Part II Planning and Problem Formulation Before everything else, getting ready is the secret to success Henry Ford Before risks are analyzed, it is necessary to lay the groundwork by defining the problem and the means to solve it This process is driven by the needs of the manager who will use the results of the assessment to make a decision, but it is constrained by the limitations of time, resources, and technical capabilities The EPA’s framework for ecological risk assessment (Section 3.1) distinguishes between planning, which involves the risk manager and potentially the stakeholders, and problem formulation, which is performed by the assessors alone The intent was to comply with the National Research Council’s (1983) mandate to keep policy and science distinct However, the distinction seldom holds up in practice because the formulation of the problem also involves the interpretation of policy The best example is the EPA’s Data Quality Objectives (DQO) process that requires the risk manager to participate in the development of the analysis plan, the last step in problem formulation (Quality Assurance Management Staff 1994) Further, the distinction is not generally recognized outside the United States Finally, this distinction is unlikely to convince skeptics that risk assessment is not biased by the policies of its sponsor External peer review is a better guard against bias in the technical aspects of an assessment Therefore, the planning of an assessment and formulation of the problem are presented here as an integrated process The distinction between routine assessments and novel assessments has more practical importance Assessments of new chemicals and similar cases may follow standard policies and procedures In such cases, goals, endpoints, and management options are set for all assessments, and little planning or problem formulation is required The assessors need not consult with the risk manager for each assessment and, after assembling the data, may proceed to the analysis and characterization of risks At the other extreme, site-specific assessments and assessments of novel agents such as acid deposition, a genetically modified fish, or global warming typically require an extended process of planning and problem ß 2006 by Taylor & Francis Group, LLC 146 Ecological Risk Assessment formulation Multiple meetings may be held with the stakeholders and the public, and the process may be iterated as assessors develop conceptual models, endpoints, and other planning products for review and revision It is the responsibility of the risk manager to determine how routine or ad hoc the planning and problem formulation should be However, it is the assessor’s responsibility to inform the managers about cases in which routine policies and methods may not be applicable For example, a new chemical may have physical or toxicological properties that suggest the need to consider novel endpoints or may have patterns of use that suggest the need for novel management options The success of risk assessments and the resulting management decisions depends on the quality of the planning and problem formulation process If the problem formulation is done in a haphazard manner, the resulting assessment is unlikely to be useful to the risk manager The process should be taken as seriously as the performance of toxicity tests or the creation of a hydrologic model and should be done with at least as much care In some cases, the completion of a problem formulation may be sufficient to drive a management decision That is, once the nature of the situation is delineated, the need for action or the appropriateness of inaction may be apparent without quantitatively assessing the risks ß 2006 by Taylor & Francis Group, LLC 10 Impetus and Mandate Assessors should understand why they are performing an ecological risk assessment and the powers and limitations of the decision maker These may be briefly and clearly defined, as in the regulation of new chemicals, or broadly and ill defined, as in the assessments of risks from climate change On the basis of the paradigm of the release of agents by sources resulting in exposures that cause effects, the impetus for assessment falls into one of three categories: sources and agents, exposures, or effects Sources and agents are the most common and familiar They include sources such as waste outfalls, new technologies or new resource management plans and hazardous entities such as chemicals, exotic organisms, and dredges Such assessments support the issuance of permits or approvals to operate a facility, market a product, import an organism, release an effluent, harvest a resource, or build a structure The usual management options are to approve, disapprove, or approve with restrictions Effect-initiated risk assessments are prompted by the observation of ecological impairments such as the fishless lakes that prompted the aquatic component of the National Acid Precipitation Assessment (Baker and Harvey 1984), or the deformed frogs that have prompted assessments of the risks to anurans (Burkhart et al 2000) These ecoepidemiological assessments are characterized by the need to perform a causal analysis before the risks from alternative actions can be considered (Section 4.3) Exposure-initiated assessments are the least common For chemicals, a risk assessment may be initiated by the observation of elevated body burdens For exotic organisms one may be initiated by observations of a new pathogen in an organism or new species in an ecosystem Exposure-initiated assessments require determination of the sources as well as the risks of the effects For example, the observation of high mercury levels in fish has led to risk assessments that determine the sources as well as the risks to humans and piscivorous wildlife (EPA 1995) In addition, the EPA (1998a) identifies value-initiated assessments These are cases in which a valued aspect of the environment is the subject of the assessment No sources, agents, exposures, or effects are identified as requiring analysis Instead, sources, agents, exposures, and effects must be hypothesized or identified through observation, measurement, or modeling Examples of such impetus include rare or high-valued species or places such as parks or watersheds A specific example is the Big Darby Creek watershed in Ohio, which was assessed because it has an exceptionally high-quality biotic community including rare and endangered fish and mussels (Cormier et al 2000; Serveiss 2002) These value-initiated assessments must evolve into one or more assessments based on a conventional impetus, i.e., effects, exposures, or sources of agents must be identified before risks can be assessed ß 2006 by Taylor & Francis Group, LLC 11 Goals and Objectives If you not know where you are going, any road will take you there The White Rabbit to Alice (Lewis Carol) The planning of an ecological risk assessment depends primarily on the goal of the management action to be supported Most environmental laws in the United States provide rather vague goals such as ‘‘protect public health and the environment’’ or ‘‘protect and restore the physical, chemical and biological integrity of the Nation’s waters.’’ Agencies that implement laws should interpret those vague goals in more concrete terms that can be evaluated For example, the International Joint Commission (1989) interpreted the biological impairment goal for Lake Superior thus: ‘‘The Lake should be maintained as a balanced and stable oligotrophic ecosystem with lake trout as the top aquatic predator of a cold-water community and with Pontoporeia hoyi as a key organism in the food chain.’’ Such specific goals, called objectives, may apply to an entire regulatory or management program, or may be assessmentspecific A programmatic example is the European Commission’s goals for their water quality objectives (WQOs) Accordingly, a WQO should be such as to permit all stages in the life of aquatic organisms to be successfully completed should not produce conditions that cause these organisms to avoid parts of the habitat where they would normally be present should not give rise to the accumulation of substances that can be harmful to the biota (including man) whether via the food chain or otherwise and should not produce conditions that alter the functioning of the ecosystem (CSTE= EEC 1994) Examples of appropriate management goals in the EPA’s guidelines include ‘‘reduce or eliminate macroalgal growth’’ and ‘‘maintain diversity of native biotic communities’’ (EPA 1998a) Goals for site-specific or ‘‘place-based’’ assessments may be generated through workshops or other consensus-building processes Goals for public lands or other natural resources are often contained in their management plans In addition to these goals for a specific law or assessment, it may be possible to define national environmental goals However, goal setting is probably the most inconsistent and ill-defined component of the ecological risk assessment process (McCarty and Power 2001) In any case, careful thought should be devoted to defining goals (Box 11.1) However derived, ecological goals provide the basis for identification of the assessment endpoints Some goals are defined in terms of desired attributes and require no comparison Examples for fish species include: (1) the endangered species should persist for at least 50 years after the action, (2) the fishery should support a median yield of 100 MT, or (3) no kills should occur However, it is often necessary to define goals relative to a reference condition As discussed in ß 2006 by Taylor & Francis Group, LLC BOX 11.1 Goals for Urban Streams A goal of the US Clean Water Act is to ‘‘protect and restore the physical, chemical and biological integrity of the Nation’s waters.’’ The biological integrity goal obviously requires some clarification The most common approach is to define relatively undisturbed streams in the least disturbed watersheds within a region as having integrity and then to develop an index or other indicator to define a scale of loss of integrity relative to the undisturbed stream (Yoder and Rankin 1995b) However, since even developments that result in only 10% to 20% impervious surfaces in a watershed cause significant changes in stream communities, it is not possible to achieve that sort of integrity in urban streams Rather, somewhat degraded standards such as ‘‘modified warm-water habitat’’ are created for such streams An alternative would be to develop definitions of biological integrity for urban streams, based on what is desirable and possible in those conditions This need not be an unambitious goal It may require expensive projects and controversial regulations to eliminate combined sewer overflows, store, treat and slowly release storm water, eliminate toxicity and high nutrient levels in effluents, reduce residential pesticide and fertilizer use, and create habitat structure However, the goal of a high quality urban stream rather than a somewhat less impaired stream could provide significant incentives and psychosocial benefits This would require more than a change in the semantics of the goal The metrics that define the degree of departure from an ideal undisturbed stream would not be the best metrics to define the departure from a high quality urban stream For example, it may be impossible to reduce temperatures sufficiently to support trout and darters, but it may be possible to sustain an urban fishery of catfish and sunfish that differs from undisturbed streams in the region but has an integrity of its own Hence, the goals would be to achieve design criteria for designated uses including recreation, flood control, aesthetics, and recreational fisheries, rather than minimal departure from a regional reference state This would require the development of a practice of urban aquatic ecology that would be equivalent to urban forestry succeeding sections, the definition of reference is usually treated as a technical problem during the definition of assessment endpoints and the development of the analysis plan However, as the urban stream example illustrates (Box 11.1), the choice of reference has policy implications The results of management will be different if the goal is to achieve an attribute of undisturbed and uncontaminated ecosystems (e.g., wilderness steams) of some percentile of all ecosystems (e.g., the tenth percentile of all streams arrayed from highest to lowest quality), of a historical reference (e.g., the community composition reported in the first records), or of high quality urban streams Hence, the bases for comparisons should be defined by decision makers during the goal setting process Ideally, the management goals would also specify the decision criteria Thresholds for effects, three-part logics, cost-effectiveness, cost–benefit, net environmental benefit, or other decision criteria should lead to different risk assessments For example, thresholds for acceptability may be based on any variety of metrics, but cost–benefit or net benefit analyses require that the expected changes in the environment be clearly specified and quantified ß 2006 by Taylor & Francis Group, LLC 12 Management Options Environmental risk managers have limited range of options that are determined by legal and regulatory constraints and by practical constraints For example, a US EPA remedial project manager for a Superfund site may mandate a variety of remedial actions, such as removal or capping of contaminated soil, but cannot order the purchase and protection or restoration of replacement ecosystems Practical constraints include monetary limitations; public acceptance; and limitations of the available science and technology for remediation, restoration, or treatment Within those constraints, the management options considered should be those that could potentially achieve the management goals These options must include not only environmental goals but also human health goals and socioeconomic goals The management options determine the range of future conditions to be assessed For a new chemical, that range may be simply the conditions with and without the use and release of that chemical For pesticides, it could be the conditions resulting from the range of uses defined by the manufacturer and those associated with various restraints on use, such as no application within a prescribed distance of a water body For forest fires it would include allowing the fire to burn or sending various sizes of fire crews and support teams, so the future conditions might include not only the extent of the fire, but also the effects of fire breaks, flame retardants, etc The result of consideration of management options should be descriptions of each alternative action that will be assessed and compared in the risk