Informing management with monitoring data: the value of Bayesian forecasting

Abstract

AbstractInventory and Monitoring Programs in the National Park Service (NPS) provide information needed to support wise planning, management, and decision making. Mathematical and statistical models play a critical role in this process by integrating data from multiple sources in a way that is honest about uncertainty. We show the utility of Bayesian hierarchical models for supporting decisions on managing natural resources of national parks. These models can assimilate monitoring data to provide true forecasts, resulting in probabilistic predictions of future states of park ecosystems accompanied by rigorous estimates of uncertainty. We discuss a novel approach for communicating these forecasts to decision makers who need to evaluate the probability that NPS goals will be met given different management actions, including the null model of no action. We illustrate how this approach has been used successfully to inform decisions on the elk (Cervus elaphus nelsoni) population management in Rocky Mountain National Park based on 47 yr of monitoring data. Forecasts from a discrete time, stage‐structured population model assimilated with annual census and sex and age classifications are being used annually to help park managers decide on actions needed to meet goals for elk and vegetation. In particular, park managers were able to determine the probability that the elk population would fall within a desired population range, which led to both population reduction actions and no action depending on the year of interest. Moreover, this approach allowed multiple survey methodologies from the last 47 years to be incorporated into a single model with associated estimates of uncertainty. Models like this one are especially useful for adaptive management where continuous improvement in models and data results in long‐term improvement in the wisdom of policy and management.