Predictability in species distributions: a global analysis across organisms and ecosystems

AimWe explored the effects of multiple species traits, spatial extent and ecosystem type on predictability in species distributions. LocationGlobal. MethodsWe assembled over 4900 published AUC (area under the curve of a receiver operating characteristic plot) values from the species distribution modelling literature. Our data covered broad variation in species characteristics such as body size or trophic position for taxa ranging from bacteria to mammals. Data covered ecosystems from freshwater to forests and encompassed geographical areas from the tropics to polar regions. We used generalized linear mixed models and boosted regression trees to analyse the AUC data. ResultsWe found that most AUC values originated from large-sized terrestrial taxa while studies considering smaller taxa, especially from aquatic ecosystems, were rare. Predictability was highest in autotrophs and in active non-flying taxa and increased with body size and study extent. There were marginal differences in predictability between ectotherms and endotherms and between taxa originating from different ecosystems. Main conclusionsOur results suggest that predictability in species distributions is related to organismal variables such as body size, dispersal mode and trophic position as well as to study extent. We also identified a gap in species distribution modelling studies for aquatic species and for small taxa.