
above represents the final baseline BAF for trophic level 4. For a given trophic level, calculating a
national BAF involves adjusting the final baseline BAF to reflect conditions that are expected to
affect the bioavailability of chemical i in ambient waters of the United States. This is
accomplished through the use of national default values for f. and ffd that are based on national
central tendency estimates. For each trophic level, the general equation for deriving a national
BAF is:
National BAFTLn = [(Final Baseline BAF)TLn - (QTLn + 1]
(See Equation 3-2)
For the purposes of this example, a national BAF is calculated for aquatic organisms at only one
trophic level (trophic level 4). In the 2000 Human Health Methodology, EPA divided the default
fish intake rate into trophic-level 2-, 3-, and 4-specific rates. Hence, in the process of deriving
national AWQC, EPA will derive national BAFs for trophic levels 2, 3, and 4.
For chemical i, the baseline BAF at trophic level 4 was calculated to be 8.7* 106 L/kg of
lipid. The freely dissolved fraction of chemical i that is estimated for all water bodies in the United
States is calculated by using Equation 4-6 and the national default values of 5 * 10"7 kg/L for POC
and 2.9x 10'6 kg/L for DOC (Section 6.3), and the Kow of chemical i which is l.Ox 10s (log Kow of
5). A value of 0.93 is calculated as shown below:
fM = i = 0.93
[1 + (S.OxlO'7 kg/L • IxlO5 L/kg) + (2.9xlo-6 kg/L • 0.08 • IxlO5 L/kg>]
(See Equation 4-6)
The national default f. for trophic level 4 is 0.03 (3%; see Section 6.2). Using the ffd calculated
above and the national default f. in the national BAF equation, the national BAF for trophic
level 4 organisms is calculated to be 2.4x 10s L/kg, as shown below:
National BAF for Trophic Level 4
[(8.7xl06 L/kg of lipid) • (0.03 kg of lipid per kg of tissue) + 1] • 0.93
2.4x 105 L/kg of wet tissue
7-3
-------
This national BAF relates the total concentration of chemical in water to the total concentration
of chemical in tissue of trophic level 4 organisms.
7.2 EXAMPLE 2: CALCULATION OF A NATIONAL BAF FROM FIELD-
MEASURED BSAFs (METHOD 2)
This example illustrates the development of a national trophic level 4 BAF using method 2
for PCB congener 126. Calculating national BAFs using method 2 requires the use of field-
measured BSAFs for reference chemicals and the chemical of interest. In section 5.2.4, it is
suggested that multiple reference chemicals be used to calculate a baseline BAF with method 2
because this results in a more accurate baseline BAF prediction (see Section 5.2.4).
In this example, data from Lake Ontario are used to derive baseline BAFs from BSAFs
for chemicals like PCB 126, which cannot be readily detected in water (USEPA, 1995b; Cook and
Burkhard, 1998). To simplify this example, a baseline BAF is derived for only one trophic level 4
organism, that is, age 5-7 lake trout. A review of the dietary preferences of the larger sizes of lake
trout that are commonly consumed by the general U.S. population confirms that these organisms
belong to trophic level 4. Previously, the PCB congeners 52, 105, and 118 have been used as the
reference chemicals for calculating baseline BAFs for PCB 126 (USEPA, 1995b; Cook and
Burkhard, 1998). These three congeners were selected because (1) they have similar
physicochemical properties, (2) they are well quantified in sediment and biota, and
(3) available data indicate they have loading histories similar to PCB 126 and thus their
(* Socw)r/(Kow)r values should be similar. In this example, the detailed, step-by-step calculations for
each component of the equation are shown only for reference PCB congener 118. In practice, the
same steps are performed for all reference congeners, but for this example, only the final baseline
BAFs are shown for PCBs 52 and 105.
7.2.1 Calculating a Field-Measured BSAF
The BSAF for PCB 126 is determined by relating lipid-normalized concentrations of the
chemical in 5 to 7-year-old lake trout (C.) to the average organic carbon-normalized concentration
of the chemical in surface sediment (Csoc), using equation 5-2. On the basis of data collected from
Lake Ontario, the C. of PCB 126 in age 5-7 lake trout is 12.3 ng/g of lipid, and the Csoc of PCB
126 in the sediment is 3.83 ng/g of organic carbon (actual calculations for these normalized values
are not shown here). Therefore:
BSAF = 12.3 ng/g lipid =
126 3.8 ng/g soc
(See Equation 5-3)
7-4
-------
7.2.2 Determining a Sediment-Water Column Concentration Quotient (• socw)r
Sediment-water column concentration quotients for reference chemicals are determined
by using equation 5-6. For this calculation, the concentration of reference chemical that is freely
dissolved in water, (C")PCB m, is needed. To calculate the (C")PCB m , the fraction of reference
chemical that is freely dissolved in water (ffd)r is needed. The (ffd)r is calculated by using equation
4-6. The measured DOC value is 2.Ox 10"6 kg/L; POC is set equal to zero (because all particulates
were removed using a filter); and the Kow for PCB 118 = 5.5* 106 (log Kow of 6.7). Using Equation
4-6, the fraction of PCB 118 that is freely dissolved in water is calculated as follows:
(f«W 11* = = 0.53
" ' 5.5X106 L/kg) + (2.0xlO-6 kg/L • 0.08 • S.SxlO6 L/kg)]
(See Equation 4-6)
For this example, the measured concentration of reference congener PCB 1 18 in filtered
Lake Ontario water is 34 pg/L. Thus, (C ™)PCB us = 34 pg/L x 0.53 = 18 pg/L or 1 .8x 10'5 |ig/L . The
average (Csoc)pCBii8 is 555 ng/kg of sediment organic carbon. By substituting these values into
Equation 5-6, • socw for the reference chemical, PCB 1 18, is calculated as:
1.8x10 5 jag/L
(See Equation 5-6)
7.2.3 Calculating a Baseline BAF
For each species having an acceptable field-measured (BSAF);, a baseline BAF for the
chemical of interest may be calculated with the following equation and an appropriate value of
\ socw/r ' v^-ow/i"
(Baseline BAfy = (BSAF)i •
(See Equation 5-11)
By using a commonly valid assumption — that Dj/r ~ 1 for PCB congeners 118 and 126;
substituting the B SAP for PCB 126(3.2),' socwforPCB 118 (3.1 xlO7), the appropriate Kow values
for PCB 126 (7.8x 106 or log Kow = 6.9) and PCB 118 (5.5x 106 or log Kow = 6.7), and 0.20 (20%)
fraction of lipid for lake trout into the baseline BAF equation (Equation 5-11), the baseline BAF
for PCB 126 may be calculated as:
7-5
-------
Baseline BAF15S = 3.2 • u; ' l->.i*i" ; ' i/.»*m ; _ J_ = lAxlQ* ^
5.5x10* -20
(See Equation 5-11)
The baseline BAFs using reference PCB congeners 52 and 105 are derived in the same
manner as for PCB 118. The predicted baseline BAFs that result are 3.7x 108 using congener 52
and 1.6xl08 using congener 105. Once all the baseline BAFs have been derived, the final baseline
BAF is derived by calculating the geometric mean of the three baseline BAFs, which in this case
is2.0x!08L/kg.
7.2.4 Calculating a National BAF
After deriving all acceptable baseline BAFs for chemical i and selecting a final baseline
BAF for trophic level 4, the next step is to calculate a national BAF for this trophic level that will
be used to derive the AWQC. In this example, it is assumed that the baseline BAF calculated
above represents the final baseline BAF for trophic level 4. For a given trophic level, calculating a
national BAF involves adjusting the final baseline BAF to reflect conditions that are expected to
affect the bioavailability of chemical i in ambient waters of the United States. This is
accomplished through the use of national default values for f. and ffd that are based on national
central tendency estimates. For each trophic level, the general equation for deriving a national
BAF is:
National BAFTLn = [(Final Baseline BAF)TLn - (QTLn + 1]
(See Equation 3-2)
For the purposes of this example, a national BAF is calculated only for aquatic organisms at one
trophic level (trophic level 4). In the 2000 Human Health Methodology, EPA divided the default
fish intake rate into trophic-level 2-, 3-, and 4-specific rates. Hence, in the process of deriving
national AWQC, EPA will derive national BAFs for trophic levels 2, 3, and 4.
For PCB 126, the baseline BAF at trophic level 4 was calculated to be 2.Ox 108 L/kg of
lipid. The freely dissolved fraction of PCB 126 that is estimated to be applicable to all water
bodies in the United States is calculated by using Equation 4-6 and the national default values of
5x ID'7 kg/L for POC and 2.9x 10'6 kg/L for DOC (Section 6.3), and the Kow of PCB 126 which is
7.8x 106, or a log Kow = 6.9. A value of 0.15 is calculated as shown below:
7-6
-------
t. = - - - = 0.15
[1 + (5.&X1&-1 kg/L • y.Kxll^ L/kg) + CZ.&X10-15 kg/L • 0.08 • 7.8x10?
(See Equation 4-6)
The national default f. for trophic level 4 is 0.03 (3%; see Section 6.2). Using the ffd calculated
above and the national default f. in the national BAF equation, the national BAF for trophic
level 4 organisms is calculated to be 9.Ox 10s L/kg, as shown below:
National BAF for Trophic Level 4
[(2.0xl08 L/kg of lipid) • (0.03 kg of lipid per kg of tissue) +1] • (0.15)
9.0x 105 L/kg of wet tissue
This example of a national BAF for PCB 126 relates the total concentration of chemical in water
to the total concentration of chemical in tissue of trophic level 4 organisms.
7.3 EXAMPLE 3: CALCULATION OF A NATIONAL BAF FOR CHEMICAL i
FROM BCFj x FCM (METHOD 3)
This example illustrates the calculation of a national trophic level 4 BAF using method 3
for a hydrophobic nonionic chemical (chemical i). Calculating national BAFs using method 3
requires the use of a BCFj (also commonly referred to as a "laboratory-measured BCFV) and a
FCM. Determination of a BCFj requires information on the total concentration of chemical i in
fish tissue and the total concentration of chemical i in the laboratory test water.
