Occurrence and environmental data
We obtained locality point data for Scutiger boulengeri from our own fieldwork and the published literatures (Li et al. 2009, Hofmann et al. 2017). Localities cover Himalayas, QTP, Hengduan Mountains, Min Mountains and adjacent Mountains (Figure 1). Our final dataset comprises 96 geo-referenced occurrence records: 42 for East represented by 4 clades (9 for E. A clade, 9 for E. B, 10 for E. C, 14 for E. D) and 54 for West represented by 2 clades (35 for W. a clade and 19 for W. b).
We compiled 19 environmental variables and elevation for each period from the WorldClim database (Hijmans et al. 2005a) with a resolution of 30 seconds (~ 1 km) for each environment layer (Hijmans et al. 2005b). We included the dissimilarity of the enhanced vegetation index variable, drawn from the Global Habitat Heterogeneity project (Tuanmu and Jetz 2015). Because strong colinearity between environmental variables could inflate model accuracy (Veloz 2009, Boria et al. 2014), we examined pairwise correlations among the 21 variables within each clade distribution. We reduced autocorrelation of input environmental data by removing highly correlated variables with the threshold of Pearson’s correlation tests |r| >0.8 (Dormann et al. 2013). Finally, we chose 8 variables with lower correlation for subsequent analyses.