To effectively retrieve large-scale daily soil moisture, this study proposed a model-level integrated approach termed co-inversion of visible, infrared and passive microwave remote sensing data. Specifically, the MODIS data are used to derive soil moisture base, and the AMSR-E data are employed to estimate daily variation of land surface soil moisture over a large area. The soil moisture information over the large area is then estimated by integrating these two parts: base and variation. Improvements inherent in the proposed approach enable daily 1 km×1 km soil moisture estimation of the entire study area, even when some areas were covered with clouds. Verification with ground truthing data in Xinjiang, China shows that the co-inversion of thermal and passive microwave remotely sensed data can achieve better estimation of soil moisture than each single data source or model. The square correlation coefficient is 0.86 and RSME is 3.99 when the estimated soil moisture is compared with the ground truthings. The results proved that the co-inversion model outperformed either the MODIS or AMSR-E inversion of soil moisture over large areas, and can meet the needs of Xinjiang's soil moisture monitoring.