characterization Hence, they must include sufficient detail to allow the analyses and characterization to be performed For example, it is not sufficient to identify wastewater treatment as an option Rather, the quality and quantity of the treated water must be specified as well as the likely frequency and consequences of treatment failures ß 2006 by Taylor & Francis Group, LLC 13 Agents and Sources The subject of most risk assessments is an agent that has been or will be imposed on the environment such as a new pesticide, an effluent, a timber harvest, or an exotic species In some cases, the impetus for the assessment is a new source of such agents These include new technologies such as a new coal liquefaction technology, facilities such as a highway, activities such as off-road motorcycle races, plans such as a timber management plan, or even cultural pressures such as suburban sprawl that will impose multiple agents on the environment The assessment must begin with a reasonably complete description of the agent and its source That information is typically provided to regulatory agencies by the applicant for a permit, and the required information is specified in regulations However, if the source is contaminated media, an established exotic species or equivalently existing ambient source, then it and its associated agents must be characterized by census, sampling, or analysis If such data are not available, the characterization of sources and agents must be included in the analysis plan 13.1 EMISSIONS Conventionally the design of a facility, including any regulatory restrictions, determines what may be released from a facility, how it is released, and how frequently This information is commonly known as the source term Source terms typically estimate the characteristics of normal operations including variance in concentrations and release rates due to variance in raw materials, product mix, and conditions However, they should go further and estimate the variance due to unintended releases Chemicals and other agents may be released by spills and other accidents, by fugitive releases (e.g., leaks around gaskets or during pouring), or by operation during startup, shutdown, or upset conditions (e.g., when operating temperatures or pressures are not achieved or when switching between products) In addition, many chemicals get into the environment even though they are not intentionally released (PCBs) or even not intentionally produced (chlorinated dioxins) In some cases, unintended releases are characterized by engineering risk assessments These assessments evaluate the results of potential accidents such as tank ruptures or upset conditions such as power loss (Rasmussen 1981; Haimes 1998; Wang and Roush 2000) There are two basic types of analysis for accidents Fault trees define the causal links from a defined failure (e.g., a pipe rupture resulting in a spill) back to all of the potential initiating events (e.g., corrosion, vandalism, impact of a vehicle) Event trees begin with an initiating event (e.g., a stuck valve) and trace all of its possible consequences (e.g., safe shutdown, pipe rupture) Probabilities are assigned to each node in a tree so that probabilities of a failure given an event or probabilities that an event caused a failure can be estimated In some cases, standard models are available for assessing risks of unintended releases to the environment For example, LANDSIM was developed to simulate release of contaminants from landfills in Britain with various design features (Environment Agency 1996) The importance of engineering risk assessment is highlighted by disastrous accidents such as the Exxon Valdez oil spill ß 2006 by Taylor & Francis Group, LLC and the dike failure on the Tisza River in Romania that released cyanide and metals resulting in mass mortality of aquatic life 13.2 ACTIVITIES AND PROGRAMS In some cases, the subject of the assessment is an activity or program rather than a chemical or emission Examples include forest management plans, military training exercises, mining activities, irrigation projects, or construction and operation of ports or other facilities In such cases, the source characterization is an analysis of the activity to determine what hazardous agents will be employed or generated It may appropriately begin by listing the phases of the activity ranging from, e.g., site exploration through construction and operation to decommissioning Other activities my have little temporal phasing but may have distinct constituent activities that may be listed For example, a military exercise may include driving tracked vehicles, live fire of weapons, aircraft overflights, and excavation of defenses In either case, listing of the constituent activities or phases focuses the source characterization and assures that no major component of the activity is overlooked For each constituent activity or phase, the potentially hazardous aspects of the activity to be assessed should be listed For example, driving a vehicle off-road involves crushing plants, striking animals, compacting soil, rutting the soil surface, and starting fires For repeated activities such as forest management plans or military exercises, a checklist based on prior assessment experience may be used For novel activities, lists must be generated by brainstorming exercises involving experts on the activity, experts on the site, and possibly stakeholders 13.3 SOURCES OF CAUSES If the impetus for the assessment is an observed biological effect, it is necessary to determine the cause of the effect and then the source of the causal agents (Chapter 4) In such assessments, the characterization of the source or sources is part of the assessment rather than an input to the assessment (Section 19.2) 13.4 PROPERTIES OF THE AGENT In addition to characterizing the source, the problem formulation must identify the properties of the agent that are potentially relevant to the risk assessment These include both exposurerelated and effects-related properties To develop the conceptual model it is necessary to know, at least qualitatively, what the agent does in the environment Where does it go, what is its fate, and what are its effects on the various organisms and processes of ecosystems? Traditional human health and ecological risk assessments have focused on identifying the hazards and characterizing stressors However, ecological risk assessments must recognize that they are assessing agents such as nutrients, fire, temperature, and habitat modification that are beneficial at typical levels and may be stressful only at exceptionally high or low levels Hence, it is important to describe the range of effects of an agent and not simply those that are potentially harmful 13.5 SOURCES OF INDIRECT EXPOSURE AND EFFECTS The sources to be identified are usually the primary anthropogenic sources: pesticide releases, waste emissions, dams, explosives, etc However, experienced ecological assessors know that the effects of those primary sources are themselves sources of secondary exposures and ß 2006 by Taylor & Francis Group, LLC Training plan Tank maneuvers Compressed and disturbed soil Crushed organisms Erosion Silt in streams Plant community Siltation of benthos Species dependent on rocky substrates Feeding Population processes Feeding Plant survival and growth Low-mobility species Habitat use Herbivorous wildlife species Alarmed