Morphological data
These data are derived from 151 specimens from our recent fieldworks and Herpetological Museum of the Chengdu Institute of Biology, CAS. The sample size varied between 8 and 42 specimens per clade (only male adults included), with a mean of 25 individuals per clade. The morphological variables including: snout–vent length (SVL), head length (HL), head width (HW), snout length (SL); inter-nasal space (INS), width of upper eyelid (UEW), inter-orbital space (IOS), diameter of eye (ED), length of lower arm and hand (LAHL), diameter of lower arm (LAD), hind-limb length (HLL), tibia length (TL), tibia width (TW), length of foot and tarsus (TFL) and foot length (FL), named system of morphological variables following Fei et al. (2005). Males within S. boulengeri predominant outstandingly in two pairs of keratinized spine patches on the chests and fingers of males and in variation adult body size and diameter of lower arm between them (Fu et al. 2007, Chen et al. 2009). In data analysis, we removed females for their insufficient quantity. Prior to all statistical analyses, the variables were log-transformed to better meet the requirements of normality and homogeneity (Rabosky and Adams 2012).
PCMs and trait correlative analyses
Biologists have long recognized that closely related species are generally more similar to one another than they are to more distantly related, which is often termed phylogenetic conservatism (Martin and Hansen 1997). Phylogenetic signal can be thought of as the degree to which similarity in trait values between species can be predicted upon their relatedness (Harvey and Rambaut 2000). If there is phylogenetic signal in the data, then PCMs are necessary for robust statistical analyses of trait correlations. To address the relationships between morphology and climate variables caused by niche evolution, we used PGLS (Grafen 1989, Martins and Hansen 1997).