In view of the low spatial resolution of passive microwave soil moisture （SM） data， statistical and physical models including random forest （RF）， polynomial fitting and DISPATCH are utilized to disaggregate the FY3B microwave SM product from 25 km to 1 km with the synergistic application of Optical/Thermal infrared （TIR） observations and surface elevation parameters. Meanwhile， considering different overpass times of FY3B and other relevant input data source observations， four data combinations are separately used to derive the spatially downscaled SM with above three downscaling method， and the optimized data combination of FY-3B downscaling is proposed by comparison and analysis. Validation is performed from April 1， 2015 to December 31， 2016 with the in-situ measurements of REMEDHUS network and the precipitation time series of ECA&D meteorological site. Experimental results show that RF-based method can achieve the highest comprehensive downscaling accuracy and the best model fitting effect. In addition， the effect of applying FY-3B ascending data to downscale turns out to be better.