7.3.1 Calculating a Laboratory-Measured BCF^
In this example, data are available from John Doe's laboratory (a hypothetical laboratory)
on the total concentration of chemical i in fish tissue (10 |ig/kg) and the laboratory test water
(3.0xl03 |ig/L). The laboratory-measured BCF calculated for chemical i is 3.3xl03 L/kg, as shown
below:
BCFt = IQng/kg 3.3xl03 L/kg wet tissue
S.OxlO'3 jig/L
(See Equation 2-8)
7.3.2 Calculating a Baseline BAF
The BCFj is converted to a baseline BAF for a specific trophic level by incorporating site-
specific information on the fraction of the chemical that is freely dissolved in the test water (ffd),
the fraction of tissue or aquatic organism tested that is lipid (f.), and a food-chain multiplier
(FCM) for the chemical. The equation for calculating a baseline BAF from BCFj is:
7-7
-------
Baseline BAF = FCM
(See Equation 5-12)
Determining the fraction of chemical i that is freely dissolved in the test water (ffd) requires
information on the POC and DOC in the test water and the Kow of chemical i. For this example,
the median POC concentration in the test water is 0.6 mg/L (6.Ox 10"7 kg/L) and the median DOC
concentration is 8.0 mg/L (8.Ox 10"6 kg/L). It is important that the POC and DOC concentrations
used in calculating the freely dissolved fraction for baseline BAFs be determined from the water
used in the BCF study. It is not appropriate to use national default POC and DOC concentrations
to derive baseline BAFs from BCFxS. The Kow for chemical i is 1 x 104, or a log Kow of 4.0. Based
on these data, the fraction of chemical i that is freely dissolved is 0.99, calculated as shown
below:
fM = 1 = 0.99
[1 + (6.0X10'7 kg/L • IxlO4 L/kg) + (8.0xlO-6 kg/L • 0.08 • IxlO4 L/kg>]
(See Equation 4-6)
The f. of the fish species sampled in the laboratory in this example is 0.08 (8%). The
FCM, based on a log Kow of 4, is 1.07, as indicated in Table 4-6 (assuming a mixed benthic and
pelagic food web structure and trophic level 4 for the tested species). Using this f. and FCM with
the BCFj and ffd calculated above, a baseline BAF of 4.5* 104 L/kg of lipid is calculated as
follows:
Baseline BAF = 1.07
3.3xlQ3
- 1
1
= 4.5xl04 L/kg of lipid
0.08
(See Equation 5-12)
For the purposes of this example, it is assumed that only one acceptable BCF study is
available for trophic level 4 organisms. Thus, the baseline BAF value for trophic level 4 is the
baseline BAF for the tested organism. If acceptable BCFxS are available for additional trophic
level 4 organisms, baseline BAFs are calculated for each of the trophic level 4 species for which
there are acceptable data and then the baseline BAF for trophic level 4 is calculated as the
geometric mean of these baseline BAFs. Recall that in EPA's BAF methodology, BAFs are
trophic-level specific. Hence, this calculation would be carried out similarly for each trophic level.
7.3.3 Calculating a National BAF
After deriving all acceptable baseline BAFs for chemical i and selecting a final baseline
BAF for trophic level 4, the next step is to calculate a national BAF for this trophic level that will
be used to derive the AWQC. In this example, it is assumed that the baseline BAF calculated
above represents the final baseline BAF for trophic level 4. For a given trophic level, calculating a
7-8
-------
national BAF involves adjusting the final baseline BAF to reflect conditions that are expected to
affect the bioavailability of chemical i in ambient waters of the United States. This is
accomplished through the use of national default values for f. and ffd that are based on national
central tendency estimates. For each trophic level, the general equation for deriving a national
BAF is:
National BAFTLn = [(Final Baseline BAF)^ - (Q^ + 1] • fc
(See Equation 3-2)
For the purposes of this example, a national BAF is calculated for aquatic organisms at
only one trophic level (trophic level 4). In the 2000 Human Health Methodology, EPA divided the
default fish intake rate into trophic-level 2-, 3-, and 4-specific rates. Hence, in the process of
deriving national AWQC, EPA will derive national BAFs for trophic levels 2, 3, and 4.
For chemical i, the baseline BAF at trophic level 4 was calculated to be 4.5* 104 L/kg of
lipid. The freely dissolved fraction of chemical i that is estimated for all water bodies in the United
States is calculated by using Equation 4-6 and the national default values of 5 * 10"7 kg/L for POC
and 2.9x 10'6 kg/L for DOC (Section 6.3), and the Kow of chemical i which is 1 x 104 (a log Kow of
4.0). A value of 0.99 is calculated as shown below:
f« =
[ 1 + (S.OxlO'7 kg/L • IxlO4 L/kg) + (2.9xlO-6 kg/L • 0.08 • IxlO4 L/kg)]
(See Equation 4-6)
The national default f. for trophic level 4 is 0.03 (3%; see Section 6.2). Using the ffd calculated
above and the national default f. in the national BAF equation, the national BAF for trophic
level 4 organisms is calculated to be 1.3 x 103 L/kg, as shown below:
National BAF for Trophic Level 4
[(4.5xl04 L/kg of lipid) • (0.03 kg of lipid per kg of tissue) + 1] • (0.99)
1.3 x 103 L/kg of wet ti ssue
This national BAF relates the total concentration of chemical in water to the total concentration of
chemical in tissue of trophic level 4 organisms.
7-9
-------
7.4 EXAMPLE 4: CALCULATION OF A NATIONAL BAF FOR CHEMICAL i
FROM Kow x FCM (METHOD 4)
This example illustrates the development of a national trophic level 4 BAF using
method 4 for a hydrophobic nonionic chemical (chemical i). Calculating national BAFs using
method 4 does not require knowing the fraction of the chemical that is freely dissolved in the test
water (ffd) or the fraction of the species sampled that is lipid (f.). This is because the Kow is
assumed to be equal to the baseline BCF, as discussed in Section 5.4 and Appendix A. Method 4
requires selection of an appropriate Kow for the chemical and that the Kow be multiplied by an
appropriate FCM to account for biomagnification.
7.4.1 Selecting a Kow and FCM
The procedures that EPA will follow in selecting chemical Kows are described in detail in
Appendix B. For the purposes of this example, a Kow value of 1 x 104 (log Kow= 4.0) has been
selected for chemical i. The FCM, based on a log Kow of 4, is 1.07, as indicated in Table 4-6
(assuming a mixed benthic and pelagic food web structure and trophic level 4 for the tested
species).
7.4.2 Calculating a Baseline BAF
Method 4 does not require adjusting a field- or laboratory-derived bioaccumulation factor
with ffd or f.. The calculation of a baseline BAF, using the selected Kow and FCM, is
straightforward, as shown below:
Baseline BAF = Kow x FCM (See Equation 5-13)
= (lxl04)xl.Q7
= l.lx!04L/kg of lipid
For this example, only one Kow is provided. As discussed in Appendix B, it is possible
that several Kow values may be located. The data quality considerations provided in Appendix B
will be used for judging the quality of various Kow values, and the procedures outlined in
Appendix B will be used for selecting among or combining Kow values of acceptable quality.
Recall that in EPA's BAF methodology, BAFs are trophic-level specific. Hence, this calculation
would be carried out similarly, with appropriate FCMs, for each trophic level.
7.4.3 Calculating a National BAF
After deriving all acceptable baseline BAFs for chemical i and selecting a final baseline
BAF for trophic level 4, the next step is to calculate a national BAF for this trophic level that will
be used to derive the AWQC. In this example, it is assumed that the baseline BAF calculated
above represents the final baseline BAF for trophic level 4. For a given trophic level, calculating a
national BAF involves adjusting the final baseline BAF to reflect conditions that are expected to
affect the bioavailability of chemical i in ambient waters of the United States. This is
accomplished through the use of national default values for f. and ffd that are based on national
7-10
-------
central tendency estimates. For each trophic level, the general equation for deriving a national
BAFis:
National BAFTLn = [(Final Baseline BAF)^ - (Q^ + 1] • fc
(See Equation 3-2)
For the purposes of this example, a national BAF is calculated for aquatic organisms at
only one trophic level (trophic level 4). In the 2000 Human Health Methodology, EPA divided the
default fish intake rate into trophic-level 2-, 3-, and 4-specific rates. Hence, in the process of
deriving national AWQC, EPA will derive national BAFs for trophic levels 2, 3, and 4.
For chemical i, the baseline BAF at trophic level 4 was calculated to be 1.1 * 104 L/kg of
lipid. The freely dissolved fraction of chemical i that is estimated to be applicable to all water
bodies in the United States is calculated by using Equation 4-6 and the national default values of
5x ID'7 kg/L for POC and 2.9x 10'6 kg/L for DOC (Section 6.3), and the Kow of chemical i which is
1 x 104 (log Kow = 4.0). A value of 0.99 is calculated as shown below:
f« = = 0.99
[1 + (S.OxlO'7 kg/L • IxlO4 L/kg) + (2.9xlo-6 kg/L • 0.08 • IxlO4 L/kg>]
(See Equation 4-4)
The national default f. for trophic level 4 is 0.03 (3%; see Section 6.2). Using the ffd calculated
above and the national default f. in the national BAF equation, the national BAF for trophic
level 4 organisms is calculated to be 1,344 L/kg, as shown below:
National BAF for Trophic Level 4
[(l.lxlO4 L/kg of lipid) • (0.03 kg of lipid per kg of tissue) + 1] • (0.99)
3.3 x 104 L/kg of wet ti ssue
This national BAF relates the total concentration of chemical in water to the total concentration
of chemical in tissue of trophic level 4 organisms.
7-11
-------
8. REFERENCES
Abbott JD, Hinton SW, Borton DL. 1995. Pilot scale validation of the RIVER/FISH
bioaccumulation modeling program for nonpolar hydrophobic organic compounds using the
model compounds 2,3,7,8-TCDD and 2,3,7,8-TCDF. Environ Toxicol Chem 14:1999-2012.
Amrhein JF, Stow CA, Wible C. 1999. Whole-fish versus filet polychlorinated-biphenyl
concentrations: An analysis using classification and regression tree models. Environ Toxicol
Chem 18:1817-1823.
American Society of Testing and Materials. 1990. Standard Practice for Conducting
Bioconcentration Tests with Fishes and Saltwater Bivalve Molluscs. Designation E 1022-84. In:
Annual Book of ASTM Standards. Section 11, Water and Environmental Technology.
11(04):606-6622. West Conshohocken, PA.
APHA. 1995. Standard Methods for the Examination of'Water andWastewater. 19thed.
American Public Health Association, American Water Works Association, Water Environment
Federation, Washington, DC.
Baird D, Ulanowicz RE. 1989. The seasonal dynamics of the Chesapeake Bay ecosystem. Ecol
Monogr 59:329-364.
Barber MC. 2000. Bioaccumulation and Aquatic System Simulator (BASS) User's Manual,
Beta Test Version 2.1. Ecosystems Research Division, U.S. Environmental Protection Agency,
Athens, GA.
Barber MG, Suarez LA, Lassiter RR. 1991. Modeling bioaccumulation of organic pollutants in
fish with an application to PCBs in Lake Ontario salmonids. Can JFish Aquat Sci 48:318-337.
Baron MG. 1990. Bioconcentration. Environ Sci Technol 24:1612-1618.
Bergen BJ, Nelson WG, Quinn JG, Jayaraman S. 2001. Relationships among total lipid, lipid
classes, and poly chlorinated biphenyl concentrations in two indigenous populations of ribbed
mussels (Geukensia demissa) over an annual cycle. Environ Toxicol Chem 20:575-581.
Black MC, McCarthy JF. 1988. Dissolved organic macro molecules reduce the uptake of
hydrophobic organic contaminants by the gills of rainbow trout (Salmo gairdneri). Environ
Toxicol Chem 7:593-600.
Bligh EG, Dyer WJ. 1959. A rapid method of total lipid extraction and purification. CanJ
Biochem Physiol 37:911-917.
8-1
-------
BromanD, CarinaN, Rolff C, Zebuhr Y. 1991. Occurrence and dynamics of poly chlorinated
dibenzo-p-dioxins and dibenzofurans and polycyclic aromatic hydrocarbons in the mix surface
layer of remote coastal and offshore waters of the Baltic. Environ Sci Technol 25:1850-1864.
Burkhard LP. 1998. Comparison of two models for predicting bioaccumulation of hydrophobic
organic chemicals in aquatic food webs. Environ Toxicol Chem 17:383-393.
Burkhard LP. 2000. Estimating dissolved organic carbon partition coefficients for nonionic
organic chemicals. Environ Sci Technol 34:4663-4668.
Burkhard LP. 2003. Factors influencing the design of BAF and BSAF field studies. Environ
Toxicol Chem 22:351-360.
