Modeling spatial dynamics of Steller sea lions (Eumetopias jubatus) using maximum likelihood and Bayesian methods: Evaluating causes for population declin

Abstract

The timing and extent of the negative population trend in the abundance of the western stock of Steller sea lions has not been geographically uniform. A stochastic metapopulation dynamics model is developed for Steller sea lions. This model allows for geographical differences in factors affecting population processes, and can be parameterized to represent a wide range of hypotheses for the decline in Steller sea lion abundance. Bayesian and maximum likelihood methods are used to fit this model to pup and non-pup count data, age structure samples, and survival estimates. Inferences from model selection criteria highlight the spatial variability in the types of impact deemed to provide most parsimonious representation of the data. Bayesian posteriors for the estimated model parameters show that many combinations of parameter values are able to provide similar fits to the data, even given a specific hypothesis for the decline. This highlights the uncertainty in the precise nature of the impact of these hypotheses. Indeed, while pup production is generally estimated consistently among models, estimates of the size of other components of the Steller sea lion population (such as total population size) depend greatly on the assumptions regarding the cause of the decline. The results demonstrate that future simulation modeling approaches will require more formal, spatial, and mechanistic descriptions of the manner in which specific hypotheses for the decline affect the population. Sea Lions of the World 405 Alaska Sea Grant College Program • AK-SG-06-01, 2006