A Bayesian Hierarchical Approach for Estimating the Probability of Detecting a New Prey Item: How Many Stomach Samples Is Enough?

Multi-species approaches are increasingly being used to gain a better understanding of ecosystem structure and population dynamics.  Food habits data are critical for such approaches; however, there are no objective techniques to determine the number of samples required to fully describe a predator’s diet.  We developed a Bayesian state-space model that estimates, given the observed prey distribution, the probability of detecting a new prey species when individual predators are sampled in clusters, such as in a trawl net.   The approach jointly models the probability of occurrence and detection.  Occurrence describes if prey species j is present in the population being sampled; occurrence probability is the proportion of the samples where prey species j occurs.  Detection is the probability of observing the species in the stomach sample cluster, given it occurs in the sampling area.  To test our approach, we simulated stomach content data for generalist and specialists and determined the number of stomachs necessary to fully describe the diet.  Additionally, we compared results from our approach to species accumulation curves.  This model will aid fisheries biologists by providing an objective approach to determine the required sample size for a food habits study.