How decisions about fitting species distribution models affect conservation outcomes

Species distribution models (SDMs) are increasingly used in conservation and land-use planning as inputs to describe biodiversity patterns. These models can be built in different ways, and decisions about data preparation, selection of predictor variables, model fitting, and evaluation all alter the resulting predictions. Commonly, the true distribution of species is unknown and independent data to verify which SDM variant to choose are lacking. Such model uncertainty is of concern to planners. We analyzed how 11 routine decisions about model complexity, predictors, bias treatment, and setting thresholds for predicted values altered conservation priority patterns across 25 species. Models were created with MaxEnt and run through Zonation to determine the priority rank of sites. Although all SDM variants performed well (area under the curve >0.7), they produced spatially different predictions for species and different conservation priority solutions. Priorities were most strongly altered by decisions to not address bias or to apply binary thresholds to predicted values; on average 40% and 35%, respectively, of all grid cells received an opposite priority ranking. Forcing high model complexity altered conservation solutions less than forcing simplicity (14% and 24% of cells with opposite rank values, respectively). Use of fewer species records to build models or choosing alternative bias treatments had intermediate effects (25% and 23%, respectively). Depending on modeling choices, priority areas overlapped as little as 10–20% with the baseline solution, affecting top and bottom priorities differently. Our results demonstrate the extent of model-based uncertainty and quantify the relative impacts of SDM building decisions. When it is uncertain what the best SDM approach and conservation plan is, solving uncertainty or considering alterative options is most important for those decisions that change plans the most.     

The Sensitivity of Population Viability Analysis to Uncertainty about Habitat Requirements: Implications for the Management of the Endangered Southern Brown Bandicoot

Abstract:  Whenever population viability analysis (PVA) models are built to help guide decisions about the management of rare and threatened species, an important component of model building is the specification of a habitat model describing how a species is related to landscape or bioclimatic variables. Model-selection uncertainty may arise because there is often a great deal of ambiguity about which habitat model structure best approximates the true underlying biological processes. The standard approach to incorporate habitat models into PVA is to assume the best habitat model is correct, ignoring habitat-model uncertainty and alternative model structures that may lead to quantitatively different conclusions and management recommendations. Here we provide the first detailed examination of the influence of habitat-model uncertainty on the ranking of management scenarios from a PVA model. We evaluated and ranked 6 management scenarios for the endangered southern brown bandicoot ( Isoodon obesulus ) with PVA models, each derived from plausible competing habitat models developed with logistic regression. The ranking of management scenarios was sensitive to the choice of the habitat model used in PVA predictions. Our results demonstrate the need to incorporate methods into PVA that better account for model uncertainty and highlight the sensitivity of PVA to decisions made during model building. We recommend that researchers search for and consider a range of habitat models when undertaking model-based decision making and suggest that routine sensitivity analyses should be expanded to include an analysis of the impact of habitat-model uncertainty and assumptions.     

Using global sensitivity analysis of demographic models for ecological impact assessment

Population viability analysis (PVA) is widely used to assess population‐level impacts of environmental changes on species. When combined with sensitivity analysis, PVA yields insights into the effects of parameter and model structure uncertainty. This helps researchers prioritize efforts for further data collection so that model improvements are efficient and helps managers prioritize conservation and management actions. Usually, sensitivity is analyzed by varying one input parameter at a time and observing the influence that variation has over model outcomes. This approach does not account for interactions among parameters. Global sensitivity analysis (GSA) overcomes this limitation by varying several model inputs simultaneously. Then, regression techniques allow measuring the importance of input‐parameter uncertainties. In many conservation applications, the goal of demographic modeling is to assess how different scenarios of impact or management cause changes in a population. This is challenging because the uncertainty of input‐parameter values can be confounded with the effect of impacts and management actions. We developed a GSA method that separates model outcome uncertainty resulting from parameter uncertainty from that resulting from projected ecological impacts or simulated management actions, effectively separating the 2 main questions that sensitivity analysis asks. We applied this method to assess the effects of predicted sea‐level rise on Snowy Plover (Charadrius nivosus). A relatively small number of replicate models (approximately 100) resulted in consistent measures of variable importance when not trying to separate the effects of ecological impacts from parameter uncertainty. However, many more replicate models (approximately 500) were required to separate these effects. These differences are important to consider when using demographic models to estimate ecological impacts of management actions.     

