Machine learning coupled structure mining method visualizes the impact of multiple drivers on ambient ozone

Abstract

Ground-level ozone affects Earth’s climate and human health. The formation of ozone is a complex process, depending on both atmospheric chemical processes and meteorological factors. In this study, machine learning coupled with a structure mining analysis was applied to investigate the ozone formation mechanism in Tianjin, China. The results showed isoprene has the greatest individual impact on local ozone generation, which suggests the biogenic emission of vegetation contribute significantly to native ozone pollution. The interaction between isoprene and nitrogen oxides is the strongest among precursors, with an obvious antagonistic effect between them. Reducing active volatile organic compounds is more effective for mitigating ozone pollution. Visualized network diagram also clearly illustrated the impacts of multiple drivers on ozone formation: isoprene, temperature and nitrogen oxides were the key drivers among all the influencing factors, other drivers (such as relative humidity) could assist the key drivers to collaboratively enhance or suppress ozone formation.