In view of the fact that the current soil moisture retrieval from remotely sensed data is low in accuracy, a new integrated approach termed “Co-inversion of land surface soil moisture by integrating optical, thermal infrared and passive microwave remote sensing data” was proposed. Specifically, the MODIS optical and thermal infrared bands are used to derive soil moisture benchmark, and the AMSR-E X-band data to estimate daily variation of land surface soil moisture. Then the two are integrated, building up a co-inversion model for soil moisture retrieval over a large area. Xinjiang was cited as experiment zone. A total of 365 in-situ measured soil moisture values were collected from a typical area and used to test the proposed inversion model. Verification analysis with the ground truthing data of the study area shows that the co-inversion of optical/thermal and microwave remotely sensed data displays higher correlation coefficient and smaller root mean square errors (RMSE) than any inversion using one single data source.