The applicability of the Alpha approximation model for retrieving soil moisture data from multi-temporal SAR data will be affected when incidence angles change. Aiming at this problem, the Alpha model is modified based on the theory of Small Perturbation model, and verified it with the time series Sentinel-1 SAR data of Heihe and Winnipeg, the two experimental areas coupled with the ground observation network therein. A correlation analysis was done of the inversion with the observation data. It was found that the correlation coefficient increased from 0.473 and 0.601 to 0.851 and 0.821 respectively, while the root mean square error (m³• m^-3) decreased from 0.053 and 0.152 to 0.023 and 0.065 respectively. In order to further evaluate quality of the model, the NSE coefficient which is commonly used in hydrology for evaluation of model was introduced for use. The NSE before and after the modification were -0.532 and -1.243, 0.713 and 0.588, respectively. The results show that though a certain relationship was found between the inversion using the original model and the measurement in data, a quite big error was too. However, once the model is modified as above mentioned, the impact of changes in incidence angle can be well corrected, making the inversion more reliable, which extends the application of the Alpha model.