Using climate variables to predict small mammal occurrence in hot, dry environments

Many species are considered to be threatened by the direct and synergistic effects of increased temperatures and less reliable rainfall caused by anthropogenic climate change. The climate of arid Australia is naturally defined by high temperatures and highly variable rainfall. In these environments the survival of small mammals is facilitated by a range of physiological, morphological and behavioural adaptive traits (e.g. insectivorous diets, torpor, nocturnal activity and sheltering in burrows). Features within the landscape associated with these traits may have a proximal influence on species occurrence and be critical for predicting the distribution of arid species. Here we explored the contribution of coarse- and fine-scale environmental variables, sampled from 150 to 10,000m from known species records, to the development of more robust species distribution models. The inclusion of environmental variables representing topography, soil and vegetation better modelled occurrence for all five small mammal species (up to 55.2% DE increase over climate-only models) and increased discrimination of species occurrence in an out-of-sample cross-validation for Antechinomys laniger, Dasyuroides byrnei and Notomys cervinus (e.g. increase in area-under-the curve by 0.4). For these species, model performance (% DE) increased with the resolution of environmental data used, reflecting a coarse-scale relationship with distance to watercourse (e.g. at 1,000m), but masking a fine-scale relationship with rainfall during winter (at 150m). The inclusion of a fine-scale variable shrub cover under 0.5m (SD) improved the representation of S. macroura occurrence. These results show improved representation of habitats and resources using course- and fine-scale variables, which may be vital to model the occurrence of some species and their refuge habitats, but as static variables may limit accurate predictions of the effect of climate change on species range dynamics.