Evaluating the main drivers of ozone pollution in a typical city of the Yangtze River Delta based on machine learning

Abstract

Ozone (O3) pollution has become more and more serious in recent years, especially in economically developed areas, such as the Yangtze River Delta region. Surface O3 is nonlinear affected by meteorological conditions, emissions of precursors and chemical reactions. Changes in atmospheric pollutants concentrations caused by emission reduction policies can be obscured by effects of dispersion and transport. Distinguish meteorological dispersion and transport from local influences is important for government to implement policy to control O3 pollution. Meteorological normalization is a technique to remove the influence of meteorological factors in the time series of pollutants concentrations. In order to better understand the local impacts on O3 production, we used random forest with meteorological normalization to remove the dispersion or transport on the concentrations of O3 and its precursors based on long-term observation data of air pollutants and meteorological parameters in Hangzhou from 2019 to 2022. After removing these meteorological effects, we found that the trend of O3 concentration was basically consistent with the observed concentration, indicating that the change of O3 in Hangzhou was mainly driven by precursor emission and chemical formation. The contribution of dispersion and transport to O3 concentration was positive in 2019 (5.2%) and 2021 (4.4%), and negative in 2022 (3.8%). Comparing the PMF results of observation and meteorological normalization, the contributions of vehicle emission and combustion to VOCs during the pollution period were underestimated, while the contributions of industry and solvent use were overestimated. Our results demonstrate that the main factors affecting local O3 generation after decoupling the effects of meteorological dispersion and transport. This work is helpful to scientifically formulate control measures of atmospheric pollution, and effectively assess the actual benefits of emission reduction policies.