Aim Extinction risk is non-randomly distributed across phylogeny and space and is influenced by environmental conditions. We quantified the relative contribution of these factors to extinction risk to unveil the underlying macroecological processes and derive predictive models.
Methods Based on the IUCN global assessments, we divided 192 carnivore species into two dichotomous classes representing different levels of extinction risk. We used spatial proximity, phylogenetic relationship and environmental variables together with phylogenetic eigenvector regression and spatial eigenvector filters to model and predict threat status.
Results Our full models explained between 57% and 96% of the variance in extinction risk. Phylogeny and spatial proximity roughly explained between 21% and 70% of the total variation in all analyses, while the explanatory power of environmental conditions was relatively weaker (up to 15%). Phylogeny and spatial proximity contributed equally to the explained variance in the lower threat level, while spatial proximity was the most important factor in the models of the higher threat level. Prediction of threat status achieved 97% correct assignments.
Main conclusions Our approach differs fundamentally from current studies of extinction risk because it does not necessarily rely on life-history information. We clearly show that instead of treating phylogenetic inertia and spatial signal as statistical nuisances, space and phylogeny should be viewed as very useful in explaining a wide range of phenomena in comparative studies.