Estimation of aboveground green biomass is essential for evaluating grassland productivity and functioning. This study aimed to explore the potential of band depth indices for estimating aboveground green biomass in grassland with low canopy cover. Field spectral and biomass measurements were conducted during 2009 and 2010 growing seasons in desert steppe of Inner Mongolia. Band depth (BD), band depth ratio (BDR), normalised band depth index (NBDI), band depth normalised to area (BNA), maximum band depth (BDmax), and area of absorption region (BDarea) extracted from red absorption region (650–740 nm) were utilised as band depth indices. Results indicated that: (1) BD at individual bands between 655 and 716 nm showed good accuracy for aboveground green biomass estimation; (2) BD at 698 nm yielded the best accuracy (R2 = 0.7, RMSECV = 29.6 g m−2 for calibration; RMSE = 32.4 g m−2, rRMSE = 26.9% for validation); (3) BDR, NBDI, and BNA at all bands were not reliable estimators of aboveground green biomass (R2 < 0.3, RMSECV > 45 g m−2 for calibration; RMSE > 46 g m−2, rRMSE > 39% for validation); (4) although the performance of BDmax (R2 = 0.65, RMSECV = 32.1 g m−2 for calibration; RMSE = 34.5 g m−2, rRMSE = 28.7% for validation) and BDarea (R2 = 0.69, RMSECV = 30.2 g m−2 for calibration; RMSE = 33.1 g m−2, rRMSE = 27.6% for validation) was slight lower than that of BD698nm, the performance was far better than that of BDR, NBDI, and BNA. Our results suggest that BD698nm has good potential to estimate aboveground green biomass in grassland with low canopy cover. The performance of BD698nm needs to be further tested using space-borne hyperspectral images.