Reanalysis-based retrieval of near-surface ozone concentrations and its implications for ozone pollution events across China

Abstract

Ozone (O₃) pollution has become a serious environmental issue in China, posing a significant threat to human health. However, existing satellite-based surface ozone concentration retrieval products are often restricted by data gaps and low spatial resolution, which limit their accuracy and applicability. Therefore, this study integrated reanalysis data from the National Centers for Environmental Prediction (NCEP) and ground-based observations, employing the XGBoost machine learning algorithm to estimate near-surface ozone concentrations across China at a high spatial resolution of 1 km. The model performed excellently in sample-based 10-fold cross-validation, achieving an R² of 0.91 and an RMSE of 13 μg/m³. Case studies of ozone pollution in two typical industrial cities (Tianjin and Dongguan) revealed the spatial heterogeneity of ozone pollution, supporting fine-scale regional monitoring. With the advantages of high spatial resolution and full coverage, our dataset successfully captured a prolonged ozone exposure event in eastern China from 12 June to July 15, 2023. This dataset holds great potential for applications in ozone exposure assessment and environmental impact studies.