[Federal Register Volume 61, Number 175 (Monday, September 9, 1996)]
[Notices]
[Pages 47552-47631]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 96-22648]



[[Page 47551]]


_______________________________________________________________________

Part II





Environmental Protection Agency





_______________________________________________________________________



Proposed Guidelines for Ecological Risk Assessment; Notice

  Federal Register / Vol. 61, No. 175 / Monday, September 9, 1996 / 
Notices  

[[Page 47552]]



ENVIRONMENTAL PROTECTION AGENCY

[FRL-5605-9]


Proposed Guidelines for Ecological Risk Assessment

AGENCY: U.S. Environmental Protection Agency.

ACTION: Notice of Availability and Opportunity to Comment on Proposed 
Guidelines for Ecological Risk Assessment.

-----------------------------------------------------------------------

SUMMARY: The U.S. Environmental Protection Agency (EPA) is today 
publishing a document entitled Proposed Guidelines for Ecological Risk 
Assessment (hereafter ``Proposed Guidelines''). These Proposed 
Guidelines were developed as part of an interoffice Guidelines 
development program by a Technical Panel of the Risk Assessment Forum. 
The Proposed Guidelines expand upon the previously published EPA report 
Framework for Ecological Risk Assessment (EPA/630/R-92/001, February 
1992), while retaining the report's broad scope. When final, these 
Proposed Guidelines will help improve the quality of ecological risk 
assessments at EPA while increasing the consistency of assessments 
among the Agency's program offices and regions.

DATES: The Proposed Guidelines are being made available for a 90-day 
public review and comment period. Comments must be in writing and must 
be postmarked by December 9, 1996. See Addresses section for guidance 
on submitting comments.

FOR FURTHER INFORMATION CONTACT: Bill van der Schalie, National Center 
for Environmental Assessment-Washington Office, telephone: 202-260-
4191.

ADDRESSES: The Proposed Guidelines will be made available in the 
following ways:
    (1) The electronic version will be accessible on EPA's Office of 
Research and Development home page on the Internet at http://
www.epa.gov/ORD/WebPubs/fedreg.
    (2) 3\1/2\'' high-density computer diskettes in Wordperfect 5.1 
format will be available from ORD Publications, Technology Transfer and 
Support Division, National Risk Management Research Laboratory, 
Cincinnati, OH; telephone: 513-569-7562; fax: 513-569-7566. Please 
provide the EPA No. (EPA/630/R-95/002B) when ordering.
    (3) This notice contains the full proposed guideline. In addition, 
copies will be available for inspection at EPA headquarters and 
regional libraries, through the U.S. Government Depository Library 
program, and for purchase from the National Technical Information 
Service (NTIS), Springfield, VA; telephone: 703-487-4650, fax: 703-321-
8547. Please provide the NTIS No. PB96-193198; Price Code A13: ($47.00) 
when ordering.

Submitting Comments

    Comments on the Proposed Guidelines should be submitted to: U.S. 
Environmental Protection Agency, Air and Radiation Docket and 
Information Center (6102), Attn: File ORD-ERA-96-01, Waterside Mall, 
401 M St. SW, Washington, DC 20460. Please submit one unbound original 
with pages numbered consecutively, and three copies. For attachments, 
provide an index, number pages consecutively, provide comment on how 
the attachments relate to the main comment(s), and submit an unbound 
original and three copies. Please identify all comments and attachments 
with the file number ORD-ERA-96-01. Mailed comments must be postmarked 
by the date indicated. Comments may also be submitted electronically by 
sending electronic mail (e-mail) to: A-and-R-D[email protected]. 
Electronic comments must be submitted as an ASCII file avoiding the use 
of special characters and any form of encryption. Comments and data 
will also be accepted on disks in WordPerfect 5.1 file format or ASCII 
file format. All comments in electronic form also must be identified by 
the file number ORD-ERA-96-01.
    The Air and Radiation Docket and Information Center is open for 
public inspection and copying between 8:00 a.m. and 5:30 p.m., 
weekdays, in Room M-1500, Waterside Mall, 401 M St. SW, Washington, DC 
20460. The Center is located on the ground floor in the commercial area 
of Waterside Mall. The file index, materials, and comments are 
available for review in the information center or copies may be mailed 
on request from the Air and Radiation Docket and Information Center by 
calling (202) 260-7548 or -7549. The FAX number for the Center is (202) 
260-4400. A reasonable fee may be charged for copying materials.
    Please note that all technical comments received in response to 
this notice will be placed in the public record. For that reason, 
commentors should not submit personal information such as medical data 
or home addresses, confidential business information, or information 
protected by copyright. Due to limited resources, acknowledgments will 
not be sent.

SUPPLEMENTARY INFORMATION: These Proposed Guidelines are EPA's first 
Agency-wide ecological risk assessment guidelines. They are broad in 
scope, describing general principles and providing numerous examples to 
show how ecological risk assessment can be applied to a wide range of 
systems, stressors, and biological/spatial/temporal scales. This 
general approach provides sufficient flexibility to permit EPA's 
offices and regions to develop specific guidance suited to their 
particular needs. Because of their broad scope, the Proposed Guidelines 
do not provide detailed guidance in specific areas nor are they highly 
prescriptive. Frequently, rather than requiring that certain procedures 
always be followed, the Proposed Guidelines describe the strengths and 
limitations of alternate approaches. Agency preferences are expressed 
where possible, but because ecological risk assessment is a relatively 
new, rapidly evolving discipline, requirements for specific approaches 
could soon become outdated. EPA is working to expand the references in 
the Proposed Guidelines to include additional review articles or key 
publications that will help provide a ``window to the literature'' as 
recommended by peer reviewers. In the future, EPA intends to develop a 
series of shorter, more detailed guidance documents on specific 
ecological risk assessment topics after these Proposed Guidelines have 
been finalized.
    These Proposed Guidelines were prepared during a time of increasing 
interest in the field of ecological risk assessment and reflect input 
from many sources outside as well as inside the Agency. Over the last 
few years, the National Research Council proposed an ecological risk 
paradigm (NRC, 1993), there has been a marked increase in discussion of 
ecological risk assessment issues at meetings of professional 
organizations, and numerous articles and books on the subject have been 
published. Agency work on the Proposed Guidelines has proceeded in a 
step-wise fashion during this time. Preliminary work began in 1989 and 
included a series of colloquia sponsored by EPA's Risk Assessment Forum 
to identify and discuss significant issues in ecological risk 
assessment (U.S. EPA, 1991). Based on this early work and on a 
consultation with EPA's Science Advisory Board (SAB), the Agency 
decided to produce ecological risk assessment guidance sequentially, 
beginning with basic terms and concepts and continuing with the 
development of source materials for these Proposed Guidelines. The 
first product of this effort was the Risk Assessment Forum report, 
Framework

[[Page 47553]]

for Ecological Risk Assessment (Framework Report; U.S. EPA, 1992a,b), 
which proposes principles and terminology for the ecological risk 
assessment process. Since then, the Agency has solicited suggestions 
for ecological risk assessment guidelines structuring (U.S. EPA, 1992c) 
and has sponsored the development of other peer-reviewed materials, 
including ecological assessment case studies (U.S. EPA, 1993a, 1994a), 
and a set of issue papers that highlight important principles and 
approaches that EPA scientists should consider in preparing these 
Proposed Guidelines (U.S. EPA, 1994b,c).
    The nature and content of these Proposed Guidelines have been 
shaped by these documents as well as numerous meetings and discussions 
with individuals both within and outside of EPA. In late 1994 and early 
1995, the Agency solicited responses to the planned nature and 
structure of these Proposed Guidelines at three colloquia with Agency 
program offices and regions, other Federal agencies, and the public. 
Draft Proposed Guidelines were discussed at an external peer review 
workshop in December, 1995 (U.S. EPA, In Press). Subsequent reviews 
have included the Agency's Risk Assessment Forum and the Regulatory and 
Policy Development Committee, and interagency comment by members of 
subcommittees of the Committee on the Environment and Natural Resources 
of the Office of Science and Technology Policy. The EPA appreciates the 
efforts of all participants in the process and has tried to address 
their recommendations in these Proposed Guidelines.
    The EPA's Science Advisory Board will review these Proposed 
Guidelines at a future meeting. Following public and SAB reviews, 
Agency staff will prepare comment summaries. Appropriate comments will 
be incorporated, and the revised Guidelines will be submitted to EPA's 
Risk Assessment Forum for review. The Agency will consider comments 
from the public, the SAB, and the Risk Assessment Forum when finalizing 
these Proposed Guidelines.
    The public is invited to provide comments to be considered in EPA 
decisions about the content of the final Guidelines. EPA asks those who 
respond to this notice to include their views on the following:
    (1) Consistent with a recent National Research Council report (NRC, 
1996), these Proposed Guidelines emphasize the importance of 
interactions between risk assessors and risk managers as well as the 
critical role of problem formulation to ensuring that the results of 
the risk assessment can be used for decision-making. Overall, how 
compatible are these Proposed Guidelines with the National Research 
Council concept of the risk assessment process and the interactions 
between risk assessors, risk managers, and other interested parties?
    (2) The Proposed Guidelines are intended to provide a starting 
point for Agency program and regional offices that wish to prepare 
ecological risk assessment guidance suited to their needs. In addition, 
the Agency intends to sponsor development of more detailed guidance on 
certain ecological risk assessment topics. Examples might include 
identification and selection of assessment endpoints, selection of 
surrogate or indicator species, or the development and application of 
uncertainty factors. Considering the state of the science of ecological 
risk assessment and Agency needs and priorities, what topics most 
require additional guidance?
    (3) Some reviewers have suggested that the Proposed Guidelines 
should provide more discussion of topics related to the use of field 
observational data in ecological risk assessments, such as selection of 
reference sites, interpretation of positive and negative field data, 
establishing causal linkages, identifying measures of ecological 
condition, the role and uses of monitoring, and resolving conflicting 
lines of evidence between field and laboratory data. Given the general 
scope of these Proposed Guidelines, what, if any, additional material 
should be added on these topics and, if so, what principles should be 
highlighted?
    (4) The scope of the Proposed Guidelines is intentionally broad. 
However, while the intent is to cover the full range of stressors, 
ecosystem types, levels of biological organization, and spatial/
temporal scales, the contents of the Proposed Guidelines are limited by 
the present state of the science and the relative lack of experience in 
applying risk assessment principles to some areas. In particular, given 
the Agency's present interest in evaluating risks at larger spatial 
scales, how could the principles of landscape ecology be more fully 
incorporated into the Proposed Guidelines?
    (5) Assessing risks when multiple stressors are present is a 
challenging task. The problem may be how to aggregate risks 
attributable to individual stressors or to identify the principal 
stressors responsible for an observed effect. Although some approaches 
for evaluating risks associated with chemical mixtures are available, 
our ability to conduct risk assessments involving multiple chemical, 
physical, and biological stressors, especially at larger spatial 
scales, is limited. Consequently, the Proposed Guidelines primarily 
discuss predicting the effects of chemical mixtures and on general 
approaches for evaluating causality of an observed effect. What 
additional principles can be added?
    (6) Ecological risk assessments are frequently conducted in tiers 
that proceed from simple evaluations of exposure and effects to more 
complex assessments. While the Proposed Guidelines acknowledge the 
importance of tiered assessments, the wide range of applications of 
tiered assessments make further generalizations difficult. Given the 
broad scope of the Proposed Guidelines, what additional principles for 
conducting tiered assessments can be discussed?
    (7) Assessment endpoints are ``explicit expression of the 
environmental value that is to be protected''. As used in the Proposed 
Guidelines, assessment endpoints include both an ecological entity and 
a specific attributes of the entity (e.g., eagle reproduction or extent 
of wetlands). Some reviewers have recommended that assessment endpoints 
also include a decision criterion that is defined early in the risk 
assessment process (e.g., no more than a 20% reduction in reproduction, 
no more than a 10% loss of wetlands). While not precluding this 
possibility, the Proposed Guidelines suggest that such decisions are 
more appropriately made during discussions between risk assessors and 
managers in risk characterization at the end of the process. What are 
the relative merits of each approach?

    Dated: August 21, 1996.
Carol M. Browner,
Administrator.

Proposed Guidelines for Ecological Risk Assessment

Contents

Lists of Figures, Tables, and Text Notes
Executive Summary
1. Introduction
    1.1. Ecological Risk Assessment in a Management Context
    1.1.1. Contributions of Ecological Risk Assessment to 
Environmental Decisionmaking
    1.1.2. Risk Management Considerations
    1.2. Scope and Intended Audience
    1.3. Guidelines Organization
2. Planning the Risk Assessment: Dialogue Between Risk Managers and 
Risk Assessors
    2.1. Establishing Management Goals
    2.2. Management Decisions

[[Page 47554]]

    2.3. Scope and Complexity of the Risk Assessment
    2.4. Planning Outcome
3. Problem Formulation Phase
    3.1. Products of Problem Formulation
    3.2. Integration of Available Information
    3.3. Selecting Assessment Endpoints
    3.3.1. Selecting What to Protect
    3.3.1.1. Ecological Relevance
    3.3.1.2. Susceptibility to Known or Potential Stressors
    3.3.1.3. Representation of Management Goals
    3.3.2. Defining Assessment Endpoints
    3.4. Conceptual Models
    3.4.1. Risk Hypotheses
    3.4.2. Conceptual Model Diagrams
    3.4.3. Uncertainty in Conceptual Models
    3.5. Analysis Plan
    3.5.1. Selecting Measures
    3.5.2. Relating Analysis Plans to Decisions
4. Analysis Phase
    4.1. Evaluating Data and Models for Analysis
    4.1.1. Strengths and Limitations of Different Types of Data
    4.1.2. Evaluating Measurement or Modeling Studies
    4.1.2.1. Evaluating the Purpose and Scope of the Study
    4.1.2.2. Evaluating the Design and Implementation of the Study
    4.1.3. Evaluating Uncertainty
    4.2. Characterization of Exposure
    4.2.1. Exposure Analyses
    4.2.1.1. Describe the Source
    4.2.1.2. Describe the Distribution of the Stressor or Disturbed 
Environment
    4.2.1.3. Describe Contact or Co-occurrence
    4.2.2. Exposure Profile
    4.3. Characterization of Ecological Effects
    4.3.1. Ecological Response Analysis
    4.3.1.1. Stressor-Response Analysis
    4.3.1.2. Establishing Cause and Effect Relationships (Causality)
    4.3.1.3. Linking Measures of Effect to Assessment Endpoints
    4.3.2. Stressor-Response Profile
5. Risk Characterization
    5.1. Risk Estimation
    5.1.1. Risk Estimates Expressed as Qualitative Categories
    5.1.2. Single-Point Estimates
    5.1.3. Estimates Incorporating the Entire Stressor-Response 
Relationship
    5.1.4. Estimates Incorporating Variability in Exposure or 
Effects
    5.1.5. Estimates Based on Process Models
    5.1.6. Field Observational Studies
    5.2. Risk Description
    5.2.1. Lines of Evidence
    5.2.2. Determining Ecological Adversity
    5.3. Reporting Risks
6. Relating Ecological Information to Risk Management Decisions
7. Text Notes
Appendix A: Changes from EPA's Ecological Risk Assessment Framework
Appendix B: Key Terms
Appendix C: Conceptual Model Examples
Appendix D: Analysis Phase Examples
Appendix E: Criteria for Determining Ecological Adversity: a 
Hypothetical Example
References

List of Figures

 Figure 1-1. The framework for ecological risk assessment.
Figure 1-2. The ecological risk assessment framework, with an 
expanded view of each phase.
Figure 3-1. Problem formulation phase.
Figure 3-2. Elements of a conceptual model diagram.
Figure 4-1. Analysis phase.
Figure 4-2. A simple example of a stressor-response relationship.
Figure 5-1. Risk characterization.
Figure 5-2. Risk estimation techniques. a. Comparison of exposure 
and stressor-response point estimates. b. Comparison of point 
estimates from the stressor-response relationship with uncertainty 
associated with an exposure point estimate.
Figure 5-3. Risk estimation techniques: comparison of point 
estimates with associated uncertainties.
Figure 5-4. Risk estimation techniques: stressor-response curve 
versus a cumulative distribution of exposures.
Figure 5-5. Risk estimation techniques: comparison of exposure 
distribution of an herbicide in surface waters with freshwater 
single species toxicity data.

List of Tables

Table 4-1. Uncertainty Evaluation in the Analysis Phase.

List of Text Notes

Text Note 1-1. Related Terminology.
Text Note 1-2. Flexibility of the Framework Diagram.
Text Note 1-3. The Iterative Nature of Ecological Risk Assessment.
Text Note 2-1. Who Are Risk Managers?
Text Note 2-2. Who Are Risk Assessors?
Text Note 2-3. Questions Addressed by Risk Managers and Risk 
Assessors.
Text Note 2-4. The Role of Interested Parties.
Text Note 2-5. Sustainability as a Management Goal.
Text Note 2-6. Management Goals for Waquoit Bay.
Text Note 2-7. Questions to Ask About Scope and Complexity.
Text Note 3-1. Avoiding Potential Shortcomings Through Problem 
Formulation.
Text Note 3-2. Uncertainty in Problem Formulation.
Text Note 3-3. Initiating a Risk Assessment: What's Different When 
Stressors, Effects, or Values Drive the Process?
Text Note 3-4. Assessing Available Information: Questions to Ask 
Concerning Source, Stressor, and Exposure Characteristics, Ecosystem 
Characteristics, and Effects.
Text Note 3-5. Salmon and Hydropower: Why Salmon Would Provide the 
Basis for an Assessment Endpoint.
Text Note 3-6. Cascading Adverse Effects: Primary (Direct) and 
Secondary (Indirect).
Text Note 3-7. Sensitivity and Secondary Effects: The Mussel-Fish 
Connection.
Text Note 3-8. Examples of Management Goals and Assessment 
Endpoints.
Text Note 3-9. Common Problems in Selecting Assessment Endpoints.
Text Note 3-10. What Are Risk Hypotheses and Why Are They Important?
Text Note 3-11. Examples of Risk Hypotheses.
Text Note 3-12. What Are the Benefits of Develoving Conceptual 
Models?
Text Note 3-13. Uncertainty in Problem Formulation.
Text Note 3-14. Examples of Assessment Endpoints and Measures.
Text Note 3-15. Selecting What to Measure.
Text Note 3-16. How Do Water Quality Criteria Relate to Assessment 
Endpoints?
Text Note 3-17. Data Quality Objectives (DQO) Process.
Text Note 4-1. Data Collection and the Analysis Phase.
Text Note 4-2. The American National Standard for Quality Assurance
Text Note 4-3. Questions for Evaluating a Study's Utility for Risk 
Assessment.
Text Note 4-4. Considering the Degree of Aggregation in Models.
Text Note 4-5. Questions for Source Description.
Text Note 4-6. Questions to Ask in Evaluating Stressor Distribution.
Text Note 4-7. General Mechanisms of Transport and Dispersal.
Text Note 4-8. Questions to Ask in Describing Contact or Co-
occurrence.
Text Note 4-9. Example of an Exposure Equation: Calculating a 
Potential Dose via Ingestion.
Text Note 4-10. Measuring Internal Dose Using Biomarkers and Tissue 
Residues.
Text Note 4-11. Questions Addressed by the Exposure Profile.
Text Note 4-12. Questions for Stressor-Response Analysis.
Text Note 4-13. Qualitative Stressor-Response Relationships.
Text Note 4-14. Median Effect Levels.
Text Note 4-15. No-Effect Levels Derived From Statistical Hypothesis 
Testing.
Text Note 4-16. General Criteria for Causality.
Text Note 4-17. Koch's Postulates.
Text Note 4-18. Examples of Extrapolations to Link Measures of 
Effect to Assessment Endpoints.
Text Note 4-19. Questions Related to Selecting Extrapolation 
Approaches.
Text Note 4-20. Questions to Consider When Extrapolating From 
Effects Observed in the Laboratory to Field Effects of Chemicals.
Text Note 4-21. Questions Addressed by the Stressor-Response 
Profile.
Text Note 5-1. Using Qualitative Categories to Estimate Risks of an 
Introduced Species.
Text Note 5-2. Applying the Quotient Method.
Text Note 5-3. Comparing an Exposure Distribution With a Point 
Estimate of Effects.
Text Note 5-4. Comparing Cumulative Exposure and Effects 
Distributions for Chemical Stressors.
Text Note 5-5. Estimating Risk With Process Models.
Text Note 5-6. An Example of Field Methods Used for Risk Estimation.

[[Page 47555]]

Text Note 5-7. What Are Statistically Significant Effects?
Text Note 5-8. Possible Risk Assessment Report Elements.
Text Note 5-9. Clear, Transparent, Reasonable, and Consistent Risk 
Characterizations.
Text Note 6-1. Questions Regarding Risk Assessment Results.
Text Note 6-2. Risk Communication Considerations for Risk Managers.

Executive Summary

    The ecological problems facing environmental scientists and 
decisionmakers are numerous and varied. Growing concern over potential 
global climate change, loss of biodiversity, acid precipitation, 
habitat destruction, and the effects of multiple chemicals on 
ecological systems has highlighted the need for flexible problem-
solving approaches that can link ecological measurements and data with 
the decisionmaking needs of environmental managers. Increasingly, 
ecological risk assessment is being suggested as a way to address this 
wide array of ecological problems.
    Ecological risk assessment ``evaluates the likelihood that adverse 
ecological effects may occur or are occurring as a result of exposure 
to one or more stressors'' (U.S. EPA, 1992a). It is a process for 
organizing and analyzing data, information, assumptions, and 
uncertainties to evaluate the likelihood of adverse ecological effects. 
Ecological risk assessment provides a critical element for 
environmental decisionmaking by giving risk managers an approach for 
considering available scientific information along with the other 
factors they need to consider (e.g., social, legal, political, or 
economic) in selecting a course of action.
    To help improve the quality and consistency of EPA's ecological 
risk assessments, EPA's Risk Assessment Forum initiated development of 
these guidelines. The primary audience for this document is risk 
assessors and risk managers at EPA, although these guidelines may be 
useful to others outside the Agency (e.g., Agency contractors, state 
agencies, and other interested parties). These guidelines are based on 
and replace the 1992 report, Framework for Ecological Risk Assessment 
(referred to as the Framework Report). They were written by a Forum 
work group and have been extensively revised based on comments from 
outside peer reviewers as well as Agency staff. The guidelines retain 
the Framework Report's broad scope, while expanding on some framework 
concepts and modifying others to reflect Agency experiences. EPA 
intends to follow these guidelines with a series of shorter, more 
detailed documents that address specific ecological risk assessment 
topics. This ``bookshelf'' approach provides the flexibility necessary 
to keep pace with developments in the rapidly evolving field of 
ecological risk assessment while allowing time to form consensus, where 
appropriate, on science policy inferences (default assumptions) to 
bridge gaps in knowledge.
    Ecological risk assessment includes three primary phases (problem 
formulation, analysis, and risk characterization). Within problem 
formulation, important areas include identifying goals and assessment 
endpoints, preparing the conceptual model, and developing an analysis 
plan. The analysis phase involves evaluating exposure to stressors and 
the relationship between stressor levels and ecological effects. In 
risk characterization, key elements are estimating risk through 
integration of exposure and stressor-response profiles, describing risk 
by discussing lines of evidence and determining ecological adversity, 
and preparing a report. The interface between risk assessors and risk 
managers at the beginning and end of the risk assessment is critical 
for ensuring that the results of the assessment can be used to support 
a management decision.
    Both risk assessors and risk managers bring valuable perspectives 
to the initial planning activities for an ecological risk assessment. 
Risk managers charged with protecting environmental values can ensure 
that the risk assessment will provide information relevant to a 
decision. Ecological risk assessors ensure that science is effectively 
used to address ecological concerns. Both evaluate the potential value 
of conducting a risk assessment to address identified problems. Further 
objectives of the initial planning process are to establish management 
goals that are agreed upon, clearly articulated, and contain a way to 
measure success; determine the purpose for the risk assessment by 
defining the decisions to be made within the context of the management 
goals; and agree upon the scope, complexity, and focus of the risk 
assessment, including the expected output and available resources.
    Problem formulation, which follows these planning discussions, 
provides a foundation upon which the entire risk assessment depends. 
Successful completion of problem formulation depends on the quality of 
three products: assessment endpoints, conceptual models, and an 
analysis plan. Since problem formulation is inherently interactive and 
iterative, not linear, substantial reevaluation is expected to occur 
within and among all products of problem formulation.
    Assessment endpoints are ``explicit expressions of the actual 
environmental value that is to be protected'' (U.S. EPA, 1992a) that 
link the risk assessment to management concerns. Assessment endpoints 
include both a valued ecological entity and an attribute of that entity 
that is important to protect and potentially at risk (e.g., nesting and 
feeding success of piping plovers or areal extent and patch size of 
eelgrass). For a risk assessment to have scientific validity, 
assessment endpoints must be ecologically relevant to the ecosystem 
they represent and susceptible to the stressors of concern. Assessment 
endpoints that represent societal values and management goals are more 
effective in that they increase the likelihood that the risk assessment 
will be used in management decisions. Assessment endpoints that fulfill 
all three criteria provide the best foundation for an effective risk 
assessment.
    Potential interactions between assessment endpoints and stressors 
are explored by developing a conceptual model. Conceptual models link 
anthropogenic activities with stressors and evaluate interrelationships 
between exposure pathways, ecological effects, and ecological 
receptors. Conceptual models include two principal components: risk 
hypotheses and a conceptual model diagram.
    Risk hypotheses describe predicted relationships between stressor, 
exposure, and assessment endpoint response. Risk hypotheses are 
hypotheses in the broad scientific sense; they do not necessarily 
involve statistical testing of null and alternative hypotheses or any 
particular analytical approach. Risk hypotheses may predict the effects 
of a stressor (e.g., a chemical release) or they may postulate what 
stressors may have caused observed ecological effects. Key risk 
hypotheses are identified for subsequent evaluation in the risk 
assessment.
    A useful way to express the relationships described by the risk 
hypotheses is through a diagram of a conceptual model. Conceptual model 
diagrams are useful tools for communicating important pathways in a 
clear and concise way and for identifying major sources of uncertainty. 
Risk assessors can use these diagrams and risk hypotheses to identify 
the most important pathways and relationships that will be evaluated in 
the analysis phase. Risk assessors justify what will be done as well as 
what will not be done in the assessment in an analysis plan.

[[Page 47556]]

The analysis plan also describes the data and measures to be used in 
the risk assessment and how risks will be characterized.
    The analysis phase, which follows problem formulation, includes two 
principal activities: characterization of exposure and characterization 
of ecological effects. The process is flexible, and interaction between 
the ecological effects and exposure evaluations is recommended. Both 
activities include an evaluation of available data for scientific 
credibility and relevance to assessment endpoints and the conceptual 
model. In exposure characterization, data analyses describe the 
source(s) of stressors, the distribution of stressors in the 
environment, and the contact or co-occurrence of stressors with 
ecological receptors. In ecological effects characterization, data 
analyses may evaluate stressor-response relationships or evidence that 
exposure to a stressor causes an observed response.
    The products of analysis are summary profiles that describe 
exposure and the stressor-response relationships. Exposure and 
stressor-response profiles may be written documents or modules of a 
larger process model. Alternatively, documentation may be deferred 
until risk characterization. In any case, the objective is to ensure 
that the information needed for risk characterization has been 
collected and evaluated.
    The exposure profile identifies receptors and exposure pathways and 
describes the intensity and spatial and temporal extent of exposure. 
The exposure profile also describes the impact of variability and 
uncertainty on exposure estimates and reaches a conclusion about the 
likelihood that exposure will occur.
    The stressor-response profile may evaluate single species, 
populations, general trophic levels, communities, ecosystems, or 
landscapes--whatever is appropriate for the assessment endpoints. For 
example, if a single species is affected, effects should represent 
appropriate parameters such as effects on mortality, growth, and 
reproduction, while at the community level, effects may be summarized 
in terms of structure or function depending on the assessment endpoint. 
The stressor-response profile summarizes the nature and intensity of 
effect(s), the time scale for recovery (where appropriate), causal 
information linking the stressor with observed effects, and 
uncertainties associated with the analysis.
    Risk characterization is the final phase of an ecological risk 
assessment. During risk characterization, risks are estimated and 
interpreted and the strengths, limitations, assumptions, and major 
uncertainties are summarized. Risks are estimated by integrating 
exposure and stressor-response profiles using a wide range of 
techniques such as comparisons of point estimates or distributions of 
exposure and effects data, process models, or empirical approaches such 
as field observational data.
    Risk assessors describe risks by evaluating the evidence supporting 
or refuting the risk estimate(s) and interpreting the adverse effects 
on the assessment endpoint. Criteria for evaluating adversity include 
the nature and intensity of effects, spatial and temporal scales, and 
the potential for recovery. Agreement among different lines of evidence 
of risk increases confidence in the conclusions of a risk assessment.
    When risk characterization is complete, a report describing the 
risk assessment can be prepared. The report may be relatively brief or 
extensive depending on the nature and the resources available for the 
assessment and the information required to support a risk management 
decision. Report elements may include:
     A description of risk assessor/risk manager planning 
results.
     A review of the conceptual model and the assessment 
endpoints.
     A discussion of the major data sources and analytical 
procedures used.
     A review of the stressor-response and exposure profiles.
     A description of risks to the assessment endpoints, 
including risk estimates and adversity evaluations.
     A summary of major areas of uncertainty and the approaches 
used to address them.
     A discussion of science policy judgments or default 
assumptions used to bridge information gaps, and the basis for these 
assumptions.

To facilitate understanding, risk assessors should characterize risks 
``in a manner that is clear, transparent, reasonable, and consistent 
with other risk characterizations of similar scope prepared across 
programs in the Agency'' (U.S. EPA, 1995c).
    After the risk assessment is completed, risk managers may consider 
whether additional follow-up activities are required. Depending on the 
importance of the assessment, confidence level in the assessment 
results, and available resources, it may be advisable to conduct 
another iteration of the risk assessment in order to facilitate a final 
management decision. Ecological risk assessments are frequently 
designed in sequential tiers that proceed from simple, relatively 
inexpensive evaluations to more costly and complex assessments. Initial 
tiers are based on conservative assumptions, such as maximum exposure 
and ecological sensitivity. When an early tier cannot sufficiently 
define risk to support a management decision, a higher assessment tier 
that may require either additional data or applying more refined 
analysis techniques to available data may be needed. Higher tiers 
provide more ecologically realistic assessments while making less 
conservative assumptions about exposure and effects.
    Another option is to proceed with a management decision based on 
the risk assessment and develop a monitoring plan to evaluate the 
results of the decision. For example, if the decision was to mitigate 
risks through exposure reduction, monitoring could help determine 
whether the desired reduction in exposure (and effects) was achieved. 
Monitoring is also critical for determining the extent and nature of 
any ecological recovery that may be occurring. Experience obtained by 
using focused monitoring results to evaluate risk assessment 
predictions can help improve the risk assessment process and is 
encouraged.
    Communicating ecological risks to the public is usually the 
responsibility of risk managers. Although the final risk assessment 
document (including its risk characterization sections) can be made 
available to the public, the risk communication process is best served 
by tailoring information to a particular audience. It is important to 
clearly describe the ecological resources at risk, their value, and the 
costs of protecting (and failing to protect) the resources (U.S. EPA, 
1995c). The degree of confidence in the risk assessment and the 
rationale for risk management decisions and options for reducing risk 
are also important (U.S. EPA, 1995c).

1. Introduction

    Ecological risk assessment is a process for organizing and 
analyzing data, information, assumptions, and uncertainties to evaluate 
the likelihood of adverse ecological effects. Ecological risk 
assessment provides a critical element for environmental 
decisionmaking. This document, which is structured by the stages of the 
ecological risk assessment process, provides Agency personnel with 
broad guidelines that can be adapted to their specific requirements.
    The full definition of ecological risk assessment is:

[[Page 47557]]

    ``The process that evaluates the likelihood that adverse ecological 
effects may occur or are occurring as a result of exposure to one or 
more stressors.'' (U.S. EPA, 1992a)
    Several terms within this definition require further explanation:
     `` * * * likelihood * * * '' Descriptions of risk may 
range from qualitative judgments to quantitative probabilities. While 
risk assessments may include quantitative risk estimates, the present 
state of the science often may not support such quantitation. It is 
preferable to convey qualitatively the relative magnitude of 
uncertainties to a decision maker than to ignore them because they may 
not be easily understood or estimated.
     `` * * * adverse ecological effects * * * '' Ecological 
risk assessments deal with anthropogenic changes that are considered 
undesirable because they alter valued structural or functional 
characteristics of ecological systems. An evaluation of adversity may 
consider the type, intensity, and scale of the effect as well as the 
potential for recovery.
     `` * * * may occur or are occurring * * * '' Ecological 
risk assessments may be prospective or retrospective. Retrospective 
ecological risk assessments evaluate the likelihood that observed 
ecological effects are associated with previous or current exposures to 
stressors. Many of the same methods and approaches are used for both 
prospective and retrospective assessments, and in the best case, even 
retrospective assessments contain predictive elements linking sources, 
stressors and effects.
     `` * * * one or more stressors * * * '' Ecological risk 
assessments may address single or multiple chemical, physical, or 
biological stressors. (See Appendix A for definitions of stressor 
types.) Because risk assessments are conducted to provide input to 
management decisions, this document focuses on stressors generated or 
influenced by anthropogenic activity.
    The overall ecological risk assessment process is shown in figure 
1-1.1 Problem formulation is the first phase of the process where 
the assessment purpose is stated, the problem defined, and the plan for 
analyzing and characterizing risk determined. In the analysis phase, 
data on potential effects of and exposures to stressor(s) identified 
during problem formulation are technically evaluated and summarized as 
exposure and stressor-response profiles. These profiles are integrated 
in risk characterization to estimate the likelihood of adverse 
ecological effects. Major uncertainties, assumptions, and strengths and 
limitations of the assessment are summarized during this phase. While 
discussions between risk assessors and risk managers are emphasized 
both at risk assessment initiation (planning) and completion 
(communicating results), these guidelines maintain a distinction 
between risk assessment and risk management. Risk assessment focuses on 
evaluating the likelihood of adverse effects, and risk management 
involves the selection of a course of action in response to an 
identified risk that is based on many factors (e.g., social, legal, 
political, or economic) in addition to the risk assessment results. 
Section 1.1 briefly discusses how risk assessments fit into a 
decisionmaking context.
---------------------------------------------------------------------------

    \1\ Changes in process and terminology from EPA's previous 
ecological risk assessment framework (U.S. EPA, 1992a) are 
summarized in Appendix A.

BILLING CODE 6560-50-P

[[Page 47558]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.000



BILLING CODE 6560-50-C

[[Page 47559]]

    The bar along the right side of figure 1-1 shows several activities 
that are associated with risk assessments: data acquisition, iteration, 
and monitoring. While the risk assessment may focus on data analysis 
and interpretation, acquiring the appropriate quantity and quality of 
data for use in the process is critical. If such data are lacking, the 
risk assessment may stop until the necessary data are acquired. As 
discussed in text note 1-3, the process is more frequently iterative 
than linear, since the evaluation of new data or information may 
require revisiting a part of the process or conducting a new 
assessment.
    Monitoring data can provide important input to all phases of the 
risk assessment process. For example, monitoring can provide the 
impetus for initiating a risk assessment by identifying changes in 
ecological condition. In addition, monitoring data can be used to 
evaluate the results predicted by the risk assessment. For example, 
follow-up studies could be used to determine whether techniques used to 
mitigate pesticide exposures in field situations in fact reduce 
exposure and effects as predicted by the risk assessment. Or, for a 
hazardous waste site, monitoring might help verify whether source 
reduction resulted in anticipated ecological changes. Monitoring is 
also critical for determining the extent and nature of any ecological 
recovery that may occur. The experience gained by comparing monitoring 
results to evaluate risk assessment predictions can help improve the 
risk assessment process and is encouraged.

1.1. Ecological Risk Assessment in a Management Context

    Ecological risk assessment is important for environmental 
decisionmaking because of the high cost of eliminating environmental 
risks associated with human activities and the necessity of making 
regulatory decisions in the face of uncertainty (Ruckelshaus, 1983; 
Suter, 1993a). Even so, ecological risk assessment provides only a 
portion of the information required to make risk management decisions. 
This section describes how ecological risk assessments fit into a 
larger management framework.
1.1.1. Contributions of Ecological Risk Assessment to Environmental 
Decisionmaking
    At EPA, ecological risk assessments provide input to a diverse set 
of environmental decisionmaking processes, such as the regulation of 
hazardous waste sites, industrial chemicals, and pesticides, or the 
management of watersheds affected by multiple nonchemical and chemical 
stressors. The ecological risk assessment process has several features 
that contribute to managing ecological risks:
     In a risk assessment, changes in ecological effects can be 
expressed as a function of changes in exposure to a stressor. This 
inherently predictive aspect of risk assessment may be particularly 
useful to the decision maker who must evaluate tradeoffs and examine 
different alternatives.
     Risk assessments include an explicit evaluation of 
uncertainties. Uncertainty analysis lends credibility and a degree of 
confidence to the assessment that can strengthen its use in 
decisionmaking and can help the risk manager focus research on those 
areas that will lead to the greatest reductions in uncertainty.
     Risk assessments can provide a basis for comparing, 
ranking, and prioritizing risks. The risk manager can use such 
information to help decide among several management alternatives.
     Risk assessments emphasize consistent use of well-defined 
and relevant endpoints. This is especially important for ensuring that 
the results of the risk assessment will be expressed in a way that the 
risk manager can use.
1.1.2. Risk Management Considerations
    Although risk assessors and risk managers interact both at the 
initiation and completion of an ecological risk assessment (sections 2, 
3, 5 and 6), risk managers decide how to use the results of an 
assessment and whether a risk assessment should be conducted. While a 
detailed review of management issues is beyond the scope of these 
guidelines, key areas are highlighted below.
     A risk assessment is not always required for management 
action. When faced with compelling ecological risks and an immediate 
need to make a decision, a risk manager might proceed without an 
assessment, depending on professional judgment and statutory 
requirements (U.S. EPA, 1992a).
     Because initial management decisions or statutory 
requirements significantly affect the scope of an assessment, it is 
important, where possible, for risk managers to consider a broader 
scope or alternative actions for a risk assessment. Sometimes a 
particular statute may require the risk assessment to focus on one type 
of stressor (e.g., chemicals) when there are other, perhaps more 
important, stressors in the system (e.g., habitat alteration). In other 
situations, however, it may be possible to evaluate a range of options. 
For example, before requesting an ecological risk assessment of 
alternative sites for the construction and operation of a dam for 
hydroelectric power, risk managers may consider larger issues such as 
the need for the additional power and the feasibility of using other 
power-generating options.
     Risk managers consider many factors in making regulatory 
decisions. Legal mandates may require the risk manager to take certain 
courses of action. Political and social considerations may lead the 
risk manager to make decisions that are either more or less 
ecologically protective. Economic factors may also be critical. For 
example, a course of action that has the least ecological risk may be 
too expensive or technologically infeasible. If cost-benefit analysis 
is applied, ecological risks may be translated into monetary terms to 
be compared against other monetary considerations. Thus, while 
ecological risk assessment provides critical information to risk 
managers, it is only part of the whole environmental decisionmaking 
process.