Burkhard LP, Sheedy BR, McCauley DJ, DeGraeve GM. 1997. Bioaccumulation factors for
chlorinated benzene, chlorinated butadienes and hexachloroethane. Environ Toxicol Chem
16:1677-1686.
Burkhard LP, Endicott DD, Cook PM, Sappington KG, Winchester EL. 2003a. Evaluation of
two methods for prediction of bioaccumulation factors. Environ Sci Technol 37:4626-4634.
Burkhard LP, Cook PM, Mount DR. 2003b. The relationship of bioaccumulative chemicals in
water and sediment to residues in fish: A visualization approach. Environ Toxicol Chem 22:351-
360.
Burkhardt MR, Brenton RW, Kammer JA, Jha VK, O'Mara-Lopez PG. 1999. Improved method
for the determination of nonpurgeable suspended organic carbon in natural water by silver filter
filtration, wet chemical oxidation, and infrared spectrometry. Water Res 35:329-334.
Butcher JB, Garvey EA, Bierman VJ Jr. 1998. Equilibrium partitioning of PCB congeners in the
water column: Field measurements from the Hudson River. Chemosphere 15:3149-3166.
Campfens J, Mackay D. 1997. Fugacity-based model of PCB bioaccumulation in complex
aquatic food webs. Environ Sci Technol 31:577-583.
Carlberg GE, Martinsen K, Kringstad A, Gjessing E, Grande M, Kallqvist T, Skare JU. 1986.
Influence of aquatic humus on the bioavailability of chlorinated micropollutants in Atlantic
salmon. Arch Environ Contain Toxicol 15:543-548.
Carter CW, Suffet IH. 1982. Binding of DDT to dissolved humic materials. Environ Sci
Technol 16:735-740.
Chefurka W, Gnidec EPP. 1987. Binding of [14C]DDT by submitochondrial particles. Comp
Biochem Physiol 88C:213-217.
8-2
-------
Chen W, Kan AT, Fu G, Vignona LC, Tomson MB. 1999. Adsorption-desorption behaviors of
hydrophobic organic compounds in sediments of Lake Charles, Louisiana, USA. Environ
ToxicolChem 18:1610-1616.
Chin Y, Gschwend PM. 1992. Partitioning of poly cyclic aromatic hydrocarbons to marine
porewater organic colloids. Environ Sci Technol 26:1621-1626.
Chiou GT. 1985. Partition coefficients of organic compounds in lipid-water systems and
correlation with fish bioconcentration factors. Environ Sci Technol 19:57-62.
Connolly JP. 1991. Application of a food chain model to poly chlorinated biphenyl
contamination of the lobster and winter flounder food chains in New Bedford Harbor. Environ
Sci Technol 25:760-770.
Connolly J, Pedersen C. 1988. A thermodynamic-based evaluation of organic chemical
accumulation in aquatic organisms. Environ Sci Technol 22:99-103.
Cook PM, Burkhard LP. 1998. Development of bioaccumulation factors for protection offish
and wildlife in the Great Lakes. In: U.S. Environmental Protection Agency. National Sediment
Bioaccumulation Conference Proceedings. Office of Water, Washington, DC. EPA/823/R-
98/002.
Dean KE, Shafer MM, Armstrong DE. 1993. Particle-mediated transport and fate of a
hydrophobic organic contaminant in southern Lake Michigan: The role of major water column
particle species. J Great Lakes Res 19:480-496.
de Boer J. 1988. Chlorobiphenyls in bound and non-bound lipids of fishes; comparison of
different extraction methods. Chemosphere 17:1803-1810.
de Wolf W, de Bruijn JHM, Seinen W, Hermens JLM. 1992. Influence of biotransformation on
the relationship between bioconcentration factors and octanol-water partition coefficients.
Environ Sci Technol 26:1197-1201.
DiToro DM, Zarba CS, Hansen DJ, Berry WJ, Swartz RC, Cowan CE, Pavlou SP, Allen HE,
Thomas NA, Paquin PR. 1991. Technical basis for establishing sediment quality criteria for
nonionic organic chemicals using equilibrium partitioning. Environ Toxicol Chem 10:1541-1583.
Eadie BJ, Morehead NR, Landrum PF. 1990. Three-phase partitioning of hydrophobic organic
compounds in Great Lake waters. Chemosphere 20:161-178.
Eadie BJ, Morehead NR, Val Klump J, Landrum PF. 1992. Distribution of hydrophobic organic
compounds between dissolved and particulate organic matter in Green Bay waters. J Great
Lakes Res 18:91-97.
8-3
-------
Endicott DD, Cook PM. 1994. Modeling the partitioning and bioaccumulation of TCDD and
other hydrophobic organic chemicals in Lake Ontario. Chemosphere 28:75-87.
Evans HE. 1988. The binding of three PCB congeners to dissolved organic carbon in
freshwaters. Chemosphere 12:2325-2338.
Ewald G, Bremle G, Karlsson A. 1998. Differences between Bligh and Dyer and Soxhlet
extractions of PCBs and lipids from fat and lean fish muscle: Implications for data evaluation.
Marine PollutBull 36:222-230.
Ewald G, Larsson P. 1994. Partitioning of 14C-labelled 2,2',4,4'-tetrachlorobiphenyl between
water and fish lipids. Environ Toxicol Chem 13:1577-1580.
Exler J. 1987. Composition of Foods: Finfish and Shellfish Products; Raw, Processed, Prepared.
U.S. Department of Agriculture, Human Nutrition Information Service. Agriculture Handbook
Number 8-15. (available at: http://www.nal.usda.gov/fnic/foodcomp/data/srl4/srl4.html.)
Fisk AT, Norstron RJ, Cymbalisty CC, Muir DCB. 1998. Dietary accumulation and depuration
of hydrophobic organochlorines: Bioaccumulation parameters and their relationship with the
octanol/water partition coefficient. Environ Toxicol Chem 17:951-961.
Flint R. 1986. Hypothesized carbon flow through the deep water Lake Ontario food web. J
Great Lakes Res 12:344-354.
Gauther TD, Shane EC, Guerin WF, Seitz WR, Grant CL. 1986. Fluorescence quenching
method for determining equilibrium constants for polycyclic aromatic hydrocarbons binding to
dissolved humic materials. Environ Sci Technol 20:1162-1166.
Gobas FAPC. 1993. A model for predicting the bioaccumulation of hydrophobic organic
chemicals in aquatic food-webs: Application to Lake Ontario. EcolMod 69:1-17'.
Gobas FAPC, Pasternak JP, Lien K, Duncan RK. 1998. Development and field-validation of a
multi-media exposure assessment model for waste load allocation in aquatic ecosystems:
Application to TCDD and TCDF in the Fraser River Watershed. Environ Sci Technol
32:2442-2449.
Gschwend PM, Wu S-C. 1985. On the constancy of sediment-water partition coefficients of
hydrophobic organic pollutants. Environ Sci Technol 19:90-96.
Gustafsson O, Haghseta F, Chan C, MacFarlane J, Gschwend PM. 1997. Quantification of the
dilute sedimentary soot-phase: Implications for PAH speciation and bioavailability. Environ Sci
Technol 31:203-209.
8-4
-------
Haitzer M, Hoss S, Traunspurger W, Steinberg C. 1998. Effects of dissolved organic matter
(DOM) on the bioconcentration of organic chemicals in aquatic organisms: A review.
Chemosphere 37:1335-1362.
Haitzer M, Akkanen J, Steinberg C, Kukkonen JVK. 2001. No enhancement in bioconcentration
of organic contaminants by low levels of DOM. Chemosphere 44:165-171.
Hamelink JL, Landrum PF, Bergman HL, Benson WH, eds. 1994. Bioavailability: Physical,
Chemical and Biological Interactions. Lewis Publishers, Boca Raton, FL. pp. 73-170.
Kara A, Radin N. 1978. Lipid extraction of tissues with a low-toxicity solvent. Anal Biochem
90:420-426.
Hassett JP, Anderson MA. 1979. Association of hydrophobic organic compounds with
dissolved organic matter in aquatic systems. Environ Sci Technol 13:1526-1529.
Havens K. 1992. Scale and structure in natural food webs. Science 257:1107-1109.
Hebert CE, Keenleyside KA. 1995. To normalize or not to normalize? Fat is the question.
Environ Toxicol Chem 14:801-808.
Henderson RJ, Tocher DR. 1987. The lipid composition and biochemistry of freshwater fish.
Prog Lipid Res 26:281-347.
Herbert BE, Bertsch PM, Novak JM. 1993. Pyrene sorption by water-soluble organic carbon.
Environ Sci Technol 27:398-403.
Herlihy AT, Larsen DP, Paulsen SG, Urquhart NS, Rosenbaum BJ. 2000. Designing a spatially
balanced, randomized site selection process for regional stream surveys: The EMAP Mid-Atlantic
Pilot Study. Environ Monitor Assess 63:95-113.
lannuzzi TJ, Harrington NW, Shear NM, Curry CL, Carlson-Lynch H, Henning MH, Su SH,
Rabbe DE. 1996. Distribution of key exposure factors controlling the uptake of xenobiotic
chemicals in an estuarine food web. Environ Toxicol Chem 15:1979-1992.
Isnard P, Lambert S. 1988. Estimating bioconcentration factors from octanol-water partition
coefficients and aqueous solubility. Chemosphere 17:21-34.
Jepsen R, Borglin S, Lick W, Swackhamer DL. 1995. Parameters affecting the adsorption of
hexachlorobenzene to natural sediments. Environ Toxicol Chem 9:1487-1497.
Kammer JA. 2000. Personal communication with Amy Benson, Abt Associates, Inc., Bethesda,
MD, January 28.
8-5
-------
Karickhoff SW. 1984. Organic pollutant sorpti on in aquatic systems. J Hydraul Div Am Soc Civ
110:707-735.
Karickhoff SW, Brown DS, Scott TA. 1979. Sorpti on of hydrophobic pollutants on natural
sediments. Water Res 13:241-24$.
Krzynowek J, Murphy J. 1987. Proximate Composition, Energy, Fatty Acid, Sodium, and
Cholesterol Content of Finfish, Shellfish, and Their Products. NOAA Technical Report NMFS
55. Department of Commerce, National Oceanic and Atmospheric Administration, National
Marine Fisheries Service. July.
Kukkonen J. 1995. The role of natural organic material on the fate and toxicity of xenobiotics in
aquatic environment. In: The Contaminants and the Nordic Ecosystem: Dynamics, Processes
and Fate. Ecovision World Monographs Series. M. Munawar and M. Luotola (eds.). Academic
Publishing, Amsterdam, pp. 95-108.
Kukkonen J, Oikari A. 1991 . Bioavailability of organic pollutants in boreal waters with varying
levels of dissolved organic material. Water Res 23:455-463.
Kukkonen J, Oikari A, Johnsen S, Gjessing E. 1989. Effects of humus concentrations on
benzo[a]pyrene accumulation from water toDaphnia magna: Comparison of natural waters and
standard preparations. Sci Total Environ 79:197-207.
Kukkonen J, Pellinen J. 1994. Binding of organic xenobiotics to dissolved organic
macromolecules: Comparison of analytical methods. Sci Total Environ 152:19-29.
Landrum PF, Nihart SR, Eadie BJ, Gardner WS. 1984. Reverse-phase separation method for
determining pollutant binding to Aldrich humic acid and dissolved organic carbon of natural
waters. Environ Sci Technol 18:187-192.
