Abstract: The estimation of soil moisture in summer of 2009 is carried out using the ENVISAT-1/ASAR VV polarization data for the Maqu area at the upper reaches of the Yellow River.The geometrical optics model (GOM)is used to compute backscattering coefficients.The effect of surface short grasses on the estimation can be eliminated from the total microwave backscatters using a semi-empirical water-cloud model.The latter is calibrated using measurements collected synchronously from the regional soil moisture network.The result shows that the root mean square error of the remotely sensed soil moisture is less than 0.02 m³/m³,while the R² (coefficient of determination)is greater than 0.86.A high value of 0.4 m³/m³ is obtained in the summer soil moisture average,indicating that the Maqu area might serve as an important source water area for the Yellow River.However,a significant heterogeneity is present in the spatial distribution of soil moisture due to the influence of terrains.The spatial correlation distance of 500 m to 1000 m is found for the soil moisture in mountainous regions,while that in flat regions is greater than 2000 m.Our result demonstrates the applicability of using remote sensing techniques for soil moisture retrieval at the regional scale.