organism Behavioral responses Species with critical behavior that may be disrupted FIGURE 17.3 A generic conceptual model for an activity, tank training, that is a source of hazardous processes The rectangles are states of entities and hexagons are processes linking those states (Previously published in Suter G.W II, Hum Ecol Risk Assess., 5, 397, 1999 With permission.) depends on ecological factors that cannot be quantified Display of each step in the elaboration can make the logic of the model clear to the reader Break the model into modules: Often, when creating a conceptual model of a complex system, the modelers can get lost in the tangle of boxes and arrows To avoid this, it is often helpful to define component modules of the system In that way, relatively simple toplevel and low-level models can be created for each of the modules (Suter 1999b) For example, a conceptual model of South Florida used a hierarchy of models of society, ecosystems (e.g., marl prairie and Biscayne Bay), and species (e.g., Florida panther) (Gentile et al 2001) A more conventional example might be an assessment of a waste site that contributes leachate to a creek The overall conceptual model of the system includes the biota of that creek as a module along with a waste module, groundwater module, terrestrial biota module, etc The creek biota would be elaborated in a conceptual submodel with uptake of the contaminant by various species or trophic groups, movement of the contaminant through the food web, and primary and secondary effects (Figure 17.5) This approach lends itself to object-oriented ß 2006 by Taylor & Francis Group, LLC Employment Catch (a) Catch Harvesting effort Employment Stock size Fleet size Implicit ecological factors (b) Growth Harvesting effort Fleet size Capital investment Stock size Catch Employment Recruitment Depreciation Natural mortality Interest rates (c) Product prices Implicit economic factors FIGURE 17.4 A conceptual model of risks to a fishery, built up in steps from the relationship that is fundamental to the decision (Adapted from Walters, C.J., Adaptive Management of Renewable Resources, Macmillan, New York, 1986 With permission.) programming that allows the independent creation and updating of component models Modules may be defined in terms of distinct spatial units, distinct processes (e.g., eutrophication), or distinct types of entities (e.g., species or ecosystems) The conceptual models of the modules may be more detailed, or may represent the mechanistic processes at a lower level of ß 2006 by Taylor & Francis Group, LLC Surface water Detritus Sediment Aquatic biota Plants emergent Plants periphytic Plants planktonic Aquatic invertebrates Fish periphyton feeders Fish detritivores Zooplankton Fish invertebrate feeders Fish plankton feeders Fish piscivores Terrestrial food web FIGURE 17.5 A stream biota module from a larger conceptual model of ecological risks from waste disposal The module is connected to inputs from contaminated water, sediment, and detritus and is in turn connected to a terrestrial food web module through feeding by piscivorous and insectivorous wildlife organization For example, a module for uptake of a chemical by an organism may simply show all of the routes of uptake, or may represent physiological toxico-kinetics (Section 22.9) Linking of standard component modules: If certain sources, endpoint entities, or processes appear repeatedly in ecological risk assessments, then generic components can be created for ß 2006 by Taylor & Francis Group, LLC Flood Frost Mortality Vegetation Food Vegetation Mice Mouse cover Prey Wintering hawk abundance FIGURE 17.6 A receptor conceptual model for influences on the abundance of wintering hawks (Modified from Andrewartha, H.G and Birch, L.C., The Distribution and Abundance of Animals, University of Chicago Press, Chicago, 1984 and previously published in Suter, G.W II, Hum Ecol Risk Assess., 5, 397, 1999 With permission.) repeated use (Suter 1999b) This is particularly likely when multiple risk assessments must be performed for the same site due to multiple activities or sources These would fall into three categories First, receptor conceptual models can be created for particular species or communities that are the objects of multiple assessments The receptor models would represent all of the significant influences on the endpoint receptors (Figure 17.6) Conceptual models of this sort, termed influence models are important tools in ecology (Andrewartha and Birch 1984) Second, source conceptual models can be created for individual categories of sources or activities that generate the agents to be assessed (Figure 17.3) Examples might include buried wastes, coal fired power plants, logging operations, livestock grazing, or sewage treatment plants The models would illustrate the release of chemicals, removal of resources, physical modification of ecosystems, and other activities that may affect the environment If the source initiates a common causal chain, such as eutrophication from nutrient releases, it should be included in the generic source model Third, site conceptual models would describe features of the site (actual or defined by a scenario) that can causally link the sources and endpoint receptors These could include hydrologic models, atmospheric models, and food web models (Figure 17.7) The generation of a conceptual model for an assessment would involve selecting appropriate source, site, and receptor models and linking them (Figure 17.8) The linkage would involve identifying the logical relations among the three types of models For example, release of any aqueous effluent or any runoff would link to the surface hydrology model, which would show how the materials moved through the streams, flood plains, etc., and any persistent chemicals would link to the aquatic food web models Components of any model that not provide a connection between source and receptors would be pruned For example, the tank training source model (Figure 17.3) might be linked to a desert ß 2006 by Taylor & Francis Group, LLC Y12 plant Poplar creek Y12 burial grounds X10 burial grounds X10 plant Bear creek White-oak creek East fork Poplar creek Oak Ridge city Poplar creek embayment K25 plant Clinch river FIGURE 17.7 A site conceptual model, hydrologic transport of contaminants on the Oak Ridge Reservation, Tennessee (Previously published in Suter, G.W II, Hum Ecol Risk Assess., 5, 397, 1999 With permission.) pronghorn receptor model at all three terrestrial community links, but the population processes path would be pruned because pronghorns would not be crushed Linkages that are peculiar to the assessment would be added Generic conceptual models: If the situation being assessed is repeated and sufficiently consistent, generic conceptual models may be used Examples might include effluents from sewage treatment plants on small rivers in the upper midwestern United States, insecticide application to irrigated cotton, or release of surfactants from domestic laundry It is important not to extend such generic conceptual models beyond their range of applicability, and to modify them as needed to reflect the peculiar aspects of individual cases Brain storming and pruning: One may begin by developing a model that includes all components and relationships that any member of the assessment team, manager, or stakeholder feels is relevant One may then