Abstract: Abstract: Soil moisture is one key factor to restrict the growth of crops in rainless regions, and it is crucial to farmland production and can significantly affect the irrigation decision-making for agricultural management. Polarimetric SAR is very sensitive to soil moisture and can penetrate smoke, fog, rain and snow, etc. Therefore, polarimetric SAR system can operate in all-weather conditions. Moreover, soil moisture of farmland can be detected the by penetrating the vegetation and soil to reach the subsurface. In this study, in-situ measurements were carried out in Yangling district, Shannxi province, an important winter wheat producing area in China. Soil moisture and crop parameters (such as leaf area index (LAI), biomass and vegetation water content) were obtained by different teams at Radarsat-2 satellite transit time. Radar vegetation index (RVI), which was less sensitive to change in environmental conditions, had strong correlations with LAI, biomass and vegetation water content, etc. Aiming at monitoring soil moisture by satellite polarimetric SAR data conveniently and accurately, an improved RVI model for soil moisture estimation was proposed via bare soil model and RVI combined with water cloud model (WCM). 108 soil moisture samples (including jointing, heading and filling stageof winter wheat) were used to estimate and verify the accuracy of the parameters in the WCM, among which 62 samples were used to estimate the empirical parameters of the model, and the remaining 46 samples to verify the accuracy of estimated parameters. Usually, the crop parameters were used to describe the water cloud in the WCM, however, it causes lots of difficulties to apply this model. The crop parameters in improved RVI model were replaced by the RVI to reduce the uncertainty of the WCM. The RVI, instead of the crop parameters, was used to describe the water cloud in WCM, then the soil moisture values estimated by LAI model and improved RVI model were compared with in-situ measured soil moisture values, respectively. The LAI model was verified by LAI obtained in ground measurement and estimated by optical remote sensing data, respectively. To prove the potential applications, the improved RVI model was compared with two LAI models using ground measurement data. Conclusions were drawn based on the comparisons of coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE) as follows: 1) The improved RVI model (combined with the WCM, bare soil model and RVI) achieved the best accuracy (R2 was 0.41, RMSE was 5.7%, MAE was4.7%), the second was the original LAI model (using on ground measured LAI for validation, R2 was 0.28, RMSE was 7.1%, MAE was 5.1%), while the original LAI model (using optical remote sensing estimated LAI for validation) was the worst (R2 was 0.14, RMSE was 7.7%, MAE was 6.1%). 2) The improved RVI model, which used polarimetric SAR data and RVI as inputs, provided a convenient and efficient way to estimate the soil moisture. 3) The improved RVI model provided an accurate and rapid method for monitoring soil moisture in various growth stages of winter wheat.