Correlative species distribution models have gained dramatic popularity in ecology, conservation and global change research. These statistical models employ environmental correlates of species presences to characterize geographic distributions. Recently, distribution models have found widespread use for projecting species occurrences into environments that are novel, either spatially or temporally. Most noticeably, such models are often used to predict the future ranges of species under global warming and have become the basis of some widely debated impact assessments of projected climate change.
Such extrapolation beyond the conditions present in the data used to calibrate the model assumes that some underlying mechanistic relationship is being captured. It is assumed that distribution models capture the environmental factors constraining distributions adequately enough to represent the environmental niche of a species. However, it has been shown that even randomly generated data can produce apparently significant correlations and statistically ‘explain’ distributions. In response to this, some modelers have attempted to build more mechanistic models of species distributions that incorporate physiology and population dynamics. Such models are highly restricted in their use due to their demand for long-term data. Correlative distribution models offer a widely usable approach, but one that lacks a well-articulated ecological foundation or paradigm to date. Given the rapid rate of global climate change and its potentially massive impacts on biota, some form of tool is needed to assess the future of biodiversity in a changing world. Distribution models have been by far the most popular tool to date, but serious challenges remain.
Results/Conclusions
We evaluate the use of environmental niche models for ecological inference. We examine how their application helps us understand and test specific hypotheses about environmental vs. biotic constraints on species distributions from regional to global scale. As examples we use ants and birds. We examine the intricate relationship between species traits and ecology and the ability of distribution models to capture their Grinnellian niche. We discuss how the fitting of dozens of environmental variables and neglect to attribute ecological mechanisms to the parameters of environmental niche models may limit their usefulness for ecological forecasting. Finally, we attempt a general outline of how and where ecological thinking would benefit from environmental niche models and vice versa.
