How to apply O3 and PM2.5 collaborative control to practical management in China: A study based on meta-analysis and machine learning

Abstract

Constant increase of atmospheric O₃ concentration is a barrier for the further air quality improvement in China. Given that PM_2.5 is still controlled as a key pollutant, managements for the collaborative reduction of O₃ and PM_2.5 are urgently required in China. In the current work, monitoring data of O₃ and PM_2.5 from 2015 to 2016 in 1464 monitoring sites (MS) was collected and cleaned. Additionally, 7 anthropogenic emission reductions were jointed with the corresponding monitoring data. According to the O₃ and PM_2.5 variation, a meta-analysis was conducted and divided regions into 4 categories via the effect size, region I: O₃ and PM_2.5 collaborative reduction, region II: PM_2.5 decreased and O₃ increased, region III: O₃ decreased and PM_2.5 increased, regions IV: both O₃ and PM_2.5 increased. Then, based on the region labels, machine learning was used to identify the pattern between region label and its precursor reductions. The findings were as follows: (1) Principal component analysis showed that NH₃ was not focused on. (2) Random forest had a well performance on region classification with the accuracy of 80.40% and the importance of the 7 precursors was in the sequence of VOCs>NH₃ > PM_2.5 > OC > SO₂ > NO_X > Coarse PM. (3) Polytomous logistic regression evaluated the critical factors that influenced the region label, which showed that the reductions of VOCs, NH₃ and PM_2.5 could achieve the collaborative reduction in a short time in most of cities in China. Based on the statistical results above, a kinetic management system including evaluation and policy-making sections was finally established, which filled the gap of the collaborative reduction in environmental management in China.

Keywords: Air pollution prevention and control action; Collaborative pollution reduction; Machine learning; Ozone; PM(2.5).