Estimation of Surface Ozone Concentration and Health Impact Assessment in China

Abstract

Within the context of PM2.5 concentrations decreasing annually, ozone concentrations have increased instead of decreased, and ozone has become one of the main pollutants in the warm season in China. Based on the idea of big data association analysis, the extreme gradient boosting (XGBoost) ozone concentration estimation model was constructed and developed to estimate the maximum daily 8 h average ozone concentration (O3_8h) in China in 2019 for human exposure assessment. The model input ground monitoring station data, high-resolution remote-sensing satellite data, meteorological data, emission inventory data, digital elevation model (DEM) data, and population data were used to capture the temporal and spatial variation of O3_8h. In this study, ten-fold cross-validation was used to evaluate the estimation performance of the model (R2=0.871, RMSE=11.7 μg·m-3). Compared to those with the random forest (RF) model and kernel ridge regression (KRR) model, due to the improvement in the algorithm itself and the advancement of parallel processing, the estimation results of the XGBoost model showed higher accuracy (RF:R2=0.864, RMSE=12.387 μg·m-3). The KRR model was as follows:R2=0.582, RMSE=23.1 μg·m-3, and the computational efficiency of the model was significantly improved. At the same time, the level of ozone exposure and the relative risk of death due to chronic obstructive pulmonary disease (COPD) in China's provinces and cities were evaluated. The results showed that the top five number of days exceeding the standard occurred in Shandong Province, Henan Province, Hebei Province, Anhui Province, and the Ningxia Hui Autonomous Region. In terms of exposure intensity, Hebei Province, Shandong Province, Shanxi Province, Tianjin City, and Jiangsu Province ranked the top five in terms of population weighted ozone concentration. In terms of health effects, the relative risk of COPD death showed seasonal changes, with the highest in summer and the lowest in winter.