Abstract: Using the forests in Pu'er City, Yunnan Province as the research subject, the forest carbon concentration was inversely inferred using the GOSAT satellite's B2, B3, and B4 spectral channels. Highly correlated factors for forest carbon stock were extracted and selected from Landsat OLI multispectral images and DEM, and a GEOS–Chem model was constructed. Carbon flux was then inversely inferred, and through the conversion between carbon flux and carbon stock, the estimation of forest carbon stock was achieved. The results showed that the highly correlated factors for forest carbon stock were the Elevation, NDVI, R9Mean from the Landsat data, and the carbon concentration inverted by GOSAT. The optimal model for estimating forest carbon stock was GEOS–Chem2, with an R² of 0.978 and P of 94.89%. Compared to the model GEOS–Chem1 constructed solely using GOSAT data (R² of 0.847, P of 85.32%), the R² and P were increased by 0.131 and 9.57% respectively. The estimated forest carbon stock in Pu'er City using the GEOS–Chem2 model was 4.253 × 10⁷ t, with an average carbon stock of 19.356 t/hm², and an overall estimation error of 4.69%. By integrating Landsat and GOSAT satellite data to construct the GEOS–Chem model, the estimation error of forest carbon stock in Pu'er City can be effectively reduced. The research results can provide reference for the exploration of high-precision remote sensing estimation methods for forest carbon stock.