The efficacy of data assimilation frameworks incorporating the Ensemble Kalman Filter (EnKF) for enhancing soil moisture estimation has been examined in prior studies. This research evaluates the capability of generating high-resolution soil moisture maps through the synergistic integration of optical remote sensing observations with a hydrological modeling system at regional scales. The proposed methodology employs a two-component operational structure: the Temperature-Vegetation Dryness Index (TVDI) serves as the observation operator, while the Distributed Hydrology-Soil-Vegetation Model (DHSVM) functions as the model operator. Specifically, the EnKF algorithm facilitates the assimilation of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) thermal infrared observations, subsequently updating surface soil moisture estimates across heterogeneous landscapes. The developed data assimilation framework was rigorously validated through a 15-day field campaign (June 1–15, 2022) conducted within the middle reaches of the Heihe River Basin, utilizing in situ soil moisture measurements as ground truth references. The results show that the data assimilation can improve the surface soil moisture estimation significantly. Both of the MBE and RMSE of assimilation soil moisture are smaller than simulation results in most of the study area, only when the soil is a little dry (mv < 0.1 g/cm3), the assimilation results are worse. In addition, it may be a practical and effective way to increase the precision of soil moisture in regional scale.