Solar radiation reaching the surface in China was analyzed on the basis of four data sets. National Centers for Environment Prediction (NCEP) reanalysis solar radiation data exceeded surface observations by 40 W/m2 to more than 100 W/m2. Satellite data sets produced by Goddard Institute for Space Studies (GISS) and University of Maryland (UMD) were correlated significantly with surface observations, but satellite estimates exceeded surface observations over much of China. The largest difference occurred in the middle reach of Yangze river and coast regions where satellite algorithms occasionally overestimated solar radiation by more than 40 W/m2. The mean bias of satellite estimates ranged from about 8 W/m2 to about 36 W/m2 in regional scale. The root-mean-square error varied from about 20 to 36 W/m2. The effects of differences in space-time sampling between satellite and surface data might account partly for the random uncertainties of satellite estimates. The systematic overestimation by satellite algorithms in China was most likely associated with aerosols and their complex interactions with clouds. Surface data indicated a significant decline in solar radiation from 1961 to 1990, but the decline trend did not persist into the 1990s. Instead, a pronounced increase trend has been observed from 1984 to 2000, ranging from 1.7% to 7.7% per decade. The decrease would not be expected given that the cloud amount decreased significantly during the same period, but this was in accordance with the increasing aerosol loading that was indicated by the visibility deterioration. NCEP data and satellite estimates indicated quite different trends in solar radiation. The marginal increase trends were derived on the basis of NCEP data from 1961 to 2000 in three regions except in south China. The trends derived from satellite estimates since the middle 1980s were much less than that of surface observations. Moreover, both satellite estimates showed decline trends over much of China from 1984 to 2000. These differences suggested that representation of aerosols and clouds should be improved in order to make satellite estimates more robust over regions with high aerosol loading.