Aim To assess whether the water availability measures commonly used in species distribution models might be misleading because they do not account for the hydrological effects of changes in vegetation structure and functioning.
Methods We compared different methods for estimating water availability in species distribution models with the soil water content predicted by a process-based ecosystem model. The latter also accounted for the hydrological effects of dynamic changes in vegetation structure and functioning, including potential physiological effects of increasing CO2.
Results All proxies showed similar patterns of water availability across Europe for current climate, but when projected into the future, the changes in the simpler water availability measures showed no correlation with those projected by the more complex ecosystem model, even if CO2 effects were switched off.
Main conclusions Results from species distribution modelling studies concerning future changes in species ranges and biodiversity should be interpreted with caution, and more process-based representations of the water balance of terrestrial ecosystems should be considered within these models.