There is an urgent need for more ecologically realistic models for better predicting the effects of climate change on species' potential geographic distributions. Here we build ecological niche models using MAXENT and test whether selecting predictor variables based on biological knowledge and selecting ecologically realistic response curves can improve cross-time distributional predictions. We also evaluate how the method chosen for extrapolation into nonanalog conditions affects the prediction. We do so by estimating the potential distribution of a montane shrew (Mammalia, Soricidae, Cryptotis mexicanus) at present and the Last Glacial Maximum (LGM). Because it is tightly associated with cloud forests (with climatically determined upper and lower limits) whose distributional shifts are well characterized, this species provides clear expectations of plausible vs. implausible results. Response curves for the MAXENT model made using variables selected via biological justification were ecologically more realistic compared with those of the model made using many potential predictors. This strategy also led to much more plausible geographic predictions for upper and lower elevational limits of the species both for the present and during the LGM. By inspecting the modeled response curves, we also determined the most appropriate way to extrapolate into nonanalog environments, a previously overlooked factor in studies involving model transfer. This study provides intuitive context for recommendations that should promote more realistic ecological niche models for transfer across space and time.

Predictions from species distribution models (SDMs) are commonly used in support of environmental decision-making to explore potential impacts of climate change on biodiversity. However, because future climates are likely to differ from current climates, there has been ongoing interest in understanding the ability of SDMs to predict species responses under novel conditions (i.e., model transferability). Here, we explore the spatial and environmental limits to extrapolation in SDMs using forest inventory data from 11 model algorithms for 108 tree species across the western United States. Algorithms performed well in predicting occurrence for plots that occurred in the same geographic region in which they were fitted. However, a substantial portion of models performed worse than random when predicting for geographic regions in which algorithms were not fitted. Our results suggest that for transfers in geographic space, no specific algorithm was better than another as there were no significant differences in predictive performance across algorithms. There were significant differences in predictive performance for algorithms transferred in environmental space with GAM performing best. However, the predictive performance of GAM declined steeply with increasing extrapolation in environmental space relative to other algorithms. The results of this study suggest that SDMs may be limited in their ability to predict species ranges beyond the environmental data used for model fitting. When predicting climate-driven range shifts, extrapolation may also not reflect important biotic and abiotic drivers of species ranges, and thus further misrepresent the realized shift in range. Future studies investigating transferability of process based SDMs or relationships between geodiversity and biodiversity may hold promise.