Portal de Periódicos da CAPES

Setting conservation management thresholds using a novel participatory modeling approach

2015; Wiley; Volume: 29; Issue: 5 Linguagem: Inglês

10.1111/cobi.12544

1523-1739

Prue Addison, Kelly de Bie, Libby Rumpff,

Coral and Marine Ecosystems Studies

We devised a participatory modeling approach for setting management thresholds that show when management intervention is required to address undesirable ecosystem changes. This approach was designed to be used when management thresholds: must be set for environmental indicators in the face of multiple competing objectives; need to incorporate scientific understanding and value judgments; and will be set by participants with limited modeling experience. We applied our approach to a case study where management thresholds were set for a mat-forming brown alga, Hormosira banksii, in a protected area management context. Participants, including management staff and scientists, were involved in a workshop to test the approach, and set management thresholds to address the threat of trampling by visitors to an intertidal rocky reef. The approach involved trading off the environmental objective, to maintain the condition of intertidal reef communities, with social and economic objectives to ensure management intervention was cost-effective. Ecological scenarios, developed using scenario planning, were a key feature that provided the foundation for where to set management thresholds. The scenarios developed represented declines in percent cover of H. banksii that may occur under increased threatening processes. Participants defined 4 discrete management alternatives to address the threat of trampling and estimated the effect of these alternatives on the objectives under each ecological scenario. A weighted additive model was used to aggregate participants' consequence estimates. Model outputs (decision scores) clearly expressed uncertainty, which can be considered by decision makers and used to inform where to set management thresholds. This approach encourages a proactive form of conservation, where management thresholds and associated actions are defined a priori for ecological indicators, rather than reacting to unexpected ecosystem changes in the future. Establecimiento de Umbrales de Manejo de Conservación con el Uso de una Estrategia Novedosa de Modelación Participativa Diseñamos una estrategia de modelación participativa para establecer umbrales de manejo que muestren cuando se requiere de intervención en el manejo para tratar con cambios indeseables en los ecosistemas. Esta estrategia se diseñó para ser usada cuando los umbrales de manejo: deban estar fijados para los indicadores ambientales de frente a múltiples objetivos en competencia; necesiten incorporar al entendimiento científico y a los juicios de valor; sean fijados por participantes con experiencia limitada en la modelación. Aplicamos nuestra estrategia a un estudio de caso en el que los umbrales de manejo estaban fijados para una alga café formadora de tapetes, Hormosira banksii, en un contexto de manejo de área protegida. Los participantes, incluidos el personal de manejo y los científicos, estuvieron involucrados en un taller para probar la estrategia y establecieron los umbrales de manejo para tratar la amenaza del pisoteo por los visitantes para un arrecife rocoso inter-mareal. La estrategia involucró el intercambio del objetivo ambiental (mantener las condiciones de las comunidades del arrecife inter-mareal) por objetivos sociales y económicos para asegurar que la intervención del manejo fuera rentable. Los escenarios ecológicos, desarrollados usando la planeación de escenarios, fueron una característica clave que proporcionó la base para ubicar los umbrales de manejo. Los escenarios desarrollados representaron declinaciones en el porcentaje de cobertura de H. banksii que podrían ocurrir bajo procesos crecientes de amenaza. Los participantes definieron cuatro alternativas discretas de manejo para tratar la amenaza del pisoteo y estimaron el efecto de estas alternativas sobre los objetivos bajo cada escenario ecológico. Se usó un modelo aditivo ponderado para añadir los estimados de consecuencia de los participantes. Los resultados del modelo (puntajes de decisión) expresaron claramente la incertidumbre, lo que puede ser considerado por quienes toman las decisiones y utilizado para informar dónde ubicar los umbrales de manejo. Esta estrategia fomenta una forma proactiva de conservación, en la cual los umbrales de manejo y las acciones asociadas son definidas a priori por los indicadores ecológicos, en lugar de reaccionar a futuros cambios inesperados en los ecosistem Conservation management agencies aspire to safeguard species, ecosystems and habitats from undesirable change. State-dependent management is one useful approach, where monitoring data are used to assess whether a protected system is shifting into an undesirable state and management intervention is required to address this change (Nichols & Williams 2006; Lyons et al. 2008). A proactive application of state-dependent management involves the use of management thresholds, which represent when management intervention is required to address undesirable ecosystem changes. Management thresholds have had widespread application in natural resource management, including species harvest management (Nichols et al. 2007) and water quality management (ANZECC 2000). However, there are fewer applications of management thresholds to enable state-dependent management in conservation (Groffman et al. 2006; Martin et al. 2009). Consequently, there have been instances where programs have monitored species of conservation value until they became extinct without triggering adequate or timely management intervention (Lindenmayer et al. 2013). The discussion around thresholds in the conservation literature has been dominated by ecological thresholds, representing a point of rapid or irreversible ecosystem change (Groffman et al. 2006). The inherent complexity of ecological systems means that ecological thresholds can be difficult to identify