Divergent Ozone Predictions in China Under Carbon Neutrality: Why Chemical Mechanisms Disagree

Uncertainty in air quality models can lead to divergent assessments of emission control policies. Here, we investigate why two widely used chemical mechanisms in the Weather Research and Forecasting model with Chemistry (WRF-Chem) predict inconsistent ozone levels and conflicting responses to emission reductions over major city clusters of China. By combining process analysis with an explainable machine learning technique, we reveal that these discrepancies primarily stem from differences in the rates of ozone-forming and -suppressing reactions involving hydroperoxy (HO 2 ) and organic peroxy (RO 2 ) radicals between the two mechanisms. This thereby underscores the need for a more accurate depiction of volatile organic compounds reactivity in models. We further quantify the impact of these discrepancies by projecting ozone levels across China from 2030 to 2060 under the carbon neutrality emission reduction scenario. Divergences peak in 2030, with the two mechanisms disagreeing on whether ozone mitigation in city clusters is achievable. Over time, their predictions begin to converge. By 2060, both mechanisms agree that nearly the entire Chinese population will experience reduced ozone levels, and support the continued reduction of nitrogen oxides (NO x ) emissions as an effective strategy for curbing ozone pollutions. However, significant differences persist in the magnitude of reductions, with one mechanism projecting greater policy efficacy. Continued efforts are therefore required to further reduce the model uncertainty.