Before coupled atmosphere-lake models can be used to study the response of large lake systems to climatic forcings, we must first evaluate how well they simulate the water balance and associated lake atmosphere interactions under present-day conditions. We evaluate the hydrology simulated by a lake model coupled to NCAR's regional climate model (RegCM2) in a study of the Aral Sea. The meteorological variables that are input to the lake model are simulated well by RegCM2. Simulated surface air temperatures closely match observed values, except during spring and fall when the simulated temperatures are too cold. The magnitude of precipitation is too high in the region surrounding the Aral Sea during summer and fall. On a yearly basis, RegCM2 produces a reasonable amount of runoff throughout the drainage basin. The lake model coupled to RegCM2 accurately simulates Aral Sea surface temperatures (SSTs). The lake model also simulates observed mid-winter ice fraction well, although the onset of ice growth occurs too late in the year and the ice melts too rapidly in the spring. The simulated annual evaporation from the Aral Sea is consistent with observed estimates; however, the simulated evaporation is greater than observed during summer and less than observed during winter. In a "stand-alone" lake model simulation, the simulated Aral Sea hydrology does not match observations as closely as in the coupled model experiment. These results suggest that a stand-alone lake model would not accurately simulate the hydrologic response of the Aral Sea to various forcings. Copyright 1999 by the American Geophysical Union.