Fooled by local robustness: an applied ecology perspective

In this short discussion, we point out that it is apparently as easy to be fooled by robustness as it is to be fooled by randomness. Our objective is to bring to the attention of applied ecologists that radius-of-stability robustness models are models of local robustness. As such, these models are utterly unsuitable for the treatment/management of a severe uncertainty characterized by a vast uncertainty space and a likelihood-free quantification of the uncertainty. This observation is particularly pertinent to applications of info-gap decision theory in ecology, conservation biology, and environmental management, where the objective is to identify decisions that are robust against a severe uncertainty of this type.     

Predicting barrier passage and habitat suitability for migratory fish species

Fish migrate to spawn, feed, seek refuge from predators, and escape harmful environmental conditions. The success of upstream migration is limited by the presence of barriers that can impede the passage of fish. We used a spatially explicit modeling strategy to examine the effects of barriers on passage for 21 native and non-native migratory fish species and the amount of suitable habitat blocked for each species. Spatially derived physical parameter estimates and literature based fish capabilities and tolerances were used to predict fish passage success and habitat suitability. Both the fish passage and the habitat suitability models accurately predicted fish presence above barriers for most common, non-stocked species. The fish passage model predicted that barriers greater than or equal to 6 m block all migratory species. Chinook salmon (Oncorhynchus tshawytscha) was expected to be blocked the least. The habitat suitability model predicted that low gradient streams with intact habitat quality were likely to support the highest number of fish species. The fish passage and habitat suitability models were intended to be used by environmental managers as strategy development tools to prioritize candidate dams for field assessment and make decisions regarding the management of migratory fish populations.     

Incorporating Uncertainty into Management Models for Marine Mammals

Abstract: Good management models and good models for understanding biology differ in basic philosophy. Management models must facilitate management decisions despite large amounts of uncertainty about the managed populations. Such models must be based on parameters that can be estimated readily, must explicitly account for uncertainty, and should be simple to understand and implement. In contrast, biological models are designed to elucidate the workings of biology and should not be constrained by management concerns. We illustrate the need to incorporate uncertainty in management models by reviewing the inadequacy of using standard biological models to manage marine mammals in the United States. Past management was based on a simple model that, although it may have represented population dynamics adequately, failed as a management tool because the parameter that triggered management action was extremely difficult to estimate for the majority of populations. Uncertainty in parameter estimation resulted in few conservation actions. We describe a recently adopted management scheme that incorporates uncertainty and its resulting implementation. The approach used in this simple management scheme, which was tested by using simulation models, incorporates uncertainty and mandates monitoring abundance and human-caused mortality. Although the entire scheme may be suitable for application to some terrestrial and marine problems, two features are broadly applicable: the incorporation of uncertainty through simulations of management and the use of quantitative management criteria to translate verbal objectives into levels of acceptable risk.     

Measuring impacts on species with models and metrics of varying ecological and computational complexity

