Aim The transferability of species distribution models requires that species show climatic equilibrium throughout their entire distribution area. We test this assumption for the case of the spotted hyena, Crocuta crocuta, a large carnivore that has shifted its distribution over the last 100,000 years from a widespread Eurasian and African range to its current geographical distribution, restricted to the Sub-Saharan areas of the African continent.
Location Western Eurasia and Africa.
Methods The current realized distribution of C. crocuta was estimated using presences and reliable absences as well as climatic, land-cover and anthropic variables as predictors. The potential distribution was estimated using presences and a set of pseudo-absences selected from localities outside climatically suitable localities, with only climatic variables serving as predictors. The current potential distribution was transferred to the Last Interglacial period (126,000 yr bp) using the palaeoclimatic data yielded by the GENESIS 2 general circulation model, and validated with European fossil data. Generalized linear models were used on all occasions.
Results Climatic variables are able to predict the current distribution of the species with high accuracy. The geographical projection of this model indicates that the species is distributed over almost all of its potential suitable area, which allows us to suppose that the current distribution of this species is in climatic equilibrium. However, the time transference of model predictions for the western Eurasian region reveals almost no suitable conditions for hyenas, despite the widespread presence of C. crocuta fossil remains on this continent during the Last Interglacial period.
Main conclusions Our results indicate that, even when model results suggest a climatic equilibrium for a species distribution, the time transferability of such models does not necessarily provide realistic results. This occurs because the current geographical range does not allow estimations of all of the environmental requirements of a species. Therefore, any model trained with current data risks underestimating the potential suitable environmental and geographical range for species in a new area or time period.