EPA/600/R-06/037F
                                        March 2008
   Application of Watershed Ecological Risk
Assessment Methods to Watershed Management
           National Center for Environmental Assessment
              Office of Research and Development
             U.S. Environmental Protection Agency
                   Washington, DC 20460

-------
                                     DISCLAIMER

       This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication.  Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
                                       ABSTRACT

       Watersheds are frequently used to study and manage environmental resources because
hydro logic boundaries define the flow of contaminants and other stressors. It is a challenge to
incorporate scientific information in watershed management and planning. Ecological assessments
of watersheds are complex because watersheds typically overlap multiple jurisdictional
boundaries, are subjected to multiple  environmental stressors, and have multiple stakeholders with
diverse environmental and socioeconomic interests. Ecological risk assessment (ERA) is an
approach that has successfully been used to increase the use of ecological science in decision
making, by evaluating the likelihood that adverse ecological effects may result from exposure to
one or more stressors, yet its application to watershed assessment is limited.  The purpose of this
report  is  to provide suggestions and examples for making scientific information more relevant to
the needs of watershed managers by using ERA principles to help structure ecological
assessments  of watersheds.
       This report supplements the Guidelines for Ecological Risk Assessment (U.S. EPA
1998a) by addressing issues commonly encountered when conducting watershed ecological
assessments.  Suggestions and examples to follow are provided based upon lessons learned from
prior watershed ERAs. This report is of potential use to ecologists, hydrologists, watershed
managers, risk assessors, landscape ecologists, and other scientists and managers seeking to
increase the use of environmental assessment data in decision making.
       Each activity and phase of the watershed ERA process is  explained sequentially in this
report. Guidance on how to involve stakeholders to generate environmental management goals
and objectives is provided. The processes for selecting assessment endpoints, developing
conceptual models, and selecting the  exposure and effects pathways to be analyzed are described.
Suggestions for predicting how multiple sources and stressors  affect assessment endpoints are
also provided; these include using multivariate analyses to compare land use with biotic
measurements.  In addition, the report suggests how to estimate,  describe, and communicate risk
and how to evaluate management alternatives.
Preferred Citation:
U.S. Environmental Protection Agency (EPA). (2007) Application of watershed ecological risk assessment
methods to watershed management. National Center for Environmental Assessment, Washington, DC;
EPA/600/R-06/037F. Available from National Technical Information Service, Springfield, VA, and online at
http ://www. epa.gov/ncea.
                                            11

-------
                             CONTENTS

LIST OF TABLES 	v
LIST OF FIGURES	v
LIST OF ABBREVIATIONS AND ACRONYMS 	vi
AUTHORS, CONTRIBUTORS, AND REVIEWERS	vii

1. EXECUTIVE SUMMARY 	1

2. INTRODUCTION	5
  2.1.  WHO WILL USE THIS REPORT	7
  2.2.  WHAT IS WATERSHED ECOLOGICAL RISK ASSESSMENT? 	7
  2.3.  THE WATERSHED ERA PROCESS 	8

3. PLANNING PHASE 	12
  3.1.  IDENTIFYING AND INCLUDING STAKEHOLDERS	12
  3.2.  SETTING ENVIRONMENTAL MANAGEMENT GOALS AND OBJECTIVES ... 14
  3.3.  DEFINING MANAGEMENT OPTIONS	17
  3.4.  DETERMINING THE FOCUS, SCOPE, AND COMPLEXITY OF THE RISK
      ASSESSMENT 	17
  3.5.  PRODUCTS OF THE PLANNING PHASE  	18

4. PROBLEM FORMULATION PHASE	19
  4.1.  IDENTIFY STRESSORS AND SOURCES	20
  4.2.  SELECTING ASSESSMENT ENDPOINTS	21
  4.3.  DEVELOP CONCEPTUAL MODEL 	22
  4.4.  EVALUATE ASSESSMENT ENDPOINTS  	26
  4.5.  CREATE ANALYSIS PLAN	27

5. RISK ANALYSIS PHASE	29
  5.1.  STRESSOR IDENTIFICATION EVALUATION (SIE) AND CAUSAL
      EVALUATION: IDENTIFY STRESSORS AND EVALUATE CAUSES OF
      IMPAIRMENT 	30
  5.2.  EVALUATE STRENGTH OF ASSOCIATION BETWEEN CAUSES
      AND EFFECTS	31
  5.3.  WORK WITH CHEMICAL STRESSORS	32
  5.4.  WORK WITH NONCHEMICAL STRESSORS	33
  5.5.  WORK WITH DATA LIMITATIONS	35
      5.5.1. Use Categorical Data	35
      5.5.2. Use Source Data	36
      5.5.3. Rank Risks	38
  5.6.  DEAL WITH UNCERTAINTY 	40
  5.7.  RECOGNIZE THAT ANALYSIS IS AN ITERATIVE PROCESS 	41
  5.8.  UNDERSTAND SECONDARY OR INDIRECT PATHWAYS  	42
                                 m

-------
                         CONTENTS (continued)
6. RISK CHARACTERIZATION PHASE	44
  6.1.  RISK DESCRIPTION  	44
  6.2.  DESCRIBING UNCERTAINTY	49
  6.3.  RISK CHARACTERIZATION COMPONENTS	51

7. RISK COMMUNICATION	53
  7.1.  THE RISK COMMUNICATION PROCESS	53
  7.2.  ADDRESS THE TARGET AUDIENCE 	54
  7.3.  COMMUNICATE EFFECTIVELY	55

8. RISK MANAGEMENT	56
  8.1.  EVALUATING MANAGEMENT ALTERNATIVES	56
  8.2.  APPLICATION TO DECISION MAKING IN THE EPA CASE STUDIES 	58
  8.3.  APPLICATION TO THE TOTAL MAXIMUM DAILY LOAD PROCESS	60
  8.4.  APPLICATION TO THE UAAPROCESS 	61

9. CONCLUSIONS 	67

GLOSSARY	69

REFERENCES	72
                                 IV

-------
                                   LIST OF TABLES
1.      Prominent aspects of the watershed approach and comparable aspects of
       the ecological risk assessment process 	8

2.      Effects matrix summarizing assumed strength of relationships between assessment
       endpoints identified (e.g., percent eelgrass cover) and stressors in the Waquoit Bay
       watershed  	21

3.      Middle Snake River case study: integration of stressors, responses, and recovery
       potential for the reproduction,  growth, and survival of the mountain whitefish
       population	46
                                  LIST OF FIGURES
1.      The ecological risk assessment framework	9

2.      Relationships between management or ecological risk assessment (ERA) goals
       and objectives, management options, the planning phase, and subsequent
       problem formulation in watershed ERA	13

3.      Elementary conceptual model	23

4.      Conceptual model of the Waquoit Bay watershed	25

5.      Significant relationships between urban land use activities and invertebrate ephemeroptera,
       plecoptera, and trichoptera (EPT) score	37

6.      Historical changes in nitrogen (N) loading predicted by the nitrogen loading model  ... 47

7.      Fish index of biotic integrity (IBI) or insect ephemeroptera, plecoptera, and
       trichoptera (EPT) values in relationship to proximity to coal mining sources
       in the Clinch Valley ecological isk assessment   	48

8.      Thermometer scale for relative ranking of management options  	57

9.      General steps involved in conducting a use attainability analysis (UAA)
       using a watershed ecological risk framework 	64

-------
                   LIST OF ABBREVIATIONS AND ACRONYMS
BP J       Best professional judgment
BMP      Best management practices
CADDIS  Causal Analysis/Diagnosis Decision Information System
COLD    Conditions suitable to support coldwater ecosystems
CWA     Clean Water Act
CWAP    Clean Water Action Plan
EIS       Environmental Impact Statement
EPA      U.S. Environmental Protection Agency
ERA      Ecological risk assessment
ESA      Endangered Species Act
EPT      Ephemeroptera, Plecopera, and Trichoptera
FIFRA    Federal Insecticide, Fungicide, and Rodenticide Act
FWS      U.S. Fish and Wildlife Service
GIS       Geographic information system
HSI       Habitat Suitability Index
IBI       Index of Biotic Integrity
ICI       Index of Community Integrity
LOE      Lines of evidence
LIT       Literature search
Miwb     Modified Index of Well-Being
MOS      Margin of safety
NCEA    National Center for Environmental Assessment
NEPA    National Environmental Policy Act
NRC      National Research Council
RRM      Relative Risk Model
SAB      Science Advisory Board
SIE       Stressor Identification Evaluation
TMDL    Total maximum daily load
TNC      The Nature Conservancy
TSCA    Toxic Substances Control Act
TVA      Tennessee  Valley Authority
UAA      Use Attainability Analysis
WQS      Water quality standards
                                         VI

-------
                  AUTHORS, CONTRIBUTORS, AND REVIEWERS
       The National Center for Environmental Assessment within EPA's Office of Research and
Development was responsible for preparing this document. This document was written by the
primary author and several co-authors from Tetra Tech, Inc., under contract number #3W-2881-
NATX. EPA staff and external experts reviewed this report.  Three of the external reviewers
were independently selected by Versar.

Project Officer:
Victor Serveiss, U.S. EPA, National Center for Environmental Assessment, Washington, DC

Authors:
Victor Serveiss, U.S. EPA, National Center for Environmental Assessment, Washington, DC
Jerome Diamond, Tetra Tech, Inc., Owings Mills, MD
Jonathon Butcher, Tetra Tech, Inc., Owings Mills, MD
Jeroen Gerritsen, Tetra Tech, Inc., Owings Mills, MD
Benjamin Jessup, Tetra Tech, Inc., Owings Mills, MD

Reviewers:
EPA Reviewers:
Barbara Mazur, Region V, Chicago, IL
Leela Rao, formerly with NCEA, Washington, DC
Gina Ferreira, Region II, New York, NY
Randy Bruins, formerly with NCEA, Cincinnati, OH
Katherine Sciera, Region VI, Dallas, TX
Jeff Frithsen, NCEA, Washington, DC

External Reviewers:
Wayne Landis, Western Washington University, Bellingham, WA
Peter DeFur, Environmental Stewardship Concepts, Richmond, VA
Bruce K. Hope, Oregon Department of Environmental Quality, Portland, OR
Mark T. Southerland, Versar, Inc., Columbia, MD
Barbara S. Washburn, Office of Environmental Health Hazard Assessment, Sacramento, CA
David Dow, National Oceanic and Atmospheric Administration, Woods Hole, MA

Acknowledgments:
Technical Editing  Contributors:
Julie Blue, The Cadmus Group, Inc., Watertown, MA
Roy Schiff, Milone and MacBroom, Inc., Cheshire, CT
Chi Ho Sham, The Cadmus Group, Inc., Watertown, MA
Karen Swetlow, KBM Group, Silver Spring,  MD
                                          Vll

-------
                              1.  EXECUTIVE SUMMARY

       Environmental managers are increasingly using a watershed approach for making
environmental decisions.  These environmental managers include state, federal, or local regulatory
or resource agency staff; local and county officials; watershed associations or councils; and river
watch citizen groups.  Many environmental decisions are made with much uncertainty, and the
quality of such decisions would increase if more science were used in the decision making
process.  Ecological risk assessment (ERA) is a process to collect, organize, and analyze scientific
information in order to evaluate the likelihood that adverse ecological effects may occur or are
occurring as a result of exposure to one or more stressors (U.S. EPA, 1998a).  This document
discusses the merits of applying ERA principles to watershed management and describes how
more science can be used in decision making.
       The watershed approach is a framework for coordinating environmental management that
focuses public- and private-sector efforts on addressing the highest priority problems within a
hydro logically-defined geographic area (U.S. EPA, 1996). Watershed ERA integrates ERA
methodologies with the watershed approach to improve the use of environmental monitoring and
assessment data in watershed and regional decision making. This document is written primarily to
provide guidance and examples for ecological risk assessors, scientists, and managers performing
ecological assessments of watersheds.  Some of the principles also should be useful to landscape
or seascape ecologists performing spatial-scale environmental assessments and to others seeking
to increase the use of ecological science in decision making. The document is structured around
the elements of watershed ERA (planning, problem formulation, risk analysis, risk
characterization, risk communication, and risk management) and is based primarily on lessons
learned from EPA-sponsored case studies.
       Watershed ERAs are complex because a watershed typically overlaps multiple
jurisdictional boundaries and has multiple stakeholders with diverse environmental and
socioeconomic interests.  Planning is essential in ensuring that watershed ERAs address locally
based socioeconomic and scientific challenges. Key elements of planning include identifying and
involving relevant stakeholders, developing environmental management goals and objectives, and
agreeing  on the focus, scope, and complexity of an assessment. Appropriate goals and objectives
should be based on existing watershed management plans, environmental organizations' mission
statements, survey results, and other stakeholder opinion data. An interdisciplinary team of
scientists and managers who understand ecological principles and possess local ecological and
socioeconomic knowledge is needed to develop a set of scientific assessment objectives and
management options.

-------
       During watershed assessments, the needs of managers and stakeholders may change and
managers may need to take action before an assessment is completed. Thus, flexibility on how the
ERA process is implemented, along with regular and recurring interactions between scientists and
managers throughout the assessment process, are essential.  ERA, and especially watershed ERA,
is an iterative process. It is unlikely that everything will proceed precisely in the chronological
manner presented in this report.
       Watershed ERA problem formulation provides an organizing framework for the entire
assessment.  Assessment endpoints are identified and conceptual models and an analysis plan are
developed. Assessment endpoint selection translates abstract environmental management
objectives into specific, well-defined, and identifiable attributes of the system.  Establishing a
linkage between abstract management objectives and specific system attributes is particularly
important in watershed assessments because of the ultimate need to choose between competing
management objectives. Several assessment endpoints may be identified, but only a subset of
them may be analyzed because resources, data availability, or realistic management control
options may be limited.  To select a subset of assessment endpoints to be analyzed, an assessment
team composed of risk assessors, scientists, engineers, policy analysts, managers, and other
professionals may use their best professional judgment to assess stressor impact and risk reduction
opportunities.
       Conceptual models are  especially valuable in watershed assessments because they describe
multiple physical,  chemical, and biological stressors, their sources, and the pathways by which
they affect various valued ecological resources. Due to the large number of pathways to be
considered, the conceptual model can make it easier to identify causal pathways and establish
priorities for analytical efforts.  The model may also evolve as the assessment progresses and a
better understanding of pathways and impacts is acquired.
       The final product of problem formulation is an analysis plan documenting the proposed
approach for the assessment. Sometimes, when assessment resources are limited or a decision
must be made quickly, it may not be practical to complete a quantitative and formal assessment.
In these instances, the qualitative information from problem formulation can be used for decision
making.
       Assuming  the assessment team proceeds with the ERA, the next step is the risk analysis
phase.  It is ideal to develop stressor-response relationships that relate the magnitude,  duration,
frequency, and timing of exposure to the biological effects. However, generating such
relationships for multiple interacting sources, stressors,  and endpoints for a watershed assessment
is not trivial.  Geospatial analysis using geographic information system (GIS) tools usually play a
pivotal role in analyzing land use, land cover, and other stressor and response data. To be cost
effective and efficient, a watershed assessment might need to focus  on assessing the relationships

-------
of one or a few major stressors (or their sources) on a small set of ecological effects rather than
on attempting to quantify each stressor's exposure and effect on each endpoint.  For instance, the
focus could be on studying the relationship between percentage of urban land and fish species
richness. Another approach is to examine relationships between groups of stressors, habitat data,
and impact data (i.e., groups of responses).  High relative risk occurs in the spatial areas where
source, habitat, and impact co-occur. Categorical data or scores from one to five may be used
rather than numerically modeled or predictive estimates.  Moreover, it may be necessary to
extrapolate results from a similar watershed as long as uncertainties are adequately documented.
When data on the assessment endpoint are not available, a measure of effect on a surrogate may
be commonly used instead.  Finally, exposure and effects data may need to be aggregated, thus
implying that the typical risk analysis phase is performed in conjunction with the next phase—risk
characterization.
       ERA risk characterization consists of risk estimation and risk description. Risk estimation
is a difficult task because stressors can affect valued ecological resources in different habitats via
multiple interacting  pathways. The exposure and effects characterizations (or their aggregated
estimates, as may be necessary in a watershed context) developed in the prior phase should be
used as lines of evidence to support particular risk estimates. Lines of evidence  also may include
models, field work,  and analyses performed elsewhere. All the lines of evidence need to be
brought together to  reach a final conclusion about the likelihood and the consequences of effects.
The spatial distribution of exposure and effect is important to consider and is often assessed and
presented using GIS data layers.
       Risk description summarizes and presents information so that choices among alternative
courses of action can be made with knowledge of the outcomes.  Descriptions can be qualitative
statements such as "option A causes more risk than option B." Effects data also can be color-
coded onto maps of the watershed, allowing stakeholders to see potential impacts in the lakes and
rivers that matter most to them. A data table can be used to show the various lines  of evidence
and to summarize the conclusions of risk and recovery potential to various life stages. Risk
assessments need to be peer reviewed to help convince skeptics.  Assumptions and uncertainty
can be summarized  in data tables or described in narrative form.
       Successful communication of risk assessment findings requires that communication occur
throughout the assessment process to ensure the results reflect the management  objectives,
decision context, and environmental conditions that exist at the time the information is needed.
Managers need to tell assessors what they need,  and scientists need to describe what they can
provide.  Effective risk communication must accurately translate the best available and most
useful scientific information in a manner understandable to managers and stakeholders.
Depending on the audience, the risk assessment information may need to be provided in a

-------
nontechnical manner. Clear presentation of results (e.g., maps, figures with simple dose-response
curves) will enhance understanding of findings and foster consensus among stakeholders.
       Watershed ERA principles can help watershed managers make more informed total
maximum daily load (TMDL) decisions or take a wiser approach to identify or attain beneficial
use designations.  ERA principles can also assist managers in resource planning, making land use
zoning decisions,  and implementing best management practices (BMPs).

-------
                                  2.  INTRODUCTION

       This report discusses the merit of applying ERA principles to watershed management and
describes how science can be better integrated into decision making. Environmental managers
are increasingly using a watershed approach for making decisions.  Environmental managers
include regulatory or resource agency (state, federal, local) staff and any party with authority to
implement a management plan (e.g., local and county officials with zoning oversight), watershed
associations or councils, and river watch citizen groups.  Many environmental management
decisions are based on uncertain information and would benefit from having a stronger scientific
basis.
       ERA is a process to collect, organize, and analyze scientific  information in order to
evaluate the likelihood that adverse ecological effects may occur or are occurring as a result of
exposure to one or more stressors (U.S. EPA, 1998a). ERA principles are relevant to
environmental management because even though most decisions are made without a risk
assessment, the decisions are fundamentally about managing ecological risks.
       EPA and others, through various regulations, have reduced the impact of point source
pollution and it is  now widely recognized that further substantial improvements in the quality of
U.S. lakes,  streams, rivers, creeks, and ponds depend on controlling nonpoint source pollution
(U.S. EPA, 2006). Nonpoint source pollution problems are not as well addressed by existing
regulations because of the difficulties in  identifying diffuse sources and ill-defined pathways.
Efforts to address nonpoint source pollution rely heavily on voluntary compliance, stakeholder
involvement, and a scientific understanding of the cumulative impacts of multiple physical,
chemical, and biological stressors over a broad range of spatial scales.
       Consistently incorporating science in watershed management decisions, however, is
challenging. Multiple physical, chemical, and biological stressors result from human activities.
These stressors, when combined with a network of interrelated environmental conditions, cause
diverse impacts on numerous ecological resources. Tradeoffs among environmental, political,
economic, and social factors often result from subjective value judgments that occur as part of the
decision process.  It is sometimes difficult to implement scientifically supportable actions because
of conflicting interests, lack of certainty (or public trust) in the science, and poor communication
of the risks and benefits, particularly to those who must face the risk. As a result of these
challenges,  data from many monitoring and assessment efforts frequently do not play a major role
in management  decisions (Ward, 1996; Ward et al, 1986).
       Helpful  suggestions and frameworks have been developed to increase the use of science in
watershed and regional environmental management (Rhoads et al., 1999; Maxwell, 1998; U.S.
EPA, 1996; Armitage, 1995; MacDonald,  1994; Slocombe, 1993; Ward et al., 1986). Guidelines

-------
for Ecological Risk Assessment (U.S. EPA, 1998a), written for all types of ecological risk
assessment, generally addresses the application of ERA to watershed and regional assessment.
Yet the application of ERA to these larger scales is still somewhat limited (Serveiss, 2002; U.S.
EPA, 2000a).  This report supplements the ERA guidelines by providing additional guidance for
those performing watershed and regional ecological assessments.
       In the mid-1990s, EPA's Risk Assessment Forum and Office of Water cosponsored the
development of five demonstration watershed ERAs to test application of the ERA Framework
(U.S. EPA, 1992) to Office of Water programs. The five watershed assessments were performed
in the Clinch and Powell Valley, VA, Middle Snake River, ID, Waquoit Bay, MA, Big Darby
Creek, OH, and the Middle [segment of the] Platte River, NE. All five selected sites had valued
ecological resources, multiple stressors, an existing data set, and willing assessment participants.
Using these assessments, many key aspects of the watershed ERA process, including challenges
encountered and lessons learned, were documented. Several researchers (Diamond et al., 2002;
Serveiss, 2002; U.S. EPA, 2000b;  Butcher et al., 1998) reviewed these assessments and
concluded that use of the risk assessment approach can add significant value to watershed
management.
       Butcher et al. (1998)  described the benefits of applying the integrated watershed risk
assessment approach in a review of these five case studies, including the following:

       •   The risk assessment framework can add significant value to watershed-scale
           management programs  that follow the watershed approach, particularly when
           addressing problems caused by multiple and nonchemical stressors.
       •   The watershed approach is expected to benefit from use of the formal and scientifically
           defensible methods of risk assessment for prioritizing and evaluating risk.
       •   Although best professional judgment may arrive at the same conclusions as an ERA,
           the process helps  people to carefully examine what led them to their conclusions and
           document their findings.
       •   Simpler methods  will be required for applying watershed ERA to environmental
           management on a widespread basis.

