Climatic suitability for Bsal
To understand how environmental conditions affect the distribution ofBsal , we used Generalized Linear Modeling (GLM) and Maximum
Entropy Modeling (MaxEnt) to estimate its habitat suitability (Breiner
et al., 2015; Phillips et al., 2006). If new positive locations ofBsal were detected in our study, they were to be combined
with the previous verified presence records in native Asia to build the
niche model. Previous present records of Bsal for Thailand,
Vietnam, China and Japan in Asia (n=34), were attained from Basanta et
al. (2019), Beukema et al. (2018), Laking et al. (2017), Lötters et al.
(2020), and Yuan et al. (2018). We used the ecoregions (Dinerstein et
al., 2017) of these Bsal occurrences as background to improve
model calibration. Nineteen bioclimatic variables representing climatic
conditions were download from WorldClim version 2.1 (Fick & Hijmans,
2017) as environmental predictors, at the spatial resolution of 30
seconds (~1 km2). To exclude the effects of high
collinearity between predictors, we used the bioclimatic variables with
Pearson’s r < 0.7, where the correlation was calculated in the
ENMTools (Warren et al., 2010). The final predictor set included 6
variables: Mean Diurnal Range (BIO2), Maximum Temperature of Warmest
Month (BIO5), Temperature Annual Range (BIO7), Annual Precipitation
(BIO12), Precipitation of Warmest Quarter (BIO18), and Precipitation of
Coldest Quarter (BIO19). To trim duplicate observation records, a single
coordinate for each grid cell was retained in our model predictions. The
model performance for Bsal was evaluated using the AUC (Lobo et
al., 2008), which was calculated by splitting the training (70%) and
testing (30%) observations. The final ensembled model contained ten
replications for each modeling, and were run in “sdm” package (Naimi
& Araújo, 2016).