Abstract: The Nujiang River Basin in Yunnan Province was taken as the research object, the second class forest resources survey data in 2016 and Landsat 8 OLI remote sensing data in the same period were selected as the data source. By extracting the statistical value of the mean value of remote sensing variables in the small class forest, the aboveground biomass per unit area of small class forest of 9 dominant tree species or tree species groups in the study area was calculated by volume-biomass conversion model. The spherical model of semivariogram was used to calculate the light saturation values of 9 dominant tree species or groups of tree species. The aboveground forest biomass of different dominant tree species or tree groups was estimated by the stepwise regression model and BP neural network model. The results show that the light saturation values of forest in different dominant species or groups of tree species are birch forest 139 t/hm², logwood forest 181 t/hm², eucalyptus forest 70 t/hm², Yunnan pine forest 182 t/hm², spruce fir forest 197 t/hm², arbor economic forest 161 t/hm², other coniferous forest 182 t/hm², other broad-leaved forest 147 t/hm², evergreen oak forest 141 t/hm²; the fitting accuracy and test index of the BP neural network model are obviously better than the multivariate stepwise linear regression model, the R² of BP neural network model of each tree species or tree group is 0.1–0.2 higher than that of multivariate stepwise linear regression model; among the stepwise regression models, other broad-leaved forests have the highest R² up to 0.744; other broad-leaved forests have the highest R² in BP neural network models, up to 0.815, and the R² of tree economic forest, evergreen oak forest and log forest are above 0.6; segmental residual analysis shows that both models have low value overestimation and high value underestimation, especially when the biomass is less than 150 t/hm², BP the estimation accuracy of neural network model is obviously improved compared with stepwise linear regression.