4
Setting Priorities for Action

RISK ASSESSMENT AND DECISION ANALYSIS BASICS

Risk assessment and decision analysis are tools for organizing and analyzing information in a systematic way and in the face of uncertainty to help identify the best way or ways of tackling a complex problem. By being systematic, these tools also encourage documentation of the methods used. This allows others to apply them in slightly different situations or using different values for a variety of variables. In the logical paradigm developed for evaluating risks to humans from exposure to various contaminants, the National Research Council (NRC 1983) described risk assessment as a technique for evaluating the probability and severity of an adverse outcome, while risk management is a technique for deciding on the best options for reducing risk.

The information available about the causes of declines of Atlantic salmon populations in Maine is incomplete. Therefore, it is not obvious what actions should be taken and in what order to reverse those declines. The Atlantic Salmon Conservation Plan for Seven Maine Rivers (Maine Atlantic Salmon Task Force 1997) is a thoughtful analysis of the causes of salmon declines and ways to reverse them. This committee agrees in general with that report, with a few exceptions. The value that we have attempted to add here is in (1) our ranking of the factors affecting salmon in terms of their likely severity, (2) our prioritizing the various management options available in terms of their likelihood of being effective and in terms of cost, (3) our suggesting a sequence for undertaking those options,



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 108
Atlantic Salmon in Maine 4 Setting Priorities for Action RISK ASSESSMENT AND DECISION ANALYSIS BASICS Risk assessment and decision analysis are tools for organizing and analyzing information in a systematic way and in the face of uncertainty to help identify the best way or ways of tackling a complex problem. By being systematic, these tools also encourage documentation of the methods used. This allows others to apply them in slightly different situations or using different values for a variety of variables. In the logical paradigm developed for evaluating risks to humans from exposure to various contaminants, the National Research Council (NRC 1983) described risk assessment as a technique for evaluating the probability and severity of an adverse outcome, while risk management is a technique for deciding on the best options for reducing risk. The information available about the causes of declines of Atlantic salmon populations in Maine is incomplete. Therefore, it is not obvious what actions should be taken and in what order to reverse those declines. The Atlantic Salmon Conservation Plan for Seven Maine Rivers (Maine Atlantic Salmon Task Force 1997) is a thoughtful analysis of the causes of salmon declines and ways to reverse them. This committee agrees in general with that report, with a few exceptions. The value that we have attempted to add here is in (1) our ranking of the factors affecting salmon in terms of their likely severity, (2) our prioritizing the various management options available in terms of their likelihood of being effective and in terms of cost, (3) our suggesting a sequence for undertaking those options,

OCR for page 108
Atlantic Salmon in Maine and (4) our suggesting a framework for others to do similar analyses when conditions change, when new information is available, or with different values attributed to various outcomes and costs. Evaluating and Ranking Threats to Maine Atlantic Salmon An intractably large number of threats to Atlantic salmon have been identified. For example, a Canadian group of experts recently identified 63 factors threatening the survival of Atlantic salmon in eastern North America (Cairns 2001). No feasible amount of time and resources could be enough to understand and mitigate such a large number of threats. Fortunately, we do know that some threats are more important than others, and furthermore, some threats must be mitigated before others can be addressed. For example, if barriers prevent salmon from ascending a river, then those barriers must be made passable before improving habitat above them could be of any use. Many documents and presentations read and heard by the committee gave the impression that perhaps the biggest difficulty in knowing how to rehabilitate salmon is seeing the forest for the trees. What are the most important things to do and which of them should be done first? To approach a solution to this problem, the committee developed a conceptual framework or risk-assessment model for thinking about it that involved identifying and ranking the threats and their contribution to salmon mortality. This framework considers a range of issues that apply across the watersheds in Maine where Atlantic salmon could potentially be restored. However, the committee has not considered in detail mitigation options for the significant issue of at-sea mortality because the committee recognizes the large knowledge gap in being able to ascribe causation. (The hatchery living gene-bank program at Maine’s Craig Brook Fish Hatchery is in part an ocean mitigation program. The parr are raised to adulthood in the freshwater of the hatchery, rather than having to become mature in the sea, where survivorship is very low.) The committee acknowledges the importance of at-sea mortality as a threat factor and strongly supports the need for further research to better understand mechanisms and possible remedial measures. The committee similarly has not attempted to evaluate the range of responses to potential threats that could be induced by climate change, because that issue is much larger than conservation planning efforts in Maine can reasonably address. As noted earlier, the committee’s initial work focused on understanding the genetic status of Atlantic salmon in Maine (NRC 2002a) in response to its charge. At the same time, the committee was gathering, organizing, assimilating, and discussing a wide and diverse range of pertinent data and information. Inevitably, we retraced the path of earlier

