Abstract: Comparison of surface solar radiation data of NCEP/NCAR, NCEP/CFSR reanalysis data and data from stations and marine observations are conducted to check whether reanalysis data can reflect the characteristics of surface solar radiation over China. The results show that NCEP/NCAR reanalysis data are less reliable before 1979, but both NCEP/NCAR and NCEP/CFSR reanalysis data are high reliable after 1979, and the reliability is higher in east and low latitude areas than that in west and high latitude areas over China. Because the daily mean data calculated by six-hour reanalysis data is lower than real data over China, it is reasonable to calculate the daily data by excluding the case of zero observation to minish the deviation.Long-term average errors between NCEP/NCAR, NCEP/CFSR and station observations of surface solar radiation are 10.37 W·m^-2 and-42.68 W·m^-2 in Chinese Mainland from 1979 to 2009, with standard deviation of absolute errors being 12.31 W·m^-2 and 4.19 W·m^-2. Average errors between NCEP/NCAR, NCEP/CFSR and marine observations of surface solar radiation are-161.19 W·m^-2 and-179.66 W·m^-2 over the sea during marine observing, with standard deviation of absolute errors being 37.07 W·m^-2 and 35.36 W·m^-2. The surface solar radiation from both NCEP/NCAR and NCEP/CFSR reanalysis data are bigger than station observations along the Yangtze River Valleys around Chongqing, where the biggest deviation appears.This maybe due to the fact that cloudiness data from reanalysis are smaller than observations here. Since the NCEP mode doesn't consider aerosols, it has also induced the bigger deviation of surface solar radiation between reanalysis data and observations.Besides, Chongqing is a mountainous region, whose terrain shelter influences surface solar radiation. It should increase surface solar radiation from 18% to 63%, when reanalysis data are used on the Yangtze River Valleys. There is some false phenomenonin long-term variation of surface solar radiation from reanalysis data. For example, a false upward trend exists notably in NCEP/NCAR reanalysis data before 1979 and the false extreme value of NCEP/CFSR reanalysis data appear in 1998. To some extent, the reanalysis data can be used for study of the surface solar radiation in Chinese Mainland. And the temporal and spatial distribution characteristics of surface solar radiation can also be reflected over the sea. For the deviation and false phenomenon of reanalysis data, more attention should be paid when using them in long-term climatological diagnosis. Compared with surface solar radiation from land station observation, there is a big deviation between the marine observations and the reanalysis data. This may due to the lack of marine observations, which will confine the evaluation and development of NCEP mode. Besides, the sea data used are constructed from marine observations from 2006 to 2007. For its insufficient sample, the data quality of surface solar radiation from NCEP/NCAR and NCEP/CFSR reanalysis data need to be studied further over the sea.