or may not exist in some systems. The pursuit of ecological threshold identification can detract from setting management thresholds that represent where timely and cost-effective management intervention is required to prevent undesirable ecosystem changes (Bennetts et al. 2007; Martin et al. 2009). There are limited published examples of management thresholds that enable state-dependent management in conservation. Identifying thresholds of potential concern is one approach that builds on the level of acceptable natural variability of ecological attributes. In South Africa, such thresholds have been used to identify when management intervention should occur and enable the adaptive management of national parks (Biggs & Rogers 2003). An alternative approach is the use of statistically derived management thresholds that promote careful consideration of statistical issues associated with interpreting monitoring data and that facilitate the proactive management of environmental systems (Field et al. 2004; Morrison 2008). These approaches focus on the environmental dimension of setting management thresholds for environmental indicators in isolation of competing socio-economic factors. They are useful in contexts where environmental objectives are the sole driver of conservation efforts; however, many complex conservation decisions must be made in the face of competing environmental, social, and economic objectives (Game et al. 2013). Structured decision making can assist with setting management thresholds for multi-objective conservation decisions. It involves a series of steps that incorporates both scientific knowledge and values into decision making and acknowledges that decision making is not a value-free process (Failing et al. 2007). Structured decision-making also promotes the involvement of decision makers, stakeholders, and experts (collectively participants) in decision making (Gregory et al. 2012). The 2 published examples of use of structured decision making to develop management thresholds involved the application of stochastic dynamic programming to model ecosystem behavior and set optimal management thresholds for environmental indicators in the face of competing objectives (Martin et al. 2009, 2011). This approach is particularly effective when quantitative modeling expertise is available and participants with some modeling experience are involved in setting management thresholds. However, such modeling approaches can appear unintuitive or illogical to those without quantitative modeling expertise, and in these situations more transparent and collaborative techniques may be more appropriate (Addison et al. 2013). We sought to introduce a novel modeling approach in response to the need for setting management thresholds in the face of competing objectives, with a focus on providing accessible tools to engage participants with limited modeling experience. A single approach to setting management thresholds will not be suitable in all situations because conservation decisions often involve different circumstances that require different modeling approaches (Addison et al. 2013). We propose this participatory modeling approach as one in a toolbox of available approaches to assist with setting management thresholds. Our approach is particularly relevant in situations when management thresholds must be set for environmental indicators in the face of multiple competing objectives; need to incorporate scientific understanding and value judgments about the management of the ecosystem; or must be set by participants with limited modeling experience. To illustrate this approach, we used a multi-objective case study aimed at setting management thresholds in a protected area management context. Management objectives should be specific, with quantifiable outcomes that reflect participants' values, and can reflect environmental, social, and economic values that are considered directly relevant to the decision context (Runge & Walshe 2014). Management objectives can be defined using value-focused thinking, which encourages participants to consider what they fundamentally care about (Keeney 1992). An objectives hierarchy can be used to facilitate the identification of fundamental objectives and can further help participants identify performance measures for each management objective (Clemen 1996). Performance measures determine whether management objectives are being achieved and include an indicator, unit of measure, and preferred direction (e.g., maximize or minimize). During this stage the key ecological indicator or indicators to be monitored and assessed using management thresholds must be defined. In the absence of targeted epirical evidence, expert judgement can be used to evaluate the consequences of management alternatives on each management objective. Participants provide quantitative consequence estimates that reflect their conceptual understanding, supported by empirical evidence when available, of the cause-and-effect relationships under each ecological scenario. During this stage the time frame of the estimated consequences of the management alternatives on objectives must be defined (i.e., x years after the implementation of management alternatives). To capture the uncertainty in participants' judgments, optimistic and pessimistic consequence estimates should be elicited together with their best estimates and participants should articulate the degree of confidence that the nominated interval encloses the truth (e.g., 4-point elicitation method) (Speirs-Bridge et al. 2010). When multiple competing objectives are encountered in decision making, a formal trade-off step is required. Techniques used in multi-attribute utility theory are useful here because these are designed to assist with evaluating complex multi-objective decisions (von Winterfeldt & Edwards 1986; Belton & Stewart 2002). To trade-off competing objectives, decision makers must weigh the relative importance of each management objective. Weights can be elicited using a variety of techniques (e.g., direct rating, swing weights, cross-attribute indifference) (von