Approaches to assess the impacts of landscape disturbance scenarios on species range from metrics based on patterns of occurrence or habitat to comprehensive models that explicitly include ecological processes. The choice of metrics and models affects how impacts are interpreted and conservation decisions. We explored the impacts of 3 realistic disturbance scenarios on 4 species with different ecological and taxonomic traits. We used progressively more complex models and metrics to evaluate relative impact and rank of scenarios on the species. Models ranged from species distribution models that relied on implicit assumptions about environmental factors and species presence to highly parameterized spatially explicit population models that explicitly included ecological processes and stochasticity. Metrics performed consistently in ranking different scenarios in order of severity primarily when variation in impact was driven by habitat amount. However, they differed in rank for cases where dispersal dynamics were critical in influencing metapopulation persistence. Impacts of scenarios on species with low dispersal ability were better characterized using models that explicitly captured these processes. Metapopulation capacity provided rank orders that most consistently correlated with those from highly parameterized and data-rich models and incorporated information about dispersal with little additional computational and data cost. Our results highlight the importance of explicitly considering species’ ecology, spatial configuration of habitat, and disturbance when choosing indicators of species persistence. We suggest using hybrid approaches that are a mixture of simple and complex models to improve multispecies assessments.     

Making conservation decisions under uncertainty for the persistence of multiple species.

Population models for multiple species provide one of the few means of assessing the impact of alternative management options on the persistence of biodiversity, but they are inevitably uncertain. Is it possible to use population models in multiple-species conservation planning given the associated uncertainties? We use information-gap decision theory to explore the impact of parameter uncertainty on the conservation decision when planning for the persistence of multiple species. An information-gap approach seeks robust outcomes that are most immune from error. We assess the impact of uncertainty in key model parameters for three species, whose extinction risks under four alternative management scenarios are estimated using a metapopulation model. Three methods are described for making conservation decisions across the species, taking into account uncertainty. We find that decisions based on single species are relatively robust to uncertainty in parameters, although the estimates of extinction risk increase rapidly with uncertainty. When identifying the best conservation decision for the persistence of all species, the methods that rely on the rankings of the management options by each species result in decisions that are similarly robust to uncertainty. Methods that depend on absolute values of extinction risk are sensitive to uncertainty, as small changes in extinction risk can alter the ranking of the alternative scenarios. We discover that it is possible to make robust conservation decisions even when the uncertainties of the multiple-species problem appear overwhelming. However, the decision most robust to uncertainty is likely to differ from the best decision when uncertainty is ignored, illustrating the importance of incorporating uncertainty into the decision-making process.     

Conserving and managing the subnivium

In regions where snowfall historically has been a defining seasonal characteristic of the landscape, warming winters have reduced the depth, duration, and extent of snowpack. However, most management and conservation has focused on how aboveground wildlife will be affected by altered snow conditions, even though the majority of species that persist through the winter do so under the snowpack in a thermally stable refugium: the subnivium. Shortened winters, forest management practices, and winter recreation can alter subnivium conditions by increasing snow compaction and compromising thermal stability at the soil–snow interface. To help slow the loss of the subnivium in the face of rapidly changing winter conditions, we suggest managers adopt regional conservation plans for identifying threatened snow‐covered environments; measure and predict the effects land cover and habitat management has on local subnivium conditions; and control the timing and distribution of activities that disturb and compact snow cover (e.g., silvicultural practices, snow recreation, and road and trail maintenance). As a case study, we developed a spatially explicit model of subnivium presence in a working landscape of the Chequamegon National Forest, Wisconsin. We identified landscapes where winter recreation and management practices could threaten potentially important areas for subnivium persistence. Similar modeling approaches could inform management decisions related to subnivium conservation. Current climate projections predict that snow seasons will change rapidly in many regions, and as result, we advocate for the immediate recognition, conservation, and management of the subnivium and its dependent species.     

Comparison of statistical and theoretical habitat models for conservation planning: the benefit of ensemble prediction

