Spatiotemporal high-resolution (daily, 1km) ground-level ozone (O3) dataset across China from 2000 to present

We developed a full-coverage, high-resolution estimation model using the XGBoost algorithm to generate daily ground-level ozone concentration data at a 1-km spatial resolution, with satellite LST serving as the primary predictor. Tested on 2019 samples from China, our model demonstrated strong predictive accuracy, achieving an R² of 0.91 and an RMSE of 13.51 μg/m³ through 10-fold cross-validation. It effectively captured short-term local variations in ozone levels. We then applied this method to estimate long-term spatiotemporal high-resolution ozone concentrations from 2000 to 2020. The 21-yr monthly data and 2019 daily are archieved in NetCDF format in Version 1. We used the same method to generate daily, 1-km O3 across China during 2020-2022 and archieved here. The random 10-fold cross-validation results at the daily scale are shown in the table below: Year R2 slope intercept RMSE (ug/m3) MAE (ug/m3) 2020 0.92 0.91 8.14 11.63 8.38 2021 0.92 0.91 8.57 11.41 8.23 2022 0.93 0.92 7.72 11.39 8.15 If you want to use this dataset, please cite the following publication. --He, Q., Cao, J., Saide, P. E., Ye, T., & Wang, W. (2024). Unraveling the Influence of Satellite-Observed Land Surface Temperature on High-Resolution Mapping of Ground-Level Ozone Using Interpretable Machine Learning. Environmental Science & Technology. [url] --He, Q., Cao, J., Saide, P.E., Ye, T., Wang, W., Zhang, M. and Huang, J., 2025. Evaluating spatiotemporal variations and exposure risk of ground-level ozone concentrations across China from 2000 to 2020 using high-resolution satellite-derived data. Atmospheric Chemistry and Physics, 25(13), pp.6663-6677.[url] We are going to archive the daily dataset between 2000 to 2020 and it will be published here! If you want to use the daily data now, please contact us via [e-mail]. We also share other atmopsheric reconstruction datasets: For full-coverage, 1-km, AOD data in China, please go to harvard dataverse. This dataset was imputed based on MODIS MAIAC 1-km AOD retrievals. For full-coverage, 1-km, CO2 data in China, please go to 10.5281/zenodo.10022904. This dataset was reconstructed based on OCO-2 XCO2 retrievals and machine learning algorithm. For full-covereage, 1-km, PM2.5 data in China, please go to 10.5281/zenodo.8437234 or 10.5281/zenodo.8347128. If you have any questions or comments, please feel free to contact us via [e-mail]