Abstract: Based on the second-class survey data of forest resources and Landsat 8 OLI remote sensing images in the same period, taking 11 dominant tree species in Yuanjiang basin of Yunnan Province as the research object, the remote sensing biomass estimation models are established by using the methods of multiple linear stepwise regression and support vector machine regression, and then the forest biomass in the basin is retrieved, and the threshold of optical remote sensing biomass saturation point is determined at the same time. The results showed that the light saturation values estimated by remote sensing of aboveground biomass of 11 dominant tree species were 83 t/hm² of Yunnan pine, 79 t/hm² of Simao Pine, 125 t/hm² of Huashan pine, 68 t/hm² of Cunninghamia lanceolata, 89 t/hm² of other conifers, 74 t/hm² of Eucalyptus, 66 t/hm² of rubber, 117 t/hm² of evergreen broad-leaved leaves, 56 t/hm² of deciduous broad-leaved leaves, 85 t/hm² of other broad-leaved trees and 55 t/hm² of other economic trees; Absolute value of average relative error and determination coefficient R² of support vector machine model. The average residuals of 11 dominant tree species in support vector machine model are less than those in multiple linear regression model. This study can provide a reference for improving the estimation accuracy of forest biomass in Yuanjiang River Basin.