Remote sensing-based estimation of soil moisture is crucial in many aspects including basin-scale water resource management, irrigation scheduling, regional scale drought monitoring and crop yield forecasting. In this study, we evaluate the potential of visible/thermal-infrared remote sensing in soil moisture estimation, by assessing the TVDI-based method and three categories of methods based on evaporative fraction/potential evaporation ratio (EFM1, EFM2 and EFM3). In combination with ASTER data set, soil moisture in middle reach of the Heihe River Basin is predicted by the above-mentioned four methods and validated by the ground-based measurements from eco-hydrological wireless sensor network and hydro meteorological observation network in the middle reach of Heihe river basin. Results indicate that uncertainties arise from the empiricism of the TVDI-based method in the process of determining dry and wet edges. On the other hand, the evaporation fraction/potential evaporation ratio methods can to some degree reduce the uncertainties, and among the three methods, EFM1 and EFM3 outperform EFM2. In addition, the thermal-infrared based methods require accurate soil parameters to reproduce the variation of soil moisture.