Solar radiation incident at the Earth's surface (R s ) is an essential component of the total energy exchange between the atmosphere and the surface. Reanalysis data have been widely used, but a comprehensive validation using surface measurements is still highly needed. In this study, we evaluated the R s estimates from six current representative global reanalyses (NCEP-NCAR, NCEP-DOE; CFSR; ERA-Interim; MERRA; and JRA-55) using surface measurements from different observation networks [GEBA; BSRN; GC-NET; Buoy; and CMA] (674 sites in total) and the Earth's Radiant Energy System (CERES) EBAF product from 2001 to 2009. The global mean biases between the reanalysis R s and surface measurements at all sites ranged from 11.25 W/m 2 to 49.80 W/m 2 . Comparing with the CERES-EBAF R s product, all the reanalyses overestimate R s , except for ERA-Interim, with the biases ranging from´2.98 W/m 2 to 21.97 W/m 2 over the globe. It was also found that the biases of cloud fraction (CF) in the reanalyses caused the overestimation of R s . After removing the averaged bias of CERES-EBAF, weighted by the area of the latitudinal band, a global annual mean R s values of 184.6 W/m 2 , 180.0 W/m 2 , and 182.9 W/m 2 were obtained over land, ocean, and the globe, respectively.