       In addition to the EPA case studies, many other watershed assessments have been
published; several of these are referenced in this report. A comprehensive review of watershed
assessment literature would require a much larger report and is beyond the scope of this effort.

-------
2.1. WHO WILL USE THIS REPORT?
       This report is principally written for environmental scientists and risk assessors seeking to
provide scientific information for watershed management.  Watershed managers, coordinators,
and others participating in watershed assessment may also  find this report useful. The report also
provides valuable information to researchers, educators, and students in the fields of aquatic
ecology, landscape and seascape ecology, ecological risk assessment, and watershed management.
Although the focus of this report is on watershed assessment, many of the principles can be
applied to any environmental monitoring and assessment activity intended to inform
decisionmakers. Finally, with its examples and descriptions of watershed ERA, this report also
serves as an easy-to-understand introduction to ERA.

2.2. WHAT IS WATERSHED ECOLOGICAL RISK  ASSESSMENT?
       Watershed ERA combines the watershed approach with ERA. Before describing EPA's
watershed ERA approach, we need to first describe the watershed approach.

Watershed Approach
       The watershed approach is organized around the guiding principles of partnerships,
geographic focus,  and management based on sound science and data. The watershed approach is
a framework for coordinating environmental management that focuses public- and private-sector
efforts on addressing the highest priority problems within a hydro logically-defined geographic
area (U.S. EPA, 1996). The watershed approach should involve the pertinent levels of
government, users of watershed resources, environmental groups, those believed to cause
environmental problems, and the public and should help them better understand the specific
problems in hand,  identify and agree on goals and priorities, and choose and implement solutions.
Because watersheds often cross political boundaries, it is important to involve stakeholders across
these  boundaries.
       EPA and other agencies are increasingly using watersheds for environmental management
because watersheds are naturally cohesive hydrologic units that are spatially appropriate for
management actions (Maxwell, 1998; U.S. EPA, 1995a). For example, the Tennessee Valley
Authority has River Action Teams, in which members share monitoring information with key
stakeholders (i.e.,  regulatory agencies, state and local governments, businesses and industries,
citizen-based action groups, and watershed residents) to obtain their support in developing and
implementing protection and mitigation plans (U.S. EPA, 2002a).  The Canaan Valley Task
Force in West Virginia produced an inventory of environmental stressors associated with
ecosystem problems and their causes, determined whether  the problems are getting worse, and
developed solutions (U.S. EPA, 1997a). In addition, many other states have implemented

                                           7

-------
watershed restoration actions in response to the Clean Water Act's Section 319, nonpoint source
pollution program (U.S. EPA, 2002b).  These strategies focus management actions on geographic
regions rather than on specific media (e.g., air or water).
       The prominent aspects of the watershed approach and comparable aspects of the ERA
process are shown in Table 1.  Assessment is one of the most critically important parts of
watershed management because it attempts to transform scientific data into policy-relevant
information that can support decision making and action. A watershed assessment may be
initiated by any concerned party in the watershed, including citizen groups, regional coalition
members, municipal officers, decisionmakers, litigants, ecological scientists, and a number of
other stakeholders regardless of title or association. Watershed ERA provides a basis for
estimating ecological effects as a function of exposure to various sources of stress in the
watershed. This approach helps environmental managers focus analyses on the highest priority
problems in the watershed.

       Table 1.  Prominent aspects of the watershed approach and comparable
       aspects of the ecological risk assessment process
 Watershed Management            Ecological Risk Assessment
 Geographic focus                    The scope of the assessment is identified during
                                     problem formulation.
 Partnerships and stakeholder          Interactions with managers and stakeholders are
 involvement                         encouraged, particularly during problem formulation
                                     and risk characterization phases.
 Continuous improvement based on    The analysis and risk characterization phases provide
 sound science                       and organize scientific information relevant to
                                     management decisions.
Source: Serveiss (2002).

2.3.  THE WATERSHED ERA PROCESS
       Watershed ERA integrates the central aspects of the watershed approach, hydro logic ally
defined geographic boundaries, stakeholder involvement, and sound management, with ERA.
Although the ERA process is presented in a linear fashion, in reality it is an iterative process that
includes a regularly occurring dialogue between scientists and managers. This iterative process is
represented by the two-sided "arrows" in Figure 1 and by the vertical box at the right of the
diagram denoting tasks that may need to be revisited along the way. Stakeholders should be
involved to ensure relevancy to their concerns. The iterations and risk communication take on

-------
    Planning
  (Risk Assessor/
  Risk Manager/
 Interested Parties
    Dialogue)
                      Ecological Risk Assessment
PROBLEM FORMULATION
                        CO
                        CO
    Characterization
          of
       Exposure
Characterization
     of
  Ecological
    Effects
                               V         V
                        RISK CHARACTERIZATION
                              Communicating Results
                                to the Risk Manager
                               Risk Management and
                             Communicating Results to
                                 Interested Parties
                                                                     CD
                                                                     O
                                                                     CD
                                                                     Cfl
                                                                     en
3
2. .0
O E.


So
CO Q)
— — -
                                                                     CD

                                                                     S-
                                                                     CD
                                                                     T)
                                                                     3
                                                                     O
                                                                     CD
                                                                     V)
                                                                     V)
Figure 1.  The ecological risk assessment framework.

Source: U.S. EPA (1998a).

-------
greater importance in a watershed assessment with multiple stressors, pathways, ecological
resources, and diverse managerial and stakeholder interests (Serveiss, 2002). Based on the
nature and timing of risk management needs, scientific findings, and resources available for the
assessment, various phases of watershed ERA may need to be revisited repeatedly or skipped
entirely.  For presentation purposes, this report covers the six steps of ERA sequentially:
planning, problem formulation, risk analysis, risk characterization, risk communication, and risk
management.

Planning
       During planning, the stakeholders are identified and assembled with the goal of sharing
information. Their input is needed to establish the focus, scope, and complexity of the risk
assessment.  The workgroup needs to identify the boundaries for study (see Chapter 3).

Problem Formulation
       During problem formulation, assessment objectives are refined and information about the
watershed and resources at risk is collected and shared. In addition, assessment endpoints are
selected. A conceptual model is created showing pathways between sources, stressors, effects,
and assessment endpoints.  The associated risk hypotheses are described and environmental
management options are developed based on stakeholder input.  Finally, an analysis plan is
developed to guide future steps of the assessment (see Chapter 4).

Risk Analysis
       The objectives of the risk analysis phase are to gain a better understanding of (1) the
extent to which ecological resources have been or will be exposed to the most important
environmental changes resulting from human activities, and (2) what effects are likely to occur
or have already occurred as a result. Characterizations of exposures and effects can take the
form of graphs, models, maps, or other illustrations of the relationships among sources, stressors,
and measures of effect. These characterizations can be complex because of multiple interactions
among the various watershed ecosystem components (see Chapter 5).

Risk Characterization
       Risk characterization describes the linkages between exposure and potential effects. The
lines of evidence supporting findings need to be presented  along with the uncertainty. Ideally,
risks are  associated with the various management alternatives to support the risk management
process (see Chapter 6).
                                           10

-------
Risk Communication
       Risk communication begins with planning, then continues through the assessment, and
concludes with the presentation of results. Scientists or assessors share information with
managers or stakeholders to ensure that the most fruitful analyses are performed to provide
meaningful information for environmental management (see Chapter 7).

Risk Management
       Risk management is the decision making step of the process. Attempts are made to select
management options that will reduce or minimize risk and achieve objectives that have societal
value (see Chapter 8).
                                           11

-------
                                 3. PLANNING PHASE

       To make sure ERA results support decision making, risk assessment participants plan their
activities at the start of the assessment process. During planning (Figure 1), scientists, managers,
and stakeholders discuss the focus, scope, and complexity of the risk assessment.  Participants
should determine whether watershed ERA is an appropriate tool to address environmental
management concerns because a sound scientific approach is not without cost.
       Scientists, managers, and stakeholders all play a role in watershed ERA.  Scientists (e.g.,
risk assessors, hydrologists, and ecologists) need to communicate what they can realistically
provide to the managers and the level of uncertainty.  Managers must describe why the risk
assessment is needed, its relevance to regulations, and what they expect to do with the
information they will receive.  They must also  identify and include the opinions of all appropriate
stakeholders, such as ecological and socioeconomic concerns. Managers then need to develop a
list of environmental goals and objectives that reflect  the stakeholder input.  Without this
information, assessment results may not be useful to decision making. The success of a watershed
ERA strongly depends on the  quality of communication that occurs during this initial planning
process.
       The watershed approach (U.S. EPA, 1995a) defines a similar, although less linear and
more problem-oriented, planning process (Figure 2).  The watershed approach is contingent on
building a project team and obtaining public support.  One of the team's initial activities is
defining the problem: generating a mutually agreeable problem statement that defines future
activities. Specific goals and environmental objectives are then developed to address perceived
problems. Goals and objectives are based on the condition or vulnerability of valued resources
and beneficial uses, the needs  of the ecosystem, and stakeholder's needs (U.S.  EPA, 1996).
       The key aspects of planning, as outlined in this chapter, include:

       •  Identifying and including stakeholders
       •  Setting environmental management goals  and objectives
       •  Defining management options
       •  Determining the focus, scope, and complexity of the risk assessment

3.1.  IDENTIFYING AND INCLUDING STAKEHOLDERS
       The planning phase for a watershed assessment is especially complex because a
watershed typically overlaps multiple jurisdictions that are managed by organizations with
divergent goals and responsibilities and inhabited by numerous stakeholders with varied
interests.

                                            12

-------
                                                Planning Phase: \
                                               Focus and Scope )
         Figure 2. Relationships between management or ecological risk
         assessment (ERA) goals and objectives, management options, the
         planning phase, and subsequent problem formulation in
         Watershed ERA.
       Depending on the perspectives of the stakeholders involved, a different set of goals and
objectives may emerge. Stakeholder participation may become complex if there are numerous
stakeholders or some of them are interested in only a limited range of issues (Glicken, 2000).
The inadvertent exclusion of a stakeholder group may influence a group's decision to accept or
reject the outcome of the process.  Excluded stakeholders could even take legal actions, such as
filing a citizen lawsuit under the Clean Water Act to challenge the process outcomes. Therefore,
stakeholder involvement needs to be balanced against the limited resources available for
watershed assessment and management (U.S. EPA , 2001a).
       Participants in the watershed ERA process should include all regulatory or resource
agencies (state, federal, local) with responsibilities for protecting and managing the water body
and any parties whose authority will be needed to implement a management plan (e.g., local and
county officials with zoning oversight). Nongovernmental organizations (such as watershed
associations or councils, river watch citizen groups, volunteer monitoring group, educational and
research institutions, industries, and agricultural associations) all have a stake in watershed
management. Besides organized groups, other stakeholders include  landowners, those who use
the watershed, and those whose participation is essential to successful management. In some
instances, stakeholders may be hundreds of miles away from the assessment (e.g., bird watchers
                                           13

-------
concerned about migratory waterfowl).  The nongovernmental and unaffiliated stakeholders may
have objectives that are very different from those of the regulatory agencies (e.g., minimizing
restrictions on land use, resource development, or waste disposal). Although such social, legal,
and economic objectives may be in conflict with some environmental objectives, they are still
relevant concerns that need to be considered (Stahl et al, 1999). To help assess these tradeoffs, it
maybe helpful to involve environmental economists.
       Participants on a watershed assessment team contribute different resources that include
socioeconomic information, historical data, scientific expertise, and assets to conduct the
assessment. Team members of the Big Darby Creek assessment project included biologists, a city
planner, and environmental scientists (Cormier et al., 2000). Team members performed literature
reviews and frequently consulted or interviewed experts in other disciplines. The team interacted
regularly with the Big Darby Partners, a group of state agencies, representatives from The Nature
Conservancy (TNC), and farmers concerned about the future of the working landscape.
       The various EPA-sponsored watershed assessment teams found that continuity  of project
contact people, face-to-face contact, and formal and informal group discussions were important to
building the trust necessary to engage stakeholders.  Continuity with federal or state leaders is
also important. A watershed assessment coordinator is generally needed to keep participants
focused and on track, and to communicate progress to the stakeholders (U.S. EPA, 1997b).
       Stakeholder exclusions may be acceptable if a trusted member of another group
adequately represents their opinion. In the Clinch and Powell Valley assessment (U.S. EPA,
2002a), for example,  an interdisciplinary workgroup was established with representatives from the
U.S. Fish and Wildlife Service (FWS), Tennessee Valley Authority (TVA), TNC, EPA, and other
state and federal agencies. Although many other stakeholders were not directly involved in the
Clinch and Powell assessment, TNC was able to represent their concerns because TNC interacts
regularly with the public and used results from a recently completed survey. EPA provides
further guidance on identifying and engaging stakeholders (U.S. EPA, 2001b, 2000c).

3.2.  SETTING ENVIRONMENTAL MANAGEMENT GOALS AND OBJECTIVES
       Managers must  implement decisions to achieve environmental management goals.
Selected decisions often begin as one of several management options identified during the
planning phase. Clear management goals must be set before the assessment team jumps into
analysis.  It is quite common that assessments often hastily include a few obvious goals and then
devote most of the effort to  data gathering and analysis (Reckhow, 1994).  Valuable data may be
collected, but if the goals are poorly defined, the data may not be useful in management decisions
(Ward, 1996).
                                           14

-------
       Before beginning the actual assessment, managers, with input from scientists and
stakeholders, need to develop watershed management goals. Some goals are clearly spelled out in
legislation and regulation (e.g., support for designated uses under the Clean Water Act). Other
times, goals may not be defined, such as "apply an ecosystem approach to management."
Elements of existing goal statements from watershed councils, conservation plans, or local growth
planning strategies should be incorporated where appropriate. Goals may relate to protecting,
restoring or maintaining a species, community, or ecosystem, and they should be explicit and
quantifiable.
       When goals  are broad, it may be useful to break them down into multiple management
objectives.  Objectives may translate the goals-which may be very general, abstract, and
impossible to measure-into specific characteristics that are useful for deciding among
management alternatives. The process of defining objectives generally reflects a stakeholder
vision of the future condition of the watershed (U.S.  EPA, 1995a), and plentiful examples are
available from successful watershed projects (U.S. EPA, 1997a).
       It may be useful to develop lists of draft objectives representing all stakeholder
perspectives using a facilitated interview process with individual members of a stakeholder
advisory group (McDaniels, 2000; Gregory and Keeney, 1994).  A watershed coordinator (or
watershed assessment coordinator) can then condense the lists of draft goals and objectives into a
single coherent set.  Frequently, stakeholders have differences, and sometimes they have
contradictory perspectives. Some goals may be more politically favorable than others.  A group
could start with general environmental goals (e.g., control eutrophication and protect salmonid
fishery) and then make them into specific objectives (e.g., prevent nuisance blue-green algal
blooms and provide habitat for successful trout spawning).  Reckhow (1994) provides an example
of the objectives hierarchy process for Lake Okeechobee. At the top level is  the overall goal to
"manage the eutrophication of Lake  Okeechobee." The next level is subdivided into nine specific
goals, ranging from  "minimize costs" to "protect threatened and endangered species." These
goals are further subdivided by defining specific objectives, such as maintaining the populations of
specific wading bird species. As the objectives become more refined, they are likely to naturally
point the way to useful measurement endpoints for use in the analysis.
       To illustrate  the process used in establishing a goal and objectives, Big Darby Partners,
composed of managers and a stakeholder group, developed a management goal, "protect and
maintain native stream communities of the Big Darby Creek ecosystem" (Cormier et al, 2000).
Three objectives for accomplishing that goal were established:

       1.  Attain water quality criteria for designated uses throughout the watershed.
                                            15

-------
       2.  Maintain exceptional warmwater criteria for stream segments that were designated
          between 1990 and 1995.
       3.  Ensure the continued existence of native species in the watershed.
       These were very useful objectives because Ohio's Water Quality Standards specifically
link water quality to the ability of a stream to support and maintain native species. For instance,
several streams in the watershed, including the Big Darby and Little Darby, are classified as
Exceptional Warmwater Habitat waters. These water bodies have species  composition, diversity,
and functional organization comparable to the 75th percentile of statewide designated least
impacted reference sites.
       The Waquoit Bay interdisciplinary and interagency assessment team developed an
environmental management goal for the watershed through a multistep planning process (Serveiss
et al, 2004).  The process included a public meeting to initiate the assessment and understand
public values, a meeting of team members to develop the goal and more specific objectives, and a
meeting with local resource managers to refine the goal and objectives.  The following goal was
mutually agreed upon: Reestablish and maintain water quality and habitat conditions in Waquoit
Bay and associated wetlands, freshwater rivers, and ponds to (1) support diverse, self-sustaining
commercial, recreational, and native fish and shellfish populations and (2) reverse ongoing
degradation of ecological resources in the watershed.
       The Waquoit Bay assessment team identified 10 objectives that more explicitly stated the
kinds of management results implied in the general goal (U.S. EPA, 2002c):

       1.   Reduce or eliminate hypoxic (low oxygen) level or anoxic (no oxygen) events.
       2.   Prevent toxic levels of contamination in water, sediments, and biota.
       3.   Restore and maintain self-sustaining native fish populations and their habitat.
       4.   Reestablish viable eelgrass meadows and associated aquatic communities in the bay.
       5.   Reestablish a self-sustaining scallop population that can support a viable fishery.
       6.   Protect shellfish beds from bacterial contamination that results in bed closures.
       7.   Reduce or eliminate nuisance macro algal growth.
       8.   Prevent eutrophication of rivers and ponds.
       9.   Maintain diversity of native bio tic communities.
      10.   Maintain diversity of water-dependent wildlife.

       Ideally, these objectives would be quantified to evaluate environmental improvements
from an established baseline.  For instance, objective 1 above could have been quantified as
                                            16

-------
reducing the number of days per year in which dissolved oxygen in any part of Waquoit Bay
declined to below 2 parts per million within 5 years. For objective 4, the quantified objective
could have been to increase eelgrass cover from 10% to 20% of the bay's water surface area
within 5 years.
       Although reaching agreement on watershed goals may delay the start of the assessment,
reaching agreement among diverse interests is valuable for performing the most useful analyses
and for selecting and implementing the most relevant management options.

3.3.  DEFINING MANAGEMENT OPTIONS
       For assessment findings to be useful, there should be a set of management actions or
options that could be implemented pending the results of the assessment. Managers usually must
take action to achieve environmental management goals, although no action and passive
approaches are also available.  A preliminary identification of management options or alternatives
should take place as part of the planning phase (U.S. EPA, 1998a).
       The Clinch and Powell assessment team identified several management actions that could
be implemented to help attain their environmental management goal (Serveiss, 2002), including
containing and treating mining runoff, implementing agricultural best management practices,
installing roadside spill protection devices, and improving treatment of wastewater discharges.
Inclusion of management options early in the ERA process helped them select the spatial scale  on
which to focus the assessment (i.e., upstream of Norris Dam, TN) and the impacts from mining
and agriculture.  For more complex watershed assessments, it may be necessary to determine
baseline risks as an initial goal, with subsequent iterations evaluating risks relevant to specific
management alternatives.
       Identification of management  alternatives in the planning phase is consistent with the
watershed management process, in which the stakeholder objectives and management options
together define the context for the management effort.  The watershed approach (U.S. EPA,
1995a) further emphasizes this linkage by combining "setting goals and identifying solutions" into
a single step  in the process.

