Radarsat-2 Synthetic Aperature Radar (SAR) remote sensing data were used to record soil surface moisture and evaluate the utility of a cross polarization (VV/VH) combination. Studies were conducted at Dingxi, in the semi-arid region of the Loess Plateau, China. We combined these data with MODIS optical data, used a Water-Cloud model to correct for the influence of vegetation, and then estimated the soil moisture under crop cover. For bare surfaces, the value of the cross polarization combination model was highly correlated to the measurement of soil moisture at 10 similar to 20 cm depth (R=0.75, P<0.01). The correlations between estimated values and the measured soil moisture at 0 similar to 10 cm and 20 similar to 30 cm depths were lower but still significant (R=0.47 and R=0.52, respectively, P<0.05). For soil surfaces covered with vegetation the model significantly underestimated soil moisture. After vegetation removal, the correlation coefficient increased from 0.30 to 0.70, the standard deviation decreased from 4.99 to 3.05, and the accuracy of the soil moisture model improved. Most soil moisture readings in the study area were 10 similar to 30% and these were consistent with the actual field moisture levels. Improving the accuracy of soil moisture readings in agricultural fields using optical and microwave remote sensing data will promote increased use of this technology.