Winterfeldt & Edwards 1986). A widely advocated technique is the swing weight method because this is considered very effective and simple to implement when working with participants (von Winterfeldt & Edwards 1986). The swing weight method involves showing participants the full range of consequence estimates for each objective (minimum pessimistic consequence estimate versus maximum optimistic consequence estimate) and asking participants to assign weights to each objective. Weights reflect the order of preference: a weight of 100 is given to the most valued objective and a weight of zero is assigned to objectives that have no perceived value. The weighted additive model outputs (decision scores) are used to determine the preferred management alternative under each ecological scenario. Following the concept of decision making under uncertainty (Chankong & Yacov 1983), decision makers can interpret decision score plots for each ecological scenario based on 1 of 3 rules: best guess, pessimistic, or optimistic. The best-guess rule involves selecting the best performing management alternative under each scenario based on the highest nominal decision score (ignores the uncertainty surrounding the nominal decision score). The pessimistic rule involves selecting the best performing management alternatives under each scenario based on the lower bound decision score (assumes the worst payoff will eventuate for a management alternative). The optimistic rule involves selecting the best performing management alternatives under each scenario based on the upper bound decision score (assumes the best payoff will eventuate for a management alternative). The preferred management alternative selected under each scenario informs where to set the management threshold implementation range or ranges. Management threshold implementation ranges illustrate where the most effective management action should be implemented, and importantly where a change in management intervention is required as the condition of the ecological indicator declines. Long-term monitoring data should be regularly evaluated against management thresholds. If a management threshold is breached, the associated management action should be implemented. If a management threshold has not been breached, monitoring should continue under the preferred management. The learn-and-review feedback loops encourage the recurrent nature of state-dependent management to assist with clarifying uncertainty associated with model parameters, evaluating the true effectiveness of management interventions, and revising the decision context where necessary (Fig. 1). This participatory modeling approach was trialed using a case study involving the management of an intertidal reef in a marine national park (MNP) in Victoria, Australia. The management agency responsible for managing Victorian MNPs intends to conduct state-dependent management of ongoing threats to intertidal rocky reefs. Eleven participants, including management staff (decision makers and rangers) and intertidal marine ecologists were involved in a 1-day workshop to test the approach. Workshop participants were identified using snowball sampling (initial candidates are asked to recommend additional candidates from their professional network and participants are selected from this broader pool of candidates [Patton 2002]). Participants were selected if their expertise was directly relevant to the decision context (i.e., natural resource management, visitor management, or intertidal reef ecology). This research was approved by the Human Ethics Committee of the University of Melbourne (HREC Project Number: 1237358.1). Conceptual models were constructed to represent the ecosystem dynamics of intertidal rocky reefs and key threatening processes (step 1) (Pocklington et al. 2012). A mat-forming brown alga, Hormosira banksii, was identified as a key ecological indicator that represented the overarching environmental objective of Victoria's MNPs "to maintain biodiversity." H. banksii provides important habitat for a diverse array of intertidal species and is particularly susceptible to trampling by visitors (a key threatening process) (Povey & Keough 1991; Keough & Quinn 1998). Participants engaged in a facilitated discussion to collectively define the temporal and spatial scale for management and management constraints (step 1). An objectives hierarchy was used to elicit environmental, social, and economic objectives and performance measures (step 2). Workshop participants defined 4 discrete management alternatives that could be implemented within the management time horizon. Ecological scenarios were developed for H. banksii based on the principles of scenario planning (step 3). The ecological scenarios developed represent the current state of H. banksii cover and 3 plausible future states of reduced cover of H. banksii cover due to increased threatening processes. The scenarios were developed prior to the workshop and were informed by long-term monitoring data, experimental research, and decision makers' expert judgment about undesirable condition of H. banksii on intertidal reefs. We elicited participants' expert judgment on the effectiveness of management alternatives on all management objectives under each ecological scenario (step 4). Participants worked in groups to provide best guess, optimistic, and pessimistic consequence estimates, which were based on the assumption that management alternatives had been in place for 3 years. To provide these estimates, participants drew on their own expertise, scientific publications, and management agency reports. Because participants were working in groups and were time limited, they agreed a priori to provide optimistic and pessimistic consequence estimates to represent 90% credible bounds (reflecting a 3-point, rather than 4-point elicitation approach [Speirs-Bridge et al. 2010]). Two decision makers responsible for intertidal reef management provided their combined value judgments to weight objectives using the swing weighting technique, where the assigned weights represented the management