Biological invasions reveal how niche change affects the transferability of species distribution models

Species distribution models (SDMs) have been widely applied to predict geographic ranges of species across space and time under the assumption of niche conservatism (i.e., species niches change very slowly). However, an increasing number of studies have reported evidence of rapid niche changes across space and time, which has sparked a widespread debate on whether SDMs can be transferred to new areas or time periods. Understanding how niche changes affect SDM transferability is thus crucial for the future application and improvement of SDMs. Biological invasions provide an opportunity to address this question due to the geographically independent distributions and diverse patterns of niche changes between species' native and introduced ranges. Here, we synthesized findings on 217 species from 50 studies to elucidate the effects of niche change on the spatial transferability of SDMs. When niche change was considered as a categorical classification (conserved vs. shifted niches) in tests of the niche conservatism hypothesis, SDM transferability was markedly lower for species with a shifted niche in their introduced range. When niche change was measured as numerical dynamics between native and introduced niches, SDM transferability was high for species occupying similar environmental conditions in both ranges and low for species with more environmental space remaining unoccupied in the introduced range. Surprisingly, the number of presence points used for developing SDMs turned out to have an even stronger effect on transferability. Our results thus reveal detrimental effects of both niche change and lack of presence points on SDM transferability. It is necessary to consider both niche change and data quality for improving the transferability of SDMs, so that they can better support conservation management and policy decisions.