Toward an Operational Machine-Learning-Based Model for Deriving the Real-Time Gapless Diurnal Cycle of Ozone Pollution in China with CLDAS Data

Abstract

An operational real-time surface ozone (O 3 ) retrieval (RT-SOR) model was developed that can provide a gapless diurnal cycle of O 3 retrievals with a spatial resolution of 6.25 km by integrating Chinese Land Data Assimilation System (CLDAS) data and multisource auxiliary information. The model robustly captures the hourly O 3 variability, with a sample-based (stationbased) cross-validation R 2 of 0.88 (0.85) and RMSE of 14.3 μg/m 3 (16.1 μg/m 3 ). An additional hindcast-validation experiment demonstrated that the generalization ability of the model is robust (R 2 = 0.75; RMSE = 21.9 μg/m 3 ). Compared with previous studies, the model performs comparably or even better at the daily scale and fills the gaps in terms of missing hourly O 3 data within the 24-hour cycle. More importantly, underpinned by the RT availability of CLDAS data, the hourly concentration of O 3 can be updated in RT, which is expected to advance our understanding of the diurnal cycle of O 3 pollution in China.