Abstract Accurate monitoring of atmospheric carbon dioxide (CO 2 ) concentrations is essential for understanding the global carbon cycle. Satellite remote sensing enables global‐scale CO 2 monitoring, providing column‐averaged dry air mole fractions of CO 2 (XCO 2 ). Current missions such as the Orbiting Carbon Observatory‐2 (OCO‐2) and Greenhouse gases Observing SATellite (GOSAT) series provide high‐quality XCO 2 but have many observation gaps because of narrow swath widths or cloud interference. To address this, we first propose a machine learning (ML) approach to estimate daily XCO 2 (ML XCO 2 ) at a 0.25 resolution using OCO‐2, Sentinel‐5 Precursor TROPOspheric Monitoring Instrument, and ERA5 reanalysis data from May 2018 to December 2023. Validation against the Total Carbon Column Observing Network measurements shows high agreement, with an R 2 of 0.95, a root mean square error of 1.05 ppm, and a mean absolute error of 0.78 ppm. The spatiotemporal variations in ML XCO 2 are generally consistent with OCO‐2, GOSAT, and the Copernicus Atmospheric Monitoring Service (CAMS) XCO 2 . Notably, during the Coronavirus disease 2019 (COVID‐19) pandemic in 2020, ML XCO 2 estimates were more consistent with OCO‐2 and GOSAT XCO 2 than CAMS XCO 2 , which was overestimated owing to unadjusted COVID‐19‐related CO 2 emission reductions. The annual mean CO 2 growth rates from ML XCO 2 (2.065–2.735 ppm/year, 2019–2023) also agree with estimates from OCO‐2 and the National Oceanic and Atmospheric Administration surface measurements, indicating the robustness of our approach. Our study demonstrates that the synergy between multiple‐satellite measurements enhances the spatial coverage of XCO 2 and improves our understanding of the global carbon cycle.
Lee et al. (Tue,) studied this question.