Accounting for multiple uncertainties in a decision-support population viability assessment

Conservation and management decisions often must be made on strict timelines, based on the “best available information” regarding a species' current and expected future status. Simulation models are valuable tools for predicting a species' future status but must incorporate multiple types of uncertainty in order to provide a complete understanding of plausible outcomes. Here we present a population viability analysis for a data-deficient species proposed for protection under the U.S. Endangered Species Act, the alligator snapping turtle. We used a matrix population model to simulate population trajectories, incorporating both parametric uncertainty and temporal variation into demographic parameters. We used expert elicitation to generate modified survival rates in the presence of specific anthropogenic threats, for which empirical estimates were unavailable. Because uncertainty in the expert elicited values was of particular interest to decision makers, we constructed a set of simulation scenarios to evaluate the sensitivity of model conclusions to the accuracy of expert elicited parameters. Our model predicted steep population declines under all scenarios with anthropogenic threats, indicating that under- or overestimation by experts would not change the overall conclusion that populations would decline. An additional sensitivity analysis revealed that a parameter related to nest survival for which there was high disagreement among experts had a negligible effect on model outcome, while other parameters (e.g., the effect of poaching) had more influence. Our analyses demonstrate the use of an expert-parameterized decision-support population viability analysis that explicitly evaluates the effects of multiple sources of uncertainty on model predictions. • We present a simulation-based population model for alligator snapping turtles, a data-deficient species • We used scenario building and sensitivity analyses to evaluate effects of uncertainty on predictions • Populations declined in all simulated scenarios; uncertainty in expert elicited parameters had a small effect on results • Populations were sensitive to threats to adult survival and insensitive to threats to nest survival