Five surface backscattering models, including Oh, integral equation model (IEM), advanced integral equation model (AIEM), Dubois, and Shi models are selected to evaluate and reproduce synthetic aperture radar backscatter coefficients based on radar configuration and ground measurements at L- and C-bands. Regardless of bands or polarizations, the Oh model can attain a better performance among the five models with a root mean square error (RMSE) of about 2 dB, with the only exception being the AIEM and Shi models in VV polarization at the C-band. The Dubois model overestimates the radar signal and an underestimation is produced using the Shi model. The estimation accuracy of AIEM is significantly higher than that of IEM. Meanwhile, the performance of the scattering models in 0 to 7.6 cm is better than that in 0 to 20 cm. The frequency distribution of soil moisture over the field site approximates the normal distribution. Nevertheless, the estimated accuracy is not satisfactory for the inversion of AIEM. A site-specific calibration parameter is used at the C-band and improves the backscatter prediction for the AIEM. After calibration, the mean differences between the AIEM and RADARSAT-2 are nearly −1 dB with RMSEs of about 1 dB in the HH and VV polarizations. This work indicates that effective calibration factors can significantly improve the estimation accuracy and precisely implement soil moisture retrieval.