• An ensemble machine learning framework was developed to robustly project future ozone by integrating multi-model climate data and emission data. • Emission reduction is the dominant driver for the decreasing ozone trend, contributing up to a 5.8 ppbv reduction at the national level. • Climate impact exhibits a pronounced north–south divergence, with slight decrease in the North but significant increases in the South. • Industrial sources currently drive ozone pollution, but residential and transportation sectors show increasing importance in the future. Despite intensive air pollution control policies in the past decade, ground-level ozone pollution in China has remained persistently high. Future ozone levels remain highly uncertain due to the complex interplay between climate change and anthropogenic emissions under carbon neutrality goals. Current chemical transport models driven by Shared Socioeconomic Pathways often rely on meteorological simulation from a single climate model, posing great uncertainty to the ozone predictions. In this study, we developed an ensemble machine learning framework integrating CMIP6 multi-model meteorological fields, current and future emissions inventories, and WRF-CMAQ simulations for ozone projection, supplemented by a weather normalization module to quantitatively distinguish the respective contributions from long-term climate and anthropogenic emission variations. Using WRF-CMAQ simulations under a suite of reduced-emission scenarios as training labels, our ensemble framework addressed the challenge of predicting ozone under future low-emission conditions. Demonstrating a robust modelling performance ( R 2 = 0.83), projections for 2060 (relative to 2020) show that the national average ozone will decrease by 1.9 ppbv, 2.7 ppbv, and 4.0 ppbv under SSP2-4.5-ECP, SSP5-8.5-BHE, and SSP1-2.6-BHE scenario, respectively. Emission-driven ozone (EDO) decreases dominate the national trend, with reductions up to 5.8 ppbv by 2060. In contrast, climate-driven ozone (CDO) shows a sharp north–south contrast: southern China experiences increase of 5.8–8.4 ppbv due to enhanced solar radiation and lower humidity, whereas northern China sees decreases of −1.2 to −2.6 ppbv by 2060. Our ensemble multi-model analysis reveals a sharply divergent ozone future across China, demanding region-specific strategies to address the climate penalty effect.
Zhang et al. (Wed,) studied this question.