Research on the construction method of sample set in CO2 retrieval based on statistical analysis method

China's carbon reduction and emission reduction under the ‘double carbon’ target requires high-precision atmospheric CO₂ inversion data. High-precision atmospheric CO2 satellite observation is the basis for accurate positioning of emission sources. Aiming at the problem that the inversion algorithm is difficult to converge in the existing inversion methods, a statistical method integrating multi-source information is explored to obtain accurate prior profiles. It provides relatively accurate initial values for subsequent high-precision CO₂ inversion algorithms in order to improve the accuracy of CO₂ inversion. The statistical inversion method establishes a regression relationship model between satellite spectrum and atmospheric CO₂ concentration, and the inversion accuracy is affected by the representativeness of the sample and the regression analysis method. According to the difference of regional characteristics, the CO₂ profile samples are counted, and the sensitivity of environmental parameters is analyzed to construct a representative sample set. It is verified that the profile obtained by inversion is very close to the real profile, which verifies the feasibility of the method.