Models
The proportion of variance explained was generally high for all models (16-83% - Table 2) and was overall greater for angiosperms than centipedes. Biogeographical affinity explained a large proportion of variance for all endemic categories (13-67% - Table 2), but not for the native angiosperms and centipedes (Table 2). Area was the most important predictor in almost all models (Table 2; Figure 4). LGM-related variables emerged as important predictors for all angiosperm chorotypes, except for the Greek endemics and the SIEs (Table 2; Figure 4). MSL area loss had an insignificant negative effect on SIE richness patterns (Table 2; Figure 4).