Abstract: Abstract: Soil moisture plays a key role in the interactions among the hydrosphere, biosphere, and atmosphere. Traditionally, soil moisture information is measured by ground-based soil moisture monitoring networks, which is accurate but time-consuming and laborious. In this study, a new empirical model is developed for estimating the soil moisture of bare surfaces by dual-polarization ASAR. The steps are as follows: first, a database linked to SAR backscattering coefficients, surface roughness parameters, and soil moisture is built by AIEM (advanced integral equation model). Through mathematical analysis of a simulated database, the influence of roughness and soil moisture are taken into account, respectively. For roughness impact, a new roughness parameter Rs=S3/L2 is defined by combining the traditional roughness parameter S with L. Then, the unknown parameters in the empirical model are only roughness parameter Rs and volumetric soil moisture mv. The soil moisture can be retrieved from dual-polarization SAR observations. Concerning the influence of soil moisture, the Fresnel reflection coefficient Г0 is brought in to take place of mv because a better relationship can be built between the Fresnel reflection coefficientГ0and the backscattering coefficient σ0. In this case, Fresnel reflection coefficient Г0 can be directly retrieved from the empirical model, not soil moisture mv. The soil dielectric constant ε can be determined by Fresnel reflection coefficient Г0 and the Dobson Model, in which soil moisture can be linked with dielectric constant ε. To estimate the accuracy of the empirical model, the results of the empirical model were compared with in-situ data in the same location collected at Heihe river basin, Zhangye, in 2008. It concluded that, when θ＞25°, S＜1.5 cm, L∈(4，18) cm, there was a good relationship between the estimated data and in-situ data. The correlation coefficient R2 could be as high as 0.745, meanwhile the RMSE (root mean square error) was 0.478. Because this method requires only dual-polarization SAR data for retrieving soil moisture, and does not need any ground roughness observations, it is suitable for soil moisture retrieval in large regions. However, this new model needs to be validated by more in-situ experiments and combined with vegetation models in order to to meet regions covered by vegetation.