1.2. Scope and Intended Audience

    These guidelines replace the EPA report, Framework for Ecological 
Risk Assessment (referred to as the Framework Report, U.S. EPA, 1992a). 
As a next step in developing Agency-wide guidance, the guidelines 
expand on and modify framework concepts to reflect Agency experience in 
the several years since the Framework Report was published (see 
Appendix A). Like the Framework Report, these guidelines are broad in 
scope, describing general principles and providing numerous examples to 
show how ecological risk assessment can be applied to a wide range of 
systems, stressors, and biological, spatial, and temporal scales. This 
approach provides flexibility to permit EPA's offices and regions to 
develop specific guidance suited to their particular needs.
    The proposed policies set out in this document are intended as 
internal guidance for EPA. Risk assessors and risk managers at EPA are 
the primary audience for this document, although these guidelines may 
be useful to others outside the Agency (e.g., Agency contractors, state 
agencies, and other interested parties). These Proposed Guidelines are 
not intended, nor can they be relied upon, to create any rights 
enforceable by any party in litigation with the United States. This 
document is not a regulation and is not intended for EPA regulations. 
These Proposed Guidelines set forth current scientific thinking and 
approaches for conducting and evaluating ecological risk

[[Page 47560]]

assessments. As with other EPA guidelines (developmental toxicity, 56 
FR 63798-63826; exposure assessment, 57 FR 22888-22938; and 
carcinogenicity, 61 FR 17960-18011), EPA will revisit these guidelines 
as experience and scientific consensus evolves.
    These guidelines do not provide detailed guidance in specific areas 
nor are they intended to be highly prescriptive. These guidelines 
describe the strengths and limitations of alternate approaches and may 
not apply to a particular situation based upon the circumstances. 
Agency preferences are expressed where possible, but because ecological 
risk assessment is a rapidly evolving discipline, requirements for 
specific approaches could soon become outdated. EPA intends to develop 
a series of shorter, more detailed guidance documents on specific 
ecological risk assessment topics after these guidelines have been 
finalized.
    These guidelines emphasize processes and approaches for analyzing 
data rather than specific data collection techniques, methods, or 
models. Also, while these guidelines discuss the interface between the 
risk assessor and risk manager, a detailed discussion of the use of 
ecological risk assessment information in the risk management process 
(e.g., the economic, legal, political, or social implications of the 
risk assessment results) is beyond the scope of these guidelines. Other 
EPA publications discuss how ecological concerns have been addressed in 
decisionmaking at EPA (U.S. EPA, 1994g) and provide an introduction to 
ecological risk assessment for risk managers (U.S. EPA, 1995b).

1.3. Guidelines Organization

    These guidelines are structured according to the ecological risk 
assessment process as shown in figure 1-2. Within problem formulation 
(section 3), important areas addressed include identifying goals and 
assessment endpoints, preparing the conceptual model, and developing an 
analysis plan. The analysis phase (section 4) involves evaluating 
exposure to stressors and the relationship between stressor levels and 
ecological effects. In risk characterization (section 5), key elements 
are estimating risk through integration of exposure and stressor-
response profiles and describing risk by discussing lines of evidence, 
interpreting adversity, and summarizing uncertainty. In addition, 
discussions between the risk assessor and risk manager at the beginning 
(section 2) and end of the risk assessment (section 6) are highlighted.

BILLING CODE 6560-50-P

[[Page 47561]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.001



BILLING CODE 6560-50-C

[[Page 47562]]

    The reader may notice that cross-cutting topics are covered in 
several sections. These include uncertainty, models, evaluating data, 
causality, linking measures of effect to assessment endpoints, and 
identifying ecological effects. Considerations appropriate to the 
different phases of ecological risk assessment are discussed.

2. Planning The Risk Assessment: Dialogue Between Risk Managers and 
Risk Assessors

    The purpose for an ecological risk assessment is to produce a 
scientific evaluation of ecological risk that enables managers to make 
informed environmental decisions. To ensure that ecological risk 
assessments meet risk managers' needs, a planning dialogue between risk 
managers and risk assessors (see text notes 2-1 and 2-2) is a critical 
first step toward initiating problem formulation and plays a continuing 
role during the conduct of the risk assessment. Planning is the 
beginning of a necessary interface between risk managers and risk 
assessors and is represented by a side box in the ecological risk 
assessment diagram (see figure 1-2). It is due to the importance of 
planning and the significant role it plays in ecological risk 
assessments that this section on planning is incorporated into 
guidelines on ecological risk assessment. However, it is imperative to 
remember that the planning process is distinct from the scientific 
conduct of an ecological risk assessment. This distinction helps ensure 
that political and social issues, while helping to define the 
objectives for the risk assessment, do not bias the scientific 
evaluation of risk.
    During the planning dialogue, risk managers and risk assessors each 
bring important perspectives to the table. In general, risk managers 
are charged with protecting societal values (e.g., human health and the 
environment) and must ensure that the risk assessment will provide 
information relevant to a decision. To meet this charge, risk managers 
describe why the risk assessment is needed, what decisions it will 
support, and what they want to receive from the risk assessor. It is 
also helpful for managers to consider what problems they have 
encountered in the past when trying to use risk assessments for 
decisionmaking. In turn, it is the ecological risk assessors' role to 
ensure that science is effectively used to address ecological concerns. 
Risk assessors describe what they can provide to the risk manager, 
where problems are likely to occur, and where uncertainty may be 
problematic. Both evaluate the potential value of conducting a risk 
assessment to address identified problems.
    Both risk managers and risk assessors are responsible for coming to 
agreement on the goals, scope, and timing of a risk assessment and the 
resources that are available and necessary to achieve the goals. 
Together they use information on the area's ecosystems, regulatory 
endpoints, and publicly perceived environmental values to interpret the 
goals for use in the ecological risk assessment. Examples of questions 
risk managers and risk assessors may address during planning are 
provided in text note 2-3.
    The first step in planning may be to determine if a risk assessment 
is the best option for making the decision required. Questions 
concerning what is known about the degree of risk, what management 
options are available to mitigate or prevent it, and the value of 
conducting a risk assessment compared with other ways of learning about 
and addressing environmental concerns are asked during these 
discussions. In some cases, a risk assessment may add little value to 
the decision process. It is important for the risk manager and risk 
assessor to explore alternative options for addressing possible risk 
before continuing to the next planning stage (see section 1.1.2).
    Once the decision is made to conduct a risk assessment, planning 
focuses on (1) establishing management goals that are agreed on, 
clearly articulated, and contain a way to measure success; (2) defining 
the decisions to be made within the context of the management goals; 
and (3) agreeing on the scope, complexity, and focus of the risk 
assessment, including the expected output and the technical and 
financial support available to complete it. To achieve these 
objectives, risk managers and risk assessors must each play an active 
role in planning the risk assessment.

2.1. Establishing Management Goals

    Management goals for a risk assessment are established by risk 
managers but are derived in a variety of ways. Many Agency risk 
assessments are conducted based on legally established management goals 
(e.g., national regulatory programs generally have management goals 
written into the law governing the program). In this case, goal setting 
was previously completed through public debate in establishing the law. 
In most cases, legally established management goals do not provide 
sufficient guidance to the risk assessor. For example, the objectives 
under the Clean Water Act to ``protect and maintain the chemical, 
physical and biological integrity of the nation's waters'' are open to 
considerable interpretation. Agency managers and staff often interpret 
the law in regulations and guidance. Significant interaction between 
the risk assessor and risk manager may be needed to translate the law 
into management goals for a particular location or circumstance.
    As the Agency increasingly emphasizes ``place-based'' or 
``community-based'' management of ecological resources as recommended 
in the Edgewater Consensus (U.S. EPA, 1994e), management goals take on 
new significance for the ecological risk assessor. Management goals for 
``places'' such as watersheds are formed as a consensus based on 
diverse values reflected in Federal, state, and local regulations; 
constituency group agendas; and public concerns. Significant 
interactions among a variety of interested parties are required to 
generate agreed-on management goals for the resource (see text note 2-
4). Public meetings, constituency group meetings, evaluation of 
resource management organization charters, and other means of looking 
for management goals shared by these diverse groups may be necessary. 
Diverse risk management teams may elect to use social scientists 
trained in consensus-building methods to help establish management 
goals. While management goals derived in this way may require further 
definition (see text note 2-5), there is increased confidence that 
these goals are supported by the audience for the risk assessment.
    Regardless of how management goals are established, goals that 
explicitly define which ecological values are to be protected are more 
easily used to design a risk assessment for decisionmaking than general 
management goals. Whenever goals are general, risk assessors must 
interpret those goals into ecological values that can be measured or 
estimated and ensure that the managers agree with their interpretation 
(see text note 2-6). Legally mandated goals generally are interpreted 
by Agency managers and staff. This interpretation may be performed once 
and then applied to the multiple similar assessments (e.g. evaluation 
of new chemicals). For other risk assessments, the interpretation is 
unique to the ecosystem being assessed and must be done on a case-by-
case basis as part of the planning process.

2.2. Management Decisions

    A risk assessment is shaped by the kind of decision it will 
support. When a management decision is explicitly

[[Page 47563]]

stated and closely aligned to management actions, the scope, focus, and 
conduct of the risk assessment are well defined by the specificity of 
the decision to be made. Some of these risk assessments are used to 
help establish national policy that will be applied consistently across 
the country (e.g., premanufacture notices for new chemicals, protection 
of endangered species). Other risk assessments are designed for a 
specific site (e.g., hazardous waste site clean-up level). When 
decision options (e.g., decision criteria in the data quality 
objectives process, U.S. EPA, 1994d; see section 3.5.2 for more 
details) are known prior to the risk assessment, a number of 
assumptions are inherent in those options that need to be explicitly 
stated during planning. This ensures that the decision criteria are not 
altering the scientific validity of the risk assessment by 
inappropriately applying assumptions or unnecessarily limiting the 
variables. For many risk assessments, there may be a range of possible 
management options for managing risk. When different management options 
have been identified (e.g., leave alone, clean up, or pave a 
contaminated site), risk assessment can be used to predict potential 
risk across the range of these management options.
    Risk assessments may be designed to provide guidance for management 
initiatives for a region or watershed where multiple stressors, 
ecological values, and political factors influence decisionmaking. 
These risk assessments require great flexibility and breadth and may 
use national risk-based information and site-specific risk information 
in conjunction with regional evaluations of risk. As risk assessment is 
more frequently used to support landscape-scale management decisions, 
the diversity, breadth, and complexity of the risk assessments increase 
significantly and may include evaluations that focus on understanding 
ecological processes influenced by a diversity of human actions and 
management options. Risk assessments used in this application are often 
based on a general goal statement and require significant planning to 
establish the purpose, scope, and complexity of the assessment.

2.3. Scope and Complexity of the Risk Assessment

    Although the purpose for the risk assessment determines whether it 
is national, regional, or local, the resources available for conducting 
the risk assessment determines how extensive and complex it can be 
within this framework and the level of uncertainty that can be 
expected. Each risk assessment is constrained by the availability of 
data, scientific understanding, expertise, and financial resources. 
Within these constraints there is much to consider when designing a 
risk assessment. Risk managers and risk assessors must discuss in 
detail the nature of the decision (e.g., national policy, local 
economic impact), available resources, opportunities for increasing the 
resource base (e.g., partnering, new data collection, alternative 
analytical tools), and the output that will provide the best 
information for decisions required (see text note 2-7).
    Part of the agreement on scope and complexity is based on the 
maximum uncertainty that is acceptable in whatever decision the risk 
assessment supports. The lower the tolerance for uncertainty, the 
greater the scope and complexity needed in the risk assessment. Risk 
assessments completed in response to legal mandates and likely to be 
challenged in court often require rigorous attention to acceptable 
levels of uncertainty to ensure that the assessment will be used in a 
decision. A frank discussion is needed between the risk manager and 
risk assessor on sources of uncertainty in the risk assessment and ways 
uncertainty can be reduced (if necessary) through selective investment 
of resources. Where appropriate, planning could account for the 
iterative nature of risk assessment and include explicitly defined 
steps. These steps may take the form of ``tiers'' that represent 
increasing levels of complexity and investment, with each tier designed 
to reduce uncertainty. The plan may include an explicit definition of 
iterative steps with a description of levels of investment and decision 
criteria for each tier. Guidance on addressing the interplay of 
management decisions, study boundaries, data needs, uncertainty, and 
specifying limits on decision errors may be found in EPA's guidance on 
data quality objectives (U.S. EPA, 1994d).

2.4. Planning Outcome

    The planning phase is complete when agreements are reached on the 
management goals, assessment objectives, the focus and scope of the 
risk assessment, resource availability, and the type of decisions the 
risk assessment is to support. Agreements may encompass the technical 
approach to be taken in a risk assessment as determined by the 
regulatory or management context and reason for initiating the risk 
assessment (see section 3.2), the spatial scale (e.g., local, regional, 
or national), and temporal scale (e.g., the time frame over which 
stressors or effects will be evaluated).
    In mandated risk assessments, planning agreements are often 
codified in regulations, and little documentation of agreements is 
warranted. In risk assessments where planning decisions can be highly 
variable, a summary of planning agreements may be important for 
ensuring that the risk assessment remains consistent with early 
agreements. A summary can provide a point of reference for determining 
if early decisions may need to be changed in response to new 
information. There is no defined format, length, or complexity for a 
planning summary. It is a useful reference only and should be tailored 
to the complexity of the risk assessment it represents. However, a 
summary is recommended to help ensure quality communication between and 
among risk managers and risk assessors and to document the decisions 
that have been agreed upon.
    Once planning is complete, the formal process of risk assessment 
begins through the initiation of problem formulation. During problem 
formulation, risk assessors should continue the dialogue with risk 
managers following assessment endpoint selection and once the analysis 
plan is completed. At these points, potential problems can be 
identified before the risk assessment proceeds.

3. Problem Formulation Phase

    Problem formulation is a formal process for generating and 
evaluating preliminary hypotheses about why ecological effects have 
occurred, or may occur, from human activities. As the first stage of an 
ecological risk assessment, it provides the foundation on which the 
entire assessment depends. During problem formulation, management goals 
developed during planning are evaluated to establish objectives for the 
risk assessment, the problem is defined, and the plan for analyzing 
data and characterizing risk is determined. Any deficiencies in problem 
formulation will compromise all subsequent work on the risk assessment 
(see text note 3-1).

3.1. Products of Problem Formulation

    Successful completion of problem formulation depends on the quality 
of three products: (1) assessment endpoints that adequately reflect 
management goals and the ecosystem they represent, (2) conceptual 
models that describe key relationships between a stressor and 
assessment endpoint or among several stressors and assessment

[[Page 47564]]

endpoints, and (3) an analysis plan. Essential to the development of 
these products are the effective integration and evaluation of 
available information.
    The following discussion focuses on the products of problem 
formulation and the information that determines the nature of those 
products. The products are featured in the problem formulation diagram 
as circles (see figure 3-1). The types of information that must be 
evaluated to generate those products are shown in the hexagon.

BILLING CODE 6560-50-P
      

[[Page 47565]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.002



BILLING CODE 6560-50-C

[[Page 47566]]

    To enhance clarity, the organization of the following discussion 
follows the above topics. However, problem formulation is not 
necessarily completed in the order presented here. First, the order in 
which products are produced is directly related to why the ecological 
risk assessment is initiated, as addressed in section 3.2. Second, 
problem formulation is inherently interactive and iterative, not 
linear. Substantial reevaluation is expected to occur within and among 
all products of problem formulation.

3.2. Integration of Available Information

    The foundation for problem formulation is the integration of 
available information on the sources of stressors and stressor 
characteristics, exposure, the ecosystem(s) potentially at risk, and 
ecological effects (see figure 3-1). When key information is of the 
appropriate type and sufficient quality and quantity, problem 
formulation can proceed effectively. When key information is 
unavailable in one or more areas, the risk assessment may be 
temporarily suspended while new data are collected. If new data cannot 
be collected, then the risk assessment will depend on what is known and 
what can be extrapolated from that information. Complete information is 
not available at the beginning of many risk assessments. When this is 
the case, the process of problem formulation assists in identifying 
where key data are missing and provides the framework for further 
research where more data are needed. Where data are few, a clear 
articulation of the limitations of conclusions, or uncertainty, from 
the risk assessment becomes increasingly critical in risk 
characterization (see text note 3-2).
    The reason why an ecological risk assessment is initiated directly 
influences what information is available at the outset, and what 
information must be found. A risk assessment can be initiated because a 
known or potential stressor may be released into the environment, an 
adverse effect or change in condition is observed, or better management 
of an important ecological value (e.g., valued ecological entities such 
as species, communities, ecosystems or places) is desired. Risk 
assessments are sometimes initiated for two or all three of these 
reasons.
    Risk assessors beginning with information about the source or 
stressor will seek available information on the effects the stressor 
might be associated with and the ecosystems that it will likely be 
found in. Risk assessors beginning with information about an observed 
effect or change in condition will need to seek information about 
potential stressors and sources. Risk assessors starting with concern 
over a particular ecological value may need additional information on 
the specific condition or effect of interest, the ecosystems 
potentially at risk, and potential stressors and sources.
    The initial use of available information is a scoping process 
similar to that used to develop environmental impact statements. During 
this process, data and information (i.e., actual, inferred, or 
estimated) are considered to ensure that nothing important is 
overlooked. A comprehensive evaluation of all information provides the 
framework for generating a large array of risk hypotheses to consider 
(see section 3.4.1). After the initial scoping process, information 
quality and applicability to the particular problem of concern are 
increasingly scrutinized as the risk assessor proceeds through problem 
formulation. When analysis plans are formed, data validity becomes a 
significant factor to consider. Issues relating to evaluating data 
quality are discussed in the analysis phase (see section 4.1).
    As the complexity and spatial scale of a risk assessment increase, 
information needs escalate. Ecosystems characteristics directly 
influence when, how, and why particular ecological entities may become 
exposed and exhibit adverse effects due to particular stressors. 
Predicting risks from multiple chemical, physical, and biological 
stressors requires an understanding of their interactions. Risk 
assessments for a region or watershed, where multiple stressors are the 
rule, require consideration of ecological processes operating at larger 
spatial scales.
    Despite limitations on what is known about ecosystems and the 
stressors influencing them, the process of problem formulation offers a 
valuable systematic approach for organizing and evaluating available 
information on all stressors and possible effects in a way that can be 
useful to risk assessors and decisionmakers. Text note 3-3 provides a 
series of questions that risk assessors should attempt to answer using 
available information, many of which were drawn from Barnthouse and 
Brown (1994). This exercise will help risk assessors identify known and 
unknown relationships, both of which are important in problem 
formulation.
    Problem formulation proceeds with the identification of assessment 
endpoints, and the development of conceptual models and the analysis 
plan (discussed below). However, the order in which these task are done 
is influenced by the reason for initiating the assessment (text note 3-
4). Early recognition that initiation effects the order of product 
generation will help facilitate the development of problem formulation.

3.3. Selecting Assessment Endpoints

    Assessment endpoints are ``explicit expressions of the actual 
environmental value that is to be protected'' (U.S. EPA, 1992a). 
Assessment endpoints are critical to problem formulation because they 
link the risk assessment to management concerns and they are central to 
conceptual model development. Their relevance to ecological risk 
assessment is determined by how well they target susceptible ecological 
entities. Their ability to support risk management decisions depends on 
how well they represent measurable characteristics of the ecosystem 
that adequately represent management goals. The selection of ecological 
concerns and assessment endpoints in EPA has traditionally been done 
internally by individual Agency program offices (U.S. EPA, 1994g). More 
recently, Agency activities such as the watershed protection approach 
and community-based environmental protection have used contributions by 
interested parties in the selection of ecological concerns and 
assessment endpoints. This section describes criteria for selecting and 
defining assessment endpoints.
3.3.1. Selecting What To Protect
    The ecological resources selected to represent management goals for 
environmental protection are reflected in the assessment endpoints that 
drive ecological risk assessments. Assessment endpoints often reflect 
environmental values that are protected by law, provide critical 
resources, or provide an ecological function that would be 
significantly impaired (or that society would perceive as having been 
impaired) if the resource were altered.
    Although many potential assessment endpoints may be identified, 
considering the practicality of using particular assessment endpoints 
will help refine selections. For example, when the attributes of an 
assessment endpoint can be measured directly, extrapolation is 
unnecessary; therefore this uncertainty is not introduced into the 
results. Assessment endpoints that cannot be measured directly but can 
be represented by measures that are easily monitored and modeled still 
provide a good foundation for the risk assessment. Assessment endpoints 
that cannot be linked with measurable attributes are not appropriate 
for a risk assessment.
    Three principal criteria are used when selecting assessment 
endpoints: (1) their

[[Page 47567]]

ecological relevance, (2) their susceptibility to the known or 
potential stressors, and (3) whether they represent management goals. 
Of these three criteria, ecological relevance and susceptibility are 
essential for selecting assessment endpoints that are scientifically 
valid. Rigorous selection based on these criteria must be maintained. 
However, to increase the likelihood that the risk assessment will be 
used in management decisions, assessment endpoints that represent 
societal values and management goals are more effective. Given the 
complex functioning of ecosystems and the interdependence of ecological 
entities, it is likely that assessment endpoints can be selected that 
are responsive to management goals while meeting scientific criteria. 
This provides a way to address changes that may occur over time in the 
public's perception of ecological value (e.g., wetlands viewed as 
infested swamps 30 years ago are considered prime wildlife habitat 
today; Suter, 1993a). Assessment endpoints that meet all three criteria 
provide the best foundation for an effective risk assessment (e.g., see 
text note 3-5).
3.3.1.1. Ecological Relevance
    Ecologically relevant endpoints reflect important characteristics 
of the system and are functionally related to other endpoints (U.S. 
EPA, 1992a). These are endpoints that help sustain the natural 
structure, function, and biodiversity of an ecosystem. For example, 
ecologically relevant endpoints may contribute to the food base (e.g., 
primary production), provide habitat, promote regeneration of critical 
resources (e.g., decomposition or nutrient cycling), or reflect the 
structure of the community, ecosystem, or landscape (e.g., species 
diversity or habitat mosaic). Changes in ecologically relevant 
endpoints can result in unpredictable and widespread effects.
    Ecological relevance becomes most important when risk assessors are 
identifying the potential cascade of adverse effects that could result 
from the loss or reduction of one or more species or a change in 
ecosystem function (see text note 3-6). Careful selection of assessment 
endpoints that address both specific organisms of concern and 
landscape-level ecosystem processes becomes increasingly important in 
landscape-level risk assessments. In some cases, it may be possible to 
select one or more species and an ecosystem process to represent larger 
functional community or ecosystem processes.
    Determining ecological relevance in specific cases requires expert 
judgment based on site-specific information, preliminary site surveys, 
or other available information. The less information available, the 
more critical it is to have informed expert judgment to ensure 
appropriate selections. If assessment endpoints in a risk assessment 
are not ecologically relevant, the results of the risk assessment may 
predict risk to the assessment endpoints selected but seriously 
misrepresent risk to the ecosystem of concern, which could lead to 
misguided management.
3.3.1.2. Susceptibility to Known or Potential Stressors
    Ecological resources are considered susceptible when they are 
sensitive to a human-induced stressor to which they are exposed. 
Sensitivity refers to how readily an ecological entity is affected by a 
particular stressor. Sensitivity is directly related to the mode of 
action of the stressors. For example, chemical sensitivity is 
influenced by individual physiology and metabolic pathways. Sensitivity 
also is influenced by individual and community life-history 
characteristics. For example, species with long life cycles and low 
reproductive rates will be more vulnerable to extinction from increases 
in mortality than those with short life cycles and high reproductive 
rates. Species with large home ranges may be more sensitive to habitat 
fragmentation when the fragment is smaller than their required home 
range compared to those with smaller home ranges within a fragment. 
However, habitat fragmentation may also affect species with small home 
ranges where migration is a necessary part of their life history and 
fragmentation prevents exchange among subpopulations.
    Sensitivity may be related to the life stage of an organism when 
exposed to a stressor. Frequently, young animals are more sensitive to 
stressors than adults. For example, Pacific salmon eggs and fry are 
very sensitive to sedimentation from forest logging practices and road 
building because they can be smothered. Age-dependent sensitivity, 
however, is not only in the young. In many species, special events like 
migration (e.g., in birds) and molting (e.g., in harbor seals) 
represent significant energy investments that make these organisms more 
vulnerable to an array of possible stressors. Finally, sensitivity may 
be increased by the presence of other stressors or natural 
disturbances. For example, the presence of insect pests and disease may 
make plants more sensitive to damage from ozone (Heck, 1993).
    Measures of sensitivity may include mortality or adverse 
reproductive effects from exposure to toxics, behavioral abnormalities, 
avoidance of significant food sources or nesting sites, or loss of 
offspring to predation because of the proximity of stressors such as 
noise, habitat alteration or loss, community structural changes, or 
other factors.
    Exposure is the other key determinant in susceptibility. Exposure 
can mean co-occurrence, contact, or the absence of contact, depending 
on the stressor and assessment endpoint (see section 4 for more 
discussion). The amount and conditions of exposure directly influence 
how an ecological entity will respond to a stressor. Thus, to determine 
what entities are susceptible, it is important to consider information 
on the proximity of an ecological resource to the stressor, the timing 
of exposure (both in terms of frequency and duration), and the 
intensity of exposure occurring during sensitive life stages of the 
organisms.
    Adverse effects of a particular stressor may be important during 
one part of an organism's life cycle, such as early development or 
reproduction. Adverse effects may result from exposure to a stressor or 
to the absence of a necessary resource during a critical life stage. 
For example, if fish are unable to find suitable nesting sites during 
their reproductive phase, risk is significant even when water quality 
is high and food sources abundant. The interplay between life stage and 
stressors can be very complex (e.g., see text note 3-7).
    Exposure may occur in one place or time, and effects may not occur 
until another place or time. Both life history characteristics, as 
described under sensitivity, and the circumstances of exposure, 
influence susceptibility in this case. For example, the temperature of 
the incubation medium of marine turtle eggs affects the sex ratio of 
the offspring. But the population impacts of a change in incubation 
temperature may not be observable until years later when the cohort of 
affected turtles begins to reproduce. Delayed effects and multiple 
stressor exposures add complexity to evaluations of susceptibility. For 
example, although toxicity tests may determine receptor sensitivity to 
one stressor, the degree of susceptibility may depend on the co-
occurrence of another stressor that significantly alters receptor 
response. Conceptual models (see section 3.4) need to reflect these 
factors. If a species is unlikely to be exposed to the stressor of 
concern, it is inappropriate as an assessment endpoint.
3.3.1.3. Representation of Management Goals
    Ultimately, the value of a risk assessment depends on whether it 
can

[[Page 47568]]

support quality management decisions. Risk managers are more willing to 
use a risk assessment for making decisions when the assessment is based 
on values and organisms that people care about. These values, 
interpreted from management goals (see section 2) into assessment 
endpoints, provide a defined and measurable entity for the risk 
assessment. Candidates for assessment endpoints might include entities 
such as endangered species, commercially or recreationally important 
species, functional attributes that support food sources or flood 
control (wetland water sequestration, for example), or aesthetic 
values, such as clean air in national parks or the existence of 
charismatic species like eagles or whales.
    Selection of assessment endpoints based on public perceptions alone 
could lead to management decisions that do not consider important 
ecological information. While being responsive to the public is 
important, it does not obviate the requirement for scientific validity 
as represented by the sections on ecological relevance and 
susceptibility. Many ecological entities and attributes meet the 
necessary scientific rigor as assessment endpoints; some will be 
recognized as valuable by risk managers and the public, and others will 
not. Midges, for example, can represent the base of a complex food web 
that supports a popular sports fishery. They may also be considered 
pests. While both midges and fish are important ecological entities in 
this ecosystem and represent key components of the aquatic community, 
selecting the fishery as the assessment endpoint and using midges as a 
critical ecological entity to measure allow both entities to be used in 
the risk assessment. This choice maintains the scientific validity of 
the risk assessment and is responsive to management concerns. In those 
cases where the risk assessor identifies a critical assessment endpoint 
that is unpopular with the public, the risk assessor may find it 
necessary to present a persuasive case in its favor based on scientific 
arguments.
3.3.2. Defining Assessment Endpoints
    Assessment endpoints provide a transition between broad management 
goals and the specific measures used in an assessment. They help 
assessors identify measurable attributes to quantify and predict 
change. Assessment endpoints also help the risk assessor determine 
whether management goals have been or can be achieved (see text note 3-
8).
    Two elements are required to define an assessment endpoint. The 
first is the valued ecological entity. This can be a species (e.g., 
eelgrass, piping plover), a functional group of species (e.g., 
raptors), an ecosystem function or characteristic (e.g., nutrient 
cycling), a specific valued habitat (e.g., wet meadows) or a unique 
place (e.g., a remnant of native prairie). The second is the 
characteristic about the entity of concern that is important to protect 
and potentially at risk. For example, it is necessary to define what is 
important for piping plovers (e.g., nesting and feeding success), 
eelgrass (e.g., areal extent and patch size), and wetlands (e.g., 
endemic wet meadow community structure and function). For an assessment 
endpoint to provide a clear interpretation of the management goals and 
the basis for measurement in the risk assessment, both an entity and an 
attribute are required.
    Assessment endpoints are distinct from management goals. They do 
not represent what the managers or risk assessors want to achieve. As 
such they do not contain words like ``protect,'' ``maintain,'' or 
``restore,'' or indicate a direction for change such as ``loss'' or 
``increase.''
    Defining assessment endpoints can be difficult. They may be too 
broad, vague, or narrow, or they may be inappropriate for the ecosystem 
requiring protection. ``Ecological integrity'' is a frequently cited, 
but vague, goal and an even more vague assessment endpoint. 
``Integrity'' can only be used effectively when its meaning is 
explicitly characterized for a particular ecosystem, habitat, or 
entity. This may be done by selecting key entities and processes of an 
ecosystem and describing characteristics that best represent integrity 
for that system. For example, general goals for Waquoit Bay were 
translated into several assessment endpoints, including ``estuarine 
eelgrass abundance and distribution'' (see text note 2-6).
    Expert judgment and an understanding of the characteristics and 
function of an ecosystem are important for translating general goals 
into usable assessment endpoints. Endpoints that are too narrowly 
defined, however, may not support effective risk management. For 
example, if an assessment is focused on protecting the habitat of an 
endangered species, the risk assessment may overlook important 
characteristics of the ecosystem and fail to include critical variables 
(see text note 3-7).
    Assessment endpoints must be appropriate for the ecosystem of 
concern. Selecting a game fish that grows well in reservoirs may meet a 
``feasible'' management goal, but would be inappropriate for evaluating 
risk from a new hydroelectric dam if the ecosystem of concern is a 
stream in which salmon spawn (see text note 3-5). Although the game 
fish will satisfy the fishable goal and may be highly desired by local 
fishermen, a reservoir species does not represent the ecosystem at 
risk. A vague ``viable fish populations'' assessment endpoint 
substituted by ``reproducing populations of indigenous salmonids'' 
could therefore prevent the development of an inappropriate risk 
assessment.
    Clearly defined assessment endpoints provide direction and 
boundaries for the risk assessment and can minimize miscommunication 
and reduce uncertainty. Assessment endpoints directly influence the 
type, characteristics, and interpretation of data and information used 
for analyses and the scale and character of the assessment. For 
example, an assessment endpoint such as ``egg production of pond 
invertebrates'' defines local population characteristics and requires 
very different types of data and ecosystem characterization compared 
with ``watershed aquatic community structure and function.'' If 
concerns are local, the assessment endpoints should not focus on 
landscape concerns. Where ecosystem processes and landscape mosaics are 
of concern, survival of a particular species would provide inadequate 
representation. Assessment endpoints that are poorly defined, 
inappropriate, or at the incorrect scale can be very problematic. 
Common problems encountered in selecting assessment endpoints are 
summarized in text note 3-9.
    The presence of multiple stressors should influence the selection 
of assessment endpoints. When it is possible to select one assessment 
endpoint that is sensitive to many of the identified stressors, yet 
responds in different ways to different stressors, it is possible to 
consider the combined effects of multiple stressors while still 
discriminating among effects. For example, if recruitment of a fish 
population is the assessment endpoint, it is important to recognize 
that recruitment may be adversely affected at several life stages, in 
different habitats, through different ways, by different stressors. The 
measures of effect, exposure, and ecosystem and receptor 
characteristics chosen to evaluate recruitment provide a basis for 
discriminating among different stressors, individual effects, and their 
combined effect.
    The assessment endpoint can provide a basis for comparing a range 
of stressors if carefully selected. For example, the National Crop Loss 
Assessment Network (Heck, 1993)

[[Page 47569]]

selected crop yields as the assessment endpoint to evaluate the 
cumulative effects of multiple stressors. Although the primary stressor 
was ozone, the crop-yield endpoint allowed them to consider the effects 
of sulfur dioxide and soil moisture. As Barnthouse et al. (1990) 
pointed out, an endpoint should be selected so that all the effects can 
be expressed in the same units (e.g., the abundance of 1-year-old fish 
to assess the effects from toxicity, fishing pressure, and habitat 
loss). These considerations are important when selecting assessment 
endpoints for addressing the combined effect of multiple stressors. 
However, in situations where multiple stressors act on the structure 
and function of aquatic and terrestrial communities in a watershed 
ecosystem, an array of assessment endpoints that represent the 
ecosystem community and processes is more effective than a single 
endpoint. When based on differing susceptibility to an array of 
stressors, the careful selection of assessment endpoints can help risk 
assessors distinguish among effects from diverse stressors. Exposure to 
multiple stressors may lead to effects at different levels of 
biological organization, for a cascade of adverse responses that should 
be considered.
    Although assessment endpoints must be defined in terms of 
measurable attributes, selection does not depend on the ability to 
measure those attributes directly or on whether methods, models, and 
data are currently available. If the response of an assessment endpoint 
cannot be directly measured, it may be predicted from responses of 
surrogate or similar entities. Although for practical reasons it is 
helpful to use assessment endpoints that have well-developed test 
methods, field measurement techniques, and predictive models (see 
Suter, 1993a), it is not necessary for these methods to be established 
protocols. Measures that will be used to evaluate assessment endpoint 
response to exposures for the risk assessment are often identified 
during conceptual model development and specified in the analysis plan. 
See section 3.5 for issues surrounding the selection of measures.
    It is important for risk assessors and risk managers to agree that 
selected assessment endpoints represent the management goals for the 
particular ecological value. The rationale for their selection should 
be clear. Assessment endpoint selection is an important risk manager-
risk assessor checkpoint during problem formulation.

3.4. Conceptual Models

    A conceptual model in problem formulation is a written description 
and visual representation of predicted responses by ecological entities 
to stressors to which they are exposed, and the model includes 
ecosystem processes that influence these responses. Conceptual models 
represent many relationships (e.g., exposure scenarios may 
qualitatively link land-use activities to sources and their stressors, 
may describe primary, secondary, and tertiary exposure pathways, and 
may describe co-occurrence between exposure pathways, ecological 
effects, and ecological receptors).
    Conceptual models for ecological risk assessments are developed 
from information about stressors, potential exposure, and predicted 
effects on an ecological entity (the assessment endpoint). Depending on 
why a risk assessment is initiated, one or more of these categories of 
information is known at the outset. The process of creating conceptual 
models helps identify the unknown elements.
    The complexity of the conceptual model depends on the complexity of 
the problem, number of stressors, number of assessment endpoints, 
nature of effects, and characteristics of the ecosystem. For single 
stressors and single assessment endpoints, conceptual models can be 
relatively simple relationships. In situations where conceptual models 
describe both the pathways of individual stressors and assessment 
endpoints and the interaction of multiple and diverse stressors and 
assessment endpoints (e.g., assessments initiated because of important 
values), several submodels normally will be required to describe 
individual pathways. Other models may then be used to explore how these 
individual pathways interact.
    Conceptual models consist of two principal products:
     A set of risk hypotheses that describe predicted 
relationships between stressor, exposure, and assessment endpoint 
response, along with the rationale for their selection.
     A diagram that illustrates the relationships presented in 
the risk hypotheses.
3.4.1. Risk Hypotheses
    Hypotheses are assumptions made in order to evaluate logical or 
empirical consequences (Merriam-Webster, 1972). Risk hypotheses are 
statements of assumptions about risk based on available information 
(see text note 3-10). They are formulated using a combination of expert 
judgment and information on the ecosystem at risk, potential sources of 
stressors, stressor characteristics, and observed or predicted 
ecological effects on selected or potential assessment endpoints. These 
hypotheses may predict the effects of a stressor event before it 
happens, or they may postulate why observed ecological effects occurred 
and ultimately what sources and stressors caused the effect. Depending 
on the scope of the risk assessment, the set of risk hypotheses may be 
very simple, predicting the potential effect of one stressor on one 
receptor, or extremely complex, as is typical in value-initiated risk 
assessments that often include prospective and retrospective hypotheses 
about the effects of multiple complexes of stressors on diverse 
ecological receptors.
    Although risk hypotheses should be developed even when information 
is incomplete, the amount and quality of data will affect the 
specificity and level of uncertainty associated with risk hypotheses 
and the conceptual models they form. When preliminary information is 
conflicting, risk hypotheses can be constructed specifically to 
differentiate among competing predictions. The predictions can then be 
evaluated systematically either by using available data during the 
analysis phase or by collecting new data before proceeding with the 
risk assessment. Hypotheses and predictions set a framework for using 
data to evaluate functional relationships (e.g., stressor-response 
curves).
    Early conceptual models are intended to be broad in scope, 
identifying as many potential relationships as possible. As more 
information is incorporated, the plausibility of specific risk 
hypotheses helps risk assessors sort through potentially large numbers 
of stressor-effect relationships and the ecosystem processes that 
influence them to identify those risk hypotheses most appropriate for 
the analysis phase. It is then that justifications for selecting and 
omitting selecting hypotheses are documented. Examples of risk 
hypotheses are provided in text note 3-11.
3.4.2. Conceptual Model Diagrams
    Conceptual model diagrams may be based on theory and logic, 
empirical data, mathematical models, or probability models. They are 
useful tools for communicating important pathways in a clear and 
concise way and can be used to ask new questions about relationships 
that help generate plausible risk hypotheses. Some of the benefits 
gained by developing conceptual models are featured in text note 3-12.
    Conceptual model diagrams frequently contain boxes and arrows to 
illustrate relationships (see figure 3-2

[[Page 47570]]

and Appendix C). When constructing these kinds of flow diagrams, it is 
helpful to use distinct and consistent shapes to distinguish stressors, 
assessment endpoints, responses, exposure routes, and ecosystem 
processes. Although flow diagrams are often used to illustrate 
conceptual models, there is no set configuration for conceptual model 
diagrams. Pictorial representations can be more effective (e.g., 
Bradley and Smith, 1989). Regardless of the configuration, a 
significant part of the usefulness of a diagram is linked to the 
detailed written descriptions and justifications for the pathways and 
relationships shown. Without this, diagrams can misrepresent the 
processes illustrated.

