Future evolution of Chinese near-surface ozone concentrations: the insight from a new two-stage model combining machine learning and chemical transport modeling

Near-surface ozone pollution is one of the biggest challenges for Chinese air quality improvement, while its future spatiotemporal evolution and driving factors have not been fully investigated. Here, we developed a two-stage model combining a machine learning technique (XGBoost) and a chemical transport model (WRF-CMAQ) to assess the ozone change till 2060 in China under three scenarios with various trajectories of climate change, energy transition and pollution controls. The new model effectively corrected overestimation and underestimation of ozone levels by CMAQ and global climate models, respectively. Anthropogenic efforts will overcome the adverse effect of climate and reduce future ozone concentration, especially in eastern China and warm season with greater ozone pollution. From a long-term perspective, energy structure transition was estimated to play a more important role than end-of-pipe emission controls, with a former to latter ratio of ozone reduction during 2017-2060 at 2.7. With observational information incorporated, our model was demonstrated to better capture the ozone response to precursor emission change than WRF-CMAQ, and corrected the underestimation of ozone reduction for developed urban areas.