In this paper, on the basis of the Heihe hydrological drought experiments, the active and passive microwave data joint inversion method of soil moisture has been explored as Linze Station for the study area. For ALOS/PALSAR data, co-polarization backscattering coefficient was calculated using the theoretical model AIEM, and Oh semi-empirical model was used to describe the characteristics of cross-polarization scattering. By a large number of back-scattering simulation data analysis, a calculation model of surface roughness was established. With the use of simulation data, changes of surface radiation brightness temperature with soil moisture and roughness were analyzed. Based on this, a neural network model was built to combine roughness calculation results and the flight data, and soil moisture of the study area was estimated with the trained model. Verified results with synchronous measurement data showed that the method can give full play to the active and passive microwave data on their respective strengths, while avoiding the main problem of scaling issues with passive and active data. And it provides a new effective way for basin-scale monitoring of soil moisture.