3.4.  DETERMINING THE FOCUS, SCOPE, AND COMPLEXITY OF THE RISK
     ASSESSMENT
       For most natural resource management decisions, ample time and resources to conduct a
rigorous assessment of the impacts of multiple stressors are rarely available, thus limiting the
potential scope of the assessment. The focus and scope of the assessment should be determined
by how sure  a manager needs to be to choose an appropriate management action.  Some
decisions, due to limited resources or urgency, may require only information obtained during

                                           17

-------
problem formulation, while more costly or controversial decisions may require more detailed
evaluations. Spatial and temporal boundaries need to be established at the start of the process,
and these should be both relevant to stakeholder objectives and practical from an analysis
standpoint. The level of effort necessary to reach a decision ideally would be agreed upon early
during planning, before conflicts arise over selecting the analyses to be performed. The key to
this challenge is obtaining consensus on where (what size watershed), when (what time period),
and how much (cost) to study.
       Using a tiered approach can help control the level of effort. Each tier would represent
increasing levels of complexity and investment and decreasing levels of uncertainty.  When tiers
are used, specific descriptions of management questions and decision criteria for each tier should
be included in the analysis plan (U.S. EPA, 1998a).
       Another approach for reducing the level of effort is to concentrate on one portion of the
watershed. Only the free-flowing upper portion of the Clinch and Powell Rivers was analyzed
because this was believed to be the major refugia for a number of native aquatic species (U.S.
EPA, 2002a).  It was also decided to focus limited resources on analyzing previously collected
data and not to collect additional samples.

3.5.  PRODUCTS OF THE PLANNING PHASE
       Planning is complete when participants have reached consensus on the following:

       •   Environmental management goals and objectives for the watershed
       •   Potential management options to be considered pending assessment results
       •   Objectives for the risk assessment, including criteria for success
       •   Focus and scope of the assessment, including the geographic boundary
       •   Technical and financial resources to be invested in the assessment

       Ongoing planning activities that will sustain the assessment include coordination of
participants, assignment of administrative or data collection tasks, scheduling regular meetings,
and setting due dates for assessment products.  Much planning occurs early, but because this is an
iterative process, planning activities are often revisited. Management goals, objectives, and
options along with the assessment, focus, scope and complexity can all change based on new
information.
                                            18

-------
                        4. PROBLEM FORMULATION PHASE

       Problem formulation develops and evaluates potential causes of ecological effects. This
phase provides the organizing framework upon which the entire ERA depends (U.S. EPA,
2000b). Problem formulation begins by reviewing available information and describing and
delineating the place and system of interest.  Relevant information includes habitat and ecosystem
type, general geography, major animal and plant groups, location, boundaries, and dominant land
uses.  The goal is to assemble information on the potential stressors contributing to either
observed or potential alterations in watershed conditions and processes. The assessment will
strive to improve the understanding of these relationships, provided resources to do so are
available.  In some cases, for complex systems such as watersheds, completing just the problem
formulation phase may yield an effective standalone management product without proceeding to
the risk analysis, risk characterization, risk communication, or risk management phases (Serveiss,
2002).
       Problem formulation leads to causal analysis (i.e., what processes could have caused
deleterious environmental effects). A comprehensive understanding of the ecological interactions
within a watershed is one of the most powerful tools  for informed planning, participatory decision
making, and logical problem formulation. The paradigms that have been developed to explain the
structure and function of stream and river ecosystems serve as examples of the basic knowledge
needed by the practitioner.  These paradigms include  the river continuum concept (Varmote et al.,
1980), the flood pulse concept (Junk et al.,  1989), influences of the riparian corridor (Stewart et
al., 2001), the nutrient spiraling concept (Newbold et al., 1982), the
serial discontinuity concept (Ward and Stanford, 1983), and the patch dynamics concept
(Townsend,  1989).
       The following basic elements of problem formulation are outlined in this chapter:

       •   Identifying stressors and sources
       •   Selecting assessment endpoints and other  measures that will be needed to quantify
          risks
       •   Developing a conceptual model describing predicted relationships among sources,
          stressors, and assessment endpoints
       •   Evaluating assessment endpoint
       •   Creating an analysis plan to guide the next phase of the assessment
                                            19

-------
4.1.  IDENTIFY STRESSORS AND SOURCES
       Stressors are defined as any chemical, physical, or biological entity that can cause an
adverse effect on an assessment endpoint. Examples of stressors include pesticides, altered
stream flow, and invasive species.  Typically a wide range of stressors may affect a given resource
or assessment endpoint.  Stressors may originate from a variety of sources, including different
human activities and natural processes.
       Watersheds are challenging from an ERA standpoint because they generally have multiple
potential stressors and sources of stressors.  Sources are actions that release or impose a stressor
on the environment. The sources of the stressors are often more manageable than the  stressors
themselves in watershed ERA and are therefore the focus of management options.
       When multiple sources are identified, it may be efficient to focus only on sources that
management can control. Nutrients (nitrogen in particular) were identified as the cause of
observed algal blooms in Waquoit Bay (U.S. EPA, 2002c), resulting in declines in commercial
fisheries. Management options focused on limiting excess nutrients originating from septic
systems or from various  land uses in the watershed, rather than from atmospheric deposition,
because nitrogen inputs within the watershed were found to be more controllable.
       Depending on the properties of the watershed and the assessment endpoints, a  large
number of stressors maybe identified.  The decision about which pathways and stressors to
analyze should be made by using the collective best professional judgment of an interdisciplinary
team with expertise encompassing the spectrum of the system under study (Foran and  Ferenc,
1999). Foran and Ferenc (1999) summarize a variety of techniques, the vast majority  of which
involve having team members score the impact of stressors on valued ecological resources. Each
individual team member  can enter a score, or groups of individuals can arrive at a consensus
through discussion. Scores can then be summed or averaged to  determine impacts, and a matrix
can be devised to determine potential remediation options (Harris et al, 1994).
       This scoring approach to stressor prioritization was used in the Waquoit Bay assessment,
in which a preliminary list was developed showing the major stressors that included nutrients,
toxic chemicals, suspended sediments, and physical habitat alterations (Serveiss et al.,  2004). An
impact matrix, with stressors as rows and assessment endpoints as columns,  is derived from the
conceptual model (Table 2). The interdisciplinary workgroup used their best professional
judgment to rank the stressors and their perceived impact on assessment endpoints on  a scale of 1
to 5.  This decision analysis method for ranking alternatives according to multiple criteria is
known as fuzzy-set logic (Harris et al., 1994; Wenger and Rong, 1987). Going through this
stressor ranking process helped the assessment team justify use of limited resources to analyze
impacts on the two stressor-assessment endpoint pathways related to  effects of nutrients on
scallops and eelgrass.  Assessment endpoints are discussed in the next section.

                                            20

-------
       Table 2. Effects matrix summarizing assumed strength of relationships
       between assessment endpoints identified (e.g., percent eelgrass cover) and
       stressors in the Waquoit Bay watershed"
Stressor
Chemical pollution
Altered freshwater
flow
Nutrient enrichment
Physical alteration
ofhabitat
Fishing pressure
Pathogens
Totals
Assessment endpoints ranking
Percent
eelgrass
1
1
5
2
1
2
12
Finfish
diversity
1
1
5
1
1
1
10
Scallop
abundance
1
1
5
2
2
1
12
Anadromous
fish
1
2
3
1
3
1
11
Wetland
birds
1
•-t
3
2
2
1
1
10
Piping
plovers
1
1
1
3
1
1
8
Fish/
shellfish
o
3
1
1
1
1
3
10
Total!
9
10
22
12
10
10
73
aEach cell represents the relative effect of a stressor on an endpoint. The ranking (1 = minor, 5 = severe) reflects
 experience with the likely effects specifically for the Waquoit Bay watershed.
Source: Adapted from Serveiss et al. (2004).

       Stressor Identification Evaluation (SIE) is a logical scientific process that has been
developed to evaluate available information, identify relevant stressors, and determine the
stressors that are most likely causing observed biological impairments. SIE (U.S. EPA, 2000d)
was specifically designed to determine probable causes(s), given that an undesirable effect has
already occurred. Drawing from standard ecological risk principles, the SIE process works
backward from the observed effect to identify possible pathways (causal linkages) that could
cause such an effect, the stressors that would lead to those causal linkages, and the sources that
would elicit those stressors. As part of this process, potential relevant stressors in the watershed
are identified, which, along with the proposed causal linkages and sources, are used to construct a
conceptual model as discussed in Section 4.3 of this report. A recent addition to SIE is the
Causal Analysis Diagnosis Decision Information System (CADDIS) which can be accessed at
http://www.epa.gov/caddis. CADDIS helps users apply SIEs to streams. CADDIS includes
information and many examples of stressor-response relationships that can help ecologists and
assessors identify possible stressors in their systems (see Section 5.1).

4.2. SELECT ASSESSMENT ENDPOINTS
       In the planning process, participants identified general goals and specific environmental
management objectives. Assessment endpoints translate these objectives into something
                                            21

-------
scientifically important and measurable. They also provide direction for future analyses.
Assessment endpoints consist of an entity and an attribute. The entity is the species (e.g.,
eelgrass), functional group (e.g., piscivores), community (e.g., benthic invertebrates), or
ecosystem (e.g., lake) that must be protected. The attribute is the characteristic of the entity that
is important to protect. For example, for eelgrass it could be acres of coverage, and for a lake it
could be frequency and severity of algal blooms.
       It is a challenge to select the most useful assessment endpoints. Watersheds have a
diverse array of stakeholders who  may have competing management objectives. To the extent
possible, assessment endpoints should represent the range of objectives of all stakeholders.  They
should be selected based on three  criteria:  their relevance to the management objectives, their
importance in the ecosystem, and their susceptibility to stressors (U.S. EPA, 1998a).
       For example, in the Clinch and Powell assessment, protection of native and endangered
mussels and fish was an ecologically relevant management objective. Two assessment endpoints
were examined in that ERA:  recruitment and reproduction of native mussels; and richness,
distribution, and abundance of native fish (Diamond et al, 2002; U.S. EPA, 2002a). Native
mussels and fish were assumed to  be sensitive to the multiple stressors in the watershed.
       It is important to tie the assessment endpoint to the overall management goal to make the
results directly applicable for resource management.  If the overall goal is protection and
restoration of native and endangered mussels, for example, the assessment endpoint might specify
a relevant well-described species and an attribute representative of the number of individuals of
this species in the area of interest.  Such specificity in the assessment endpoint can reveal the types
of analyses or models needed to determine causality and risk in the risk analysis phase, and will
help diffuse potential controversies that may arise among different stakeholders.
       In the Waquoit Bay ERA,  all of the assessment endpoints were potentially vulnerable to
certain watershed stressors (see Table 2), and they reflect the management objectives and overall
goals: (1) support diverse, self-sustaining commercial, recreational, and native fish and shellfish
populations; and (2) reverse ongoing degradation of ecological resources in this watershed
(Serveiss et al., 2004).  However,  only two assessment endpoints (eelgrass cover and scallop
abundance) were analyzed in detail because of their relationship to the predominant stressor.

4.3. DEVELOP CONCEPTUAL MODEL
       Conceptual models describe key causal relationships that will be evaluated in the  risk
assessment.  They include a diagram of the predicted relationships between ecological effects and
stressors, and ideally they also include a written description of the relationships. Conceptual
models are necessary for watershed and landscape assessments because they describe the
multiple physical, chemical, and biological stressors and their sources in a system, as well as the

                                            22

-------
                   Source
         Measurable Change in
            Endpoint Attribute
     FigureS.  Elementary conceptual model
pathways by which they are likely to affect
assessment endpoints (Suter,  1999) (see Figure 3).
At this point in the watershed ERA, these potential
causal relationships are based on professional
judgments and knowledge and are not yet
quantified.
       Sources, stressors, and effects are
interconnected; thus, it is a challenge to evaluate
their relationships to one another. Furthermore,
each pathway and each endpoint may have
professional advocates and detractors. Therefore,
achieving a focus for the assessment may be
difficult.  Conceptual models help address this
challenge.
       Developing the conceptual model provides
a forum for discussing causal pathways, a
framework for explaining the hypothetical
relationships and the scope of the assessment, and a structure for the forthcoming analyses.  Many
specialists may find it easier to see the big picture with the aid of a conceptual model.  This
systematic process provides a forum and documentation to management actions and helps to
efficiently bring new personnel up to speed. U.S. EPA (2005) provides a more in-depth
discussion on the applicability of conceptual models
       Conceptual model development has been identified as the single most valuable component
of EPA's watershed ERA prototype assessments (Serveiss, 2002; Butcher et al, 1998). For
instance, in the assessment of the middle segment of the Snake River, the conceptual model and
associated risk hypotheses provided a common basis for coordinating the concerns of EPA, FWS,
the Federal Energy Regulatory Commission, and the public (U.S. EPA, 2002d). In the Big Darby
Creek assessment, group efforts to develop the conceptual model and risk hypothesis were
particularly valuable for communicating expectations within the technical workgroup and to the
stakeholders.
       An interdisciplinary scientific team, including individuals with local ecological
knowledge, should develop the conceptual models for the watershed based on previously
collected information and their best professional judgment.  Early models may be simple, but they
provide a basis for predicting relationships between sources, stressors, and effect, and for
identifying knowledge gaps.  The predicted relationships provide qualitative forecasts of the
impact of stressors on resources, and qualitative hypotheses may also be helpful for decision
23

-------
making. These predictions may be used to prioritize problems and help address TMDLs, which
are pollution budgets to reduce loadings of pollutants that exceed water quality criteria.
       Figure 4 is an example of a complex conceptual model with various pathways organized
around stressors. Each stressor has a patterned line that illustrates a pathway connecting its
sources to possible effects and selected endpoints.  Each model component is represented by a
different geometric shape to aid interpretation. This is a broad-based model that provides a
framework for the risk assessment and an overview of ecosystem processes.  The diagram shows
only stressors and effects thought to be potentially important in the Waquoit Bay watershed. The
exposure pathways, from the source of stressors to valued resources, are the possible risk
hypotheses to be analyzed as part of the risk assessment.
       Initial efforts at producing a conceptual model may result in a highly complex set of
pathways and linkages. These models are necessarily complex to capture all the pathways
between sources, stressors, and effects.  It is a challenge to balance complexity with the clarity
necessary to understand a diagram with a complicated web of boxes and arrows.  One way to
achieve this balance is to use hierarchical nested conceptual models (Suter, 1999). Hierarchical
nested conceptual models are more detailed and may contain more pathways or information than
space permits viewers to see clearly in a general model. Depending on the need, a hierarchical
model could be constructed for just one stressor or just one assessment endpoint. These single-
stressor and single-endpoint models can be used to provide viewers with a simplified overview
along with the capability to examine particular pathways in more detail.
       Another option for achieving clarity and complexity is to have  the model evolve as more
information is developed during an assessment (U.S. EPA, 2005).  The initial conceptual model in
the Clinch and Powell assessment provided a framework that was used to track the progress of
analyses. Areas in need of further data collection or analysis  were  identified and prioritized.
Additional complexity was added over time as the assessment team developed a better
understanding of source-stressor-endpoint relationships.
       Different forms of communication need to be used for those involved. Scientists will need
all the details.  For the nontechnical audience, it is better to produce a simplified version of the
conceptual model showing only significant linkages. For example, the generalized version of the
Waquoit Bay conceptual model (U.S. EPA, 2002c) proved to be a powerful communication tool
during the risk assessment phase of the process. Its perceived value to stakeholders is illustrated
by the decision to display the conceptual model at the Waquoit Bay National Estuarine Research
Reserve Visitor Center.  The more complex conceptual model can be held in reserve and used to
help answer more specific questions.
       Another advantage of developing a conceptual model, such as  Figure 4, is that it helps
elucidate cascading or secondary effects that otherwise may not be immediately apparent.

                                           24

-------
    Agriculture      Urbanization and industrial development     Marine activities
Pesticides/
Herbicides

Fertilizers

Lawnsf
gardens

Itnperv.
surfaces
'Chemical v
  pollution/
                                          Pathogen
                 Altered
               freshwater
                  flow
                     Siltation
                                Nutrient  '
                               enrichment
                                                                       Fishing
                                                                       mortality
                                                            ^hysical alteratio?
                                                                of habitat
                              /Increased
                              macroalgae,'
                                         / Increased
                                        :phytoplankton-;
      Finfish and
       shellfish   i
   I contamination/
/
Eelgrass/'scallop
   habitat loss
                                                   rBarrier beach
                                                       erosion
                                                                         Reduction
                                                                         of stocks
                                                             'wetland habitat/
                                                                  loss     /
      Brook trout
     reproduction
           issue
       contaminant
           levels
                                                                     Finfish
                                                                  diversity and
                                                                   abundance
                            Eel grass
                            cover
            lover distribution
            and abundance
                    -XT/Vetland bir
                   (   distribution and
                                 Scallop
                               abun dance
                                                                                        05
                                                                                        O
                                                                                        c
                                                                                        o
                                                                                        (D
                                                                                     (D
                                                                                     V)
 m
 i
 o
 r-f
 V)
 m
 3
 Q.
13
 g
 3"
Figure 4. Conceptual model of the Waquoit Bay watershed.
Rectangles represent sources of stressors, hexagons are the specific stressors to the system,
trapezoids represent the effects of those stressors, and the ellipses indicate specific endpoints that
are affected.

Source: U.S. EPA (2002c).
                                           25

-------
For example, in Waquoit Bay, excess nutrient input may exert indirect effects on commercial
fisheries by causing algal blooms that reduce light levels to the point that some submerged
aquatic plants important as habitat for juvenile fish cannot survive (Serveiss et al, 2004).  In this
case, habitat loss is the stressor impacting the fish, but knowledge of the whole chain of events is
necessary to take cost-effective corrective action (U.S. EPA, 2002d).

4.4. EVALUATE ASSESSMENT ENDPOINTS
       Assessment endpoints need to be evaluated using observations or modeled predictions.
Choosing between management options typically involves a prediction of the response of an
assessment endpoint to the proposed management action. However, watershed assessment
endpoints themselves are not easily measured,  and the response of endpoints to changing
conditions is often difficult to predict. For instance, an assessment endpoint of a fishery might be
the spawning success of salmonids. This is difficult to measure directly for an entire watershed,
although associated information, such as counts of young-of-year from a set of monitoring sites,
may be feasible.
       Instead of measuring and predicting responses of the assessment endpoint directly, it is
often necessary to identify surrogate measures or measures of effect that can stand in for the
assessment endpoint. Measures of effect are quantifiable changes in an attribute of an
assessment endpoint or its surrogate in response to a stressor to which it is exposed. Meaningful
measures of effect will have the  following characteristics (Butcher and Craig, 1998; U.S. EPA,
1998a):

       •   They will link objectives to stressors (and changes in stressors) and thus correspond to
          a risk hypothesis defined in the conceptual model.
       •   They will be sensitive to changes in the condition of the  objective.
       •   They will reflect susceptibility to individual stressors.
       •   They will be meaningful to decisionmakers.
       •   They will be measurable,  or at a minimum, risk assessors will be able to semi-
          quantitatively rank them.
       •   Risk assessors will be able to predict them in response to management options.

       Measures of effect are selected for their suitability in reflecting changes likely to occur in
the assessment endpoint and for their ability to be measured accurately, consistently, and
economically. For example, in the Clinch and Powell assessment, an assessment endpoint was
                                           26

-------
the recruitment and reproduction of native and endangered mussel species. Because data were
not available in many parts of the watershed, a surrogate measure of effect was used, mussel
species richness. In the Big Darby watershed ERA (Cormier et al, 2000), an assessment endpoint
was "biotic integrity of the macroinvertebrate community." The applied measure of effect was the
macroinvertebrate Index of Community Integrity (ICI), a quantifiable indices with a standardized
protocol that defines the quality of the macroinvertebrate community.
       Quantified target values of assessment endpoints should be specified, where applicable, as
a means of evaluating and communicating whether a given management alternative meets
management objectives.  This is particularly important in watershed management, where defined
target values can serve as a shorthand to communicate how different management options
compare in their ability to meet multiple management objectives.  Target values for some
measures are already established directly in regulation.  For instance, numeric water quality
criteria must be met for a management plan to be acceptable, although it may be determined that a
more stringent target is appropriate for specific objectives.  Where target values are not already
defined, a variety of mechanisms can be used to determine an appropriate value.  In many cases,
the target value will emerge from the ERA itself. Sometimes a narrative standard can be used. It
is also possible to set target values by comparing the conditions in the watershed to conditions in
an appropriate reference (relatively unimpacted) watershed (U.S. EPA, 1999a), conducting user
surveys (e.g., Smeltzer and Heiskary,  1990), or using an existing classification system (e.g.,
Vollenweider and Kerekes, 1980).
       It is desirable to select assessment endpoints and determine whether alternative measures
or target values are available before collecting additional data or analyzing them. The next step is
to plan the analysis before moving into the risk analysis phase.

