Bayesian analysis of wildlife age-at-harvest data

State and federal natural resource management agencies often collect age-structured harvest data. These data represent finite realizations of stochastic demographic and sampling processes and have long been used by biologists to infer population trends. However, different sources of data have been combined in ad hoc ways and these methods usually failed to incorporate sampling error. In this article, we propose a "hidden process" (or state-space) model for estimating abundance, survival, recovery rate, and recruitment from age-at-harvest data that incorporate both demographic and sampling stochasticity. To this end, a likelihood for age-at-harvest data is developed by embedding a population dynamics model within a model for the sampling process. Under this framework, the identification of abundance parameters can be achieved by conducting a joint analysis with an auxiliary data set. We illustrate this approach by conducting a Bayesian analysis of age-at-harvest and mark-recovery data from black bears (Ursus americanus) in Pennsylvania. Using a set of reasonable prior distributions, we demonstrate a substantial increase in precision when posterior summaries of abundance are compared to a bias-corrected Lincoln-Petersen estimator. Because demographic processes link consecutive abundance estimates, we also obtain a more realistic biological picture of annual changes in abundance. Because age-at-harvest data are often readily obtained, we argue that this type of analysis provides a valuable strategy for wildlife population monitoring.