2.2 Environmental variables
A total of 22 variables were studied including 19 climatic variables, 3
topographical variables, and 22 initial environmental variables with a
resolution of 2.5’ (Table S1). The three topographic parameters (alt,
slope, and asp) were downloaded with a resolution of 2.5’ from World
Climate data (http://www.worldclim.org/). Bioclimatic characteristics
obtained from the WorldClim (version 2.0) database were utilized to
build distribution prediction models (11). The predictions for the
Geographical distributions of S. ciliatum , S. nepalense ,
and S. yunnanense were carried out by the BCC-CSM2-MR climate
system model, developed by the National Climate Center (12). The World
Climate Database (http://www.worldclim.org/ ) was used for
downloading the future climate data 2050 (2041–2060) and 2070
(2061–2080). New future pathways based on socioeconomic assumptions are
called Shared Socioeconomic Pathways (SSPs) that describe various levels
of socioeconomic development. These were used in the study (13). The
SSPs include the high-forcing scenario (SSP5-8.5), the medium-forcing
scenario (SSP2-4.5), and the low-forcing scenario (SSP1-2.6). The high
and low emission scenarios are represented by the SSP5-8.5 and SSP1-2.6
in the study , respectively.
It is necessary to analyze environmental variables before they can be
used for niche simulation calculations to avoid the multicollinearity of
the variables causing over-fitting of the model because many bioclimatic
variables are spatially related (14). ArcGIS10.2 was employed to examine
the relationships between these 22 bioclimatic variables (Table S1).
When the correlation coefficient between two climatic variables was more
than 0.80, climate variables of greater ecological relevance were kept.
For the prediction of the distribution of S. ciliatum, S.
nepalense, and S. yunnanense , a total of 7, 9, and 8 variables
with coefficients greater than 0.8 (15) were maintained, respectively
(Fig. 2).