Nonlinear Thresholds of Meteorological and Chemical Determinants of Summertime Surface Ozone across China, the United States, and Europe

Abstract Surface ozone (O3) poses significant threats to human health, ecosystems, and climate. In recent years, summertime ozone pollution has remained severe in China and has shown emerging risks worldwide, creating urgent pressure for more effective early warning and region-specific mitigation. Yet thresholds of key meteorological and chemical drivers controlling surface ozone pollution remain poorly constrained, and the mechanisms underlying regional differences in dominant drivers are not fully understood. Here, we combine interpretable machine learning, Double Machine Learning, and high versus low O3 composite diagnostics to identify dominant drivers controlling surface ozone pollution, quantify region-specific thresholds, and interpret the underlying formation regimes of summertime maximum daily 8 h average ozone (MDA8 O3) across China, the United States, and Europe. Results reveal that meteorological predictors make the largest absolute SHAP contribution to summertime MDA8 O3 in most regions, whereas aerosol-related predictors rank second in many regions and first among predictor groups in the Pearl River Delta. Our approach further identifies region-specific threshold combinations associated with elevated MDA8 O3 and conditional exceedance probabilities. High O3 days consistently occur under hot, dry, weakly cloudy, low-precipitation, and strongly irradiated conditions, but the chemical and dynamical pathways differ substantially among regions.