prune the model by deleting components that are not actually part of the system, relationships that are not possible (e.g., movement of contaminants into an isolated aquifer), components or relationships that would not influence the decision, and components or relationships that cannot be quantified and are not of sufficient importance to prompt research or testing (e.g., dermal exposure of reptiles) This approach is inefficient but is likely to serve the heuristic and communication functions discussed previously Representing implemented models: The fundamental purpose of conceptual models is to direct the selection and development of the implemented mathematical models of exposure and effects Conversely, as those models are selected and developed, the conceptual model should be modified to represent the way that risks are quantitatively modeled For example, if plant uptake from contaminated soil by both roots and leaves is simulated, it is important to ß 2006 by Taylor & Francis Group, LLC Activity Activity Activity conceptual models Consequences Sources Hydrology Food web Atmosphere Site conceptual models Transport and consequences Exposure routes Endpoint receptor Endpoint receptor Endpoint receptor Receptor conceptual models FIGURE 17.8 A representation of the process of linking modular conceptual models to create a complete conceptual model (Previously published in Suter, G.W II, Hum Ecol Risk Assess., 5, 397, 1999 With permission.) represent the two pathways However, if an empirical soil–plant uptake factor is used, the conceptual model would show the soil–plant transfer without separating the routes 17.4 LINKAGE TO OTHER CONCEPTUAL MODELS The conceptual model for ecological risk assessment should be consistent with other conceptual models developed for assessments to inform the same decision There will always be a human health risk model and sometimes an engineering risk model, or a socioeconomic model At least, the conceptual models should be consistent They should show the same sources, the same routes of transport, etc In addition, the ecological risk model should link to the other risk models Output from one should be input to another One way to achieve this is to create an integrated conceptual model with engineering, human health, ecological, socioeconomic, and common constituents The common constituents would be modules such as a conceptual model of chemical transport and fate that provides input to multiple assessments Such integration of conceptual models can lead to efficiencies in assessment (e.g., the human health and ecological assessors not duplicate effort) and better-informed decisions because the results of the various assessments are consistent (Chapter 37) ß 2006 by Taylor & Francis Group, LLC 18 Analysis Plans A limited data base is no excuse for not conducting a sound risk assessment On the contrary, with less knowledge of a system, the need for risk assessment and management becomes imperative Haimes (1998) The last step in problem formulation is the development of a plan for performing the assessment It should define the data that will be collected, the models and statistical or logical analyses that will be applied, and the type of results that will be presented It should also describe the methods that will be used to generate data and approaches that will be used to develop new models It should define the milestones and completion date of the assessment, the level of effort, and the cost It should also provide a quality assurance plan that specifies the measures taken to reduce uncertainty and the expected magnitudes of uncertainty in the results (Chapter 9) For routine assessments, such as those for new chemicals, the analysis plan may be defined by a standard protocol However, for novel assessments and site-specific assessments, it will be necessary to develop a plan de novo Typically, the analysis plan must be reviewed and approved by the risk manager, who will seek assurance that the planned work will provide the information needed for the decision, and by any others who may control the budget, staffing, and schedule The assessment team generates the analysis plan by synthesizing the other components of planning and problem formulation They begin by considering what changes in what attributes of ecological entities (assessment endpoints) must be estimated, how those changes might be brought about (source descriptions and conceptual models) and under what circumstances (environmental descriptions and scenarios) Then they must determine, given the limitations of time and resources, the best way to estimate those changes That is, given the combination of existing information and the possibilities for testing, measurement, modeling, and inference, what is the best combination of methods to apply for this assessment? The analysis plan may include plans for tiering the assessment (Section 3.3) These plans include a screening assessment using existing data (if not already performed in the problem formulation), a screening assessment performed with data from preliminary sampling and analysis, and a definitive assessment with a full data set Each tier should have its own analysis plan that describes the performance of that tier in some detail, and general plans for subsequent tiers Analysis plans for screening tiers should assure that they provide the basis for the next tier’s problem formulation, focusing the assessment on critical pathways, receptors, and locations, and providing a basis to design sampling or testing activities 18.1 CHOOSING MEASURES OF EXPOSURE, EFFECTS, AND ENVIRONMENTAL CONDITIONS Development of analysis plans is largely a matter of choosing the best measures of exposure, effects, and conditions This is typically begun by identifying measures of exposure for each of ß 2006 by Taylor & Francis Group, LLC the agents and exposure–response measures, which are appropriate to the measure of exposure and the endpoint entities and attributes Types of measures of commonly assessed categories of agents in aquatic ecosystems are presented in Table 18.1 and are described in the following chapters Ideally, the measures would be chosen so as to provide the most accurate and complete estimate of risks The Data Quality Objectives (DQO) process, described in Chapter 9, is a method for identifying such ideal data sets In fact, the DQO process may be considered an alternative to problem formulation, as a means of developing an analysis plan in cases that are as simple as health risk assessments However, the selection process is often constrained by resource limitations, the desire to use standard or conventional measures, the need to accommodate existing data, and the data requirements of existing models For example, the best measure of effect may be a time-to-death model, TABLE 18.1 Associations Between Measures of Exposure and Measures of Effects from Controlled Studies for Different Types of Aquatic Stressors Agents Characterization of Exposure Chemical External concentration in medium Effluent Internal concentration in organism Biomarker Dilution of effluent Contaminated ambient media Location and time of collection Analysis of medium Habitat Structural attributes Water level or flow Thermal energy Hydrograph and associated summary statistics (e.g., 7Q10) Temperature Silt (suspended) Suspended solids concentration Silt (bed load) Degree of embeddedness, texture