2.1 | Species data and environmental variables
We collected data on the spatial occurrence of François’ langur and White-headed langur from the published literature between 2000 and 2019 (Huang et al., 2002; Huang & Pan, 2010; Han et al., 2011; Insua-Cao, 2012; Niu et al., 2016; Yao et al., 2016). In addition, from 2017 to 2019, we conducted field investigations in all known locations of these two species in Guangxi, Guizhou, and Chongqing Provinces, China. For each wild population we recorded the presence and spatial location of each langur species based on direct observation and fecal remains. These locations were calculated using R “SpThin” in R package excluding those locations that within 1.5 km of each other (Zhao et al., 2019). In total, 110 locations for François’ langur and 50 locations for White-headed langur were used in our analyses (Figure 2).
We downloaded 19 bioclimatic variables from CHELSA (Climatologies at high resolution for the earth’s land surface areas) (https://chelsa-climate.org; Karger et al., 2017). The bioclimatic variables for the current period (1970-2013) were simulated based on a quasi-mechanistical statistical downscaling of the ERA-interim global circulation model (GCM) (Karger et al., 2017). The bioclimatic variables used for predicting climate change in the future (2050s, average of 2041-2060) are based on MIROC5 (Model for Interdisciplinary Research on Climate) in the RCP4.5 scenario. The MIROC5 model is a subset of the CMIP5 global circulation models (GCMs). We choose RCP4.5 because it represents a moderate path, that is, it assumes that emissions peak around the year 2040 and then decline (Rehnus et al., 2018). The simulation effect of MIROC5 on extreme and summer precipitation in South Asia is more accurate than other GCMs (Hasson et al., 2016). We also downloaded 19 bioclimatic variables from the Last Interglacial (LIG, 140,000-120,000 years ago) and the Last Glacial Maximum (LGM, 21,000-19,000 years ago) from the PaleoClim data base (www.paleoclim.org; Fordham et al., 2017). These bioclimate variables are at a 30 arc-second (~1 km) spatial resolution. We also downloaded human activity variables data from the Hydes3.2 dataset, including cropland, pasture, and population density (Popd) (Goldewijk, 2016).