Selection of a modeling approach is an important step in the conservation planning process, but little guidance is available. We compared two statistical and three theoretical habitat modeling approaches representing those currently being used for avian conservation planning at landscape and regional scales: hierarchical spatial count (HSC), classification and regression tree (CRT), habitat suitability index (HSI), forest structure database (FS), and habitat association database (HA). We focused our comparison on models for five priority forest-breeding species in the Central Hardwoods Bird Conservation Region: Acadian Flycatcher, Cerulean Warbler, Prairie Warbler, Red-headed Woodpecker, and Worm-eating Warbler. Lacking complete knowledge on the distribution and abundance of each species with which we could illuminate differences between approaches and provide strong grounds for recommending one approach over another, we used two approaches to compare models: rank correlations among model outputs and comparison of spatial correspondence. In general, rank correlations were significantly positive among models for each species, indicating general agreement among the models. Worm-eating Warblers had the highest pairwise correlations, all of which were significant (P < 0.05). Red-headed Woodpeckers had the lowest agreement among models, suggesting greater uncertainty in the relative conservation value of areas within the region. We assessed model uncertainty by mapping the spatial congruence in priorities (i.e., top ranks) resulting from each model for each species and calculating the coefficient of variation across model ranks for each location. This allowed identification of areas more likely to be good targets of conservation effort for a species, those areas that were least likely, and those in between where uncertainty is higher and thus conservation action incorporates more risk. Based on our results, models developed independently for the same purpose (conservation planning for a particular species in a particular geography) yield different answers and thus different conservation strategies. We assert that using only one habitat model (even if validated) as the foundation of a conservation plan is risky. Using multiple models (i.e., ensemble prediction) can reduce uncertainty and increase efficacy of conservation action when models corroborate one another and increase understanding of the system when they do not.     

Rapid assessment of management options for promoting stock rebuilding in data-poor species under climate change

The development of species recovery plans requires considering likely outcomes of different management interventions, but the complicating effects of climate change are rarely evaluated. We examined how qualitative network models (QNMs) can be deployed to support decision making when data, time, and funding limitations restrict use of more demanding quantitative methods. We used QNMs to evaluate management interventions intended to promote the rebuilding of a collapsed stock of blue king crab (Paralithodes platypus) (BKC) around the Pribilof Islands (eastern Bering Sea) to determine how their potential efficacy may change under climate change. Based on stakeholder input and a literature review, we constructed a QNM that described the life cycle of BKC, key ecological interactions, potential climate-change impacts, relative interaction strengths, and uncertainty in terms of interaction strengths and link presence. We performed sensitivity analyses to identify key sources of prediction uncertainty. Under a scenario of no climate change, predicted increases in BKC were reliable only when stock enhancement was implemented in a BKC hatchery-program scenario. However, when climate change was accounted for, the intervention could not counteract its adverse impacts, which had an overall negative effect on BKC. The remaining management scenarios related to changes in fishing effort on BKC predators. For those scenarios, BKC outcomes were unreliable, but climate change further decreased the probability of observing recovery. Including information on relative interaction strengths increased the likelihood of predicting positive outcomes for BKC approximately 5–50% under the management scenarios. The largest gains in prediction precision will be made by reducing uncertainty associated with ecological interactions between adult BKC and red king crab (Paralithodes camtschaticus). Qualitative network models are useful options when data are limited, but they remain underutilized in conservation.     

Incorporating Climate Science in Applications of the U.S. Endangered Species Act for Aquatic Species

Aquatic species are threatened by climate change but have received comparatively less attention than terrestrial species. We gleaned key strategies for scientists and managers seeking to address climate change in aquatic conservation planning from the literature and existing knowledge. We address 3 categories of conservation effort that rely on scientific analysis and have particular application under the U.S. Endangered Species Act (ESA): assessment of overall risk to a species; long‐term recovery planning; and evaluation of effects of specific actions or perturbations. Fewer data are available for aquatic species to support these analyses, and climate effects on aquatic systems are poorly characterized. Thus, we recommend scientists conducting analyses supporting ESA decisions develop a conceptual model that links climate, habitat, ecosystem, and species response to changing conditions and use this model to organize analyses and future research. We recommend that current climate conditions are not appropriate for projections used in ESA analyses and that long‐term projections of climate‐change effects provide temporal context as a species‐wide assessment provides spatial context. In these projections, climate change should not be discounted solely because the magnitude of projected change at a particular time is uncertain when directionality of climate change is clear. Identifying likely future habitat at the species scale will indicate key refuges and potential range shifts. However, the risks and benefits associated with errors in modeling future habitat are not equivalent. The ESA offers mechanisms for increasing the overall resilience and resistance of species to climate changes, including establishing recovery goals requiring increased genetic and phenotypic diversity, specifying critical habitat in areas not currently occupied but likely to become important, and using adaptive management. Incorporación de las Ciencias Climáticas en las Aplicaciones del Acta Estadunidense de Especies en Peligro para Especies Acuáticas     