4.5.  CREATE ANALYSIS PLAN
       The analysis plan must describe the objectives of the analysis phase, which are to gain a
better understanding of (1) the extent to which assessment endpoints are exposed to important
environmental changes from human activities and (2) the effects that are likely to occur as result
of exposure. The pathways more likely to impact assessment endpoints or achieve management
objectives, along with those more amenable to corrective action, should be investigated. The plan
should describe possible results and uncertainties and how this information will be communicated.
An integrated and comprehensive data analysis plan also will ensure that all the data needed to
define exposure scenarios and pathways to  assess ecological risk are clearly identified (Cook et
al., 1999). The costs of data collection  and analysis have to be weighed against the anticipated
risk reduction benefits.
                                            27

-------
       The analysis plan provides a scientifically defensible process to select specific pathways
between sources and assessment endpoints for further study in developing meaningful
conclusions. The analysis plan should describe the quantifiable (either ordinal or continuous)
measures (i.e., both assessment endpoints and stressors/sources) that will be analyzed.
       The analysis plan could also include a pilot study intended to test the proposed approach.
The Copper Creek subwatershed was studied before the risk assessment team moved on to assess
the entire Clinch and Powell watershed (U.S. EPA, 2002a). The analysis plan was modified in
response to the pilot study (see Section 5.7 for more details).
       When problem formulation is complete, the risk assessment team and stakeholders should
have a clear understanding for the assessment and a plan for the analysis phase.  Even if the
remaining assessment phases are not carried out, problem formulation alone is extremely
valuable to watershed management because it summarizes existing relevant ecological
information  for the watershed,  significant data gaps, and potential risks in an organized manner
(Serveiss, 2002).
                                            28

-------
                               5.  RISK ANALYSIS PHASE

       The risk analysis phase investigates the most important stressors and exposure pathways
and predicts how assessment endpoints respond to stressors under different exposure scenarios.
Risk analysis tasks may include collecting data, interpreting historical data, modeling, and
statistical analysis. The risk analysis phase ultimately characterizes exposure and effects.
       During exposure characterization, the assessor seeks to describe the intensity of the
contact or co-occurrence between the stressor and the assessment endpoint (U.S. EPA, 1998a).
This involves linking stressors with their sources and then describing where, how, and when the
stressors occur in the environment. In traditional ERAs,  exposure is commonly estimated by
measuring or modeling amounts of stressors and combining these estimates with assumptions
about the route of exposure to the assessment endpoint.  Spatial and temporal distributions of
both the endpoint and the stressors are considered.
       During ecological effects characterization, stressor-response data and cause-and-effect
relationships are evaluated and extrapolations are made from one set of data to the system being
studied.  Statistical techniques or mathematical models may be used to quantify and summarize
the relationship between stressor and effect.  No matter which approach is used, biological
monitoring data are very pertinent in identifying the biological consequences of human actions and
provide an essential foundation for assessing ecological risks (Karr and Chu, 1997).
       The analysis phase of watershed ERA is challenging, because watershed assessments must
deal with direct and indirect effects on multiple endpoints from multiple physical, chemical, or
biological stressors (Suter,  1999; Hunsaker et al, 1990).  The traditional risk assessment
paradigm was developed for evaluating risks to single species, or risks from specific chemicals,
for which quantitative information is often available or feasible to collect as part of the ERA. For
these kinds of single stressor and effect assessments, during the risk analysis phase calculated
relationships may be expressed as a stressor-response curve.  Although desirable, it may be
impractical to develop such a curve in a watershed assessment because of a host of confounding
factors, including the presence or absence of other stressors,  abiotic influences on the stressor
effect (e.g., dissolved carbon and pH affects the toxicity of many metals), and spatial and temporal
heterogeneity of the assessment endpoint itself.  In many cases, it will  be too costly to attain the
data for each exposure and effects pathway (Serveiss, 2002). Thus, risk analysis for watershed
ecological assessment is likely to require some deviation from the ideal ERA process.
       For watershed ERAs, rather than assessing exposure and effects separately, these analyses
are usually aggregated. This can lead to merging risk analysis with risk estimation (typically part
of the risk characterization  phase) because the assessor frequently compares source or stressor
data with biological effects  data to draw conclusions.  As  a result, in watershed ERA the

                                            29

-------
distinction between risk analysis and risk characterization can become blurred.  Many of the
points presented in this chapter on risk analysis could also apply to risk characterization.
       Ecological effects are analyzed by describing stressor-response relationships, evaluating
evidence for causality, and linking measurable effects to the assessment endpoints identified
during problem formulation (Norton et al, 2002a).  These components can be developed in any
order, and the emphasis may be different depending on whether the objective of the assessment is
to predict the effects associated with future change or to retrospectively analyze the causal factors
influencing the current state of ecological resources. Tools and challenges associated with
watershed risk analysis include the following:

       •   Identifying stressors and evaluating causes of impairment and associations between
           causes and effects (Sections 5.1 and 5.2)
       •   Working with chemical or nonchemical stressors (Sections 5.3 and 5.4)
       •   Using categorical data or source data or ranking risks because data are limited
           (Section 5.5)
       •   Dealing with lots of uncertainty (Section 5.6)
       •   Recognizing that analysis is an iterative process (Section 5.7)
       •   Understanding secondary and indirect pathways on valued resources (Section 5.8)

5.1.  STRESSOR IDENTIFICATION EVALUATION (SIE) AND CAUSAL
     EVALUATION: IDENTIFY STRESSORS AND EVALUATE CAUSES OF
     IMPAIRMENT
       SIE is a tool that helps organize scientific information and provides a system to help
document the stressors and pathways considered in a risk analysis. The SIE process (U.S. EPA,
2000d), provides a clear and consistent method to identify and compare candidate causes and
determine which causes are best supported by the evidence.  The process uses a series of three
methods.  First, candidate causes that are not logically possible given the evidence are
eliminated. Then, conventional medical/veterinary diagnostic approaches are used to identify any
causal agents that have produced characteristic symptoms. For the remaining causes, a strength-
of-evidence analysis is used to evaluate and score all available evidence for each candidate.
Based on current experiences in using SIE in several state TMDLs, diagnostic information is
often lacking (i.e., data demonstrating a specific stressor or mechanism of effect are often
unavailable).  Instead, different types of association analysis (temporal and spatial associations
between a given stressor and effects), as well as relevant data from the published literature (e.g.,
tolerance values for a given species and stressor, Habitat Suitability Index [HSI] information, or

                                            30

-------
toxicity endpoints for a given species and chemical) are used to help support or refute
hypothesized causal linkages between candidate stressors and observed effects.
       Thus, like ERA, SIE promotes risk analyses that are transparent to all stakeholders.  In
addition, SIE helps ensure that stressors and pathways are not overlooked or prematurely
dismissed from consideration, forcing the assessors to consider and evaluate all reasonable
sources, stressors, and pathways.  SIE has been incorporated into a causal assessment to
determine probable causes of biological impairment (Suter et al, 2002), and this methodology has
been applied in several places, including the Little Scioto River in Ohio (Norton et al.,
2002a, b).
       Causal evaluation can identify factors that might be controlled by management actions to
improve environmental conditions (Norton et al., 2002a, b; Suter et al., 2002). CADDIS (U.S.
EPA, 2004) is a Web-based decision support system that can help scientists in the regions, states,
and tribes find, access, use, and share information to determine the causes of biological
impairments in aquatic systems. CADDIS is a tool that applies the SIE process to streams and
enables scientists to make causal determinations more quickly, cheaply, and defensibly.  It helps
scientists bring together relevant knowledge on physical, chemical, and biological stressors that
may be affecting the aquatic system of concern. This tool is particularly useful for watershed  and
regional ecological assessments in which scientists must often rely on stressor-effect relationships
reported for other sites or in laboratory studies. CADDIS can also help scientists more easily
identify relevant data from scientific studies conducted elsewhere and help them organize and
compare the evidence across all the candidate causes. Finally, some scientists must be able to
clearly communicate the logic of their causal conclusions.

5.2.  EVALUATE STRENGTH OF ASSOCIATION BETWEEN CAUSES AND EFFECTS
       Both SIE and watershed ERAs evaluate the strength of associations between a particular
source or stressor and the effects of concern. The strength of the association (typically evaluated
using multivariate analysis) between the stressors (or sources) and measures of effect may
indicate cause-and-effect relationships, especially if supported by other lines of evidence
(Serveiss, 2002; U.S. EPA, 2000b). In most retrospective watershed ERAs, causes of effects are
often inferred from spatial relationships derived using relatively current information. For
example, in the Clinch Valley ERA, a strong negative relationship existed between coal mining
activities (particularly coal preparation plant discharges) and both fish IBI and number of native
mussel species. The particular stressors in this  relationship were believed to be both pollutants
and sedimentation from coal fines, based on published research in the watershed. The analyses
did not suggest that every coal mining activity in the watershed would cause a certain effect on
native species, but they did suggest a relatively high probability of risk to  native species living in

                                            31

-------
close proximity to coal mining activities. Landis et al. (2000), using a relative risk model
(RRM), also inferred cause-effect relationships by associating sources with important habitats in
various subareas of the Willamette River Basin in Oregon.  This model of analysis yields a
ranking of stressors and habitats based on the relative likelihood of occurrence, not on the
relative consequence (or effect) of occurrence (Landis et al., 2000).
       The Big Darby Creek assessment sought associations between stressors and impacts
(Cormier et al., 2000) by relying on current and past land use practices and biological
measurements taken at specific sites.  Researchers used the ICI for macroinvertebrates and the
fish IBI to represent ecological status within stream segments in the watershed.  Multivariate
analyses were used to determine relationships between index results, instream stressors, and land
use patterns in the watershed. The analysis identified community components that were
associated with specific types of stress.  For example, the percent of Tanytarsini midges and
Glypotendipes increased at sites with low and high biological oxygen demand, respectively.
Also, the percentage  of darters increased at sites with high scores for stream corridor structure
and low concentrations of inorganic nutrients.

5.3.  WORK WITH CHEMICAL STRESSORS
       Exposure and effects analysis that relate to multiple chemical stressors in a watershed
ERA can be handled  in much the same way as in single-chemical ERAs.  Both types of ERA use
nationally developed  chemical criteria (e.g., U.S. EPA, 1995b), if available, or effects thresholds
obtained from laboratory or field studies are used. A challenge in many watershed ERAs is the
likelihood of having multiple chemicals present simultaneously, any or all of which may have
caused, or could cause, effects on valued resources.  Laboratory-based chemical criteria or
thresholds are generally developed independently of other chemicals, or any other stressors in
percent, and therefore do not necessarily take into account synergistic or additive effects of
multiple chemicals. Some  field-based chemical thresholds, such as sediment quality thresholds
(Field et al., 2002; Long et al., 1998), may incorporate chemical interactions in their threshold
values. Several different tools have been developed to help address the issue of multiple chemical
contaminants, the most common of which is to express each chemical concentration as a
proportion of its criterion or threshold concentration. These proportions are then added; i.e., the
assessor assumes that the effects of the different chemicals are additive.  If the sum of the
proportions is >1.0, some  assessors have interpreted this as a potentially toxic condition. Limited
research suggests that additivity may be a reasonable assumption in many cases, but a fair degree
of uncertainty is associated with the results of this approach.
       One major uncertainty is the concentration of each chemical that is being compared to its
respective criterion or threshold.  Ideally, one has sufficient data to calculate frequency, duration,

                                            32

-------
and magnitude of concentrations exceeding a given threshold so that the entire record of
measurements is used to calculate risk. However, such comprehensive data are rarely available,
and the assessor often needs to estimate an average or maximum value with which to analyze risk
for the suite of toxic chemicals that might be present.
       Another source of uncertainty in analyzing risk of multiple chemicals is the often large
variability in sensitivity to the chemicals for different species of interest.  While the measures of
effect selected to represent the assessment endpoint may help to focus this issue in risk analysis,
chemical criteria and thresholds are typically based on many species, and the sensitivity of the
types of species that occur in the watershed may be fundamentally different. One tool that has
been beneficial, particularly in ERAs involving chemicals for which there are criteria, is a species
sensitivity distribution approach (Bossuyt et al., 2005; Newman et al, 2000).  This approach
examines the sensitivity of all species for which relevant effects data exist and evaluates the
distribution of species sensitivity to help predict the portion of the distribution of species that is
likely to be affected by a given chemical concentration. For multiple chemicals, such species
distributions of risk can be overlaid to obtain an estimate of risk due to chemicals to the aquatic
assemblage. The species distribution can also be tailored in some cases to a given region by
including only those species, or those representative species, that could reasonably occur in the
watershed. In practice, however, this approach is often limited due to a paucity of relevant
toxicological data for species of interest. In addition, certain types of species, such as freshwater
filter feeders (e.g., unionid mussels) are currently under-represented in toxicological databases
because reliable test methods were unavailable. Thus, the database of species sensitivity values
available for certain chemicals may be limited and may not include many species or genera of
concern in a given watershed.

5.4.  WORK WITH NONCHEMICAL STRESSORS
       Another challenge regarding stressors in risk analysis, particularly for watersheds, is the
fact that many common stressors are nonchemical (e.g., sediment/turbidity, habitat),  or are
chemicals in media for which readily available criteria or thresholds may not exist (e.g.,
pesticides in sediment), or are chemicals that are naturally available and necessary for life (e.g.,
nutrients).  Thresholds for these types of pollutants typically require some type of reference
condition approach.  A reference approach has been commonly used to set thresholds or criteria
for biological assemblage condition (e.g., U.S. EPA, 1999a; Karr and  Chu, 1997) and is now
being extended to other types of parameters that are strongly influenced by natural regional
conditions, such as nutrient criteria (U.S. EPA, 2000e). The reference condition approach relies
on having measures for the parameter of interest from multiple sites in the same ecological
region or ecoregion (Omernick, 1987) and that are minimally or least affected by human-caused

                                            33

-------
activities. Ideally these sites are located in the same watershed or basin as the one being
investigated. Thus, these measures ideally reflect natural variability, without the stressor of
interest present.
       The  threshold developed using reference conditions is a statistical attribute of the
measures obtained from the various reference sites.  A variety of statistical parameters have been
used to formulate a threshold. For example, EPA's Rapid Bioassessment Protocols (U.S. EPA,
1999a) recommend, and many states use, the 25th percentile IBI value from the distribution of
reference site IBIs as  the threshold for reference biological condition.  IBIs from a site that are
significantly lower than this threshold suggest an undesirable biological condition (i.e., an effect).
Clearly, this approach is sensitive to proper selection of reference sites. If more truly impaired
sites are identified as reference sites,  the established threshold will be less sensitive, and the
associated risk analysis will be more uncertain.  Also, this referenced threshold approach works
best when there are many appropriate reference sites with which to measure natural variability.
Biased estimates of natural variability are more likely as fewer reference sites are available.
Finding reference sites is a common challenge in certain regions of the United States (e.g.,
agricultural  areas of the Midwest and highly urbanized areas in southern California or the mid-
Atlantic coast).
       As described in Chapter 3, biocriteria thresholds used in the Big Darby watershed ERA
were based on ecoregional reference  conditions (Cormier et al, 2000). EPA's nutrient criteria
guidance for streams and rivers suggests using a reference approach for developing nutrient
criteria (U.S. EPA, 2000f). The EPA report recommends using either the 75th percentile of
concentrations observed in ecoregional reference sites or the 25th percentile of measurements
obtained from all sites in an ecoregion as a concentration threshold for a given nutrient.
       In some ecoregions and watersheds, reasonable reference conditions maybe difficult to
identify as a result of large-scale human-caused changes to the landscape, hydrology, and aquatic
habitats. In these cases, appropriate  stressor thresholds may need to be developed using other
tools, such as  modeled reference conditions, based on multivariate analyses of existing stressor
and biological condition data.  For example, using quantile regression (Cade et al., 1999;
Koenker and Bassett, 1978) of macroinvertebrate Ephemeroptera, Plecoptera, and Trichoptera
(EPT) taxa richness (a metric that includes many of the pollution-sensitive macroinvertebrate
species) and associated urban intensity (a combination of urban land cover and population and
road density),  resource agencies in southern California were able to identify best attainable
biological conditions despite the occurrence of high urbanization and associated stressors.  Such
analyses can help elucidate causes of biological impairments and provide the means to  identify
working "thresholds" for a given stressor.
                                             34

-------
       Risk analysis can often take the form of simulation or numerical modeling, depending on
the stressors of concern, availability of appropriate criteria or thresholds, and extent of stressor
data available. For the middle segment of the Snake River in Idaho, a life cycle analysis was
conducted for three highly valued coldwater fish (rainbow trout, mountain whitefish, and white
sturgeon), aquatic macrophytes, and endemic molluscs (U.S. EPA, 2002d).  These representative
species from three major trophic levels (fish, invertebrates, and plants) were chosen as assessment
endpoints for the analysis. A simulation model was used to develop quantitative risk estimates of
the likelihood of exceeding water quality criteria, macrophyte biomass, and habitat suitability
indices for the fish species.  The system dynamics were simulated using information on
meteorological conditions, hydrological and hydraulic conditions, biological oxygen demand,
dissolved oxygen, phytoplankton biomass, various forms of nitrogen and phosphorus,
temperature, coliform bacteria, and water depth, all obtained from several locations in the river.
Risks to fish and macro invertebrates were estimated by determining the likelihood of being above
or below coldwater biota tolerance limits. Tolerance limits are generally the natural level to  which
most native species have adapted. In addition, qualitative analyses of exposure and effect were
completed using best professional judgment and comparisons with field studies under similar
conditions. The results of these individual analyses were combined to draw conclusions in the risk
characterization phase.

5.5.  WORK WITH DATA LIMITATIONS
5.5.1. Use Categorical Data
       All of the EPA-sponsored case studies, and  most other large-scale ERAs that have been
reported, had to work with some types of data limitations, and therefore uncertainties in their risk
analyses. Given the multiple stressors in watersheds, and the relatively large spatial scale of
watersheds, most  analytical approaches will require fairly large sample sizes to account for the
many different potential sources of variation. As a  result, data requirements can be enormous for
developing meaningful associations between sources, stressors, and effects in watersheds.
       While regression-based relationships between a stressor and effect or a source and a
stressor are desirable in any ERA, watershed analyses often lack sufficient and appropriate data
with which to derive meaningful "stressor-response" curves. One alternative that has been used in
many watershed ERAs is to categorize the data before conducting analyses. Categorical data may
depict relationships more effectively than continuous (gradient) data in some cases.  Both stressor
and effect data may be categorized.  For  example, in the Clinch and Powell ERA, few if any
gradient relationships were evident between sources, stressors, and effects.  However,
relationships became more evident when the measure of effect  (i.e., fish IBI), was categorized as
impaired or unimpaired (based on ecoregional reference conditions) rather than using the IBI

                                            35

-------
values themselves (Diamond et al., 2002).  Similarly, relationships between habitat quality factors,
such as sedimentation, and effects were more apparent if habitat quality (the stressor) was
categorized as "satisfactory" or "unsatisfactory," based on the resource agency rating system.
Categorizing data for a factor, as opposed to relying on the individual data values themselves,
often results in increasing the power of the analysis; i.e., there is a higher probability of seeing a
relationship if one is there. However, stressor-response relationships may or may not be feasible
using categorical data. If only two or three categories of a source, stressor, or effect are
appropriate, analyses will often take the form of an analysis of variance (e.g., ANOVA,
MANOVA, ANCOVA) or analogous nonparametric analysis to determine statistical differences
between categories.
       Categorizing data works best when either criteria or consensus-based thresholds for the
parameter are available. For example, if the bioassessment data available are based on an index
that discriminates four or five categories of condition such as the fish IBI, then it is appropriate to
categorize the bioassessment data accordingly for analysis. Analysis of more categories in this
case would not be appropriate and could yield misleading results because the index was not
calibrated to distinguish between a greater number of categories. Categorizing stressor or source
data can be more uncertain because consensus thresholds are typically unavailable. For example,
effects on assessment endpoints as a function of distance from a potential source (a surrogate for
stressor intensity) is often a useful analysis in watershed ERAs.  However, distance from a source
may not lend itself to obvious categories unless much is known about  the fate and transport of the
stressors assumed to be present.  In this case, it may be useful to examine the cumulative
distribution of the distance of each relevant biological sampling location to its nearest source type
of concern (e.g., distance from the nearest active coal mine was examined in the Clinch Valley
ERA). Based on this distribution of distances, the assessor can then identify a few categories
(e.g.,  <1 km, 1-5 km, and >5 km), for which a sufficient number of sites exist and can be
statistically analyzed in a meaningful way. Again, care should be taken to ensure that the levels or
categories are appropriate for the measure being analyzed and the particular watershed.

