The usefulness of elevation as a predictor variable in species distribution modelling

Species distribution models (SDMs) are increasingly used to forecast impacts of climate change on species geographic distributions, but the reliability of predictions is scrutinized. The main limitation of SDMs lies in their assumption that species’ ranges are determined mostly by climate, which is arguable. For instance, biotic interactions, habitat and elevation may affect species ranges. The inclusion of habitat-related variables as predictors in SDMs is generally accepted, but there is no consensus regarding the inclusion of elevation. A review of randomly chosen literature revealed that elevation is used as a predictor variable by just over half of the papers studied with no apparent trends as to why, except that papers predicting mammal species distributions for large regions included elevation more often than not, and that papers that predicted mammal ranges for small regions tended to exclude elevation. In addition, we compared the performance of SDMs with and without elevation as a predictor variable for the distribution of north European mammals and plants and found that the difference between their performances is statistically significant for mammals, slightly favouring exclusion of elevation. No differences were found for plants.

► Elevation is randomly used as a predictor in species distribution models. ► Predictions for mammals are generally more precise without elevation. ► Predictions for plants are not affected by including or excluding elevation. ► The appropriateness of including elevation should be assessed before each study.