Adaptively Managing Protected Areas for Climate Change Tony Prato, Professor of Ecological Economics, and Co-Director Center for Agricultural, Resource, and Environmental Systems University of Missouri-Columbia “…protected areas will themselves need to be changed and adapted if they are to meet the challenges posed by global warming” World Wildlife Fund 2003 Background There is considerable uncertainty regarding the biophysical impacts of climate change, social and ecological responses to those impacts, and the effectiveness of alternative strategies in reducing vulnerability of protected areas to climate change. Protected area managers could respond to this situation by doing nothing or by developing and implementing adaptation strategies aimed at reducing the social and ecological vulnerability of protected areas to climate change. Objective Describe a modeling system that allows protected area managers to reduce the social and ecological vulnerability of protected areas to climate change. Adaptively Managing Protected Areas for Climate Change (AMPAC) modeling system. AMPAC incorporates a conceptual framework for evaluating and selecting adaptation strategies to increase the biophysical and social resilience of protected areas to climate change. AMPAC is designed for use in individual protected area ecosystems. However, the framework is suitable for managing other natural resource-based lands for the impacts of climate change. Explanation of Models The five models comprising AMPAC are described in the context of the ecosystem in which Glacier National Park is located. The emphasis is on the models that contribute most directly to the unique aspects of the AMPAC modeling system. Climate Change Model Climate change model: Used to develop plausible climate scenarios for a protected ecosystem. The Climate Change Model for Glacier National Park would specify plausible climate scenarios for the park ecosystem in terms of changes in annual and seasonal patterns of precipitation, temperature, and other factors. Evaluating Adaptation Strategies Effects of adaptation strategies designed to reduce adverse impacts of climate scenarios on indicator variables can be evaluated using: – The Ecosystem Response Model, if is suitable for this purpose – The Adaptive Management Model, or – A combination of the two models Ecosystem Response Model Climate response model: Simulates likely biophysical and social responses to climate scenarios in the absence of adaptation strategies, and modifies those responses for the potential effects of alternative adaptation strategies. The Ecosystem Response Model simulates likely biophysical and social responses to climate scenarios using models, such as the Regional Hydro-Ecological Simulation System (RHESSys), and social surveys. Simulated responses of indicator variables to climate scenarios in the absence of adaptation strategies are used construct original probability distributions for indicator variables. Effects of adaptation strategies on indicator variables are modeled using a Delphi method and the results used to modify the original probability distributions. To illustrate the Ecosystem Response Model, suppose a climate scenario for Glacier National Park indicates higher temperatures and associated decreases in plant and animal species favored by grizzly bear. Other things equal, these changes are likely to reduce grizzly bear population in the park ecosystem, which could adversely effect biodiversity and visitor enjoyment of the park. Suppose the original probability distribution for grizzly bear population for a particular climate scenario is represented by the triangular probability distribution in the top diagram: Pr Original probability distribution Pr Modified probability distribution Bear population Consider an adaptation strategy aimed at grizzly bear that involves decommissioning selected roads in Flathead National Forest, east of the park. Research shows a negative relationship between road density and bear mortality. Hence, road decommissioning is expected to increase grizzly bear population (other things equal), as illustrated in the bottom diagram. Pr Original probability distribution Pr Modified probability distribution Bear population This adaptation strategy could counteract some but probably not all adverse impacts of climate change on grizzly bear population. A similar procedure is used to construct modified probability distributions for all indicator variables. Adaptive Management Model Adaptive management model: Experimentally tests hypotheses about how indicator variables are likely to respond to adaptation strategies (active adaptive management). Model results are used to construct the modified probability distributions for indicator variables. The Adaptive Management Model is illustrated for testing hypotheses about the effects of lower road density on grizzly bear population. Two hypotheses are tested: – H1: bear mortality decreases as road density decreases vs. – H2: bear mortality increases or remains the same as road density decreases. Hypotheses are sequentially tested using Bayesian statistical methods. H1 and H2 are illustrated in terms of the triangular probability distributions: The ratio is estimated from data collected in the adaptive management experiments. H1 implies that bear mortality decreases as road density decreases (R>0). Pr H1 0 Pr H2 implies that bear mortality increases as road density decreases (R<0). H2 R = ∆ bear mortality/∆ road density If results from the first round of experiments indicates H1 should not be rejected, then bear mortality after road density is decreased equals bear mortality before road density is decreased times one minus the ratio selected from the triangular distribution for H1. Since ratios selected from the H1 distribution are likely to exceed zero, bear mortality is expected to decrease as road density decreases. In this case, reducing road density is judged to be an effective adaptation strategy for maintaining bear population, for the initial round of experiments. If H1 is rejected in favor of H2, then reducing road density would not be considered an effective adaptation strategy for maintaining bear population. Similar adaptive management experiments can be done for other indicator variables. Hypotheses should be periodically retested to determine how the effectiveness of a particular strategy