The hydrologic performance of a five-year GFDL R30 simulation is evaluated using observed temperature, precipitation, and runoff for the continental U.S.. While the model reproduces the major east-west structure of seasonal variations in the temperature and precipitation, it is too cold and wet in the winter, and warm and dry in the summer, and has too little runoff, for much of the continental U.S. As an alternative to the bucket hydrology used in the GFDL (and most other) GCM's, the performance of a model that represents variations in infiltration capacity within a GCM grid, as well as baseflow, was evaluated. Comparisons were made with GFDL and GISS bucket hydrology using five years of daily GFDL R30-simulated precipitation and temperature for a southeastern U.S. grid cell. More dynamic and apparently realistic short-term variations in soil moisture were achieved for the alternative hydrologic model, but the most significant differences in the hydrologic dynamics were attributable to the inclusion of baseflow to simulate between-storm runoff.