OCR for page 108
Atlantic Salmon in Maine teams of scientists and managers in the United States, Canada, and elsewhere in the world where threatened and endangered anadromous fish (particularly Atlantic and Pacific salmon) have been studied (Cairns 2001; Maine Atlantic Salmon Task Force 1997; NMFS and FWS 1999; NRC 1996a). At the end of each path, we encountered the same obstacles met by earlier efforts—a seemingly intractable number of variables, some of which were quantified with detailed data and others that were clouded by uncertainty. Like our predecessors, we could not rely solely on deductive reasoning and reductionist methods to understand a complex (in both space and time) environmental problem that had taken shape over several centuries. In other words, we faced information overload in some areas and daunting gaps in others with no readily apparent means of reaching sound and timely conclusions. Our individual and collective frustration was increased by the rapid declines in Atlantic salmon populations, despite the best efforts of dedicated scientists and managers, and the corresponding urgency of research and restoration efforts. Once the application of risk assessment and decision analysis methods was proposed by a small subset of our members, the entire committee was unified and energized by the prospect of breaking the impasse and fostering the use of the adaptive management paradigm for conservation of Atlantic salmon in Maine—to help see the forest for the trees. It has been observed that all models are wrong, but some are useful. In this case, we believe the strengths of risk assessment and decision analysis methods substantially outweigh their shortcomings and weaknesses. Some key strengths include (1) the systematic yet flexible process for diagramming complex systems, (2) the need to consider proportional influences and interaction effects at different levels, and (3) the impetus for improving input data and conducting sensitivity analyses to update and refine estimates. The primary weaknesses are held in common with virtually all modeling methods. First, the process yields an incomplete mathematical abstraction of the environment (natural and anthropogenic). Second, the weakest parameter estimate(s) limits accuracy and utility of the results. Nevertheless, risk assessment and decision analysis methods help to guide and fuel adaptive management efforts. A few notes on the mechanics of the risk assessment process may be helpful. The bubble diagram of Figure 4-1 illustrates the committee’s view of the relationship between humans and a viable wild salmon population, through ecological, direct, and genetic influences. As the group described by Cairns (2001) did, our committee drew on its expert judgment based on personal insights and experience, the information in the literature, the information in the many briefs and presentations received at committee meetings, and so forth, to assign proportional values to the impact factors. For example, the committee estimated that more than half of all the hu-

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-1 Categories of human impacts on wild salmon. man influence on the viability of wild salmon populations is through ecological factors and assigned 0.6 or 60% of the total influence on wild salmon viability to ecological factors. We estimated that direct and genetic influences have roughly equal influences, and therefore each received 0.2 or 20% of the total impact on viability. The fact that the influences sum to 1.0 (0.6 + 0.2 + 0.2) and account for all of the means by which humans affect the viability of a wild salmon population can give the appearance of an artificially high accuracy of assignment (nothing left unexplained; precise allocation to alternatives). But this appearance of precision is not intended by the committee. Instead, our view is that Figure 4-1 provides nothing more than an informed estimate of the relative weighting of impact factors, and that later investigators or new information may well lead to the revision of these estimates. Indeed, this capacity for revision based on improved data is one of the strengths of the risk assessment technique. To fully understand the mechanics of the process, it is useful to see how the analyses are structured and how proportions or probabilities are multiplied and then added to generate the final estimates for the relative importance of impact factors.

OCR for page 108
Atlantic Salmon in Maine Analyzing the Information The committee has based much of its analysis and many of its conclusions on the bubble diagrams (Figures 4-1, 4-2, 4-3, and 4-4). Before providing additional detail about the process and use of the bubble diagrams, we emphasize again that the numbers and letters in the bubble diagrams were developed for heuristic purposes. They are not random numbers pulled from the air: they are informed estimates. The numbers cannot be considered data but rather help to identify where the greatest impacts to salmon might be and where data are most likely to be useful. The committee began by asking what the known and potential sources of human-caused salmon mortality are. Using its own experience, general biological judgment, and many publications and other sources of information, the committee listed human-caused threats to salmon. We then categorized them into ecological factors, genetic factors, and direct factors. Ecological factors act by degrading the environment’s ability to support salmon productivity (survival and reproductive success, or “fitness”). They include such items as water quality and quantity, obstructions to passage, changes in availability and quality of spawning and rearing habitat, presence of nonnative species that likely compete with or prey on FIGURE 4-2 Subcategories of ecological impacts on wild salmon.

