Scale dependency of metapopulation models used to predict climate change impacts on small mammals

Abstract

To investigate potential range shifts in a changing climate it is becoming increasingly common to develop models that account for demographic processes. Metapopulation models incorporate the spatial configuration of occupied habitat (i.e. arrangement, size and quality), population demographics, and inter-patch dispersal making them suitable for investigating potential threats to small mammal range and abundance. However, the spatial scale (resolution) used to represent species–environment dynamics may affect estimates of range shift and population resilience by failing to realistically represent the spatial configuration of suitable habitat, including stepping stones and refugia. We aimed to determine whether relatively fine-scale environmental information influenced predictions of metapopulation persistence and range shift. Species distribution models were constructed for four small terrestrial mammals from southern Australia using environmental predictors measured at 0.1 × 0.1 km (0.01 km²) or 1.0 × 1.0 km (1 km²) resolution, and combined with demographic information to parameterise coupled niche-population models. These models were used to simulate population dynamics projected over 40-yr under a stable and changing climate. Initial estimates of the area of available habitat were similar at both spatial scales. However, at the fine-scale, habitat configuration comprised a greater number of patches (ca 12 times), that were more irregular in shape (ca 8 times the perimeter:area), and separated by a tenth of the distance than at the coarse-scale. While small patches were not more prone to extinction, populations generally declined at a higher rate and were associated with a lower expected minimum abundance. Despite increased species vulnerability at the fine-scale, greater range shifts were measured at the coarse-scale (for species illustrating a shift at both scales). These results highlight the potential for range shifts and species vulnerability information to be misrepresented if advanced modelling techniques incorporating species demographics and dispersal inadequately represent the scale at which these processes occur.