Dissolved oxygen and oxygen-demanding contaminants Excess mineral nutrients Dissolved oxygen Concentration Pathogen Nonindigenous species Presence or abundance of pathogens Presence or abundance of the species Characterization of Exposure–Response Concentration–response or time–response relationships from laboratory or field Effluent dilution-response tests Field tests in plume Lab or in situ tests using the medium Medium dilution-response Medium gradient-response Empirical models (e.g., habitat suitability models) Instream flow models Thermal tolerances Temperature–response relationships Concentration–response relationships from laboratory or field studies Empirical siltation–response relationships from laboratory or field Oxygen concentration–response relationships from laboratory or field Empirical concentration–response relationships from laboratory or field Eutrophication models Disease or symptoms Ecological models (food web, energetic, predator–prey, etc.) Source: Modified from US Environmental Protection Agency, Stressor Identification Guidance Document, EPA= 822=B-00=025, Washington, DC, Office of Water, 2000; and Suter, G.W., II, Norton, S.B., and Cormier, S.M., A methodology for inferring the causes of observed impairments in aquatic ecosystems, Environ Toxicol Chem., 21, 1101–1111, 2002 With permission ß 2006 by Taylor & Francis Group, LLC but, if the only data for acute lethality is an LC50 and there are no resources for testing, then the LC50 is used, and the exposure data must be related to that test As suggested by the example, the analysis plan should assure that the dimensions of the problem are reflected in the measures and that the different measures are concordant (Chapter 23) The dimensions of exposure and effects for chemicals and similar agents are discussed in Section 6.4 Once exposure ends, time to recovery is another relevant component of the temporal dimension, but it does not relate to the risk of an effect It becomes relevant when considering net benefits of an action or when comparing the total effects of heterogeneous alternative actions (Section 33.1.5) Hence, the choice of measures of exposure and effects should be based on an understanding of which of these dimensions must be quantified, how they are to be scaled and quantified, and how they are related (Chapter 6) 18.2 REFERENCE SITES AND REFERENCE INFORMATION When performing a risk assessment for a site or region, information concerning the baseline state or a nominally uncontaminated or undisturbed state is needed for purposes of comparison and normalization For example, one may need to know the concentrations of contaminant metals in uncontaminated media or the number of fish species in unchannelized streams This is referred to as reference information Background is a special case of reference sites Background sites are those that are believed to represent an uncontaminated or undisturbed state In many cases, this ideal cannot be achieved For example, the atmospheric deposition of anthropogenic mercury is ubiquitous Therefore, one can identify negative reference sites for mercury with respect to a specific mercury source, but not true background sites In practice, the distinction between reference and background is unclear Nevertheless, background concentration is an important concept in risk assessment, because they are used to determine what concentrations of metals and other naturally occurring chemicals should not be remediated and are not even worth assessing (Chapter 31) Therefore, it is important, in the analysis plan, to identify sites that will be treated as effective background sites The sources of reference information must be clearly specified in the analysis plan Several sources of such information can be recognized 18.2.1 INFORMATION CONCERNING THE PRECONTAMINATION OR PREDISTURBANCE STATE For the assessment of future disturbances or contamination, the proposed site may serve as its own reference For example, if a dam or a wastewater treatment plant is proposed for a stream, then the current biota of the stream serves as a reference against which the predicted future conditions are compared Surveys of the stream could provide the basis for that comparison, focusing in particular on the state of endpoint entities For the assessments of ongoing disturbance or contamination, studies may have been done at the site prior to the contamination or disturbance Such historic information may have been developed for permitting or other regulatory purposes, for resource management, for research, for a monitoring program, for environmental impact assessments, or for evaluation by private conservation organizations Hence, it is important to contact past landowners, local universities, regulatory and resource management agencies, and private conservation organizations to determine whether the information is available However, such information must be used with caution Information may not be of expected quality for risk assessment, it may be so old as to no longer be relevant, or it may have been obtained using inappropriate methods For example, chemical analyses may not have achieved adequate limits of detection ß 2006 by Taylor & Francis Group, LLC Some useful data may be truly historical When agriculture, resource harvesting, fire suppression, or similar activities have long modified an ecosystem, reference information may come from the writings of early visitors, early photographs, or museum collections from early biologists Such records have documented the conversion of grasslands to shrub lands and of open forests to closed canopy forests However, it must be noted that these records reflect the management practices of indigenous or premodern peoples, not a natural state (Redman 1999) 18.2.2 MODEL-DERIVED INFORMATION In some cases, no measurements or observations can supply adequate reference information For example, there may be no undisturbed or uncontaminated streams in the area In such cases, it may be possible to use a model to estimate characteristics of the uncontaminated state In particular, habitat models may be used to estimate the biota of a site in the absence of contamination or physical disturbance (US Fish and Wildlife Service 1987; Wright et al 1989) The results of such models are seldom sufficiently accurate to determine that a decline in abundance has occurred, but they can be used to indicate that species are absent or severely depleted that should be present in abundance 18.2.3 INFORMATION CONCERNING OTHER SITES The most common source of reference information is the study of reference sites, i.e., sites that resemble the site being assessed except for the presence of the disturbance or contaminants Commonly, the reference is a single site For streams and rivers, a single upstream site may be used and for terrestrial sites, a single location that is clearly beyond the contaminated or disturbed area is typically chosen This approach is inexpensive, and, if exposure and effects are clear, it may be sufficient However, a pair of sites may differ for any number of reasons other than disturbance or contamination For example, upstream and downstream sites differ due to stream gradients In addition, the use of a single reference site does not provide an adequate estimate of variance in the reference condition, because repeated samples from a single reference site are pseudoreplicates (Hurlbert 