Improving the accuracy of small vertebrate-based palaeoclimatic reconstructions derived from the Mutual Ecogeographic Range. A case study using geographic information systems and UDA-ODA discrimination methodology

Understanding past climate and the mechanisms of climate change remain major challenges in scientific research. The Mutual Ecogeographic Range (MER) method for climatic reconstruction uses the current geographical distribution of fossil assemblages to infer palaeoclimatic conditions. Current species distributions used in the MER method are usually obtained from biogeographic atlases that record the absence/presence of species in a 10 x 10 km grid. A 10 x 10 km area is quite broad and the method only records presence/absence, without considering the real area occupied by any given species. Thus, the method overlooks the fact that environmental heterogeneity is strongly related to topography, and climatic conditions may change abruptly over a few kilometres. To improve the accuracy of small vertebrate palaeoclimatic reconstructions derived from the MER method, we developed the UDA-ODA discrimination methodology, which is applied to stratigraphic units Xb and Upper Unit V from El Salt (Alcoi, Spain). We identified the Occupied Distribution Area (ODA), which denotes realistic occupied areas where the species may be present and the UDA (Uncertain Distribution Area) where the species may be absent. We achieved this by combining the species' distribution recorded in the atlases and the species environmental requirements described in the literature. Climate values resulting from the MER method and our UDA-ODA discrimination methodology showed significant differences. Extrapolating values from areas that are more similar to the areas occupied by the different species today yielded more representative climatic parameters for the assemblage tested, resulting in a more accurate palaeoclimatic reconstruction. Moreover, the discrimination analysis proposed here allows us to work with species whose distribution is currently disturbed, which was not possible using the MER method. (C) 2019 Elsevier Ltd. All rights reserved.