Landrum PF, Reinhold MD, Nihart SR, Eadie BJ. 1985. Predicting the bioavailability of organic
xenobiotics to Pontoporeia hoyi in the presence of humic and fulvic materials and natural
dissolved organic carbon. Environ Toxicol Chem 4:459-467.
LeBlanc GA. 1995. Trophic-level differences in the bioconcentration of chemicals: Implications
in assessing environmental biomagnification. Environ Sci Technol 29:154-160.
Leversee GJ, Landrum PF, Giesy JP, Fannin T. 1983. Humic acids reduce bioaccumulation of
some polycyclic aromatic hydrocarbons. Can J Fish Aquat Sci 40:63-69.
Lick W, Rapaka V. 1996. A quantitative analysis of the dynamics of the sorption of
hydrophobic organic chemicals to suspend the sediments. Environ Toxicol Chem 15:1038-1040.
Mackay D. 1982. Correlation of bioconcentration factors. Environ Sci Technol 16:274-278.
8-6
-------
MackayD. 1991. Multimedia Environmental Models: The Fugacity Approach. Lewis
Publishers, Chelsea, MI.
Martinez ND. 1991. Artifacts or attributes? Effects of resolution on the Little Rock Lake food
web. EcologMonogr 61:367-392.
McCarthy JF, Jimenez BD. 1985a. Interactions between poly cyclic aromatic hydrocarbons and
dissolved humic material: Binding and dissociation. Environ Sci Technol 19:1072-1076.
McCarthy JF, Jimenez BD. 1985b. Reduction in bioavailability to bluegills of polycyclic
aromatic hydrocarbons bound to dissolved humic material. Environ Toxicol Chem 4:511-521.
McCarthy JF, Jimenez BD, Barbee T. 1985. Effect of dissolved humic material on accumulation
of polycyclic aromatic hydrocarbons: Structure-activity relationships. Aquat Toxicol7:l5-24.
Morrison HA, Gobas FAPC, Lazar R, Whittle DM, Haffner GD. 1997. A food web
bioaccumulation model for organic contaminants in western Lake Erie. Environ Sci Technol
31:3267-3273.
Niimi AJ. 1985. Use of laboratory studies in assessing the behavior of contaminants in fish
inhabiting natural ecosystems. Wat Poll Res J Can 20:79-88.
Oliver BG, Niimi AJ. 1983. Bioconcentration of chlorobenzenes from water by rainbow trout:
Correlations with partition coefficients and environmental residues. Environ Sci Technol
17:287-291.
Oliver BG, Niimi AJ. 1988. Trophodynamic analysis of polychlorinated biphenyl congeners and
other chlorinated hydrocarbons in the Lake Ontario ecosystem. Environ Sci Technol 22:388-397'.
Pereira WE, Rostad CE, Chiou CT, Brinton TI, Barber LB II, Demcheck DK, Demas CR. 1988.
Contamination of estuarine water, biota, and sediment by halogenated organic compounds: A
field study. Environ Sci Technol 22:772-778.
Perminova IV, Grechishcheva NY, Petrosyan VS. 1999. Relationships between structure and
binding affinity of humic substances for polycyclic aromatic hydrocarbons: Relevance of
molecular descriptors. Environ Sci Technol 33:3781-3787.
Pignatello JJ, Xing B. 1996. Mechanisms of slow sorption of organic chemicals to natural
particles. Environ Sci Technol 30:1-11.
Poerschmann J, Zhang Z, Kopinke F-D, Pawliszyn J. 1997. Solid phase microextraction for
determining the distribution of chemicals in aqueous matrices. Anal Chem 69:597-600.
QEA. 1999. PCBs in the Upper Hudson River. Volume 2 -A Model ofPCB Fate, Transport,
and Bioaccumulation. Prepared for General Electric, Albany, New York.
8-7
-------
Ramos EU, Meijer SN, Vaes WHJ, Verhaar HJM, Hermans JLM. 1998. Using solid phase
microextraction to determine partition coefficients to humic acids and bioavailable concentrations
of hydrophobic chemicals. Environ Sci Technol 32:3430-3435.
Randall RC, Lee IIH, Ozretich RJ, Lake JL, Pruell RJ. 1991. Evaluation of selected lipid
methods for normalizing pollutant bioaccumulation. Environ Toxicol Chem 10:1431-1436.
Randall RC, Young DR, Lee II H, Echols SF. 1998. Lipid methodology and pollutant
normalization relationships for neutral nonpolar organic pollutants. Environ Toxicol Chem
17:788-791.
RasmussenD. 1998. Toxic Substances Monitoring Program (TSMP): DataBase Description.
California Environmental Protection Agency, State Water Resources Control Board, Water
Quality Division. 16pp.
Rasmussen JB, Rowan DJ, Lean DRS, Carey JH. 1990. Food chain structure and Ontario Lakes
determines PCB levels in lake trout {Salvelinus namaycush) and other pelagic fish. Can JFish
AquatSci 47:2030-2038.
ReinertRE. 1969. Insecticides and the Great Lakes. Limnos 2:3-9.
Reinert RE, Stewart D, Seagran HL. 1972. Effects of dressing and cooking on DDT
concentrations in certain fish from Lake Michigan. J Fish Res Board Can 29:525-529.
RickerWE. 1973. Linear regressions in fishery research. J Fish Res Board Can 30:409-434.
Russell RW, Gobas FAPC, Haffner GD. 1999. Role of chemical and ecological factors in trophic
transfer of organic chemicals in aquatic food webs. Environ Toxicol Chem 18:1250-1257.
Schlautman MA, Morgan JJ. 1993. Effects of aqueous chemistry on the binding of poly cyclic
aromatic hydrocarbons by dissolved humic materials. Environ Sci Technol 27:961-969.
Schwarzenbach RP, Gschwend PM, Imboden DM. 1993. Environmental Organic Chemistry.
Wiley, New York.
Servos MR, Muir DCG. 1989. Effect of dissolved organic matter from Canadian shield lakes on
the bioavailability of 1,3,6,8-tetrachlorodibenzo-p-dioxin to the amphipod Cranconyx
laurentianus. Environ Toxicol Chem 8:141-150.
Servos MR, Muir DCG, Webster GRB. 1989. The effect of dissolved organic matter on the
bioavailability of poly chlorinated dibenzo-p-dioxins. Aquatic Toxicol 14:169-184.
Seth R, MacKay D, Muncke J. 1999. Estimating the organic carbon partition coefficient and its
variability for hydrophobic chemicals. Environ Sci Technol 33:2390-2394.
-------
Sidwell VD. 1981. Chemical and Nutritional Composition of Finfishes, Whales, Crustaceans,
Mollusks, and Their Products. NOAA Technical Memorandum NMFS F/SEC-11. Department of
Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries
Service. January.
Stow CA, Jackson LJ, Amrhein JF. 1997. An examination of the PCB:lipid relationship among
individual fish. Can JFish Aquat Sci 54:1031-1038.
Suffet ffl, Jafvert CT, Kukkonen J, Servos MR, Spacie A, Williams LL, Noblet JA. 1994.
Chapter 3: Synopsis of Discussion Session: Influences of Particulate and Dissolved Material on
the Bioavailability of Organic Compounds. In: Bioavailability: Physical, Chemical and
Biological Interactions, Hamelink, J.L., Landrum P.F., Bergman H.L., Benson W.H. (eds.).
Lewis Publishers, Boca Raton, FL. pp. 155-170.
Swackhamer DL, Kites, RA. 1988. Occurrence and bioaccumulation of organochlorine
compounds in fishes from Siskiwit Lake, Isle Royale, Lake Superior. Environ Sci Technol
22:543-548.
Thomann RV. 1989. Bioaccumulation model of organic chemical distribution in aquatic food
chains. Environ Sci Technol 23:699-707.
Thomann RV, Connolly JP, Parkerton TF. 1992. An equilibrium model of organic chemical
accumulation in aquatic food webs with sediment interaction. Environ Toxicol Chem 11:615-
629.
Thurman EM. 1985. Organic Geochemistry of Natural Waters, Martinus Nijhoff/Dr. W. Junk,
Dordrecht, The Netherlands. 497 pp.
Tomson MB, Pignatello JJ. 1999. Editorial: Causes and effects of resistant sorption in natural
particles. Environ Toxicol Chem 8:1609.
USDA(U.S. Department of Agriculture). 1998. 1994-1996 Continuing Survey of Food Intakes
by Individuals and 1994-1996 Diet and Health Knowledge Survey. U.S. Department of
Agriculture, Agricultural Research Service. (Available from National Technical Information
Service, Springfield VA, NTIS Document Number: PB98-500457).
USEPA (U.S. Environmental Protection Agency). 1980. Appendix C—Guidelines and
Methodology Used in the Preparation of Health Effect Assessment Chapters of the Consent
Decree Water Criteria Documents. Federal Register 45:79347-79357. November 28.
USEPA (U.S. Environmental Protection Agency). 1992a. National Study of Chemical Residues
in Fish (Volumes I and II). Office of Water, Washington, DC. EPA/823/R-92/008a and
EPA/823/R-92/008b.
8-9
-------
USEPA (U.S. Environmental Protection Agency). 1992b. Development and application of a
model of PCBs in the Green Bay, Lake Michigan Walleye and Brown Trout and their food webs.
Manhattan College, Riverdale, New York. USEPA Large Lakes Research Station, Grosse He,
Michigan. (Technical Report prepared by J.P. Connolly et al., Manhattan College, under
cooperative agreement No. CR-815396).
USEPA (U. S. Environmental Protection Agency). 1993. Interim Report on Data and Methods
for Assessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin Risks to Aquatic Life and Associated
Wildlife. Office of Research and Development, Washington, DC. EPA/600/R-93/055.
USEPA (U.S. Environmental Protection Agency). 1995a. Final Water Quality Guidance for the
Great Lakes System. Federal Register 60:15366-15425. March 23.
USEPA (U.S. Environmental Protection Agency). 1995b. Great Lakes Water Quality Initiative
Technical Support Document for the Procedure to Determine Bioaccumulation Factors. Office
of Water. Washington, DC. EPA/820/B-95/005.
USEPA (U.S. Environmental Protection Agency). 1995c. Addendum to Green Bay Final Report
Food Chain Model Projections. USEPA Large Lakes Research Station, Grosse He, MI.
(Technical Report prepared by HydroQual Inc., Mahwah, NJ., Contract No. 68-C3-0332).
USEPA (U.S. Environmental Protection Agency). 1996. Ecological Effects Test Guidelines
OPPTS 850.1730 FishBCF. Office of Prevention, Pesticides, and Toxic Substances.
Washington, DC. EPA-512-C-96-129.
USEPA (U.S. Environmental Protection Agency). 1997. Further site characterization and
analysis. Volume 2C - Data evaluation and interpretation report. Hudson River PCBs
Reassessment RI/FS. USEPA, Region II, New York.
USEPA (U.S. Environmental Protection Agency). 1998a. Ambient Water Quality Criteria
Derivation Methodology-Human Health, Technical Support Document. Final Draft. Office of
Water, Washington, DC. EPA/822/B-98/005. (Report available at:
http://www.epa.gov/waterscience/humanhealth/).
USEPA (U.S. Environmental Protection Agency). 1998b. Database for the Hudson River PCBs
Reassessment RI/FS - Phase 2 Report; Release 4.1. USEPA, Region II, New York.
