2.2. Environmental and human-influence data
Many recent studies used various environmental variables representing
different hypotheses to explain the pattern species richness (e.g.,
Bailey et al. 2017; Liu et al. 2017; Shrestha et
al. 2018; Liu et al. 2018b). We used 23 predictor variables
representing 7 hypotheses (Table 1) to explain the pattern of conifer
richness. GTOPO30 digital elevation model was used to generate the
standard deviation of elevation (ELER-SD) and elevation range (ELER) in
ArcGIS 10.3.
The data of Human Influence Index (HII) was downloaded at 30 arcseconds
spatial resolution of (~ 1 km at the equator) (Wildlife
Conservation Society‐WCS 2005). MAT, MAP, MTCQ, PS, TS, and solar
radiation data obtained from WorldClim
(http://www.worldclim.org; Fick &
Hijmans 2017) at 30 arcseconds resolution. PET, AET and AI data were
downloaded from CGIAR-CSI Global database
(www.cgiar-csi.org, Trabucco et al.
2008; Fisher et al. 2011) at a spatial resolution 30 arcseconds. The
eight quantitative soil variables (SND, SLT, CRF, AW, C_stock,
C_cotent and CEC) were obtained from the ISRIC-World database
(ftp://ftp.soilgrids.org/data/aggregated; Hengl et al.2014) at depths 0-2 m and 30 arcseconds resolution. The means of depth
raster-layers were generated using ArcGIS10.3 (ESRI). Soil nitrogen and
phosphorus data were downloaded from the soil database of China
(http://globalchange.bnu.edu.cn/research/soil2d.jsp; Shangguan et
al. 2013) at the same resolution and depth and then converted from
NetCDF format to raster using the packages “Raster” and “ncdf4” in R
3.6.1 software (Hijmans et al. 2016; Pierce 2017). All the
environmental and human-influence layers were projected to “WGS 1984
UTM zone 48 N” and then their values of a 50 × 50 km2grid cells were extracted by averaging all 1 × 1 km grid cells using
zonal statistics tool in ArcGIS 10.3 (Appendix B).