5.5.2. Use Source Data
       Another common data limitation in watershed ERA is that stressor data may not be
available or in a useful form.  To deal with this limitation, risk assessors have often relied on
source data as a  substitute. Source data include such parameters as the number of a certain type
of point source or the percentage of a particular type of land use. These data are typically fairly
easy to obtain for a watershed, and they are readily mapped for analysis and presentation
purposes.  Source-effect analyses were used both for the Big Darby Creek and the Clinch Valley
assessments where land use (source)  data were compared with fish and macroinvertebrate

                                            36

-------
              (effects) data.  Originally, the assessment team for the Clinch and Powell Valley assessment
              sought associations between fish or mussel data, and stressor data such as sedimentation or water
              chemistry data, yet data were too sparse for quantitative analyses (U.S. EPA, 2002a). Instead,
              emphasis was focused more on probable stressor sources and their effects on resources (Diamond
              et al, 2002) (see Figure 5).
                      20
                      16
                      12
                       8
                       4
                       0
                             Percent urban land us. EPT
                                                                    Urban Land vs.EPT
                                12
                              °-  8
                              LU
20    40    GO    80
    % Urban / Barren
                                                     100
                                                     T
                                                                   <15     15-40     >40
                                                                   Percent Urban Land
                                      EPT Assessment us Percent Urban Area
                                                            m M»an«8D
                                                               M*an-8D
                                                                                             H»an
                                        GO
                                        50
                                      | m

                                      I 20
                                      f 10
                                         0
                 T
T
                                               Poor
                            Good
_L H«an*SD
   H«an-ID
CZl H»an*SE
   M«an-SE
 1  H*an
                                                    EPT
                     Figure 5. Significant relationships between urban land use activities and
                     invertebrate ephemeroptera, plecoptera, and trichoptera (EPT) score.
                     Relationships between the biological measure (EPT in this case) and land use cover
                     were often difficult to evaluate using linear stressor-response analyses (upper left
                     graph).  Relationships were more apparent when examined on a categorical or
                     threshold basis (upper right and lower graphs).

                     Source: U.S. EPA (2002c).
                                                          37

-------
       By comparing a group of sources or stressors to a group of effects, potential exposure of
biota to sources of stress can be examined in a variety of ways, depending on the types of data
and tools available and the types of sources being analyzed. As watershed ERAs are necessarily
place based, spatial analysis tools and GIS are essential to watershed assessment work. Land use
(e.g., agricultural and urban) constitutes a major sources of stress identified in watersheds, and
these data can be estimated using digital satellite images (e.g., Landsat) and GIS. Typically,
percent area cover in a watershed or other defined spatial area is used to quantify the "exposure"
of a given land use (e.g., urban development).  The way in which land use percentages are
calculated may need to be considered carefully. The Clinch and Powell ERA demonstrated that
land uses within the riparian corridor were more indicative of exposure to aquatic life than land
uses calculated for the watershed as a whole (Diamond et al, 2002). Others  have shown similar
significance of riparian land uses in watershed studies (Stewart et al., 2001; Hunsaker et al.,
1990).
       Exposure stemming from different land uses may be inferred from available data in the
watershed and exposure-effects information from the literature. For instance, in the Clinch and
Powell assessment, the fish community was consistently poor when the surrounding riparian
zone included all four main sources of stress: mining, urbanization, major roads, and pasture
areas (Diamond and Serveiss, 2001).  The strong association of adverse effects on fish in the
presence of these nearby land uses was meaningful, even in the absence of more specific,
quantitative profiles of exposure and effects (Diamond et al., 2002). However, assessing
relationships between land use and ecological resources is especially complex in a watershed,
and interactions among land uses could mask relationships between a particular land use and
ecological effects. The effects of pasture land could be masked by coal mining effects (Diamond
and Serveiss, 2001).  In addition, land cover estimates have a certain amount of error associated
with them, both spatially and thematically.  Thus, depending on when effects  or stressor
characterization data were collected, land coverages may not be an accurate reflection of source
information.  These uncertainties need to be recognized and discussed either  during this risk
analysis phase or during the next phase, risk characterization (Chapter 6). Land coverage data
may require "ground-truthing" to ensure that risk assessors are dealing with correct estimates of
specific land coverages in a given location or region.

5.5.3.  Rank Risks
       Occasionally, insufficient data are available to analyze an assessment endpoint or to assess
a surrogate measure of effect. In these instances, less precise measures (scalar, ranked, or
qualitative) could be used. Relative ranking of risks associated with management options may be
sufficiently robust to provide meaningful information to decision makers.  The RRM (Landis and

                                            38

-------
Wiegers, 1997) is one approach that produces a "quantitative" assessment using best professional
judgment of exposure and effect rankings.
       While the focus of this report primarily has been on EPA-sponsored case studies, a
number of regional ecological risk assessments have been developed using the RRM (Walker et
al, 2001; Landis et al, 2000; Wiegers et al, 1998).  Rather than looking at the traditional risk
assessment components of stressor,  receptor, and response, the relative risk regional approach
examines relationships between sources  (groups of stressors), habitat (groups of receptors), and
impacts (groups of responses).  Stressors are ranked in terms of their occurrence and importance.
High relative risk occurs where a source, an impact, and a habitat strongly overlap; where there is
little overlap, the relative risk is low. Relative risk can also be summarized via numeric indices, as
shown in Wiegers et al. (1998) for the Fjord of Port Valdez. This approach has the clear
advantage of prioritizing what are thought to be the main stressors and major habitat types at risk.
This analysis can then focus subsequent  data collection and analytical efforts on managers
concerns.
       With the RRM approach, a source can serve as  a group of stressors, a habitat can serve as
a location for a set of organisms or receptors, and an ecological impact can serve as a group of
receptor responses.  Each source and habitat is ranked for each area to indicate high, moderate,
low, or no risk within the context of the study site. Ranks are assigned using site-specific criteria.
RRM ranks sources or stressors in different locations in a region, evaluates or filters if the habitat
is affected by sources or stressor,  and ranks the impacts. The region is divided  into several similar
areas based on land use, basins, clusters of similar stressors, habitats, or land uses.  Ranks or
scores to represent high, medium, low, and none are used to reflect the magnitude of the source
and the amount of habitat that could be affected in each subarea.  The ranks assigned for each
combination of source, habitat, and  subarea are multiplied together. If the interaction between the
exposure and effect is likely to occur, the two scores are multiplied by one another. The results
for each subarea are summed to determine the relative ranking of risk within the region.  Thus, by
quantitatively determining the relative risks in subregions, the  contribution of risks  from sources
or stressors, risks in various habitats, and risks to assessment endpoints can be calculated.  The
results can then be portrayed easily in graphical form with maps depicting the relative risks within
the study region.
       This approach is simple to use and can include stressors about which little is known
quantitatively. This approach uses a qualitative, graphical approach for sorting out the many risk
pathways that may be present in a regional-scale ERA.
                                            39

-------
5.6.  DEAL WITH UNCERTAINTY
       Another source of uncertainty in using source data as a surrogate for stressors is that some
land uses or habitat features may be correlated to natural physiographic factors that may
themselves have a pronounced effect on biota.  In the Clinch Valley ERA, for example, elevation
was related to the number of aquatic species present, even in the absence of stressors. Thus,
elevation effects had to be "factored out" of the analysis to examine source-effect relationships.
This made it necessary for risk assessors to use a smaller subset of the available sites that were
within a relatively narrow (i.e., nonsignificant) elevation range for analyses. Further partitioning
to account for other factors, such as land uses, would further reduce the number of appropriate
sites available for analysis.  Depending on the stakeholder concerns, predicted results, and
acceptable levels of uncertainty, data may be insufficient for any type of statistical analysis.
       The inclusion of natural variability in risk analyses is especially critical at the  watershed
or larger spatial scales. Aquatic life habitats, for example, are known to vary in quality and
quantity  depending on natural factors, such as precipitation patterns and geology. At larger
spatial scales, these natural factors (including the  elevation example described above) are likely
to vary across the watershed, necessitating proper assessment of these factors.  Typically, natural
variability is addressed through the identification of appropriate reference conditions as discussed
previously. Measurements of the assessment endpoint, as well as natural factors that affect that
endpoint, across several reference sites under a given set of conditions (e.g., precipitation range,
elevation), will yield an estimate of natural variability. This variability becomes the basis  for
determining whether a given source or stressor tends to be associated with a non-natural (i.e.,
impaired) effect.  In some regions, biological data collected over more than a year or season are
needed at reference sites to adequately characterize natural variability.  This is particularly the
case for more arid regions where "wet" years can yield a different biological condition than "dry"
years.  Because habitat features are often sensitive to natural regional factors, it is important to
define natural variability for the watershed in which one is working.  The assessor should
critically evaluate reference condition data obtained elsewhere to be confident that the
information is appropriate for  the watershed being evaluated.
       An additional uncertainty associated with using source data as a surrogate of stressor
information is that all members of a given source class may not have an equal probability of being
associated with a hypothesized stressor (or condition). For example, certain agricultural land uses
may be associated with sedimentation and excess nutrient input into streams.  However,
depending on particular BMPs that may have been implemented at a given site, the intensity of
these stressors (and therefore predicted ecological effects) may not differ much from that
observed under reference conditions (where few or no stressors are present).  If such
differences among sites are not identified and taken into account in risk  analysis, results will be

                                             40

-------
"noisy," masking any real relationships that might exist between certain source types, stressors,
and effects.  A solution to this challenge is to involve people in the ERA who are knowledgeable
about the watershed and can help "ground-truth" risk analyses.
       Literature reviews often play an important role because site-specific information about
ecological effects, or interpretive benchmarks (e.g., effect or impairment thresholds), is often
unavailable for the watershed under study.  For example, researchers elsewhere may have
demonstrated relationships between the ecological measure of interest (e.g., the abundance and
diversity of certain wetland plants) and certain stressors (e.g., urban stormwater runoff) that also
occur in the watershed of concern. It may be appropriate to extrapolate results obtained in those
other locations to the subject watershed as long as uncertainties are adequately documented.

5.7.  RECOGNIZE THAT  ANALYSIS IS AN ITERATIVE PROCESS
       Another issue of importance is that the collection and analysis of data in the analysis phase
may be iterative. Analyses may bring new information to light (e.g., one pathway between
stressor and effect may have  much greater impact than originally predicted). In addition,
management's needs may change or the environmental situation may become dramatically
different (e.g., following an oil spill).  Many recent recommendations have emphasized that
stakeholder and manager involvement needs to be initiated in the planning step, and recurring
rounds of deliberations and analysis are necessary throughout the process to make the findings
most useful (U.S. EPA, 2000a; Foran and Ferenc,  1999; PCCRARM, 1997; NRC, 1996).  New
information obtained through literature review, field data, peer review, or environmental changes
since the beginning of the assessment may trigger these iterative loops.
       As the analysis proceeds, the assessors should present interim findings to managers to
ensure that the assessment is targeting the appropriate problems. A lesson learned from the
prototype assessments is that regular discussions between the assessors, managers, and
stakeholders during the analysis phase are very useful for refocusing the ERA and for
reprioritizing pathways for evaluation (Serveiss, 2002). The Clinch Valley ERA demonstrates
two particular concepts related to the benefit of dialogue among participants during analysis. The
initial risk analysis showed that certain measures were not available or were spatially distributed in
a way that made some of the proposed analyses infeasible.  Active dialogue among assessors,
managers, and stakeholders during this process resulted in modifying the analysis plan and
switching to determining associations between land use and biotic measures.
       The value of pilot studies also illuminates how watershed ERA is an iterative process.
Given the often complex interactions in watersheds, data limitations, and costs, it  also may be
useful to perform a pilot analysis first to determine that the analytical approach is likely to  be
completed successfully. Analyzing a subwatershed is one approach for pilot testing the analysis

                                            41

-------
plan (Diamond et al, 2002; MacDonald, 1994). In the Clinch and Powell Valley assessment, the
Copper Creek subwatershed was examined first to assess how to apply the analysis plan to the
entire watershed (U.S. EPA, 2002a). Copper Creek was chosen for this pilot analysis because it
was the most data-rich subwatershed and because it was a relatively simple case in that
agricultural uses were the major source of anthropogenic activity. The pilot study was needed to
address two analysis objectives central to this assessment: (1) to identify the appropriate spatial
scale to test relationships between land use activities or stressors and measures of effect and (2) to
identify whether the benthic macroinvertebrate measure (i.e., the EPT index) or the fish IBI would
be a reliable surrogate measure of effect for predicting the status of native mussel assemblages.
Achieving the latter objective was especially desirable because it was known at the outset of this
study that available native mussel data were more limited than either EPT or IBI values.

5.8. UNDERSTAND SECONDARY OR INDIRECT PATHWAYS
       In watershed ERAs, more than one source of a given stressor may lead to secondary
pathways. For example, toxic chemicals can originate from nonpoint sources such as agriculture
(pesticides), point sources (wastewater treatment plants), urban runoff (oils, metals), and many
other human activities. The conceptual model should identify the major sources of a given
stressor to reveal all pathways. Some stressors, such as temperature, acid runoff, and low
dissolved oxygen, may have a seasonal component or may be associated with certain geological
or hydro logical characteristics of the watershed.  The analysis phase should attempt to address
such variation if possible to better reflect stressor-response associations.
       Secondary stressors may be even a greater concern than primary stressors.  Some stressors
may have direct results that impair valued resources, such as nutrients resulting in nuisance algal
blooms.  Nutrients may have secondary effects, however, such as decreased dissolved oxygen,
due to increased bacterial respiration and algal die-off. In this case, a secondary stressor from
one source can be similar to a primary stressor from another source; e.g., decreased dissolved
oxygen may also be a primary stressor resulting from release of oxygen-demanding material
(biological oxygen demand) from wastewater effluents or agricultural runoff  These secondary
stressors  may need to be considered as well. For instance, in Waquoit Bay, nutrient input is a
stressor that caused excess macroalgal growth (a secondary stressor) to shade eelgrass and inhibit
its growth (Figure 4). Fate and transport models were then used to predict environmental
nitrogen  loading from the watershed into the bay (U.S. EPA, 2002c).  In a similar manner, a
model of phosphorus loading in Wister Lake in Oklahoma allowed estimation of changes in lake
trophic conditions (Hession et al., 1996).
                                           42

-------
       Risk analysis is complete when quantitative stressor-response characterizations have been
completed and assumptions, uncertainty, and data limitations have been discussed.  A few
iterations maybe required to progress from start to finish. Finally, since watershed ERA
combines risk analysis with risk characterizations, the necessary iterations may become blurred.
                                            43

-------
                        6. RISK CHARACTERIZATION PHASE

       The exposure and effects data generated in risk analysis are integrated to form
conclusions in the risk characterization phase. Risk characterization should use multiple lines of
evidence from models, field work, and statistical extrapolations to increase the confidence in
conclusions (U.S. EPA,1998b) and to help provide answers for risk management (see Chapter 8).
Finally, uncertainties originating in both the problem formulation and analysis phases should be
described (Suter, 1998). The final product of this phase is the risk assessment report prepared for
managers to support defensible management decisions.
       Risk characterization includes two major steps: risk estimation and risk description.
Risk estimation integrates the exposure data and stressor-response estimates from the analysis
phase while addressing uncertainties that arose throughout the assessment.  Methods used to
integrate exposure and effects data include comparing single values of effect and exposure,
comparing statistical distributions of exposure and effect values, or conducting simulation
modeling. As previously presented,  in watershed ERA different techniques may be used to
accomplish the integration of exposure and effects data (e.g., causal analysis, risk ranking, using
source data) and risk analysis may merge with risk estimation. As a result, risk estimation was
discussed in the preceding chapter, and risk description is the focus of this chapter.

6.1.  RISK DESCRIPTION
       During risk description, risk assessors prepare a summary of ecological risk and interpret
ecological significance.  Summarizing risk involves making a bottom-line estimate of risk.
Ideally, a risk summary would be in the form of a quantitative statement:  for example, "There is
an 80% chance of a 50% die-off of certain valued fish species in the watershed due to
sedimentation from nonpoint sources under current conditions."  Such a statement, when
combined with information on the reliability of the statement (i.e., the uncertainty), would allow
analysis of the relative costs and benefits of one of several management options. Unfortunately,
often it is not possible to produce  a precise quantitative estimate for a watershed ecological
assessment because risks typically involve multiple stressors, from multiple sources, across a
range of habitats. Instead, probability estimates can be expressed in high, medium, or low terms.
For example, there is a high probability that shorebird populations in the immediate vicinity of a
filled wetland will decline as a result of habitat elimination. Every effort should be made to state
the likelihood of the outcome. Risk descriptions also can be expressed in relative terms, such as
option A poses more risk to the assessment endpoint than option B. This is the kind of risk
description derived from the RRM that ranks risks by habitat areas and stressors (Wiegers et al,
1998).

                                            44

-------
       Because exposure to stressors is typically complex and possibly synergistic in watersheds,
these relative data may be more useful in characterizing risks than single dose-response estimates
derived from laboratory data. The nature and magnitude of predicted effects should be
considered with regard to the structure and function of the  affected ecological system, its
resistance and resilience, spatial and temporal effects, indirect effects or cascading interactions,
and the life-cycle  of organisms.
       The risk description interprets ecological significance, translating possible risk estimates
into a discussion of their consequences for the watershed. The description may address the nature
and magnitude of effects, the spatial and temporal patterns  of effects, and the potential for
ecosystem recovery if effects have already occurred. For instance, the factors limiting mountain
whitefish survival and reproduction include elevated water temperatures, loss of lotic habitat,
fluctuating water  flows,  and excessive sedimentation (U.S.  EPA, 2002d). The evidence for these
effects is demonstrated by evaluating favorable spawning conditions at two conditions in the river,
comparison with habitat suitability indices, and review  of the literature.  The risks are described in
Table 3.
       The significance  of the consequences of predicted effects may vary considerably for
different types of ecological systems or habitat  types.  For example, in the Waquoit Bay
assessment (U.S.  EPA, 2002c), atmospheric deposition of nitrogen on loam soils far from a water
body was deemed less of an impact than deposition occurring on sandy soils near an estuary.  The
effect of a herbicide may be quite  different in a stream that derives most of its organic carbon
energy from plants as compared with the effect in a stream  that utilizes predominantly
detrital-based organic carbon.  The loss of wetland area maybe highly significant if it represents
the only habitat available in an area for waterfowl but maybe negligible if it occurs among a much
greater wetland area.
       Sometimes risks can be described in relationship to  a management option that is being
considered. A review of historical data was used to describe the relationship between increased
nitrogen loads and decreasing levels of eelgrass cover over  time (U.S. EPA, 2002c). Managers
then could select  a particular percentage of eelgrass bed coverage as a management target. For
example, if the restoration target was to support the growth of eelgrass on 30% of the main
estuary, this would be comparable with 1970 conditions in Waquoit Bay. Thus, the nitrogen load
would need to be reduced to 18,000 kg N per year (Figure  6).
       In watershed assessments, the spatial distribution of exposure and effect is important to
consider and can be assessed using GIS overlays of the two types of information.  GIS maps can
be helpful as analytical and communication tools. When large areas are parsed into small subunits
or representative  subsets are selected for analyses, key differences can be lost. GIS
                                            45

-------
       Table 3. Middle Snake River case study:  integration of stressors, responses, and recovery potential for the
       reproduction, growth, and survival of the mountain whitefish population
Factor
Number of
spawning
fish
Spawning
Fry
Food supply
Movement
Stressor
Loss of adult habitat,
restrictions on
movement to feeding
and overwinter areas
Water temperature too
high for successful egg
development, loss of
lotic habitat
High water temperature,
fluctuating water levels,
loss of lotic habitat, and
predation
Sedimentation and
increased water
temperature
Dams prohibit seasonal
migratory movement by
adults
Line of
evidence
LIT, BPJ
HSI, LIT,
WQS, BPJ
HSI, LIT,
WQS, BPJ
LIT, Field,
BPJ
LIT, Field,
BPJ
Risk
An increase in the
population size is not
possible with low or no
reproduction
Population cannot
recover without
successful reproduction
Population cannot
recover without
successful recruitment
Invertebrate fauna not
adequate to support an
increased mountain
whitefish population
Adults unable to reach
upstream tributaries or
refugia used for
spawning or rearing
Uncertainty
Low, very few adult fish
are present in the Middle
Snake River or nearby
tributaries
Low, water temperature
requirements for spawning
are known
Low, temperature
requirements are known,
but rearing habitat not
identified
Moderate, adequate
analysis of sampling
information has not been
completed
Moderate, limited
information on migratory
population in the Middle
Snake River
Assumption
Available fish surveys
are a true indication of
the scarcity of fish in
this area
Low spawning success
attributed to poor water
quality conditions
Fluctuating water levels
push fry into deeper
water where increased
predation occurs
Food supply is limited
Adult fish reported at
Lower Salmon Falls
Dam in the 1950s
represented a migratory
population
Recovery potential
Low, without improving
lotic conditions and
lowering water temperature
Low, the present annual
water temperature regime
does not support successful
spawning
Low, the present annual
water temperature regime
does not support successful
fry development
Low, without improving
lotic conditions and
lowering water temperature
Low, without providing
adequate fish passage
facilities at the dams
BPJ = best professional judgment
Field = field survey in the Middle Snake River
HSI = habitat suitability index

Source: U.S. EPA (2002d).
LIT = literature search
WQS = water quality standards
                                                                 46

-------
                   24  -
                       1940    1950   1960   1970   1980   1990   2000
       Figure 6.  Historical changes in nitrogen (N) loading predicted by the
       nitrogen loading model. If the management target is a return to 30%
       eelgrass cover, then managers must reduce N loads to 18,000 kg N per year,
       loads comparable to those of around 1970.

