Application of Bayesian robust design model to assess the impacts of a hurricane on shorebird demography

Abstract

The increasing use of Bayesian inference in population demography requires rapid advancements in modeling frameworks to approach the rigor and flexibility of the current suite of maximum-likelihood models. We developed an unbiased, Jolly–Seber robust design (JSRD) model that is both accessible and generalizable in a Bayesian hierarchical multistate framework. We integrated band and age-classification data to estimate site entry, temporary emigration, and apparent survival rates, as well as estimate age-class specific abundances. The complete model parameterization is provided in the Appendix S1, as well as tools for simulating capture histories and an assessment of model fit. We applied this model to determine whether these demographic processes in non-breeding population of American oystercatchers (Haematopus palliatus) were affected by a major hurricane event (Hurricane Matthew) in coastal Georgia. The JSRD model was demonstrably unbiased at relatively small sample sizes, and the majority of parameters were identifiable in the fully saturated model parameterization. In the model application, we found that Hurricane Matthew temporarily altered local population abundances of American oystercatchers through increased movements of individuals into and out of the observable population, but mortality rates were largely unaffected. Together, our results suggest that American oystercatchers were largely able to avoid the immediate demographic consequences (i.e., reduced survival) of Hurricane Matthew. Integrating age and band ratios from survey data allowed for more descriptive and potentially less biased estimates of age-specific abundance, relative to estimates generated solely from either mark–resight or survey data.