Species distributions reflect many factors that constrain any single species’ distribution from its fundamental to realized niche. Realized niches reflect constraints by different factors over small spatial extents (e.g., lower- vs. upper-elevation edges of occupancy, within mountains); additionally, species-climate relationships may reflect clinal variability in local adaptation, ecological context, and climate. Despite these facts, innumerable species-distribution models (SDMs) forecast with spatially-unvarying predictors. Because American pikas are known to respond to variability in climatic conditions at multiple temporal and spatial scales, we are using species distribution modeling (with ensembles of multiple modeling frameworks) to challenge the assertion that climate constrains distribution via the same climate-aspect and functional form across all parts of a species’ range. Using data from >70 contributors, we have designed novel ways to rigorously: a) define the ‘background’ available to individual pika detections; b) assess the degree of overlap in importance of climatic predictors, across subunits; c) quantify the consequences of correcting for spatial, temporal, and spatio-temporal bias; and d) compare predictive ability of several schemes of subdividing the species range into subunits.
Results/Conclusions
Ecological niche models (ENMs) framed on the different range-subdivision schemes provided vastly different predictions of currently suitable distribution of Ochotona princeps. Furthermore, different division schemes best explained relationships to climate depending on the spatial extent at which relationships were examined, reflecting differences between long- and short-term processes of climate adaptation and suggesting the influence of local versus broader-scale drivers of species-climate relationships. Although we hypothesized a priori that evolutionary history (i.e., subspecies designations) would perform best, we found that heterogeneity in species-climate relationships was typically best explained by the ecoregional scheme across all three classes of analyses, especially at longer time scales. In some cases, responses to some important climate variables changed sign depending on the unit of analysis (e.g., Sierran vs. other clades), even within the same division scheme. Elevational subdivisions performed worst, in nearly every analysis, even though climatic variables often covary with elevation. In addition to revolutionizing the analysis of ENMs via our novel methods, this study holds great promise for informing climate-adaptation management actions because it identifies the particular mechanisms limiting a species’ distribution at the local extents at which such actions are typically implemented.