1984) (Box 5.1) Pseudoreplication is perhaps best explained here by a relevant example The San Joaquin kit fox population on the Elk Hills Naval Petroleum Reserve, California, crashed in the early 1980s, and it was suspected that the oil field wastes were responsible A study was conducted, which analyzed the elemental content of fur from the foxes to determine whether they had elevated exposures to the wastes (Suter et al 1992) The sponsor initially insisted that reference information be limited to analyses of fur from foxes trapped on the periphery of the oil field However, a pilot study using that reference indicated that the concentrations of several elements were statistically significantly higher in the fur from the oil field than on reference fur A subsequent study, which included areas away from oil fields and another oil field, found that the initial comparison had been misleading The metal concentrations in fur from Elk Hills oil field foxes were not high relative to these other reference sites Rather, the fur from foxes at the peripheral reference site was unusually low in metals Without true replication of reference sites, an incorrect conclusion might have been drawn about the cause of the observed decline The best solution to this problem is the one employed in the Elk Hills study; obtain information from a number of reference sites sufficient to define the nature of and variance in the reference conditions In addition to the obvious problem of increased cost, the chief problem in this approach is finding enough suitable sites Selection of reference sites involves two potentially conflicting goals First, relevant properties of the reference sites should resemble those of the assessed site, except for the presence ß 2006 by Taylor & Francis Group, LLC of the contaminants or disturbance The relevance of properties of the sites depends on the comparisons to be made For example, if a contaminated stream reach has a natural channel, but upstream reaches are channelized, that difference is probably not relevant to comparisons of metal concentrations, but it precludes using the upstream reach as a reference for population or community properties Second, the reference sites should be independent of the assessed site and of each other For example, if animals move between the sites, measurements of body burdens, abundances, or any other attribute involving those mobile organisms cannot be used for comparison, because the sites are not independent In most cases, the sites that resemble the assessed site most closely are those that are adjoining or nearby, but those are also sites that are least likely to be independent of the assessed site The problem of deciding whether reference and assessed sites differ in any relevant way requires knowledge of physics, chemistry, or biology controlling the properties to be compared For example, the occurrence of heavy metals in soils is a function of the cation exchange capacity, so that factor must be considered, but that relationship does not depend on the identity of the clay minerals providing the capacity For ecological comparisons the relevant habitat parameters must be similar, but the parameters that must be similar depend on the taxa being assessed For example, the fish species composition of a water body is affected by the diversity of habitat types, but is unlikely to be significantly affected by moderate differences in production However, condition measures such as weight and length ratios are likely to be affected by production As a result of considerations such as these, the availability of appropriate reference sites may constrain the measures of exposure and effect that are used It is pointless to measure a property of a site if it is not possible to obtain relevant reference information against which to compare it The EPA has suggested parameters that should be included when determining the similarity of sites, but that guidance cannot substitute for site-specific consideration of the relevance of differences in site characteristics to the types of comparisons among sites that will be part of the assessment (Office of Emergency and Remedial Response 1994) In some cases, suitable criteria for similarity may already exist The best example is the soil classification systems developed for the United States by the Natural Resource Conservation Service and elsewhere by other agencies For the most part, a reference site with the same soil type as an assessed site is suitable for measures of soil properties such as metal concentrations However, for other properties such as abundance of soil invertebrates, additional factors such as compaction, organic matter content, and vegetation must be considered In any case, the soil type is a good starting point in selecting reference soils Another example would be the use of the same stream characteristics that have been used to establish regional references (see below) to select reference stream reaches for local comparisons The magnitude of difference, i.e., necessary to declare a potential reference site to be acceptably similar to assessed sites, is determined by the magnitude of exposure or effects to be discriminated, and by judgment and experience For example, experienced invertebrate ecologists know approximately how much difference in sediment texture is likely to cause a 25% difference in the species composition of the community The magnitude of difference to be detected should be determined by interaction with the risk manager The independence problem requires that we consider not only factors that introduce common biases in the sites, but also factors that inappropriately reduce intersite variance For example, it is inappropriate to simply establish reference sites at different distances upstream and to use them to estimate reference variance The Oak Ridge Reservation’s Biological Monitoring and Abatement Program addressed this problem by monitoring a suite of reference streams both on and off the reservation This use of multiple reference streams establishes information equivalent to a regional reference, but more localized There are technical and policy elements in the selection of a set of reference sites, because it is ß 2006 by Taylor & Francis Group, LLC necessary to decide what variance is relevant and what properties must be independent For example, all soils in Oak Ridge Reservation are slightly contaminated with mercury due to atmospheric deposition of mercury, used at the Y-12 plant, and coal combustion from nearby power plants However, it was decided that low-level contamination was not relevant to determining which sites could be considered background, and therefore the lack of independence of mercury concentrations among reference sites on the reservation was judged to be irrelevant 18.2.4 INFORMATION CONCERNING A REGIONAL REFERENCE Regional reference information is a special case of reference information from other sites It is derived from a population of sites within a region that encompasses the assessed site and that is deemed to be acceptably uniform and undisturbed with respect to the properties of interest One example is the use of the US Geological Survey’s summaries of elemental analysis of soils from the eastern and western United States as reference soil concentrations for sites in those regions (Shacklette and Boerngen 1984) Some states have compiled background concentrations