From sensitivity regimes to policy action: Evaluating EKMA curve effectiveness with WRF-Chem, OBM, and machine learning

Despite significant progress in reducing primary pollutants in China, surface ozone pollution remains a pressing challenge. This study evaluates the effectiveness of Empirical Kinetic Modeling Approach (EKMA) curves—constructed using Weather Research and Forecasting model coupled with Chemistry (WRF-Chem), Observation-Based Model (OBM), and Empirical Kinetic Modeling Approach for Machine Learning (MLEKMA)—in guiding ozone control strategies in Luoyang. The levels of O₃ and precursors between two sampling campaigns present two main paradoxes, that reduction in source inventory do not guarantee the decrease of atmospheric NO_x and VOC levels, and reduction of precursors do not always result in O₃ decline. As indicated by all three models, O₃ sensitivity in Luoyang was in a transition regime in both 2019 and 2021. However, only the WRF-Chem model yielded ozone concentration predictions (169.0–171.0 μg/m³) that closely matched the observed values (168.8 μg/m³), while OBM and MLEKMA significantly underestimated ozone levels—the observed concentrations were 2–3 times and 3–5 times higher than the respective model predictions. Each method presents its limitations: the WRF-Chem model lacks real-time responsiveness, while OBM and MLEKMA exhibit deficiencies in accurately predicting ozone concentrations. Therefore, integrating multiple EKMA-derived methods is recommended for practical ozone management. By leveraging the timeliness of OBM and MLEKMA alongside the predictive accuracy of WRF-Chem, a rapid and robust multi-model approach for urban ozone sensitivity assessment can be achieved, enabling governments to formulate more effective ozone control policies.