Prediction of the Maxent distribution
We choose the best model and find the corresponding RM and FCs parameters for Maxent model to facilitate increased rigour in the development of Maxent models (Table 2). Finally, the distributions based on Maxent across S. boulengeri clades were characterized by high AUC statistics, indicating that these ENMs successfully discriminated real occurrences from background locations. Jackknife tests on variable importance for S. boulengeri clades revealed that Precipitation of driest month in E. A, E. B and W. a clades, while Annual precipitation in E. C and E. D and Annual mean temperature in W. b produced the greatest decrease in gain when excluded from the model, suggesting these climate factors limit distributions of clades correspondingly, which likely to be the most important reasons for the next step niche divergence.
Hypothesis tests based on ENMs and PCA-env approaches
ENMs based niche equivalency tests reveal that 4 paired comparisons do not reject null distributions, seem equivalent as the values of observed niche overlap fall well in the middle of the null distributions (Supplementary material Appendix Figure A1). ENMs based background similarity tests indicate greater niche divergence between pairwise comparisons (2/15): E. B vs. W. b and E. A vs. E. B (Supplementary material Appendix Figure A2a, A2i). Interestingly, some pairwise comparisons (7/15) just show one-side significant divergence. The last part (6/15) niche overlap falls within the 95% confidence limits of the null distributions, leading to non-rejection of the hypothesis of retained niche conservatism.
The results of PCA-env based niche equivalency test (Figure 3d, 3e and Supplementary material Appendix Figure A3) show that pairwise comparisons of 80% (12/15) less than expected null distribution ranges and reject null hypotheses, indicating closely related clades are not equivalent to most related clades, and most clades have undergone significant alteration of their environmental niche and clades may be more resilient to climate changes than their native ranges suggest. Niche background test indicates that 11 paired comparisons show a very limited niche overlapping values (scores < 0.3). Only 4 paired comparisons show niche overlap categorized as a moderate overlap (0.307–0.527). On the other hand, the background similarity tests indicate generally little niche similarity among the pairwise comparisons in six clades (Table 3 and Supplementary material Appendix Figure A4). The ordination null tests of niche similarity show that niches are less similar than random expectations in 30% (5 paired comparison cases) in reciprocal directions, while 26% (4 paired comparisons) are significantly not to reject the null hypothesis of niche conservatism in bi-directions.