BILLING CODE 6560-50-P
      

[[Page 47571]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.003



BILLING CODE 6560-50-C

[[Page 47572]]

    When developing diagrams to represent a conceptual model, factors 
to consider include the number of relationships depicted, the 
comprehensiveness of the information, the certainty surrounding a 
pathway, and the potential for measurement. The number of relationships 
that can be depicted in one flow diagram depends on how comprehensive 
each relationship is. The more comprehensive, the fewer relationships 
that can be shown with clarity. Flow diagrams that highlight where data 
are abundant or scarce can provide insights on how the analyses should 
be approached and can be used to show the degree of confidence the risk 
assessor has in the relationship. Such flow diagrams can also help 
communicate why certain pathways were pursued and others were not.
    Diagrams provide a working and dynamic representation of 
relationships. They should be used to explore different ways of looking 
at a problem before selecting one or several to guide analysis. Once 
the risk hypotheses are selected and flow diagrams drawn, they set the 
framework for final planning for the analysis phase.
3.4.3. Uncertainty in Conceptual Models
    Conceptual model development may account for one of the most 
important sources of uncertainty in a risk assessment. If important 
relationships are missed or specified incorrectly, risks could be 
seriously under- or overestimated in the risk characterization phase. 
Uncertainty can arise from lack of knowledge on how the ecosystem 
functions, failing to identify and interrelate temporal and spatial 
parameters, not describing a stressor or suite of stressors, or not 
recognizing secondary effects. In some cases, little may be known about 
how a stressor moves through the environment or causes adverse effects. 
In most cases, multiple stressors are the norm and a source of 
confounding variables, particularly for conceptual models that focus on 
a single stressor. Opinions of experts on the appropriate conceptual 
model configuration may differ. While simplification and lack of 
knowledge may be unavoidable, risk assessors should document what is 
known, justify the model, and rank model components in terms of 
uncertainty (see Smith and Shugart, 1994).
    Uncertainty associated with conceptual models can be reduced by 
developing alternative conceptual models for a particular assessment to 
explore possible relationships. In cases where more than one conceptual 
model is plausible, the risk assessor must decide whether it is 
feasible to follow separate models through the analysis phase or 
whether the models can be combined into a better conceptual model. It 
is important to revisit, and if necessary revise, conceptual models 
during risk assessments to incorporate new information and recheck the 
rationale. It is valuable to present conceptual models to risk managers 
to ensure the models communicate well and address key concerns the 
managers have. This check for completeness and clarity provides an 
opportunity to assess the need for changes before analysis begins.
    Throughout the process of problem formulation, ambiguities, errors, 
and disagreements will occur, all of which contribute to uncertainty. 
Wherever possible, these sources of uncertainty should be eliminated 
through better planning. Because all uncertainty cannot be eliminated, 
a clear description of the nature of the uncertainties should be 
clearly summarized at the close of the problem formulation. Text note 
3-13 provides recommendations for describing uncertainty in problem 
formulation.
    The hypotheses considered most likely to contribute to risk are 
pursued in the analysis phase. As discussed previously, it is important 
to provide the rationale for selecting and omitting risk hypotheses and 
to acknowledge data gaps and uncertainties.

3.5. Analysis Plan

    An analysis plan can be a usual final stage of problem formulation, 
particularly in the case of complex assessments. Here, risk hypotheses 
are evaluated to determine how they will be assessed using available 
and new data. The analysis plan can also delineate the assessment 
design, data needs, measures, and methods for conducting the analysis 
phase of the risk assessment. The analysis plan may be relatively brief 
or extensive depending on the nature of the assessment.
    The analysis plan includes the most important pathways and 
relationships identified during problem formulation that will be 
pursued in the analysis phase. It is important for the risk assessor to 
describe what will be done and, in particular, what will not be done. 
It is important to address issues concerning the level of confidence 
needed for the management decision relative to the confidence that can 
be expected from an analysis in order to determine data needs and 
evaluate whether one analytical approach may be better than another. 
When new data are needed to conduct analyses, the feasibility of 
obtaining the data should be taken into account.
    The selection of critical relationships in the conceptual model to 
pursue in analysis is based on several criteria, including:
     Availability of information.
     Strength of information about relationships between 
stressors and effects.
     The assessment endpoints and their relationship to 
ecosystem function.
     Relative importance or influence and mode of action of 
stressors.
     Completeness of known exposure pathways.
    In situations where data are few and new data cannot be collected, 
it is possible to combine existing data with extrapolation models so 
that alternative data sources may be used. This allows the use of data 
from other locations or on other organisms where similar problems exist 
and data are available. For example, the relationship between nutrient 
availability and algal growth is well established. Although there will 
be differences in how the relationship is manifested based on the 
dynamics of a particular ecosystem, the relationship itself will tend 
to be consistent. When using data that require extrapolation, it is 
important to identify the source of the data, justify the extrapolation 
method and discuss major uncertainties apparent at this point.
    Where data are not available, recommendations for new data 
collection should be part of problem formulation. An iterative, phased, 
or tiered approach (see text note 1-3) to the risk assessment may be 
selected to provide an opportunity for early management decisions on 
issues that can be addressed using available data. A decision to 
conduct a new iteration is based on the results of any previous 
iteration and proceeds using new data collected as specified in the 
analysis plan. When new data collection cannot be obtained, pathways 
that cannot be assessed are a source of uncertainty and should be 
described in the analysis plan.
3.5.1. Selecting Measures
    It is in the analysis planning stage that measures are identified 
to evaluate the risk hypotheses. There are three categories of 
measures. Measures of effect are measures used to evaluate the response 
of the assessment endpoint when exposed to a stressor (formerly 
measurement endpoints). Measures of exposure are measures of how 
exposure may be occurring, including how a stressor moves through the 
environment and how it may co-occur with the assessment endpoint. 
Measures of

[[Page 47573]]

ecosystem and receptor characteristics include ecosystem 
characteristics that influence the behavior and location of assessment 
endpoints, the distribution of a stressor, and life history 
characteristics of the assessment endpoint that may affect exposure or 
response to the stressor. These diverse measures increase in importance 
as the complexity of the assessment increases and are particularly 
important for risk assessments initiated to protect ecological values 
(see text notes 3-14 and 3-15 for more information).
    Text note 3-16, which describes water quality criteria, provides 
one example of how goals, endpoints, and measures are related. Although 
water quality criteria are often considered risk-based, they do not 
measure exposure. Instead, the water quality criteria provide an 
effects benchmark for decisionmaking. Within that benchmark there are a 
number of assumptions about significance (e.g., aquatic communities 
will be protected by achieving a benchmark derived from individual 
species' toxicological responses to a single chemical) and exposure 
(e.g., 1-hour and 4-day exposure averages). Assumptions embedded in 
decision rules should be articulated (see section 3.5.2).
    The analysis plan provides a synopsis of measures that will be used 
to evaluate risk hypotheses. Potential extrapolations, model 
characteristics, types of data (including quality), and planned 
analyses (with specific tests for different types of data) are 
described. The plan should discuss how the results will be presented 
upon completion. The analysis plan provides the basis for making 
selections of data sets that will be used for the risk assessment.
    The plan includes explanations of how data analyses will 
distinguish among hypotheses, an explicit expression of the approach to 
be used, and justifications for the elimination of some hypotheses and 
selection of others. It includes the measures selected, analytical 
methods planned, and the nature of the risk characterization options 
and considerations that will be generated (e.g., quotients, narrative 
discussion, stressor-response curve with probabilities). An analysis 
plan is enhanced if it contains explicit statements for how measures 
were selected, what they are intended to evaluate, and which analyses 
they support. During analysis planning, uncertainties associated with 
selected measures and analyses are articulated and, where possible, 
plans for addressing them are made.
3.5.2. Relating Analysis Plans to Decisions
    The analysis plan is a risk manager-risk assessor checkpoint and an 
appropriate time for technical review. Discussions between the risk 
assessors and risk managers can help ensure that the analyses will 
provide the type and extent of information that the manager can use for 
decisionmaking. These discussions may also identify what can and cannot 
be done based on the preliminary evaluation of problem formulation, 
including which relationships to portray for the risk management 
decision. A reiteration of the planning discussion is important to 
ensure that the appropriate balance among the requirements for the 
decision, data availability, and resource constraints is established 
for the risk assessment.
    The elements of an analysis plan share significant similarities 
with the data quality objectives (DQO) process (see text note 3-17), 
which emphasizes identifying the problem by establishing study 
boundaries and determining necessary data quality, quantity, and 
applicability to the problem being evaluated. The DQO guidance is a 
valuable reference for risk assessors (U.S. EPA, 1994d).
    The most important difference between problem formulation and DQO 
is the presence of a decision rule that defines a benchmark for a 
management decision before the risk assessment is completed. The 
decision rule step specifies the statistical parameter that 
characterizes the population, specifies the action level for the study, 
and combines outputs from the previous DQO steps into an ``if * * * 
then'' decision rule that defines conditions under which the decision 
maker will choose alternative options. This approach provides the basis 
for establishing null and alternative hypotheses appropriate for 
statistical testing for significance. While this approach is 
appropriate for some risk assessments, many risk assessments are not 
based on benchmark decisions. Presentation of stressor-response curves 
with uncertainty bounds will be more appropriate than statistical 
testing of decision criteria where risk managers must evaluate the 
range of stressor effects to which they compare a range of possible 
management options.
    The analysis plan is the final synthesis before the risk assessment 
proceeds. It summarizes what has been done during problem formulation, 
shows how the plan relates to management decisions that must be made, 
and indicates how data and analyses will be used to estimate risks. 
When it is determined that the problem is clearly defined and there are 
enough data to proceed, analysis begins.

4. Analysis Phase

    The analysis phase consists of the technical evaluation of data to 
reach conclusions about ecological exposure and the relationships 
between the stressor and ecological effects. During analysis, risk 
assessors use measures of exposure, effects, and ecosystem and receptor 
attributes to evaluate questions and issues that were identified in 
problem formulation. The products of analysis are summary profiles that 
describe exposure and the stressor-response relationship. When 
combined, these profiles provide the basis for reaching conclusions 
about risk during the risk characterization phase.
    The conceptual model and analysis plan developed during problem 
formulation provide the basis for the analysis phase. By the start of 
analysis, the assessor should know which stressors and ecological 
effects are the focus of investigation and whether secondary exposures 
or effects will be considered. In the analysis plan, the assessor 
identified the information needed to perform the analysis phase. By the 
start of analysis, these data should be available (text note 4-1).
    The analysis phase is composed of two principal activities, the 
characterization of exposure and characterization of ecological effects 
(figure 4-1). Both activities begin by evaluating data (i.e., the 
measures of exposure, ecosystem and receptor characteristics, and 
effects) in terms of their scientific credibility and relevance to the 
assessment endpoint and conceptual model (discussed in section 4.1). In 
exposure characterization (section 4.2), these data are then analyzed 
to describe the source, the distribution of the stressor in the 
environment, and the contact or co-occurrence of the stressor with 
ecological receptors. In ecological effects characterization (section 
4.3), data are analyzed to describe the relationship between the 
stressor and response and to evaluate the evidence that exposure to the 
stressor causes the response (i.e., stressor-response analyses). In 
many cases, extrapolation will be necessary to link the measures of 
effect with the assessment endpoint.

BILLING CODE 6560-50-P

[[Page 47574]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.004



BILLING CODE 6560-50-C

[[Page 47575]]

    Conclusions about exposure and the relationship between the 
stressor and response are summarized in profiles. The exposure and 
stressor-response profiles (sections 4.2.2 and 4.3.2, respectively) 
provide the opportunity to review what has been learned during the 
analysis phase and summarize this information in the most useful format 
for risk characterization. Depending on the risk assessment, these 
profiles may take the form of a written document or modules of a larger 
process model. Alternatively, documentation may be deferred until risk 
characterization. In any case, the purpose of these profiles is to 
ensure that the information needed for risk characterization has been 
collected and evaluated.
    This process is intended to be flexible, and interaction between 
the ecological effects characterization and exposure characterization 
is recommended. When secondary stressors and effects are of concern, 
exposure and effects analyses are conducted iteratively for different 
ecological entities, and the analyses can become so intertwined that 
they are difficult to differentiate. The bottomland hardwoods example 
(Appendix D) illustrates this type of assessment. This assessment 
examined potential changes in the plant and animal communities under 
different flooding scenarios. The stressor-response and exposure 
analyses were combined within the FORFLO model for primary effects on 
the plant community and within the Habitat Suitability Index for 
secondary effects on the animal community.
    In addition, the distinction between the analysis phase and risk 
estimation can become blurred. For example, the model results developed 
for the bottomland hardwoods example were used directly in risk 
characterization.
    The nature of the stressor (that is, whether it is chemical, 
physical, or biological) will influence the types of analyses conducted 
and the details of implementation. Thus, the results of the analysis 
phase may range from highly quantitative to qualitative, depending on 
the stressor and the scope of the assessment. The estimation of 
exposure to chemicals emphasizes contact and uptake into the organism, 
and the estimation of effects often entails extrapolation from test 
organisms to the organism of interest. For physical stressors, the 
initial disturbance may be most closely related to the assessment 
endpoint (e.g., change of wetland to upland). In many cases, however, 
secondary effects (e.g., effects on wildlife that use the wetland) are 
the principal concern. The point of view taken during the analysis 
phase will depend on the assessment endpoints identified during problem 
formulation. Because adverse effects can occur even if receptors do not 
physically contact disturbed habitat, exposure analyses may emphasize 
co-occurrence with physical stressors rather than contact. For 
biological stressors, exposure analysis evaluates entry, dispersal, 
survival, and reproduction (Orr et al., 1993). Because biological 
stressors can reproduce, interact with other organisms, and evolve over 
time, exposure and effects cannot be quantified with confidence. 
Accordingly, exposure and effects are often assessed qualitatively by 
eliciting expert opinion (Simberloff and Alexander, 1994).

4.1. Evaluating Data and Models for Analysis

    In problem formulation, the assessor identifies the information 
needed to perform the analysis phase and plans for collecting new data. 
The first step of the analysis phase is the critical evaluation of data 
and models to ensure that they can support the risk assessment. The 
sources and evaluation of data and models are discussed in sections 
4.1.1 and 4.1.2, respectively. The evaluation of uncertainty, an 
important consideration when evaluating data and also throughout the 
analysis phase, is discussed in section 4.1.3.
4.1.1. Strengths and Limitations of Different Types of Data
    The analysis phase relies on the measures identified in the 
analysis plan; these may come from laboratory or field studies or may 
be produced as output from a model. Data may have been developed for a 
specific risk assessment or for another purpose. A strategy that builds 
on the strengths of each type of data can improve confidence in the 
conclusions of a risk assessment.
    Both laboratory and field studies (including field experiments and 
observational studies) can provide useful data for risk assessment. 
Because conditions can be controlled in laboratory studies, responses 
can be less variable and smaller differences easier to detect. However, 
the controls may limit the range of responses (for example, animals 
cannot seek alternate food sources), so they may not reflect responses 
in the environment. Field surveys are usually more representative of 
both exposures and effects (including secondary effects) found in 
natural systems than are estimates generated from laboratory studies or 
theoretical models. However, because conditions are not controlled, 
variability may be higher and it may be difficult to detect 
differences. Field studies are most useful for linking stressors with 
effects when stressor and effect levels are measured concurrently. In 
addition, the presence of confounding stressors can make it difficult 
to attribute observed effects to specific stressors. Preferred field 
studies use designs that minimize effects of potentially confounding 
factors. Intermediate between laboratory and field are studies that use 
environmental media collected from the field to conduct studies of 
response in the laboratory. Such studies may improve the power to 
detect differences and may be designed to provide evidence of 
causality.
    Most data will be reported as measurements for single variables 
such as a chemical concentration or the number of dead organisms. In 
some cases, however, variables are combined into indices, and the index 
values are reported. Several indices are used to evaluate effects, for 
example, the rapid bioassessment protocols (U.S. EPA, 1989a) and the 
Index of Biotic Integrity, or IBI (Karr, 1981; Karr et al., 1986). 
These have several advantages (Barbour et al., 1995), including the 
ability to:
     Provide an overall indication of biological condition by 
incorporating many attributes of system structure and function, from 
individual to ecosystem levels.
     Evaluate responses from a broad range of anthropogenic 
stressors.
     Minimize the limitations of individual metrics for 
detecting specific types of responses.
    Although indices are very useful, they have several drawbacks, many 
of which are associated with combining heterogeneous variables. For 
example, the final value may depend strongly on the function used to 
combine variables. Some indices (e.g., the IBI) combine only measures 
of effects. Differential sensitivity or other factors may make it 
difficult to attribute causality when many response variables are 
combined. Such indices may need to be separated into their components 
to investigate causality (Suter, 1993b; Ott, 1978). Interpretation 
becomes even more difficult when an index combines measures of exposure 
and effects because double-counting may occur or changes in one 
variable can mask changes in another. Exposure and effects measures may 
need to be separated in order to make appropriate conclusions. For 
these reasons, professional judgment plays a critical role in 
developing and applying indices.
    Experience from similar situations is also an important data source 
that is particularly useful when predicting effects of stressors that 
have not yet

[[Page 47576]]

been released. For example, lessons learned from past experiences with 
related organisms are often critical in trying to predict whether an 
organism will survive, reproduce, and disperse in a new environment. 
Another example is the evaluation of toxicity of new chemicals through 
the use of structure-activity relationships, or SARs (Auer et al., 
1994; Clements and Nabholz, 1994). The simplest application of SARs is 
to identify a suitable analog for which data are available to estimate 
the toxicity of the compound for which data are lacking. More advanced 
applications involve the use of quantitative structure-activity 
relationships (QSARs). QSARs describe the relationships between 
chemical structures and specific biological effects and are derived 
using information on sets of related chemicals (Lipnick, 1995; Cronin 
and Dearden, 1995). The use of analogous data without knowledge of the 
underlying processes may substantially increase the uncertainty in the 
risk assessment (e.g., Bradbury, 1994); however, these data may be the 
only option available.
    While models are often developed and used as part of the risk 
assessment, sometimes the risk assessor relies on output of a 
previously developed model as input to the risk assessment. Models are 
particularly useful when measurements cannot be taken, for example when 
the assessment is predicting the effects of a chemical yet to be 
manufactured. Models can also provide estimates for times or locations 
that are impractical to measure and provide a basis for extrapolating 
beyond the range of observation. Starfield and Bleloch (1991) caution 
that ``the quality of the model does not depend on how realistic it is, 
but on how well it performs in relation to the purpose for which it was 
built.'' Thus, the assessor must review the questions that need to be 
answered and then ensure that a model can answer those questions. 
Because models are simplifications of reality, they may not include 
important processes for a particular system and may not reflect every 
condition in the real world. In addition, a model's output is only as 
good as the quality of its input variables, so critical evaluation of 
input data is important, as is comparing model outputs with 
measurements in the system of interest whenever possible.
    Data and models for risk assessment are often developed in a tiered 
fashion (also see text note 1-3). For example, simple models that err 
on the side of conservatism may be used first, followed by more 
elaborate models that provide more realistic estimates. Effects data 
may also be collected by using a tiered approach. Short-term tests 
designed to evaluate effects such as lethality and immobility may be 
conducted first. If the chemical exhibits high toxicity or a 
preliminary characterization indicates a risk, then more expensive, 
longer-term tests that measure sublethal effects such as changes to 
growth and reproduction can be conducted. Later tiers may employ 
multispecies tests or field experiments. It is important to evaluate 
tiered data in light of the decision they are intended to support; data 
collected for early tiers may not be able to support more sophisticated 
needs.
4.1.2. Evaluating Measurement or Modeling Studies
    Much of the information used in the analysis phase is available 
through published or unpublished studies that describe the purpose of 
the study, the methods used to collect data, and the results. 
Evaluating the utility of these studies relies on careful comparison of 
the objectives of the studies with the objectives of the risk 
assessment. In addition, study methods are examined to ensure that the 
intended objectives were met and that the data are of sufficient 
quality to support the risk assessment. Confidence in the information 
and the implications of using different studies should be described 
during risk characterization, when the overall confidence in the 
assessment is discussed. In addition, the risk assessor should identify 
areas where existing data do not meet risk assessment needs. In these 
cases, we recommend collecting new data.
    EPA is in the process of adopting the American Society for Quality 
Control's E-4 guidelines for assuring environmental data quality 
throughout the Agency (ASQC, 1994) (text note
4-2). These guidelines describe procedures for collecting new data and 
provide a valuable resource for evaluating existing studies. (Readers 
are also referred to Smith and Shugart, 1994; U.S. EPA, 1994f; and U.S. 
EPA, 1990, for more information on evaluating data and models.)
    A study's documentation directly influences the ability to evaluate 
its utility for risk assessment. Studies should contain sufficient 
information so that results can be reproduced, or at least so the 
details of the author's work can be accessed and evaluated. An 
additional advantage is the ability to access findings in their 
entirety; this provides the opportunity to conduct additional analyses 
of the data, if needed. For models, a number of factors increase the 
accessibility of methods and results. These begin with model code and 
documentation availability. Reports describing model results should 
include all important equations, tables of all parameter values, a 
description of any parameter estimation techniques, and tables or 
graphs of results.
    Papers or reports describing studies may not provide all of the 
information needed to evaluate a study's utility for risk assessment. 
Assessors are encouraged to communicate with the principal investigator 
or other study participants to gain information on study plans and 
their implementation. Questions useful for evaluating studies are shown 
in text note 4-3.
4.1.2.1. Evaluating the Purpose and Scope of the Study
    The assessor must often evaluate the utility of a study that was 
designed for a purpose other than risk assessment. In these cases, it 
is important that the objectives and scope of the original study be 
examined to evaluate their compatibility with the objectives and needs 
of the current risk assessment.
    An examination of objectives can identify important uncertainties 
and ensure that the information is used appropriately in the 
assessment. An example is the evaluation of studies that measure 
condition (e.g., stream surveys, population surveys). While the 
measurements used to evaluate condition may be the same as the effects 
measures identified in problem formulation, to support a causal 
argument, effects measures must be linked with stressors. In the best 
case, this means that the stressor should be measured at the same time 
and place as the effect.
    Similarly, a model may have been developed for purposes other than 
risk assessment. The model description should include the intended 
application, theoretical framework, underlying assumptions, and 
limiting conditions. This information can help assessors identify 
important limitations in its application for risk assessment. For 
example, a model developed to evaluate chemical transport in the water 
column alone may have limited utility for a risk assessment of a 
chemical that partitions readily into sediments.
    The variables and conditions examined by studies should also be 
compared with those variables and conditions identified during problem 
formulation. In addition, the range of variability explored in the 
study should be compared with the range of variability of interest for 
the risk assessment. For example, a study that examines habitat needs 
of an animal during the winter may miss important breeding-season 
requirements. In

[[Page 47577]]

general, studies that minimize the amount of extrapolation needed are 
preferred. These are the studies that are designed to represent:
     The measures identified in the analysis plan (i.e., 
measures of exposure, effects, and ecosystem and receptor 
characteristics).
     The time frame of interest, considering seasonality and 
intermittent events.
     The ecosystem and location of interest.
     The environmental conditions of interest.
     The exposure route of interest.
4.1.2.2. Evaluating the Design and Implementation of the Study
    The design and implementation of the study are evaluated to ensure 
that the study objectives were met and that the information is of 
sufficient quality to support the purposes of the risk assessment. The 
study design provides insight into the sources and magnitude of 
uncertainty associated with the results (see section 4.1.3 for further 
discussion of uncertainty). Among the most important design issues for 
studies of effects is whether a study had sufficient power to detect 
important differences or changes. Because this information is rarely 
reported (Peterman, 1990), the assessor may need to calculate the 
magnitude of an effect that could be detected under the study 
conditions (Rotenberry and Wiens, 1985).
    Risk assessors should evaluate evidence that the study was 
conducted properly. For laboratory studies, this may mean determining 
whether test conditions were properly controlled and control responses 
were within acceptable bounds. For field studies, issues include the 
identification and control of potentially confounding variables and the 
careful selection of reference sites. For models, issues include the 
program's structure and logic and the correct specification of 
algorithms in the model code (U.S. EPA, 1994f).
    Study evaluation is easier if a standard method or standard quality 
assurance/quality control (QA/QC) protocols are available and followed 
by the study. However, the assessor still needs to consider whether the 
precision and accuracy goals identified in the standard method were 
achieved and whether these goals are appropriate for the purposes of 
the risk assessment. For example, detection limits identified for one 
environmental matrix may not be achievable for another and may be 
higher than concentrations of interest for the risk assessment. Study 
results can still be useful even if a standard method was not used. 
However, it does place an additional burden on both the authors and the 
assessors to provide and evaluate evidence that the study was conducted 
properly.
4.1.3. Evaluating Uncertainty
    Uncertainty evaluation is an ongoing theme throughout the analysis 
phase. The objective is to describe, and, where possible, quantify what 
is known and not known about exposure and effects in the system of 
interest. Uncertainty analyses increase credibility by explicitly 
describing the magnitude and direction of uncertainties, and they 
provide the basis for efficient data collection of or application of 
refined methods.
    U.S. EPA (1992d) discusses sources of uncertainty that arise during 
the evaluation of information and conceptual model development 
(combined under the subject of scenario uncertainty), when evaluating 
the value of a parameter (e.g., an environmental measurement or the 
results of a toxicity test), and during the development and application 
of models. Uncertainty in conceptual model development is discussed in 
section 3.4.3. Many of the sources of uncertainty discussed by EPA 
(U.S. EPA, 1992d) are relevant to characterizing both exposure and 
ecological effects; these sources and example strategies for the 
analysis phase are shown in table 4-1.

                            Table 4-1.--Uncertainty Evaluation in the Analysis Phase                            
----------------------------------------------------------------------------------------------------------------
      Source of uncertainty           Example analysis phase strategies               Specific example          
----------------------------------------------------------------------------------------------------------------
Unclear communication............  Contact principal investigator or       Clarify whether the study was        
                                    other study participants if             designed to characterize local      
                                    objectives and methods of literature    populations or regional populations.
                                    studies are unclear.                                                        
                                   Document decisions made during the      Discuss rationale for selecting the  
                                    course of the assessment.               critical toxicity study.            
Descriptive errors...............  Verify that data sources followed       Double-check calculations and data   
                                    appropriate QA/QC procedures.           entry.                              
Variability......................  Describe heterogeneity using point      Display differences in species       
                                    estimates (e.g., central tendency and   sensitivity using a cumulative      
                                    high end) or by constructing            distribution function.              
                                    probability or frequency                                                    
                                    distributions.                                                              
                                   Differentiate from uncertainty due to                                        
                                    lack of knowledge.                                                          
Data gaps........................  Describe approaches used for bridging   Discuss rationale for using a factor 
                                    gaps and their rationales.              of 10 to extrapolate between a LOAEL
                                   Differentiate science-based judgments    and a NOAEL.                        
                                    from policy-based judgments.                                                
Uncertainty about a quantity's     Use standard statistical methods to     Present the upper confidence limit on
 true value.                        construct probability distributions     the arithmetic mean soil            
                                    or point estimates (e.g., confidence    concentration, in addition to the   
                                    limits).                                best estimate of the arithmetic     
                                   Evaluate power of designed experiments   mean.                               
                                    to detect differences.                 Ground-truth remote sensing data.    
                                   Consider taking additional data if                                           
                                    sampling error is too large.                                                
                                   Verify location of samples or other                                          
                                    spatial features                                                            
Model structure uncertainty        Discuss key aggregations and model      Discuss combining different species  
 (process models).                  simplifications.                        into a group based on similar       
                                   Compare model predictions with data      feeding habits.                     
                                    collected in the system of interest.                                        

[[Page 47578]]

                                                                                                                
Uncertainty about a model's form   Evaluate whether alternative models     Present results obtained using       
 (empirical models).                should be combined formally or          alternative models.                 
                                    treated separately.                                                         
                                   Compare model predictions with data     Compare results of a plant uptake    
                                    collected in the system of interest.    model with data collected in the    
                                                                            field.                              
----------------------------------------------------------------------------------------------------------------

    Sources of uncertainty that are factors primarily when evaluating 
information include unclear communication of the information to the 
assessor, unclear communication about how the assessor handled the 
information, and errors in the information itself (descriptive errors). 
These sources are usually characterized by critically examining sources 
of information and documenting the rationales for the decisions made 
when handling it. The discussion should allow the reader to make an 
independent judgment about the validity of the decisions reached by the 
assessor.
    Sources of uncertainty that arise primarily when estimating the 
value of a parameter include variability, uncertainty about a 
quantity's true value, and data gaps. The term variability is used here 
to describe the true heterogeneity in a characteristic influencing 
exposure or effects. Examples include the variability in soil organic 
carbon, seasonal differences in animal diets, or differences in 
chemical sensitivity among different species. This heterogeneity is 
usually described during uncertainty analysis, although heterogeneity 
may not reflect a lack of knowledge and cannot usually be reduced by 
further measurement. Variability can be described by presenting a 
distribution or specific percentiles from it (e.g., mean and 95th 
percentile).
    Uncertainty about a quantity's true value may include uncertainty 
about its magnitude, location, or time of occurrence. This uncertainty 
can usually be reduced by taking additional measurements. Uncertainty 
about a quantity's true magnitude is usually described by sampling 
error (or variance in experiments) or measurement error. When the 
quantity of interest is biological response, sampling error can greatly 
influence the ability of the study to detect effects. Properly designed 
studies will specify sample sizes that are sufficiently large to detect 
important signals. Unfortunately, many studies have sample sizes that 
are too small to detect anything but gross changes (Smith and Shugart, 
1994; Peterman, 1990). The discussion should highlight situations where 
the power to detect difference is low. Meta-analysis has been suggested 
as a way to combine results from different studies to improve the 
ability to detect effects (Laird and Mosteller, 1990; Petitti, 1994). 
However, these approaches have been applied primarily in the arena of 
human epidemiology and are still controversial (Mann, 1990).
    Interest in quantifying spatial uncertainty has increased with the 
increasing use of geographic information systems. Strategies include 
verifying the locations of remotely sensed features, ensuring that the 
spatial resolution of data or a method is commensurate with the needs 
of the assessment, and using methods to describe and use the spatial 
structure of data (e.g., Cressie, 1993).
    Nearly every assessment encounters situations where data are 
unavailable or where information is available on parameters that are 
different from those of interest for the assessment. Examples include 
using laboratory animal data to estimate a wild animal's response or 
using a bioaccumulation measurement from an ecosystem other than the 
one interest. These data gaps are usually bridged based on a 
combination of scientific data or analyses, scientific judgement, and 
policy judgement. For example, in deriving an ambient water quality 
criterion (text note 3-16), data and analyses are used to construct 
distributions of species sensitivity for a particular chemical. 
Scientific judgment is used to infer that species selected for testing 
will adequately represent the range of sensitivity of species in the 
environment. Policy judgment is used to define the extent to which 
individual species should be protected (e.g., 90 vs 95 percent of the 
species). It is important to differentiate among these elements when 
key assumptions and the approach used are documented.
    In some circumstances scientists may disagree on the best way to 
bridge data gaps. This lack of consensus can increase uncertainty. 
Confidence can be increased through consensus building techniques such 
as peer reviews, workshops, and other methods to elicit expert opinion. 
Data gaps can often be filled by completing additional studies on the 
unknown parameter. Opportunities for reducing this source of 
uncertainty should be noted and carried through to risk 
characterization. Data gaps that preclude the analysis of exposure or 
ecological effects should also be noted and discussed in risk 
characterization.
    An important objective of characterizing uncertainty in the 
analysis phase is to distinguish variability from uncertainties arising 
from lack of knowledge (e.g., uncertainty about a quantity's true 
value) (U.S. EPA, 1995c). This distinction facilitates the 
interpretation and communication of results. For example, in their food 
web models of herons and mink, MacIntosh et al. (1994) separated 
variability expected among feeding habits of individual animals from 
the uncertainty in the mean concentration of chemical in prey species. 
In this way, the assessors could place error bounds on the distribution 
of exposure among the animals using the site and estimate the 
proportion of the animal population that might exceed a toxicity 
threshold.
    Sources of uncertainty that arise primarily during the development 
and application of models include the structure of process models and 
the description of the relationship between two or more variables in 
empirical models. Process model description should include key 
assumptions, simplifications, and aggregations of variables (see text 
note 4-4). Empirical model descriptions should include the rationale 
for selection, and statistics on model performance (e.g., goodness of 
fit). Uncertainty in process or empirical models can be quantitively 
evaluated by comparing model results to measurements taken in the 
system of interest or by comparing the results obtained using different 
model alternatives.
    Methods for analyzing and describing uncertainty can range from 
simple to complex. The calculation of one or more point estimates is 
one of the most common approaches to presenting analysis results; point 
estimates that reflect different aspects of uncertainty can have great 
value if appropriately developed and communicated. Classical 
statistical methods (e.g., confidence limits, percentiles) can be 
readily applied to describing uncertainty in parameters. When a 
modeling approach

[[Page 47579]]

is used, sensitivity analysis can be used to evaluate how model output 
changes with changes in input variables, and uncertainty propagation 
can be analyzed to examine how uncertainty in individual parameters can 
affect the overall uncertainty of the assessment. The availability of 
software for Monte-Carlo analysis has greatly increased the use of 
probabilistic methods; readers are encouraged to follow best practices 
that have been suggested (e.g., Burmaster and Anderson, 1994; Haimes et 
al. 1994). Other methods (e.g., fuzzy mathematics, Bayesian 
methodologies) are available, but have not yet been extensively applied 
to ecological risk assessment (Smith and Shugart, 1994). These 
guidelines do not endorse the use of any one method over others and 
note that the poor execution of any method can obscure rather than 
clarify the impact of uncertainty on an assessment's results. No matter 
what technique is used, the sources of uncertainty discussed above 
should be addressed.