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-3 Subcategories of direct impacts on wild salmon. salmon, and so on. Genetic factors act to reduce salmon productivity by reducing the quality of their genetic adaptations and thus reducing inherent capacity to respond to their environment within their lifetimes (e.g., appropriate predator avoidance) and, in some cases, the population’s ability to respond to environmental change evolutionarily, across generations. They include inbreeding, domestication selection, breakdown of co-adapted gene complexes through lack of mate choice, genetic drift due to small population size, the incorporation of genes into the population from nonnative or nonlocal populations, and so on. Finally, direct factors are human actions that directly kill adult or juvenile fish. They include incidental and targeted fishing, turbines in hydroelectric plants, the killing of fish through research, and so on. These three categories cover all major sources of salmon mortality. Thus, the committee was able to take a rather extensive list of threats to salmon and compile these into three categories, which could then be considered for their proportional impor-

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-4 Subcategories of genetic impacts on wild salmon. tance as impact factors on salmon viability. The committee considered what reasonable limits could be assigned to the contributions of individual factors in each category. It seemed reasonable that no impact factor should be less than 5% of the total to avoid having to consider too many very small factors. Given that more than half the original spawning habitat of Atlantic salmon in Maine is no longer available to them because of obstructions to passage, and given the presence of additional ecological factors, it seems clear that ecological factors contribute more than half of all sources of human-caused salmon mortality in Maine. The committee therefore assigned the ecological category a total contribution of 0.6. Given that much direct mortality of salmon has been reduced or even eliminated, especially fishing, it seemed appropriate to allow that factor to be one-third as big as the ecological factor (0.2). Similarly, the genetic factors that have affected salmon are likely to be important but not nearly as important as the ecological factors, so they also were given an overall value of one-third that of the ecological factors (0.2). These three values become level 3 in our analysis—they portion out the relative contribution that all other factors at level 2 and level 1 above them can make to the viability of a wild

OCR for page 108
Atlantic Salmon in Maine salmon population. For instance, the myriad of ecological factors will, when considered in sum, account for 0.6 or 60% of the loss of viability of wild salmon. The committee then addressed the many ecological factors that have been identified in other similar studies (e.g., Cairns 2001, Maine Atlantic Salmon Task Force 1997, NRC 1996a) and tried to position them in a relative sense. It did this by considering the various pathways outlined in the Figure 4-2 bubble diagrams. At level 2, the level that feeds directly into level 3, it was possible to identify three major abiotic factors (water quality, habitat, passage) and three major biotic factors (predator-prey relationships, disease, competition). The committee judged that changes to salmon habitat, physical passage (adults upstream, juveniles downstream), and predator-prey dynamics had large ecological effects, and each received a weighting of 0.28 or 28% of the total contribution to ecological impacts. By contrast, the impact of changes in water quality suggested only half the importance and received a weighting of 0.14 or 14% of the ecological impact on salmon viability. Similar considerations led to assignment of 0.12 to disease and 0.06 to competition. Whenever possible, our estimates were based on the available literature, but in many cases, the values are illustrations of our analysis so that others can provide better numbers if and when better information is available. Finally, at level one, we identified nine immediate consequence of human activity, including dams, irrigation, roads, agriculture, acidity, logging, hatcheries, introduction of exotics, and aquaculture activities, that feed into level two. Dams, for example, were estimated to account for 0.4 or 40% of the changes to water quality, 0.6 of the changes to habitat other than water quality, 0.85 of the passage problems, and 0.15 of the changes in predator-prey dynamics. By contrast, dams did not appear to contribute to ecological impacts through disease or competition. Similar methods were used to develop relative values for the other factors at level one. Each factor at level two could thus be attributed to the inputs from level 1. We were therefore able to account for 100% of the ecological impact on viability of wild salmon (0.6 of the total impact on wild salmon population, Figure 4-1) through the action of humans. Humans also directly affect wild salmon, as illustrated in Figure 4-3. Human activity can directly increase the mortality of adults and that of juveniles, with our rough estimate at level two being 0.9 through adults and 0.1 juveniles. At level one, the impact on adults is through fisheries, including poaching (0.6); incidental activities, such as impacts by boating (0.15); research activities, such as tagging (0.15); and involvement in hatcheries, such as handling (0.1). Juvenile mortality is not affected by directed fisheries, but juveniles can suffer incidental captures (0.33), are frequently in research programs (0.33), and experience direct mortality from hatch-

