2.2 Environmental variables and study
area
We used 19 bioclimatic variables with a resolution of 2.5 minutes
obtained from WORLDCLIM as predictors
(www.worldclim.org). Although
there was low spatial autocorrelation and dependence of points in
environmental space (Olivier Broennimann et al., 2012; Guisan &
Zimmermann, 2000), to compare the relationship between environmental
factors and species, we used consistent variables used in two different
space. Pearson correlation analysis was used to remove the data of 19
environmental layers by correlation value | r |
> 0.7, considering Asian openbill habits, 10 environmental
variables were retained bio1, bio3, bio8, bio9, bio10, bio12, bio15,
bio17, bio18, and bio19. These nonlinear data were then converted into
niche quantification analysis through principal component analysis. We
used the first three principal components, which summarized 84.8% of
the overall variation in the environmental data, to represent
environmental variability across the landscape for model analysis. PC1
was mainly composed of precipitation factors related to annual average
temperatures and low temperatures, PC2 mainly represented changes in low
temperatures and temperature differences, and PC3 was related to annual
precipitation and temperature in the rainy season (see Table S1).
The extent of the model calibration area had key impacts on the ENM
results (Barve et al., 2011). Considering the distribution status,
dispersal trends, dispersal ability, and ecological habits of the Asian
openbill, we decided to use four different study areas. 1) The native
range consisted of 9 administrative borders of India, Nepal, Sri Lanka,
Bangladesh, Myanmar, Vietnam, Laos, Thailand, and Cambodia. 2) For the
new distribution areas, we delimited the administrative boundaries of
Yunnan, Guangxi and Guizhou provinces of China as the new northern
range, while we chose the geographical division of the Malay Peninsula
as the new southern range. 3) The calibration range, based on the native
range, added accessible areas of the species including southwest China,
the Malay Peninsula, Sumatra Island, and Pakistan. 4) The modeling
prediction extent was comprised of the boundaries of east Asia,
southeast Asia, South Asia, and Australia (Fig.1). The native range and
new distribution range were used for comparative ecological niche
analysis in environmental space, while the calibration range and project
range were used for modeling and prediction in geographical space.