4.2. Characterization of Exposure

    Exposure characterization describes the contact or co-occurrence of 
stressors with ecological receptors. The characterization is based on 
measures of exposure and of ecosystem and receptor characteristics (the 
evaluation of this information is discussed in section 4.1). These 
measures are used to analyze stressor sources, their distribution in 
the environment, and the extent and pattern of contact or co-occurrence 
(discussed in section 4.2.1). The objective is to produce a summary 
exposure profile (section 4.2.2) that identifies the receptor (i.e., 
the exposed ecological entity), describes the course a stressor takes 
from the source to the receptor (i.e., the exposure pathway), and 
describes the intensity and spatial and temporal extent of co-
occurrence or contact. The profile also describes the impact of 
variability and uncertainty on exposure estimates and reaches a 
conclusion about the likelihood that exposure will occur.
    The exposure profile is combined with an effects profile (discussed 
in section 4.3.2) to estimate risks. For the results to be useful, they 
must be compatible with the stressor-response relationship generated in 
the effects characterization.
4.2.1. Exposure Analyses
    Exposure is analyzed by describing the source and releases, the 
distribution of the stressor in the environment, and the extent and 
pattern of contact or co-occurrence. The order of discussion of these 
topics is not necessarily the order in which they are evaluated in a 
particular assessment. For example, the assessor may start with 
information about tissue residues, and attempt to link these residues 
with a source.
4.2.1.1. Describe the Source
    A source description identifies where the stressor originates, 
describes what stressors are generated, and considers other sources of 
the stressor. Exposure analyses may start with the source when it is 
known, but some analyses may begin with known exposures and attempt to 
link them to sources, while other analyses may start with known 
stressors and attempt to identify sources and quantify contact. The 
source is the first component of the exposure pathway and significantly 
influences where and when stressors eventually will be found. In 
addition, many management alternatives focus on modifying the source. 
Text note 4-5 provides some useful questions.
    A source can be defined in several ways--as the place where the 
stressor is released (e.g., a smoke stack, historically contaminated 
sediments) or the management practice or action (e.g., dredging) that 
produces stressors. In some assessments, the original source no longer 
exists and the source is defined as the current origin of the 
stressors. For example, the source may be defined as contaminated 
sediments because the industrial plant that produced the contaminants 
no longer operates.
    In addition to identifying the source, the assessor describes the 
generation of stressors in terms of intensity, timing, and location. 
The location of the source and the environmental medium that first 
receives stressors are two attributes that deserve particular 
attention. In addition, the source characterization should consider 
whether other constituents emitted by the source influence transport, 
transformation, or bioavailability of the stressor of interest. For 
example, the presence of chloride in the feedstock of a coal-fired 
power plant influences whether mercury is emitted in divalent (e.g., as 
mercuric chloride) or elemental form (Meij, 1991). In the best case, 
stressor generation is measured or modeled quantitatively; however, 
sometimes it can only be qualitatively described.
    Many stressors have natural counterparts or multiple sources, and 
the characterization of these other sources can be an important 
component of the analysis phase. For example, many chemicals occur 
naturally (e.g., most metals), are generally widespread due to other 
sources (e.g., polycyclic aromatic hydrocarbons in urban ecosystems), 
or may have significant sources outside the boundaries of the current 
assessment (e.g., atmospheric nitrogen deposited in Chesapeake Bay). 
Many physical stressors also have natural counterparts. For example, 
construction activities may add fine sediments to a stream in addition 
to those from a naturally undercut bank. In addition, human activities 
may change the magnitude or frequency of natural disturbance cycles. 
For example, development may decrease the frequency but increase the 
severity of fires or may increase the frequency and severity of 
flooding in a watershed.
    The way multiple sources are evaluated during the analysis phase 
depends on the objectives of the assessment articulated during problem 
formulation. Options include (in order of increasing complexity):
     Focus only on the source under evaluation and calculate 
incremental risks attributable to that source (common for assessments 
initiated with an identified source or stressor).
     Consider all sources of a stressor and calculate total 
risks attributable to that stressor. Relative source attribution can be 
accomplished as a separate step (common for assessments initiated with 
an observed effect or an identified stressor).
     Consider all stressors influencing an assessment endpoint 
and calculate cumulative risks to that endpoint (common for assessments 
initiated because of concern for an ecological value).
    Source characterization can be particularly important for new 
biological stressors, since many of the strategies for reducing risks 
focus on preventing entry in the first place. Once the source is 
identified, the likelihood of entry may be characterized qualitatively. 
For example, in their analysis of risks from importation of Chilean 
logs, the assessment team concluded that the beetle Hylurgus ligniperda 
had a high potential for entry into the United States. They based this 
conclusion on the fact that they are attracted to freshly cut logs and 
tend to burrow under the bark and thus would be protected during 
transport (USDA, 1993).
    The description of the source can set the stage for the second 
objective of exposure analysis, which is describing the distribution of 
the stressor in the environment.
4.2.1.2. Describe the Distribution of the Stressor or Disturbed 
Environment
    The second objective of exposure analyses is to describe the 
spatial and temporal distribution of the stressor in

[[Page 47580]]

the environment. For physical stressors that directly alter or 
eliminate portions of the environment, the assessor describes the 
temporal and spatial distribution of the disturbed environment. Because 
exposure occurs where receptors co-occur with or contact stressors in 
the environment, characterizing the spatial and temporal distribution 
of a stressor is a necessary precursor to estimating exposure. The 
stressor's distribution in the environment is described by evaluating 
the pathways that stressors take from the source as well as the 
formation and subsequent distribution of secondary stressors.
    Evaluating Transport Pathways. There are many pathways by which 
stressors can be transported in the environment (see text note 4-7). An 
evaluation of transport pathways can help ensure that measurements are 
taken in the appropriate media and locations and that models include 
the most important processes.
    For chemical stressors, the evaluation of pathways usually begins 
by determining into which media a chemical will partition. Key 
considerations include physicochemical properties such as solubility 
and vapor pressure. For example, lipophilic chemicals tend to be found 
in environmental compartments with higher proportions of organic 
carbon, such as soils, sediments, and biota. From there, the evaluation 
may examine the transport of the contaminated medium. Because 
constituents of chemical mixtures may have different properties, it is 
important to consider how the composition of a mixture may change over 
time or as it moves through the environment. Guidance on evaluating the 
fate and transport of chemicals is beyond the scope of these 
guidelines; readers are referred to the exposure assessment guidelines 
(U.S. EPA, 1992d) for additional information.
    The attributes of physical stressors may also influence where the 
stressors will go. For example, the size of silt particles determines 
where they will eventually deposit in a stream. Physical stressors that 
eliminate ecosystems or portions of them (e.g., logging activity or the 
construction of dams or parking lots) may require no modeling of 
pathways--the wetland is filled, the fish are harvested, or the valley 
is flooded. For these direct disturbances, the challenge is usually to 
evaluate the formation of secondary stressors and the effects 
associated with the disturbance.
    The dispersion of biological stressors has been described in two 
ways, as diffusion and jump-dispersal (Simberloff and Alexander, 1994). 
Diffusion involves a gradual spread from the establishment site and is 
a function primarily of reproductive rates and motility. The other 
movement pattern, jump-dispersal, involves erratic spreads over periods 
of time, usually by means of a vector. The gypsy moth and zebra mussel 
have spread this way; the gypsy moth via egg masses on vehicles and the 
zebra mussel via boat ballast water. Biological stressors can use both 
diffusion and jump-dispersal strategies, and often one or more 
mechanisms are important. This makes dispersal rates very difficult to 
predict. Key considerations include the availability of vectors, 
whether the organism has natural attributes that enhance dispersal 
(e.g., ability to fly, adhere to objects, disperse reproductive units), 
and the habitat or host needs of the organism.
    For biological stressors, assessors must consider the additional 
factors of survival and reproduction. There is a wide range of 
strategies organisms use to survive in adverse conditions, for example, 
fungi form resting stages such as sclerotia and chlamydospores and some 
amphibians became dormant during drought. The survival of some 
organisms can be measured to some extent under laboratory conditions. 
However, it may be impossible to determine how long some resting stages 
(e.g., spores) can survive under adverse conditions; many can remain 
viable for years. Similarly, reproductive rates may vary substantially, 
depending on specific environmental conditions. Therefore, while life-
history data such as temperature and substrate preferences, important 
predators, competitors or diseases, habitat needs, and reproductive 
rates are of great value, they must be interpreted with caution.
    Ecosystem characteristics influence the transport of all types of 
stressors. The challenge is to determine the particular aspects of the 
ecosystem that are most important. In some cases, ecosystem 
characteristics that influence distribution are known. For example, 
fine sediments tend to accumulate in areas of low energy in streams 
such as pools and backwaters. In other cases, much more professional 
judgment is needed. For example, when evaluating the likelihood that an 
introduced organism will become established, it is useful to know 
whether the ecosystem is generally similar to or different from the one 
where the biological stressor originated. In this case, professional 
judgment is needed to determine which characteristics of the current 
and original ecosystems should be compared.
    Evaluating Secondary Stressors. The creation of secondary stressors 
can greatly alter conclusions about risk. Secondary stressors can be 
formed through biotic or abiotic transformation processes and may be of 
greater or lesser concern than the primary stressor. Evaluating the 
formation of secondary stressors is usually done as part of exposure 
characterization; however, coordination with the ecological effects 
characterization is important to ensure that all potentially important 
secondary stressors are evaluated.
    For chemicals, the evaluation of secondary stressors usually 
focuses on metabolites or degradation products or chemicals formed 
through abiotic processes. For example, microbial action increases the 
bioaccumulation of mercury by transforming it from inorganic form to 
organic forms. Many azo dyes are not toxic because of their large 
molecular size but, in an anaerobic environment, the polymer is 
hydrolyzed into more toxic water-soluble units. In addition, secondary 
stressors can be formed through ecosystem processes. For example, 
nutrient inputs into an estuary can decrease dissolved oxygen 
concentrations because they increase primary production and subsequent 
decomposition. While the possibility and rates of transformation can be 
investigated in the laboratory, rates in the field may differ 
substantially, and some processes may be difficult or impossible to 
replicate in a laboratory. When evaluating field information, though, 
it may be difficult to distinguish between transformation processes 
(e.g., degradation of oil constituents by microorganisms) and transport 
processes (e.g., loss of oil constituents through volatilization).
    Disturbances can also generate secondary stressors, and identifying 
the specific consequences that will affect the assessment endpoint can 
be a difficult task. For example, the removal of riparian vegetation 
can generate many secondary stressors, including increased nutrients, 
stream temperature, sedimentation, and altered stream flow. However, it 
may be the resulting increase in stream temperature that is the primary 
cause of adult salmon mortality in a particular stream.
    The distribution of stressors in the environment can be described 
using measurements, models, or a combination of the two. If stressors 
have already been released, direct measurements of environmental media 
or a combination of modeling and measurement is preferred. However, a 
modeling approach may be necessary if the assessment is intended to 
predict future scenarios or if measurements are

[[Page 47581]]

not possible or practicable. Considerations for evaluating data 
collection and modeling studies are discussed in section 4.1. For 
chemical stressors, we also refer readers to the exposure assessment 
guidelines (U.S. EPA, 1992d). For biological stressors, the 
distribution in the environment is difficult to predict quantitatively. 
If measurements in the environment cannot be taken, distribution can be 
evaluated qualitatively by considering the potential for transport, 
survival, and reproduction (see above).
    By the end of this step, the environmental distribution of the 
stressor or the disturbed environment should be described. This 
description can be an important precursor to the next objective of 
exposure analysis--estimating the contact or co-occurrence of the 
stressor with ecological entities. In cases where the extent of contact 
is known, describing the environmental distribution of the stressor can 
help identify potential sources, and ensure that all important 
exposures have been addressed. In addition, by identifying the pathways 
a stressor takes from a source, the second component of an exposure 
pathway is described.
4.2.1.3. Describe Contact or Co-occurrence
    The third objective of the exposure analysis is to describe the 
extent and pattern of co-occurrence or contact between a stressor and a 
receptor (i.e., exposure). The objective of this step is to describe 
the intensity and temporal and spatial extent of exposure in a form 
that can be compared with the stressor-response profile generated in 
the effects assessment. The description of exposure is a critical 
element of estimating risk--if there is no exposure, there can be no 
risk. Questions for describing contact or co-occurrence are shown in 
text note
4-8.
    Exposure can be described in terms of co-occurrence of the stressor 
with receptors, of the actual contact of a stressor with receptors, or 
of the uptake of a stressor into a receptor. The terms by which 
exposure is described depend on how the stressor causes adverse 
effects. Co-occurrence is particularly useful for evaluating stressors 
that can cause effects without actually contacting ecological 
receptors. For example, whooping cranes use sandbars in rivers for 
their nesting areas, and they prefer sandbars with unobstructed views. 
Manmade obstructions, such as bridges, can interfere with nesting 
behavior without ever actually contacting the birds. Most stressors, 
however, must contact receptors to cause an effect. For example, flood 
waters must contact tree roots before their growth is impaired. 
Finally, some stressors must not only be contacted, but also must be 
internally absorbed. For example, a toxicant that causes liver tumors 
in fish must be absorbed through the gills and reach the target organ 
to cause the effect.
    Co-occurrence is evaluated by comparing the distribution of the 
stressor with the distribution of the ecological receptor. For example, 
maps of the stressor may be overlaid with maps of ecological receptors 
(e.g., the placement of bridges overlaid on maps showing habitat 
historically used for crane nests). The increased availability of 
geographic information systems (GIS) has provided new tools for 
evaluating co-occurrence.
    Contact is a function of the amount of a stressor in an 
environmental medium and activities or behavior that brings receptors 
into contact with the stressor. For biological stressors, this step 
relies extensively on professional judgment; contact is often assumed 
to occur in areas where the two overlap. For chemicals, contact is 
quantified as the amount of a chemical ingested, inhaled, or in 
material applied to the skin (i.e., the potential dose). In its 
simplest form, it is quantified as an environmental concentration, with 
the assumptions that the chemical is well mixed and that the organism 
contacts a representative concentration. This approach is commonly used 
for respired media (e.g., water for aquatic organisms, air for 
terrestrial organisms). For ingested media (e.g., food, soil), another 
common approach combines modeled or measured concentrations of the 
contaminant with assumptions or parameters describing the contact rate 
(U.S. EPA, 1993c) (see text note 4-9).
    Uptake is evaluated by considering the amount of stressor that is 
internally absorbed into an organism. Uptake is a function of the 
stressor (e.g., a chemical's form or valence state), the medium (e.g., 
sorptive properties or presence of solvents), the biological membrane 
(e.g., integrity, permeability), and the organism (e.g., sickness, 
active uptake) (Suter et al., 1994). Because of interactions among 
these four factors, uptake will vary on a situation-specific basis. 
Uptake is usually assessed by modifying an estimate of contact with a 
factor indicating the proportion of the stressor that is available for 
uptake (i.e., the bioavailable fraction) or actually absorbed. 
Absorption factors and bioavailability measured for the chemical, 
ecosystem, and organism of interest are preferred. Internal dose can 
also be evaluated by using a pharmacokinetic model or by measuring 
biomarkers or residues in receptors (see text note 4-10). Most 
stressor-response relationships express the amount of stressor in terms 
of media concentration or potential dose rather than internal dose; 
this limits the utility of using estimates of uptake for risk 
estimation. However, biomarkers and tissue residues can provide 
valuable confirmatory evidence that exposure has occurred, and tissue 
residues in prey organisms can be used for estimating risks to their 
predators.
    The characteristics of the ecosystem and receptors must be 
considered to reach appropriate conclusions about exposure. Abiotic 
attributes may increase or decrease the amount of a stressor contacted 
by receptors. For example, the presence of naturally anoxic areas above 
contaminated sediments in an estuary may reduce the amount of time that 
bottom-feeding fish spend in contact with the contaminated sediments 
and thereby reduce exposure to the contamination. Biotic interactions 
can also influence exposure. For example, competition for high-quality 
resources may force some organisms to utilize disturbed areas. The 
interaction between exposure and receptor behavior can influence both 
the initial and subsequent exposures. For example, some chemicals 
reduce the prey's ability to escape predators and thereby may increase 
predator exposure to the chemical as well as the prey's risk of 
predation. Alternatively, organisms may avoid areas, food, or water 
with contamination they can detect. While avoidance can reduce exposure 
to chemicals, it may increase other risks by altering habitat usage or 
other behavior.
    Three dimensions must be considered when estimating exposure: 
intensity, time, and space. Intensity is the most familiar dimension 
for chemical and biological stressors and may be expressed as the 
amount of chemical contacted per day or the number of pathogenic 
organisms per unit area.
    The temporal dimension of exposure has aspects of duration, 
frequency, and timing. Duration can be expressed as the time over which 
exposure occurs, exceeds some threshold intensity, or over which 
intensity is integrated. If exposure occurs as repeated, discrete 
events of about the same duration (e.g., floods), frequency is the 
important temporal dimension of exposure. If the repeated events have 
significant and variable durations, both duration and frequency must be 
considered. In addition, the timing of exposure, including the order or 
sequence of events, can be an important factor to describe. For 
example, in the Northeast, lakes receive high concentrations of 
hydrogen ions and aluminum during

[[Page 47582]]

snow melt; this period also corresponds to the sensitive life stages of 
some aquatic organisms.
    In chemical assessments, the dimensions of intensity and time are 
often combined by averaging intensity over time. The duration over 
which intensity is averaged is determined by considering both the 
ecological effects of concern and the likely pattern of exposure. For 
example, an assessment of bird kills associated with granular 
carbofuran focused on short-term exposures because the effect of 
concern was acute lethality (Houseknecht, 1993). Because toxicological 
tests are usually conducted using constant exposures, the most 
realistic comparisons between exposure and effects are made when 
exposure in the real world does not vary substantially. In these cases, 
the arithmetic average exposure over the time period of toxicological 
significance is the appropriate statistic to use (U.S. EPA, 1992d). 
However, as concentrations or contact rates become more episodic or 
variable, the arithmetic average may not reflect the toxicologically 
significant aspect of the exposure pattern. In extreme cases, averaging 
may not be appropriate at all, and assessors may need to use a toxic 
dynamic model to assess chronic effects.
    Spatial extent is another dimension of exposure. It is most 
commonly expressed in terms of area (e.g., hectares of filled wetland, 
square meters that exceed a particular chemical threshold). At larger 
spatial scales, however, the shape or arrangement of exposure may be an 
important issue, and area alone may not be the appropriate descriptor 
of spatial extent for risk assessment. A general solution to the 
problem of incorporating pattern into ecological assessments has yet to 
be developed; however, the emerging field of landscape ecology and the 
increased availability of geographic information systems have greatly 
expanded the options for analyzing and presenting the spatial dimension 
of exposure.
    This step completes exposure analysis. Exposure should be described 
in terms of intensity, space, and time, in units that can be combined 
with the effects assessment. In addition, the assessor should be able 
to trace the paths of stressors from the source to the receptors, 
completing the exposure pathway. The results of exposure analysis are 
summarized in the exposure profile, which is discussed in the next 
section.
4.2.2. Exposure Profile
    The final product of exposure analysis is a summary profile of what 
has been learned. Depending on the risk assessment, the profile may be 
a written document, or a module of a larger process model. 
Alternatively, documentation may be deferred until risk 
characterization. In any case, the objective is to ensure that the 
information needed for risk characterization has been collected and 
evaluated. In addition, compiling the exposure profile provides an 
opportunity to verify that the important exposure pathways identified 
in the conceptual model were evaluated.
    The exposure profile identifies the receptor and describes the 
exposure pathways and intensity and spatial and temporal extent of co-
occurrence or contact. It also describes the impact of variability and 
uncertainty on exposure estimates and reaches a conclusion about the 
likelihood that exposure will occur (text note 4-11).
    The profile should describe the relevant exposure pathways. If 
exposure can occur through many pathways, it may be useful to rank 
them, perhaps by contribution to total exposure. For example, consider 
an assessment of risks to grebes feeding on a mercury-contaminated 
lake. The grebes may be exposed to methyl mercury in fish that 
originated from historically contaminated sediments. They may also be 
exposed by drinking lake water, but comparing the two exposure pathways 
may show that the fish pathway contributes the vast majority of 
exposure to mercury.
    The profile should describe the ecological entity that is exposed 
and represented by the exposure estimates described below. For example, 
the exposure profile may focus on the local population of grebes 
feeding on a specific lake during the summer months.
    The assessor should state how each of the three general dimensions 
of exposure (intensity, time, and space) was treated and why that 
treatment is necessary or appropriate. Continuing with the grebe 
example, exposure might be expressed as the daily potential dose 
averaged over the summer months and over the extent of the lake.
    The profile should also describe how variability in receptor 
attributes or stressor levels can change exposure. For example, 
variability in receptor attributes of the grebes may be addressed by 
using data on how the proportion of fish in the diet varies among 
individuals. If several lakes were the subject of the assessment and 
individual grebes tended to feed on the same lake throughout the 
season, variability in stressor levels could be addressed by comparing 
exposures among the lakes.
    Variability can be described by using a distribution or by 
describing where a point estimate is expected to fall on a 
distribution. Cumulative-distribution functions (CDFs) and probability-
density functions (PDFs) are two common presentation formats; (see 
Appendix B, figures B1 and B2). Figures 5-4 to 5-6 show examples of 
cumulative frequency plots of exposure data. The point estimate/
descriptor approach is used when there is not enough information to 
describe a distribution. We recommend using the descriptors discussed 
in U.S. EPA, 1992d, including central tendency to refer to the mean or 
median of the distribution, high end to refer to exposure estimates 
that are expected to fall between the 90th and 99.9th percentile of the 
exposure distribution, and bounding estimates to refer to those higher 
than any actual exposure.
    The exposure profile should summarize important uncertainties 
(i.e., lack of knowledge) (see section 4.1.3 for a discussion of the 
different sources of uncertainty). In particular, the assessor should:
     Identify key assumptions and describe how they were 
handled.
     Discuss (and quantify if possible) the magnitude of 
sampling and/or measurement error.
     Identify the most sensitive variables influencing 
exposure.
     Identify which uncertainties can be reduced through the 
collection of more data.
    Uncertainty about a quantity's true value can be shown by 
calculating error bounds on a point estimate, as shown in figure 5-2.
    All of the above information is synthesized to reach a conclusion 
about the likelihood that exposure will occur. The exposure profile is 
one of the products of the analysis phase. It is combined with the 
stressor-response profile (the product of the ecological effects 
characterization discussed in the next section) during risk 
characterization.

4.3. Characterization of Ecological Effects

    Characterization of ecological effects describes the effects that 
are elicited by a stressor, links these effects with the assessment 
endpoints, and evaluates how the effects change with varying stressor 
levels. Ecological effects characterization begins by evaluating 
effects data (discussed generally in section 4.1) to further specify 
the effects that are elicited, confirm that the effects are consistent 
with the assessment endpoints, and confirm that the conditions under 
which they occur are

[[Page 47583]]

consistent with the conceptual model. Once the effects of interest are 
identified, then an ecological response analysis (section 4.3.1) is 
conducted to evaluate how the magnitude of the effects change with 
varying stressor levels, evaluate the evidence that the stressor causes 
the effect, and link the effects with the assessment endpoint. The 
conclusions of the ecological effects characterization are summarized 
in a stressor-response profile (section 4.3.2).
4.3.1. Ecological Response Analysis
    Ecological response analysis has three primary elements: 
determining the relationship between stressor levels and ecological 
effects (section 4.3.1.1), evaluating the plausibility that effects may 
occur or are occurring as a result of exposure to stressors (section 
4.3.1.2), and linking measurable ecological effects with the assessment 
endpoints when assessment endpoints cannot be directly measured 
(section 4.3.1.3).
4.3.1.1. Stressor-Response Analysis
    Evaluating ecological risks requires an understanding of the 
relationships between stressor levels and resulting ecological 
responses. The stressor-response relationships used in a particular 
assessment depend on the scope and nature of the ecological risk 
assessment as defined in problem formulation and reflected in the 
analysis plan. For example, an assessor may need a point estimate of an 
effect (such as an LC50) to compare with point estimates from 
other stressors. The shape of the stressor-response curve may be 
critical for determining the presence or absence of an effects 
threshold or for evaluating incremental risks, or stressor-response 
curves may be used as input for ecological effects models. If 
sufficient data are available, the risk assessor may construct 
cumulative distribution functions using multiple point estimates of 
effects. Or the assessor may use process models that already 
incorporate empirically derived stressor-response relationships 
(section 4.3.1.3). Some questions for stressor-response analysis are 
provided in text note 4-12.
    This section describes a range of stressor-response approaches 
available to risk assessors following a theme of variations on the 
classical stressor-response relationship (e.g., figure 4-2). While 
quantifying this relationship is encouraged, qualitative stressor-
response evaluations are also possible (text note 4-13). In addition, 
many stressor-response relationships are more complex than the simple 
curve shown in this figure. Ecological systems frequently show 
responses to stressors that may involve abrupt shifts to new community 
or system types (Holling, 1978).

BILLING CODE 6560-50-P

[[Page 47584]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.005



BILLING CODE 6560-50-C

[[Page 47585]]

    In simple cases, the response will be one variable (e.g., 
mortality, incidence of abnormalities), and most quantitative 
techniques have been developed for univariate analysis. If the response 
of interest is composed of many individual variables (e.g., species 
abundances in an aquatic community), multivariate statistical 
techniques may be useful. These techniques have a long history of use 
in ecology (see texts by Gauch, 1982; Pielou, 1984; Ludwig and 
Reynolds, 1988) but have not yet been extensively applied in risk 
assessment.
    Stressor-response relationships can be described using any of the 
dimensions of exposure (i.e., intensity, time, or space). Intensity is 
probably the most familiar dimension and is often used for chemicals 
(e.g., dose, concentration). The duration of exposure is also commonly 
used for chemical stressor-response relationships; for example, median 
acute effects levels are always associated with a time parameter (e.g., 
24 hr, 48 hr, 96 hr). As noted in text note 4-13, the timing of 
exposure was the critical dimension in evaluating the relationship 
between seed germination and flooding (Pearlstine et al., 1985). The 
spatial dimension is often of concern for physical stressors. For 
example, the spatial extent of suitable habitat was related to the 
probability of sighting a spotted owl (Thomas et al., 1990), and water-
table depth was related to the growth of tree species by Phipps (1979).
    Single-point estimates and stressor-response curves can be 
generated for some biological stressors. For pathogens such as bacteria 
and fungi, inoculum levels (e.g., spores per ml; propagules per unit of 
substrate) may be related to the level of symptoms in a host (e.g., 
lesions per area of leaf surface, total number of plants infected) or 
actual signs of the pathogen (asexual or sexual fruiting bodies, 
sclerotia, etc.). For other biological stressors such as introduced 
species, developing simple stressor-response relationships may be 
inappropriate.
    Data from individual experiments can be used to develop curves and 
point estimates both with and without associated uncertainty estimates 
(see figures 5-2 and 5-3). The advantages of curve-fitting approaches 
include using all of the available experimental data and the ability to 
interpolate to values other than the data points measured. If 
extrapolation outside the range of experimental data is required, risk 
assessors should justify that the observed experimental relationships 
remain valid. A disadvantage of curve fitting is that the number of 
data points required to complete an analysis may not always be 
available. For example, while standard toxicity tests with aquatic 
organisms frequently contain sufficient experimental treatments to 
permit regression analysis, frequently this is not the case for 
toxicity tests with wildlife species.
    Risk assessors sometimes use curve-fitting analyses to determine 
particular levels of effect for evaluation. These point estimates are 
interpolated from the fitted line. Point estimates may be adequate for 
simple assessments or comparative studies of risk and are also useful 
if a decision rule for the assessment was identified during the 
planning phase (see section 2). Median effect levels (text note 4-14) 
are frequently selected because the level of uncertainty is minimized 
at the midpoint of the regression curve. While a 50% effect for an 
endpoint such as survival may not be appropriately protective for the 
assessment endpoint, median effect levels can be used for preliminary 
assessments or comparative purposes, especially when used in 
combination with uncertainty modifying factors (see text note 5-2). 
Selection of a different effect level (10%, 20%, etc.) can be arbitrary 
unless there is some clearly defined benchmark for the assessment 
endpoint. Thus, it is preferable to carry several levels of effect or 
the entire stressor-response curve forward to risk estimation.
    When risk assessors are particularly interested in effects at lower 
stressor levels, they may seek to establish ``no-effect'' levels of a 
stressor based on comparisons between experimental treatments and 
controls. Statistical hypothesis testing is frequently used for this 
purpose. (Note that statistical hypotheses are different from the risk 
hypotheses discussed in problem formulation; see text note 3-10). An 
example of this approach for deriving chemical no-effect levels is 
provided in text note 4-15. An advantage of statistical hypothesis 
testing is that the risk assessor is not required to pick a particular 
effect level of concern. The no-effect level is determined instead by 
experimental conditions such as the number of replicates as well as the 
variability inherent in the data. Thus it is important to consider the 
level of effect detectable in the experiment (i.e., its power) in 
addition to reporting the no-effect level. Another drawback of this 
approach is that it is difficult to evaluate effects associated with 
stressor levels other than the actual treatments tested. Several 
investigators (Stephan and Rogers, 1985; Suter, 1993a) have proposed 
using regression analysis as an alternative to statistical hypothesis 
testing.
    In observational field studies, statistical hypothesis testing is 
often used to compare site conditions with a reference site(s). The 
difficulties of drawing proper conclusions from these types of studies 
(which frequently cannot employ replication) have been discussed by 
many investigators, including Hurlbert (1984), Stewart-Oaten et al. 
(1986), Wiens and Parker (1995), and Eberhardt and Thomas (1991). Risk 
assessors should examine whether sites were carefully matched to 
minimize differences other than the stressor and consider whether 
potential covariates should be included in any analysis. An advantage 
of experimental field studies is that treatments can be replicated, 
increasing the confidence that observed differences are due to the 
treatment.
    Data available from multiple experiments can be used to generate 
multiple point estimates that can be displayed as cumulative 
distribution functions. Figure 5-6 shows an example of a cumulative 
distribution function for species sensitivity derived from multiple 
point estimates (EC5s) for freshwater algae exposed to a herbicide. 
These distributions facilitate identification of stressor levels that 
affect a minority or majority of species. A limiting factor in the use 
of cumulative frequency distributions is the amount of data needed as 
input. Cumulative effects distribution functions can also be derived 
from models that use Monte Carlo or other methods to generate 
distributions based on measured or estimated variation in input 
parameters for the models.
    When multiple stressors are present, stressor-response analysis is 
particularly challenging. Stressor-response relationships can be 
constructed for each stressor separately and then combined. 
Alternatively, the relationship between response and the suite of 
stressors can be combined in one analysis. It is preferable to directly 
evaluate complex chemical mixtures present in environmental media 
(e.g., wastewater effluents, contaminated soils; U.S. EPA, 1986b), but 
it is important to consider the relationship between the samples tested 
and the potential spatial and temporal variability in the mixture. The 
approach taken for multiple stressors depends on the feasibility of 
measuring the suite of stressors and whether an objective of the 
assessment is to project different stressor combinations.
    In some cases, multiple regression analysis can be used to 
empirically relate multiple stressors and a response. Detenbeck (1994) 
used this approach to evaluate change in the water quality of

[[Page 47586]]

wetlands resulting from multiple physical stressors. Multiple 
regression analysis can be difficult to interpret if the explanatory 
variables (i.e., the stressors) are not independent. Principal 
components analysis can be used to extract independent explanatory 
variables formed from linear combinations of the original variables 
(Pielou, 1984).
4.3.1.2. Establishing Cause and Effect Relationships (Causality)
    Causality is the relationship between cause (one or more stressors) 
and effect (assessment endpoint response to one or more stressors). 
Without a sound basis for linking cause and effect, uncertainty in the 
conclusions of an ecological risk assessment is likely to be high. 
Developing causal relationships is especially important for risk 
assessments driven by observed adverse ecological effects such as bird 
or fish kills or a shift in the species composition of an area. This 
section proposes considerations for evaluating causality based on 
criteria primarily for observational data developed by Fox (1991) and 
additional criteria for experimental evaluation of causality modified 
from Koch's postulates (e.g. see Woodman and Cowling, 1987).
    Evidence of causality may be derived from observational evidence 
(e.g., bird kills are associated with field application of a pesticide) 
or experimental data (e.g., laboratory tests with the pesticides in 
question show bird kills at levels similar to those found in the 
field), and causal associations can be strengthened when both types of 
information are available. But since not all situations lend themselves 
to formal experimentation, scientists have looked for other criteria, 
based largely on observation rather than experiment, to support a 
plausible argument for cause and effect. Text note 4-16 provides 
criteria based on Fox (1991) that are very similar to others reviewed 
by Fox (U.S. Department of Health, Education, and Welfare, 1964; Hill, 
1965; Susser, 1986a,b). While data to support some criteria may be 
incomplete or missing for any given assessment, these criteria offer a 
useful way of evaluating available information.
    The strength of association between stressor and response is often 
the main reason that adverse effects (such as bird kills) are first 
noticed. A stronger response to a hypothesized cause is more likely to 
indicate true causation. Additional strong evidence of causation is 
when a response follows after a change in the hypothesized cause 
(predictive performance).
    The presence of a biological gradient or stressor-response 
relationship is another important criterion for causality. The 
stressor-response relationship need not be linear. It can be a 
threshold, sigmoidal, or parabolic phenomenon, but in any case it is 
important that it can be demonstrated. Biological gradients, such as 
decreasing effects downstream of a toxic discharge, are frequently used 
as evidence of causality. To be credible, such relationships should be 
consistent with current biological or ecological knowledge (biological 
plausibility).
    A cause-effect relationship that is demonstrated repeatedly 
(consistency of association) provides strong evidence of causality. 
Consistency may be shown by a greater number of instances of 
association between stressor and response, occurrences in diverse 
ecological systems, or associations demonstrated by diverse methods 
(Hill, 1965). Fox (1991) adds that in ecoepidemiology the occurrence of 
an association in more than one species and species population is very 
strong evidence for causation. An example would be the numerous species 
of birds that were killed as a result of carbofuran application 
(Houseknecht, 1993). Fox (1991) also believes that causality is 
supported if the same incident is observed by different persons under 
different circumstances and at different times.
    Conversely, inconsistency in association between stressor and 
response is strong evidence against causality (e.g., the stressor is 
present without the expected effect, or the effect occurs but the 
stressor is not found). Temporal incompatibility (i.e., the presumed 
cause does not precede the effect) and incompatibility with 
experimental or observational evidence (factual implausibility) are 
also indications against a causal relationship.
    Two other criteria may be of some help in defining causal 
relationships: specificity of an association and probability. The more 
specific the effect, the more likely it is to have a consistent cause. 
However, Fox (1991) argues that effect specificity does little to 
strengthen a causal claim. Disease can have multiple causes, a 
substance can behave differently in different environments or cause 
several different effects, and biochemical events may result in a 
diverse array of biological responses. But in general, the more 
specific or localized the effects, the easier it is to identify the 
cause. Sometimes, a stressor may have a distinctive mode of action that 
suggests its role. Yoder and Rankin (1995) found that patterns of 
change observed in fish and benthic invertebrate communities could 
serve as indicators for different types of anthropogenic impact (e.g., 
nutrient enrichment vs. toxicity).
    For some pathogenic biological stressors, the causal evaluations 
proposed by Koch (text note 4-17) may be useful. For chemicals, 
ecotoxicologists have slightly modified Koch's postulates to provide 
evidence of causality (Adams, 1963; Woodman and Cowling, 1987). The 
modifications are:
      The injury, dysfunction, or other putative effect of the 
toxicant must be regularly associated with exposure to the toxicant and 
any contributory causal factors.
      Indicators of exposure to the toxicant must be found in 
the affected organisms.
      The toxic effects must be seen when normal organisms or 
communities are exposed to the toxicant under controlled conditions, 
and any contributory factors should be manifested in the same way 
during controlled exposures.
      The same indicators of exposure and effects must be 
identified in the controlled exposures as in the field.
    These modifications are conceptually identical to Koch's 
postulates. While useful, this approach may not be practical if 
resources for experimentation are not available or if an adverse effect 
may be occurring over such a wide spatial extent that experimentation 
and correlation may prove difficult or yield equivocal results.
    Experimental techniques are frequently used for evaluating 
causality in complex chemical mixtures. Options include evaluating 
separated components of the mixture, developing and testing a synthetic 
mixture, or determining how the toxicity of a mixture relates to the 
toxicity of individual components. The choice of method depends on the 
goal of the assessment and the resources and test data that are 
available.
    Laboratory toxicity identification evaluations (TIEs) can be used 
to help determine which components of a chemical mixture are causing 
toxic effects. By using fractionation and other methods, the TIE 
approach can help identify chemicals responsible for toxicity and show 
the relative contributions to toxicity of different chemicals in 
aqueous effluents (U.S. EPA, 1988a, 1989b, c) and sediments (e.g., 
Ankley et al., 1990).
    Risk assessors may utilize data from synthetic chemical mixtures if 
the individual chemical components are well characterized. This 
approach

[[Page 47587]]

allows for manipulation of the mixture and investigation of how varying 
the components that are present or their ratios may affect mixture 
toxicity but also requires additional assumptions about the 
relationship between effects of the synthetic mixture and those of the 
environmental mixture.
    When the modes of action of chemicals in a mixture are known to be 
similar, an additive model has been successful in predicting combined 
effects (Konemann, 1981; Hermens et al., 1984a; McCarty and Mackay, 
1993; Sawyer and Safe, 1985; Broderius et al., 1995). In this 
situation, the contribution of each chemical to the overall toxicity of 
the mixture can be evaluated. However, the situation is more 
complicated when the modes of action of the chemical constituents are 
unknown or partially known (see additional discussion in section 
5.1.2).
4.3.1.3. Linking Measures of Effect to Assessment Endpoints
    Assessment endpoints express the environmental values of concern 
for a risk assessment, but they cannot always be measured directly. 
When measures of effect differ from assessment endpoints, sound and 
explicit linkages between the two are needed. Risk assessors may make 
these linkages in the analysis phase or, especially when linkages rely 
on expert judgment, risk assessors may work with measures of effect 
through risk estimation (in risk characterization) and then make the 
connection with the assessment endpoints. Common extrapolations used to 
link measures of effect with assessment endpoints are shown in text 
note 4-18.
    General Considerations. During the preparation of the analysis plan 
in problem formulation, risk assessors identify the extrapolations 
required between assessment endpoints and measures of effect. During 
the analysis phase, risk assessors should revisit the questions listed 
in text note 4-19 before proceeding with specific extrapolation 
approaches to use.
    The scope and nature of the risk assessment and the environmental 
decision to be made help determine the degree of uncertainty (and type 
of extrapolation) that is acceptable. At an early stage of a tiered 
risk assessment, extrapolations from minimal data that involve large 
uncertainties are acceptable when the primary purpose is to determine 
whether a risk exists given worst-case exposure and effects scenarios. 
To define risk further at later stages of the assessment, additional 
data and more sophisticated extrapolation approaches are usually 
required.
    The scope of the risk assessment also influences extrapolation 
through the nature of the assessment endpoint. Preliminary assessments 
that evaluate risks to general trophic levels, such as fish and birds, 
may extrapolate among different genera or families to obtain a range of 
sensitivity to the stressor. On the other hand, assessments concerned 
with management strategies for a particular species may employ 
population models.
    Analysis phase activities may suggest additional extrapolation 
needs. Evaluation of exposure may indicate different spatial or 
temporal scales than originally anticipated. If spatial scales are 
broadened, additional receptors may need to be included in 
extrapolation models. If a stressor persists for an extended time in 
the environment, it may be necessary to extrapolate short-term 
responses over a longer period of exposure, and population level 
effects may become more important.
    Whatever methods are employed to link assessment endpoints with 
measures of effect, it is important to apply the methods in a manner 
consistent with sound ecological principles and the availability of an 
appropriate database. For example, it is inappropriate to use 
structure-activity relationships to predict toxicity from chemical 
structure unless the chemical under consideration has a similar mode of 
toxic action to the reference chemicals (Bradbury, 1994). Similarly, 
extrapolations from upland avian species to waterfowl may be more 
credible if factors such as differences in food preferences, body mass, 
physiology, and seasonal behavior (e.g., mating and migration habits) 
are considered. Extrapolations made in a rote manner or that are 
biologically implausible will erode the overall credibility of the 
assessment.
    Finally, many extrapolation methods are limited by the availability 
of suitable databases. Although these databases are generally largest 
for chemical stressors and aquatic species, data do not exist for all 
taxa or effects. Chemical effects databases for mammals, amphibians, or 
reptiles are extremely limited, and there is even less information on 
most biological and physical stressors. Risk assessors should be aware 
that extrapolations and models are only as useful as the data on which 
they are based and should recognize the great uncertainties associated 
with extrapolations that lack an adequate empirical or process-based 
rationale.
    The rest of this section addresses the approaches used by risk 
assessors to link measures of effect to assessment endpoints, as noted 
below.
     Linkages based on expert judgment. This approach is not as 
desirable as empirical or process-based approaches, but is the only 
option when data are lacking.
     Linkages based on empirical or process models. Empirical 
extrapolations use experimental or observational data that may or may 
not be organized into a database. Process-based approaches are based on 
some level of understanding of the underlying operations of the system 
under consideration.
    Judgment Approaches for Linking Measures of Effect to Assessment 
Endpoints. Expert judgment approaches rely on the professional 
expertise of risk assessors, expert panels, or others to relate changes 
in measures of effect to changes in the assessment endpoint. They are 
essential when databases are inadequate to support empirical models and 
process models are unavailable or inappropriate. Expert judgment 
linkages between measures of effect and assessment endpoints can be 
just as credible as empirical or process-based expressions, provided 
they have a sound scientific basis. This section highlights expert 
judgment extrapolations between species, from laboratory data to field 
effects, and between geographic areas.
    Because of the uncertainties in predicting the effects of 
biological stressors such as introduced species, expert judgment 
approaches are commonly used. For example, there may be measures of 
effect data on a foreign pathogen that attacks a certain tree species 
not found in the United States, but the assessment endpoint concerns 
the survival of a commercially important tree found only in the United 
States. In this case, a careful evaluation and comparison of the life 
history and environmental requirements of both the pathogen and the two 
tree species may contribute toward a useful determination of potential 
effects, even though the uncertainty may be high. Expert panels are 
typically used for this kind of evaluation (USDA, 1993).
    Risks to organisms in field situations are best estimated from 
studies at the site of interest. However, such data are not always 
available. Frequently, risk assessors must extrapolate from laboratory 
toxicity test data to field effects. Text note 4-20 summarizes some of 
the considerations for risk assessors when extrapolating from 
laboratory toxicity test results to field situations for chemical 
stressors. Factors altering exposure in the field are among the most 
important factors limiting extrapolations from laboratory test results, 
but indirect effects on exposed

