China’s Fengyun-3D meteorological satellite launched in December 2016 carries the high-resolution greenhousegases absorption spectrometer(GAS) aimed at providing global observations of carbon dioxide(CO2). To date, GAS is one of the few instruments measuring CO2from the near-infrared spectrum. On orbit, the oxygen(O2) A band suffers a disturbance, and the signal-to-noise ratio(SNR) is significantly lower than the nominal specification. This leads to difficulties in the retrieval of surface pressure and hence a degradation of the retrieval of the column-averaged CO2dry air mole fraction(XCO2) if a full physics retrieval algorithm is used. Thus, a fast CO2inverse method, named semi-physical statistical algorithm, was developed to overcome this deficiency. The instrument characteristics, the semi-physical statistical algorithm, and the results of comparison with ground-based measurements over land were introduced in this paper. XCO2can be obtained from three bands, namely, the O2A, weak CO2, and strong CO2bands, with compensation from the Medium Resolution Spectral Imager-2(MERSI-2) products, ECMWF Reanalysis v5(ERA-5) data, and Total Carbon Column Observing Network(TCCON) data. The eigenvectors of covariance matrices and the least square fits were used to derive retrieval coefficients and yield cloud-free solutions. In addition to the GAS radiance, some key factors necessary for the accurate estimations of XCO2were also taken as input information(e.g., air mass, surface pressure, and a priori XCO2). The global GAS XCO2restricted over land was compared against the simultaneously collocated observations from TCCON. The retrieval algorithm can mitigate the issue caused by the low SNR of the O2A band to a certain extent. Overall, through site-by-site comparisons, GAS XCO2agreed well with the average precision(1σ) of 1.52 ppm and bias of-0.007 ppm. The seasonal variation trends of GAS XCO2can be clearly seen at TCCON sites on the 1-yr timescale.