Not-so-splendid isolation: modeling climate-mediated range collapse of a montane mammal Ochotona princeps across numerous ecoregions

We modeled current and future distribution of suitable habitat for the talus-obligate montane mammal Ochotona princeps (American pika) across the western USA under increases in temperature associated with contemporary climate change, to: a) compare forecasts using only climate variables vs using those plus habitat considerations; b) identify possible patterns of range collapse (center vs margins, and large- vs small-sized patches); and c) compare conservation and management implications of changes at two taxonomic resolutions, and using binned- vs binary-probability maps. We used MaxEnt to analyze relationships between occurrence records and climatic variables to develop a bioclimatic-envelope model, which we refined by masking with a deductive appropriate-habitat filter based on suitable land-cover types. We used this final species-distribution model to predict distribution of suitable habitat under range-wide temperature increases from 1 to 7 degrees C, in 1 degrees C increments; we also compared these results to distribution under IPCC-forecasted climates for 2050 and 2080. Though all currently recognized lineages and traditionally defined subspecies were predicted to lose increasing amounts of habitat as temperatures rose, the most-dramatic range losses were predicted to occur among traditional subspecies. Nineteen of the 31 traditional US pika subspecies were predicted to lose > 98% of their suitable habitat under a 7 degrees C increase in the mean temperature of the warmest quarter of the year, and lineages were predicted to lose 88 95% of suitable habitat. Under a 4 degrees C increase, traditional subspecies averaged a predicted 73% (range = 4499%) reduction. The appropriate-habitat filter removed 406% of the predicted climatically suitable pixels, in a stepped and monotonically decreasing fashion as predicted temperatures rose. Predicted range collapse proceeded until only populations in island-biogeographic mainlands remained, which were not in the geographic range center. We used this model system to illustrate possible distributional shifts under stepped changes in biologically relevant aspects of climate, importance of land cover and taxonomic level in species-distribution forecasts, and impact of using a single threshold vs multiple categories of persistence probability in predicted range maps; we encourage additional research to further investigate the generality of these patterns.