[[Page 47588]]

organisms due to predation, competition, or other biotic or abiotic 
factors not evaluated in the laboratory may also be significant. 
Variations in direct chemical effects between laboratory tests and 
field situations may not contribute as much to the overall uncertainty 
of the extrapolation.
    In addition to single-species tests, laboratory multiple species 
tests are sometimes used to predict field effects. While these tests 
have the advantage of evaluating some aspects of a real ecological 
system, they also have inherent scale limitations (e.g., lack of top 
trophic levels) and may not adequately represent features of the field 
system important to the assessment endpoint.
    Extrapolations based on expert judgment are frequently required 
when assessors wish to use field data obtained from one geographic area 
and apply them to a different area of concern, or to extrapolate from 
the results of laboratory tests to more than one geographic region. In 
either case, risk assessors should consider variations between regions 
in environmental conditions, spatial scales and heterogeneities, and 
ecological forcing functions (see below).
    Variations in environmental conditions in different geographic 
regions may alter stressor exposure and effects. If exposure to 
chemical stressors can be accurately estimated and are expected to be 
similar (e.g., see text note 4-20), the same species in different areas 
may respond similarly. For example, if the pesticide granular 
carbofuran were applied at comparable rates throughout the country, 
seed-eating birds could be expected to be similarly affected by the 
pesticide (Houseknecht, 1993). Nevertheless, the influence of 
environmental conditions on stressor exposure and effects can be 
substantial.
    For biological stressors, environmental conditions such as climate, 
habitat, and suitable hosts play major roles in determining whether a 
biological stressor becomes established. For example, climate would 
prevent establishment of the Mediterranean fruit fly in the much colder 
northeastern United States. Thus, a thorough evaluation of 
environmental conditions in the area versus the natural habitat of the 
stressor is important. Even so, many biological stressors can adapt 
readily to varying environmental conditions, and the absence of natural 
predators or diseases may play an even more important role than abiotic 
environmental conditions.
    For physical stressors that have natural counterparts, such as 
fire, flooding, or temperature variations, effects may depend on the 
natural variations in these parameters for a particular region. Thus, 
the comparability of two regions depends on both the pattern and range 
of natural disturbances.
    Spatial scales and heterogeneities affect comparability between 
regions. Effects observed over a large scale may be difficult to 
extrapolate from one geographical location to another mainly because 
the spatial heterogeneity is likely to differ. Factors such as number 
and size of land-cover patches, distance between patches, connectivity 
and conductivity of patches (e.g., migration routes), and patch shape 
may be important. Extrapolations can be facilitated by using 
appropriate reference sites, such as sites in comparable ecoregions 
(Hughes, 1995).
    Ecological forcing functions may differ between geographic regions. 
Forcing functions are critical abiotic variables that exert a major 
influence on the structure and function of ecological systems. Examples 
include temperature fluctuations, fire frequency, light intensity, and 
hydrologic regime. If these differ significantly between sites, it may 
be inappropriate to extrapolate stressor effects from one system to 
another.
    The following references may be useful when assessing effects over 
different geographical areas: Bedford and Preston (1988), Detenbeck et 
al. (1992), Gibbs (1993), Gilbert (1987), Gosselink et al. (1990), 
Preston and Bedford (1988), and Risser (1988).
    Empirical and Process-Based Approaches for Linking Measures of 
Effect to Assessment Endpoints. There are a variety of empirical and 
process-based approaches available to risk assessors depending on the 
scope of the assessment and the data and resources available. Empirical 
and process-based approaches include numerical extrapolations between 
effects measures and assessment endpoints. These linkages range in 
sophistication from applying an uncertainty factor to using a complex 
model requiring extensive measures of effects and measures of ecosystem 
and receptor characteristics as input. But even the most sophisticated 
quantitative models involve qualitative elements and assumptions and 
thus require professional judgment for evaluation. Individuals who use 
models and interpret their results should be familiar with the 
underlying assumptions and components contained in the model.
    Empirical Approaches. Empirically based uncertainty factors or 
taxonomic extrapolations may be used when adequate effects databases 
are available but the understanding of underlying mechanisms of action 
or ecological principles is limited. When sufficient information on 
stressors and receptors is available, process-based approaches such as 
pharmacokinetic/pharmacodynamic models or population or ecosystem 
process models may be used. Regardless of the options used, risk 
assessors should justify and adequately document the approach selected.
    Uncertainty factors are used to ensure that effects measures are 
sufficiently protective of assessment endpoints. Uncertainty factors 
are empirically derived numbers that are divided into measure of 
effects values to give an estimated stressor level that should not 
cause adverse effects to the assessment endpoint. Uncertainty factors 
have mostly been developed for chemicals because of the extensive 
ecotoxicologic databases available, especially for aquatic organisms. 
Uncertainty factors are useful when decisions must be made about 
stressors in a short time and with little information.
    Uncertainty factors have been used to compensate for assessment 
endpoint/effect measures differences between endpoints (acute to 
chronic effects), between species, and between test situations (e.g., 
laboratory to field). Typically, uncertainty factors vary inversely 
with the quantity and type of effects measures data available (Zeeman, 
1995). Uncertainty factors have been used in screening-level 
assessments of new chemicals (Nabholz, 1991), in assessing the risks of 
pesticides to aquatic and terrestrial organisms (Urban and Cook, 1986), 
and in developing benchmark dose levels for human health effects (U.S. 
EPA, 1995d).
    In spite of their usefulness, uncertainty factors can also be 
misused, especially when used in an overly conservative fashion, as 
when chains of factors are multiplied together without sufficient 
justification. Like other approaches to bridging data gaps, uncertainty 
factors are often based on a combination of scientific analysis, 
scientific judgement and policy judgement (see section 4.1.3). It is 
important to differentiate among these three elements when documenting 
the basis for the uncertainty factors used.
    Empirical data can be used to facilitate extrapolations between 
species to species, genera, families, or orders or functional groups 
(e.g., feeding guilds) (Suter, 1993a). Suter et al. (1983), Suter 
(1993a), and Barnthouse et al. (1987, 1990) developed methods to 
extrapolate toxicity among freshwater and marine fish and arthropods. 
As noted by Suter

[[Page 47589]]

(1993a), the uncertainties associated with extrapolating between 
orders, classes, and phyla tend to be very high. However, 
extrapolations can be made with fair certainty between aquatic species 
within genera and genera within families. Further applications of this 
approach (e.g., for chemical stressors and terrestrial organisms) are 
limited by a lack of suitable databases.
    Dose-scaling or allometric regression has also been used to 
extrapolate the effects of a chemical stressor to another species. The 
method is used for human health risk assessment but has not been 
applied extensively to ecological effects (Suter, 1993a).
    Allometric regression has been used with avian species (Kenaga, 
1973) and to a limited extent for estimating effects to marine 
organisms based on their length. For chemical stressors, allometric 
relationships can enable an assessor to estimate toxic effects to 
species not commonly tested, such as native mammalian species. It is 
important that the assessor consider the taxonomic relationship between 
the known species and the species of interest. The closer the two are 
related, the more likely that the toxic response will be similar. 
Allometric approaches should not be applied to species that differ 
greatly in uptake, metabolism, or depuration of a chemical.
    Process-Based Approaches. Process models for extrapolation are 
representations or abstractions of a system or process (Starfield and 
Bleloch, 1991) that incorporate causal relationships and provide a 
predictive capability that does not depend on the availability of 
existing stressor-response information as empirical models do (Wiegert 
and Bartell, 1994). Process models enable assessors to translate data 
on individual effects (e.g., mortality, growth, and reproduction) to 
potential alterations in specific populations, communities, or 
ecosystems. Such models can be used to evaluate risk hypotheses about 
the duration and severity of a stressor on an assessment endpoint that 
cannot be tested readily in the laboratory.
    There are two major types of models: single-species population 
models and multispecies community and ecosystem models. Population 
models describe the dynamics of a finite group of individuals through 
time and have been used extensively in ecology and fisheries management 
and to assess the impacts of power plants and toxicants on specific 
fish populations (Barnthouse et al., 1987; Barnthouse et al., 1990). 
Population models are useful in answering questions related to short- 
or long-term changes of population size and structure and can be used 
to estimate the probability that a population will decline below or 
grow above a specified abundance (Ginzburg et al., 1982; Ferson et al., 
1989). This latter application may be useful when assessing risks 
associated with biological stressors such as introduced or pest 
species. Excellent reviews of population models are presented by 
Barnthouse et al. (1986) and Wiegert and Bartell (1994). Emlen (1989) 
has reviewed population models that can be used for terrestrial risk 
assessment.
    Proper use of the population models requires a thorough 
understanding of the natural history of the species under 
consideration, as well as knowledge of how the stressor influences its 
biology. Model input can include somatic growth rates, physiological 
rates, fecundity, survival rates of various classes within the 
population, and how these change when the population is exposed to the 
stressor and other environmental factors. In addition, the effects of 
population density on these parameters may be important (Hassell, 1986) 
and should be considered in the analysis of uncertainty.
    Community and ecosystem models (e.g., Bartell et al., 1992; O'Neill 
et al., 1982) are particularly useful when the assessment endpoint 
involves structural (e.g., community composition) or functional (e.g., 
primary production) elements of the system potentially at risk. These 
models can also be useful when secondary effects are of concern. 
Changes in various community or ecosystem components such as 
populations, functional types, feeding guilds, or environmental 
processes can be estimated. By incorporating submodels describing the 
dynamics of individual system components, these models permit 
evaluation of risk to multiple assessment endpoints within the context 
of the larger environmental system.
    Risk assessors should evaluate the degree of aggregation in 
population or multispecies model parameters that is appropriate based 
both on the input data available and on the desired output of the 
model. For example, if a decision is required about a particular 
species, a model that lumps species into trophic levels or feeding 
guilds will not be very useful. Assumptions concerning aggregation in 
model parameters should be included in the discussion of uncertainty.
4.3.2. Stressor-Response Profile
    The final product of ecological response analysis is a summary 
profile of what has been learned. Depending on the risk assessment, the 
profile may be a written document, or a module of a larger process 
model. Alternatively, documentation may be deferred until risk 
characterization. In any case, the objective is to ensure that the 
information needed for risk characterization has been collected and 
evaluated. A useful approach in preparing the stressor-response profile 
is to imagine that it will be used by someone else to perform the risk 
characterization. Using this approach, the assessor may be better able 
to extract the information most important to the risk characterization 
phase. In addition, compiling the stressor-response profile provides an 
opportunity to verify that the assessment and measures of effect 
identified in the conceptual model were evaluated.
    Risk assessors should address several questions in the stressor-
response profile (text note 4-21). Depending on the type of risk 
assessment, affected ecological entities could include single species, 
populations, general trophic levels, communities, ecosystems, or 
landscapes. The nature of the effect(s) should be germane to the 
assessment endpoint(s). Thus if a single species is affected, the 
effects should represent parameters appropriate for that level of 
organization. Examples include effects on mortality, growth, and 
reproduction. Short- and long-term effects should be reported as 
appropriate. At the community level, effects could be summarized in 
terms of structure or function depending on the assessment endpoint. At 
the landscape level, there may be a suite of assessment endpoints and 
each should be addressed separately.
    Examples of different approaches for displaying the intensity of 
effects as stressor-response curves or point estimates were provided in 
section 4.3.1.1. Other information such as the spatial area or time to 
recovery may be appropriate, depending on the scope of the assessment. 
Causal analyses are important, especially for assessments that include 
field observational data.
    While ideally the stressor-response profile should express effects 
in terms of the assessment endpoint, this will not always be possible. 
Especially where it is necessary to use qualitative extrapolations 
between assessment endpoints and measures of effect, the stressor-
response profile may only contain information on measures of effect. 
Under these circumstances, risk will be estimated using the measures of 
effects, and extrapolation to the assessment endpoints will occur 
during risk characterization.
    Risk assessors need to be descriptive and candid about any 
uncertainties

[[Page 47590]]

associated with the ecological response analysis. If it was necessary 
to extrapolate from measures of effect to the assessment endpoint, 
describe both the extrapolation and its basis. Similarly, if a 
benchmark or similar reference dose or concentration was calculated, 
discuss the extrapolations and uncertainties associated with its 
development. For additional information on establishing reference 
concentrations, see Nabholz (1991), Urban and Cook (1986), Stephan et 
al. (1985), Van Leeuwen et al. (1992), Wagner and Lkke (1991), 
and Okkerman et al. (1993). Finally, the assessor should clearly 
indicate major assumptions and default values used in models.
    At the end of the analysis phase, the stressor-response and 
exposure profiles are used to estimate risks. These profiles provide 
the opportunity to review what has been learned and to summarize this 
information in the most useful format for risk characterization. 
Whatever form the profiles take, they ensure that the necessary 
information is available for risk characterization.

5. Risk Characterization

    Risk characterization (figure 5-1) is the final phase of ecological 
risk assessment. Its goals are to use the results of the analysis phase 
to estimate risk to the assessment endpoints identified in problem 
formulation (section 5.1), interpret the risk estimate (section 5.2), 
and report the results (section 5.3).

BILLING CODE 6560-50-P

[[Page 47591]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.006



BILLING CODE 6560-50-C

[[Page 47592]]

    Risk characterization is a major element of the risk assessment 
report. To be successful, it should provide clear information to the 
risk manager to use in environmental decision making (NRC, 1994; see 
section 6). If the risks are not sufficiently defined to support a 
management decision, the risk manager may elect to proceed with another 
iteration of the risk assessment process. Additional research or a 
monitoring program may improve the risk estimate or help to evaluate 
the consequences of a risk management decision.

5.1. Risk Estimation

    Risk estimation determines the likelihood of adverse effects to 
assessment endpoints by integrating exposure and effects data and 
evaluating any associated uncertainties. The process uses exposure and 
stressor-response profiles which are developed according to the 
analysis plan (section 3.5). Risks can be estimated by one or more of 
the following approaches: (1) estimates expressed as qualitative 
categories, (2) estimates comparing single-point estimates of exposure 
and effects, (3) estimates incorporating the entire stressor-response 
relationship, (4) estimates incorporating variability in exposure and 
effects estimates, (5) estimates based on process models that rely 
partially or entirely on theoretical approximations of exposure and 
effects, and (6) estimates based on empirical approaches, including 
field observational data.
5.1.1. Risk Estimates Expressed as Qualitative Categories
    In some cases, best professional judgment may be used to express 
risks qualitatively using categories such as low, medium, and high or 
yes and no. This approach is most frequently used when exposure and 
effects data are limited or not easily expressed in quantitative terms. 
A U.S. Forest Service assessment used qualitative categories because of 
limitations on both the exposure and effects data for the introduced 
species of concern as well as the resources available for the 
assessment. (text note 5-1)
5.1.2. Single-Point Estimates
    When sufficient data are available to quantify exposure and effects 
estimates, the simplest approach for comparing the estimates is to use 
a ratio of two numbers (figure 5-2a). Typically, the ratio (or 
quotient) is expressed as an exposure concentration divided by an 
effects concentration. Quotients are commonly used for chemical 
stressors, where reference or benchmark toxicity values are widely 
available (text note
5-2).

BILLING CODE 6560-50-P

[[Page 47593]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.007



BILLING CODE 6560-50-C

[[Page 47594]]

    The principal advantages of the quotient method are that it is 
simple and quick to use and risk assessors and managers are familiar 
with its application. The quotient method provides an efficient, 
inexpensive means of identifying high or low risk situations that can 
allow risk management decisions to be made without the need for further 
information.
    Quotients have also been used to integrate the risks of multiple 
chemical stressors. In this approach, quotients for the individual 
constituents in a mixture are generated by dividing each exposure level 
by a corresponding toxicity endpoint (e.g., an LC50). Although the 
toxicity of a chemical mixture may be greater (synergism) or less 
(antagonism) than predicted from the toxicities of individual 
constituents of the mixture, a quotient addition approach assumes that 
toxicities are additive or close to additive, which may be true when 
the modes of action of chemicals in a mixture are similar (e.g., 
Konemann, 1981; Broderius et al., 1995; Hermens et al., 1984a,b; 
McCarty and Mackay, 1993; Sawyer and Safe, 1985).
    For mixtures of chemicals having dissimilar modes of action, there 
is some evidence from fish acute toxicity tests with industrial organic 
chemicals that strict additivity or less-than-strict additivity is 
common, while antagonistic and synergistic responses are rare 
(Broderius, 1991). These experiences suggest that caution should be 
used when predicting that chemicals in a mixture will act independently 
of one another. However, these relationships observed with aquatic 
organisms may not be relevant for other endpoints, exposure scenarios, 
and species. When the mode of action for constituent chemicals are 
unknown, the assumptions and rationale concerning chemical interactions 
must be clearly stated.
    The application of the quotient method is restricted by a number of 
limitations (see Smith and Cairns, 1993; Suter, 1993a). While a 
quotient can be useful in answering whether risks are high or low, it 
may not be helpful to a risk manager who needs to make a decision 
requiring a quantification of risks. For example, it is seldom useful 
to say that a risk mitigation approach will reduce a quotient value 
from 25 to 12, since this reduction cannot by itself be clearly 
interpreted in terms of effects on an assessment endpoint.
    Another potential difficulty with the quotient method is that the 
point estimate of effect may not reflect the appropriate intensity of 
effect or exposure pattern for the assessment. For example, an 
LC50 derived from a 96-hour laboratory test using constant 
exposure levels may not be appropriate for an assessment of effects on 
reproduction resulting from short-term, pulsed exposures.
    The quotient method cannot evaluate secondary effects. Interactions 
and effects beyond what is predicted from the simple quotient may be 
critical to characterizing the full extent of impacts from exposure to 
the stressors (e.g., bioaccumulation).
    Finally, in most cases, the quotient method does not explicitly 
consider uncertainty (e.g., extrapolation from tested species to the 
species or community of concern). However, some uncertainties can be 
incorporated into single-point estimates to provide a statement of 
likelihood that the effects point estimate exceeds the exposure point 
estimate (figures 5-2b and 5-3). If exposure variability is quantified, 
then the point estimate of effects can be compared with a cumulative 
exposure distribution as described in text note 5-3. Further discussion 
of comparisons between point estimates of effects and distributions of 
exposure may be found in Suter et al., 1983.

BILLING CODE 6560-50-P

[[Page 47595]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.008



BILLING CODE 6560-50-P

[[Page 47596]]

    In view of the advantages and limitations of the quotient method, 
it is important for risk assessors to consider the points listed below 
when evaluating quotient method estimates.
     How does the effect concentration relate to the assessment 
endpoint?
     What extrapolations are involved?
     How does the point estimate of exposure relate to 
potential spatial and temporal variability in exposure?
     Are data sufficient to provide confidence intervals on the 
endpoints?
5.1.3. Estimates Incorporating the Entire Stressor-Response 
Relationship
    If the stressor-response profile described a curve relating the 
stressor level to the magnitude of response, then risk estimation can 
examine risks associated with many different levels of exposure (figure 
5-4). These estimates are particularly useful when the risk assessment 
outcome is not based on exceedance of a predetermined decision rule 
such as a toxicity benchmark level.

BILLING CODE 6560-50-P

[[Page 47597]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.009



BILLING CODE 6560-50-C

[[Page 47598]]

    There are both advantages and limitations to comparing a stressor-
response curve with an exposure distribution. The steepness of the 
effects curve shows the magnitude of change in effects associated with 
incremental changes in exposure, and the capability to predict changes 
in the magnitude and likelihood of effects for different exposure 
scenarios can be used to compare different risk management options. 
Also, uncertainty can be incorporated by calculating uncertainty bounds 
on the stressor-response or exposure estimates. While comparing 
exposure and stressor-response curves provides a predictive ability 
lacking in the quotient method, this approach shares the quotient 
method's limitations of not evaluating secondary effects, assuming that 
the exposure pattern used to derive the stressor-response curve is 
comparable to the environmental exposure pattern, and not explicitly 
considering uncertainties, such as extrapolations from tested species 
to the species or community of concern.
5.1.4. Estimates Incorporating Variability in Exposure or Effects
    If the exposure or stressor-response profiles describe the 
variability in exposure or effects, then many different risk estimates 
can be calculated. Variability in exposure can be used to describe 
risks to moderately or highly exposed members of a population being 
investigated, while variability in effects can be used to describe 
risks to average or sensitive population members.
    A major advantage of this approach is the capability to predict 
changes in the magnitude and likelihood of effects for different 
exposure scenarios, thus providing a means for comparing different risk 
management options. As noted above, comparing distributions also allows 
one to identify and quantify risks to different segments of the 
population. Limitations include the increased data requirements 
compared with previously described techniques and the implicit 
assumption that the full range of variability in the exposure and 
effects data is adequately represented. As with the quotient method, 
secondary effects are not readily evaluated with this technique. Thus, 
it is desirable to corroborate risks estimated by distributional 
comparisons with field studies or other lines of evidence. Text note 5-
4 and figure 5-5 illustrate the use of cumulative exposure and effects 
distributions for estimating risk.

BILLING CODE 6560-50-P

[[Page 47599]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.010



BILLING CODE 6560-50-C

[[Page 47600]]

5.1.5. Estimates Based on Process Models
    Process models are mathematical expressions that represent our 
understanding of the mechanistic operation of a system under 
evaluation. They can be useful tools both in the analysis phase (see 
section 4.1.2.) and the risk characterization phase of ecological risk 
assessment. For illustrative purposes, we distinguish between process 
models used for risk estimation that integrate exposure and effects 
information (text note 5-5) and process models used in the analysis 
phase that focus on either exposure or effects evaluations.
    A major advantage of using process models for risk estimation is 
the ability to consider ``what if'' scenarios and to forecast beyond 
the limits of observed data that constrain risk estimation techniques 
based on empirical data. The process model can also consider secondary 
effects, unlike other risk estimation techniques such as the quotient 
method or comparisons of exposure and effect distributions. In 
addition, some process models may be capable of forecasting the 
combined effects of multiple stressors (e.g., Barnthouse et al., 1990).
    Process model outputs may be point estimates or distributions. In 
either case, risk assessors should interpret these outputs with care. 
Process model outputs may imply a higher level of certainty than is 
appropriate and all too often are viewed without sufficient attention 
to underlying assumptions. The lack of knowledge on basic life 
histories for many species and incomplete knowledge on the structure 
and function of a particular ecosystem is often lost in the model 
output. Since process models are only as good as the assumptions on 
which they are based, they should be treated as hypothetical 
representations of reality until appropriately tested with empirical 
data. Comparing model results to field data provides a check on whether 
our understanding of the system was correct (Johnson, 1995) with 
respect to the risk hypotheses presented in problem formulation.
5.1.6. Field Observational Studies
    Field observational studies (surveys) can serve as risk estimation 
techniques because they provide direct evidence linking exposure to 
stressors and effects. Field surveys measure biological changes in 
uncontrolled situations through collection of exposure and effects data 
at sites identified in problem formulation. A key issue with field 
surveys is establishing causal relationships between stressors and 
effects (section 4.3.1.2).
    A major advantage of field surveys is that they provide a reality 
check on other risk estimates, since field surveys are usually more 
representative of both exposures and effects (including secondary 
effects) found in natural systems than are estimates generated from 
laboratory studies or theoretical models (text note 5-6). On the other 
hand, field data may not constitute reality if they are flawed due to 
poor experimental design, biased in sampling or analytical techniques, 
or fail to measure critical components of the system or random 
variations (Johnson, 1995). A lack of observed effects in a field 
survey may occur because the measurements are insufficiently sensitive 
to detect ecological effects, and, unless causal relationships are 
carefully examined, effects that are observed may be caused by factors 
unrelated to the stressor(s) of concern. Finally, field surveys taken 
at one point in time are usually not predictive; they describe effects 
associated with only one scenario (i.e., the one that exists).

5.2. Risk Description

    After risks have been estimated, risk assessors need to integrate 
and interpret the available information into conclusions about risks to 
the assessment endpoints. In some cases, risk assessors may have 
quantified the relationship between assessment endpoints and measures 
of effect in the analysis stage (section 4.3.1.3). In other situations, 
qualitative links to assessment endpoints are part of the risk 
description. For example, if the assessment endpoints are survival of 
fish, aquatic invertebrates, and algae, risks may be estimated using a 
quotient method based on LC50c. Regardless of the risk estimation 
technique, the technical narrative supporting the estimates is as 
important as the risk estimates themselves.
    Risk descriptions include an evaluation of the lines of evidence 
supporting or refuting the risk estimate(s) and an interpretation of 
the adverse effects on the assessment endpoint.
5.2.1. Lines of Evidence
    Confidence in the conclusions of a risk assessment may be increased 
by using several lines of evidence to interpret and compare risk 
estimates. These lines of evidence may be derived from different 
sources or by different techniques relevant to adverse effects on the 
assessment endpoints, such as quotient estimates, modeling results, 
field experiments, or field observations. (Note that the term ``weight 
of evidence'' is sometimes used in legal discussions or in other 
documents, e.g., Urban and Cook, 1986; Menzie et al., 1996. We use the 
phrase lines of evidence to emphasize that both qualitative evaluation 
and quantitative weightings may be used.)
    Some of the factors that the risk assessor should consider when 
evaluating separate lines of evidence are:
 The relevance of evidence to the assessment endpoints
 The relevance of evidence to the conceptual model
 The sufficiency and quality of data and experimental designs 
used in key studies
 The strength of cause/effect relationships
 The relative uncertainties of each line of evidence and their 
direction.

This process involves more than just listing the factors that support 
or refute the risk. The risk assessor should carefully examine each 
factor and evaluate its contribution to the risk assessment.
    For example, consider the two lines of evidence described for the 
carbofuran example (text notes 5-2 and 5-6): quotients and field 
studies. Both approaches are relevant to the assessment endpoint 
(survival of birds that forage in agricultural areas where carbofuran 
is applied), and both are relevant to the exposure scenarios described 
in the conceptual model (figure 3-2). However, the quotients are 
limited in their ability to express incremental risks (e.g., how much 
greater risk is expressed by a quotient of ``2'' versus a quotient of 
``4''), while the field studies had some design flaws (text note 5-6). 
Nevertheless, because of the great preponderance of the data, the 
strong evidence of causal relationships from the field studies, and the 
consistency between these two lines of evidence, confidence in a 
conclusion of high risk to the assessment endpoint is supported.
    Sometimes lines of evidence do not point toward the same 
conclusion. When they disagree, it is important to distinguish between 
true inconsistencies and those related to differences in statistical 
powers of detection. For example, a model may predict adverse effects 
that were not observed in a field survey. The risk assessor should ask 
whether the experimental design of the field study had sufficient power 
to detect the predicted difference or whether the endpoints measured 
were comparable with those used in the model.

[[Page 47601]]

Conversely, the model may have been unrealistic in its predictions. 
While it may be possible to use numerical weighting techniques for 
evaluating various lines of evidence, in most cases qualitative 
evaluations based on professional judgment are appropriate for sorting 
through conflicting lines of evidence. While iteration of the risk 
assessment process and collection of additional data may help resolve 
uncertainties, this option is not always available.
5.2.2. Determining Ecological Adversity
    At this point in risk characterization, the changes expected in the 
assessment endpoints have been estimated and described. The next step 
is to interpret whether these changes are considered adverse. Adverse 
changes are those of concern ecologically or socially (section 1). 
Determining adversity is not always an easy task and frequently depends 
on the best professional judgment of the risk assessor.
    Five criteria are proposed for evaluating adverse changes in 
assessment endpoints:
     Nature of effects.
     Intensity of effects.
     Spatial scale.
     Temporal scale.
     Potential for recovery.
    The extent to which the five criteria are evaluated depends on the 
scope and complexity of the ecological risk assessment. However, 
understanding the underlying assumptions and science policy judgments 
is important even in simple cases. For example, when exceedance of a 
previously established decision rule such as a benchmark stressor level 
is used as evidence of adversity (e.g., see Urban and Cook, 1986, or 
Nabholz, 1991), the reasons why exceedences of the benchmark are 
considered adverse should be clearly understood.
    To distinguish ecological changes that are adverse from those 
ecological events that are within the normal pattern of ecosystem 
variability or result in little or no significant alteration of biota, 
it is important to consider the nature and intensity of effects. For 
example, for an assessment endpoint involving survival, growth, and 
reproduction of a species, do predicted effects involve survival and 
reproduction or only growth? If survival of offspring will be affected, 
by what percentage will it diminish?
    It is important for risk assessors to consider both the ecological 
and statistical contexts of an effect when evaluating intensity. For 
example, a statistically significant 1% decrease in fish growth (text 
note 5-7) may not be relevant to an assessment endpoint of fish 
population viability, and a 10% decline in reproduction may be worse 
for a population of slowly reproducing trees than for rapidly 
reproducing planktonic algae.
    Natural ecosystem variation can make it very difficult to observe 
(detect) stressor-related perturbations. For example, natural 
fluctuations in marine fish populations are often large, with intra- 
and interannual variability in population levels covering several 
orders of magnitude. Furthermore, cyclic events (e.g., bird migration, 
tides) are very important in natural systems. Predicting the effects of 
anthropogenic stressors against this background of variation can be 
very difficult. Thus, a lack of statistically significant effects in a 
field study does not automatically mean that adverse ecological effects 
are absent. Rather, risk assessors must consider factors such as 
statistical power to detect differences, natural variability, and other 
lines of evidence in reaching their conclusions.
    Spatial and temporal scales need to be considered in assessing the 
adversity of the effects. The spatial dimension encompasses both the 
extent and pattern of effect as well as the context of the effect 
within the landscape. Factors to consider include the absolute area 
affected, the extent of critical habitats affected compared with a 
larger area of interest, and the role or use of the affected area 
within the landscape.
    Adverse effects to assessment endpoints vary with the absolute area 
of the effect. A larger affected area may be (1) subject to a greater 
number of other stressors, increasing the complications from stressor 
interactions; (2) more likely to contain sensitive species or habitats; 
or (3) more susceptible to landscape-level changes because many 
ecosystems may be altered by the stressors.
    Nevertheless, a smaller area of effect is not always associated 
with lower risk. The function of an area within the landscape may be 
more important than the absolute area. Destruction of small but unique 
areas, such as critical wetlands, may have important effects on local 
wildlife populations. Also, in river systems, both riffle and pool 
areas provide important microhabitats that maintain the structure and 
function of the total river ecosystem. Stressors acting on some of 
these microhabitats may present a significant risk to the entire 
system.
    Spatial factors are important for many species because of the 
linkages between ecological landscapes and population dynamics. 
Linkages between one or more landscapes can provide refugia for 
affected populations, and species may require adequate corridors 
between habitat patches for successful migration.
    The temporal scale for ecosystems can vary from seconds 
(photosynthesis, prokaryotic reproduction) to centuries (global climate 
change). Changes within a forest ecosystem can occur gradually over 
decades or centuries and may be affected by slowly changing external 
factors such as climate. When interpreting ecological adversity, risk 
assessors should recognize that the time scale of stressor-induced 
changes operates within the context of multiple natural time scales. In 
addition, temporal responses for ecosystems may involve intrinsic time 
lags, so that responses from a stressor may be delayed. Thus, it is 
important to distinguish the long-term impacts of a stressor from the 
immediately visible effects. For example, visible changes resulting 
from eutrophication of aquatic systems (turbidity, excessive macrophyte 
growth, population decline) may not become evident for many years after 
initial increases in nutrient levels.
    Considering the temporal scale of adverse effects leads logically 
to a consideration of recovery. Recovery is the rate and extent of 
return of a population or community to a condition that existed before 
the introduction of a stressor. (While this discussion deals with 
recovery as a result of natural processes, risk mitigation options may 
include restoration activities to facilitate or speed up the recovery 
process.) Because ecosystems are dynamic and even under natural 
conditions are constantly changing in response to changes in the 
physical environment (weather, natural catastrophes, etc.) or other 
factors, it is unrealistic to expect that a system will remain static 
at some level or return to exactly the same state that it was before it 
was disturbed (Landis et al., 1993). Thus, the attributes of a 
``recovered'' system must be carefully defined. Examples might include 
productivity declines in an eutrophic system, reestablishment of a 
species at a particular density, species recolonization of a damaged 
habitat, or the restoration of health of diseased organisms.
    Recovery can be evaluated in spite of the difficulty in predicting 
events in ecological systems (e.g., Niemi et al., 1990). For example, 
it is possible to distinguish changes that are usually reversible 
(e.g., recovery of a stream from sewage effluent discharge), frequently 
irreversible (e.g., establishment of introduced species), and always 
irreversible (e.g., species extinction). It is important for risk 
assessors to consider whether significant structural or functional

[[Page 47602]]

changes have occurred in a system that might render changes 
irreversible. For example, physical alterations such as deforestation 
in the coastal hills of Venezuela in recent history and Britain in the 
Neolithic period changed soil structure and seed sources such that 
forests cannot easily grow again (Fisher and Woodmansee, 1994).
    Risk assessors should note natural disturbance patterns when 
evaluating the likelihood of recovery from anthropogenic stressors. 
Ecosystems that have been subjected to repeated natural disturbances 
may be more vulnerable to anthropogenic stressors (e.g., overfishing, 
logging of old-growth forest). Alternatively, if an ecosystem has 
become adapted to a disturbance pattern, it may be affected when the 
disturbance is removed (fire-maintained grasslands). The lack of 
natural analogues make it difficult to predict recovery from novel 
anthropogenic stressors (e.g., synthetic chemicals).
    The relative rate of recovery can also be estimated. For example, 
fish populations in a stream are likely to recover much faster from 
exposure to a degradable chemical than from habitat alterations 
resulting from stream channelization. Risk assessors can use knowledge 
of factors such as the temporal scales of organisms' life histories, 
the availability of adequate stock for recruitment, and the 
interspecific and trophic dynamics of the populations in evaluating the 
relative rates of recovery. A fisheries stock or forest might recover 
in several decades, a benthic infaunal community in years, and a 
planktonic community in weeks to months.
    Appendix E illustrates how the criteria for ecological adversity 
(nature and intensity of effects, spatial and temporal scales, and 
recovery) might be used in evaluating two cleanup options for a marine 
oil spill. This example also shows that recovery of a system depends 
not only on how quickly a stressor is removed but also on how any 
cleanup efforts affect the recovery.

5.3. Reporting Risks

    When risk characterization is complete, the risk assessors should 
be able to estimate ecological risks, indicate the overall degree of 
confidence in the risk estimates, cite lines of evidence supporting the 
risk estimates, and interpret the adversity of ecological effects. 
Usually this information is included in a risk assessment report 
(sometimes referred to as a risk characterization report because of the 
integrative nature of risk characterization). This section describes 
elements that risk assessors should consider when preparing a risk 
assessment report.
    Like the risk assessment itself, a risk assessment report may be 
brief or extensive depending on the nature of and the resources 
available for the assessment. While it is important to address the 
elements described below, risk assessors must judge the appropriate 
level of detail required. The report need not be overly complex or 
lengthy, depending on the nature of the risk assessment and the 
information required to support a risk management decision. In fact, it 
is important that information be presented clearly and concisely.
    While the breadth of ecological risk assessment precludes providing 
a detailed outline of reporting elements, the risk assessor should 
consider the elements listed in text note 5-8 when preparing a risk 
assessment report.
    To facilitate mutual understanding, it is critical that the risk 
assessment results are properly presented. Agency policy requires that 
risk characterizations be prepared ``in a manner that is clear, 
transparent, reasonable, and consistent with other risk 
characterizations of similar scope prepared across programs in the 
Agency'' (U.S. EPA 1995c). Ways to achieve such characteristics are 
described in text note 5-9.
    After the risk assessment report is prepared, the results are 
discussed with risk managers. Section 6 provides information on 
communication between risk assessors and risk managers, describes the 
use of the risk assessment in a risk management context, and briefly 
discusses communication of risk assessment results from risk managers 
to the public.

6. Relating Ecological Information to Risk Management Decisions

    After characterizing risks and preparing a risk assessment report 
(section 5), risk assessors discuss the results with risk managers 
(figure 5-1). Risk managers use risk assessment results along with 
other factors (e.g., economic or legal concerns) in making 
environmental decisions. The results also provide a basis for 
communicating risks to the public.
    Mutual understanding between risk assessors and risk managers can 
be facilitated if the questions listed in text note 6-1 are addressed. 
Risk managers need to know what the major risks (or potential risks) 
are with respect to assessment endpoints and have an idea of whether 
the conclusions are supported by a large body of data or if there are 
significant data gaps. When there is insufficient information to 
characterize risk at an appropriate level of detail due to a lack of 
resources, a lack of a consensus on how to interpret information, or 
other reasons, the issues, obstacles, and correctable deficiencies 
should be clearly articulated for the risk manager's consideration.
    In making a decision regarding ecological risks, risk managers use 
risk assessment results along with other information that may include 
social, economic, political, or legal issues. For example, the risk 
assessment may be used as part of a risk/benefit analysis, which may 
require translating resources (identified through the assessment 
endpoints) into monetary values. One difficulty with this approach is 
that traditional economic considerations may not adequately address 
things that are not considered commodities, intergenerational resource 
values or issues of long-term or irreversible effects (U.S. EPA, 
1995b). Risk managers may also consider risk mitigation options or 
alternative strategies for reducing risks. For example, risk mitigation 
techniques such as buffer strips or lower field application rates can 
be used to reduce the exposure (and risk) of a new pesticide. Further, 
risk managers may consider relative as well as absolute risk, for 
example, by comparing the risk of a new pesticide to other pesticides 
currently in use. Finally, risk managers consider public opinion and 
political demands in their decisions. Taken together, these other 
factors may render very high risks acceptable or very low risks 
unacceptable.
    Risk characterization provides the basis for communicating 
ecological risks to the public. This task is usually the responsibility 
of risk managers. Although the final risk assessment document 
(including its risk characterization sections) can be made available to 
the public, the risk communication process is best served by tailoring 
information to a particular audience. It is important to clearly 
describe the ecological resources at risk, their value, and the 
monetary and other costs of protecting (and failing to protect) the 
resources (U.S. EPA, 1995b).
    Managers should clearly describe the sources and causes of risks, 
the potential adversity of the risks (e.g., nature and intensity, 
spatial and temporal scale, and recovery potential). The degree of 
confidence in the risk assessment, the rationale for the risk 
management decision, and the options for reducing risk are also 
important (U.S. EPA, 1995b). Other risk communication considerations 
are provided in text note 6-2.
    Along with the discussions of risk and communications with the 
public, it is

[[Page 47603]]

important for risk managers to consider whether additional follow-on 
activities are required. Depending on the importance of the assessment, 
confidence level in the assessment results, and available resources, it 
may be advisable to conduct another iteration of the risk assessment 
(starting with problem formulation or analysis) in order to facilitate 
a final management decision. Another option is to proceed with the 
decision and develop a monitoring plan to evaluate the results of the 
decision (see section 1). For example, if the decision was to mitigate 
risks through exposure reduction, monitoring could help determine 
whether the desired reduction in exposure (and effects) was achieved.