USEPA (U.S. Environmental Protection Agency). 2000a. Methodology for Deriving Ambient
Water Quality Criteria for the Protection of Human Health (2000). Office of Water,
Washington, DC. EPA-822-B-00-004.
USEPA (U.S. Environmental Protection Agency). 2000b. Draft Technical Basis for the
Derivation of Equilibrium Partitioning Sediment Benchmarks (ESBs) for the Protection of
Benthic Organisms: Nonionic Organics. Draft. Office of Water, Washington, DC. EPA-822-R-
02-047.
8-10
-------
USEPA(U.S. Environmental Protect!on Agency). 2000c. AQUATOXfor Windows: A Modular
Fate and Effects Model for Aquatic Ecosystems, Release 1. Vol. 1: User's Manual. Office of
Water, Washington, DC.
USEPA(U.S. Environmental Protect! on Agency). 2000d. AQUATOX for Windows: A Modular
Fate and Effects Model for Aquatic Ecosystems, Release 1. Vol. 2: Technical Documentation.
Office of Water, Washington, DC.
USEPA (U. S. Environmental Protection Agency). 2000e. Trophic Level and Exposure Analyses
for Selected Piscivorous Birds and Mammals. Volume I: Analyses of Species for the Great
Lakes. Draft. Office of Water, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2000f. Trophic Level and Exposure Analyses
for Selected Piscivorous Birds and Mammals. Volume II: Analyses of Species in the
Conterminous United States. Draft. Office of Water, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2000g. Trophic Level and Exposure Analyses
for Selected Piscivorous Birds and Mammals. Volume III: Appendices. Draft. Office of Water,
Washington, DC.
USEPA (U.S. Environmental Protection Agency). 200la. National Sediment Quality Survey
Database, 1980-1999: Fact Sheet. Office of Water, Office of Science and Technology. EPA-
823-F-01-002. Database available at http://www.epa.gov/ost/cs/nsidbase.html.
USEPA (U.S. Environmental Protection Agency). 2002. Draft Action Plan for the Development
of a Framework for Metals Assessment and Guidance for Characterizing and Ranking Metals
(External Review Draft). Office of Research and Development, Washington, DC. 55 pp.
EPA/630/P-02/003A.
USGS (U.S. Geological Survey ). 1997. Methods of Analysis by the U.S. Geological Survey
National Water Quality Laboratory—Determination ofNonpurgeable Suspended Organic
Carbon by Wet-Chemical Oxidation and Infrared Spectrometry. Open-File Report 97-380.
U.S. Geological Survey, National Water Quality Laboratory, Denver, CO.
van den Heuvel MR, McCarty LS, Lanno RP, Hickie BE, Dixon DG. 1991. Effect of total body
lipid on the toxicity and toxicokinetics of pentachlorophenol in rainbow trout (Oncorhynchus
mykiss). Aquat Toxicol20:235-252.
van Wezel AP, Opperhuizen A. 1995. Narcosis due to environmental pollutants in aquatic
organisms: Residue-based toxicity, mechanisms, and membrane burdens. Crit Rev Toxicol
25:255-279.
Vander Zanden MJ, Rasmussen JB. 1996. A trophic position model of pelagic food webs:
Impact on contaminant bioaccumulation in lake trout. EcolMonogr 66:451-477.
8-11
-------
Veith GD, Austin NM, Morris RT. 1979a. A rapid method for estimating log P for organic
chemicals. Water Res 13:43-47.
Veith GD, DeFoe DFL, Bergstedt BV. 1979b. Measuring and estimating the bioconcentration
factor in fish. JFish Res Board Can 36:1040-1045.
Verhaar HJM, de Jongh J, Hermans JLM. 1999. Modeling the bioconcentration of organic
compounds by fish: A novel approach. Environ Sci Technol 33:4069-4072.
Wu S-C, Gschwend PM. 1986. Sorption kinetics of hydrophobic organic compounds to natural
sediments and soils. Environ Sci Technol 20:717-725.
Yin C, Hassett JP. 1986. Gas-partitioning approach for laboratory and field studies of Mirex
fugacity in water. Environ Sci Technol 20:1213-1217.
Yin C, Hassett JP. 1989. Fugacity and phase distribution of Mirex in Oswego River and Lake
Ontario waters. Chemosphere 19:1289-1296.
8-12
-------
APPENDIX A
DERIVATION OF THE BASIC BIOCONCENTRATION AND BIO ACCUMULATION
EQUATIONS FOR ORGANIC CHEMICALS
A-l
-------
1. DERIVATION OF THE BASIC BIOCONCENTRATION AND
BIO ACCUMULATION EQUATIONS FOR ORGANIC CHEMICALS
This appendix provides a detailed presentation of the derivation of the basic
bioconcentration and bioaccumulation equations for organic chemicals that are the basis for the
methods for deriving BCFs and BAFs in EPA's BAF methodology. The equations are based on
widely accepted and peer-reviewed scientific principles and theories, as referenced in this
appendix. This appendix was developed to provide additional background for the TSD.
Therefore, some additional notations (e.g., subscripts and superscripts) have been added to the
equation terms to provide clarity to the discussion of the equation derivations.
A. 1.1 Bioconcentration
The basic BCF applicable to all classes of chemicals is defined as:
BCFTl = =^
(Equation A-1)
where:
BCFj = total bioconcentration factor (i.e., a BCF that is based on the total
concentrations of the chemical in the water and in the aquatic biota)
CB = total concentration of the chemical in the aquatic biota, based on the wet
weight of the aquatic biota
C^ = total concentration of the chemical in the water around the aquatic biota
As more bioconcentration information was generated and reviewed by scientists, it was
realized that extrapolation of BCFs for organic chemicals from one species to another would be
more accurate if the BCFs were normalized on the basis of the amount of lipid in the aquatic
biota exposed in the original bioconcentration test, because many nonpolar organic chemicals are
hydrophobic, accumulating in direct proportion to the amount of lipid in a given aquatic
organism (Mackay, 1982; Connolly andPederson, 1988; Thomann, 1989). It was also realized
that extrapolation of BCFs for organic chemicals from one water to another would be more
accurate if the BCFs were calculated on the basis of the freely dissolved concentration of the
organic chemical in the water around the aquatic biota. Thus, two additional BCFs were defined
and used:
A-2
-------
BCF,'=
(Equation A-2)
BCF/d = =L
(Equation A-3)
where:
BCF = lipid-normalized total BCF (i.e., normalized to 100% lipid and based on the
total concentration of the chemical in the water around the biota)
C. = lipid-normalized concentration of the chemical in the aquatic biota
BCF? = lipid-normalized and freely dissolved-based BCF
C ™ = freely dissolved concentration of chemical in the water around the aquatic
biota
The experimental definition of C. is:
„ = the total amount of chemical in the aquatic biota
the amount of lipid in the aquatic biota
(BXC.J) (B)(C.h C^
=
-------
If ffd = the fraction of the chemical in the water around the aquatic biota that is freely dissolved,
then:
(Equation A-6)
Using Equations 4 and 6 to substitute for C. and C " in Equation 3 and then using Equation 1:
/•• t i»/""i3 i
fi l_in Dl_iFip
BCF/d = B _ T
(Equation A-7)
Equations 1, 5, and 7 show the relationships among the three different BCFs.
Theoretical justification for use of both lipid normalization and the freely dissolved
concentration of the organic chemical in the ambient water is based on the concept of equilibrium
partitioning, whereas practical justification is provided by the general similarity of the value of
BCF? for an organic chemical across both species and waters. This concept of equilibrium
partitioning is discussed further in the following section. It will be demonstrated in Section A.2,
however, that a more complete application of equilibrium partition theory shows that BCF?
extrapolates well only for chemicals whose Kows are greater than 1,000, whereas a different BCF
(BCF£d) extrapolates well for organic chemicals whose Kows are greater than 1,000 as well as for
chemicals whose Kows are less than 1,000.
A.2 PARTITION THEORY AND BIOCONCENTRATION
Equilibrium partition theory provides the understanding necessary to ensure proper use of
Kows, BCFs, and BAFs in the derivation of water quality criteria for organic chemicals. For the
purpose of applying partition theory, aquatic biota can be modeled as consisting of water, lipid,
and nonlipid organic matter (Barber et al., 1991). In this model, an organic chemical in aquatic
biota partitions into three phases:
1. The chemical that is freely dissolved in the water that is in the biota.
2. The chemical that is partitioned to the lipid that is in the biota.
3. The chemical that is partitioned to nonlipid organic matter in the biota.
The total concentration of chemical in the water inside the biota includes chemical that is
partitioned to lipid and nonlipid organic matter in the water.
A-4
-------
According to this model:
(Equation A-8)
where:
fw = fraction of the aquatic biota that is water
C^B = freely dissolved concentration of the organic chemical in the water in the
aquatic biota
f. = fraction of the aquatic biota that is lipid
CL = concentration of the organic chemical in the lipid
fN = fraction of the aquatic biota that is nonlipid organic matter
CN = concentration of the organic chemical in the nonlipid organic matter in the
aquatic biota
The most important partitioning of the organic chemical within the aquatic biota is between the
lipid and the water, which is described by the following equation:
where:
(Equation A-9)
KLW = the lipid-water partition coefficient
"KLW" (Gobas, 1993) is used herein because it is more descriptive than "KL" which is
used by DiToro et al. (1991). This partition coefficient is central to the equilibrium partition
approach that is used to derive sediment quality criteria (DiToro et al., 1991), the food chain
multipliers based on the Gobas model, and the equations given here that are used to derive BCFs
and BAFs for the national BAF methodology.
In order for Equations 8 and 9 to be correct, partition theory requires that the
concentration of the organic chemical in the lipid, CL, be defined as:
c _ the amount of chemical partitioned to lipid in aquatic biota
L the amount of lipid in the aquatic biota
A-5
-------
It is difficult to determine CL experimentally because it is not easy to measure only the
chemical that is partitioned to the lipid (i.e., it is not easy to separate the three different
compartments within aquatic biota that the chemical partitions into according to the model).
Because all of the organic chemical in the biota is measured when C. is determined, C. can be
determined easily, and C. is higher than CL.
It is useful to define another BCF as:
BCF/d = —L
(Equation A-10)
because CL is lower than C., BCF£d < BCF?.
The only difference between KLW and BCF™ is that the denominator in KLW is C^B ,
whereas the denominator in BCF™ is C^B- When partition theory applies, however, all phases
are in equilibrium, and so:
fd _ £ fd
(Equation A-11)
Therefore, when the organic chemical is not metabolized by the aquatic biota and when growth
dilution is negligible:
BCFLfd = KLW
(Equation A-12)
In laboratory experiments it has been shown that the chemical octanol is a useful surrogate for
lipid, thus a reasonable approximation is that:
(Equation A- 13)
where:
Kow = the w-octanol-water partition coefficient.
