Table 1: Results from a spatial autoregressive model (SAR) relating species richness to environmental, climatic, and historical variables. Spatial analysis was performed with statistical package  and spdep library  (see the Supplementary Appendix for further explanations). The spatial structure was implemented by a neighbourhood matrix of the drainage basins (see  and the Supplementary Appendix for further explanations) and assuming that the autoregressive process occurs in the error term (i.e., the “spatial error model” described by Dormann et al. ). Further methodological details on species richness, environmental variables computing, and modelling procedure are available in the Supplementary Appendix. Habitat heterogeneity was estimated by applying Shannon’s diversity index to proportions of biomes (i.e., vegetation types associated with regional variations in climate) within drainage basins. Temperature anomaly represents the Quaternary climate variability measured as the change in mean annual temperature between the present and the Last Glacial Maximum (LGM, circa 21 thousand years ago). Following Oberdorff et al.  we also considered whether or not a drainage basin was on a land mass, a peninsula, or an island (LPI; continental mass = 0; peninsula = 1; island = 2). All other variables are fully explained in the Supplementary Appendix. The Moran’s I value represents the remaining autocorrelation on the residuals of the model for the first distance class, that is, neighbour drainages (the values for the remaining distance classes are also nonsignificant).
Habitat size and diversity
Historical climatic stability and geographic isolation