The increase of greenhouse gas is one of the most important factors leading to global climate change. Differential absorption lidar (DIAL) is considered to be the tool with the most potential to measure CO2 remotely. However, it is difficult to obtain accurate CO2 retrievals and determine carbon fluxes with the traditional dual-wavelength differential absorption inversion method, which is characterized by a low signal-to-noise ratio (SNR). Therefore, a multi-wavelength differential absorption inversion framework is proposed in this work. Based on the measurement of the absorption optical depths (ODs) of a single line at 30 different wavelengths, we tried to minimize the differences between the simulated and measured absorption ODs through an iterative process, which consisted of forward simulation and reverse inversion processes. The retrievals are determined when an optimal solution is obtained. The simulation experiments show that the precision of the nonuniform sampling wavelength method is better than that of the dual-wavelength method. The precision of the proposed method was improved by more than 60% for an SNR of 30-40 dB compared to the traditional method. Furthermore, the average error of this method is about 1/9th the traditional method.