Long-term reconstruction of NO2 photolysis rate coefficients using machine learning and its impact on secondary pollution: A case study in a megacity of the Sichuan Basin, China

The NO₂ photolysis rate coefficient (J_NO2) is a critical parameter for assessing the intensity of atmospheric photochemical reactions. However, continuous long-term measurements of J_NO2 are scarce. In this study, we developed a machine learning-based method to reconstruct hourly J_NO2 values, applying it to a megacity in the Sichuan Basin from 2015 to 2023. The model exhibited strong performance with cross-validation R² = 0.854 and RMSE = 8.15 × 10⁻⁴ s⁻¹. Utilizing the Shapley Additive Explanations (SHAP) method, we identified solar activity and pollutant levels both as significant predictors for J_NO2. Our long-term J_NO2 reconstructions indicate a strong correlation between J_NO2 and ozone concentration, highlighting its important role in secondary pollution. This study illustrates the effectiveness of machine learning in reconstructing hourly J_NO2 values, providing a valuable enhancement to traditional models. The findings are crucial for understanding regional photochemical processes and for analyzing trends and causes of ozone and aerosol pollution.