Cost‐effective conservation of an endangered frog under uncertainty

How should managers choose among conservation options when resources are scarce and there is uncertainty regarding the effectiveness of actions? Well‐developed tools exist for prioritizing areas for one‐time and binary actions (e.g., protect vs. not protect), but methods for prioritizing incremental or ongoing actions (such as habitat creation and maintenance) remain uncommon. We devised an approach that combines metapopulation viability and cost‐effectiveness analyses to select among alternative conservation actions while accounting for uncertainty. In our study, cost‐effectiveness is the ratio between the benefit of an action and its economic cost, where benefit is the change in metapopulation viability. We applied the approach to the case of the endangered growling grass frog (Litoria raniformis), which is threatened by urban development. We extended a Bayesian model to predict metapopulation viability under 9 urbanization and management scenarios and incorporated the full probability distribution of possible outcomes for each scenario into the cost‐effectiveness analysis. This allowed us to discern between cost‐effective alternatives that were robust to uncertainty and those with a relatively high risk of failure. We found a relatively high risk of extinction following urbanization if the only action was reservation of core habitat; habitat creation actions performed better than enhancement actions; and cost‐effectiveness ranking changed depending on the consideration of uncertainty. Our results suggest that creation and maintenance of wetlands dedicated to L. raniformis is the only cost‐effective action likely to result in a sufficiently low risk of extinction. To our knowledge we are the first study to use Bayesian metapopulation viability analysis to explicitly incorporate parametric and demographic uncertainty into a cost‐effective evaluation of conservation actions. The approach offers guidance to decision makers aiming to achieve cost‐effective conservation under uncertainty.     

Assigning trend-based conservation status despite high uncertainty

Estimates of temporal trends in species’ occupancy are essential for conservation policy and planning, but limitations to the data and models often result in very high trend uncertainty. A critical source of uncertainty that degrades scientific credibility is that caused by disagreement among studies or models. Modelers are aware of this uncertainty but usually only partially estimate it and communicate it to decision makers. At the same time, there is growing awareness that full disclosure of uncertainty is critical for effective translation of science into policies and plans. But what are the most effective approaches to estimating uncertainty and communicating uncertainty to decision makers? We explored how alternative approaches to estimating and communicating uncertainty of species trends could affect decisions concerning conservation status of freshwater fishes. We used ensemble models to propagate trend uncertainty within and among models and communicated this uncertainty with categorical distributions of trend direction and magnitude. All approaches were designed to fit an established decision-making system used to assign species conservation status by the New Zealand government. Our results showed how approaches that failed to fully disclose uncertainty, while simplifying the information presented, could hamper species conservation or lead to ineffective decisions. We recommend an approach that was recently used effectively to communicate trend uncertainty to a panel responsible for setting the conservation status of New Zealand's freshwater fishes.     

Sensitivity of population viability to spatial and nonspatial parameters using GRIP.