       Source: U.S. EPA (2002c).

approaches help address that problem by providing a mapping tool that can capture basic
descriptive information and link calculations for subareas to produce an overall picture.
       With a GIS, data layers can be added to identify other more specific potential sources of
stressors, such as wastewater discharges and mining activities.  Although certain types of land use
may not be amenable to calculation on a percent area basis, they may be analyzed on a distance or
density basis (i.e., how close a given mine is to aquatic life habitat or how many mines per square
mile are within a certain area) (see Figure 7). Either of these calculations could provide useful
estimates of mine "exposure." Before using GIS, ground-truthing of data is necessary.
       The RRM projects described previously rely on GIS-based approaches to evaluate
potentially important stressors in watersheds or regions.  These approaches use maps to display
major stressors and associated degree of potential biological impacts on maps. The maps greatly
facilitate communication of the information to the public and managers.  Stressors can also
                                           47

-------
        55
        50
        45

        40

      s35
        30
        25
        20
        15
        18

        16

        14

        12

        10

         8

         6

         4
 *
N = 31
             = 25
                         = 81
                   < 1 Km    1 -2 Km    >2 Km
               Distance from nearest Mining Activity
                 * = significantly different from sites > 2km from mining
                                       ZC ±Std. Dev,
                                           ±Std,Err,
                                           Mean
 = 23
                   < 1 Km
         1 - 2 Km
>2 Km
                                           ±Std. Dev.
                                           ±Sld. Err.
                                        n   Mean
               Distance from nearest Mining Activity
                * = significantly different from sites > 2km from mining
Figure 7. Fish index of biotic integrity (IBI) or insect ephemeroptera,
plecoptera, and trichoptera (EPT) values in relationship to proximity to coal
mining sources in the Clinch Valley ecological risk assessment. If the location
of a source of potential stressors can be clearly defined (as in the case of point
sources or those that can be spatially defined), it may be useful to relate measures
of effect to the distance from the source (in this case active coal mines) to assess
potential effects of a given type of source.

Source: U.S. EPA (2002a).
                                       48

-------
combine to influence the quantity and quality of habitat for species, and therefore habitat can be
used as a basis for analyzing and showing how effects are integrated.  Because habitat
requirement vary by species, the range occupied by particular species, or biological communities
needs to be considered.  For example, land use impacts can be analyzed and shown according to
the various habitats occurring across the landscape.
       Besides GIS, graphs are also a helpful analytical and communication tool.  In assessments
where timing of events is critical (e.g., as in acid precipitation, or in the assessment of episodic
events), graphs that show the timing and distribution of stressor events in comparison to an
effects threshold may be a useful way to present the information. For prospective risk analyses
supporting watershed planning, risk estimation derives the estimates of management measures for
comparison between different management options. In EPA's TMDL protocols, this is referred
to as a "linkage analysis" (U.S. EPA, 1999b). When exposure and effects data are limited or are
not easily expressed in quantitative terms, qualitative evaluation techniques may be used to rank
risks using subjective judgment and categories such as low, medium, and high.
       It can be useful to incorporate results of risk analyses, and the probable co-occurrence of
stressors and assessment endpoints, into the  initial conceptual model.  Pathways can be shaded or
color-coded (e.g., green = no or few risks likely;  yellow = risk possible; red = risk likely) on the
conceptual model to help characterize the risk to assessment endpoints (Serveiss and Ohlson,
2006).  This "updated" conceptual model could also be used to highlight where major
uncertainties lie and/or additional sampling or analyses would be especially worthwhile.
       Agreement among multiple lines of evidence increases the confidence  in the conclusions,
although any discrepancies warrant discussion.  The Clinch Valley assessment used a table to
show the various lines of evidence and to summarize the conclusions of risk and recovery
potential to various life stages of mussels and fish (U.S. EPA, 2002a).  In the Middle Snake River
ecological risk assessment (U.S. EPA, 2002d), a water quality simulation model was used to
predict attainment of water quality standards and habitat requirements for the life-cycle stages of
coldwater fish (Table 3). Simulations were used to modify the model or provide information on
uncertainty. The data from the model were supplemented with field data and literature review.

6.2.  DESCRIBING UNCERTAINTY
       There are two general sources of error in ecological assessments—uncertainty and
variability.  Uncertainty from lack  of knowledge can be expressed in many ways during the
analysis phase, such as insufficient data or spatial  coverage of required data, inaccurate model
inputs, GIS spatial resolution errors, and investigator bias (intended or unintended). Natural
variability includes spatial and temporal heterogeneity in measures of interest, background
                                            49

-------
ecological system factors (e.g., geology, precipitation), and natural disturbances (e.g., floods,
fires, hurricanes), all of which can affect the ecological effects and relationships observed.
       Uncertainties encountered throughout the assessment, including the basic data set, should
be summarized in an uncertainty analysis (Suter,  1998; U.S. EPA, 1998a). There are likely to be
many sources of uncertainty,  such as measurement error (inappropriate, imprecise, inconsistent,
or sparse measurements), conditions of observation (such as extrapolating from laboratory tests to
field predictions), and limitations of models (oversimplifying complex ecological pro cesses,
incorrect assumptions). Uncertainties exist in every aspect of a risk assessment, especially in
estimates of exposure, estimates of effects, and the integration of exposure and effect.  The
relationship between the measure of effect and the actual assessment endpoint is always subject to
uncertainty. Watershed and regional assessments may contain more uncertainty than smaller scale
assessments because of difficulties in obtaining sufficient data for multiple stressors, sources, and
measures of effect for the entire geographic area.
       Uncertainty can be described in either narrative form, with statistical distributions, or with
sensitivity analysis. Narrative descriptions are especially suited for qualitative types of
uncertainties, such as unclear communications, descriptive errors, and professional judgment and
any assumptions made throughout the assessment. Depending on how data gaps are addressed,
their associated uncertainties  may be quantifiable. Natural variability and sampling error may be
described statistically.  Some  uncertainties, such as those associated with modeling, can be
addressed in terms of sensitivity analyses, in which certain model parameters (or inputs) are
adjusted and the outputs compared (Suter, 1998; Hession et al, 1996).  Other uncertainties, such
as those associated with sparse data, cannot be reduced without additional data collection but may
lead to identification of those data that greatly reduce uncertainties for future risk assessments.
Uncertainties may increase if the appropriate temporal scale is not evaluated. Episodic pollution
events, for example, can have dramatic, persistent ecological effects that might not be recognized
using a short temporal horizon.  In the Clinch and Powell ERA,  episodic chemical spills from
transportation or industrial accidents resulted in long-term reductions in native mussels
downstream (Diamond and Serveiss, 2001).  Long-term data sets, if available, are desirable to
address such issues (Hunsaker et al., 1990).
       The Middle Snake River assessment (U.S. EPA, 2002d) used Table 3 not only to
summarize risk descriptions but also to describe the uncertainties and assumptions associated with
each effect. In successive model simulations, input parameters were varied to provide information
on uncertainty in model output. The data from the model were supplemented with field data and
literature review.  Sources of uncertainty included limited knowledge of stressor-effect
relationships, variability in field measurements, model  error, and estimation error. Each source of
uncertainty was further described in narrative form.

                                            50

-------
       When data are sufficient to proceed with a quantitative assessment, as when focusing on
single pathways of the conceptual model (e.g., Serveiss et al., 2004; Hession et al., 1996), then
quantitative measures of uncertainty are possible for the exposure and effect measures.
Uncertainties associated with interactions among stressors and sources that are not considered in
models only can be addressed qualitatively.  RRM addresses multiple stressors and effects by
using qualitative rankings in quantitative models.  Uncertainty of model output  can be represented
quantitatively in sensitivity analyses, although the input is qualitative (Walker et al., 2001;
Wiegers et al., 1998). Describing risk and uncertainty are two of the major features of risk
characterization.  Since there will always be uncertainty, it is important to have assessments peer-
reviewed by other scientists in the field. Peer-review comments and a description of how the
report was revised in response to comments should be made available.

6.3.  RISK CHARACTERIZATION COMPONENTS
       Risk assessors prepare a report for the risk managers using concepts outlined in Chapter 7,
Risk Communication. Risk managers use this information to make decisions (see Chapter 8).
Although these steps are presented chronologically in this report, these activities may occur
throughout a watershed assessment. A watershed manager may need to make a relatively quick
decision to control the impact of a spill and may ask the assessor for interim information before
the assessment is  completed. Thus, some elements  of risk characterization and risk
communication described in this and subsequent chapters may apply to the manner in which
interim findings are presented earlier in the assessment.
       Risk characterization can take the form of an isolated product by summarizing the earlier
phases or it can be one part of the entire report. Either way, somewhere in the  risk assessment
report, the following information should be included:

       •    A description of risk assessor/risk manager planning results
       •    A discussion of the major data sources and analytical procedures used
       •    A review of the stressor-response relationships
       •    Estimates and descriptions of risks to the assessment endpoints
       •    A summary evaluation of the relative risk of different management scenarios for use as
           a basis for risk management decision making
       •    The major sources of uncertainty and the approaches used to address them
       •    Documentation of science policy judgments or default  assumptions used to bridge
           information gaps and the basis for these assumptions
                                            51

-------
       •  Lines of evidence and the rationale for reaching the conclusions that were reached
       •  A description of the peer review process and a list of peer reviewers


       Effective risk characterization must accurately translate the best available information
about a risk into a language nonscientists can understand (NRC, 1996). The risk characterization
should clearly communicate to the risk manager the major risks at some level of biological
organization (organism, population, community, ecosystem), the ecological significance of the
findings, and the level of uncertainty. It is important to clearly describe the ecological resources
at risk, their value, and the costs of protecting (or failing to protect) the ecological resources.
The ultimate success of risk characterization depends on successful risk communication.
                                            52

-------
                              7. RISK COMMUNICATION

       Risk communication is the translation of scientific information in a form that is useful to
managers and the public.  Risk communication should occur throughout the assessment process,
but ultimately what counts is how the results of the assessment are used.  Hence, this is one of the
latter chapters in this report.
       When risk characterization is complete, a report is prepared showing estimates of risk.
Clear presentation of the strengths, limitations (including uncertainties), and conclusions of the
risk assessment will greatly enhance the assessment's usefulness in decision making (Serveiss,
2002).  The challenges for risk communication include establishing a process for risk
communication, targeting the audience, and communicating effectively.

7.1. THE RISK COMMUNICATION PROCESS
       Regular and recurring communication between scientists and managers is needed to
refine management objectives and to ensure that the risk assessment is relevant to the decision
problems. The management  objectives, and the relationships between management objectives,
stressors, and management options, are often poorly defined at the start of the process (Reckhow,
1994).  When the assessment is completed, the manager and key stakeholders should have  a full
and complete understanding  of the assessment.
       Involving stakeholders throughout the process increases the likelihood that findings will
be useM and helps prevent confrontation or litigation. Watershed management requires an
interactive, participatory approach (Glicken, 2000; U.S. EPA, 1995a), in which risk assessors
both impart information to and gain information from stakeholders. In the ERA paradigm,
stakeholders work as partners with scientific risk assessors at every stage of the
process—defining the problem to be solved, prioritizing the risks posed, and evaluating and
ranking the remedies (NRC,  1996).
       Many types of risk cannot be assessed in a completely objective way because the
evaluation of risk depends on the interpretation of goals and objectives, which is a value-laden
process. A stakeholder-based decision process provides a means to make value judgments that
are supported by the scientific evidence. The Delaware River TMDL for polychlorinated
biphenyls was established as such a structured decision process. A Web site was used to describe
the time and place for business and public meetings and the topics to be discussed (Delaware
River Basin, 2006).
       While stakeholder involvement is crucial, there are also reasons to refrain from involving
all participants.  An adequate treatment of science is possible in a stakeholder-driven
process—but only if sufficient high-quality staff, time, and resources are available (U.S. EPA,

                                           53

-------
200la). Where resources are not sufficient, or when a decision must be made quickly to address
an imminent threat, stakeholder processes may not be the solution. The SAB also warns against
handing off processes to a stakeholder group as a means to address mandates for which an agency
has insufficient resources.
       If stakeholder groups are too large (or intractable in their opinions or desires), it may be
difficult to reach closure. There are likely to be different stakeholder groups with widely varying
levels of interest in the process, and varying ability to commit time to the process. One solution
appropriate for many risk assessments is to form a stakeholder advisory group consisting of a
smaller group of well-respected and representative stakeholders (about 10 members).  The
advisory group serves as a communication pathway between the risk assessment team and the full
set of interested parties in the community, serving both to identify values and concerns of the
community at large and to communicate progress in risk assessment and risk management efforts
back to the community.
       An important challenge for the participatory approach lies in balancing meaningful
stakeholder participation with the need to make progress in managing risks.  Rather than allowing
the stakeholders to have access to the risk management team and process at all times, it is
important to strike a participation/management balance by providing well-defined, strategic time
frames for stakeholder input.

7.2. ADDRESS THE TARGET AUDIENCE
       Watershed assessments have multiple audiences and stakeholders. There are managers or
groups with defined decision making authority, such as a governmental agency with regulatory
authority over water quality or a county staffer charged with developing a land use plan. In
addition, individuals or groups may actively participate in the watershed management process
(e.g., watershed associations, river watch citizen groups, volunteer monitoring groups, industries,
and agricultural associations).  Other stakeholders include participants who are influenced by and
have the ability to affect (if only through lifestyle and land use choices) the area of interest (e.g.,
land owners in the watershed).  Communication needs differ accordingly with each stakeholder
group. Glicken (2000) makes the further point that stakeholders or interested parties are
identified and defined relative to specific issues.  A characteristic of most watershed management
exercises is that there are multiple issues to be addressed, and thus the participating stakeholders
do not constitute a single, well-defined set but, rather, a complex network that shifts in
relationship to issues. Efforts should be made to engage the participants on all issues and at every
phase, as neglect may lead to alienation and contention (also refer to Section 3.1).
                                            54

-------
7.3.  COMMUNICATE EFFECTIVELY
       Risk characterizations should 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, 1998a).  For effective risk communication, scientific
information should be characterized in the manner most appropriate for addressing the major
concerns. Since watershed risk assessments are place-based studies, maps are often important
tools to aid in summarizing and conveying information  (Serveiss, 2002). With the advent of
powerful and accessible GIS and database tools, it has become feasible to analyze and display
spatial information.  Maps can provide easily grasped visual relationships between sources,
stressors, and effects to which the public can relate. In the Clinch and Powell assessment, maps
were especially useful in visualizing cumulative sources of stress in relationship to locations of
threatened and endangered mussel species (U.S.  EPA, 2002c). Such maps can then be used to
evaluate vulnerability and prioritize management strategies.
       Communication can be misleading if uncertainty is not described. At each step of the
assessment process, what is not known should be presented along with what is known.
       Clear presentation of the strengths, limitations, and conclusions of the risk assessment will
greatly enhance the assessment's usefulness in decision making. Describing  the association
between a range of stressor levels and various risk estimates can be especially useful for managers
who may need to choose among options with varying capabilities for stressor reduction.
Successful risk communication leads to a higher  likelihood of selecting optimal risk management
actions.
                                            55

-------
                               8. RISK MANAGEMENT

       The purpose of a watershed ERA is to provide managers with information to help address
environmental risks (i.e., make risk management decisions).  The goal of risk management is to
take scientifically sound and cost-effective actions to reduce or prevent risks, while taking into
account social, cultural, ethical, political, and legal considerations (PCCRARM, 1997).

8.1.  EVALUATING MANAGEMENT ALTERNATIVES
       Decisionmakers must select a particular management action, including no action.  When
environmental management goals and risk assessment endpoints are clearly defined and agreed
upon by assessment participants, the risk assessment process is more likely to produce useful
information for choosing an action.
       Watershed management typically includes multiple objectives, and watershed degradation
typically involves multiple sources and multiple risks.  In an ideal case, the final management
plan would be endorsed by all participants. However, when conflict arises between stakeholder
groups with opposing goals, selecting appropriate management alternatives may be difficult.
Some watershed projects may prove too controversial and reach a stalemate from time to time.
In the worst cases, some stakeholders may choose to selectively ignore parts of the science  or to
withhold relevant information. Despite these challenges, some stakeholder involvement is
always necessary (as discussed in Sections 7.1 and 7.2), especially before money is committed to
possible solutions.
       An important characteristic of many watershed risk management processes is the role of
stakeholders in identifying and evaluating potential management options.  It is generally not
appropriate to use open stakeholder processes to make regulatory decisions that are explicitly
mandated to agencies (U.S. EPA, 200la).  However, when selecting among management
alternatives that are consistent with the law, stakeholder involvement increases acceptability of the
outcome—particularly where social value judgments are required  to choose the best level and
approach for environmental protection.
       Evaluation of the relative performance of candidate management options can be made on
the basis of their abilities to achieve management objectives, usually through the use of an
appropriate indicator or measurement endpoint.  In some cases, this is  a relatively straightforward
process. For  instance, the objective might be to reduce fish mortality from ammonia inputs, with
an associated indicator of predicted un-ionized ammonia concentrations.  If the modeling tools are
present to evaluate un-ionized ammonia concentrations associated with each management
alternative, then a direct comparison may be made. Designing assessment endpoints by relating
them to management objectives enables them to  be useful in risk management decisions.

                                           56

-------
            BEST
100%
  50 %
   0%
       Quantitative management objectives are always desirable and sometimes management
options can be tested in a prospective risk analysis. In these instances, there may be a wide array
of management options that can be considered (e.g., limit types of development, limit
development density, use stormwater controls). In circumstances where there are no competing
objectives, and if all possible management objectives can be ranked quantitatively using an agreed-
upon scale, selecting the best management options is relatively straightforward. However, this
rarely happens.
       In most watershed management exercises, there are competing objectives and a lack of
consensus.  Objectives may be pre-established and
unfocused (e.g., control future nutrient-related
eutrophication in a lake). The management options also
may change after the  results of early analyses are
presented.
       When a quantifiable management objective may
not be available or agreed upon by various stakeholder
groups, a subjective ranking approach may be used. A
simple value scoring approach that requires each
stakeholder group to  identify the best and worst of the
candidate scenarios (combinations of management
alternatives) can be used (Stewart and Scott,  1995).
Best and worst scenarios are defined as the top and
bottom of a subjective scale running from 100% to 0%,
with intermediate scores relating to the relative "gaps"
between the two well-defined extremes. The resulting
scores obtained by a stakeholder group for a set of
scenarios may then be displayed on a "thermometer"
scale (Figure 8).  In the figure, the stakeholder group
ranked five scenarios, judging Scenario II the best and
Scenario III the worst. A compilation of rankings
submitted by multiple stakeholder groups would reveal the alternatives that are least objectionable
to the participants.
       While the actual decision process is likely to vary on a case-by-case basis, some
guidelines may help promote a successful outcome. First, as stated earlier, it is important to
ensure that stakeholder groups are present at the table from the start of the process, particularly
those with the potential to present obstacles to management.  Second, the decision context and
the role of stakeholders in the process should be made clear from the onset. Third,  it is important
          WORST
Figure 8.  Thermometer scale for
relative ranking of management
options.
                                           57

-------
for all parties to realize that evaluating alternatives is an iterative process and that scientific
information that emerges during the process may require rethinking the nature and significance of
risk pathways and, potentially, the management goals for a watershed.  Finally, one must
recognize that risk management may take time and money.