OCR for page 108
Atlantic Salmon in Maine ery programs in which they are collected from the wild (0.34). For example, some statistics are available on fishing mortality due to catch-and-release fishing, landing statistics are available, rough estimates of removal are available, and so on. Finally, humans can have strong impacts on the viability of wild salmon through their genetics (Figure 4-4). The genetic quality of wild salmon probably has four components (level 2; see also Appendix D), including good genes (best genes in the population), complementary genes, (ideal matches at a locus within an individual), co-adapted gene complexes (ideal matches among loci within an individual), and diverse genes (heterozygosity across loci). Since diverse genes are so often the target of conservation biology, the committee gave this category twice the weighting (0.4) of the others (0.2 each). In turn, at level one, human activity operates primarily through programs of aquaculture and hatcheries. Aquaculture, for example, has very strong effects on good genes as what is good within the fish farm (e.g., delayed maturity) is often different from what is good in nature. Since hatchery production does not use targeted selection to maximize survival in the artificial environment, its good genes impact (0.1) is less than that of aquaculture (0.9). Aquaculture and hatcheries may have similar impacts on complementary and coadapted genes, as neither allow mate choice, but aquaculture will have a large impact on diverse genes, as there is usually less attention paid to maintaining heterozygosity and more effort to producing a specific strain of fish. Level 0 in Table 4-3 represents the “penetrance ratio” of hatcheryfish genes as compared with farm-fish genes. For the table, the assumption is that the ratio is 9:1 in favor of hatchery fish, i.e., the value of level 0 for hatcheries is 9 and for aquaculture, or farms, it is 1 (but see Ranking the Threats, below, for a more detailed discussion of this assumption). Level 1 is described above. Level 1b results from calculating level 0 times level 1 using the formula level 1b = (level 0 times level 1) divided by [(level 0 times level 1) aquaculture + (level 0 times level 1) hatcheries]. Throughout its analysis, the committee chose as its target a viable wild populations with a genetically effective population size (Ne) of 1,000 or greater and a probability of surviving for 100 years from now without reliance on a hatchery of 95% or greater. The outcome of the analyses are tabulated in Tables 4-1 for ecology, 4-2 for direct impacts, and 4-3 for genetics. As an example, consider the ecological impacts of dams (Table 4-1). The effects of dams originate at level 1, feeding into level 2 and level 3. Thus, to understand the full impact of dams on viable wild salmon through impacts on water quality, we take the value of 0.4 at level 1 (impact of dams on water quality) and multiply this by 0.14, which is the relative impact of water quality on ecology. We multiply again by 0.6, which is the relative impact of ecology on salmon viability. This suggests

OCR for page 108
Atlantic Salmon in Maine TABLE 4-1 Proportional Impacts of Ecological Components on Viable Wild Atlantic Salmon Populations Based on a Risk Assessment (refer to Figures 4-1 and 4-2) Component Level 1 Level 2 Level 3 Product (1x2x3) Impact (Sum) Dams Water quality 0.4 0.14 0.6 0.0336   Habitat 0.6 0.28 0.6 0.1008   Passage 0.85 0.28 0.6 0.1428   Predator/Prey 0.15 0.12 0.6 0.0108 0.29 Withdrawal Water Quality 0.05 0.14 0.6 0.0042   Habitat 0.15 0.28 0.6 0.0252 0.03 Roads Water Quality 0.2 0.14 0.6 0.0168   Passage 0.11 0.28 0.6 0.0185 0.04 Agriculture Water Quality 0.2 0.14 0.6 0.0168   Habitat 0.15 0.28 0.6 0.0252 0.04 Acidity Water Quality 0.1 0.14 0.6 0.0084 0.01 Logging Water Quality 0.05 0.14 0.6 0.0042   Habitat 0.1 0.28 0.6 0.0168   Passage 0.02 0.28 0.6 0.0034   Predator/Prey 0.05 0.12 0.6 0.0036 0.03 Hatcheries Passage 0.02 0.28 0.6 0.0034   Disease 0.3 0.12 0.6 0.0216   Competition 0.3 0.06 0.6 0.0108 0.04 Exotics Predator/Prey 0.8 0.12 0.6 0.0576   Disease 0.05 0.12 0.6 0.0036   Competition 0.6 0.06 0.6 0.0216 0.08 Aquaculture Disease 0.65 0.12 0.6 0.0468   Competition 0.1 0.06 0.6 0.0036 0.05 Totals       0.6000 0.60

