Abstract: 【Objective】 In order to explore the ability of multi-source data to estimate aboveground biomass (AGB) of D.tianschanicum. 【Methods】 A drone platform equipped with visible light and multispectral sensors was used to collect information on blooming and heading stages and obtain visible light vegetation index, multispectral vegetation index, and a combination of the two vegetation indices.Multiple linear regression (MLR), stepwise linear regression (SMLR) Random Forest Regression (RF) were applied to establish an AGB estimation model for single vegetation index and fused vegetation index by using the determination coefficient (R²), and to valuate the estimation model with the adjusted coefficient of determination (R_adj^2) and root mean square error (RMSE). 【Results】 The vegetation index in the combination of near-infrared and red edge bands was more sensitive to the AGB of D. tianschanicum, so selecting;The peak flowering period had the best estimation effect among different growth stages, and the fitting effect was the best in the multiple linear stepwise regression estimation model based on the fusion vegetation index. The model's R², R_adj^2 and RMSE were 0.837, 0.831, and 7.357;Compared with vegetation index estimation models based on a single type, estimation models based on fused vegetation indices hadthe best fitting accuracy and better stability. 【Conclusion】 The fusion of vegetation index can effectively increase spectral information and improve model prediction accuracy.