Thus:
predicted BCFLfd = KLW = Kw
(Equation A- 14)
A-6
-------
By using Equations 9 and 1 1 to substitute for CL and C^B in Equation 8
wfd) + (Q(BCFLfd)(Cjd) +
By using Equation 6 to substitute for C™ in Equation 15:
Dividing by C^ gives:
Using Equation 1 and rearranging gives:
BCFTl = (ffd)[
Using Equation 6:
BCFTl = (ffd)[
(Equation A-15)
(Equation A-16)
(Equation A-17)
(Equation A-18)
(Equation A-19)
Substituting x = fw + (fN)(CN/C" ) and rearranging gives:
BCFTl =
(Equation A-20)
The term "(f.)(BCF£d)" accounts for the amount of organic chemical that is partitioned to the lipid
in the biota, whereas in "x," the term "fw" accounts for the amount of organic chemical that is
freely dissolved in the water in the biota and the term "(fN)(CN/Cwd)" accounts for the amount of
organic chemical that is partitioned to nonlipid organic matter in the biota. The relative
magnitudes of these three terms depend on the following:
A-7
-------
Because of bones and other inorganic matter that make up the total mass of an
organism, the sum of fw + f . + fN must be less than 1 .
fw is usually about 0.7 to 0.9.
f. must be measured in the organism in question if the BAF or BCF is to be useful,
or estimated from other similar biota; it is usually between 0.03 and 0.15.
The term "(CN/C ")" is similar to BCF£d (see Equation 1 0) and is therefore
probably related to Kow (see Equation 14), although the affinity of the chemical for
nonlipid organic matter is probably much less than its affinity for lipid.
Although such considerations aid in understanding "x" in Equation 20, the magnitude of
"x" is important only for chemicals whose log Kows are in the range of 1 to 3. For organic
chemicals whose log Kows are about 1, ffd is about 1. In addition, such chemicals distribute
themselves so as to have similar concentrations in water and in the different organic phases in the
aquatic biota, which means that ECF{ will be approximately 1 if both metabolism and growth
dilution are negligible. An organic chemical whose log Kow is less than 1 will also have a ECF{ on
the order of 1 because water is the predominant component in aquatic biota. Setting "x" equal to
1 is about right in the range of log Kows in which it is not negligible (see also McCarty et al.,
1992).
Substituting x = 1 into Equation 20:
BCFTl =
(Equation A-21)
Rearranging gives:
BCF" =
(Equation A-22)
Because BCF" is normalized for both the aquatic biota lipid content and freely dissolved fraction
of the chemical, it is called the "baseline BCF." The baseline BCF is the most useful BCF for
extrapolating from one species to another and from one water to another for organic chemicals
with both high and low Kows. The baseline BCF is intended to reference bioconcentration of
organic chemicals to partitioning between lipid and water.
Equations 12, 13, and 22 demonstrate that both Kow and
A-8
-------
are useful approximations of the baseline BCFs. It will probably be possible to improve both
approximations within a few years, but such improvements might not affect the BCFs
substantially and probably will not require changes in the rest of the equations or the
terminology.
When BCFx is greater than 1,000, the "-1" in Equation 22 is negligible, and so this
equation becomes equivalent to Equation 7 (i.e., when BCFj is large it generally indicates that the
chemical readily partitions to and accumulates in the lipid portion of aquatic biota, and thus the
BCF? is a useful approximation of the baseline BCF).
A.3 DERIVATION OF THE BASIC BAF AND BASELINE BAF EQUATIONS
As has been previously mentioned, bioaccumulation represents uptake and retention of a
chemical from all routes of exposure, including water only (i.e., bioconcentration) and the food
chain, therefore by analogy and substituting BAF for BCF in Equations 21 and 22:
BAFTl =
BAFLfd = (
(Equation A-23)
(Equation A-24)
As with the BCF, the BAF" can be called the "baseline BAF" because it normalizes the factor to
the lipid content of the aquatic biota and the freely dissolved fraction of the chemical in water. It
too is the most useful BAF for extrapolating from one species to another and from one water to
another for chemicals with both high and low Kows.
A.4 CALCULATION OF CRITERIA
Baseline BCFs and BAFs can be extrapolated between species and waters, but they
cannot be used directly in the calculation of criteria that are based on the total concentration of
the chemical in the water. The BCFs and BAFs that are needed to calculate such criteria can be
calculated from measured and predicted baseline BCFs and BAFs using the following equations,
which are derived from Equations 21 and 23:
1 + (Baseline BCF)($) ]
(Equation A-25)
A-9
-------
BAFT = (*fd>[ l + (Baseline BAF)($) ]
(Equation A-26)
A.5 DERIVATION OF THE BASIC FCM AND BMF EQUATIONS
Food chain multipliers (FCM) are used in BAF methods 3 and 4 (see Sections 5.3 and 5.4)
for estimating B AFs for chemicals that biomagnify up the food chain. The FCM can be defined
as:
FCM = Baseline BAF _ BAFLfd
Baseline BCF BCFfd
(Equation A-27)
Some of the consequences of Equation 27 are:
1. Substituting Equations 22 and 24 into Equation 27:
_ BAFTl - ffd
FCM =
RPU * — f
Bi>rT 1^
(Equation A-28)
Therefore, BAF^ = (FCM) (BCF^) only when ffd is much less than BAF^ and BCF^.
2. When FCM = 1 (as for trophic level 2 in the Gobas model):
Baseline BAF = Baseline BCF
(Equation A-29)
3. Predicted Baseline BAFs can be obtained using FCMs and the following rearrangement of
Equation 27:
predicted Baseline BAF = (FCM)(Baseline BCF) (Equation A-30)
a. Using a laboratory-measured BCF in Equation 22:
predicted Baseline BAF = (FCM)(measured Baseline BCF) (Equation A-31)
BCF,1 i
= (FCM)( 1 - 1 )(i) (Equation A-32)
A-10
-------
b. Using a predicted BCF in Equation 14:
predicted Baseline BAF = (FCM)(predicted BCF£d)
= (FCM)(KOW)
(Equation A-33)
(Equation A-34)
The FCMs used to calculate predicted baseline BAFs must be appropriate for the trophic
level of the aquatic biota to which the predicted baseline BAF is intended to apply.
Although BAFs can be related to BCFs using FCMs, BAFs and BCFs can also be related
using biomagnification factors (BMFs). The two systems are entirely compatible, but confusion
can result if the terms are not used consistently and clearly. Because both FCMs and BMFs are
used in the Guidance document and elsewhere, it is appropriate to explain the relation between
the two here. The basic difference is that FCMs always relate back to trophic level one (TL1),
whereas BMFs always relate back to the next trophic level. In the FCM system:
BAF
TLl
In the BMF system:
= BCF
= (FCMTL2)(BAFTL1)
= (FCMTL3)(BAFTL1)
= (FCMTL4)(BAFTL1)
= BCF
= (BMFTL2)(BAFTL1)
= (BMFTL3)(BAFTL2)
= (BMFTL4)(BAFTL3)
Therefore:
= FCM
-TL2
= (FCMTL3)/(FCMTL2)
= (FCMTL4)/(FCMTL3)
Both metabolism and growth dilution can cause BMFs to be less than 1.
A-ll
-------
A.6 REFERENCES
Barber MG, Suarez LA, Lassiter RR. 1991. Modeling bioaccumulation of organic pollutants in
fish with an application to PCBs in Lake Ontario salmonids. Can JFish Aquat Sci 48:318-337.
Connolly J, Pedersen C. 1988. A thermodynamic-based evaluation of organic chemical
accumulation in aquatic organisms. Environ Sci Technol 22:99-103.
DiToro DM, Zarba CS, Hansen DJ, Berry WJ, Swartz RC, Cowan CE, Pavlou SP, Allen HE,
Thomas NA, Paquin PR. 1991. Technical basis for establishing sediment quality criteria for
nonionic organic chemicals using equilibrium partitioning. Environ Toxicol Chem 10:1541-1583.
Gobas, FAPC. 1993. A model for predicting the bioaccumulation of hydrophobic organic
chemicals in aquatic food-webs: Application to Lake Ontario. EcolMod 69:1-17'.
Mackay D. 1982. Correlation of bioconcentration factors. Environ Sci Technol 16:274-278.
McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG. 1992. Residue-based
interpretation of toxicity and bioconcentration QSARs from aquatic bioassays: neutral narcotic
organics. Environ Toxicol Chem 11:917-930.
Thomann RV. 1989. Bioaccumulation model of organic chemical distribution in aquatic food
chains. Environ Sci Technol 23:699-707.
A-12
-------
APPENDIX B
PROTOCOL FOR DETERMINING
OCTANOL-WATER PARTITION COEFFICIENTS (Kow)
FOR COMPOUNDS WITH LOG Kow VALUES > 5
B-l
-------
PROTOCOL FOR DETERMINING
OCTANOL-WATER PARTITION COEFFICIENTS (Kow)
FOR COMPOUNDS WITH LOG Kow VALUES > 5
1. INTRODUCTION
The octanol -water partition coefficient (Kow) is one of the most widely used chemical
parameters. The Kow of a chemical has been found to be representative of a chemical's propensity
to partition into biotic and abiotic components of the environment as well as a chemical's
propensity to accumulate in living organisms. Because of these associations, the Kow is widely
used to predict a chemical's behavior in the environment and to evaluate a chemical's impact on
human health.
The octanol-water partition coefficient (Kow) is a unitless measure and is defined as the
ratio of the equilibrium concentrations, C, of a chemical in the two phases of a system consisting
of n -octanol and water at standard temperature and pressure (STP, 25° C, 1 atm)
where Coc, represents the concentration in the n -octanol phase, and Cw represents the
concentration in the water. The concentrations in the respective phases are expressed in the same
volume-referenced units (i.e., mg/mL, mole/L, etc.), therefore, the Kow is a unitless property.
Since the value of the partition coefficient spans orders of magnitude, it is frequently expressed
on a log scale (base ten) such that a given chemical has a log Kow value which may range from 1
to >8. This parameter is also called the log P value.
Some specific applications of the Kow within the U.S. EPA include: evaluation of a
chemical's potential to bioaccumulate in aquatic life, wildlife and humans; modeling the fate,
transport and distribution of a chemical in the environment; prediction of the distribution of a
contaminant in a living organism; classification of persistent bioaccumulators for regulatory
actions; derivation of soil screening levels; calculation of water quality benchmarks; and
derivation of Sediment Quality Advisory Levels.
Although a seemingly simple experimental determination, Kow measurement is beset with
difficulties. The appropriateness and accuracy of laboratory methods to directly measure a Kow
are influenced by a number of factors which include the magnitude of the value itself. For
chemicals with log Kow values at or exceeding 5, common sources of error include: (1) failure to
achieve equilibrium; (2) incomplete phase separation or interphase mixing during sampling;
(3) emulsion effects derived from "excessive" mixing or induced by contaminants; (4) propensity
of the chemical to self-associate, tautomerize or form hydrates; and (5) the presence of small
quantities of contaminants with a lower Kow value. These errors tend not to be random, but to
give measured numbers lower than the true value, frequently by an order of magnitude or more.
The likelihood and degree of error increases with increasing Kow and also seems to be more
prevalent for certain classes of chemicals (such as halogenated compounds or phthalate esters).
As a result, in addition to direct experimental measurement methods, techniques to indirectly
experimentally measure or estimate Kow values have been developed.
B-2
-------
1.1 Experimental Measurement Techniques
Direct experimental measurement techniques include the shake-flask approach,
generator column, and slow-stir methods.
The shake-flask method is the classical approach and fairly straight-forward for
chemicals with log Kow values below 5. For chemicals with higher log Kow values,
the shake-flask approach requires large volumes of water and formation of
emulsions becomes a significant impediment to accurate measurements.
The generator-column approach was developed to measure the partition
coefficients of more hydrophobic chemicals (those with larger log Kow values). This
is a laborious method which results in more reliable data than the shake-flask
approach for chemicals with higher log Kow values, but some discontinuities in the
data for higher-chlorinated PCB congeners have been observed.
