Abstract: The national weather service modernization is the core and key to the modernization of the national weather, which is an important symbol to enhance China meteorological technology level and professional ability. China Meteorological Administration publishes the national meteorological modernization objectives and evaluation plan (2014-2020), which clearly proposes the development of multi-source data fusion data set, and the land surface data fusion is one of the most important parts. Using the technique of multi-source data fusion, China Meteorological Administration Land Data Assimilation System (CLDAS) integrates observation of ground, satellite and numerical model to obtain the high-quality temperature, pressure, humidity, wind speed, the grid point data of precipitation and radiation and other factors, and then to drive land surface model to simulate different depths of soil temperature and moisture.CLM3.5 land surface model is used to simulate land surface soil moisture of different depths, and then results are assessed using 3 ground datasets. The first is the automatic soil moisture observation of CMA in 2013, which is checked strictly by quality control process, the second is CTP-SMTMN data, and the last is GLDAS soil moisture and ERA-Interim Reanalysis. A comprehensive assessment for soil moisture is conducted and it shows that the correlation coefficient reaches a high level in most provinces, which can better reflect the objective change of soil moisture and has a strong guiding role. In statistical analysis of time series by selecting the representative station, surface soil moisture changing rates are higher than deeper layers, because the interaction between surface soil and the atmospheric boundary layer feedback is more sensitive, and the water heat exchange are more frequent. On the Tibetan Plateau, using Taylor diagrams comparison, it's found simulation results in the appraisal process of different indices are better than the other two kinds of foreign soil moisture data. In summary, the correlation coefficient is up to 0.8, and deviation is about-0.04 mm³·mm^-3 to 0.04 mm³·mm^-3 and root mean square error is lower than 0.04-0.05 mm³·mm^-3 from stations of provinces average and all over the country.