agency's organizational priorities. A weighted additive model combined the consequence estimates and objective weights (Eqs. 1 and 2), which generated decision scores for each management alternative under each ecological scenario. The pessimistic rule was used to interpret decision scores and select the best performing management alternatives under each ecological scenario (step 5). This provided the basis for setting a management threshold implementation range, which guided the decision on where to set the management threshold to manage the threat of trampling on H. banksii. At the completion of the workshop all participants were asked to fill out a questionnaire consisting of 2 open ended questions that sought feedback on what aspects of the workshop participants thought did and did not work well. Seven management objectives were defined; however, not all these objectives were independent. To avoid double counting, a subset of the original objectives was identified: maximize rocky intertidal reef communities, minimize resources spent on management, maximize visitor satisfaction, and maximize visitor numbers to the intertidal reef (Table 1). The fundamental objective to maximize rocky intertidal reef communities was represented by the proxy performance measure, H. banksii percent cover, and all other performance measures were direct measures of fundamental objectives. Consequence estimates and objective weights were combined using the weighted additive model (Eqs. 1 and 2) to calculate decision scores for each management alternative under each ecological scenario (Fig. 2). The decision scores reflect the substantial uncertainty of participants' consequence estimates for some management alternatives (Table 3; Fig. 2). Following the pessimistic rule, a decision maker would select the following best performing management alternatives under each ecological scenario: status quo under the 70% cover scenario; medium protection under the 42% and 30% cover scenarios; and status quo under the 15% cover scenario. Two exploratory checks were conducted of the model outputs interpreted using the pessimistic rule (Supporting Information): a breakdown of decision scores and one-dimensional sensitivity analysis. These checks illustrated the variable influence of most objectives (except for visitor numbers) on the final decision scores under the ecological scenarios (Supporting Information). The model outputs were sensitive to changing weights for 2 objectives (rocky intertidal reef communities and resources [Supporting Information]). These exploratory checks can prompt re-defining or re-estimating model parameters in steps 2 and 4; however, this was not deemed necessary in this case study. Following management threshold selection, future monitoring data should be regularly assessed against the 42% cover medium protection management threshold. If the management threshold is breached, then the medium protection management approach should be implemented. Setting management thresholds remains a challenging task in conservation (Bennetts et al. 2007; Lindenmayer et al. 2013). This novel participatory modeling approach provides an accessible and effective method to set conservation management thresholds. It allows for competing objectives, the incorporation of scientific understanding, and value judgments about the management of the ecosystem and does not require participants to have in-depth quantitative modeling skills. This approach assisted in setting management thresholds for H. banksii, a key ecological indicator, so that the ongoing threat of trampling on the intertidal reef could be managed. The multiple competing objectives of this case study (Table 1) reflect increasingly considered environmental, social, and economic components of conservation decisions (e.g., Martin et al. 2011; Runge 2011). The novelty of this approach lies in the unique combination of techniques used to set management thresholds. The management threshold determined through this approach encapsulated empirical evidence, participants' scientific knowledge, and value judgments. Empirical evidence includes research or monitoring results, and participants' scientific knowledge represents their interpretation of cause-and-effect relationships within ecosystems, whereas values represent individual or organizational priorities (Keeney 1992; Failing et al. 2007). These sources of information are valued by conservation decision makers (Cook et al. 2012) and highlight that conservation decision making is not a value-free process (Failing et al. 2007). The ecological scenarios developed in step 3 are a key feature of this approach; they provide the foundation for where to set management thresholds. In step 4, participants estimate the consequences of management alternatives on management objectives under each ecological scenario. In step 5, decision makers select the preferred management alternative under each ecological scenario and use this information to set a management threshold implementation range for the preferred management alternatives. This range is used by decision makers to set management thresholds. We recommend scenario planning as a technique to assist participants in developing ecological scenarios that represent plausible future states of the key ecological indicator. Scenario planning has a long history of use in business strategic planning and has been advocated over the last decade to support conservation decisions (Peterson et al. 2003; Game et al. 2013). It is particularly useful in facilitating the creative exploration of complex natural systems, the uncertainty of future trajectories of these systems, and the development of management alternatives to address future unfavorable states. The most positively reported aspect of the workshop was that it was well-organized and facilitated (75% of participants [Supporting Information]). Well facilitated participatory modeling approaches can improve communication and build trust between scientists and decision makers (Addison et al. 2013). Part of the success of the workshop came from using the participatory