Prediction of Spatial Distribution of PM2.5 Pollution Based on Machine Learning Random Forest Optimized LUR Model

With the acceleration of urbanization and industrialization in China in recent years, the adverse effects of air pollution on public health and ecological environment are becoming more and more significant, and PM2.5 concentration is regarded as an important indicator of pollution level. Land use regression (LUR) model is one of the classical methods to simulate the spatial variation of air pollution. Because the nonlinear relationship between PM2.5 concentration and its influencing factors cannot be fully explained by LUR, the optimization of LUR model by random forest algorithm can make up for the defects. In this study, the LUR model was constructed with meteorological factors, three-dimensional urban morphology, land use, terrain elevation and socioeconomic factors as input variables, and the machine learning algorithm random forest (RF) was used to optimize LUR to predict the spatial distribution characteristics of PM2.5 concentration in Fenwei Plain. The results show that the LURRF model has high simulation accuracy and superior performance with correlation coefficient at 0.98, root mean square error at 5.36 and mean absolute error at 3.46. The areas with high PM2.5 concentration are mainly concentrated in Xi’an, Weinan, Linfen, Yuncheng, Jinzhong and Luoyang. The research results can provide important scientific basis for formulating air pollution control strategies.