Optimization of Hourly PM_2.5 Inversion Model Integrating Upper-Air Meteorological Elements

Abstract

PM2.5 is directly related to the air quality and poses a threat to human health, thus high-precision monitoring of PM2.5 is necessary. The dispersion and accumulation of PM2.5 are affected by meteorological elements near the ground and in the upper air. Nevertheless, the current PM2.5 inversion models based on the deep neural network only consider ground elements. To further optimize the model and improve the inversion accuracy, a PM2.5 hourly inversion model integrating upper-air meteorological data was proposed, and a parametric rectified linear unit was used as the activation function of the model. The results showed that among the input elements, PM2.5 had the highest correlation with AOD, with a correlation coefficient of 0.33. The proposed model achieved the highest accuracy on the test set, with RMSE, MAE, and $\text{R}^{2}$ of $14.39 \mu \text{g}/\text{m}^{3}$ , $9.67 \mu \text{g}/\text{m}^{3}$ , and 0.83, respectively. Compared with the deep neural network models for surface meteorological data and surface+850hPa meteorological data, the RMSE of the proposed model on the test set was reduced by 23.13% and 17.05%, respectively. Meanwhile, the RMSE of the proposed model on the test set was reduced by 56.15%, 39.99%, 14.60% and 5.76%, respectively, compared with adaptive boosting, gradient boosting regression, random forest, and the integrated model of these three models. During the heating season in Shanxi Province, the high-value areas of PM2.5 were mainly distributed in the basin area, the PM2.5 concentration reached the highest in November and peaked at 11 a.m. during the day.

Keywords

• Aerosol optical depth
• air pollution
• deep neural network
• PM2.5
• remote sensing