8.2.  APPLICATION TO DECISION MAKING IN THE EPA CASE STUDIES
       The three completed EPA case study reports each include a section on how the
assessment assists with decision making.  For the Clinch and Powell Valley assessment (U.S.
EPA, 2002a), the ERA process advanced a better understanding of environmental problems.  In
particular, the conceptual model and multivariate analysis helped clarify the interrelationships
between various components of the ecosystem and the manner in which human activities
contribute to  environmental problems within the watershed. In assessing environmental risk,  a
number of federal, state, and local environmental agencies and organizations came together to
share data,  explore and develop solutions, and undertake actions within the watershed. Risk
assessment findings can also help direct the efforts of the Upper Tennessee River Basin
Roundtable, which is composed of various individuals, agencies, and organizations  that have an
interest in protecting the watershed.  ERA findings will be useful to the roundtable as it begins
comprehensive strategic planning for watershed protection.  Additionally, the numerous
watershed coalition groups within the basin can use the risk assessment findings to direct their
efforts to protect and improve water quality within their watershed.
       The process of risk assessment helped lend further credence to what many professional
resource managers had long conjectured about problems within the watershed, thereby providing
more scientific support for taking actions to address identified problems.  For example, there is
now a better understanding of what sediment from cattle grazing contributes to the river.  Such
risk assessment findings will be useful to FWS and TVA personnel, who can now share this
information with farmers and encourage them to take actions, such as building fences to keep
cattle out of streams.
       Examples of management actions that FWS and TVA have considered on the basis of the
overall risk assessment findings include the following:

       •   Restoring additional abandoned mine lands throughout the watershed
       •   Further study of the chemical makeup of discharges from coal mining and processing
           facilities and the toxicity of these discharges to aquatic species
       •   Increasing the extent of forested riparian areas adjacent to and upstream of critical
           aquatic habitat sites for mussels and fish
                                           58

-------
       •   Implementing better spill control mechanisms on roadways and railroads near sensitive
          streams and more spill contingency plans for the watershed, which will enable the
          Virginia Department of Transportation and other agencies involved in constructing
          highway projects on or near waterways to  design those projects to reduce catastrophic
          events and minimize impacts of accidental  spills
       •   Installing BMPs for pasture and agricultural land to reduce sediment loading and
          implementing better treatment of wastewater discharges

       In the Waquoit Bay ERA (U.S. EPA, 2002c),  nitrogen loading from the watershed led to
water quality problems and diminished populations of valued biotic resources. Nitrogen loading
and estuarine loading models were developed as part of the assessment. These models helped
managers to better  understand nitrogen impacts and to make more informed decisions in
addressing the problem. The ERA process also helped engage local citizens in the dialogue on
mitigation strategies (including validating model predictions, long-term monitoring,  and adaptive
management).
       Another valuable product of the Waquoit Bay ERA is the conceptual model, which allows
citizens and local managers to view the watershed stressors and their associated ecological effects
from a more holistic perspective.  It emphasizes both point and nonpoint sources of pollution and
shows how valued  ecological resources are impacted by multiple stressors. Before formulation of
the model, the focus was on analyzing single stressors or single species or on managing fisheries
on a stock-by-stock basis.  The conceptual model helped develop evaluation and management
strategies that considered the wider effects of pollution on habitats.
       EPA used simulation techniques developed for the middle segment of the Snake River
(U.S.  EPA, 2002d) and the associated conceptual model to determine impacts from a proposed
new hydroelectric facility.  The Federal Energy Regulatory Commission was convinced by
EPA's risk assessment that cumulative impacts should be addressed  when reviewing new or
continued licensing for hydroelectric facilities.  The assessment also provided community groups
with a more robust analysis of the possible impact  of impoundments on the river. Risk
assessment participants were able to gain an improved understanding of the interrelationships
between various components of the ecosystem and how human activities contribute to
environmental problems within the watershed.  For example, there is now  a better understanding
of the contribution  of sediment to changes in the ecosystem, including eutrophication and decline
offish species. This information is useful to managers as they begin to regulate the  sediment
sources.
       The simulation models developed for the Snake River can be used  to test management
scenarios. Those interested in land use activities and resource protection can better explore
                                           59

-------
alternative management options for the watershed.  The methods developed through this risk
analysis can also be applied to other watersheds with similar conditions.
       The benefits of watershed ERA for the three EPA-sponsored watershed assessments have
been summarized and demonstrated above. Watershed ERA can also improve the organization
and quality of several other types of watershed analyses, including TMDL analyses and use
attainability analyses (UAAs), as described in the following sections.

8.3.  APPLICATION TO TOTAL MAXIMUM DAILY LOAD PROCESS
       The TMDL process includes activities similar to ERA risk analysis. For instance, the
objective of source assessment is to "characterize the types, magnitudes, and locations of sources
of sediment loading to the waterbody," while the objective of linkage analysis is to "Define a
linkage between the selected water quality targets and  the identified sources to determine total
assimilative capacity for sediment loading or total load reduction needed" (U.S. EPA,  1999b).
The source assessment objective is consistent with  exposure characterization, while the linkage
analysis is similar to effects characterization.
       An example of how watershed ERA can be applied to a TMDL watershed management
exercise is the clean sediment TMDL for Redwood Creek, CA (U.S. EPA, 1999b), which was
summarized in Serveiss et al. (2005).  Redwood Creek had a designated use to support a
coldwater fishery.  The water body was listed as impaired on the state's 303(d) list because
salmonid stocks were declining. Upland sediment loading from forestry was believed to have
salmonid impaired habitat quality. Changes to pool, substrate, and channel structure were
apparent and also contributing to increased temperatures.  Collectively, these changes affected the
spawning and food production of salmonids.  The TMDL listing and associated problem
statement mimic ERA problem formulation activities.
       Numeric water quality standards for sediment do not exist in California's regulations, and
the narrative criteria (support of the coldwater fishery) and a pollutant (clean sediment) made it
necessary to select  measures and  target values for the analysis.  The TMDL identified eight
instream and seven "hillslope" indicators and targets. The instream indicators were measures of
exposure, and they have been selected in a way that provides reasonable surrogates for effects.
For instance, the first instream indicator was "Percent  fines <0.85 mm in riffle crests offish-
bearing streams."  This indicator represents a critical node along a stressor-response pathway: that
excess fine sediment clogs spawning gravels  and results in reduced spawning success of
salmonids.  For the purpose of the analysis, the first instream indicator (and the seven other
indicators) served as  a surrogate for what would be the assessment endpoint (spawning success of
salmonids) and thus was also used as a measure of effect.  The hillslope indicators (e.g., number
of road crossings with diversion potential) served as measures of exposure. Management

                                           60

-------
(implementation) scenarios could then be compared and defined in terms of the hillslope indicators
and their estimated effect on sediment loading.
       Target values for the instream indicators were based on a combination of literature values,
professional judgment, and comparison of impaired reaches within the watershed to unimpaired
reference reaches, both within the watershed and in similar terrain. For instance, the target value
for percent fines in riffle crests was determined to be less than 14%.
       The linkage analysis for this TMDL was performed by estimating upland sediment loading
rates for the impaired water body and the unimpaired reference streams. Based on this
comparison, it was estimated that a 60% reduction in average annual sediment loading in
Redwood Creek was needed to support uses. Specific allocations were made by comparing
management scenarios and their impacts on erosion processes, which were mostly associated with
specific land use activities and were expressed as long-term average annual loads per square mile.
The individual load allocations were based on assessment of the controllability of different source
categories, that is, the extent to which sources are associated with human activity and will
respond to mitigation. The analysis indicated that the application of reasonable reductions to the
controllable load would be adequate to meet the TMDL.
       The Redwood Creek TMDL did not follow all the ERA steps because of time and budget
constraints. Yet, this TMDL has most of the components of a watershed risk assessment,
including problem formulation; identification of environmental management goals, options, and
objectives; assessment endpoints and associated measures; and the equivalent of exposure
characterization and effects characterization.  In other words, the TMDL was performed in a
manner consistent with the watershed ERA approach.  The scientific underpinnings of TMDL
development can be improved using several features of watershed assessment (Serveiss et al,
2005).

8.4.  APPLICATION TO THE UAA PROCESS
       Beneficial uses of water bodies are designated by states under the  Clean Water Act.
Common beneficial uses include warm or coldwater aquatic life uses, which are ecologically based
goals of a watershed that are subject to enforcement, if necessary. The designated use dictates the
physicochemical water quality standards that need to be maintained in a water body. Therefore,
proper identification of uses is a critical component of all water quality programs. A UAA is
defined as the process of determining attainable beneficial uses, or modifying designated uses that
are inappropriate (U.S. EPA, 1994).  The Federal Water Quality Standards Regulation, 40 CFR
131.10(g), defines a UAA as "a structured scientific assessment of the factors affecting the
attainment of a use which may include physical, chemical, biological, and economic factors."
                                           61

-------
       A UAA consists of three general steps:  (1) identifying and defining existing proposed or
designated beneficial uses that were present as of November 1975, (40 CFR 131.3(e)); (2)
determining whether uses are appropriate for the water body, and (3) identifying constraints on
these uses by examining physical, chemical, and biological characteristics of the water body as
well as so cio economic constraints.  The Federal Water Quality Standards Regulation defines six
factors, any one of which, if demonstrated, could be used to remove an inappropriate use or
establish subcategories of a use (see below).  The documentation necessary to support any of
these factors has been unclear and a source of confusion in UAAs to date, limiting the use of this
essential tool.  This is especially true for substantially modified watersheds (e.g., urban or
agricultural watersheds and streams that are effluent dominated) in which "human caused
conditions" predominate (see item 3 in list below). The following six factors are associated with
modifying or removing a 10 l(a) use within a use attainability analysis (UAA), as cited in 40 CFR
       1 .  Naturally occurring pollutant concentrations prevent attainment of the use.

       2.  Natural, ephemeral, intermittent, or low- flow conditions or water levels
          prevent the attainment of the use.

       3.  Human-caused conditions or sources of pollution prevent attainment of the use and
          cannot be remedied or would cause more environmental damage to correct than to
          leave in place.

       4.  Dams, diversions, or other types of hydrologic modifications preclude attainment of
          the use, and it is not feasible to restore the water body to its original condition or to
          operate such modification in a way that would result in attainment of the use.

       5.  Physical conditions related to the natural features of the water body, such as the lack
          of a proper substrate, cover, flow depth. . .unrelated to water quality preclude
          attainment of aquatic life protection uses.

       6.  Controls more stringent than those required by Sections 301(b)(l)(A) and (B) and 306
          of the Clean Water Act would result in substantial and widespread economic and
          social imp act.

       The watershed ERA framework offers many advantages for conducting UAAs, because:
(1) stakeholders  in the watershed (including resource agencies and wastewater dischargers)
participate in defining the assessment process; (2) the assessment explicitly treats multiple
stressors, including physical and biological, and other water quality stressors; and (3) the
assessment is necessarily place based because it deals with a specific watershed or basin.  UAAs
have been thwarted in many cases because it was often difficult to reach consensus on the types

                                            62

-------
of information needed and how multiple lines of evidence should be interpreted.  Watershed ERA
can greatly improve the performance and transparency of UAAs by providing the scientific
structure necessary to evaluate relevant information in a comprehensive and efficient manner.
       Figure 9 shows a generalized watershed ERA process that lends itself well to UAAs.
Using the watershed ERA approach, the UAA methodology relies on the identification and
evaluation of specific indicators (i.e., assessment endpoints) and measures that represent salient
characteristics of a given use (i.e., measures of effect) as defined by the standards.  Through
objective analysis of these measures (i.e., risk analysis), and comparison of those measures with
known minimum criteria, thresholds, or requirements needed to support the use, an assessment is
made as to the attainability of that use (i.e., risk characterization). Explicit in these analyses is a
determination of specific water body attributes  that are preventing attainability of a given use
(stressor identification evaluation or causal analyses). These attributes are evaluated to determine
whether certain modifications or controls would allow the use to be attainable and, if so, the
feasibility or reasonableness of those options (i.e., risk management).
       The first step in a UAA is problem formulation, which includes determination of the
assessment endpoints indicative of the use  in question, and conceptual model development.
Assessment endpoint selection (as well as measures of effect) should be identified using a
stakeholder process as described earlier in  this document. Because UAAs depend on at least one
of the six factors to be legally defensible, these  factors become central to conceptual model
development.  In practice, socioeconomic constraints (item 6 listed above) are handled differently
from ecological constraints (items 1 through 5 listed above) and are addressed outside the ERA
process.  The remaining five natural or anthropogenic factors then become categories of stressors
that help define relevant potential stressors and their sources in the watershed.  The assessment
endpoints identified during problem formulation give rise to specific measures of effect that can be
analyzed and compared with what is minimally needed to maintain that endpoint in the water
body. An example endpoint would be salmonid abundance and distribution for a beneficial use
that identifies coldwater aquatic life habitat.
                                            63

-------
   Step 1: Problem Formulation:
      •   Translate proposed use into a goal and objectives.
      •   Develop conceptual model.
      •   Identity assessment endpoints and measures of effect.
   Step 2: Risk Analysis
      •   Evaluate stressors, pathways, and measures of effect.
      •   Use reference conditions and watershed data to determine effects on
         assessment endpoints.
   Step 3: Risk Characterization
      •   Compare stressor levels and physicochemical regimes with minimum
         thresholds or criteria for assessment endpoints.
      •   Determine relevant causes of nonattainability; stressor identification
         evaluation.
   Step 4: Risk Management
      •  Identify control options if applicable.
      •  Consider feasibility of meeting use given candidate causes.
      •  Conduct stakeholder discussions and management option evaluation.
      •  Modify use if appropriate.
Figure 9.  General steps involved in conducting a use attainability analysis
(UAA) using a watershed ecological risk framework.
                                    64

-------
       In a UAA context, risk is analyzed in terms of the ability of an indicator or assemblage
maintaining itself given the five ecological factors as they exist in the water body.  Later steps in
risk characterization and risk management will determine whether present constraints in the water
body can and should be remediated to better meet the use.
       A UAA of two  small central valley streams in California followed a watershed ERA
framework. The conditions suitable to support coldwater ecosystems (COLD) include
preservation or enhancement of aquatic habitats, vegetation, fish, invertebrates, or wildlife. Both
salmonid habitat suitability indices and available habitat information for coldwater stoneflies were
used to assess suitability of the COLD use.  Measures were compared with criteria or "threshold
values" to determine whether each indicator is supported and the use is attainable.
       Using published habitat suitability parameters for salmonids, a simple conceptual model
was constructed to delineate potentially important pathways for determining attainability of the
COLD use with respect to fish. Physical habitat and chemical data indicative of habitat
requirements for assessment endpoints were collected and combined with available literature
information for the region to identify  likely stressors affecting the maintenance of obligate cool- or
coldwater aquatic life.  The relative strength of each cause-effect relationship was then determined
by considering the degree to which the indicator measurements differed from minimum thresholds
needed to maintain coldwater fauna.  For aquatic life uses, such as COLD, comparisons between
measured values and thresholds could be quantitatively determined in part using fish HSIs and
their various component values. The greater the disparity between the measured  HSI component
value for a given environmental factor (e.g., sediment particle size) and the minimum threshold, as
defined by the HSI,  the more limiting that particular factor is in preventing attainment of the use.
These quantitative assessments were often supplemented with qualitative evaluations.
       A decision tree was developed that presents the process used to determine whether the
COLD use is attainable. If analyses indicated that water body conditions do not meet minimum
thresholds for a particular use, the causes for nonattainment were then evaluated.  The poorer a
particular factor actually is, in comparison with minimum threshold criteria, the more limiting that
factor is in attaining the use and, therefore, the less  likely it is that the use is attainable. A weight-
of-evidence approach was used to synthesize results of the conceptual model approach to
determine overall attainability of the COLD use.  In this example, indicator results (i.e., trout  and
stonefly results for COLD use) were consistent. More consistency in results among indicators
(e.g., all indicators demonstrate a low probability of beneficial use attainment) suggests a higher
degree of certainty in the overall assessment regarding use attainability.  The more factors that do
not meet minimum threshold criteria,  the more certain that the use is not attainable under current
conditions.
                                            65

-------
       Using a watershed ERA framework, regulatory and resource agencies were able to quickly
and efficiently weigh management options and the feasibility of making the COLD use attainable
in these streams.  These UAAs confirmed that COLD use was not attainable in the streams
examined, given current hydrological and physical conditions.
                                          66

-------
                                   9.  CONCLUSIONS

       This report advocates integrating the watershed approach with ecological risk assessment
to make more informed environmental management decisions.  This integration offers numerous
benefits for those faced with the challenge of structuring environmental monitoring and analysis
efforts and using such information for decision making. The process fosters regular interactions
between scientists and managers to increase the likelihood that the optimal suite of scientific data
is collected, analyzed, and considered in decision making.
       Developing a management goal for a watershed ERA encourages organizations and the
community to work together to develop a common vision, share information, and understand
problems and ecological concerns.  The usefulness of a risk assessment is enhanced when
scientists and managers communicate regularly throughout the process.  Agreeing on the focus,
scope, and complexity of an assessment increases the likelihood that the collected and analyzed
data will be of value to and used by managers.
       The analysis phase of risk assessment addresses, to the  degree technology and resources
permit, the spatial and temporal co-occurrence of exposure of ecological resources to stressors
and the direct and indirect effects of such exposures to valued ecological resources. Based on a
few watershed ecological risk assessments conducted to date, risk analysis is the most challenging
phase of the process because there are typically multiple sources and stressors that may vary over
time and space. Tools such as GIS and CADDIS can assist with deciphering causes of ecological
degradation. Often it is useful to simplify the analytical approach in a watershed setting by
focusing on the impacts of one stressor or to use multivariate analysis to describe associations
between sources or stressors and effects. Risk characterization integrates multiple  lines of
evidence into a  clear summary of the risks to ecological resources. The risk characterization may
predict the magnitude of changes in environmental quality associated with alternative management
actions. It also  identifies critical knowledge gaps and describes the level of certainty of the
findings.
       Interim  findings need to be presented to managers to provide a clear view of the
significance of the findings and to enable conceptualizing  and possibly implementing potential
solutions.  The  analysis can also be refined based on new information. Maps, graphs, and tables
are among the best analytical tools for explaining relationships between stressors and effects.  The
presentation needs to be targeted for a specific audience, and more than one final presentation
format may be necessary.
       Documenting the impacts to valued ecological resources and the importance of their
protection helps justify management actions.  Even without direct regulatory authority,
environmental improvements can be attained because the increased awareness that occurs as a

                                            67

-------
result of documenting the consequences of human activities on the environment will encourage
positive behavioral changes and management actions. In addition, improvements in the
communication and coordination associated with the process can reduce duplication of effort.
       Elements of the watershed ERA process can be used alone or in conjunction with other
analytical decision making frameworks (e.g., TMDLs and UAAs).  Scientists, managers, and
stakeholders should communicate to determine the goals, objectives, focus, scope and complexity
of an assessment. Developing a conceptual model requiring an interdisciplinary team of scientists,
preferably with some local knowledge helps to better understand and communicate how potential
stressors might affect valued ecological resources. The analysis plan helps to focus the
assessment and to determine which data to collect and analyze.  It should be based on perceived
importance of stressors on assessment endpoints, corrective action potential, and resource
constraints.  The analysis plan should be considered an iterative process, where modifications are
made as results are obtained.
       Scientists, managers, and stakeholders should meet periodically to discuss interim findings
and modify or refine the thrust of analysis efforts based on intermittent deliberations as necessary.
With multiple stressors and multiple pathways, multivariate analyses or categorical ranking
schemes maybe needed to compare land use, stressors, and biotic measures. Conclusions need to
be based on integrating multiple lines of evidence and discussing the degree of uncertainty in the
findings.
       Presentation of assessment findings needs to be targeted for particular audiences. As a
result, more than one report may be needed.  Maps, GIS, and other visual-based graphics are
often very useful for communicating risk assessment results. All these steps will increase the
likelihood that environmental monitoring and assessment data will be used in decision making.
                                            68

-------
                                      GLOSSARY


       The following definitions are derived from EPA's Watershed Academy training module on
watershed ecological risk assessment (U.S. EPA, 2000b), Guidelines for Ecological Risk
Assessment (U.S. EPA, 1998a), and Terminology Reference System
(http://www.epa.gov/trs/index.htm).

Adverse ecological effects - Changes that are considered undesirable because they alter valued
structural or functional characteristics of ecosystems or their components. An evaluation of
adversity may consider the type, intensity, and scale of the effect as well as the potential for
recovery.

Assessment endpoint - An explicit expression of the environmental value that is to be protected,
operationally defined by an ecological entity and its attributes. For example, salmon are valued
ecological entities; reproduction and age class structure are some of their important attributes.
Together salmon reproduction and age class structure form an assessment endpoint.

Characterization of ecological effects - A portion of the analysis phase of ecological risk
assessment that evaluates a stressor's ability 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.

Coal fines - Fine particulate manner from coal mining operations.

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. 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 with the greatest risk reduction potential.

Conceptual model - A conceptual model in problem formulation for ecological risk assessment is
a written description and visual representation of predicted relationships between assessment
endpoints and the stressors to which they may be exposed. The conceptual model also describes
the sources of stressors, the ecosystem potentially at risk, the relationships between measures of
effect and assessment endpoints, and exposure scenarios.