• A third direct measurement technique is the slow-stir method. In this method,
careful stirring and close temperature control can prevent or limit the formation of
emulsions and reliable very high partition coefficients can be obtained relatively
easily.
1.2 Indirect Experimental Measurement Techniques
An indirect experimental measurement technique is to incorporate a radioactive label into
the chemical and use a radiotracer assay to evaluate the compound's distribution between the
octanol and water phases. This approach can be used when you have small amounts of the
compound. However, radiotracer assays do not directly measure the compound, and low Kow
values frequently result from the presence of impurities or instability of the compound.
1.3 Computer-based Estimation Techniques
Because of the difficulty of directly and accurately measuring Kow values, various
computer-based estimation methods exist. These can be divided into two types, those based
upon fundamental chemical thermodynamics, and those requiring a training set of chemicals with
measured Kows.
1.3.1 Technique based on fundamental thermodynamics
Computer methods based on fundamental chemical structure theory and are not limited
by nor do they require a training set of chemicals with measured Kows. For example, the SPARC3
model consists of a set of core models describing intra- and inter-molecular interactions. These
3 SPARC (SPARC Performs Automated Reasoning in Chemistry) is a mechanistic model developed at the
Ecosystems Research Division of the National Exposure Research Laboratory of the Office of Research and
Development of the U.S. Environmental Protection Agency by Sam Karickhoff, Lionel Carreira, and co-workers. A
prototype version was used for which no performance data for Kow estimation is available. The model complements
the aforementioned models because development, training, and testing were done away from Kow data. (See Hilal,
Carreira, and Karickhoff, 1994, for model description.)
B-3
-------
models are linked by appropriate thermodynamic relationships to provide estimates of reactivity
parameters under desired conditions (e.g., temperature, pressure, solvent).
1.3.2 Techniques using a training set of chemicals
Methods requiring a training set of chemicals use Quantitative Property-Property
Relationships (QPPRs) or Quantitative Structure Activity Relationships (QSARs) to derive Kows.
In QPPRs, Kow values are correlated with the values for other chemical parameters—either
measured or calculated—using data available from a training set of chemicals. In QSARs, Kow
values are derived from fragment constants obtained from a training set of chemicals.
One application of QPPRs is estimating Kows indirectly from other experimental
measurements. In this approach, the Kow is correlated with another measured property. These
techniques include the use of reversed-phase high performance liquid chromatography (HPLC)
and reversed-phase thin-layer chromatography (TLC). In applying these approaches, Kows are
estimated from linear equations relating retention times on the reversed-phase column to the Kow
values. The equations are developed based on a set of reference chemicals for which Kow values
are well established. These are relatively efficient methods because they do not require
quantification of concentrations, but the linear equations can not be extrapolated beyond the Kow
range represented by the reference chemicals from which the equation was derived. In
application, values for the reference chemicals are usually shake-flask values obtained from the
literature, resulting in unreliable Kow estimates for chemicals with higher log Kow values.
In addition to direct and indirect measurement methods, QPPRs are also used to establish
correlations between the Kow and calculated properties. For example, Hawker and Connell (1988)
developed a correlative relationship between log Kow and molecular surface area using
approximately two dozen PCBs. They then estimated log Kows for the remaining PCBs by
inputting the molecular surface area of each PCB. This technique is limited to estimating Kows for
chemicals which are similar to the chemicals used in developing the relationship.
In QSARs, hydrophobic fragment values are derived from a large database of measured
Kows. These fragments constants are used to estimate Kow in two ways: (1) One approach is to
estimate the Kow by adding up the values for all the fragments composing the chemical, either by
atom or by functional group. (2) The other approach is to start with a measured Kow value for a
structurally similar compound and add or subtract the fragment constants for functional groups
or atoms to estimate the Kow for the specific compound. In both these cases, the calculated Kow
value must also be corrected for proximity effects between structurally close substituent groups,
and the Kow value derived is only as good as the data associated with the training set of chemicals.
This method is also limited to predicting Kows for chemicals with structures similar to those within
B-4
-------
the training set. Computer-based models exist which apply QSAR approaches to estimate Kows.
CLOGP4 and LOGKOW5 databases are both applications of this approach.
1.4 Recommendations
Given the numerous techniques available to determine the Kow and its numerous and
important applications across the Agency, the U.S. EPA has formed an Agency Kow Work Group
to determine recommended Kow values for chemicals of concern to various EPA programs. In
determining these recommended Kow values, the preferable option would be to recommend actual
measured values. For chemicals with log Kow values below 5, the classical shake-flask approach is
adequate to obtain these measurements. However, there is a serious shortage of reliable measured
data for compounds with higher log Kow values (Log Kow > 5) and these chemicals frequently
exhibit a propensity to accumulate in living tissues or bind to soils and sediments. For these
reasons, this protocol has been restricted to chemicals with log Kow values equal to or
exceeding 5.
2. PROTOCOL FOR DETERMINING RECOMMENDED Kow VALUES
Measured values are preferable to estimated values for determining recommended Kow
values. However, the absence or scarcity of reliable data necessitates the use of estimation
methods in evaluating data and in assigning Kows. Kow estimates used in this exercise include:
calculation methods (e.g., CLOGP, LOGKOW, SPARC, and fragment additions or subtractions)
and QPPRs (e.g., HPLC and TLC methods). All of these approaches except SPARC, an
estimation method based on fundamental chemical structure theory, require measured Kow values
for a training set of chemicals.
Assigning a Kow from these data will necessarily involve scientific judgement in evaluating
not only the reliability of all data inputs but also the accretion/concretion of evidence in support
of the recommended Kow value. Supporting rationale will be provided for each recommended
value.
2.1 Operational Guidelines
"High quality" measured value are preferred over estimates. For chemicals with
log Kow> 5, it is highly unlikely to find multiple "high quality" measurements.
(Note: "high quality" is data judged to be reliable based on the guidelines
4 CLOGP is a molecular fragment-based model developed at Pomona College by Albert Leo, Corwin Hansch, and co-
workers. This model has undergone extensive development and exhaustive testing; version 3.1 was used in this
exercise. (See Hansch and Leo, 1995, for model description and performance data.)
5 LOGKOW is essentially an expanded CLOGP with more recent training data and additional fragment constants. The
developers were Philip Howard, William Meylan and co-workers at Syracuse Research Corporation; Version 1.51
was used in this exercise. (See Meylan and Howard, 1994, for model details and performance information.)
B-5
-------
presented in Appendix I). Due to the paucity of "high quality" data, assigning
Kow's from estimation techniques may be necessary.
Kow measurements by slow stir are extendable to 108. Shake flask Kow
measurements are extendable to 106 with sufficient attention to micro emulsion
effects; for classes of chemicals that are not highly sensitive to emulsion effects
(i.e., polycyclic nuclear aromatic hydrocarbons) this range may extend to 106 5.
What is considered reasonable agreement in log Kow data (measured or estimated)
depends primarily on the magnitude of the log Kow value. Therefore, the following
ranges of acceptable variation have been established for this exercise: 0.5 for log
Kow > 7; 0.4 for 6 • log Kow • 7 ; 0.3 for log Kow < 6.
Statistical methods should be applied to data as appropriate. However, it is
recognized that application is limited by the paucity of data and the
determinate/methodic nature of most measurement error(s).
2.2 Tiered Procedure for Selecting Kow Values
I. Assemble/evaluate experimental and calculated data (e.g.,CLOGP, LOGKOW,
SPARC).
II. If calculated log Kow' s > 8:
A. Develop independent estimates
1. Liquid Chromatography (LC) methods with "appropriate"
standards. (See Appendix I for guidelines for LC application.)
2. Structure Activity Relationship (SAR) estimates extrapolated from
similar chemicals where "high quality" measurements are available.
3. Property Reactivity Correlation (PRC) estimates based on other
measured properties (solubility, etc.)
B. If calculated data are in reasonable agreement (as defined in section 2.1)
and are supported by independent estimates described above, report the
average calculated value.
C. If calculated/estimated data do not agree, use professional judgement to
evaluate/blend/weight calculated and estimated data to assign the Kow
value.
D. Document rationale including relevant statistics.
B-6
-------
III. If calculated log Kow' s in range 6 - 8:
A. Look for "high quality" measurements. These will generally be slow stir
measurements, the exception being certain classes of compounds where
micro emulsions tend to be less of a problem (i.e., polycyclic nuclear
aromatic hydrocarbons; for these compounds, shake flask measurements
are good to log Kow of 6.5).
B. If measured data are available with reasonable agreement (both
measurements and calculations), report average measured value.
C. If measured data are in reasonable agreement, but differ from calculated
values, develop independent estimates and apply professional judgement
to evaluate/blend/weight the measured, calculated and estimated data to
assign the Kow value.
D. If measured data are not in reasonable agreement (or if only one
measurement is available), use II A, B, and C to produce a "best estimate";
use this value to evaluate/screen the measured Kow data. Report the average
value of the screened data. If no measurements agree with the "best
estimate," apply professional judgement to evaluate/blend/weight the
measured, calculated and estimated data to assign the Kow value.
E. If measured data are unavailable, proceed through II A, B, C and report the
"best estimate".
F. Document rationale including relevant statistics.
IV. If calculated log Kow' s < 6:
A. Proceed as in III. Slow stir is the preferred method but shake flask data can
be considered for all chemicals if sufficient attention has been given to
emulsion problems in the measurement.
3. REFERENCES
Hansch, C. and A. Leo, Exploring QSAR, American Chemical Society, 1995.
Hawker, D. W. andD.W. Connell. Environ. Sci. Technol.. 1988, 22, 382-387.
Hilal, S. H., L. A. Carreira and S. W. Karickhoff, Quantitative Treatments of Solute/Solvent
Interactions, Theoretical and Computational Chemistry, Vol. 1, 291-353, 1994 Elsevier Science.
Meylan, W. M. and P. H. Howard, J. Pharm. Sci. 1995, 84, 83-92.
B-7
-------
ATTACHMENT I: GUIDELINES FOR EVALUATING MEASURED AND ESTIMATED
VALUES
1. ASSESSMENT OF MEASURED Kow VALUES
1 . 1 Molecular Speciation. In order to interpret measured data, it is necessary to understand
the molecular species present in both the octanol and water phases including ionization,
self-association, tautomerization, and hydrate formation. For these reasons, it is difficult
to conduct or interpret such measurements for mixtures of unknown composition or for
single molecules of unknown structure. Solutes composed of more than one molecular
species may also show substantial temperature dependence of Kow reflecting relative
change in speciation in the octanol and water phases.
1.1.1 Ionization. This protocol is directed primarily towards assigning a log Kow value
for neutral (non-ionizable) organic compounds. In the case of weakly acidic or
basic compounds a portion of the molecules may be ionized at environmental pH,
and partitioning into biota or abiota will be correspondingly reduced. For weakly
ionizable molecules, shake flask measurements are conducted in solutions of a
stable, non-extractable buffer to suppress ionization. Measurements for weakly
ionizable molecules can also be performed using a potentiometric titration method
(Avdeef, 1992; 1993, Slater et al. 1994).
