Snow cover is one of the active components of the cryosphere. Snow cover has a very important im-pact on the natural environment and human activities. Snow parameters (snow area, snow depth and snow water equivalent) inversion has practical significance to hydrological models and climate change research. However, the accuracy of snow depth inversion of remote sensing in the forest area should be further improved at present. Northeast is one of Chinas largest natural forest areas and important seasonal snow areas. This paper used L1 level brightness temperature data and L2 level snow water equivalent data of Microwave Radiation Imager (MWRI) mounted on FY3B satellite, and used field snow depth data in Northeast typical forest regions. Chang algorithm, NASA 96 algorithm and FY3B operational inversion algorithm were validated and analyzed. The re-sults showed that, in Northeast typical forest regions, the retrieved snow depth of Chang algorithm and NASA 96 algorithm had large fluctuations. The performance of NASA 96 algorithm was better than Chang algorithm and FY3B operational inversion algorithm when fractional forest cover (f ) was 0.6 or less, because the root mean square error value of NASA 96 algorithm was smaller than the other two algorithms. However, NASA96 algo-rithm had serious distortion when f was bigger than 0.6. Considering the pure forest pixel (f=1), Chang algo-rithm underestimated the snow depth of 47%. When f was 0.3 or less, FY3B operational inversion algorithm is better than Chang algorithm. On the whole, FY3B operational algorithm was relatively stable, and FY3B opera-tional algorithm had higher accuracy compared with Chang algorithm and NASA 96 algorithm.