for environmental media (Slayton and Montgomery 1991; Webb 1992; Toxics Cleanup Program 1994) Regional reference information is available for biotic communities in many states due to bioassessment programs (Davis and Simon 1995; Barbour et al 1996) For example, the state of Ohio has been divided into ecoregions, and reference fish and benthic invertebrate community properties have been defined for each (Yoder and Rankin 1995) The use of regional reference information has advantages Regional reference information is available from the responsible agency, eliminating the need for reference sampling; it is based on multiple independent sites so it accounts for variance; and, when the reference information is generated by a regulatory agency, it is likely to be acceptable to that agency Regional reference information has some disadvantages One is that the data used to establish the regional reference may not be suitable for risk assessment Detection limits for analyses of reference media may be too high or the quality assurance and quality control may not be up to the standards required by regulators or responsible parties For example, the results of metal analyses of uncontaminated waters performed prior to the early 1990s were often too high because of inadvertent contamination during sample handling and analysis (Windom et al 1991) And the other is that, the regional reference may not include the parameters that are needed to estimate the exposure or the assessment endpoint, as specified by the problem formulation For example, the abundance or production of a particular fish species may be an assessment endpoint for a site, but regional reference values for fishes are commonly defined in terms of community properties In addition, the ubiquity of contamination and disturbance may result in reference sites that are degraded This is the case in Ohio, where the 25th percentile of the distribution of reference community indices is the threshold for acceptability (Yoder and Rankin 1998) That is, 25% of the best streams in Ohio are declared unacceptable Finally, the risk managers often prefer site-specific reference information This preference is particularly likely when the region is large, making the relevance of the reference information questionable 18.2.5 GRADIENTS AS REFERENCE An alternative to the use of multiple reference sites to provide reliable reference information is the use of gradients This is particularly useful where there is a gradient of contamination or where there is a biological or physical gradient One use of gradient analysis is to establish a reference concentration of a contaminant in terms of the asymptotic concentration This approach is particularly useful for sites where the extent of contamination is unknown ß 2006 by Taylor & Francis Group, LLC An analogous use is to establish the natural gradient of stream community properties as a reference against which to compare the state of the community in assessed reaches For example, fish species richness naturally increases along downstream gradients due to increased habitat quantity and diversity Therefore, sampling fish on a gradient from well upstream of the site to well downstream can establish both the natural gradient and any deviations from it associated with the site Gradient analyses eliminate the problem of balancing the need for similarity against the need for independence, by eliminating both That is, if the relevant factors follow a gradient, then the lack of similarity and the lack of independence are incorporated into the gradient model Gradients are relatively seldom used in ecological risk assessments, partly because their existence is not recognized, and partly because confounding disturbances or contaminant sources may disrupt the natural gradient 18.2.6 POSITIVE REFERENCE INFORMATION In addition to the conventional or negative reference information discussed above, positive reference information may be needed Positive reference information concerns contaminated or disturbed sites that are not the sites being assessed While the purpose of a negative reference is obvious, the uses of positive references are more diverse and difficult The most common is the upstream reference on a stream that has sources of contamination upstream of the source being assessed For example, Poplar Creek, Tennessee, is contaminated by various sources on the Oak Ridge Reservation that were assessed for possible remediation, but also by upstream sources including coal mines and a small city Analyses and tests performed on water and sediment samples from Poplar Creek upstream of the reservation provided a measure of the level of contamination and toxicity to which Reservation’s sources were an addition Therefore, they provided an indication of the improvement that might be expected if remediation was implemented Another use of positive reference is to determine whether a hypothesized cause of observed effects is plausible For example, Poplar Creek receives mercury from its East Fork If the effects observed in Poplar Creek were due to that mercury input, the same or greater effects with the same symptomology should have been observed in lower East Fork, where the mercury is less dilute The fact that aquatic toxic effects were not observed in the East Fork indicates that mercury was at most a minor contributor to the observed effects in Poplar Creek Positive controls, tests of toxic levels of well-characterized chemicals, serve a similar function in toxicity testing They should be used with site-specific toxicity tests to demonstrate that the tests, as performed, are sensitive to concentrations of chemicals that are known to cause relevant toxic effects Cadmium chloride and other well-studied chemicals have been recommended as positive control chemicals, but for tests at contaminated sites, chemicals that are contaminants of potential concern at the site or at least have the same mode of action should be selected 18.2.7 GOALS AS AN ALTERNATIVE TO REFERENCE In some cases, no appropriate reference is available because contamination or disturbance is ubiquitous, because a desirable ecological condition no longer exists in the region, or because the desired state may have never existed (e.g., a high-quality urban stream—Box 11.1) In such cases, management goals serve in place of reference conditions as a standard for comparison As discussed in Chapter 11, goals may be set in a variety of ways However, if they are to serve in place of reference conditions, goals must be very specifically and clearly defined, like an assessment endpoint For example, 30% eelgrass cover might be a goal for an estuarine bay and reproducing populations of three fish species might be a goal for an urban stream ß 2006 by Taylor & Francis Group, LLC ... Environmental Protection Agency, Generic Ecological Assessment Endpoints (GEAEs) for Ecological Risk Assessment, EPA=630=P- 02= 004B, Washington, DC, Risk Assessment Forum, 20 03 Specific definition of structural... Assessment Endpoints for Human and Ecological Risk Assessments Entities Organism-level attributes Individual organism Human Health Risk Assessment Ecological Risk Assessment Population of organisms... Reagan (20 02) has made this ß 20 06 by Taylor & Francis Group, LLC TABLE 16 .2 The US EPA’s (20 03c) Generic Ecological Assessment Endpointsa Entity Organism-level endpoints Organisms (in an assessment