2.3 Optimization of model parameters and model building
Because the MaxEnt model is sensitive to sampling deviation and prone to overfitting, directly running the default parameters of the MaxEnt model may lead to unreliable prediction results. Therefore, the regulation magnification (RM) and feature combination (FC) parameters of the MaxEnt model were adjusted using the ENMeval data package developed by R language (Phillips et al. , 2006). The values of RM were set to 0.5, 1, 1.5, 2, 2.5, 3, 3.5, and 4. There are five types of FC: linear (L), quadratic (Q), hinge (H), product (P), and threshold (T) and it was set as L, LQ, H, LQH, LQHP, and LQHPT 6 combinations. The minimum value of AICc was selected as the optimal setting and the final model was set (Muscarella et al. , 2014).
After 647 pine wood nematode distribution points were collected and imported into the MaxEnt model software, 75% of the samples were selected as the training subset, and the remaining 25% of the samples were used to verify the model. The maximum number of iterations was set to 10,000, and the operation was repeated 10 times for modeling. The contribution rate obtained by the MaxEnt model and knife-cutting test was used to evaluate the importance of environmental factors limiting the potential geographical distribution of PWD in China. The area under the receiver operating characteristic curve (AUC) was used to evaluate the accuracy of the model. The value of AUC was 0 - 1. The closer the AUC value is to 1, the more accurate the prediction result of the model is (Yan et al. , 2021). Documents containing the prediction results of PWD were reclassified using ArcGIS 10.5. Jenks natural breaks were used (Zhao et al. , 2021). The suitable area of PWD can be divided into four levels: highly suitable habitat (values ranging from 0.50 to 1.00), moderately suitable habitat (values ranging from 0.30 to 0.50), poorly suitable habitat (values ranging from 0.10 to 0.30), and unsuitable habitat (with values < 0.10). The area of each suitable area was counted. After 0.14 was set as the threshold, a suitable grade distribution map of PWD was transformed into binary format. Based on the binary map of PWD distribution under the current and future climate change scenarios, the COGravity function in the SDMTools package of R language was used to calculate the centroid position of the highly suitable area of PWD disease under current and future climate change, and the changes in the centroid in the highly suitable area of PWD under different climate scenarios were compared.