*2.3. Methods*

### 2.3.1. MaxEnt Modeling

MaxEnt is a popular open-access species distribution model. It employs the maximum entropy algorithm and species occurrence points to predict the probability of species occurrence in potential distribution areas. In the study, we used MaxEnt 3.4.4 to model the current habitat distribution of typical *Meconopsis* species and the whole genus. Species occurrence data were divided into two data sets, 75% of total occurrence records were used for training the model, while the remaining 25% were used for validating the model. The number of iterations to run the algorithm is set to 1000 and the model was repeated for 10 runs, which allows adequate time for the model to converge [59], and the rest of the settings were the default settings.

Receiver operating characteristic (ROC) analysis was used to calibrate and validate the robustness of the evaluation of the MaxEnt model [58]. The area under the curve (AUC) of the ROC curve was used to estimate the model performance, which varies from 0 to 1 and is a diagnostic evaluation index with high recognition at present. Generally, the model performance can be categorized according to the value of AUC as fair (0.6–0.7), good (0.7–0.8), very good (0.8–0.9), and excellent (>0.9) [58]. A jackknife test was applied to identify variables that made a significant contribution to model output. The results of this test showed the training gains of each variable when the model was run in isolation and was compared with the training gains of all variables. Additionally, species response curves were created to explore the relationships between the probability of occurrence and environmental variables.

A threshold value of 25% was set based on the logistic output from the Maxent model, below which the probability of species occurring can be considered almost negligible. The final suitable habitat predictions were regrouped into three classes: low suitable (25–50%), medium suitable (50–75%), high suitable (>75%), and the values below 25% as nonsuitable.

### 2.3.2. Spatial Pattern Distribution of Species Richness

A total of 62 *Meconopsis* species were included in this study, which represents the species richness of *Meconopsis* genus in the study area. The species distribution pattern was affected by different spatial scales or areas, and to eliminate their influence, when calculating the species richness, the fishnet tool was used to divide the study area according to the equal area grid of 50 km × 50 km, and a total of 1314 grids were obtained. Then, the species distribution points of *Meconopsis* were associated with the grids, and the number of species in each equal-area grid was counted based on the actual recorded points of *Meconopsis*, and this value was taken as the species richness value of each grid. Finally, the species richness was divided into 15 levels using the classification function of ArcGIS 10.8, and the spatial pattern of species richness was visualized.
