Abstract: Forest above-ground biomass (AGB) in Linhai city, Taizhou city, Zhejiang province was estimated with vegetation indexes, texture feature factors of Sentinel-2 optical remote sensing data during late October and mid November of 2017, and backscattering coefficient of Sentinel-1 SAR. Moreover, the inventory data for forest management survey in 2017 and digital elevation model data were integrated. Based on Random Forest (RF), Adaptive Boosting (AdaBoost) and Category Boosting (CatBoost) methods, AGB estimation models of 6 forest types in Lin’an were established. The coefficient of determination (R2) and root mean squared error (RMSE) are used to evaluate the performance of the models. The results show that CatBoost is better than RF and AdaBoost. The performance indexes generated by CatBoost are: R?=0.78, RMSE=7.26 t/ha for broad-leaved mixed forest; R?=0.72, RMSE=11.94 t/ha for conifer and broadleaves mixed forest; R?=0.60, RMSE=12.65 t/ha for coniferous mixed forest; R?=0.82, RMSE=9.22 t/ha for other hard broadleaf forest; R?=0.74, RMSE=10.12 t·hm-2 for Pinus massoniana forest; R?=0.75, RMSE=8.93 t·hm-2 for Cunninghamia lanceolata forest.