The Study on Spatial Distribution and Spatial Model Establishing Based on GWR of Carbon Content of Individual Tree for Pinus kesiya var. langbianensis in Natural Forest

The artical takes individual tree from the Pinus kesiya var. langbianensis natural forest survey in Lancang County, Yunnan Province as the research object, the carbon content of different organs of the analytical trees was measured, and the branches, leaves, trunks, and bark of the individual P. kesiya var. langbianensis were constructed using a geographically weighted regression model. The topographic factor and dendrometric factor model of the carbon content of the root and the whole tree, and compare the least square model to analyze the pros and cons of the model. The results show that there is spatial autocorrelation in the carbon content of each organ of P. kesiya var. langbianensis within 250 m. Therefore, when studying the carbon content of P. kesiya var. langbianensis, the influence of spatial autocorrelation cannot be ignored. The semivariogram can quantitatively analyze the spatial heterogeneity of the carbon content of each organ of P. kesiya var. langbianensis. The spatial heterogeneity in the 200 m range is mainly caused by the random part. When studying similar problems, appropriate research should be determined scale. The geographically weighted regression model has a good fitting effect. The coefficient of determination of the carbon content model in each dimension is above 0.8. Although the AIC of the carbon content model for some organs is slightly larger than that of the OLS, the RMSE of geographically weighted regression model is less than that of the least squares model. The residual error of the geographically weighted regression model is also less than that of the least squares model. Obviously, the model accuracy is better than that of the least squares model. At the same time, it shows that the geographically weighted regression model can effectively explain the spatial heterogeneity that the least squares model cannot explain.