Abstract: Based on the standard survey data of 671 random plots of Liangshui National Nature Reserve and the landsat9 remote sensing image, three global models of Poisson model, Logistic model and Gaussian model based on least squares were established, and three local models based on the geographically weighted regression model（GWR） were established, including the geograrhically weighted Poisson（GWPR） model, the geograrhically weighted Logistic（GWLR） model and the geograrhically weighted Gaussian（GWGR） model, to predict the distribution of natural Korean pines in Liangshui National Nature Reserve. The results showed that the most significant factors affecting the distribution of natural Korean pines in this area were slope and altitude. By comparing the residual spatial correlation of the global model and the GWR model, it was found that the GWR model can produce more ideal model residuals, and the spatial correlation of the model residuals was significantly smaller than that of the global model. Therefore, the GWR model can be used to solve the problem of spatial heterogeneity between plots, which was conducive to improving the prediction accuracy of the distribution of Korean pines. Both the global model and the GWR model had good fitting effects, but the evaluation indicators of the GWR model were better than the global model, and the fitting results were better. Natural Korean pines was most distributed in the northern part of the Liangshui National Nature Reserve and the least in the middle strip area. This study can provide a theoretical basis for estimating the distribution of natural Korean pines in large-scale forest management.