Prediction of PM2.5 concentration in Guangxi region, China based on MLR-ARIMA

Abstract

In recent years, the problem of atmospheric pollution has received more and more attention. Combining the concentration data of various air pollutants monitored by the air quality monitoring stations in Nanning, Guilin, and Baise in Guangxi province in 2017 and the precipitable water vapor (PWV) data obtained by sounding stations in the three cities, analyzed the changes of PM2.5 and PWV in major cities in Guangxi and build the multiple linear regression-differential autoregressive moving average (MLR-ARIMA) models respectively make short-term predictions for the changes in PM2.5 concentration in the three cities. Among them, the mean absolute error (MAE) of the prediction results of Nanning, Guilin and Baise are 7.57μg/m3, 12.75μg/m3 and 7.67μg/m3, compared with the multivariate linear regression model and the neural network model, the prediction accuracy of this model in Nanning is 43.55% and 46.50% higher than that of the multiple linear regression model and neural network model, respectively, and in Baise is 21.41% and 26.32% higher accordingly, The model prediction effect in Guilin is optimal for the neural network model, which improves 24.46% and 11.84% compared with MLR and MLR-ARIMA models, respectively, where MLR-ARIMA model still has 14.31% accuracy improvement compared with MLR model. This study has some reference value for PM2.5 prediction work in major cities in Guangxi, China.