Bayesian networks as a framework to step-down and support Strategic Habitat Conservation of data-poor species: A case study with King Rail (Rallus elegans) in Eastern North Carolina and Southeastern Virginia

Although Bayesian network (BN) models have been promoted to the conservation community as models well-suited to support adaptive management strategies, there have been few tests of these claims. To test the value of BNs to support U.S. Fish and Wildlife Service and U.S. Geological Service's Strategic Habitat Conservation approach to adaptive management, we modeled habitat occupancy of breeding King Rail, Rallus elegans, in Eastern North Carolina and Southeastern Virginia. The limited regional empirical data for this species, combined with its priority conservation status, made it an ideal candidate to explore strengths and weaknesses of an expert-based Bayesian modeling approach. Specifically, we evaluated whether BN models initiated with expert knowledge and incrementally updated with empirical data could effectively support the definition of population and habitat objectives at regional and local (e.g., refuge) scales. Following two years of field surveys, we compared occupancy predictions from the original expert-only BN model, using a variety of BN models updated with different methods and with different data, and empirically derived detection-adjusted occupancy estimates calculated in the program PRESENCE. To interpret differences among these models, we considered the relative contribution of spatial data error, expert error, and uncertainty to overall model error. Our results demonstrate how BN models can advance conservation for poorly documented species. We also provide recommendations to maximize the utility of expert knowledge within BNs designed to support the U.S Fish and Wildlife Service's and U.S. Geological Service's adaptive management processes. ii ACKNOWLEDGEMENTS We would like to thank all members of the Eastern North Carolina Southeastern Virginia Strategic Habitat Conservation Team for their support and feedback throughout this project. We especially acknowledge the contribution of Back Bay, Mackay Island, and Pocosin Lakes National Wildlife Refuges for hosting our field crew during the empirical studies. This work could not have been accomplished without the assistance of our colleagues at the Biodiversity and Information Center of North Carolina State University and the exemplary work of our student (S. Rogers), field crew