OCR for page 108
Atlantic Salmon in Maine TABLE 4-2 Proportional Impacts of Direct Sources of Mortality on Viable Wild Atlantic Salmon Populations Based on a Risk Assessment (refer to Figures 4-1 and 4-3) Component Level 1 Level 2 Level 3 Product (1x2x3) Impact (Sum) Fisheries Adult Mortality 0.6 0.9 0.2 0.1080 0.11 Incidental Adult Mortality 0.15 0.9 0.2 0.0270   Juvenile Mortality 0.33 0.1 0.2 0.0066 0.03 Research Adult Mortality 0.15 0.9 0.2 0.0270   Juvenile Mortality 0.33 0.1 0.2 0.0066 0.03 Hatcheries Adult Mortality 0.1 0.9 0.2 0.0180   Juvenile Mortality 0.34 0.1 0.2 0.0068 0.02 Totals       0.2000 0.20 that dams, through water quality, reduce the viability of salmon by a relative magnitude of 0.03 or 3%. Similar calculations for the impact of dams through habitat (0.1), passage (0.14), and predator-prey dynamics (0.01) result in a cumulative impact of 0.29. This suggests that 29% of the total impact on viable wild salmon populations by humans is through dams, and we can now map the role that dams have on disrupting the ecology of salmon. The tables summarize similar kinds of analyses for all the possible impact factors that the committee identified. The analytic procedure described above—multiplying fractions—leads to the appearance of greater precision than is intended. For example, a more appropriate characterization of the importance of dams as a threat is that they are the largest single factor but are responsible for less than half of human effects on salmon. Finally, the committee performed a very rough sensitivity analysis. How would the results change if a category changed relative size and if factors within a category changed relative sizes? These exercises led the committee to conclude with considerable confidence that the single largest human-caused factor affecting salmon mortality is obstruction to passage. At the same time, the committee concluded that obstruction to passage probably accounts for less than half of all human-caused mortality.

OCR for page 108
Atlantic Salmon in Maine Zoning. Move aquaculture sites away from the migratory path of adult salmon returning to DPS and other rivers and smolts leaving them. Bioconfinement. Sterilize farm fish to minimize mating success between escapees and wild fish. Tagging/weir. Mark all farm fish so escapees can be identified and removed at weirs where all upstream migrants are captured. Land farming. Move salmon farms onshore. Solid confinement. Enclose farm fish pens to prevent escapes. Status quo. Current default option. Remove aquaculture pens. Discontinue farming of Atlantic salmon in Maine. Table 4-4 provides the committee’s estimates of the likelihood that each strategy would be successful in meeting objectives 1 and 2 and the effects of factors that could influence the ability of each strategy to achieve the desired objective. The estimates of the probability of success were made TABLE 4-4 Strategic Options and Committee’s Estimates of Success Factors for Meeting Aquaculture Escapee Management Objectivesa   Strategies Success Factors 1 2 3 4 5 6 7 Permitting X (–) (–) X (–) X 0 Political acceptance X 0 0 X X X XX Socioeconomic effects X 0 0 (+) X 0 XX Technical feasibilityb 1.0 1.0 1.0 1.0 0.6 1.0 1.0 Survival impairment 0 0 X 0 0 X 0 Capital costs X (–) X XX XX (–) 0 Management costs 0/Xc (–) X X XX (–) 0 Legal liabilityd (–) X X (+) (+) XX (–) Probability of success Objective 1 (ecological/genetic) 0.9 0.9 0.99 0.95 0.9 0 1.0 Objective 2 (disease) 0.99 0.5 0.5 0.99 0.9 0.5 1.0 aStrategy 1: zoning; strategy 2: bioconfinement; strategy 3: tag/weir; strategy 4: land farming; strategy 5: solid confinement; strategy 6: status quo; strategy 7: remove aquaculture pens. 0 = no real problem or issue. X = significant problem or issue. (–) = minor problem or issue. (+) = improvement from status quo with respective to objectives. bProbability of success. cDepends on relocation site. dPotential for ESA, CWA, or other violation.

