Data for Paper of LEarning Surface Ozone from satellite columns (LESO): A regional daily estimation framework for surface ozone monitoring in China

The dataset is generated for depicting the surface ozone concentrations over the regions of Jing-Jin-Ji (JJJ), Yangtze River Delta (YRD), and Pearl River Delta (PRD) in China. The surface ozone concentrations were estimated based on the satellite measurements and the meteorological data using deep learning techniques. The estimation framework called LESO, which contains a Deep Forest 21 (DF21) model to interpolate ozone concentration by learning spatial patterns and a Long Short-Term Memory (LSTM) model to forecast ozone concentration by learning data from the past. The training data includes the Sentinel-5P/TROPOMI Level-3 gridded near-real-time trace gas columns (ozone, NO2, HCHO), in-situ hourly ozone measurements from China National Environmental Monitoring Center (CNEMC), and the ERA5 hourly-reanalyzed meteorological data (incl. solar radiation, fraction of cloud cover, ozone VMR, relative humidity, rainwater content, temperature, and U-/V- component of wind).