changes over time. Details on this method are given in: Prato, T. 2005. Bayesian adaptive management of ecosystems. Ecological Modelling 183: 147-156. Ecosystem Vulnerability Model Ecosystem vulnerability model: Assesses extent to which alternative adaptation strategies reduce vulnerability of the park ecosystem to climate change by synthesizing the outputs of the previous three models. Synthesis Indicator variables and their modified probability distributions Indicator elements Ecosystem conditions Park’s vulnerability to climate change Example for Glacier Park Four indicator variables: floral diversity, population of grizzly bear, water quality, and opportunities for viewing glaciers. Modified probability distributions for all indicator variables under each climate scenario are constructed as explained earlier. Indicator elements are statements about indicator variables, e.g., grizzly bear population is moderately higher than the target recovery population. Four ecosystem conditions: R1, R2, R3, and R4. – R1: heavy loss in floral diversity, grizzly bear absent from area, very high water quality degradation, and no opportunities for viewing glaciers (all glaciers melted) – R2: moderate loss in floral diversity, grizzly bear population moderately below the target recovery population, moderately high water quality degradation, and moderately limited opportunities for viewing glaciers. – R3: good floral diversity, grizzly bear population moderately higher than the target recovery population, very little water quality degradation, and moderately good opportunities for viewing glaciers; and – R4: excellent floral diversity, grizzly bear population significantly higher than the target recovery population, no water quality degradation, and excellent opportunities for viewing glaciers. Four ecosystem states of vulnerability to climate change: M1 = very low; M2 = moderately low; M3 = moderately high; and M4 = very high. M1 and M2 are desirable ecosystem states (they imply resilience to climate change), and M3 and M4 are undesirable ecosystem states. Subjective prior probabilities of ecosystem states are p(M1), p(M2), p(M3), and p(M4), which sum to one. Evaluation of Indicator Elements – Grizzly bear is absent from the park if the probability the population is between zero and 25 percent of the target recovery population is greater than or equal to 0.75; – Grizzly bear population is moderately below the target recovery population if the probability the population is between 75 percent and 100 percent of the target recovery population is greater than or equal to 0.75; – Grizzly bear population is moderately higher than the target recovery population if the probability the population is between 100 percent and 125 percent of the target recovery population is greater than or equal to 0.75; and – Grizzly bear population is significantly higher than the target recovery population if the probability the population is between 100 percent and 125 percent is greater than or equal to 0.95. Pr Modified probability distribution a c Bear population b P* ∫ implies grizzly bear population is f(x)dx ≥ 0.75 moderately below the target recovery population (P*) .75P* f(x) = [2(x-a)/(b-a)(c-a)] for a≤x≤c, and = [2(b-x)/(b-a)(b-c)] for c≤x≤b, Inferring Ecosystem Conditions and States Inferring ecosystem conditions: Determine the ecosystem condition from the combination of indicator elements simulated for each adaptation strategy and climate scenario, and the evaluation of those elements using modified probability distributions of indicator elements. Inferring ecosystem states: Use Bayes theorem to calculate posterior probabilities for all ecosystem states conditional on that ecosystem condition, and determine the most likely ecosystem state, i.e., the one with the highest posterior probability. Errors in Inferring Ecosystem States Two mutually exclusive errors are possible: – 1. Decide ecosystem is vulnerable to climate change when it is not. Causes implementation of an adaptation strategy when it is not necessary-Type I error. – 2. Decide ecosystem is not vulnerable to climate change when, in fact, it is. Causes no implementation of an adaptation strategy when one is needed. Runs the risk that the ecosystem remains in or changes to a less desirable state-Type II error. Bayes theorem can be used to minimize inferential errors. Optimal Strategy Model Optimal strategy model: Determines the best adaptation strategy for each climate scenario by applying the minimax regret criterion to the posterior probabilities for ecosystem states and the maximum expected costs for alternative strategies. Minimax regret criterion selects the adaptation strategy for each climate scenario that minimizes the maximum expected cost unless the social cost of that strategy is unacceptably high. Example (for a particular climate scenario) E(s1) = s11p(M1|R3) + s12p(M2|R3) + s13p(M3|R3) + s14p(M4|R3) For example, T1 is the best adaptation strategy for a particular climate scenario if E(a1) + E(s1) < E(a2) + E(s2) and E(a1) + E(s1) < E(a3) + E(s3). If scientists can assign subjective probabilities to climate scenarios, then the minimax regret criterion can be used to identify the best adaptation strategy across all climate scenarios. Conclusions AMPAC has high informational and scientific requirements that are not likely to be satisfied in most protected areas at this time. This limitation might be diminishing due to advances in environmental monitoring and assessment, climate research, and adoption of science-based management, e.g., National Park Service’s Inventorying and Monitoring Program and Natural Resource Challenge, USGS’ global change research program, and the Cooperative Ecosystem Studies Units. Agencies that manage protected areas may not consider it worthwhile and/or may lack the financial and technical resources to implement AMPAC. This concern can be addressed by initiating a pilot program that implements AMPAC for a sample of protected areas containing a range of natural resource and environmental conditions, human uses and values, and availabilities of scientific knowledge and technical expertise. Information obtained from such a pilot program can be used to assess the expected benefits and expected costs of using AMPAC in protected areas, and identify the conditions under which its use is economically feasible. Questions? Questions?