Model performance
The AUC values of training data and test data were 0.962 and 0.961 respectively (Fig. 2), and the performance level of the model was “excellent”.
Analysis of the importance of environmental variables
The results showed that the altitude (52.6 %, q = 0.45) and the human activity intensity (29.2 %, q = 0.39) were the two most important variables determining the distribution of FCB, which accounted for 90 % of the variations (Table 2). The annual precipitation (q = 0.37) explained 8.7 % of the contribution. The min temperature of coldest month were the variable with least impacts on FCBdistribution (0.1 %, q = 0.20).
The results of Jackknife test (Fig. 3) showed that when altitude, human activity intensity (HAI) and annual precipitation (bio12) were used to model separately, their regularized training gain were significantly higher than other variables, which indicated that they contained unique information affecting the distribution of FCB.
Relationship between environmental variables and probability of presence
In order to clarify the relationship between key environmental variables and the probability of presence of FCB, we used MaxEnt to draw the response curve using only a single environmental variable (Fig. 4). The results showed that the suitable ranges of altitude, human activity intensity, annual precipitation and were 2083.7-4081.9 m, 390.2-3825 mm, 4.8-20.1 and -7.9-8.0 ℃, respectively.
Risk detector was also used to measure the suitable range of each variable, and the results showed that the suitable ranges of altitude, human activity intensity, annual precipitation and mean temperature of coldest quarter were 2347-3612.7 m, 6.9-16.2, 566.0-912.9 mm, and -4.7-1.8 ℃, respectively.
Simulation of the geographical distribution of FCB in China under current climate condition
Fig. 5 showed the geographical distribution of FCB in China under current climate conditions predicted by MaxEnt. The results showed that the highly suitable areas were mainly located in the east of Qinghai Tibet Plateau, including western Sichuan, southeastern Tibet, southern Gansu, northwestern Yunnan and eastern Qinghai, with a total area of 31.47×104 km2, accounting for 3.26 % of China's land area. Among them, Sichuan had large areas, reaching 14.23×104 km2. The moderately suitable areas were mainly located in eastern Tibet, southern and northwestern Sichuan, northeastern Qinghai and southern Gansu, with a total area of 30.38×104 km2, accounting for 3.15 % of China's land area. Among them, Tibet and Sichuan had larger areas, which were 9.79×104 km2and 8.51×104 km2, respectively. The lowly suitable areas were located in eastern and southern Tibet, eastern and southern Qinghai, central and eastern Gansu, northwestern Sichuan, northern Yunnan and western Xinjiang, with a total area of 39.88×104 km2, accounting for 4.13 % of China's land area. Among them, the area of Tibet was the largest, reaching 17.69×104 km2.
GAP analysis of FCB in China
In order to identify the protection gaps of FCB, the highly suitable areas and the boundaries of Natural Conservation Area in China were overlapped in ArcGIS. The results showed that the highly suitable area of the medicine was 31.47×104 km2, and the area within the nature reserve was 7.13×104 km2, accounting for 22 % of the highly suitable area, indicating that there was a large protection gap of FCB, mainly in western Sichuan, eastern Tibet, southern Gansu, northwest Yunnan and eastern Qinghai (Fig. 6).
Potential distribution of FCB in China under climate change scenarios
Fig. 7 showed the changes of the suitable area of FCB in the future SSP1-2.6, SSP2-4.5 and SSP5-8.5 scenarios. Under the three climate change scenarios, the areas of the highly and poorly suitable areas of FCB showed a decreasing trend, while the areas of the moderately and total suitable areas showed a increasing trend.
By 2050s, the areas of the highly suitable areas would be reduced to 31.22×104 km2 (SSP1-2.6), 31.71×104 km2 (SSP2-4.5) and 28.88×104 km2 (SSP5-8.5), while by 2070s, the areas would be reduced to 31.81×104 km2 (SSP1-2.6), 30.59×104 km2 (SSP2-4.5) and 20.25×104 km2 (SSP5-8.5).
By 2050s, the areas of the total suitable areas would increase to 106.3×104 km2 (SSP1-2.6), 107.75×104 km2 (SSP2-4.5) and 107.81×104 km2 (SSP5-8.5). Under SSP5-8.5 scenario, compared with the current simulation results, the increase and decrease areas were the highest, the increase areas were mainly located in eastern Tibet, southern and eastern Qinghai and northwestern Xinjiang, and the decrease areas were mainly located in southeastern Gansu, southern Sichuan and northern Yunnan (Fig. 8). By 2070s, the areas would increase to 106.89×104 km2 (SSP1-2.6), 106.14×104 km2 (SSP2-4.5) and 108.9×104 km2 (SSP5-8.5). Under SSP5-8.5 scenario, the areas reduced most (10.52×104 km2), mainly distributed in Sichuan, Shandong and Shaanxi. Under SSP5-8.5 scenario, compared with the current simulation results, the increase and decrease areas were the highest, the increase areas were mainly located in southeastern Gansu, southwestern Sichuan, northern Yunnan and southeastern Tibet (Fig. 8).
Variations of the geometric center of the suitable areas under climate change scenarios
Under SSP1-2.6, the geometric center of the total suitable areas of FCB would move 96.19 km from Jiangda (Current) to northwest to Yushu (2050s), then 7.02 km to northwest to Yushu (2070s). By 2070s, the center will generally displaced 101.51 km to the northwest. Under SSP2-4.5, the geometric center of the total suitable areas of the medicine would move 117.26 km from Jiangda (Current) to northwest to Yushu (2050s), then 11.12 km to northwest to Nangqian (2070 s). By 2070s, the center will generally displaced 128.04 km to the northwest. Under SSP5-8.5, the geometric center of the total suitable areas of FCB would move 133.52 km from Jiangda (Current) to northwest to Nangqian (2050s), then 37.28 km to northwest to Nangqian (2070 s). By 2070 s, the center will generally displaced 170.72 km to the northwest (Fig. 9).