Assessing the Significance of Regional Transport in Ozone Pollution through Machine Learning: A Case Study of Hainan Island

Abstract

Regional transport of air pollutants is a serious challenge to outdoor O₃ pollution control. Characterizing the transport of air pollutants by traditional air quality models heavily relies on accurate precursor emission inventories, chemical reaction mechanisms, and meteorological factors. In this study, the pollutant concentrations of upwind cities were incorporated as features into a random forest regression model (Traj-RF) to investigate the contribution of regional transport to local O₃ pollution. Hainan island was selected as the target area in this study, due to its air quality being affected significantly by regional transport. The Traj-RF model shows good predictive performance for O₃ with a coefficient of determination (R²) of 0.68 on the independent test set based on only observed air pollutants concentrations and meteorological data. The results of the Traj-RF model show that direct O₃ transport from upwind areas contributes approximately 27.5% to the O₃ concentration in Hainan, effectively highlighting the substantial role of regional transport in Hainan’s O₃ pollution. This refined machine learning method may have the potential to assess the impact of pollutant transport on regional air quality.