Extinction risk assessment
Population growth and temperature increases in China have been highly correlated in the last few hundred years, which leads to uncertainty regarding which variable plays a stronger role in local extinctions. In total, 604 occurrence records (87% of the historical observations) of Chinese pangolin were documented in 1970-2000 across China, and more detailed climatic data are available for this period. Therefore, we lowered the timescale and combined 19 climatic variables, anthropogenic variables from HYDE, elevation and identified extinction records of Chinese pangolin to construct a model to assess extinction risk with MaxEnt. We compared those occurrences with the current distribution range of Chinese pangolins assessed by the IUCN expert group and identified 94 occurrences outside the distribution area (IUCN, 2019; Fig. 2). We collected 162 rescue and observation records of Chinese pangolins during 2000-2020 from the wildlife recuse departments, news reports and GBIF database in China. We set up circular buffer zones (r=50 km) according to extant occurrences of Chinese pangolin to exclude the potential distribution area (Fig. 2). Considering the rapid development of wildlife monitoring technology and emphasis on biodiversity conservation of the Chinese government, 220 occurrences out of buffer zones were supposed to be extinct (Fig. 2). In total, 314 occurrence records out of distribution range and buffer zones were considered extinction locations (Fig. 2). Those occurrences were dispersed (no highly spatial autocorrelation) based on the analysis of Ripley’s K function. The elevation data were derived from the SRTM. Higher-resolution and multidimensional climate data from 1970-2000 were available from WorldClim (download fromhttps://worldclim.org/). These biological variables are related to various aspects of temperature and precipitation affecting the geographic distribution of Chinese pangolins and their prey (mainly ants and termites).
The resolution of the environmental variables was uniform at 5 arc minutes, and each grid retained only one occurrence record. We input extinction records and environmental variables into MaxEnt (version 3.4.1) (Phillips et al., 2006) and ran the Model 25 iterations (preexperiment) to exclude insignificant variables with 0% contribution and 0 permutation importance value. To avoid multicollinearity, we calculated the Pearson correlation coefficient (r) between variables; when r >0.7, the variable with the lower contribution rate was discarded. Finally, eight variables, including population density, elevation, cropland, grazing, temperature seasonality (bio4), mean temperature of the driest quarter (bio9), precipitation seasonality (bio15) and precipitation of the warmest quarter (bio19), were used to construct the model.
We ran the Algorithm 100 times, and the average of the predicted results was output in logistic. Maximum training sensitivity plus specificity was used as the threshold value to distinguish extinction, and Nature Breaks methods were used to further assess the levels of extinction risk. Finally, we extracted the extinction risk of extant populations of Chinese pangolins according to the risk map.