We perform a 20-year, global-scale simulation with a land surface model (LSX) and compare its annual hydrologic budget with observations at continental and global scales. The model is forced with observed climate data, including temperature, relative humidity, precipitation, cloudiness, and wind. We find that LSX produces a satisfactory simulation of the annual hydrologic budget on both global and continental scales. The differences between model output and observations suggest the need for improved boundary conditions and land surface parameterizations, as well as the need for more complete data sets of global hydrologic variables. This study presents a methodology for the evaluation of land surface models for their more effective use in climate modeling studies. ¿ American Geophysical Union 1996