1.1.2 Self-Association - This protocol is directed primarily towards neutral (non-
ionizable) organic compounds where self-association is generally not expected to
be of concern. In some situations, self association can arise for very high log Kow
solutes because rather high concentrations of the solute in very small amounts of
octanol will be required for the delivery of sufficient solute to the water phase for a
successful measurement. For these types of molecules, it will be very difficult, if
not impossible, to be sure that no self-association occurs in the octanol phase. Self
association can also arise for molecules which have H-bonding donor and
acceptor groups that could participate in such self-association at high
concentration in the octanol phase. For the latter group of molecules, e.g., amines,
carboxylic acids, and phenols, especially if cyclic dimers can form, measurements
need to be conducted at a sufficiently low concentration so that Kow reflects only
the unassociated form of the molecule in both water and octanol phases. In either
of the cases of self-association, it is recommended that measurements be
performed using several solute concentrations in the octanol and water phases. No
change in Kow with differing solute concentrations provides an indication that
measurements have been performed using the unassociated form of the molecule.
If the KOW decreases with decreasing solute concentration in octanol, extrapolation
to infinite dilution is suggested.
1.1.3 Tautomerization - This protocol is directed primarily towards neutral (non-
ionizable) organic compounds where tautomerization is generally not expected to
be of concern. The most common tautomerism (keto-enol tautomerism) involves
B-8
-------
structures with a -OH attached to a doubly-bonded carbon (enol) which rapidly
convert to the keto structure where the -OH becomes -C=O group and the
hydrogen attaches to the other carbon of previously existing doubly-bonded
carbon. If the molecule is likely to exist in more than one tautomeric forms, the
ratio of tautomers is often quite different in the octanol and water phases. The
measured value for a tautomeric chemical is meaningful. However, this value will
in most cases lie somewhere among the values for the individual tautomeric
forms; in essence, an average value for the ratio of tautomers is measured. The
individual values for the tautomeric forms most often will have to be calculated
because measurements can not be performed for the individual tautomeric forms
since individual tautomeric forms rapidly requilibrate to the tautomeric mixture.
Sometimes molecules exhibit both ionization and tautomerization, leading to
further complications.
1.1.4 Hydrate Formation - Similar to the case of tautomerization, hydrates may exist to
different degrees in the water and octanol phases thus confounding the
interpretation of the measured value.
1.1.5 Photodegradation - If the compound is expected to be light-sensitive and subject
to photodegradation, care should be taken to protect the substance from light
during the experiment.
1.2 Shake Flask or Slow-Stirring Considerations. (1) Water and octanol phases should be free
of impurities; (2) mixing should be of sufficient duration (e.g., 7 days for dioctyl
phthalate) to reach steady state equilibrium, particularly for very hydrophobic chemicals;
(3) when using volatile solutes, it is particularly important that both phases are analytically
measured; (4) avoid formation of emulsions during mixing and centrifuge before
measuring; (5) experimental protocol should be particularly scrutinized for Kow
measurements 4-6; (6) the ratio of octanol to water should be reduced for high Kow
chemicals; and (7) sorption to glass (e.g., for pyrethroids) during workup can be a
problem.
1.3 General Considerations. Solute should be stable to hydrolysis during the course of the
experiment. If stability can not be ensured, a calculated value may be used. Solutes
should be of high purity as the presence of a less lipophilic impurity exerts a dominant
effect in the measured Kow value. Mixtures such as chlorinated paraffins (containing
thousands of isomers, congeners, and degrees of chlorination) therefore cannot be
determined except by chromatographic methods.
1.4 Indicators of Potential Concern. Inconsistency with other measured values, with
estimated value, or inconsistency among estimated values. The importance of
professional judgement and knowledge of chemistry cannot be overemphasized in
making the best Kow assignments. For example, inconsistency between measured and
predicted may reflect only problems in the training set used based upon poor
experimental values when better data have since become available.
B-9
-------
1.5 References. Listed below are references for the shake flask and slow stir methods for
determining Kow used by various governmental agencies and specified in EPA testing
protocols. Selected references for measurement of ionizable compounds using
potentiometric titration are also included.
OECD. 1994. OECD Guidelines for Testing of Chemicals, 2nd Ed. Method: 107. Partition
Coefficient (n-octanol/water)(Flask-shaking Method (30 May 89).
US-Environmental Protection Agency. 1996. Product Properties Test Guidelines. OPPTS
830.7550. Partition Coefficient (n-Octanol/water), shake flask method. EPA 712-C-96-310
(1701). Public Draft.
OECD. 1998. OECD Guidelines for Testing of Chemicals, Partition Coefficient
(n-octanol/water), Slow-stirring method, Review Draft.
Avdeef, A. 1992. Quant. Struct.-Act. Relat. 11:510-517.
Avdeef, A. 1993. J. Pharm. Sci. 82:183-190.
Slater, B., A. McCormack, A.Avdeef, I.E.A. Comer. 1994. J. Pharm. Sci. 83:1280-1283.
2. ASSESSMENT OF Kow VALUES ESTIMATED FROM LIQUID
CHROMATOGRAPHIC TECHNIQUES
An estimated Kow value would be considered "appropriate" provided the following
experimental conditions existed during its determination:
2.1 Kow"s used for the reference compounds consist of "high quality" slow stir
measurements.
2.1.1 Better estimates for Kow's are obtained when reference and test chemicals are
similar.
2.1.2 When solutes have hydrogen accepting and/or amphiprotic substituents,
predictions of the log Kows from the log capacity factor (using relationships
developed with non-hydrogen bonding solutes) will generally result in predicted
log Kows which are too large and too small, respectively (Yamagami et al. 1994).
The chromatographic behavior for solutes containing hydrogen accepting and/or
amphiprotic substituents for the prediction of log Kow has been extensively studied
by Yamagami and coworkers. These studies concluded that "corrections for
hydrogen-bond effects are required in most cases when polar functional groups
are present" and that solvent composition in the chromatography system can
greatly change the capacity factors for these chemicals relative to non-hydrogen
bonding chemicals.
B-10
-------
2.2 A minimum of five chemicals are used in developing the log capacity factor (k1)- log Kow
calibration relationship. The Kow's of the reference chemicals should be evenly distributed
and should span 3 to 4 orders of magnitude.
2.3 The log k' - log Kow calibration curve is linear and has a correlation coefficient greater than
0.95.
2.4 The Kow estimated for the test chemical is within the range of Kows for the reference
compounds or does not exceed the upper end of the range of Kows for the reference
compounds beyond 0.5 log units without adequate justification.
2.5 Chemical speciation must be accounted for in performing the measurements. For
example, with ionizable chemicals, measurements must be performed on the unionized
form by using an appropriate buffer with a pH below the pK for an acid and above the pK
for a base.
2.6 Reference and test chemicals are of known purity and structure. Independent
confirmation of the identity and purity of the reference and test chemicals is required or
highly desirable.
2.7 Chemical mixtures can be used as the source of test chemicals provided accurate
identities can be assigned to individual chromatographic components.
2.8 References. These references for liquid chromatography techniques include methods
recommended by various governmental agencies that would provide "appropriate" Kows
when the reference compounds used in the determination are similar to the compound of
interest.
ASTM. 1997. Standard test method for partition coefficient (n-octanol/water) estimation
by liquid chromatography, Designation: E 1147 - 92. Annual Book of ASTM Standards,
Section 11, Water and Environmental Technology, Volume 11.05. ASTM, West
Conshohocken, PA.
OECD. 1994. OECD Guidelines for Testing of Chemicals, 2nd Ed. Method: 117. Partition
Coefficient (n-octanol/water), HPLC method (30 May 89).
U.S. Environmental Protection Agency. 1995. Product Properties Test Guidelines. OPPTS
830.7570. Partition Coefficient (n-Octanol/H2O), estimation by liquid chromatography.
EPA 712-C-95-040 (1701). Public Draft.
Yamagami, C., T. Ogura, N. Tako. 1990. J. Chromatogr. 514:123-126.
Yamagami, C., M. Yokota. 1991. Chem. Pharm. Bull. 39:1217-1221.
Yamagami, C., M. Yokota. 1991. Chem. Pharm. Bull. 39:2924-2929.
B-ll
-------
Yamagami, C., M. Yokota. 1992. Chem. Pharm. Bull. 40:925-929.
Yamagami, C., M. Yokota. 1993. Chem. Pharm. Bull. 41:694-698.
Yamagami, C., M. Yokota, N. Tako. 1994. J. Chromatogr. 662:49-60.
Yamagami, C., M. Yokota. 1994. Chem. Pharm. Bull. 42:907-912.
Yamagami, C., M. Yokota. 1995. Chem. Pharm. Bull. 43:2238-2242.
Yamagami, C., M. Yokota. 1996. Chem. Pharm. Bull. 44:1338-1343.
B-12
-------
ATTACHMENT II: ESTIMATION OF Kow FROM MOLECULAR FRAGMENTS
For computing thermodynamic properties it is often useful to consider a molecule as a
collection of molecular fragments, each making a distinct contribution to the property of interest,
which is relatively independent of the rest of the molecule. The rationale behind the method is
that a large number of structures can be generated from a relatively small number of fragments,
and thus a large number of estimates can be derived from a small number of experimentally
determined fragment constants. The accuracy of the estimation, however, necessarily improves as
the specificity of the fragment environment increases, which entails an increase in the number of
fragments or corrective factors that must be considered. This approach is applied at different
levels of sophistication. One user may employ a few fragment constants and generate 'first order'
estimates whereas another may make numerous corrections or adjustments reflecting more
fragment specificity for a given molecular environment. For a more complete discussion of group
fragment methods one should consult Hansch and Leo, 1995. For this exercise, these methods
will be used for molecule-to-molecule extrapolation (via addition or subtraction of
fragments\substituents) rather than a priori estimation.
1. Addition of ring fragments
For condensed ring aromatics, the addition of rings is given by
/" = log K (anthracene) - log K (naphthalene) « 1.20
= 0.5 ,log K^ (pyrene) - log K^ (naphthalene^ « 0.85
L = log K (pyrene) - log K (phenanthrene) « 0.50
where /« „ , /£. „ , fL are the fragment addition constants for • , • , and •
f"**j) f^ ^
condensation respectively.
2. Addition of substituents
The addition of a substituent, S (replacing a H atom) is a primary application of this
method. In this case
- H
B-13
-------
where R is the base molecule and TLg is a substituent constant, which is experimentally
determined. Tables for common substituents are readily available or can be easily determined
from measured data. One must distinguish (i.e., have different substituent constants for)
attachment to aliphatic, ethylenic, acetylenic, and aromatic carbon atoms in 'R'. Also corrections
must be made for multiple substitution if attachment is to the same or adjacent carbons. The
following is an example of Kow estimation. The fragment constant for Cl attached to aromatic
carbon can be derived from:
= log K-jj (chlorobenzen») - log K penzenfy * 0.71
With this constant, one can derive
log K-j (1,3.5- trichlorobenzen») * log K. (benzene^ + 3 ( 0.71) = 4.26
An exhaustive list of substituent constants is included in the aforementioned Hansch and Leo
(1995) reference.
It should be noted that • -values are most often illustrated, as above, by replacing a hydrogen
atom on a benzene 'parent' molecule. However, substituents that are strong electron donors or
acceptors such as chlorine have different • -values when placed on other 'parent' aromatic
molecules. For example, differences in log Kows of 0.71, 0.99, 0.71 and 0.85 are obtained between
the 4-chloro-analogues and their parent molecules for benzene, aniline, nitrobenzene, and
phenoxyacetic acid, respectively. The literature is replete with calculations making this type of
error and the importance of using the correct • -values for the 'parent' molecule can not be under
emphasized.
B-14
------- |