modeling approach that did not require in-depth quantitative modeling skills. Widely accessible spreadsheet software was used to construct the weighted additive model, and model outputs were created during the workshop and viewed immediately by participants (e.g., Fig. 2). Thus, this approach is accessible to management agencies and participants with limited modeling expertise. Our approach was not without some challenges; in particular participants can be subject to a substantial elicitation burden when involved in participatory modeling approaches (Speirs-Bridge et al. 2010). Sixty-three percent of participants found the elicitation of quantitative consequence estimates under each ecological scenario particularly challenging (Supporting Information). Participants believed there was a lack of opportunity to document the assumptions and rationale behind the estimates effectively. Some also thought unqualified to provide quantitative estimates and noted a desire to revisit the quantitative modeling stage. These limitations can be overcome by documenting participants' assumptions to reveal differences in perspectives; documenting supporting information so that consequence estimates can be re-visited and updated with new scientific knowledge when available; or using rapid prototyping to increase participants' familiarity with the modeling process and to refine, improve, or expand modeling steps as necessary (Blomquist et al. 2010). Uncertainty is an important component to consider in any decision (Burgman 2005; Regan et al. 2005), and explicit consideration of uncertainty is encouraged throughout this participatory modeling approach. First, ecological scenarios, the foundation for setting management thresholds, are developed to represent uncertainty associated with the future trajectories of ecological systems. Second, the outputs of the weighted additive model carry through the uncertainty associated with participants' consequence estimates, which reflected participants' best guess and 90% credible bounds (Fig. 2). This is unlike other approaches, such as stochastic dynamic programming, which provide only an optimal solution to setting management thresholds without considering uncertainty (e.g., Martin et al. 2009). Third, management threshold implementation ranges are set in the face of uncertainty with one of three decision rules in step 5. For example, a decision maker may take a precautionary approach and use the pessimistic rule to set a management threshold implementation range for 15–42% cover of H. banksii for the medium protection alternative (Fig. 3). Where a management threshold will be set within the management threshold implementation range reflects a decision makers' risk tolerance associated with the decision context. Just like the need for transparently incorporating uncertainty, decision makers' should openly state their appetite for risk in conservation decision making (Burgman 2005; Regan et al. 2005). In our case study, we illustrated how a risk-averse decision maker may err on side of caution and set the medium protection management threshold at 42% cover of H. banksii. If the preferred management alternative is not apparent due to uncertainty, there are several techniques to help clarify uncertainty that can be incorporated into the modeling approach. Sensitivity analysis can be used to highlight whether any model parameters need to be re-defined or re-estimated (demonstrated in Supporting Information). Expected value of perfect information analysis can be used to assess the benefits of clarifying uncertainty before investing in potentially expensive data gathering exercises (e.g., Runge et al. 2011; Moore & Runge 2012). If warranted, investment in additional research can complement expert judgment and experimental adaptive management could be used to trial management alternatives and determine their effectiveness (Walters & Holling 1990). The feedback loops of this approach highlight the recurrent nature of state-dependent management and encourage revisiting modeling steps where necessary (Fig. 1). The review feedback loop provides an opportunity to clarify uncertainty associated with model parameters and revise the decision context where necessary. The learn feedback loop ensures that the effectiveness of management actions is assessed using empirical evidence and if necessary modifications are made to the management thresholds to ensure future management is more effective. This type of knowledge updating is considered passive, compared with experimental adaptive management (Walters & Holling 1990; Runge 2011); nevertheless, it is vital to ensure the on-going effectiveness of state-based management. This participatory modeling approach encourages a proactive form of conservation management, where management thresholds and associated management actions are defined a priori for ecological indictors, rather than reacting to unexpected future ecosystem changes. Future budgets and organizational priorities may prevent implementing a particular management action, but at a minimum, breaching a management threshold will trigger a more informed investigation to determine what new management strategies are obtainable. We thank the workshop participants who generously donated their time to participate in this research. This research was part of P.F.E.A.'s PhD at the University of Melbourne. P.F.E.A. was funded by an Australian Postgraduate Award scholarship, with generous support from the Centre of Excellence for Biosecurity Risk Analysis, Parks Victoria's Research Partners Panel program, and the Australian Government's National Environment Research Program, Marine Biodiversity Research Hub. L.R. is funded by LP110100321 and supported by the ARC Centre of Excellence for Environmental Decisions. K.dB. is funded by Parks Victoria's Research Partners Panel program and supported by the National Environment Research Program and ARC Centre of Excellence for Environmental Decisions. Disclaimer: Supplementary materials have been peer-reviewed but not copyedited. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.