7. Text Notes

Text Note 1-1. Related Terminology

    The following terms overlap to varying degrees with the broad 
concept of ecological risk assessment used in these guidelines (see 
Appendix B for definitions):
     Hazard assessment.
     Comparative risk assessment.
     Cumulative ecological risk assessment.
     Environmental impact statement.

Text Note 1-2. Flexibility of the Framework Diagram

    The framework process (figure 1-1) is a general representation of a 
complex and varied group of assessments, but this diagram should not be 
viewed as rigid and prescriptive. Rather, as illustrated by the 
examples below, broad applicability of the framework requires a 
flexible interpretation of the process.
     In problem formulation, an assessment may begin with a 
consideration of endpoints, stressors, or ecological effects. Problem 
formulation is frequently interactive and iterative rather than linear.
     In the analysis phase, it may be difficult to maintain a 
clear distinction between exposure and effects analyses in all but the 
simplest systems. Exposure and effects frequently become intertwined, 
as when an initial exposure leads to a cascade of additional exposures 
and effects. It is important that a risk assessment is based on an 
understanding of these complex relationships.
     Analysis and risk characterization are shown as separate 
phases. However, some models may combine the analysis of exposure and 
effects data with the integration of these data that occurs in risk 
characterization.

Text Note 1-3. The Iterative Nature of Ecological Risk Assessment

    The ecological risk assessment process is by nature iterative. For 
example, it may take more than one pass through problem formulation to 
complete planning for the risk assessment, or information gathered in 
the analysis phase may suggest further problem formulation activities 
such as modification of the endpoints selected.
    To maximize efficient use of limited resources, ecological risk 
assessments are frequently designed in sequential tiers that proceed 
from simple, relatively inexpensive evaluations to more costly and 
complex assessments. Initial tiers are based on conservative 
assumptions, such as maximum exposure and ecological sensitivity. When 
an early tier cannot define risk to support a management decision, a 
higher assessment tier is used that may require either additional data 
or applying more refined analysis techniques to available data. 
Iterations proceed until sufficient information is available to support 
a sound management decision, within the constraints of available 
resources.
    Because a tiered approach can incorporate standardized decision 
points and supporting analyses, it can be particularly useful for 
multiple assessments of similar stressors or situations. However, it is 
difficult to generalize further concerning tiered risk assessments 
because they are used to answer so many different questions. Examples 
of organizations that use, are considering, or have advocated using 
tiered ecological risk assessments include the Canadian government 
(proposed, Gaudet, 1994), the European Community (E.C., 1993), industry 
(Cowan et al., 1995), the Aquatic Dialogue Group (SETAC 1994a), and the 
U.S. EPA Offices of Pesticide Programs (Urban and Cook, 1986), 
Pollution Prevention and Toxics (Lynch et al., 1994), and Superfund 
(document in preparation).

Text Note 2-1. Who Are Risk Managers?

    Risk managers are individuals and organizations that take 
responsibility for, or have the authority to take action or require 
action, to mitigate an identified risk. The expression ``risk manager'' 
is often used to represent a decisionmaker in agencies like EPA or 
state environmental offices who has the authority to protect or manage 
a resource. However, risk managers often represent a diverse group of 
interested parties that influence the outcome of resource protection 
efforts. Particularly as the scope of environmental management expands 
to communities, the meaning of risk manager significantly expands to 
include decision officials in Federal, state, and local governments, as 
well as private-sector leaders in commercial, industrial, and private 
organizations. Risk managers may also include constituency groups, 
other interested parties, and the public. In situations where a complex 
of ecosystem values (e.g., watershed resources) is at risk from 
multiple stressors, many of these groups may act together as risk 
management teams. For additional insights on risk management and 
manager roles, see text notes 2-3 and 2-4.

Text Note 2-2. Who Are Risk Assessors?

    Risk assessors are a diverse group of professionals who bring a 
needed expertise to a risk assessment. When a specific risk assessment 
process is well defined through regulations and guidance, one trained 
individual may be able to complete a risk assessment if needed 
information is available (e.g., premanufacture notice of a chemical). 
However, as more complex risk assessments become common, it will be 
rare that one individual can provide the necessary breadth of 
expertise. Every risk assessment team should include at least one 
professional who is knowledgeable and experienced in using the risk 
assessment process. Other team members bring specific expertise 
relevant to the location, the stressors, the ecosystem, and the 
scientific issues and other expertise as determined by the type of 
assessment.

Text Note 2-3. Questions Addressed by Risk Managers and Risk Assessors

Questions Principally for Risk Managers:
    What is the nature of the problem and the best scale for the 
assessment?
    What are the management goals and decisions needed, and how will 
risk assessment help?
    What are the ecological values of concern?
    What are the policy considerations (law, corporate stewardship, 
societal concerns, environmental justice)?
    What precedents are set by previous risk assessments and decisions?
    What is the context of the assessment (e.g., industrial, national 
park)?
    What resources (e.g., personnel, time, money) are available?
    What level of uncertainty is acceptable?
Questions Principally for Risk Assessors
    What is the scale of the risk assessment?

[[Page 47604]]

    What are the critical ecological endpoints and ecosystem and 
receptor characteristics?
    How likely is recovery and how long will it take?
    What is the nature of the problem: past, present, future?
    What is our state of knowledge on the problem?
    What data and data analyses are available and appropriate?
    What are the potential constraints (e.g., limits on expertise, 
time, availability of methods and data)?

Text Note 2-4. The Role of Interested Parties

    The involvement of all interested and affected parties, which 
``stakeholder'' is commonly used to represent, is important to the 
development of management goals for some risk assessments. The greater 
the involvement, the broader the base of consensus about those goals. 
With strong consensus on management goals, decisions are more likely to 
be supported by all community groups during implementation of 
management plans. However, the context of this involvement can vary 
widely, and the ability to achieve consensus often decreases as the 
size of the management team increases. Where large diverse groups need 
to come to consensus, social science professionals and methods for 
consensus building become increasingly important. Interested parties 
become risk managers when they influence risk reduction. See additional 
discussion in text note 2-1 and section 2.2.

Text Note 2-5. Sustainability as a Management Goal

    Sustainability is used repeatedly as a management goal in a variety 
of settings (see U.S. EPA, 1995b). To sustain is to prolong, to hold up 
under, or endure (Merriam-Webster, 1972). Sustainability and other 
concepts such as biotic or community integrity are very useful as 
guiding principles for management goals. However, in each case these 
principles must be explicitly interpreted to support a risk assessment. 
To do this, key questions need to be addressed: What does 
sustainability or integrity mean for the particular ecosystem? What 
must be protected to meet sustainable goals or system integrity? Which 
ecological resources and processes are to be sustained and why? How 
will we know we have achieved it? Answers to these questions serve to 
clarify the goals for a particular ecosystem. Concepts like 
sustainability and integrity do not meet the criteria for an assessment 
endpoint (see section 3.3.2).

Text Note 2-6. Management Goals for Waquoit Bay

    Waquoit Bay is a small estuary on Cape Cod showing signs of 
degradation, including loss of eelgrass, fish, and shellfish and 
increasing macroalgae mats and fish kills. The management goal for 
Waquoit Bay was established through public meetings, preexisting goals 
from local organizations, and state and Federal regulations:
    Reestablish and maintain water quality and habitat conditions in 
Waquoit Bay and associated freshwater rivers and ponds to (1) support 
diverse self-sustaining commercial, recreational, and native fish and 
shell fish populations, and (2) reverse ongoing degradation of 
ecological resources in the watershed.
    To define this goal, it was interpreted into 10 objectives, two of 
which are:
     Reestablish a self-sustaining scallop population in the 
bay that can support a viable sport fishery.
     Reduce or eliminate nuisance macroalgal growth.
    From these objectives, specific ecological resources in the bay 
were identified to provide the basis for the risk assessment, one of 
which is:
    Areal extent and patch size of eelgrass beds.
    Eelgrass was selected because scallops are dependent directly on 
eelgrass beds for survival and eelgrass is highly sensitive to excess 
macroalgal growth.

Text Note 2-7. Questions to Ask About Scope and Complexity

    Is this risk assessment legally mandated, addressing a court-
ordered decision, or providing guidance to a community?
    Are decisions more likely based on assessments of a small area 
evaluated in-depth or a large-scale area in less detail?
    What are the spatial and temporal boundaries of the problem?
    What kinds of information are already available compared to what is 
needed?
    How much time can be taken and how many resources are available?
    What practicalities constrain data collection?
    Is a tiered approach an option?

Text Note 3-1. Avoiding Potential Shortcomings Through Problem 
Formulation

    The importance of problem formulation has been shown repeatedly in 
the Agency's analysis of ecological risk assessment case studies and in 
interactions with senior EPA managers and regional risk assessors (U.S. 
EPA, 1993a, 1994a). Consistent shortcomings identified in the case 
studies include (1) absence of clearly defined goals, (2) endpoints 
that are ambiguous and difficult to define and measure, and (3) failure 
to identify important risks. These and other shortcomings can be 
avoided through rigorous development of the products of problem 
formulation as described in this section of the guidelines.

Text Note 3-2. Uncertainty in Problem Formulation

    In each product of problem formulation there are elements of 
uncertainty, a consideration of what is known and not known about a 
problem and its setting. The explicit treatment of uncertainty during 
problem formulation is particularly important because it will have 
repercussions throughout the remainder of the assessment. Uncertainty 
is discussed in section 3.4, Conceptual Models, because uncertainty in 
problem formulation is articulated in these models.

Text Note 3-3. Assessing Available Information: Questions to Ask 
Concerning Source, Stressor, and Exposure Characteristics, Ecosystem 
Characteristics, and Effects

Source and Stressor Characteristics
     What is the source? Is it anthropogenic, natural, point 
source, or diffuse nonpoint?
     What type of stressor is it: chemical, physical, or 
biological?
     What is the intensity of the stressor (e.g., the dose or 
concentration of a chemical, the magnitude or extent of physical 
disruption, the density or population size of a biological stressor)?
     What is the mode of action? How does the stressor act on 
organisms or ecosystem functions?
Exposure Characteristics
     With what frequency does a stressor event occur (e.g., is 
it isolated, episodic, or continuous; is it subject to natural daily, 
seasonal, or annual periodicity)?
     What is its duration? How long does it persist in the 
environment (e.g., for chemical, what is its half-life, does it 
bioaccumulate; for physical, is habitat alteration sufficient to 
prevent recovery; for biological, will it reproduce and proliferate)?
     What is the timing of exposure? When does it occur in 
relation to critical organism life cycles or ecosystem events (e.g., 
reproduction, lake overturn)?
     What is the spatial scale of exposure? Is the extent or 
influence of the stressor local, regional, global, habitat-specific, or 
ecosystemwide?

[[Page 47605]]

     What is the distribution? How does the stressor move 
through the environment (e.g., for chemical, fate and transport; for 
physical, movement of physical structures; for biological, life history 
dispersal characteristics)?
Ecosystems Potentially at Risk
     What are the geographic boundaries? How do they relate to 
functional characteristics of the ecosystem?
     What are the key abiotic factors influencing the ecosystem 
(e.g., climatic factors, geology, hydrology, soil type, water quality)?
     Where and how are functional characteristics driving the 
ecosystem (e.g., energy source and processing, nutrient cycling)?
     What are the structural characteristics of the ecosystem 
(e.g., species number and abundance, trophic relationships)?
     What habitat types are present?
     How do these characteristics influence the susceptibility 
(sensitivity and likelihood of exposure) of the ecosystem to the 
stressor(s)?
     Are there unique features that are particularly valued 
(e.g., the last representative of an ecosystem type)?
     What is the landscape context within which the ecosystem 
occurs?
Ecological Effects
     What are the type and extent of available ecological 
effects information (e.g., field surveys, laboratory tests, or 
structure-activity relationships)?
     Given the nature of the stressor (if known), which effects 
are expected to be elicited by the stressor?
     Under what circumstances will effects occur?

Text Note 3-4. Initiating a Risk Assessment: What's Different When 
Stressors, Effects, or Values Drive the Process?

    The reasons for initiating a risk assessment also influence how the 
risk assessor proceeds through the process of problem formulation. When 
the assessment is initiated due to concerns about stressors, risk 
assessors use what is known about the characteristics of the stressor 
and its source to focus the assessment. Goals are articulated based on 
how the stressor is likely to cause risk to possible receptors that may 
become exposed. This information forms the basis for developing 
conceptual models and selecting assessment endpoints. When an observed 
effect is the basis for initiating the assessment, endpoints are 
normally established first. Often these endpoints involve affected 
ecological entities and their response. Goals for protecting the 
assessment endpoints are then established, which support the 
development of conceptual models. The models aid in the identification 
of the most likely stressor(s). Value-initiated risk assessments are 
driven up front by goals for the ecological value of concern. These 
values might involve ecological entities such as species, communities, 
ecosystems, or places. Based on these goals, assessment endpoints are 
selected first to serve as an interpretation of the goals. Once 
selected, the endpoints provide the basis for identifying an array of 
stressors that may be influencing them, and describing the diversity of 
potential effects. This information is then captured in the conceptual 
model(s).

Text Note 3-5. Salmon and Hydropower: Salmon as the Basis for an 
Assessment Endpoint

    A hydroelectric dam is to be built on a river in the Pacific 
Northwest where anadromous fish such as salmon spawn. Assessment 
endpoints must be selected to assess potential ecological risk. Of the 
anadromous fish, salmon that spawn in the river are an appropriate 
choice because they meet the criteria for good assessment endpoints. 
Salmon fry and adults are important food sources for a multitude of 
aquatic and terrestrial species and are major predators of aquatic 
invertebrates (ecological relevance). Salmon are sensitive to changes 
in sedimentation and substrate pebble size, require quality cold water 
habitats, and have difficulty climbing fish ladders. Hydroelectric dams 
represent significant and normally fatal habitat alteration and 
physical obstacles to successful salmon breeding and fry survival 
(susceptibility). Finally, salmon support a large commercial fishery, 
some species are endangered, and they have ceremonial importance and 
are key food sources for Native Americans (basis for management goals). 
``Salmon reproduction and population maintenance'' is a good assessment 
endpoint for this risk assessment, and if salmon populations are 
protected, other anadromous fish populations are likely to be protected 
as well. However, one assessment endpoint can rarely provide the basis 
for a risk assessment of complex ecosystems. These are better 
represented by a set of assessment endpoints.

Text Note 3-6. Cascading Adverse Effects: Primary (Direct) and 
Secondary (Indirect)

    The interrelationships among entities and processes in ecosystems 
result in the potential for cascading effects: as one population, 
species, process, or other entity in the ecosystem is altered, other 
entities are affected as well. Primary, or direct, effects occur when a 
stressor acts directly on the assessment endpoint and causes an adverse 
response. Secondary, or indirect, effects occur when the response of an 
ecological entity to a stressor becomes a stressor to another entity. 
Secondary effects are not limited in number. They often are a series of 
effects among a diversity of organisms and processes that cascade 
through the ecosystem. For example, application of an herbicide on a 
wet meadow results in direct toxicity to plants. Death of the wetland 
plants leads to secondary effects such as loss of feeding habitat for 
ducks, breeding habitat for red-winged black birds, alteration of 
wetland hydrology that changes spawning habitat for fish, and so forth.

Text Note 3-7. Sensitivity and Secondary Effects: The Mussel-Fish 
Connection

    Native freshwater mussels are endangered in many streams. 
Management efforts have focused on maintaining suitable habitat for 
mussels because habitat loss has been considered the greatest threat to 
this group. However, larval unionid mussels must attach to the gills of 
a fish host for one month during development. Each species of mussel 
must attach to a particular host species of fish. In situations where 
the fish community has been changed, perhaps due to stressors to which 
mussels are insensitive, the host fish may no longer be available. 
Mussel larvae will die before reaching maturity as a result. Regardless 
of how well managers restore mussel habitat, mussels will be lost from 
this system unless the fish community is restored. In this case, 
exposure to the absence of a critical resource is the source of risk.

[[Page 47606]]



                      Text Note 3-8.--Examples of Management Goals and Assessment Endpoints                     
----------------------------------------------------------------------------------------------------------------
                Case                        Regulatory context/management goal            Assessment endpoint   
----------------------------------------------------------------------------------------------------------------
Assessing Risks of New Chemical      Protect ``the environment'' from ``an            Survival, growth, and     
 Under Toxic Substances Control Act   unreasonable risk of injury'' (TSCA Sec.  2[b]   reproduction of fish,    
 (Lynch et al., 1994).                [1] and [2]); protect the aquatic environment.   aquatic invertebrates,   
                                      Goal was to exceed a concentration of concern    and algae.               
                                      by no more than 20 days a year.                                           
Special Review of Granular           Prevent * * * ``unreasonable adverse effects on  Individual bird survival. 
 Carbofuran Based on Adverse          the environment'' (FIFRA Secs.  3[c][5] and                               
 Effects on Birds (Houseknecht,       3[c][6]); using cost-benefit considerations.                              
 1993).                               Goal was no regularly repeated bird kills.                                
Modeling Future Losses of            National Environmental Policy Act may apply to   (1) Forest community      
 Bottomland Forest Wetlands (Brody    environmental impact of new levee                structure and habitat    
 et al., 1993).                       construction; also Clean Water Act Sec.  404.    value to wildlife        
                                                                                       species.                 
                                                                                      (2) Species composition of
                                                                                       wildlife community.      
Pest Risk Assessment on Importation  This assessment was done to help provide a       Survival and growth of    
 of Logs From Chile (USDA, 1993).     basis for any necessary regulation of the        tree species in the      
                                      importation of timber and timber products into   western United States.   
                                      the United States.                                                        
Baird and McGuire Superfund Site     Protection of the environment (CERCLA/SARA)....  (1) Survival of soil      
 (terrestrial component);                                                              invertebrates.           
 (Burmaster et al., 1991; Callahan                                                    (2) Survival and          
 et al., 1991; Menzie et al., 1992).                                                   reproduction of song     
                                                                                       birds.                   
Waquoit Bay Estuary Watershed Risk   Clean Water Act--wetlands protection; water      (1) Estuarine eelgrass    
 Assessment.                          quality criteria--pesticides; endangered         habitat abundance and    
                                      species. National Estuarine Research Reserve,    distribution.            
                                      Massachusetts, Area of Critical Environmental   (2) Estuarine fish species
                                      Concern. Goal was to reestablish and maintain    diversity and abundance. 
                                      water quality and habitat conditions to         (3) Freshwater pond       
                                      support diverse self-sustaining commercial,      benthic invertebrate     
                                      recreational, and native fish, water-dependent   species diversity and    
                                      wildlife, and shellfish, and reverse ongoing     abundance.               
                                      degradation.                                                              
----------------------------------------------------------------------------------------------------------------

Text Note 3-9. Common Problems in Selecting Assessment Endpoints

     Endpoint is a goal (e.g., maintain and restore endemic 
populations).
     Endpoint is vague (e.g., estuarine integrity instead of 
eelgrass abundance and distribution).
     Ecological entity is better as a measure (e.g., measure 
emergence of midges for endpoint on feeding of fish).
     Ecological entity may not be as sensitive to the stressor 
(e.g., catfish versus salmon for sedimentation).
     Ecological resource is not exposed to the stressor (e.g., 
using insectivorous birds for avian risk of pesticide application to 
seeds).
     Ecological resources are irrelevant to the assessment 
(e.g., lake fish in salmon stream).
     Value of a species or attributes of an ecosystem are not 
fully considered (e.g., mussel-fish connection, see text note
3-7).
     Attribute is not sufficiently sensitive for detecting 
important effects (e.g., survival compared with recruitment for 
endangered species).

Text Note 3-10. What Are Risk Hypotheses and Why Are They Important?

    Risk hypotheses are proposed answers to questions risk assessors 
have about what responses assessment endpoints (and measures) will show 
when they are exposed to stressors and how exposure will occur. Risk 
hypotheses clarify and codify relationships that are posited through 
the consideration of available data, information from scientific 
literature, and the best professional judgment by risk assessors 
developing the conceptual models. This explicit process opens the risk 
assessment to peer review and evaluation to ensure the scientific 
validity of the work. Risk hypotheses are not equivalent to statistical 
testing of null and alternative hypotheses. However, predictions 
generated from risk hypotheses can be tested in a variety of ways, 
including standard statistical approaches.

Text Note 3-11. Examples of Risk Hypotheses

    Hypotheses include known information that sets the problem in 
perspective and the proposed relationships that need evaluation.
    Stressor-initiated: Chemicals with a high Kow tend to 
bioaccumulate. Premanufacture notice (PMN) chemical A has a Kow of 
5.5 and similar molecular structure as known chemical stressor B. 
Hypotheses: Based on the Kow of chemical A, the mode of action of 
chemical B, and the food web of the target ecosystem, when the PMN 
chemical is released at a specified rate, it will bioaccumulate 
sufficiently in 5 years to cause developmental problems in wildlife and 
fish.
    Effects-initiated: Bird kills were repeatedly observed in golf 
courses following the application of the pesticide carbofuran, which is 
highly toxic. Hypotheses: Birds die when they consume recently applied 
granulated carbofuran; as the level of application increases, the 
number of dead birds increases. Exposure occurs when dead and dying 
birds are consumed by other animals. Birds of prey and scavenger 
species will die from eating contaminated birds.
    Ecological value-initiated: Waquoit Bay, Massachusetts, supports 
recreational boating and commercial and recreational shellfishing and 
is a significant nursery for fish. Large mats of macroalgae clog the 
estuary, most of the eelgrass has died, and scallops are gone. 
Hypotheses: Nutrient loading from septic systems, air pollution, and 
lawn fertilizers cause eelgrass loss by shading from algal growth, and 
direct toxicity from nitrogen compounds. Fish and shellfish populations 
are decreasing because of loss of eelgrass habitat and periodic 
hypoxia.

Text Note 3-12. What Are the Benefits of Developing Conceptual Models?

     The process of creating a conceptual model is a powerful 
learning tool.
     Conceptual models can be improved as knowledge increases.

[[Page 47607]]

     Conceptual models highlight what we know and don't know 
and can be used to plan future work.
     Conceptual models can be a powerful communication tool. 
They provide an explicit expression of our assumptions and 
understanding of a system for others to evaluate.
     Conceptual models provide a framework for prediction and 
are the template for generating more risk hypotheses.

Text Note 3-13. Uncertainty in Problem Formulation

    Uncertainties in problem formulation are manifested in the quality 
of conceptual models. To describe uncertainty:
     Be explicit in defining assessment endpoints; include both 
entity and measurable attributes.
     Reduce or define variability by carefully defining 
boundaries for the assessment.
     Be open and explicit about the strengths and limitations 
of pathways and relationships depicted in the conceptual model.
     Identify and describe rationale for key assumptions made 
because of lack of knowledge, model simplification, approximation, or 
extrapolation.
     Describe data limitations.

Text Note 3-14. Examples of Assessment Endpoints and Measures (see also 
section 3.5.1)

    Assessment Endpoint: Coho salmon breeding success and fry survival.
Measures of Effects
     Egg and fry response to low dissolved oxygen.
     Adult behavior in response to obstacles.
     Spawning behavior and egg survival in response to 
sedimentation.
Measures of Ecosystem and Receptor Characteristics
     Water temperature, water velocity, and physical 
obstructions.
     Abundance and distribution of suitable breeding substrate.
     Abundance and distribution of suitable food sources for 
fry.
     Feeding, resting, and reproductive cycles.
     Natural population structure (proportion of different size 
and age classes).
     Laboratory evaluation of reproduction, growth, and 
mortality.
Measures of Exposure
     Number and height of hydroelectric dams.
     Toxic chemical concentrations in water, sediment, and fish 
tissue.
     Nutrient and dissolved oxygen levels in ambient waters.

Text Note 3-15. Selecting What To Measure

    Direct measurement of assessment endpoint responses is often not 
possible. Under these circumstances, the selection of a surrogate 
response measure is necessary. The selection of what, where, and how to 
measure determines whether the risk assessment is still relevant to 
management decisions about an assessment endpoint. For example, a risk 
assessment may be conducted to evaluate the potential risk of a 
pesticide used on seeds. Birds and mammals may be selected as the 
entities for assessment endpoints. However, to ensure that the 
organisms selected are susceptible to the pesticide, only those that 
eat seeds should be chosen. While insectivorous birds may serve as a 
good surrogate measure for determining the sensitivity of birds to the 
pesticide, they do not address issues of exposure. To evaluate 
susceptibility, the appropriate assessment endpoints in this case would 
be seed-eating birds and mammals. Problem formulations based on 
assessment endpoints that are both sensitive and likely to be exposed 
to the stressor will be relevant to management concerns. If assessment 
endpoints are not susceptible, their use in assessing risk can lead to 
poor management decisions.

Text Note 3-16. How Do Water Quality Criteria Relate to Assessment 
Endpoints?

    Water quality criteria (U.S. EPA, 1986a) have been developed for 
the protection of aquatic life from chemical stressors. This text note 
shows how the elements of a water quality criterion correspond to 
management goals, assessment endpoints, and measures.
Regulatory Goal
     Clean Water Act, Sec. 101: Protection of the chemical, 
physical, and biological integrity of the Nation's waters.
Program Management Objective
     Protect 99% of individuals in 95% of the species in 
aquatic communities from acute and chronic effects resulting from 
exposure to a chemical stressor.
Assessment Endpoints
     Survival of fish, aquatic invertebrate, and algal species 
under acute exposure.
     Survival, growth, and reproduction of fish, aquatic 
invertebrate, and algal species under chronic exposure.
Measures of Effect
     Laboratory LC50s for at least eight species meeting 
certain requirements.
     Chronic NOAELs for at least three species meeting certain 
requirements.
Measures of Ecosystem and Receptor Characteristics
     Water hardness (for some metals).
     pH.
    The water quality criterion is a benchmark level derived from a 
distributional analysis of single-species toxicity data. It is assumed 
that the species tested adequately represent the composition and 
sensitivities of species in a natural community.

Text Note 3-17. Data Quality Objectives (DQO) Process

    The DQO process combines elements of both planning and problem 
formulation in its seven-step format.
    Step 1--State the problem. Review existing information to concisely 
describe the problem to be studied.
    Step 2--Identify the decision. Determine what questions the study 
will try to resolve and what actions may result.
    Step 3--Identify inputs to the decision. Identify information and 
measures needed to resolve the decision statement.
    Step 4--Define study boundaries. Specify time and spatial 
parameters and where and when data should be collected.
    Step 5--Develop decision rule. Define statistical parameter, action 
level, and logical basis for choosing alternatives.
    Step 6--Specify tolerable limits on decision errors. Define limits 
based on the consequences of an incorrect decision.
    Step 7--Optimize the design. Generate alternative data collection 
designs and choose most resource-effective design that meets all DQOs.

Text Note 4-1. Data Collection and the Analysis Phase

    Data needs are identified during problem formulation (the analysis 
plan step), and data are collected before the start of the analysis 
phase. These data may be collected for the specific purpose of a 
particular risk assessment, or they may be available from previous 
studies. If additional data needs are identified as the assessment 
proceeds, the analysis phase may be temporarily halted while they are 
collected or the assessor may choose to iterate the problem formulation 
again. Data collection methods are not described in these guidelines. 
However, the evaluation of data for the purposes of

[[Page 47608]]

risk assessment is discussed in section 4.2.

Text Note 4-2. The American National Standard for Quality Assurance

    The Specifications and Guidelines for Quality Systems for 
Environmental Data Collection and Environmental Technology Programs 
(ASQC, 1994) recognizes several areas that are important to ensuring 
that environmental data will meet study objectives, including:
     Planning and scoping.
     Design of data collection operations.
     Implementation and monitoring of planned operations.
     Assessment and verification of data usability.

Text Note 4-3. Questions for Evaluating a Study's Utility for Risk 
Assessment

    How do study objectives compare with those of the risk assessment?
    Are the variables and conditions the study represents compared to 
those important to the risk assessment?
    Was the study design adequate to meet its objectives?
    Was the study conducted properly?
    How were variability and uncertainty treated and reported?

Text Note 4-4. Considering the Degree of Aggregation in Models

    Wiegert and Bartell (1994) suggest the following considerations for 
evaluating the proper degree of aggregation or disaggregation:
    (1) Do not aggregate components with greatly disparate rates of 
fluxes;
    (2) Do not greatly increase the disaggregation of the structural 
aspects of the model without a corresponding increase in the 
sophistication of the functional relationships and controls; and
    (3) Disaggregate models only insofar as required by the goals of 
the model to facilitate testing.

Text Note 4-5. Questions for Source Description

    Where does the stressor originate?
    What environmental medium first receives stressors?
    Does the source generate other constituents that will influence a 
stressor's eventual distribution in the environment?
    Are there other sources of the same stressor?
    Are there background sources?
    Is the source still active?
    Does the source produce a distinctive signature that can be seen in 
the environment, organisms or communities?
Additional Questions for Introduction of Biological Stressors
    Is there an opportunity for repeated introduction or escape into 
the new environment?
    Will the organism be present on a transportable item?
    Are there mitigation requirements or conditions that would kill or 
impair the organism before entry, during transport, or at the port of 
entry?

Text Note 4-6. Questions To Ask in Evaluating Stressor Distribution

    What are the important transport pathways?
    What characteristics of the stressor influence transport?
    What characteristics of the ecosystem will influence transport?
    What secondary stressors will be formed?
    Where will they be transported?

Text Note 4-7. General Mechanisms of Transport and Dispersal

Physical, Chemical and Biological Stressors
     By air current.
     In surface water (rivers, lakes, streams).
     Over and/or through the soil surface.
     Through ground water.
Primarily Chemical Stressors
     Through the food web.
Primarily Biological Stressors
     Splashing or raindrops.
     Human activity (boats, campers).
     Passive transmittal by other organisms.
     Biological vectors.

Text Note 4-8. Questions To Ask in Describing Contact or Co-Occurrence

    Must the receptor actually contact the stressor for adverse effects 
to occur?
    Must the stressor be taken up into a receptor for adverse effects 
to occur?
    What characteristics of the receptors will influence the extent of 
contact or co-occurrence?
    Will abiotic characteristics of the environment influence the 
extent of contact or co-occurrence?
    Will ecosystem processes or community-level interactions influence 
the extent of contact or co-occurrence?

Text Note 4-9. Example of an Exposure Equation: Calculating a Potential 
Dose via Ingestion
[GRAPHIC] [TIFF OMITTED] TN09SE96.020

Where:

ADDpot=Potential average daily dose (e.g., in mg/kg-day)
Ck=Average contaminant concentration in the kth type of food 
(e.g., in mg/kg wet weight)
FRk=Fraction of intake of the kth food type that is from the 
contaminated area (unitless)
NIRk=Normalized ingestion rate of the kth food type on a wet-
weight basis (e.g., in g food/g body-weight-day).
m=Number of contaminated food types

Source: U.S. EPA, 1993c

Text Note 4-10. Measuring Internal Dose Using Biomarkers and Tissue 
Residues

    Biomarkers, tissue residues, or other bioassessment methods may be 
useful in estimating or confirming exposure in cases where 
bioavailability is expected to be a significant issue, but the factors 
influencing it are not known. They can also be very useful when the 
metabolism and accumulation kinetics are important factors (McCarty and 
Mackay, 1993). These methods are most useful when they can be 
quantitatively linked to the amount of stressor originally contacted by 
the organism. In addition, they are most useful when the stressor-
response relationship expresses the amount of stressor in terms of the 
tissue residues or biomarkers. Additional information and some 
considerations for their development can be found in Huggett et al. 
(1992).

Text Note 4-11. Questions Addressed by the Exposure Profile

    How does exposure occur?
    What is exposed?
    How much exposure occurs? When and where does it occur?
    How does exposure vary?
    How uncertain are the exposure estimates?
    What is the likelihood that exposure will occur?

Text Note 4-12. Questions for Stressor-Response Analysis

    Does the assessment require point estimates or stressor-response 
curves?
    Does the assessment require the establishment of a ``no-effect'' 
level?
    Would cumulative effects distributions be useful?

Text Note 4-13. Qualitative Stressor-Response Relationships

    The relationship between stressor and response can be described 
qualitatively, for instance, using categories of high, medium, and low, 
to describe the intensity of response given exposure to a stressor. For 
example, Pearlstine et al. (1985) assumed that seeds would not

[[Page 47609]]

germinate if they were inundated with water at the critical time. This 
stressor-response relationship was described simply as a yes or no. In 
most cases, however, the objective is to describe quantitatively the 
intensity of response associated with exposure, and in the best case, 
to describe how intensity of response changes with incremental 
increases in exposure.

Text Note 4-14. Median Effect Levels

    Median effects are those effects elicited in 50% of the test 
organisms exposed to a stressor, typically chemical stressors. Median 
effect concentrations can be expressed in terms of lethality or 
mortality and are known as LC50 or LD50, depending on whether 
concentrations (in the diet or in water) or doses (mg/kg) were used. 
Median effects other than lethality (e.g., effects on growth) are 
expressed as EC50 or ED50. The median effect level is always 
associated with a time parameter (e.g., 24 or 48 hr). Because these 
tests seldom exceed 96 hr, their main value lies in evaluating short-
term effects of chemicals. Stephan (1977) discusses several statistical 
methods to estimate the median effect level.

Text Note 4-15. No-Effect Levels Derived From Statistical Hypothesis 
Testing

    Statistical hypothesis tests have typically been used with chronic 
toxicity tests of chemical stressors that evaluate multiple endpoints. 
For each endpoint, the objective is to determine the highest test 
concentration for which effects are not statistically different from 
the controls (the no observed adverse effect concentration, NOAEC) and 
the lowest concentration at which effects were statistically 
significant from the control (the lowest observed adverse effect 
concentration, LOAEC). The range between the NOAEC and the LOAEC is 
sometimes called the maximum acceptable toxicant concentration, or 
MATC. The MATC, which can also be reported as the geometric mean of the 
NOAEC and the LOAEC, provides a useful reference with which to compare 
toxicities of various chemical stressors.
    Reporting the results of chronic tests in terms of the MATC or 
GMATC has been widely used within the Agency for evaluating pesticides 
and industrial chemicals (e.g., Urban and Cook, 1986; Nabholz, 1991).

Text Note 4-16. General Criteria for Causality (Adapted From Fox, 1991)

    Criteria strongly affirming causality:
     Strength of association.
     Predictive performance.
     Demonstration of a stressor-response relationship.
     Consistency of association.
    Criteria providing a basis for rejecting causality:
     Inconsistency in association.
     Temporal incompatibility.
     Factual implausibility.
    Other relevant criteria:
     Specificity of association.
     Theoretical and biological plausibility.

Text Note 4-17. Koch's Postulates (Pelczar and Reid, 1972)

     A pathogen must be consistently found in association with 
a given disease.
     The pathogen must be isolated from the host and grown in 
pure culture.
     When inoculated into test animals, the same disease 
symptoms must be expressed.
     The pathogen must again be isolated from the test 
organism.

Text Note 4-18. Examples of Extrapolations to Link Measures of Effect 
to Assessment Endpoints

    Every risk assessment has data gaps that must be addressed, but it 
is not always possible to obtain more information. When there is a lack 
of time, monetary resources, or a practical means to acquire more data, 
extrapolations such as those listed below may be the only way to bridge 
gaps in available data. Extrapolations may be:
     Between taxa (e.g., bluegill to rainbow trout).
     Between responses (e.g., mortality to growth or 
reproduction).
     From laboratory to field.
     Between geographic areas.
     Between spatial scales.
     From data collected over a short timeframe to longer-term 
effects.

Text Note 4-19. Questions Related to Selecting Extrapolation Approaches

    How specific is the assessment endpoint?
    Does the spatial or temporal extent of exposure suggest the need 
for additional receptors or extrapolation models?
    Are the quantity and quality of the data available sufficient for 
planned extrapolations and models?
    Is the proposed extrapolation technique consistent with ecological 
information?
    How much uncertainty is acceptable?

Text Note 4-20. Questions To Consider When Extrapolating From Effects 
Observed in the Laboratory to Field Effects of Chemicals

    Exposure factors:
    How will environmental fate and transformation of the chemical 
effect exposure in the field?
    How comparable are exposure conditions and the timing of exposure?
    How comparable are the routes of exposure?
    How do abiotic factors influence bioavailability and exposure?
    How likely are preference or avoidance behaviors?
    Effects factors:
    What is known about the biotic and abiotic factors controlling 
populations of the organisms of concern?
    To what degree are critical life stage data available?
    How may exposure to the same or other stressors in the field have 
altered organism sensitivity?

Text Note 4-21. Questions Addressed by the Stressor-Response Profile

    What ecological entities are affected?
    What is the nature of the effect(s)?
    What is the intensity of the effect(s)?
    Where appropriate, what is the time scale for recovery?
    What causal information links the stressor with any observed 
effects?
    How do changes in measures of effects relate to changes in 
assessment endpoints?
    What is the uncertainty associated with the analysis?

Text Note 5-1. Using Qualitative Categories to Estimate Risks of an 
Introduced Species

    The importation of logs from Chile required an assessment of the 
risks posed by the potential introduction of the bark beetle, Hylurgus 
ligniperda (USDA, 1993). Experts to judged the potential for 
colonization and spread of the species, and their opinions were 
expressed as high, medium, or low as to the likelihood of establishment 
(exposure) or consequential effects of the beetle. Uncertainties were 
similarly expressed. A ranking scheme was then used to sum the 
individual elements into an overall estimate of risk (high, medium, or 
low). Narrative explanations of risk accompanied the overall rankings.

