Analysis of the Gridded Influencing Factors of the PM2.5 Concentration in Sichuan Province Based on a Stacked Machine Learning Model

Scientific identification of the factors that influence PM2.5 concentrations is crucial for accurate management of the regional atmosphere. In this paper, we selected boundary layer height (BLH), a low vegetation cover index (CVL), atmospheric pressure (PS), temperature (TEMP), rainfall (RAIN), a high vegetation cover index (CVH), wind speed (WS), relative humidity (RH), land use data (LUCC), normalized vegetation index (NDVI), a digital elevation model (DEM) and a nighttime light index (NL) as influencing factors and used a 5 x 5 km grid as the evaluation unit to analyze the degree of influence of different influencing factors on the change in PM2.5 concentrations on a monthly time scale based on a stacked machine learning model. The results reveal that (i) compared with a convolutional neural network (CNN), a deep belief network (DBN), adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost) and random forest (RF), the stacked machine learning model has a better ability to simulate the relationship between PM2.5 concentrations and influencing factors, with higher accuracy and a lower error rate. (ii) There was significant spatial autocorrelation and spatial heterogeneity of the PM2.5 concentration in Sichuan Province from October 2015 to September 2020, and the regional and clustering characteristics of air pollution are obvious, with PM2.5 concentrations mainly showing high-high clustering and low-low clustering. (iii) The key influencing factors of PM2.5 concentrations differ by month. Therefore, identifying the key influencing factors of PM2.5 concentrations in Sichuan Province in different months is crucial for the accurate management of air quality. (iv) The influence of interacting factors on PM2.5 concentrations varied by month. Therefore, the identification of key interaction factors influencing PM2.5 concentrations in Sichuan Province on a monthly time scale can provide some scientific evidence to support PM2.5 pollution prevention and control.