Diversity and Distributions (2017)

DOI: 10.1111/ddi.12663


Conventional spatial density models assume a constant relationship between densities and habitat covariates over some time period, typically a survey season. The estimated population size must change whenever total habitat availability changes. For highly mobile long‐lived species, however, density–habitat relationships likely adjust more rapidly than population size. We developed an integrated population‐redistribution model based on a more ecologically plausible alternative hypothesis: (1) population size is effectively constant over each survey season; (2) if habitat availability changes, then the population redistributes itself following an ideal free distribution process. Thus, the estimated relationship between densities and habitat covariates adjusts rather than population size. We constructed Bayesian hierarchical models corresponding to the conventional and alternative hypotheses and applied them to distance sampling data for Dall’s porpoise (Phocoenoides dalli), a highly mobile cetacean with distribution patterns closely tied to cool sea‐surface temperatures.

The Dall’s porpoise data provided strong support for the hypothesis based on an ideal free redistribution process. Our results indicate that the population size of Dall’s porpoise within the survey region was relatively stable over each summer/fall survey season, but the distribution expanded and contracted with the extent of suitable habitat. Over multiple survey seasons, the model partitioned variation in observed densities among three sources: variation in population size, the density–habitat relationship and measurement error, leading to lower and more ecologically plausible estimates of interannual variation in population size.

We conclude that the integrated population‐redistribution model (IPRM) presented here represents an ecologically plausible model for use in future assessments of the population size and dynamics of cetaceans and other highly mobile long‐lived species with variable spatial distributions.