Text Note 5-2. Applying the Quotient Method

    When applying the quotient method to chemical stressors, the 
effects concentration or dose (e.g., an LC50, LD50, 
EC50, ED50, NOAEL, or LOAEL) is frequently adjusted by 
uncertainty modifying factors prior to division into the exposure 
number (U.S. EPA, 1984; Nabholz, 1991; Urban and Cook, 1986; see 
section 4.3.1.3), although EPA used a slightly different approach in

[[Page 47610]]

estimating the risks to the survival of birds that forage in 
agricultural areas where the pesticide granular carbofuran is applied 
(Houseknecht, 1993). In this case, EPA calculated the quotient by 
dividing the estimated exposure levels of carbofuran granules in 
surface soils (number/ft2) by the granules/LD50 derived from 
single-dose avian toxicity tests. The calculation yields values with 
units of LD50/ft2. It was assumed that a higher quotient 
value corresponded to an increased likelihood that a bird would be 
exposed to lethal levels of granular carbofuran at the soil surface. 
Minimum and maximum values for LD50/ft2 were estimated for 
songbirds, upland game birds, and waterfowl that may forage within or 
near 10 different agricultural crops.

Text Note 5-3. Comparing an Exposure Distribution With a Point Estimate 
of Effects

    The EPA Office of Pollution Prevention and Toxics uses a 
Probabilistic Dilution Model (PDM3) to generate a distribution of daily 
average chemical concentrations based on estimated variations in stream 
flow in a model system. The PDM3 model compares this exposure 
distribution with an aquatic toxicity test endpoint to estimate how 
many days in a 1-year period the endpoint concentration is exceeded 
(Nabholz et al., 1993; U.S. EPA, 1988b). The frequency of exceedance is 
based on the duration of the toxicity test used to derive the effects 
endpoint. Thus, if the endpoint was an acute toxicity level of concern, 
an exceedance would be identified if the level of concern was exceeded 
for 4 days or more (not necessarily consecutive). The exposure 
estimates are conservative in that they assume instantaneous mixing of 
the chemical in the water column and no losses due to physical, 
chemical, or biodegradation effects.

Text Note 5-4. Comparing Cumulative Exposure and Effects Distributions 
for Chemical Stressors

    Exposure distributions for chemical stressors can be compared with 
effects distributions derived from point estimates of acute or chronic 
toxicity values derived from different species (e.g., HCN, 1993; 
Cardwell et al., 1993; SETAC, 1994a; Solomon et al., 1996). Figure 5-5 
shows a distribution of exposure concentrations of an herbicide 
compared with single-species algal toxicity data for the same chemical. 
The degree of overlap of the curves indicates the likelihood that a 
certain percentage of species may be adversely affected. For example, 
figure 5-5 indicates that the 10th percentile of algal species' 
EC5 values is exceeded less than 10% of the time.
    The predictive value of this approach is evident. The degree of 
risk reduction that could be achieved by changes in exposure associated 
with proposed risk mitigation options can be readily determined by 
comparing modified exposure distributions with the effects distribution 
curve.
    When using effects distributions derived from single-species 
toxicity data, risk assessors should consider the following questions:
     Does the subset of species for which toxicity test data 
are available represent the range of species present in the 
environment?
     Are particularly sensitive (or insensitive) groups of 
organisms represented in the distribution?
     If a criterion level is selected--e.g., protect 95% of 
species--does the 5% of potentially affected species include organisms 
of ecological, commercial, or recreational significance?

Text Note 5-5. Estimating Risk With Process Models

    Models that integrate both exposure and effects information can be 
used to estimate risk. During risk estimation, it is important that 
both the strengths and limitations of a process model approach be 
highlighted. Brody et al. (1993; see Appendix D) linked two process 
models to integrate exposure and effects information and forecast 
spatial and temporal changes in forest communities and their wildlife 
habitat value. While the models were useful for projecting long-term 
effects based on an understanding of the underlying mechanisms of 
change in forest communities and wildlife habitat, they could not 
evaluate all possible stressors of concern and were limited in the 
plant and wildlife species they could consider. Understanding both the 
strengths and limitations of models is essential for accurately 
representing the overall confidence in the assessment.

Text Note 5-6. An Example of Field Methods Used for Risk Estimation

    Along with quotients comparing field measures of exposure with 
laboratory acute toxicity data (text note 5-2), EPA evaluated the risks 
of granular carbofuran to birds based on incidents of bird kills 
following carbofuran applications. Over 40 incidents involving nearly 
30 species of birds were documented. Although reviewers identified 
problems with individual field studies (e.g., lack of appropriate 
control sites, lack of data on carcass-search efficiencies, no 
examination of potential synergistic effects of other pesticides, and 
lack of consideration of other potential receptors such as small 
mammals), there was so much evidence of mortality associated with 
carbofuran application that the study deficiencies did not alter the 
conclusions of high risk found by the assessment (Houseknecht, 1993).

Text Note 5-7. What Are Statistically Significant Effects?

    Statistical testing is the ``statistical procedure or decision rule 
which leads to establishing the truth or falsity of a hypothesis. * * 
*'' (Alder and Roessler, 1972). Statistical significance is based on 
the number of data points, the nature of their distribution, whether 
inter-treatment variance exceeds intra-treatment variance in the data, 
and the a priori significance level (). The types of 
statistical tests and the appropriate protocols (e.g., power of test) 
for these tests should be established as part of the analysis plan 
during problem formulation.

Text Note 5-8. Possible Risk Assessment Report Elements

     Describe risk assessor/risk manager planning results.
     Review the conceptual model and the assessment endpoints.
     Discuss the major data sources and analytical procedures 
used.
     Review the stressor-response and exposure profiles.
     Describe risks to the assessment endpoints, including risk 
estimates and adversity evaluations.
     Review and summarize major areas of uncertainty (as well 
as their direction) and the approaches used to address them.
     Discuss the degree of scientific consensus in key areas 
of uncertainty.
     Identify major data gaps and, where appropriate, 
indicate whether gathering additional data would add significantly to 
the overall confidence in the assessment results.
     Discuss science policy judgments or default assumptions 
used to bridge information gaps, and the basis for these assumptions.

Text Note 5-9. Clear, Transparent, Reasonable, and Consistent Risk 
Characterizations

    For clarity:
     Be brief; avoid jargon.
     Make language and organization understandable to risk 
managers and the informed lay person.
     Fully discuss and explain unusual issues specific to a 
particular risk assessment.
    For transparency:

[[Page 47611]]

     Identify the scientific conclusions separately from policy 
judgments.
     Clearly articulate major differing viewpoints of 
scientific judgments.
     Define and explain the risk assessment purpose (e.g., 
regulatory purpose, policy analysis, priority setting).
     Fully explain assumptions and biases (scientific and 
policy).
    For reasonableness:
     Integrate all components into an overall conclusion of 
risk that is complete, informative, and useful in decision making.
     Acknowledge uncertainties and assumptions in a forthright 
manner.
     Describe key data as experimental, state of the art, or 
generally accepted scientific knowledge.
     Identify reasonable alternatives and conclusions that can 
be derived from the data.
     Define the level of effort (e.g., quick screen, extensive 
characterization) along with the reason(s) for selecting this level of 
effort.
     Explain the status of peer review.
    For consistency with other risk characterizations:
     Describe how the risks posed by one set of stressor(s) 
compare with the risks posed by a similar stressor(s) or similar 
environmental conditions.
     Indicate how the strengths and limitations of the 
assessment compare with past assessments.

Text Note 6-1. Questions Regarding Risk Assessment Results (Adapted 
From U.S. EPA, 1993d)

    Questions principally for risk assessors to ask:
     Are the risks sufficiently well defined (and data gaps 
small enough) to support a risk management decision?
     Was the right problem analyzed?
     Was the problem adequately characterized?
    Questions principally for risk managers to ask:
     What effects might occur?
     How adverse are the effects?
     How likely is it that effects will occur?
     When and where do the effects occur?
     How confident are you in the conclusions of the risk 
assessment?
     What are the critical data gaps, and will information be 
available in the near future to fill these gaps?
     Are more ecological risk assessment iterations required?
     How could monitoring help evaluate results of the risk 
management decision?

Text Note 6-2. Risk Communication Considerations for Risk Managers 
(U.S. EPA, 1995c)

     Plan carefully and evaluate the success of your 
communication efforts.
     Coordinate and collaborate with other credible sources.
     Accept and involve the public as a legitimate partner.
     Listen to the public's specific concerns.
     Be honest, frank, and open.
     Speak clearly and with compassion.
     Meet the needs of the media.

Text Note A-1. Stressor vs. Agent

    Agent has been suggested as an alternative for the term stressor 
(Suter et al., 1994). Agent is thought to be a more neutral term than 
stressor, but agent is also associated with certain classes of 
chemicals (e.g., chemical warfare agents). In addition, agent has the 
connotation of the entity that is initially released from the source, 
whereas stressor has the connotation of the entity that causes the 
response. Agent is used in EPA's Guidelines for Exposure Assessment 
(U.S. EPA, 1992d) (i.e., with exposure defined as ``contact of a 
chemical, physical, or biological agent''). These two terms are 
considered to be nearly synonymous, but stressor is used throughout 
these guidelines for internal consistency.

Appendix A--Changes From EPA'S Ecological Risk Assessment Framework

    EPA has gained much experience with the ecological risk assessment 
process since the publication of the Framework Report (U.S. EPA, 1992a) 
and has received many suggestions for modifications of both the process 
and the terminology. While EPA is not recommending major changes in the 
overall ecological risk assessment process, proposed modifications are 
summarized here to assist those who may already be familiar with the 
Framework Report. Changes in the diagram are discussed first, followed 
by changes in terminology and definitions.

A.1. Changes in the Framework Diagram

    The revised framework diagram is shown in figure 1-2. Within each 
phase, rectangular boxes are used to designate inputs, hexagon-shaped 
boxes indicate actions, and circular boxes represent outputs. There 
have been only minor changes in the wording for the boxes outside of 
the risk assessment process (planning and communications between risk 
assessors and risk managers; acquire data, iterate process, monitor 
results). ``Iterate process'' was added to emphasize the iterative (and 
frequently tiered) nature of risk assessment.
    The new diagram of problem formulation contains several changes. 
The hexagon encloses information about stressors, sources, and 
exposures, ecological effects, and the ecosystem at risk to better 
reflect the importance of integrating this information before selecting 
assessment endpoints and building conceptual models. The three products 
of problem formulation are enclosed in circles. Assessment endpoints 
are shown as a key product that drives conceptual model development. 
The conceptual model remains a central product of problem formulation. 
The analysis plan has been added as an explicit product of problem 
formulation to emphasize the need to plan data evaluation and 
interpretation before analyses begin. It is in the analysis plan that 
measures of ecological effects (measurement endpoints) are identified.
    In the analysis phase, the left-hand side of figure 1-2 shows the 
general process of characterization of exposure, and the right-hand 
side shows the characterization of ecological effects. These two 
aspects of analysis must closely interact to produce compatible output 
that can be integrated in risk characterization. The dotted line and 
hexagon that includes both the exposure and ecological response 
analyses emphasize this interaction. In addition, the first three boxes 
in analysis now include the measures of exposure, effects, and 
ecosystem and receptor characteristics that provide input to the 
exposure and ecological response analyses.
    Experience with the application of risk characterization as 
outlined in the Framework Report suggests the need for several 
modifications in this process. Risk estimation entails the integration 
of exposure and effects estimates along with an analysis of 
uncertainties. The process of risk estimation outlined in the Framework 
Report separates integration and uncertainty. The original purpose for 
this separation was to emphasize the importance of estimating 
uncertainty. This separation is no longer needed since uncertainty 
analysis is now explicitly addressed in most risk integration methods.
    The description of risk is similar to the process described in the 
Framework Report. Topics included in the risk description include the 
lines of evidence that support causality and a determination of the 
ecological adversity of observed or predicted effects. Considerations 
for reporting risk assessment results are also described.

[[Page 47612]]

A.2. Changes in Definitions and Terminology

    Except as noted below, these guidelines retain definitions used in 
the Framework Report (see Appendix B). Some definitions have been 
revised, especially those related to endpoints and exposure. Some 
changes in the classification of uncertainty from the Framework Report 
are also described in this section. It is likely that these terms will 
continue to generate considerable discussion among risk assessors.

A.2.1. Endpoint Terminology

    The Framework Report uses the assessment and measurement endpoint 
terminology of Suter (1990) but offers no specific terms for 
measurements of stressor levels or ecosystem attributes. Experience has 
shown that stressor measurements are sometimes inappropriately called 
measurement endpoints; measurement endpoints should be ``* * * 
measurable responses to a stressor that are related to the valued 
characteristics chosen as assessment endpoints'' (U.S. EPA, 1992a; 
Suter, 1990; emphasis added). These guidelines replace measurement 
endpoint with measure of effect, which is defined as a measurable 
ecological characteristic that is related to the valued characteristic 
chosen as the assessment endpoint (Suter, 1990; U.S. EPA, 1992a). (An 
assessment endpoint is ``an explicit expression of the environmental 
value to be protected'' [U.S. EPA, 1992a].) Since data other than those 
required to evaluate responses (i.e., measures of effects) are required 
for an ecological risk assessment, two additional types of measures are 
used. Measures of exposure include stressor and source measurements, 
while measures of ecosystem and receptor characteristics include, for 
example, habitat measures, soil parameters, water quality conditions, 
or life history parameters that may be necessary to better characterize 
exposure or effects. Any of the three types of measures may be actual 
data (e.g., mortality), summary statistics (e.g., an LC50), or 
estimated values (e.g., an LC50 estimated from a structure-
activity relationship).

A.2.2. Exposure Terminology

    These guidelines define exposure in a manner that is relevant to 
any chemical, physical, or biological entity. While the broad concepts 
are the same, the language and approaches vary depending on whether a 
chemical, physical, or biological entity is the subject of assessment. 
Key exposure-related terms and their definitions are:
     Source. A source is an entity or action that releases to 
the environment or imposes on the environment a chemical, physical, or 
biological stressor or stressors. Sources may include a waste treatment 
plant, a pesticide application, a logging operation, introduction of 
exotic organisms, or a dredging project.
     Stressor. A stressor is any physical, chemical, or 
biological entity that can induce an adverse response. This term is 
used broadly to encompass entities that cause primary effects and those 
primary effects that can cause secondary (i.e., indirect) effects. 
Stressors may be chemical (e.g., toxics or nutrients), physical (e.g., 
dams, fishing nets, or suspended sediments), or biological (e.g., 
exotic or genetically engineered organisms). While risk assessment is 
concerned with the characterization of adverse responses, under some 
circumstances a stressor may be neutral or produce effects that are 
beneficial to certain ecological components (see text note A-1). 
Primary effects may also become stressors. For example, a change in a 
bottomland hardwood plant community affected by rising water levels can 
be thought of as a stressor influencing the wildlife community. 
Stressors may also be formed through abiotic interactions; for example, 
the increase in ultraviolet light reaching the earth's surface results 
from the interaction of the original stressors released 
(chlorofluorocarbons) with the ecosystem (stratospheric ozone).
     Exposure. As discussed above, these guidelines use the 
term exposure broadly after the common definition of expose: ``to 
submit or subject to an action or influence'' (Merriam-Webster, 1972). 
Used in this way, exposure applies to physical and biological stressors 
as well as to chemicals (organisms are commonly said to be exposed to 
radiation, pathogens, or heat). Exposure is also applicable to higher 
levels of biological organization, such as exposure of a benthic 
community to dredging, exposure of an owl population to habitat 
modification, or exposure of a wildlife population to hunting. Although 
the operational definition of exposure, particularly the units of 
measure, depends on the stressor and receptor (defined below), the 
following general definition is applicable: Exposure is the contact or 
co-occurrence of a stressor with a receptor.
     Receptor. The receptor is the ecological component exposed 
to the stressor. This term may refer to tissues, organisms, 
populations, communities, and ecosystems. While either ``ecological 
component'' (U.S. EPA, 1992a) or ``biological system'' (Cohrssen and 
Covello, 1989) are alternative terms, ``receptor'' is usually clearer 
in discussions of exposure where the emphasis is on the stressor-
receptor relationship. As discussed below, both disturbance and stress 
regime have been suggested as alternative terms for exposure. Neither 
term is used in these guidelines, which instead use exposure as broadly 
defined above.
     Disturbance. A disturbance is any event or series of 
events that disrupts ecosystem, community, or population structure and 
changes resources, substrate availability, or the physical environment 
(modified slightly from White and Pickett, 1985). Defined in this way, 
disturbance is clearly a kind of exposure (i.e., an event that subjects 
a receptor, the disturbed system, to the actions of a stressor). 
Disturbance may be a useful alternative to stressor specifically for 
physical stressors that are deletions or modifications (e.g., logging, 
dredging, flooding).
     Stress Regime. The term stress regime has been used in at 
least three distinct ways: (1) To characterize exposure to multiple 
chemicals or to both chemical and nonchemical stressors (more clearly 
described as multiple exposure, complex exposure, or exposure to 
mixtures), (2) as a synonym for exposure that is intended to avoid 
overemphasis on chemical exposures, and (3) to describe the series of 
interactions of exposures and effects resulting in secondary exposures, 
secondary effects, and, finally, ultimate effects (also known as risk 
cascade [Lipton et al., 1993]) or causal chain, pathway, or network 
(Andrewartha and Birch, 1984). Because of the potential for confusion 
and the availability of other clearer terms, this term is not used in 
these guidelines.

A.2.3. Uncertainty Terminology

    The Framework Report divided uncertainty into conceptual model 
formation, information and data, stochasticity, and error. These 
guidelines discuss uncertainty throughout the process, focusing on the 
conceptual model (section 3.4.3), the analysis phase (section 4.1.3), 
and the incorporation of uncertainty in risk estimates (section 5.1). 
The bulk of the discussion appears in section 4.1.3, where the 
discussion is organized according to the following sources of 
uncertainty:
     Unclear communication.
     Descriptive errors.
     Variability.
     Data gaps.
     Uncertainty about a quantity's true value.

[[Page 47613]]

     Model structure uncertainty (process models).
     Uncertainty about a model's form (empirical models).

Appendix B.--Key Terms (Adapted From U.S. EPA, 1992a)

    Agent--Any physical, chemical, or biological entity that can induce 
an adverse response (synonymous with stressor).
    Assessment endpoint--An explicit expression of the environmental 
value that is to be protected. An assessment endpoint includes both an 
ecological entity and specific attributes of that entity. For example, 
salmon are a valued ecological entity; reproduction and population 
maintenance of salmon form an assessment endpoint.
    Characterization of ecological effects--A portion of the analysis 
phase of ecological risk assessment that evaluates the ability of a 
stressor to cause adverse effects under a particular set of 
circumstances.
    Characterization of exposure--A portion of the analysis phase of 
ecological risk assessment that evaluates the interaction of the 
stressor with one or more ecological entities. Exposure can be 
expressed as co-occurrence or contact, depending on the stressor and 
ecological component involved.
    Community--An assemblage of populations of different species within 
a specified location in space and time.
    Comparative risk assessment--A process that generally uses an 
expert judgment approach to evaluate the relative magnitude of effects 
and set priorities among a wide range of environmental problems (e.g., 
U.S. EPA, 1993b). Some applications of this process are similar to the 
problem formulation portion of an ecological risk assessment in that 
the outcome may help select topics for further evaluation and help 
focus limited resources on areas having the greatest risk reduction 
potential. In other situations, a comparative risk assessment is 
conducted more like a preliminary risk assessment. For example, EPA's 
Science Advisory Board used expert judgment and an ecological risk 
assessment approach to analyze future ecological risk scenarios and 
risk management alternatives (U.S. EPA, 1995a).
    Conceptual model--The conceptual model describes a series of 
working hypotheses of how the stressor might affect ecological 
entities. The conceptual model also describes the ecosystem potentially 
at risk, the relationship between measures of effect and assessment 
endpoints, and exposure scenarios.
    Cumulative distribution function (CDF)--Cumulative distribution 
functions are particularly useful for describing the likelihood that a 
variable will fall within different ranges of x. F(x) (i.e., the value 
of y at x in a CDF plot) is the probability that a variable will have a 
value less than or equal to x (figure B-1).

BILLING CODE 6560-50-P
[GRAPHIC] [TIFF OMITTED] TN09SE96.011



[[Page 47614]]


BILLING CODE 6560-50-C
    Cumulative ecological risk assessment--A process that involves 
consideration of ``the aggregate ecologic risk to the target entity 
caused by the accumulation of risk from multiple stressors'' (Bender, 
1996).
    Disturbance--Any event or series of events that disrupts ecosystem, 
community, or population structure and changes resources, substrate 
availability, or the physical environment (modified from White and 
Pickett, 1985).
    Ecological entity--A general term that may refer to a species, a 
group of species, an ecosystem function or characteristic, or a 
specific habitat. An ecological entity can be one component of an 
assessment endpoint.
    Ecological risk assessment--The process that evaluates the 
likelihood that adverse ecological effects may occur or are occurring 
as a result of exposure to one or more stressors.
    Ecosystem--The biotic community and abiotic environment within a 
specified location in space and time.
    Environmental impact statement--Assessments are required under the 
National Environmental Policy Act (NEPA) to fully evaluate 
environmental effects associated with proposed major Federal actions. 
Like ecological risk assessments, environmental impact statements (EIS) 
typically require a ``scoping process'' analogous to problem 
formulation, an analysis by multidisciplinary teams, and a presentation 
of uncertainties (CEQ, 1986, cited in Suter, 1993a). By virtue of 
special expertise, EPA may cooperate with other agencies by preparing 
EISs or otherwise participating in the NEPA process.
    Exposure--The contact or co-occurrence of a stressor with a 
receptor.
    Exposure profile--The product of characterization of exposure in 
the analysis phase of ecological risk assessment. The exposure profile 
summarizes the magnitude and spatial and temporal patterns of exposure 
for the scenarios described in the conceptual model.
    Exposure scenario--A set of assumptions concerning how an exposure 
may take place, including assumptions about the exposure setting, 
stressor characteristics, and activities that may lead to exposure.
    Hazard assessment--This term has been used to mean either (1) 
evaluating the intrinsic effects of a stressor (U.S. EPA, 1979) or (2) 
defining a margin of safety or quotient by comparing a toxicologic 
effects concentration with an exposure estimate (SETAC, 1987).
    Lines of evidence--Information derived from different sources or by 
different techniques that can be used to interpret and compare risk 
estimates. While this term is similar to the term ``weight of 
evidence,'' it does not necessarily imply assignment of quantitative 
weightings to information.
    Lowest observed adverse effect level (LOAEL)--The lowest level of a 
stressor evaluated in a test that causes statistically significant 
differences from the controls.
    Maximum acceptable toxic concentration (MATC)--For a particular 
ecological effects test, this term is used to mean either the range 
between the NOAEL and the LOAEL or the geometric mean of the NOAEL and 
the LOAEL for a particular test. The geometric mean is also known as 
the chronic value.
    Measure of ecosystem and receptor characteristics--A measurable 
characteristic of the ecosystem or receptor that is used in support of 
exposure or effects analysis.
    Measure of effect--A measurable ecological characteristic that is 
related to the valued characteristic chosen as the assessment endpoint.
    Measure of exposure--A measurable stressor characteristic that is 
used to help quantify exposure.
    Measurement endpoint--See ``measure of effect.''
    Median lethal concentration (LC50)--A statistically or 
graphically estimated concentration that is expected to be lethal to 
50% of a group of organisms under specified conditions (ASTM, 1990).
    No observed adverse effect level (NOAEL)--The highest level of a 
stressor evaluated in a test that does not cause statistically 
significant differences from the controls.
    Population--An aggregate of individuals of a species within a 
specified location in space and time.
    Primary effect--An effect where the stressor acts on the ecological 
component of interest itself, not through effects on other components 
of the ecosystem (synonymous with direct effect; compare with 
definition for secondary effect).
    Probability density function (PDF)--Probability density functions 
are particularly useful in describing the relative likelihood that a 
variable will have different particular values of x. The probability 
that a variable will have a value within a small interval around x can 
be approximated by multiplying f(x) (i.e., the value of y at x in a PDF 
plot) by the width of the interval (figure B-2).

BILLING CODE 6560-50-P

[[Page 47615]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.012



BILLING CODE 6560-50-C
    Receptor--The ecological entity exposed to the stressor.
    Recovery--The rate and extent of return of a population or 
community to a condition that existed before the introduction of a 
stressor. Due to the dynamic nature of ecological systems, the 
attributes of a ``recovered'' system must be carefully defined.
    Relative risk assessment--A process similar to comparative risk 
assessment. It involves estimating the risks associated with different 
stressors or management actions. To some, relative risk connotes the 
use of quantitative risk techniques, while comparative risk approaches 
more often rely on expert judgment. Others do not make this 
distinction.
    Risk characterization--A phase of ecological risk assessment that 
integrates the exposure and stressor response profiles to evaluate the 
likelihood of adverse ecological effects associated with exposure to a 
stressor. The adversity of effects is discussed, including 
consideration of the nature and intensity of the effects, the spatial 
and temporal scales, and the potential for recovery.
    Secondary effect--An effect where the stressor acts on supporting 
components of the ecosystem, which in turn have an effect on the 
ecological component of interest (synonymous with indirect effects; 
compare with definition for primary effect).
    Source--An entity or action that releases to the environment or 
imposes on the environment a chemical, physical, or biological stressor 
or stressors.
    Source term--As applied to chemical stressors, the type, magnitude, 
and patterns of chemical(s) released.
    Stress regime--The term stress regime has been used in at least 
three distinct ways: (1) to characterize exposure to multiple chemicals 
or to both chemical and nonchemical stressors (more clearly described 
as multiple exposure, complex exposure, or exposure to mixtures), (2) 
as a synonym for exposure that is intended to avoid overemphasis on 
chemical exposures, and (3) to describe the series of interactions of 
exposures and effects resulting in secondary exposures, secondary 
effects, and, finally, ultimate effects (also known as risk cascade 
[Lipton et al., 1993]) or causal chain, pathway, or network 
(Andrewartha and Birch, 1984).
    Stressor--Any physical, chemical, or biological entity that can 
induce an adverse response (synonymous with agent).
    Stressor-response profile--The product of characterization of 
ecological effects in the analysis phase of ecological risk assessment. 
The stressor-response profile summarizes the data on the effects of a 
stressor and the relationship of the data to the assessment endpoint.
    Trophic levels--A functional classification of taxa within a 
community that is based on feeding relationships (e.g., aquatic and 
terrestrial green plants comprise the first trophic level and 
herbivores comprise the second).

Appendix C.--Conceptual Model Examples

    Conceptual model diagrams are visual representations of the 
conceptual

[[Page 47616]]

models. They may be based on theory and logic, empirical data, 
mathematical models, and probability models. These diagrams are useful 
tools for communicating important pathways in a clear and concise way. 
They can be used to ask new questions about relationships that help 
generate plausible risk hypotheses. Further discussion of conceptual 
models is found in section 3-4.
    Flow diagrams like those shown in figures C-1 through C-3 are 
typical conceptual model diagrams. When constructing flow diagrams like 
these, it is helpful to use distinct and consistent shapes to 
distinguish among stressors, assessment endpoints, responses, exposure 
routes, and ecosystem processes. Although flow diagrams are often used 
to illustrate conceptual models, there is no set configuration for 
conceptual model diagrams. Pictorial representations of the processes 
of an ecosystem can be more effective (e.g., Bradley and Smith, 1989).

BILLING CODE 6560-50-P
      

[[Page 47617]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.013



[[Page 47618]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.014



[[Page 47619]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.015



[[Page 47620]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.016



BILLING CODE 6560-50-C

[[Page 47621]]

    Figure C-1 illustrates the relationship between a primary physical 
stressor (logging roads) and an effect on an assessment endpoint 
(fecundity in insectivorous fish). This simple diagram illustrates that 
building logging roads (which could be considered a stressor or a 
source) in ecosystems where slope, soil type, low riparian cover, and 
other ecosystem characteristics lead to the erosion of soil, which 
enters streams and smothers the benthic organisms (exposure pathway is 
not explicit in this diagram). Because of the dependence of 
insectivorous fish on benthic organisms, the fish are believed to be at 
risk from the building of logging roads. Each arrow in this diagram 
represents a hypothesis about the proposed relationship (e.g., human 
action and stressor, stressor and effect, primary effect to secondary 
effect). Each risk hypothesis provides insights into the kinds of data 
that will be needed to verify that the hypothesized relationships are 
valid.
    Figure C-2 is a conceptual model used by Kendall et al. (1996) to 
track a contaminant through upland ecosystems. In this example, upland 
birds are exposed to lead shot when it becomes embedded in their tissue 
after being shot and by ingesting lead accidentally when feeding on the 
ground. Both are hypothesized to result in increased morbidity (e.g., 
lower reproduction and competitiveness and higher predation and 
infection) and mortality, either directly (lethal intoxication) or 
indirectly (effects of morbidity leading to mortality). These effects 
are believed to result in changes in upland bird populations and, due 
to hypothesized exposure of predators to lead, to increase predator 
mortality. This example shows multiple exposure pathways for effects on 
two assessment endpoints. Each arrow contains within it assumptions and 
hypotheses about the relationship depicted that provide the basis for 
identifying data needs and analyses.
    Figure C-3 is a conceptual model adapted from the Waquoit Bay 
watershed risk assessment. At the top of the model, multiple human 
activities that occur in the watershed are shown in rectangles. Those 
sources of stressors are linked to stressor types depicted in ovals. 
Multiple sources are shown to contribute to an individual stressor, and 
each source may contribute to more than one stressor. The stressors 
then lead to multiple ecological effects depicted again in rectangles. 
Some rectangles are double-lined to indicate effects that can be 
directly measured for data analysis. Finally, the effects are linked to 
particular assessment endpoints. The connections show that one effect 
can result in changes in many assessment endpoints. To fully depict 
exposure pathways and types of effects, specific portions of this 
conceptual model would need to be expanded to illustrate those 
relationships.

Appendix D.--Analysis Phase Examples

    The analysis phase process is illustrated here for a chemical, 
physical, and biological stressor. These examples do not represent all 
possible approaches but illustrate the analysis phase process using 
information from actual assessments.

D.1. Special Review of Granular Formulations of Carbofuran Based on 
Adverse Effects on Birds

    Figure D-1 is based on an assessment of the risks of carbofuran to 
birds under the Federal Insecticide, Fungicide, and Rodenticide Act 
(FIFRA) (Houseknecht, 1993). Carbofuran is a broad-spectrum insecticide 
and nematicide applied primarily in granular form on 27 crops as well 
as forests and pineseed orchards. The assessment endpoint was survival 
of birds that forage in agricultural areas where carbofuran is applied.

BILLING CODE 6560-50-P

[[Page 47622]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.017



BILLING CODE 6560-50-C

[[Page 47623]]

    The analysis phase focused on birds that may incidentally ingest 
granules as they forage or that may eat other animals that contain 
granules or residues. Measures of exposure included application rates, 
attributes of the formulation (e.g., size of granules), and residues in 
prey organisms. Measures of the ecosystem and receptors included an 
inventory of bird species that may be exposed following applications 
for 10 crops. The birds' respective feeding behaviors were considered 
in developing routes of exposure. Measures of effect included 
laboratory toxicity studies and field investigations of bird mortality.
    The source of the chemical was application of the pesticide in 
granular form. The distribution of the pesticide in agricultural fields 
was estimated based on the application rate. The number of exposed 
granules was estimated from literature data. Based on a review of avian 
feeding behavior, seed-eating birds were assumed to ingest any granules 
left uncovered in the field. The intensity of exposure was summarized 
as the number of exposed granules per square foot.
    The stressor-response relationship was described using the results 
of toxicity tests. These data were used to construct a toxicity 
statistic expressed as the number of granules needed to kill 50% of the 
test birds (i.e., granules per LD50), assuming 0.6 mg of active 
ingredient (AI) per granule and average body weights for the birds 
tested. Field studies were used to document the occurrence of bird 
deaths following applications and provide further causal evidence. 
Carbofuran residues and cholinesterase levels were used to confirm that 
exposure to carbofuran caused the deaths.

D.2. Modeling Losses of Bottomland-Forest Wetlands

    Figure D-2 is based on an assessment of the ecological consequences 
(risks) of long-term changes in hydrologic conditions (water-level 
elevations) for three habitat types in the Lake Verret Basin of 
Louisiana (Brody et al., 1989, 1993; Connor and Brody, 1989). The 
project was intended to provide a habitat-based approach for assessing 
the environmental impacts of Federal water projects under the National 
Environmental Policy Act and Section 404 of the Clean Water Act. Output 
from the models provided risk managers with information on how changes 
in water elevation might alter the ecosystem. The primary anthropogenic 
stressor addressed in this assessment was artificial levee construction 
for flood control, which contributes to land subsidence by reducing 
sediment deposition in the floodplain. Assessment endpoints included 
forest community structure and habitat value to wildlife species and 
the species composition of the wildlife community.

BILLING CODE 6560-60-P

[[Page 47624]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.018



BILLING CODE 6560-50-C

[[Page 47625]]

    The analysis phase began by considering primary (direct) effects of 
water-level changes on plant community composition and habitat 
characteristics. Measures of exposure included the attributes and 
placement of the levees and water-level measurements. Ecosystem and 
receptor measures included location and extent of bottomland-hardwood 
communities, plant species occurrences within these communities, and 
information on the historic flow regimes. Effects measures included 
laboratory studies of plant response to moisture and field measurements 
along moisture gradients.
    While the principal stressor under evaluation was the construction 
of levees, the decreased gradient of the river due to sediment 
deposition at its mouth also contributed to increased water levels. The 
extent and frequency of flooding were simulated by the FORFLO model 
based on estimates of net subsidence rates from levee construction and 
decreased river gradient. Seeds and seedlings of the tree species were 
assumed to be exposed to the altered flooding regime. Stressor-response 
relationships describing plant response to moisture (e.g., seed 
germination, survival) were embedded within the FORFLO model. This 
information was used by the model to simulate changes in plant 
communities: The model tracks the species type, diameter, and age of 
each tree on simulated plots from the time the tree enters the plot as 
a seedling or sprout until it dies. The FORFLO model calculated changes 
in the plant community over time (from 50 to 280 years). The spatial 
extent of the three habitat types of interest--wet bottomland 
hardwoods, dry bottomland hardwoods, and cypress-tupelo swamp--was 
mapped onto a Geographic Information System (GIS) along with the 
hydrological information. Then the changes projected by FORFLO were 
manually linked to the GIS to show how the spatial distribution of 
different communities would change. Evidence that flooding would 
actually cause these changes included comparisons of model predictions 
with field measurements, the laboratory studies of plant response to 
moisture, and knowledge of the mechanisms by which flooding elicits 
changes in plant communities.
    Secondary (indirect) effects on wildlife associated with changes in 
the habitat provided by the plant community formed the second part of 
the analysis phase. Important measures included life-history 
characteristics and habitat needs of the wildlife species. Effects on 
wildlife were inferred by evaluating the suitability of the plant 
community as habitat. Specific aspects of the community structures 
calculated by the FORFLO model provided the input to this part of the 
analysis. For example, the number of snags was used to evaluate habitat 
value for woodpeckers. Resident wildlife (represented by five species) 
were assumed to co-occur with the altered plant community. Habitat 
value was evaluated by calculating the Habitat Suitability Index (HSI) 
for each habitat type multiplied by the habitat type's area.
    A combined exposure and stressor-response profile is shown in 
figure D-2; these two elements were combined with the models used for 
the analysis and then used directly in risk characterization.

D.3. Pest Risk Assessment of Importation of Logs From Chile

    Figure D-3 is based on the assessment of potential risks to U.S. 
forests due to the incidental introduction of insects, fungi, and other 
pests inhabiting logs harvested in Chile and transported to U.S. ports 
(USDA, 1993). This risk assessment was used to determine whether 
actions to restrict or regulate the importation of Chilean logs were 
needed to protect U.S. forests and was conducted by a team of six 
experts under the auspices of the U.S. Department of Agriculture Forest 
Service. Stressors include insects, forest pathogens (e.g., fungi), and 
other pests. The assessment endpoint was the survival and growth of 
tree species (particularly conifers) in the western United States. 
Damage that would affect the commercial value of the trees as lumber 
was clearly of interest.

BILLING CODE 6560-50-P

[[Page 47626]]

[GRAPHIC] [TIFF OMITTED] TN09SE96.019



BILLING CODE 6560-50-C

[[Page 47627]]

    The analysis phase was carried out by eliciting professional 
opinions from a team of experts. Exposure measures used by the team 
included distribution information for the imported logs and attributes 
of the insects and pathogens such as dispersal mechanisms and life 
history characteristics. Ecosystem and receptor measures included the 
climate of the United States, location of geographic barriers, 
knowledge of host suitability, and ranges of potential host species. 
Effect measures included knowledge of the infectivity of these pests in 
other countries and the infectivity of similar pests on U.S. hosts.
    This information was used by the risk assessment team to evaluate 
the potential for exposure. They began by evaluating the likelihood of 
entry of infested logs into the United States. The distribution of the 
organisms given entry was evaluated by considering the potential for 
colonization and spread beyond the point of entry as well as the 
likelihood of organisms surviving and reproducing. The potential for 
exposure was summarized by assigning each of the above elements a 
judgment-based value of high, medium, or low.
    The evaluation of ecological effects was also conducted based on 
collective professional judgment. Of greatest relevance to this 
guidance was the consideration of environmental damage potential, 
defined as the likelihood of ecosystem destabilization, reduction in 
biodiversity, loss of keystone species, and reduction or elimination of 
endangered or threatened species. (The team also considered economic 
damage potential and social and political influences; however, these 
guidelines consider those factors to be part of the risk management 
process.) Again, each consideration was assigned a value of high, 
medium, or low to summarize the potential for ecological effects.

