2.3.3. Distribution Pattern of Species Richness along Environmental Gradients

We extracted the topographic factor data of 1445 occurrence data points for 62 species of the genus *Meconopsis* in the study area, including elevation, slope, and aspect, and explored the relationship among topographic factors, longitude and latitude, and species richness. In this study, elevation was divided into 32 bands, each with an interval of 100 m, and the species richness was counted according to the altitudinal band. In terms of slope, aspect, longitude, and latitude, we divided them into 30, 36, 33, and 14 bands at 1◦ intervals, and calculated the species richness of each band in turn. The influence of environmental factors on *Meconopsis* plant species richness geospatial patterns were determined by running polynomial regression or nonlinear regression (GaussMod) models.

### 2.3.4. Landscape Fragmentation

Four landscape indices were used to measure landscape fragmentation in this paper (Table 2). Based on the reclassified land cover data grid-by-grid, the landscape indices in the grid based on the landscape scale were calculated, the spatial data were processed in ArcGIS 10.8, and the landscape indices were calculated by software Fragstats 4.2.1.


**Table 2.** Selected landscape indices and their ecological significance.

### **3. Results**

*3.1. Model Performance and Key Variables to Predict Typical Meconopsis Species*

Models for the *Meconopsis*, *M. integrifolia*, *M. horridula*, *M. racemosa*, and *M. punicea* performed better than random, with the given set of training and test data. The average AUC values for *Meconopsis*, *M. integrifolia*, *M. horridula*, *M. racemosa*, and *M. punicea* were 0.847, 0.888, 0.852, 0.914, and 0.951, respectively, indicating these five models performed well and generated very good (excellent) evaluations.

The results showed that precipitation of warmest quarter (Bio 18, 43.8%), temperature seasonality (Bio 4, 20%), elevation (13.5%) and annual mean temperature (Bio 1, 7.8%) made the greatest contributions to the distribution model for *Meconopsis* relative to other variables (Table 3), and the cumulative contributions of these factors reached values as high as 85.1%. Among the 11 environmental variables, precipitation of the warmest quarter (Bio 18, 59.2%), annual mean temperature (Bio 1, 18.9%), precipitation seasonality (Bio 15, 8.3%), and precipitation of the coldest quarter (Bio 19, 5.5%) made a greater contribution to the species distribution model for *M. integrifolia* than other environmental variables (Table 2), accounting for 91.9% of variation in total. For *M. horridula*, the most important factors were precipitation of the warmest quarter (Bio 18, 36%), elevation (11.8%), temperature seasonality (Bio 4, 11.5%), and isothermality (Bio 3, 10.2%) (Table 2); the cumulative contribution value accounted for 69.5% of the total contribution value of all environmental factors to the model. As for *M. racemosa*, precipitation of the warmest quarter (Bio 18, 37.7%), temperature seasonality (Bio 4, 24.4%), annual mean temperature (Bio 1, 16.7%), and precipitation of the coldest quarter (Bio 19, 4.4%) totally contributed 83.2% in the model (Table 2), which means these four variables contain the most significant and useful information to predict species distribution. Similarly, precipitation of the warmest quarter (Bio 18, 43.6%) had the highest contribution in the *M. punicea* model and, followed by precipitation seasonality (Bio 15, 15.8%), temperature seasonality (Bio 4, 8.6%), and elevation (8.5%) (Table 3), they totally accounted for 76.5% of the contribution value and were identified as main factors influencing the species' spatial distribution.

The thresholds (presence probability > 0.25) of the main environmental parameters were obtained from the response curve (Figure 4). The presence probability of all species is the greatest with 300–400 mm precipitation of the warmest quarter (Figure 4). In *Meconopsis*, temperature seasonality (>500) and annual mean temperature (>8 ◦C) beyond thresholds affect the habitat suitability, and the most suitable elevation was about 3700 m (Figure 4a). *M. integrifolia* had the highest probability of existence when the annual mean temperature

was about 10 ◦C and precipitation seasonality was 60–95. With the maximum probability of species occurrence when precipitation value reached 20 mm in the coldest quarter for *M. integrifolia* (Figure 4b). *M. horridula* was most likely to occur at elevations between 3200 and 4800 m, with the most suitable temperature seasonality ranging from 500 to 800, and isothermality had a sigmoid trend with the maximum species presence probability when its value up to 54 (Figure 4c). For *M. racemosa*, the optimum ranges of temperature seasonality, annual mean temperature, and precipitation of the coldest quarter were 540–690, 0–13 ◦C, and 10–30 mm, respectively (Figure 4d). The response curves of *M. punicea* indicated that precipitation seasonality, temperature seasonality and elevation followed a Gaussian shape, and they were within a certain range when the probability of species occurrence was higher (Figure 4e).

**Figure 4.** *Cont*.

**Figure 4.** *Cont*.

**Figure 4.** Relationships between key predictor variables and probability of presence of (**a**) *Meconopsis* species, (**b**) *M. integrifolia*, (**c**) *M. horridula*, (**d**) *M. racemose*, and (**e**) *M. punicea*.

**Table 3.** Relative contribution (%) of environmental variables to the MaxEnt model output for (a) *Meconopsis* species, (b) *M. integrifolia*, (c) *M. horridula*, (d) *M. racemosa*, and (e) *M. punicea*.