OCR for page 108
Atlantic Salmon in Maine as functions of factors such as permitting complexity and success, political acceptability, availability of suitable alternative sites, effects on jobs and local commerce, technical feasibility, impacts on survival of wild fish, capital and management costs including those for monitoring and effectiveness evaluation, and legal liability. Some conclusions can be drawn from Table 4-4. The first is that bioconfinement, tag/weir, and status quo are not very likely to achieve objective 2. Even if the standard for objective 2 were dropped or significantly lowered, each option carries potential legal liability regarding possible violations of Section 9 of the ESA and perhaps the Clean Water Act. They are therefore unlikely to be sufficient as single approaches for meeting the challenge posed by aquaculture releases. Among the strategies likely to be successful in meeting both objectives, high capital and management costs and low technical feasibility (0.6) work against solid confinement. Land farming entails higher economic costs than zonal relocation, but the greater availability of suitable sites, lower potential legal liability, and possible socioeconomic benefits argue in its favor. The need to find suitable estuarine or offshore sites for relocating pens where escapees would not threaten wild stocks in DPS and other rivers is a major consideration for relocation in an aquatic setting. Siting factors include an ice-free environment, protection from storms, adequate depth, flushing, ready access, and community acceptance. The impact of displacing an industry and employees from an area of existing operations will also influence its political acceptability. Eliminating aquaculture of Atlantic salmon in Maine altogether would clearly meet both objectives (except of course for any effects of salmon farms in Canada), but would also eliminate employment and economic benefit. Enhancing Habitat Availability A more complex example of decision analysis concerns improving access to habitat blocked by dams. Assume that a major goal for recovery efforts is to increase available spawning and juvenile rearing habitat on two rivers that empty into the Gulf of Maine and are separated by about 100 miles. On River A, a large dam located near the mouth blocks upstream passage of returning adults except in occasional years of unusually high stream flow. River A historically supported a substantial run of Atlantic salmon. Although a few fish have ventured upstream to the base of the dam in recent years, spawning is sporadic if it occurs at all because of the lack of suitable habitat below the dam. Salmon found in River A are not protected under the Endangered Species Act (ESA). On River B, where the Atlantic salmon stock is listed under the ESA as a DPS, two moderate sized dams impede access. The dam located further downstream has a

OCR for page 108
Atlantic Salmon in Maine marginally effective fish ladder but spawning habitat is poor between the dams because of the quiet water and lack of gravel for redds. The upstream dam is a complete barrier to further migration. On River B a 2-mile reach below the downstream dam is poor habitat for spawning because it is used for gravel mining, but it could be restored, especially if mining were to cease. All three dams provide water for irrigated agriculture and generate electric power during periods of adequate flow. The watershed surrounding River A is mostly forest subject to long-term harvest rotation. Mixed land use consisting of small forest plots, pasture crops, several widely separated small towns, low density rural housing with septic tanks, and cranberry farms occupy the watershed landscape along River B. Riparian corridors that protect the river from adverse impacts of human activity in the watershed only occur along 50% of the length of River B and its tributaries. Consequently the quality of habitat that could be made available above the dam on River A is apt to be greater than that on River B if the dams were breached. Restoration of riparian buffers and control of non-point-source pollution would be needed to maximize the habitat potential on River B. A well-defined objective for these rivers might be to increase available spawning and rearing habitat by a specified number of habitat units over a certain period of time. A unit of habitat is 100 square meters (see Table 1-1 for habitat units in Maine Rivers). The general term “salmon habitat” refers to riffles and runs. A second success metric could stipulate that some level of spawning by Atlantic salmon should be attained on the newly available habitat. The objective could further distinguish between the relative values of new habitat units on DPS rivers vs. non-DPS rivers. Those expected outcomes that meet the threshold can be further sorted by other measures such as costs, while those that do not reach the standard can be ignored (Peterman and Peters 1998). It is also possible to create an objective that calls for maximizing available habitat independently of time or costs, but this is less realistic in terms of public agency budgetary policy. The process starts with developing an influence diagram (Figure 4-6) to show important variables and relationships affecting expected outcomes, in this case new habitat units. Other tools, such as a decision hierarchy, ensure that the focus of the decision will be on strategic elements and not directed toward aspects of the problem that are givens or can be resolved later as tactical details (Chevron Strategic Decisions Group, unpublished material, 1991). The main advantage of the influence diagram is to understand the basic structure of a problem (Clemen 1991) and to be able to communicate the essential elements to stakeholders. The next step is to identify workable alternative strategies that represent choices for action by the decision maker. For this exercise, dam removal should be considered as a possible option in order to attain the

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-6 Influence diagram for factors affecting strategic choices. biological objective. Although there can be many obstacles for implementing such a strategy (Heinz Center 2002), the recent success on the Kennebec suggests that it could be a viable approach. A strategy diagram shows the range of choices for the series of decisions needed to implement each strategic theme (Figure 4-7). For example, supporting decisions for the habitat augmentation strategies might include whether supplemental stocking is needed, and if so, what life stage should be used. Questions of the preferred sequence for dam removal on River B and whether further research would improve the chances of success might also need to be explored further. The alternative of removing all three dams was not included because we judged the projected costs exceed available funding within the specified time period. A decision table that represents the various uncertainties shown in the influence diagram can then be used to model the different strategies

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-7 Strategy table for habitat improvement strategies for Maine salmon.

