Five (or so) challenges for species distribution modelling

Species distribution modelling is central to both fundamental and applied research in biogeography. Over the past 10 years species distribution models have become commonplace in studies of biogeography, conservation biology, ecology, palaeoecology and wildlife management. Model fitting is usually based on pattern-recognition approaches, whereby associations between geographic occurrence of a species and a set of predictor variables are explored to allow or support statements of the mechanisms governing species’ distributions. These models allow estimation of species’ ecological requirements, although the degree to which causal relationships between species distributions and the predictor variables are unveiled depends on the adequacy of the predictors used for model building. In spite of the widespread use of niche-based species distribution models (e.g. Guisan & Thuiller, 2005), important conceptual, biotic and algorithmic uncertainties still need to be investigated if models are to make important contributions for conservation and biogeographical research. As with climate change research (e.g. Smith, 2002), the species’ distribution modelling community needs to deepen the ongoing debate, where the strengths and limitations of available approaches are investigated, fuelled by more rigorous assessments of the sensitivity of model outcomes to initial assumptions and parameters (Whittaker et al., 2005). In this essay, we identify five high-priority areas of inquiry for niche-based species distribution modelling: (1) clarification of the niche concept; (2) improved designs for sampling data for building models; (3) improved parameterization; (4) improved model selection and predictor contribution; and (5) improved model evaluation. The challenges discussed in this essay do not preclude the need for development of other areas of species distribution modelling research, but are critical for allowing this science to move forward.