2.3.3. Soil Property Data

Soil data used in this study were collected from Harmonized World Soil Database (HWSD, http://www.fao.org/soils-portal, accessed on 15 November 2021).

### 2.3.4. Processing and Selection of Environment Variables

All environment variables were resampled to 30" spatial resolution and were processed to the same geographic bounds. In the modeling process, high correlation variables and environment variables that contribute less to the model were removed to improve the accuracy of the model [28,29]. The correlation coefficient was calculated to account for the influence of collinearity on the model accuracy. The variables with r below 0.8 were selected [30]. The 16 variables with low correlation coefficients and high contribution rates were selected for distribution modeling (Table 1).

**Table 1.** The selection of environmental variables used in this study.



**Table 1.** *Cont.*

### *2.4. Distribution Modeling*

MaxEnt with advantages in performance and stability was used to predict the potential distribution of three species (Figure 2) [11,31]. In addition, MaxEnt has the advantage of utilizing continuous and classified data and integrating the interaction between variables [14]. MaxEnt software version 3.4.4k was used to identify the species potential habitat distribution. The MaxEnt was set to run 500 iterations with a maximum of 10,000 background points, a convergence threshold (0.00001), a regularization multiplier of 1, a logistic output grid format, and the algorithm parameters set to "auto feature". The other parameter values were kept in the default settings [32]. A total of 70% of the distribution point data were selected for training, and the rest were used for testing [33]. The Jackknife was used for testing the importance of environmental variables in a model with a small amount of the distribution point records [34].

### 2.4.1. Accuracy Assessment

The value of the area under the receiver operating characteristic curve (AUC) was selected to assess model accuracy [11]). Model performance can be regarded as fail when it is between 0.5 and 0.6, poor when it is between 0.6 and 0.7, fair when it is between 0.7 and 0.8, good when it is between 0.8 and 0.9; and excellent when it is between 0.9 and 1 [35].

### 2.4.2. The Area and Elevation Changes of the Habitat Suitability

SDMtoolbox of ArcGIS 10.7 was used to convert the current and future results. The asc format files in the model result were converted to the raster format and reclassified into four suitable habitats, and we calculated the area and average altitude of potential distribution by zonal statistic tool [36]. The area and average altitude changes in the suitable habitat for species distribution were used as an indicator to evaluate the impact of climate change on the distribution of species [8]. The intersection distributions of the three species were obtained through the raster calculator and extracted by attributes tools.

**Figure 2.** Flow diagram of methodology adopted.
