Comparison of maps of species richness and endemism produced directly from species occurrences with those produced by stacked species distribution modelling

Knowledge of the spatial patterns of biodiversity is essential for identifying conservation hotspots. Two commonly used methods of biodiversity assessments are point-to-grid maps and habitat suitability maps generated by stacked species distribution modelling (S-SDM). We compared these two types of maps using three metrics: species richness, weighted endemism (WE), and corrected weighted endemism (CWE) produced for three monocot genera (Allium, Tulipa and Eremurus) in Uzbekistan. Also, we asked how close the patterns of species richness and weighted endemism predicted by S-SDM will be if the sampling density is significantly reduced. To answer this question, we compared the maps of species richness and WE produced using the entire dataset, with one in which species occurrences were thinned to no more than ten occurrences per species. The patterns of species richness and WE in point-to-grid maps and maps generated by S-SDM were very similar for all three genera. However, the spatial distribution of CWE generated by S-SDM was dramatically different from that of point-to-grid maps, and this difference was consistent across the three genera. For species richness and WE, the maps produced by S-SDM from the reduced and complete datasets were similar and close to the analogous point-to-grid maps. These results suggest that even incomplete surveys can provide valuable data for elucidating spatial biodiversity patterns and that for conservationists, S-SDM should be an essential part of their toolbox. However, the use of CWE in S-SDM applications should be avoided.