In highly stratified soils as on the Tibetan Plateau, uncertainty associated with a vertical profile of soil and hydraulic properties largely restricts the performance of Soil Vegetation Atmosphere Transfer (SVAT) model. In lieu of commonly used pedotransfer functions (PTFs) or artificial neural networks (ANNs), soil hydraulic properties in this study were inverted from an Ensemble Kalman filter (EnKF) analysis of Synthetic Aperture Radar (SAR) surface soil moisture. The calibrated SVAT scheme using inverted soil hydraulic variables C1 and θgeq was better matched with in situ field measurements than the uncalibrated SVAT scheme using soil maps–based PTFs on a local point scale. It was shown that the inverse calibration of two soil hydraulic variables solved the forecast bias (underestimation) in surface soil moisture due to the assumption of vertical homogeneity and the site-specificity of empirical PTFs. Additionally, at a SAR spatial scale, the calibrated SVAT scheme appropriately captured a high vertical gradient between surface and subsurface soil moisture, while the uncalibrated SVAT scheme could not. This suggests that it is possible to infer the SVAT soil hydraulic variables that are the main error source in SVAT scheme from the SAR soil moisture data assimilation analysis.