OCR for page 108
Atlantic Salmon in Maine TABLE 4-5 Decision Table for Habitat Improvement Strategies Alternative Technical Feasibility Regulatory Acceptance Legal Liability A. Remove dam on river A High Low to Moderate Low (non-DPS river) B. Remove both dams on river B High Low to Moderate Moderate (potential for take during demolition) C. Remove upstream dam on river A; improve access on B High to Moderate Moderate Moderate D. Improve passage on all three dams Low to Moderate High Low E. Improve habitat below downstream dam on river B Low High High (potential for jeopardy during permit renewal if access not improved and possible lawsuits over incidental take) (Table 4-5). Although it is possible to calculate value measures for each strategy if the table is set up deterministically, qualitative estimates provide an informative summary of how key factors are likely to affect outcomes. In this case, the high number of habitat units gained by removing the dam on River A might be tempered by the effect of high costs and high economic impact on the political acceptability of the strategy and the longer time to implement it. The next step is to understand the critical uncertainties that need to be modeled in the subsequent phases. These uncertain states of nature (Peterman and Peters 1998) are considered by setting a range of values for each uncertainty parameter, whether costs, habitat units, or the likelihood of success (or failure) during different stages of strategy implementation. The influence diagram and the strategy table provide key input for this step of the analysis. In the example, we can identify three main uncertainties that will directly affect success of the various strategies. These are technical feasibility of the alternatives, regulatory acceptance, and the prospect that the new habitat will become occupied for spawning and juvenile rearing by Atlantic salmon. Other factors such as the potential for adverse legal action and political support could also affect outcomes, but

OCR for page 108
Atlantic Salmon in Maine Implementation Cost ($ in millions) Economic Impact Time to Implement (Years) Political Acceptance Habitat Units Gained (Base Value) 12 High 5–10 Low-Moderate (high cost and high economic impact, but precedent on Kennebec River High 8 Moderate- High 3–5 Moderate Moderate to high 4 Moderate 2–4 Moderate Moderate 4 Low 1–4 High Low 2 Low 2–3 Moderate (could be pressure from possible legal liability) Low their influence may be unpredictable or less direct through their impact on costs, timing or permitting success. For each uncertainty variable, it is necessary to assign a probabilistic estimate to the different states we choose to analyze. In most cases, sufficient data are not available to develop precise probability estimates with a high degree of confidence. For natural resource problems, the decision analysis process often uses subjective estimates, usually developed by a cross-section of stakeholders. A variety of sources can be used to inform these subjective evaluations—performance history, experimental results, trend analysis, extrapolation, correlations to other variables, scenario modeling, and so forth. As a practical matter, they are often based on personal experience and professional judgment of the team conducting the decision analysis. Even though hard data are often lacking, decision analysis allows decision makers to consider a range of values to gain a better overall picture of the effect of different uncertainty variables. One of the strengths of decision analysis is that making quantified judgments about uncertainties promotes clear communication (Clemen 1991) and helps to resolve disagreements that can result from differences in belief systems, experi-

OCR for page 108
Atlantic Salmon in Maine ence, and biases (Stewart 2000). For example, one of the problems inherent in setting public policy objectives based on imperfect information is to establish whether a decision should favor possible false-positive outcomes or should lean to false negatives. Because these two results are mutually exclusive (Stewart 2000), favoring one or the other usually means having to make trade-offs, e.g., conservation value vs. short-term economic impact, that shift depending on which type of error is more acceptable. Decision analysis can also be used to quantify the value of a trade-off that attends the question of whether to spend more time and money gathering additional information in order to reduce uncertainty about outcomes (Clemen 1991). After assigning probabilities, a decision tree (Figure 4-8) displays the strategic options, the uncertainty variables, their probability of occurrence, and the outcomes in terms of specified value measures (Peterman and Peters 1998). The estimated probabilities in the example are assigned to illustrate how the process works and are intended to reflect how the hypothetical facts might drive probability estimates. For example, the chance of successful colonization of new habitat on River A would be lower than that on River B because salmon only occasionally occupy the reach below the dam to be removed. However, they consistently appear below the lower dam on River B even though they do not successfully reproduce in the gravel quarry. The decision tree provides a convenient way of ranking the alternatives according to their expected or “net” value. The technique is to weigh the base value of habitat units for each option by the probability factor at each branch of the tree corresponding to the three uncertainty variables. Some management actions lead to more than one outcome that are then summed to give the total net value (NV) for that strategy. In the example, the strategy to improve fish passage on all three dams gives a net value of According to the decision tree, Strategy 1, removal of the large dam on River A, would create the greatest net value of new habitat units (960 HUs). This is followed by Strategy 2, which calls for removing both dams on River B (800 HUs). This ranking would be reversed, however, if the new habitat units on River A were discounted by 25% because it is not a DPS river, which illustrates the role of perception and relative quality of different options in establishing and choosing among preferences. River B would further benefit if instream habitat improvements were undertaken to maximize the value gained by breaching the dams. This option could

OCR for page 108
Atlantic Salmon in Maine FIGURE 4-8 Decision tree for habitat improvement strategies.

