One of the gases responsible for the greenhouse effect is carbon dioxide (CO₂). For climate prediction, human production, and human life, the spatial distribution of CO₂ concentrations must be considered. Controlling the spatial and temporal distribution of CO₂ worldwide requires an accurate inversion of CO₂ concentrations. However, in the near-infrared band, surface reflectance uncertainty affects the inversion of CO₂ concentration. The ratio method is introduced to process the satellite ground-to-ground radiation spectrum, and it is verified that the absorption band radiance ratio and CO₂ concentration are related; moreover, inverting the CO₂ concentration using the relationship is feasible. The data source was the MODTRAN4-simulated radiation spectrum, and the four absorption peaks' spectral radiance ratio and CO₂ concentrations were selected for analysis. The results show that the spectral radiance ratio and CO₂ concentration have an approximately linear relationship, and the linear relationship is evident at 6310 cm^-1, with an error of only 1.15%. The ratio of radiance ratio to concentration is further investigated using various atmospheric and aerosol models. The results show the radiance ratio and concentration are highly correlated in the range of 0.1-0.9 reflectivities, the correlation coefficient is up to 0.98, and the average error is less than 2%. The measured data is subjected to the same processing as the simulation data, and the results are compared with simulated data. The linear relationship at 6334 cm^-1 is the best among the four bands, and the linear relationship reaches 0.99. It shows that the spectral radiance ratio and CO₂ concentrations have a linear relationship, and this relationship can be effectively applied to the inversion of CO₂ concentration, effectively eliminating surface reflectance.