Solar radiation is a critical variable in global change sciences. While most of the current global datasets provide only the total downward solar radiation, we aim to develop a method to estimate the downward global land surface solar radiation and its partitioned direct and diffuse components, which provide the necessary key meteorological inputs for most land surface models. We developed a simple satellite-based computing scheme to enable fast and reliable estimation of these variables. The global Moderate Resolution Imaging Spectroradiometer (MODIS) products at 1° spatial resolution for the period 2003–2011 were used as the forcing data. Evaluations at Baseline Surface Radiation Network (BSRN) sites show good agreement between the estimated radiation and ground-based observations. At all the 48 BSRN sites, the RMSE between the observations and estimations are 34.59, 41.98 and 28.06 W∙m−2 for total, direct and diffuse solar radiation, respectively. Our estimations tend to slightly overestimate the total and diffuse but underestimate the direct solar radiation. The errors may be related to the simple model structure and error of the input data. Our estimation is also comparable to the Clouds and Earth’s Radiant Energy System (CERES) data while shows notable improvement over the widely used National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) Reanalysis data. Using our MODIS-based datasets of total solar radiation and its partitioned components to drive land surface models should improve simulations of global dynamics of water, carbon and climate.