Diurnal variations of ozone formation sensitivity in southeastern China based on GEMS: Insights from gridded thresholds and machine learning.

Tropospheric ozone (O3) formation is often limited by NOx, VOCs, or both, necessitating detailed understanding of its spatiotemporal sensitivity for effective control. This study used hourly Geostationary Environment Monitoring Spectrometer (GEMS) data to investigate the diurnal variations of O3 formation sensitivity over Fujian Province, Southeast China, during the 2024 warm season. GEMS observations reveal pronounced diurnal and spatial heterogeneity in O3 precursor indicators. Gridded (0.25°) HCHO/NO2 thresholds for regime transitions were derived from a steady-state O3-NOx-VOC model with GEOS Composition Forecast (GEOS-CF) simulations. The NOx-Limited transition threshold () declines notably from 08:45 to 13:45 LT and is substantially higher in coastal than inland regions. eXtreme Gradient Boosting (XGBoost) machine learning with Shapley Additive Explanations (SHAP) analysis indicates that is dominantly governed by VOC reactivity (VOCR) across scales. After applying the gridded thresholds, we find that warm-season O3 production across Fujian is predominantly NOx-Limited, with the province-wide fraction rising from ∼71% at 08:45 LT to 92% at 14:45 LT. Inland areas remain NOx-Limited all day, whereas coastal cities display morning VOC-Limited conditions that shrink toward urban cores-many of which stay VOC-Limited during peak photochemistry hours. Thus, in high-emission coastal urban centers, NOx-only controls may be ineffective or counterproductive, requiring prioritized VOC reductions. In contrast, NOx abatement remains a critical priority for inland regions. This work highlights the unique capability of GEMS to resolve diurnal precursor variability and demonstrates that spatially and temporally differentiated control strategies are essential for effective O3 mitigation in Fujian and similar coastal regions.