We collected hourly ground ozone observations (more than 6 million records), meteorological data, remote sensing products, and social-economic information, and applied the Long Short-Term Memory (LSTM) recurrent neural networks to map surface ozone concentrations data (HrSOD) at an hourly time-step and a 0.1°× 0.1° resolution across China during 2005-2020.The gridded ozone concentration data are provided in NetCDF format in ppb. The daily data is a NetCDF file and the file is named "YYYYMMDD.nc", where "YYYY", "MM" and "DD'' refer to the year, month and day of the file.