Appendix E.--Criteria for Determining Ecological Adversity: A 
Hypothetical Example (Adapted from Harwell et al., 1994)

    As a result of a collision at sea, an oil tanker releases 15 
million barrels of #2 fuel oil 3 km offshore. It is predicted that 
prevailing winds will carry the fuel onshore within 48 to 72 hours. The 
coastline has numerous small embayments that support an extensive 
shallow, sloping subtidal community and a rich intertidal community. A 
preliminary assessment determined that if no action were taken, 
significant risks to the communities would result. Additional risk 
assessments were conducted to determine which of two options should be 
used to clean up the oil spill.
    Option 1 is to use a dispersant to break up the slick, which would 
reduce the likelihood of extensive onshore contamination but would 
cause extensive mortality to the phytoplankton, zooplankton, and 
ichthyoplankton, which are important for commercial fisheries. Option 2 
is to try to contain and pump off as much oil as possible; this option 
anticipates that a shift in wind direction will move the spill away 
from shore and allow for natural dispersal at sea. If this does not 
happen, the oil will contaminate the extensive sub- and intertidal mud 
flats, rocky intertidal communities, and beaches and pose an additional 
hazard to avian and mammalian fauna. It is assumed there will be a 
demonstrable change beyond natural variability in the assessment 
endpoints (e.g., structure of planktonic, benthic, and intertidal 
communities). What is the adversity of each option?
     Nature and severity of the effect. For both options, the 
magnitude of change in the assessment endpoints is likely to be severe. 
Planktonic populations often are characterized by extensive spatial and 
temporal variability. Nevertheless, within the spatial boundaries of 
the spill, the use of dispersants is likely to produce complete 
mortality of all planktonic forms within the upper 3 m of water. For 
benthic and intertidal communities that generally are stable and have 
less spatial and temporal variability than planktonic forms, oil 
contamination will likely result in severe impacts on survival and 
chronic effects lasting for several years. Thus, under both options, 
changes in the assessment endpoints will probably exceed the natural 
variability for threatened communities in both space and time.
     Spatial scale. The areal extent of impacts is similar for 
each of the options. While extensive, the area of impact constitutes a 
small percentage of the landscape. This leaves considerable area 
available for replacement stocks and creates significant fragmentation 
of either the planktonic or inter- and subtidal habitats. Ecological 
adversity is reduced because the area is not a mammalian or avian 
migratory corridor.
     Temporal scale and recovery. Based on experience with 
other oil spills, it is assumed that the effects are reversible over 
some time period. The time needed for reversibility of changes in 
phytoplankton and zooplankton populations should be short (days to 
weeks) given their rapid generation times and easy immigration from 
adjacent water masses. Similarly, although ichthyoplankton do not 
reproduce, they typically experience extensive natural mortality, and 
immigration is readily available from surrounding water masses. On the 
other hand, the time needed for reversibility of changes in benthic and 
intertidal communities is likely to be long (years to decades). First, 
the stressor (oil) would be likely to persist in sediments and on rocks 
for several months to years. Second, the life histories of the species 
comprising these communities span 3 to 5 years. Third, the 
reestablishment of benthic intertidal community and ecosystem structure 
(hierarchical composition and function) often requires decades.
    Both options result in (1) assessment endpoint effects that are of 
great severity, (2) exceedances of natural variability for those 
endpoints, and (3) similar estimates of areal impact. What 
distinguishes the two options is temporal scale and reversibility. In 
this regard, changes to the benthic and intertidal ecosystems are 
considerably more adverse than those to the plankton. On this basis, 
the option of choice would be to disperse the oil, effectively 
preventing it from reaching shore where it would contaminate the 
benthic and intertidal communities.

References

    Adams, D.F. (1963) Recognition of the effects of fluorides on 
vegetation. J. Air Pollut. Control Assoc. 13: 360-362. Cited in: 
Suter, G.W., II. (1993) Ecological risk assessment. Boca Raton, FL: 
Lewis Publishers.
    Alder, H.L.; Roessler, E.B. (1972) Introduction to probability 
and statistics. San Francisco, CA: W.H. Freeman and Co.
    American Society for Testing and Materials (ASTM). (1990). 
Standard terminology relating to biological effects and 
environmental fate. E943-90. In: ASTM; 1990 Annual Book of ASTM 
Standards, Section 11, Water and Environmental Technology, ASTM, 
Philadelphia, PA.
    American Society for Quality Control (ASQC) (1994) American 
National Standard: Specifications and Guidelines for Quality Systems 
for Environmental Data Collection and Environmental Technology 
Programs. ANSI/ASQC E4-1994. Milwaukee, WI: American Society for 
Quality Control.
    Andrewartha, H.G.; Birch, L.C. (1984) The ecological web. 
Chicago, IL: University of Chicago Press.
    Ankley, G.T.; Katko, A.; Arthur, J.W. (1990) Identification of 
ammonia as an important sediment-associated toxicant in the Lower 
Fox River and Green Bay, Wisconsin. Environ. Toxicol. Chem. 9: 313-
322.
    Auer, C.M.; Zeeman, M.; Nabholz, J.V.; Clements, R.G. (1994) 
SAR--the U.S. regulatory perspective. SAR and QSAR Environ. Res. 
2:29-38
    Barbour, M.T.; Stribling, J.B.; Karr, J.R. (1995) Multimetric 
approach for establishing biocriteria and measuring biological 
condition. In: Davis, W.S.; Simon, T.P., eds. Biological assessment 
and criteria, tools for

[[Page 47628]]

water resource planning and decision making. Boca Raton, FL: Lewis 
Publishers; pp. 63-77.
    Barnthouse, L.W.; Brown, J. (1994) Issue paper on conceptual 
model development. In: Ecological risk assessment issue papers. 
Washington, DC: Risk Assessment Forum, U.S. Environmental Protection 
Agency; pp. 3-1 to 3-70. EPA/630/R-94/009.
    Barnthouse, L.W.; O'Neill, R.V.; Bartell, S.M.; Suter, G.W., II. 
(1986) Population and ecosystem theory in ecological risk 
assessment. In: Poston, T.M.; Purdy, R., eds. Aquatic ecology and 
hazard assessment, 9th symposium. Philadelphia, PA: American Society 
for Testing and Materials; pp. 82-96.
    Barnthouse, L.W.; Suter, G.W., II; Rosen, A.E.; Beauchamp, J.J. 
(1987) Estimating responses of fish populations to toxic 
contaminants. Environ. Toxicol. Chem. 6: 811-824.
    Barnthouse, L.W.; Suter, G.W., II; Rosen, A.E. (1990) Risks of 
toxic contaminants to exploited fish populations: influence of life 
history, data uncertainty, and exploitation intensity. Environ. 
Toxicol. Chem. 9: 297-311.
    Bartell, S.M.; Gardner, R.H.; O'Neill, R.V. (1992) Ecological 
risk estimation. Boca Raton, FL: Lewis Publishers.
    Bedford, B.L.; Preston, E.M. (1988) Developing the scientific 
basis for assessing cumulative effects of wetland loss and 
degradation on landscape functions: status, perspectives, and 
prospects. Environ. Manag. 12: 751-772.
    Bender, E. (1996) Personal communication. Science Policy 
Council, U.S. Environmental Protection Agency, Washington, DC.
    Bradbury, S.P. (1994) Predicting modes of toxic action from 
chemical structure: an overview. SAR and QSAR Environ. Res. 2:29-38.
    Bradley, C.E.; Smith, D.G. (1989) Plains cottonwood recruitment 
and survival on a prairie meandering river floodplain. Milk River, 
southern Alberta in Northern Canada. Can. J. Botany 64: 1433-1442.
    Broderius, S.J. (1991) Modeling the joint toxicity of 
xenobiotics to aquatic organisms: basic concepts and approaches. In: 
Mayes, M.A.; Barron, M.G., eds. Aquatic toxicology and risk 
assessment: fourteenth volume. Philadelphia, Pa: American Society 
for Testing and Materials; pp. 107-127.ASTM STP 1124.
    Broderius, S.J.; Kahl, M.D.; Hoglund, M.D. (1995) Use of joint 
toxic response to define the primary mode of toxic action for 
diverse industrial organic chemicals. Environ. Toxicol. Chem. 9: 
1591-1605.
    Brody, M.; Conner, W.; Pearlstine, L.; Kitchens, W. (1989) 
Modeling bottomland forest, and wildlife habitat changes in 
Louisiana's Atchafalaya Basin. In: Sharitz, R.R.; Gibbons, J.W., 
eds. Freshwater wetlands and wildlife. Oak Ridge, TN: Office of 
Science and Technical Information, U.S. Department of Energy, U.S. 
Department of Energy Symposium Series, No. 61. CONF-8603100.
    Brody, M.S.; Troyer, M.E.; Valette, Y. (1993) Ecological risk 
assessment case study: modeling future losses of bottomland forest 
wetlands and changes in wildlife habitat within a Louisiana basin. 
In: A review of ecological assessment case studies from a risk 
assessment perspective. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency; pp. 12-1 to 12-39. EPA/630/R-92/
005.
    Burmaster, D.E.; Anderson, P.D. (1994) Principles of good 
practice for the use of Monte Carlo techniques in human health and 
ecological risk assessments. Risk Anal. 14: 477-481.
    Burmaster, D.E.; Menzie, C.A.; Freshman, J.S.; Burris, J.A.; 
Maxwell, N.I.; Drew, S.R. (1991) Assessment of methods for 
estimating aquatic hazards at Superfund-type sites: a cautionary 
tale. Environ. Toxicol. Chem. 10:827-842.
    Callahan, C.A.; Menzie, C.A.; Burmaster, D.E.; Wilborn, D.C.; 
Ernst, T. (1991) On-site methods for assessing chemical impacts on 
the soil environment using earthworms: A case study at the Baird and 
McGuire Superfund site, Holbrook, Massachusetts. Environ. Toxicol. 
Chem. 10:817-826.
    Cardwell, R.D.; Parkhurst, B.R.; Warren-Hicks, W.; Volosin, J.S. 
(1993) Aquatic ecological risk. Water Environ. Technol. 5: 47-51.
    Clements, R.G.; Nabholz, J.V. (1994) ECOSAR: a computer program 
for estimating the ecotoxicity of industrial chemicals based on 
structure activity relationships, user's guide. Washington, DC: 
Environmental Effects Branch, Health and Environmental Review 
Division (7402), U.S. Environmental Protection Agency; EPA/748/R-93/
002.
    Cohrssen, J.; Covello, V.T. (1989) Risk analysis: a guide to 
principles and methods for analyzing health and ecological risks. 
Washington, DC: Council on Environmental Quality.
    Conner, W.H.; Brody, M. (1989) Rising water levels and the 
future of southeastern Louisiana swamp forests. Estuaries 12(4):318-
323.
    Council on Environmental Quality (CEQ). (1986) Regulations for 
implementing the procedural provisions of the National Environmental 
Policy Act. 40 CFR Parts 1500-1508. Washington, DC: U.S. Government 
Printing Office. Cited in: Suter, G.W., II. (1993) Ecological risk 
assessment. Boca Raton, FL: Lewis Publishers.
    Cowan, C.E.; Versteeg, D.J.; Larson, R.J.; Kloepper-Sams, P.J. 
(1995) Integrated approach for environmental assessment of new and 
existing substances. Regul. Toxicol. Pharmacol. 21: 3-31.
    Cressie, N.A.C. (1993) Statistics for spatial data. New York, 
NY: John Wiley.
    Cronin, M.T.D.; Dearden, J.C. (2995) QSAR in toxicology. 1. 
Prediction of Aquatic toxicity. Quant. Struct. Relat. 14:1-7.
    Detenbeck, N. (1994) Ecological risk assessment case study: 
effects of physical disturbance on water quality status and water 
quality improvement function of urban wetlands. In: A review of 
ecological assessment case studies from a risk assessment 
perspective, v. II. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency; pp. 4-1 to 4-58. EPA/630/R-94/003.
    Detenbeck, N.E.; DeVore, P.W.; Niemi, G.J.; Lima, A. (1992) 
Recovery of temperate-stream fish communities from disturbance: a 
review of case studies and synthesis of theory. Environ. Manag. 
16(1): 33-53.
    Eberhardt, L.L.; Thomas, J.M. (1991) Designing environmental 
field studies. Ecol. Mono. 61(1): 53-73.
    Emlen, J.M. (1989) Terrestrial population models for ecological 
risk assessment: a state-of-the-art review. Environ. Toxicol. Chem. 
8: 831-842.
    European Community. (1993) Technical guidance document in 
support of the risk assessment directive (93/67/EEC) for new 
substances notified in accordance with the requirements of council 
directive 67/548/EEC. Brussels, Belgium.
    Ferson, S.; Ginzburg, L.; Silvers, A. (1989) Extreme event risk 
analysis for age-structured populations. Ecol. Model. 47: 175-187.
    Fisher, S.G.; Woodmansee, R. (1994) Issue paper on ecological 
recovery. In: Ecological risk assessment issue papers. Washington, 
DC: Risk Assessment Forum, U.S. Environmental Protection Agency; pp. 
7-1 to 7-54. EPA/630/R-94/009.
    Fox, G.A. (1991) Practical causal inference for 
ecoepidemiologists. J. Toxicol. Environ. Health 33: 359-373.
    Gauch, H.G. (1982) Multivariate anlaysis in community ecology. 
Cambridge, MA: Cambridge University Press.
    Gaudet, C. (1994) A framework for ecological risk assessment at 
contaminated sites in Canada: review and recommendations. Ottawa, 
Canada: Environment Canada.
    Gibbs, J.P. (1993) Importance of small-wetlands for the 
persistence of local populations of wetland-associated animals. 
Wetlands 13(1): 25-31.
    Gilbert, R.O. (1987) Statistical methods for environmental 
pollution monitoring. New York, NY: Van Nostrand Reinhold.
    Ginzburg, L.R.; Slobodkin, L.B.; Johnson, K.; Bindman, A.G. 
(1982) Quasiextinction probabilities as a measure of impact on 
population growth. Risk Anal. 2: 171-182.
    Gosselink, J.G.; Shaffer, G.P.; Lee, L.C.; Burdick, D.L.; 
Childer, N.C.; Leibowitz, N.C.; Hamilton, S.C.; Boumans, R.; 
Cushmam, D.; Fields, S.; Koch, M.; Visser, J.M. (1990) Landscape 
conservation in a forested wetland watershed: can we manage 
cumulative impacts? Bioscience 40(8): 588-600.
    Haimes, Y.Y.; Barry, T.; Lambert, J.H., eds. (1994) Workshop 
proceedings: when and how can you specify a probability distribution 
when you don't know much. Risk Anal. 14: 661-706.
    Harwell, M.J.; Norton, B.; Cooper, W.; Gentile, J. (1994) Issue 
paper on ecological significance. In: Ecological risk assessment 
issue papers. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency; pp. 2-1 to 2-49. EPA/630/R-94/009.
    Hassell, M.P. (1986) Detecting density dependence. Trends Ecol. 
Evol. 1: 90-93.
    Health Council of the Netherlands (HCN). (1993) Ecotoxicological 
risk assessment and policy-making in the Netherlands--dealing with 
uncertainties. Network 6(3)/7(1): 8-11.
    Heck, W.W. (1993) Ecological assessment case study: the National 
Crop Loss Assessment Network. In: A review of ecological assessment 
case studies from a risk assessment perspective. Washington, DC: 
Risk Assessment Forum, U.S. Environmental Protection Agency; pp. 6-1 
to 6-32. EPA/630/R-92/005.

[[Page 47629]]

    Hermens, J.; Canton, H.; Janssen, P.; De Jong, R. (1984a) 
Quantitative structure-activity relationships and toxicity studies 
of mixtures of chemicals with anaesthetic potency: acute lethal and 
sublethal toxicity to Daphnia magna. Aquatic Toxicol. 5: 143-154.
    Hermens, J.; Canton, H.; Steyger, N.; Wegman, R. (1984b) Joint 
effects of a mixture of 14 chemicals on mortality and inhibition of 
reproduction of Daphnia magna. Aquatic Toxicol. 5: 315-322.
    Hill, A.B. (1965) The environment and disease: association or 
causation? Proc. Royal Soc. Med. 58: 295-300.
    Holling, C.S., ed. (1978) Adaptive environmental assessment and 
management. Chichester, U.K.; John Wiley & Sons.
    Houseknecht, C.R. (1993) Ecological risk assessment case study: 
special review of the granular formulations of carbofuran based on 
adverse effects on birds. In: A review of ecological assessment case 
studies from a risk assessment perspective. Washington, DC: Risk 
Assessment Forum, U.S. Environmental Protection Agency; pp. 3-1 to 
6-25. EPA/630/R-92/005.
    Huggett, R.J.; Kimerle, R.A.; Merhle, P.M., Jr.; Bergman, H.L., 
eds. (1992) Biomarkers: biochemical, physiological, and histological 
markers of anthropogenic stress. Boca Raton, FL: Lewis Publishers.
    Hughes, R.M. (1995) Defining acceptable biological status by 
comparing with reference conditions. In: Davis, W.S.; Simon, T.P., 
eds. Biological assessment and criteria: tools for water resource 
planning and decision making. Boca Raton, FL: Lewis Publishers; pp. 
31-47.
    Hurlbert, S.H. (1984) Pseudoreplication and the design of 
ecological field experiments. Ecol. Mono. 54: 187-211.
    Johnson, B.L. (1995) Applying computer simulation models as 
learning tools in fishery management. N. Amer. J. Fisheries Manag. 
15: 736-747.
    Karr, J.R. (1981) Assessment of biotic integrity using fish 
communities. Fisheries 6: 21-27.
    Karr, J.R.; Fausch, K.D.; Angermeier, P.L.; Yant, P.R.; 
Schlosser, I.J. (1986) Assessing biological integrity in running 
waters: a method and its rationale, Illinois Natural History Survey, 
Special Publication 5, Champaign, IL.
    Kenaga, E.E. (1973) Factors to be considered in the evaluation 
of the toxicity of pesticides to birds in the environment. Environ. 
Qual. Saf. 2: 166-181.
    Kendall, R.J.; Lacher, T.E.; Bunck, C.; Daniel, B.; Driver, C.; 
Grue, C.E.; Leighton, F.; Stansley, W.; Watanabe, P.G.; Whitworth, 
M. (1996) An ecological risk assessment of lead shot exposure in 
non-waterfowl avian species: upland game birds and raptors. Environ. 
Toxicol Chem. 15:4-20.
    Konemann, H. (1981) Fish toxicity studies with mixtures of more 
than two chemicals: a proposal for a quantitative approach and 
experimental results. Aquatic Toxicol. 19: 229-238.
    Laird, N.M.; Mosteller, R. (1990) Some statistical methods for 
combining experimental results. Int. J. Technol. Asses. Health Care 
6: 5-30.
    Landis, W.G.; Matthews, R.A.; Markiewicz, A.J.; Matthews, G.B. 
(1993) Multivariate analysis of the impacts of the turbine fuel JP-4 
in a microcosm toxicity test with implications for the evaluation of 
ecosystem dynamics and risk assessment. Ecotoxicology 2: 271-300.
    Lipnick, R.L. (1995) Structure-activity relationships. In: Rand, 
G.M.; Petrocelli, S.R., eds. Fundamentals of aquatic toxicology. 
London, UK: Taylor and Francis; pp. 609-655.
    Lipton, J.; Galbraith, H.; Burger, J.; Wartenberg, D. (1993) A 
paradigm for ecological risk assessment. Environ. Manag. 17:1-5.
    Ludwig, J.A.; Reynolds, J.F. (1988) Statistical ecology. New 
York, NY: Wiley-Interscience, 337 pp.
    Lynch, D.G.; Macek, G.J.; Nabholz, J.V.; Sherlock, S.M.; Wright, 
R. (1994) Ecological risk assessment case study: assessing the 
ecological risks of a new chemical under the Toxic Substances 
Control Act. In: A review of ecological assessment case studies from 
a risk assessment perspective, v. II. Washington, DC: Risk 
Assessment Forum, U.S. Environmental Protection Agency; pp. 1-1 to 
1-35. EPA/630/R-94/003.
    MacIntosh, D.L.; Suter G.W., II; Hoffman, F.O. (1994) Uses of 
probabilistic exposure models in ecological risk assessments of 
contaminated sites. Risk Anal. 14(4): 405-419.
    Mann, C. (1990) Meta-analysis in the breech. Science 249: 476-
480.
    McCarty, L.S.; Mackay, D. (1993) Enhancing ecotoxicological 
modeling and assessment: body residues and modes of toxic action. 
Environ. Sci. Technol. 27: 1719-1728.
    Meij, R. (1991) The fate of mercury in coal-fired power plants 
and the influence of wet flue-gas desulphurization. Water Air Soil 
Pollut. 56: 21-33.
    Menzie, C.A.; Burmaster, D.E.; Freshman, J.S.; Callahan, C.A. 
(1992) Assessment of methods for estimating ecological risk in the 
terrestrial component: a case study at the Baird & McGuire Superfund 
Site in Holbrook, Massachusetts. Environ. Toxicol. Chem. 11:245-260.
    Menzie, C.; Henning, M.H.; Cura, J.; Finkelstein, K.; Gentile, 
J.; Maughan J.; Mitchell, D.; Petron, S.; Potocki, B.; Svirsky, S.; 
Tyler, P. (1996) Special report of the Massachusetts weight-of-
evidence workgroup: a weight of evidence approach for evaluating 
ecological risks. Human Ecol. Risk Assess. 2:277-304.
    Merriam-Webster Dictionary. (1972) New York, NY: G. & C. Merriam 
Co., Simon & Schuster, Inc.
    Nabholz, J.V. (1991) Environmental hazard and risk assessment 
under the United States Toxic Substances Control Act. Science Total 
Environ. 109/110: 649-665.
    Nabholz, J.V.; Miller, P.; Zeeman, M. (1993) Environmental risk 
assessment of new chemicals under the Toxic Substances Control Act 
(TSCA) section five: In: Landis, W.G.; Hughes, S.G.; Lewis, M., eds. 
Environmental toxicology and risk assessment. Philadelphia, PA: 
American Society for Testing and Materials; pp. 40-55. ASTM STP 
1179.
    National Research Council. (1993) A paradigm for ecological risk 
assessment. In: Issues in risk assessment. Washington, DC: National 
Academy Press.
    National Research Council. (1994) Science and judgment in risk 
assessment. Washington, DC: National Academy Press.
    National Research Council. (1996) Understanding risk: Informing 
decisions in a democratic society. Washington, DC: National Academy 
Press.
    Niemi, G.J.; DeVore, P.; Detenbeck, N.; Taylor, D.; Lima, A.; 
Pastor, J.; Yount, J.D.; Naiman, R.J. (1990) Overview of case 
studies on recovery of aquatic systems from disturbance. Environ. 
Mgmt. 14: 571-587.
    O'Neill, R.V.; Gardner, R.H.; Barnthouse, L.W.; Suter, G.W.; 
Hildebrand, S.G.; Gehrs, C.W. (1982) Ecosystem risk analysis: a new 
methodology. Environ. Toxicol. Chem. 1: 167-177.
    Okkerman, P.C.; Plassche, E.J.V.D.; Emans, H.J.B. (1993) 
Validation of some extrapolation methods with toxicity data derived 
from multispecies experiments. Ecotoxicol. Environ. Saf. 25: 341-
359.
    Orr, R.L.; Cohen, S.D.; Griffin, R.L. (1993) Generic non-
indigenous pest risk assessment process. Beltsville, MD: USDA Animal 
and Plant Health Inspection Service.
    Ott, W.R. (1978) Environmental indices--theory and practice. Ann 
Arbor, MI: Ann Arbor Science. Cited in: Suter, G.W., II. (1993) A 
critique of ecosystem health concepts and indexes. Environ. Toxicol. 
Chem. 12: 1533-1539.
    Pearlstine, L.; McKellar, H.; Kitchens, W. (1985) Modelling the 
impacts of a river diversion on bottomland forest communities in the 
Santee River Floodplain, South Carolina. Ecol. Model. 29: 281-302.
    Pelczar, M.J.; Reid, R.D. (1972) Microbiology. New York, NY: 
McGraw-Hill Company.
    Peterman, R.M. (1990) The importance of reporting statistical 
power: the forest decline and acidic deposition example. Ecology 71: 
2024-2027.
    Petitti, D.B. (1994) Meta-analysis, decision analysis and cost-
effectiveness analysis: methods for quantitative synthesis in 
medicine. Monographs in epidemiology and biostatistics, v. 24. New 
York, NY: Oxford University Press.
    Phipps, R.L. (1979) Simulation of wetlands forest vegetation 
dynamics. Ecol. Model. 7: 257-288.
    Pielou, E.C. (1984) The interpretation of ecological data. A 
primer on classification and ordination. New York, NY: Wiley-
Interscience, 263 pp.
    Preston, E.M.; Bedford, B.L. (1988) Evaluating cumulative 
effects on wetland functions: a conceptual overview and generic 
framework. Environ. Manag. 12(5): 565-584.
    Risser, P.G. (1988) General concepts for measuring cumulative 
impacts on wetland ecosystems. Environ. Manag. 12(5): 585-590.
    Rotenberry, J.T.; Wiens, J.A. (1985) Statistical power analysis 
and community-wide patterns. Am. Naturalist 125: 164-168.
    Ruckelshaus, W.D. (1983) Science, risk, and public policy. 
Science 221: 1026-1028.
    Sawyer, T.W.; Safe, S. (1985) In vitro AHH induction by 
polychlorinated biphenyl and dibenzofuran mixtures: additive 
effects. Chemosphere 14: 79-84.
    Simberloff, D.; Alexander, M. (1994) Issue paper on biological 
stressors. In: Ecological

[[Page 47630]]

risk assessment issue papers. Washington, DC: Risk Assessment Forum, 
U.S. Environmental Protection Agency; pp. 6-1 to 6-59. EPA/630/R-94/
009.
    Smith, E.P.; Cairns J., Jr. (1993) Extrapolation methods for 
setting ecological standards for water quality: statistical and 
ecological concerns. Ecotoxicology 2: 203-219.
    Smith, E.P.; Shugart, H.H. (1994) Issue paper on uncertainty in 
ecological risk assessment. In: Ecological risk assessment issue 
papers. Washington, DC: Risk Assessment Forum, U.S. Environmental 
Protection Agency; pp. 8-1 to 8-53. EPA/630/R-94/009.
    Society of Environmental Toxicology and Chemistry (SETAC). 
(1987) Research priorities in environmental risk assessment. Report 
of a workshop held in Breckenridge, CO, August 16-21, 1987. 
Washington, DC: Society of Environmental Toxicology and Chemistry.
    Society of Environmental Toxicology and Chemistry (SETAC). 
(1994a) Aquatic risk assessment and mitigation dialogue group. Final 
report. Pensacola, FL: SETAC Foundation for Environmental Education.
    Solomon, K.R.; Baker, D.B.; Richards, R.P.; Dixon, K.R.; Klaine, 
S.J.; La Point, T.W.; Kendall, R.J.; Weisskopf, C.P.; Giddings, 
J.M.; Geisy, J.P.; Hall, L.W.; Williams, W.M. (1996) Ecological risk 
assessment of atrazine in North American surface waters. Environ. 
Toxicol. Chem. 15(1): 31-76.
    Starfield, A.M.; Bleloch, A.L. (1991) Building models for 
conservation and wildlife management. Edina, MN: Burgess 
International Group, Inc.
    Stephan, C.E. (1977) Methods for calculating an LC50. 
In: ASTM Special Technical Publication 634. Philadelphia, PA: 
American Society for Testing and Materials; pp. 65-88.
    Stephan, C.E.; Rogers, J.R. (1985) Advantages of using 
regression analysis to calculate results of chronic toxicity tests. 
In: Bahner, R.C.; Hanse, D.J., eds. Aquatic toxicology and hazard 
assessment. Eighth symposium. Philadelphia, PA: American Society for 
Testing and Materials; pp. 328-339.
    Stephan, C.E.; Mount, D.I.; Hansen, D.J.; Gentile, J.H.; 
Chapman, G.A.; Brungs, W.A. (1985) Guidelines for deriving numerical 
national water quality criteria for the protection of aquatic 
organisms and their uses. Duluth, MN: Office of Research and 
Development, U.S. Environmental Protection Agency. PB85-227049.
    Stewart-Oaten, A.; Murdoch, W.W.; Parker, K.R. (1986) 
Environmental impact assessment: ``pseudoreplication'' in time? 
Ecol. 67(4): 929-940.
    Susser, M. (1986a) Rules of inference in epidemiology. Reg. 
Toxicol. Pharmacol. 6: 116-128.
    Susser, M. (1986b) The logic of Sir Carl Popper and the practice 
of epidemiology. Am. J. Epidemiol. 124: 711-718.
    Suter, G.W., II. (1990) Endpoints for regional ecological risk 
assessments. Environ. Manag. 14: 19-23.
    Suter, G.W., II. (1993a) Ecological risk assessment. Boca Raton, 
FL: Lewis Publishers.
    Suter, G.W., II. (1993b) A critique of ecosystem health concepts 
and indexes. Environ. Toxicol. Chem. 12: 1533-1539.
    Suter, G.W., II; Vaughan, D.S.; Gardner, R.H. (1983) Risk 
assessment by analysis of extrapolation error. A demonstration for 
effects of pollutants on fish. Environ. Toxicol. Chem. 1: 369-377.
    Suter, G.W., II; Gillett, J.W.; Norton, S.B. (1994) Issue paper 
on characterization of exposure. In: Ecological risk assessment 
issue papers. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency; pp. 4-1 to 4-64. EPA/630/R-94/009.
    Thomas, J.W.; Forsman, E.D.; Lint, J.B.; Meslow, E.C.; Noon, 
B.R.; Verner, J. (1990) A conservation strategy for the spotted owl. 
Interagency Scientific Committee to Address the Conservation of the 
Northern Spotted Owl. 1990-791/20026. Washington, DC: U.S. 
Government Printing Office.
    Urban, D.J.; Cook, J.N. (1986) Ecological risk assessment. 
Hazard Evaluation Division standard procedure. Washington, DC: 
Office of Pesticide Programs, U.S. Environmental Protection Agency. 
EPA-54019-83-001.
    U.S. Department of Agriculture (1993) Pest risk assessment of 
the importation of Pinus radiata, Nothofagus dombeyi, and Laurelia 
philippiana logs from Chile. Forest Service Miscellaneous 
Publication 1517.
    U.S. Department of Health, Education, and Welfare. (1964) 
Smoking and health. Report of the Advisory Committee to the Surgeon 
General. Washington, DC: U.S. Department of Health, Education, and 
Welfare. Public Health Service Publication 1103.
    U.S. Environmental Protection Agency. (1979) Toxic Substances 
Control Act. Discussion of premanufacture testing policies and 
technical issues: request for comment. FR 44: 16240-16292.
    U.S. Environmental Protection Agency. (1984) Estimating concern 
levels for concentrations of chemical substances in the environment. 
Washington, DC: U.S. Environmental Protection Agency, Health and 
Environmental Review Division, Environmental Effects Branch.
    U.S. Environmental Protection Agency. (1986a) Quality criteria 
for water. Washington, DC: Office of Water, U.S. Environmental 
Protection Agency. EPA/440/5-86/001.
    U.S. Environmental Protection Agency. (1986b) Guidelines for the 
health risk assessment of chemical mixtures. FR 52: 34014-34025.
    U.S. Environmental Protection Agency. (1988a) Methods for 
aquatic toxicity identification evaluations: phase I toxicity 
characterization procedures. Duluth, MN: Environmental Research 
Laboratory, U.S. Environmental Protection Agency. EPA/600/3-88/034.
    U.S. Environmental Protection Agency. (1988b) User's guide to 
PDM3, final report; EPA Contract No. 68-02-4254, task no. 117. 
Exposure Assessment Branch, Exposure
    U.S. Environmental Protection Agency. (1989a) Rapid 
bioassessment protocols for use in streams and rivers: benthic 
macroinvertebrates and fish. Washington, DC: Office of Water, U.S. 
Environmental Protection Agency. EPA/440/4-89/001.
    U.S. Environmental Protection Agency. (1989b) Methods for 
aquatic toxicity identification evaluations: phase II toxicity 
identification procedures. Duluth, MN: Environmental Research 
Laboratory, U.S. Environmental Protection Agency. EPA/600/3-88/035.
    U.S. Environmental Protection Agency. (1989c) Methods for 
aquatic toxicity identification evaluations: phase III toxicity 
confirmation procedures. Duluth, MN: Environmental Research 
Laboratory, U.S. Environmental Protection Agency. EPA/600/3-88/035.
    U.S. Environmental Protection Agency. (1990) Guidance for data 
useability in risk assessment. Washington, DC: U.S. Environmental 
Protection Agency. EPA/540/G-90/008.
    U.S. Environmental Protection Agency. (1991) Summary report on 
issues in ecological risk assessment. Washington, DC: Risk 
Assessment Forum, U.S. Environmental Protection Agency. EPA/625/3-
91/018.
    U.S. Environmental Protection Agency. (1992a) Framework for 
ecological risk assessment. Washington, DC: Risk Assessment Forum, 
U.S. Environmental Protection Agency. EPA/630/R-92/001.
    U.S. Environmental Protection Agency. (1992b) Peer review 
workshop report on a framework for ecological risk assessment. 
Washington, DC: Risk Assessment Forum, U.S. Environmental Protection 
Agency. EPA/625/3-91/022.
    U.S. Environmental Protection Agency. (1992c) Ecological risk 
assessment guidelines strategic planning workshop. Washington, DC: 
Risk Assessment Forum, U.S. Environmental Protection Agency. EPA/
630/R-92/002.
    U.S. Environmental Protection Agency. (1992d) Guidelines for 
exposure assessment: notice. FR 57: 22888-22938.
    U.S. Environmental Protection Agency. (1993a) A review of 
ecological assessment case studies from a risk assessment 
perspective. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency. EPA/630/R-92/005.
    U.S. Environmental Protection Agency. (1993b) A guidebook to 
comparing risks and setting environmental priorities. Washington, 
DC: Office of Policy, Planning, and Evaluation, U.S. Environmental 
Protection Agency. EPA/230/B-93/003.
    U.S. Environmental Protection Agency. (1993c) Wildlife exposure 
factors handbook. Washington, DC: Office of Research and 
Development, U.S. Environmental Protection Agency. EPA/600/R-93/187a 
and 187b.
    U.S. Environmental Protection Agency. (1993d) Communicating risk 
to senior EPA policy makers: a focus group study. Research Triangle 
Park, NC: Office of Air Quality Planning and Standards, U.S. 
Environmental Protection Agency.
    U.S. Environmental Protection Agency. (1994a) A review of 
ecological assessment case studies from a risk assessment 
perspective, volume II. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency. EPA/630/R-94/003.
    U.S. Environmental Protection Agency. (1994b) Ecological risk 
assessment issue papers. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency. EPA/630/R-94/009.

[[Page 47631]]

    U.S. Environmental Protection Agency. (1994c) Peer review 
workshop report on ecological risk assessment issue papers. 
Washington, DC: Risk Assessment Forum, U.S. Environmental Protection 
Agency. EPA/630/R-94/008.
    U.S. Environmental Protection Agency. (1994d) Guidance for the 
data quality objectives process. Washington, DC: Quality Assurance 
Management Staff. EPA QA/G-4.
    U.S. Environmental Protection Agency. (1994e) Ecosystem 
protection. Memorandum from Robert Perciaspe, David Gardiner, and 
Johnathan Cannon to Carol Browner, March 1994.
    U.S. Environmental Protection Agency. (1994f) Environmental 
Services Division guidelines. Hydrogeologic modeling. Seattle, WA: 
Region X, U.S. Environmental Protection Agency.
    U.S. Environmental Protection Agency. (1994g) Managing 
ecological risks at EPA: issues and recommendations for progress. 
Washington, DC: Center for Environmental Research Information, U.S. 
Environmental Protection Agency. EPA/600/R-94/183.
    U.S. Environmental Protection Agency. (1995a) An SAB report: 
ecosystem management--imperative for a dynamic world. Washington, 
DC: Science Advisory Board. EPA-SAB-EPEC-95-003.
    U.S. Environmental Protection Agency. (1995b) Ecological risk: a 
primer for risk managers. Washington, DC: U.S. Environmental 
Protection Agency. EPA/734/R-95/001.
    U.S. Environmental Protection Agency. (1995c) Memo to EPA 
managers from administrator Carol Browner, ``EPA risk 
characterization program''; March 1995.
    U.S. Environmental Protection Agency. (1995d) The use of the 
benchmark dose approach in health risk assessment. Washington, DC: 
Risk Assessment Forum, U.S. Environmental Protection Agency. EPA/
630/R-94/007.
    U.S. Environmental Protection Agency. (In Press) Peer review 
workshop report on draft proposed guidelines for ecological risk 
assessment. Washington, DC: Risk Assessment Forum, U.S. 
Environmental Protection Agency. EPA/630/R-96/002.
    Van Leeuwen, C.J.; Van der Zandt, P.T.J.; Aldenberg, T.; Verhar, 
H.J.M.; Hermens, J.L.M. (1992) Extrapolation and equilibrium 
partitioning in aquatic effects assessment. Environ. Toxicol. Chem. 
11: 267-282.
    Wagner, C.; Lkke, H. (1991) Estimation of 
ecotoxicological protection levels from NOEC toxicity data. Water 
Res. 25: 1237-1242.
    White, P.S.; Pickett, S.T.A. (1985) Natural disturbance and 
patch dynamics: an introduction. In: Pickett, S.T.A.; White, P.S., 
eds. The ecology of natural disturbance and patch dynamics. Orlando, 
FL: Academic Press; pp. 3-13.
    Wiegert, R.G.; Bartell, S.M. (1994) Issue paper on risk 
integration methods. In: U.S. Environmental Protection Agency. 
Ecological risk assessment issue papers. Washington, DC: Risk 
Assessment Forum, Environmental Protection Agency; pp. 9-1 to 9-66. 
EPA/630/R-94/009.
    Wiens, J.A.; Parker, K.R. (1995) Analyzing the effects of 
accidental environmental impacts: approaches and assumptions. Ecol. 
Appl. 5(4): 1069-1083.
    Woodman, J.N.; Cowling, E.B. (1987) Airborne chemicals and 
forest health. Environ. Sci. Technol. 21: 120-126.
    Yoder, C.O.; Rankin, E.T. (1995) Biological response signatures 
and the area of degradation value: new tools for interpreting multi-
metric data. In: Davis, W.S.; Simon, T.P., eds. Biological 
assessment and criteria: tools for water resource planning and 
decision making. Boca Raton, FL: Lewis Publishers.
    Zeeman, M. (1995) EPA's framework for ecological effects 
assessment. In: Screening and testing chemicals in commerce. 
Washington, DC: Office of Technology Assessment, OTA-BP-ENV-166; pp. 
69-78.

[FR Doc. 96-22648 Filed 9-6-96; 8:45 am]
BILLING CODE 6560-50-P