Surface Ozone Estimation over the Beijing–Tianjin–Hebei Region: A Case Study Using EMI-II Total Ozone Observations and Machine Learning Integration

Surface ozone monitoring remains challenging due to sparse ground networks and limited satellite boundary-layer sensitivity. This study evaluates, for the first time, China’s Environmental Trace Gases Monitoring Instrument II (EMI-II) for estimating surface ozone over the Beijing–Tianjin–Hebei (BTH) region. EMI-II total ozone columns (TOCs) are retrieved using the differential optical absorption spectroscopy (DOAS) algorithm and validated against the TROPOspheric Monitoring Instrument (TROPOMI) (R = 0.96), Geostationary Environment Monitoring Spectrometer (GEMS) (R = 0.97), and the World Ozone and Ultraviolet Radiation Data Centre (WOUDC) ground measurements (R > 0.92, bias < 4%). TOCs are then combined with ERA5 meteorology, satellite NO2/HCHO, and surface observations within machine learning models, achieving cross-validated R2 of 0.94 and RMSE of 12.05 μg/m3 for surface ozone estimation. EMI-II estimates show strong agreement with independent observations (R = 0.91, RMSE = 10.83 μg/m3) and reproduce seasonal gradients, with summer concentrations (131 μg/m3) more than double winter levels (61 μg/m3). Estimation skill is regime-dependent: performance comparable to TROPOMI occurs under strong photochemical activity, while reduced sensitivity occurs under weak radiation and stable boundary layers—consistent with averaging kernel diagnostics. This first comprehensive validation demonstrates that EMI-II, despite vertical sensitivity limitations, provides meaningful surface ozone constraints under favorable atmospheric conditions. The framework is potentially applicable to other regions and sensors under similar conditions, providing a case study for integrating national satellite products into multi-source surface ozone estimation.