Quantitative Estimation of the Impacts of Precursor Emissions on Surface O 3 and PM 2.5 Collaborative Pollution in Three Typical Regions of China via Multi-Task Learning

The coordinated control of PM 2.5 and O 3 pollution has become a critical factor restricting the improvement of air quality in China. In this work, precursors and related influencing factors were utilized to establish PM 2.5 and O 3 estimation models in the North China Plain (NCP), the Yangzi River Delta (YRD), and the Pearl River Delta (PRD) using a multi-task-learning (MTL) model. The prediction accuracy of these three MTL models was high, with R 2 values ranging from 0.69 to 0.83. Subsequently, these MTL models were used to quantitatively reveal the relative importance of each factor to PM 2.5 and O 3 collaborative pollution simultaneously. Precursors and meteorological factors were the two most critical influencing factors for PM 2.5 and O 3 pollution in three regions, with their relative importance values larger than 29.99% and 15.89%, respectively. Furthermore, these models were used to reveal the response of PM 2.5 and O 3 to each precursor in each region. In the NCP and the YRD, the two most important precursors of PM 2.5 pollution are SO 2 and HCHO, while the two most critical factors for O 3 pollution are HCHO and NO 2 . Therefore, SO 2 and VOC emissions reduction is the most important measure for PM 2.5 pollution, while VOC and NO 2 emission reduction is the most critical measure for O 3 pollution in these two regions. In terms of the PRD, SO 2 and NO 2 are the most important precursors of PM 2.5 pollution, while the most important precursors for O 3 pollution are HCHO and SO X , respectively. Thus, NO 2 , SO 2 , and VOC emission reduction is the most critical measure for PM 2.5 pollution, while VOC and NO 2 emission reduction is the most critical measure for O 3 pollution in the PRD. Overall, this study provides clues and references for the control of PM 2.5 and O 3 collaborative pollution in the NCP, the YRD, and the PRD.