OCR for page 108
Atlantic Salmon in Maine be the subject of a separate decision analysis designed to evaluate its merits in terms of extra costs, time to implement, and likelihood of gaining the landowner support needed for success. If the objective is to create the most new HUs at the lowest cost, Strategy E, buying out the gravel mining rights and improving habitat below the downstream dam on River B, is most cost effective. Even though the possibility of failure for Strategy E is 30%, it would still represent a preferred approach if cost is a key success driver. If the exact costs to carry out the strategies are not known, they can be incorporated as an uncertainty factor in the analysis. A common approach is to assign probabilities to a high (90%), medium (50%), and low (10%) range of costs for each alternative. Reviewing the outcomes from another angle, we conclude that Strategies D and E also warrant strong consideration if the objective calls for maximizing occupation of new habitat at the earliest time. Another way to scale the value outcome is to combine the new HUs with the cumulative time that they are available, giving a metric called habitat service years. In this case, a decision maker would include time to implement and time for the new habitat to become occupied as uncertainty variables in the decision tree. The strategy yielding the most habitat service years within a specified period following the decision would rank highest. The decision tree in Figure 4-8 is based on primary factors influencing all the strategies being ranked. These are determined from an influence diagram like Figure 4-6, and are the variables that must be evaluated to distinguish among the various options. It is possible that factors such as conducting further research and implementing (or disbanding) supplemental stocking programs could improve the success of individual options. But the committee did not include them as part of the primary decision tree, because each would have its own set of variables to evaluate (e.g., life stage, number of individuals, seasonality, and stocking location for supplemental stocking). Thus, they are considered secondary factors in Figure 4-7 to explore maximizing habitat utilization (the primary decision) in Figure 4-8 and Table 4-5. If decisions about whether to implement a stocking plan or to conduct research in conjunction with the strategy to increase habitat availability might differentially affect strategic outcomes, they can be included in the decision tree as decision nodes with yes/no branches. After the alternative strategies are ranked according to a decision tree, a variety of sensitivity analysis techniques can be used to answer the question, “what matters in this decision?” (Clemen 1991). The primary purpose of this more introspective look is to ensure that the analysis is focusing on the right question to satisfy the original objective. The idea is to avoid making Type III errors, as opposed to the familiar Type I and

OCR for page 108
Atlantic Salmon in Maine Type II errors in statistics (Clemen 1991). Type III errors give rise to the wrong question being asked, given the available information. In the above example, decision makers could be short-sighted if they decided to restrict water withdrawals in an effort to improve water quality above dams that limit access unless plans to improve habitat occupation were also in the works. These exercises are only illustrative. People with in-depth knowledge of and experience with physical, biological, social, and political environments need to undertake these risk-management decision processes. In addition, people who must live with the consequences of these management decisions should be involved, otherwise the decisions will be difficult to implement. Conclusions and Recommendations This discussion has attempted to show how decision analysis could be a helpful tool in sorting through the myriad choices of potential recovery strategies for Atlantic salmon. The approach could be used to understand the value of gaining additional information through baseline assessment, research, and monitoring. The potential value of missing information would become apparent in considering specific decision choices. Another application would be to clarify the role of different stocking strategies. The value of expanding fisheries on non-DPS rivers could be evaluated against the chance of attaining recovery goals on listed rivers. The issue of number, location, and controls on aquaculture facilities also needs to be examined. Habitat restoration measures designed to mitigate the adverse effects of erosion and sedimentation, reduced in-stream flow and elevated temperatures, and pollutant loading should be investigated for their potential contribution to recovery. The committee recommends that recovery planning efforts for Atlantic salmon in Maine rivers employ structured, systematic, strategically focused decision making processes for developing conservation and recovery objectives and analyzing the optional approaches for achieving them. All stakeholders need to be involved in this process to ensure its validity and acceptability. The committee further recommends that recovery planners engage the services of an expert in the field of strategic decision analysis, especially someone experienced in its application for natural resource problems, to advise them in their endeavors. These activities will need to be repeated when changes in environmental conditions or human interventions change conditions relevant to the analysis. The committee also recommends research on the socioeconomic effects to changes in aquaculture (discussed in Chapter 5).