Dose-response - How a biological organism's response to a toxic susbstance quantitatively shifts
as its overall exposure to the substance changes (e.g., a small dose of carbon monoxide may cause
drowsiness;  a large dose maybe fatal).
                                           69

-------
Ecological risk assessment - A process for analyzing environmental problems intended to
increase the use of ecological science in decision making.  The process evaluates the likelihood
that adverse ecological effects may occur or are occurring as a result of exposure to one or more
stressors.

Exposure - Co-occurrence of or contact between a stressor and a receptor.

Exposure pathway - The course a chemical or physical agent takes from a source to an exposed
organism. Each exposure pathway includes a source, an exposure point, and an exposure route.
Air or water maybe transport/exposure media if the source and endpoint are not in contact.

Lines of evidence - Information derived from different sources or by different techniques that
can be combined to reduce uncertainty and increase defensibility of a finding.

Linkage analysis - Terminology used in EPA's total maximum daily load (TMDL) protocols to
describe a quantitative analysis of stressor response, typically accomplished with a simulation
model.

Management scenario - A structured combination of management options selected for
evaluation.

Measure of effect - A change in an attribute of an assessment endpoint or its surrogate in
response to a stressor to which it is exposed.

Measure of exposure - A measure of stressor existence and movement in the environment and its
contact or co-occurrence with the assessment endpoint.

Problem formulation - The first phase of ecological risk assessment, which includes a
preliminary description of exposure and ecological effects, scientific data and data needs, key
factors to be considered, and the  scope and objectives of the assessment.  This phase produces
the risk hypotheses, conceptual model, and analysis plan, which serve as the basis for the
assessment.

Receptor - The ecological entity that is exposed to the stressor.

Risk analysis phase - A phase of ecological risk assessment consisting of two main parts: (1)
characterization of exposure—evaluating the interaction of the stressor with one or more
ecological entities, and (2) characterization of ecological effects—evaluating the ability of a
stressor(s) to cause adverse effects to an assessment endpoint under a particular set of
circumstances.

Risk characterization phase - 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. Lines of evidence and the adversity of effects are
discussed.
                                           70

-------
Risk estimation - Ideally, the conclusions of the risk characterization phase preferably expressed
in a quantitative manner (e.g., there is a 20% chance of 50% mortality under the circumstances
assessed), but also if expressed as a qualitative statement (e.g., there is a high likelihood of
mortality occurring).

Risk hypothesis - A candidate description of the relationship between a particular stressor and an
assessment endpoint. A risk hypothesis formulates proposed answers to questions risk assessors
have about what responses assessment endpoints will show when they are exposed to stressors
and how exposure will occur.

Risk management - The process of evaluating and selecting action alternatives in response to
risk assessment findings.  The goal of risk management is to recommend scientifically sound, cost-
effective actions that reduce or prevent risks while taking into account social, cultural, ethical,
political, and legal considerations.

Seascape  - The marine equivalent of landscape ecology and tied into the designation of special
areas as marine protected areas.

Source - An action that releases to the environment or imposes on the environment a chemical,
physical, or biological stressor or stressors.

Stakeholder - An individual or group influenced by, and with an ability to significantly affect
(either directly or indirectly), the topical area of interest.

Stressor - Any physical, chemical, or biological entity (agent) that can induce an adverse
response.

Stressor-response curve - A graphic, quantitative representation of the relationship between a
stressor (such as a pesticide concentration in the water column) and an ecological effect (such as
mortality of a given fish species if exposed to a certain concentration of that pesticide).

Tiered approach - An approach to assessment that commences with simple scoping analyses and
proceeds to additional tiers of more complex analyses as needed based on the results of earlier
tiers or information that becomes available as the assessment progresses.

Watershed approach - A framework for coordinating environmental management that focuses
public- and private-sector efforts on addressing the highest priority problems within
hydrologically defined geographic areas, taking into consideration both ground-water and surface-
water flow.

Watershed ecological risk assessment - An ecological risk assessment carried out at the
watershed level, usually in support of watershed management efforts.
                                            71

-------
                                           REFERENCES

Armitage, D. (1995) An integrative methodological framework for sustainable environmental planning and
management. Environ Manage 19(4):469-479.

Bossuyt, B; Muyssen, B; Janssen, C. (2005) Relevance of generic and site-specific species sensitivity distributions
in the current risk assessment procedures for copper and zinc. Environ Toxicol Chem 24:470-478.

Butcher, JB; Craig, J. (1998) Indicators for watershed management. In: Watershed management: moving from
theory to implementation. Water Environment Federation Specialty Conference, Denver, CO., May 3-6, 1998.
Alexandria, VA: Water Environment Federation; pp. 1063-1071.

Butcher, JB; Creager, CS; Clements, JT; et al. (1998) Watershed-scale ecological risk assessment. Alexandria,
VA: Water Environment Research Foundation; Project 93-IRM-4(A).

Cade, B; Terrell, J; Schroeder, R. (1999) Estimating effects of limiting factors with regression quantiles. Ecology
80:311-323.

Cook, RB; Suter II, GW; Sain ER.  (1999) Ecological risk assessment in a large river-reservoir:  1. Introduction
and background. Environ Toxicol Chem 18(4):581-588.

Cormier, SM; Smith, M; Norton, S; et al. (2000) Assessing ecological risk in watersheds: a case study of problem
formulation in the Big Darby Creek watershed, Ohio, USA. Environ Toxicol Chem 19(2): 1082-1096.

Delaware River Basin Commission. (2006) Commission meeting and public hearing. West Trenton, NJ. Available
online at www.nj.gov/drbc/commeet.htm.

Diamond,  JM; Serveiss, VB. (2001) Identifying sources of stress to native aquatic fauna using a watershed
ecological risk assessment framework. Environ Sci Technol 35:4711-4718.

Diamond, JM; Bressler, DW; Serveiss, VB. (2002) Assessing relationships between human land uses and the
decline of native mussels, fish, and macroinvertebrates in the Clinch and Powell River Watershed, USA. Environ
Toxicol Chem 21:1147-1155.

Field, J; MacDonald, D; Norton, S; et al. (2002) Predicting amphipod toxicity from sediment chemistry using
logistic regression. Environ Toxicol Chem 21:1993 -2005.

Foran, JA; Ferenc, SA. (1999) Multiple stressors in ecological risk and impact assessment.  SETAC Pellston
workshop on multiple stressors in ecological risk and impact assessment; September 13-18, 1997, Pellston, MI.
Pensacola, FL: Society of Environmental Toxicology and Chemistry (SETAC). 100 p.

Glicken, J. (2000) Getting stakeholder participation 'right': a discussion of participatory processes and possible
pitfalls. Environ Sci Policy 3:305-310.

Gregory, R; Keeney, R. (1994) Creating policy alternatives with stakeholder values. Manage Sci 40:1035-1048.

Harris, H J; Wenger, RB; Harris, VA; et al. (1994) A method for assessing environmental risk: a case study of
Green Bay, Lake Michigan, USA. Environ Manage 18:295-306.

Hession, WC; Storm, DE; Haan, CT; et al. (1996) A watershed-level ecological risk assessment methodology.
Water Res Bull  32(5): 1039-1054.

Hunsaker, CT; Graham, RL; Suter II, GW; et al. (1990) Assessing ecological risk on a regional scale. Environ
Manage 14(3):325-332.


                                                  72

-------
Junk, WJ; Bayley, PB; Sparks, RE. (1989)  The flood pulse concept in river- floodp lain systems. Canadian Journal of
Fisheries and Aquatic Sciences.

Karr, JR; Chu, EW. (1997) Biological monitoring: Essential foundation for ecological risk assessment. Human and
Ecological Risk Assessment 3:(6):993-1004.

Koenker, R; Bassett JG. (1978) Regression quantiles. Econometrica 46:33-50.

Landis, WG; Wiegers, JK. (1997) Design considerations and a suggested approach for regional and comparative
ecological risk assessment. Human Ecol Risk Assess 3(3):287-297.

Landis, WG; Luxon, M; Bodensteiner, LR. (2000) Design of a relative rank method regional -scale risk assessment with
confirmational sampling for the Willamette and MacKenzie River, Oregon. In: Price, FT; Brix, KV; Lane, NK; eds. Ninth
Symposium on environmental toxicology and risk assessment: recent achievements in environmental fate and transport.
West Conshohocken, PA: American Society of Testing and Materials; ASTM STP1381.

Long, E; Field, J; MacDonald, D. (1998) Predicting toxicity of marine sediments with numerical sediment quality
guidelines. Environ Toxicol Chem 17:714-727.

MacDonald, LH. (1994) Developing a monitoring project.  J Soil Water Conserv 49(3):221-227.

Maxwell, J. (1998) Ecosystem management by watersheds. Lakeview, CO: USDA Forest Service. Available online at
http://watershed.org/news/spr_93/ecoman.html.

McDaniels, TL. (2000) Creating  and using objectives for ecological risk assessment and management. Environ Sci Policy
3:299-304.

Newman, M; Ownby, D; Mezin, L; et al. (2000) Applying Species-sensitivity distributions in ecological risk assessment:
assumptions of distribution type and sufficient numbers of species. Environ Toxicol Chem 19:508-515.

Newbold, JD; O'Neill, RV; Elwood, JW; et al. (1982) Nutrient spiraling in streams: implications for nutrient limitation
and invertebrate activity. Am Natural 120:628-652.

Norton, SB; Cormier, SM; Smith, M; et al. (2002a)  Predicting levels of stress from biological assessment data: empirical
models from the eastern corn belt plains, Ohio, USA. Environ Toxicol Chem 21(6):1 168-1 175.

Norton, SB; Cormier, SM; Suter, GW; et al. (2002b) Determining probable causes of ecological impairment in the little
Scioto River, Ohio, USA: part 1. Listing candidate causes and analyzing evidence. Environ Toxicol Chem
NRC (National Research Council). (1996) Understanding risk. Informing decisions in a Democratic society. Washington,
DC: National Academy Press; 250 pp.

PCCRARM (Presidential/Congressional Commission on Risk Assessment and Risk Management). (1 997) Framework for
environmental health risk management. Washington, DC: The Presidential/Congressional Commission on Risk
Assessment and Risk Management.

Reckhow, KH. (1994) A decision analytic framework for environmental analysis and simulation modeling. Environ
Toxicol Chem 13(12):1901-1906.

Rhoads, BL; Wilson, D; Urban, M; et al. (1999) Interaction between scientists and nonscientists in community-based
watershed management: emergence of the concept of stream naturalization. Environ Manage 24(3):297-308.

Serveiss, VB. (2002) Applying ecological risk principles to watershed assessment and management. Environ
Manage 29(2):145-154.
                                                     73

-------
Serveiss, VB; Ohlson, DW. (2006) Using ecological risk assessment principles to protect drinking water in British
Columbia, Canada. Human Ecol Risk Assess: accepted.

Serveiss, VB; Bowen, JL; Dow, D; et al. (2004) Using ecological risk assessment to identify the major
anthropogenic stressor in the Waquoit Bay Watershed, Cape Cod, MA.  Environ Manage  33(5):730-740.

Serveiss, VB; Butcher, JB; Diamond, J; et al. (2005) Improving the TMDL process using watershed risk
assessment principles. Environ Manage 36(1):143-151.

Slocombe, DS. (1993) Environmental planning, ecosystem science, and ecosystem approaches for integrating
environment and development. Environ Manage  17(3):289-303.

Smeltzer, E; Heiskary, SA. (1990) Analysis and applications of lake user survey data. Lake Reserv Manage
Stahl, RG; Bachman, RA; Barton, AL; et al. (1999) A multi-stakeholder framework for ecological risk
management: Summary of a SETAC technical workshop.  Environ Toxicol and Chem 18(2):l-22.

Stewart, JS; Wang, L; Lyons, J; et al. (2001) Influences of watershed, riparian-corridor, and reach-scale
characteristics on aquatic biota in agricultural watersheds. J Am Water Resour Assoc 37:1475-1487.

Stewart, TJ; Scott, L. (1995)  A scenario-based framework for muilti criteria decision-analyis water resources
planning. Water Resour Res 31(ll):2835-2843.

Suter II, GW. (1998) An overview perspective of uncertainty. In: Warren-Hicks W J; Moore, DRJ. Uncertainty
analysis in ecological risk assessment. Pensacola, FL: SETAC Press; pp. 121-130.

Suter II, GW. (1999) Developing conceptual models for complex ecological risk assessments. Human Ecol Risk
Assess 5:375-396.

Suter II, GW; Norton, SB; Cormier, SM. (2002) A methodology for inferring the causes of observed impairments
in aquatic ecosystems. Environ Toxicol Chem 21(6):1 101-1 111.

Townsend, CR.  (1989) The patch dynamics concept of stream community ecology. J North Am Benthol Soc
8:36-50.

U.S. EPA (Environmental Protection Agency). (1992) A framework for ecological risk assessment. Risk
Assessment Forum, Washington, DC;  EPA/6 3 0/R- 92/001. Available online at http://www.epa.gov/ncea/raf

U.S. EPA (Environmental Protection Agency). (1994) Water quality standards handbook, 2nd ed. Office of Water,
Washington, DC; EPA/823/B-94/005a. Available online at
http ://www.epa.gov/waterscience/library/wqstandards/handbook. pdf.

U.S. EPA (Environmental Protection Agency). (1995a) Watershed protection: a project focus.  Office of Water,
Washington, DC; E PA/ 84 1/R- 95/003. Available online at http://www.epa.gov/ow.

U.S. EPA (Environmental Protection Agency). (1995b) Guidance for Assessing chemical contaminant data for use
in fish advisories. Vol. IV, Risk Communication. Office of Water, Washington, DC; EPA/823/R-95/001.
Available online at http://www.epa.gov/waterscience/library/fish/fishvolume4.pdf.

U.S. EPA (Environmental Protection Agency). (1996) Watershed approach framework. Office of Water,
Washington, DC; EPA/840/S-96/001.  Available online at http://www.epa.gov/owow/watershed/framework.html.
                                                  74

-------
U.S. EPA (Environmental Protection Agency). (1997a) Community-based environmental protection: a resource
book for protecting ecosystems and communities.  Office of Policy, Planning, and Evaluation, Washington, DC;
EPA/230/B-96/003. Available online at http://www.epa.gov/ecocommunity/pdf/coverrev.pdf.

U.S. EPA (Environmental Protection Agency). (1997b)  Top 10 watershed lessons learned.  Office of Wetlands,
Oceans, and Watersheds (450IF), Washington, DC; EPA/840/F-97/001.  Available online at
http ://www. epa.gov/owow/lessons/.

U.S. EPA (Environmental Protection Agency). (1998a) Guidelines for ecological risk assessment. Risk
Assessment Forum, Washington, DC; EPA/630/R-95/002F. Available online at http://www/epa.gov/raf.

U.S. EPA (Environmental Protection Agency). (1998b) Restoring and protecting America's waters. Interagency
report marking 25 years of Clean Water Act benefits. Office of Water, Washington, DC; EPA/840/R-98/001.
Available online at http://www.epa.gov/owow/tmdl/sediment/pdf/sediment.pdf.

U.S. EPA (Environmental Protection Agency). (1999a) Rapid bioassessment protocols for use in streams and
wadeable rivers: periphyton, benthic macroinvertebrates and fish. 2nd ed. Office of Water, Washington, DC;
EPA/84l/B-99/002. Available online at http://www.epa.gov/owow/monitoring/rbp/.

U.S. EPA (Environmental Protection Agency). (1999b) Protocol for developing sediment TMDLs.  Office of
Water, Washington, DC; EPA/841/B-99/004. Available online at
http://www.epa.gov/owow/tmdl/sediment/pdf/sediment.pdf.

U.S. EPA (Environmental Protection Agency). (2000a) Report on the watershed ecological risk characterization
workshop.  National Center for Environmental Assessment, Washington, DC; EPA/600/R-99/111.  Available
online at http://www.epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2000b) Watershed ecological risk assessment.  Office of Wetlands,
Oceans, and Watersheds.  Available online at
http ://www. epa.gov/OWOW/watershed/wacademy/acad2000/ecorisk/.

U.S. EPA (Environmental Protection Agency). (2000c) The model plan for public participation. Office of
Enforcement and Compliance Assurance, Washington, DC; EPA/300/K-00/001. Available online at
http ://www.epa.gov/compliance/resources/publications/ej/model_public_part _plan.pdf.

U.S. EPA (Environmental Protection Agency). (2000d) Stressor identification guidance document.  Office of
Water, Washington, DC; EPA/822/B-00/025. Available online at
http://www.epa.gov/waterscience/biocriteria/stressors/stressorid.pdf.

U.S. EPA (Environmental Protection Agency). (2000e) Ecoregional nutrient criteria.  Office of Water, Office of
Science and Technology, Washington, DC; EPA/822/F-00/007. Available online at
http://www.epa.gov/waterscience/criteria/nutrient/ecoregions/.

U.S. EPA (Environmental Protection Agency). (2000f) Nutrient criteria technical guidance manual. Rivers and
streams. Office of Water, Office of Science and Technology, Washington, DC; EPA/822/B-00/002. Available
online at http://www.epa.gov/waterscience/criteria/nutrient/guidance/rivers/index.html.

U.S. EPA (Environmental Protection Agency). (2001a) Improved science-based environmental  stakeholder
process, a commentary by the EPA science advisory board. EPA Science Advisory Board, Washington, DC; EPA-
SAB-EC-COM-01-006. Available online at http://epa.gov/sab/pdf/eccm01006.pdf.

U.S. EPA (Environmental Protection Agency). (200Ib) Stakeholder involvement & public participation at the U.S.
EPA, lessons learned,  barriers, & innovative approaches.  Office of Policy, Economics, and Innovation,
Washington, DC; EPA/100/R-00/040. Available online at http://www.epa.gov/publicinvolvement/pdf/sipp.pdf.
                                                   75

-------
U.S. EPA (Environmental Protection Agency). (2002a) Clinch and Powell Valley watershed ecological risk
assessment.  National Center for Environmental Assessment, Washington, DC; EPA/600/R-01/050.  Available
from the National Technical Information Service, Springfield, VA, and online at http://www.epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2002b) Section 319 success stories, volume III: the successful
implementation of the Clean Water Act's Section 319 non point source pollution program.  Office of Water,
Washington, DC; EPA 841-S-01-001.  Available online at http://www.epa.gov.ncepihom.

U.S. EPA (Environmental Protection Agency). (2002c) Waquoit Bay watershed ecological risk assessment.
National Center for Environmental Assessment, Washington, DC; EPA/600/R-02/079. Available from the
National Technical Information Service, Springfield, VA, and http://www.epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2002d) Ecological risk assessment for the Middle Snake River
watershed.  Office of Research and Development, Washington, DC, and Region  10, Seattle, WA; EPA/600/R-
01/017. Available online at http://www.epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2004) Development Plan  for the causal analysis, diagnosis decision
information system (CADDIS). National Center for Environmental Assessment, Washington, DC; EPA/600/R-
03/074. Available from the National Technical Information Service, Springfield, VA, and
http ://www. epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2005) Conceptual models for watershed and regional risk
assessments [External Review Draft]. National Center for Environmental Assessment, Washington, DC;
EPA/600/P-04/100A. Available from: National Technical Information Service,  Springfield, VA, and
http ://www. epa.gov/ncea.

U.S. EPA (Environmental Protection Agency). (2006) Nonpoint source pollution (polluted runoff): A great threat
to our environment.  Region 2, Water, New York, NY. Available online at
http://www.epa.gov/region02/water/npspage.htm

Vannote, RL; Minshall, GW; Cummins, KW; et al. (1980) The river continuum concept. Can J Fish Aquat Sci
37:130-137.

Vollenweider, RA; Kerekes, JJ. (1980) OECD eutrophication program, synthesis report. Paris, France:
Organization for Economic Cooperation and Development.

Walker, R;  Landis, W; Brown P. (2001) Developing a regional ecological risk assessment: a case study of a
Tasmanian agricultural catchment. Human Ecol Risk Assess 7(2):417-439.

Ward, JV; Stanford, JS. (1983) The serial discontinuity concept of lotic ecosystems. Ch. 2. In: Fontaine, TD;
Bartell, SM; eds. Dynamics of lotic ecosystems. Ann Arbor, MI: Ann Arbor Science Publishers.

Ward, RC.  (1996) Water quality monitoring: Where's the beef? Water Resour Bull 32(4):673-680.

Ward, RC;  Loftis, JC; Mcbride, GB. (1986) The "Data-rich but information poor" syndrome in water quality
monitoring. Environ Manage 10(3):291-297.

Wenger, RB; Rong, Y. (1987) Two fuzzy-set models for comprehensive environmental decision-making. J Environ
Manage 25:167-180.

Wiegers, JK; Feder,  HM; Mortensen, LS; et al. (1998) A regional multiple-stressor rank-based ecological risk
assessment for the Fjord of Port Valdez, Alaska. Human Ecol Risk Assess 4(5):1125-1173
                                                  76

-------