Metapopulation dynamics are influenced by spatial parameters including the amount and arrangement of suitable habitat, yet these parameters may be uncertain when deciding how to manage species or their habitats. Sensitivity analyses of population viability analysis (PVA) models can help measure relative parameter influences on predictions, identify research priorities for reducing uncertainty, and evaluate management strategies. Few spatial PVAs, however, include sensitivity analyses of both spatial and nonspatial parameters, perhaps because computationally efficient tools for such analyses are lacking or inaccessible. We developed GRIP, a program to facilitate sensitivity analysis of spatial and nonspatial input parameters for PVAs created in RAMAS Metapop, a widely applied software program. GRIP creates random sets of input files by varying parameters specified in the PVA model including vital rates and their correlations among populations, the number and configuration of populations, dispersal rates, dispersal survival, initial population abundances, carrying capacities, and the probability, intensity, and spatial extent of catastrophes, while drawing on specified parameter distributions. We evaluated GRIP's performance as a tool for sensitivity analysis of spatial PVAs and explored the consequences of varying spatial input parameters for predictions of a published PVA model of the sand lizard (Lacerta agilis). We used GRIP output to generate standardized regression coefficients (SRCs) and nonparametric correlation coefficients as indices of the relative sensitivity of predicted conservation status to input parameters. GRIP performed well; with a single analysis we were able to rank the relative influence of input parameters identified as influential by the PVA's original author, S. A. Berglind, who used three separate forms of sensitivity analysis. Our analysis, however, also underscored the value of exploring the relative influence of spatial parameters on PVA predictions; both SRCs and correlation coefficients indicated that the most influential parameters in the sand lizard model were spatial in nature. We provide annotated code so that GRIP may be modified to reflect particular species biology, customized for more complex spatial PVA models, upgraded to incorporate features added in newer versions of RAMAS Metapop, used as a template to develop similar programs, or used as it is for computationally efficient sensitivity analyses in support of conservation planning.     

Impacts of Citrus Development on Habitats of Southwest Florida

Examining resource management needs at the landscape level has become critical for the conservation of ecosystems and the preservation of species. Geographic information systems (GIS) that allow for the integration of spatially referenced databases are a powerful tool that can be used by resource managers to examine potential impacts and develop strategies for regional planning. We applied a landscape-level approach to examine the potential impacts of citrus development on habitats and species in southwest Florida. We developed GIS models for panthers, Sandhill Cranes, and wading birds that reflect changes in potential habitats under a series of development scenarios. The models indicate that, under the maximum development scenario, 63% of potential panther habitat, 66% of potential Sandhill Crane habitat, and 67% and 33% of potential wading bird nesting and foraging habitats could be lost. In addition, the habitat that would remain would be severely fragmented. Several key areas were identified that will be critical to the continued existence of these species and to maintenance of regional biodiversity. The areas identified are habitats not represented on the existing public lands concentrated in the southern portion of the study area and/or that provide connections among existing natural areas.     

Knowledge coproduction on the impact of decisions for waterbird habitat in a changing climate

Scientists, resource managers, and decision makers increasingly use knowledge coproduction to guide the stewardship of future landscapes under climate change. This process was applied in the California Central Valley (USA) to solve complex conservation problems, where managed wetlands and croplands are flooded between fall and spring to support some of the largest concentrations of shorebirds and waterfowl in the world. We coproduced scenario narratives, spatially explicit flooded waterbird habitat models, data products, and new knowledge about climate adaptation potential. We documented our coproduction process, and using the coproduced models, we determined when and where management actions make a difference and when climate overrides these actions. The outcomes of this process provide lessons learned on how to cocreate usable information and how to increase climate adaptive capacity in a highly managed landscape. Actions to restore wetlands and prioritize their water supply created habitat outcomes resilient to climate change impacts particularly in March, when habitat was most limited; land protection combined with management can increase the ecosystem's resilience to climate change; and uptake and use of this information was influenced by the roles of different stakeholders, rapidly changing water policies, discrepancies in decision-making time frames, and immediate crises of extreme drought. Although a broad stakeholder group contributed knowledge to scenario narratives and model development, to coproduce usable information, data products were tailored to a small set of decision contexts, leading to fewer stakeholder participants over time. A boundary organization convened stakeholders across a large landscape, and early adopters helped build legitimacy. Yet, broadscale use of climate adaptation knowledge depends on state and local policies, engagement with decision makers that have legislative and budgetary authority, and the capacity to fit data products to specific decision needs.