A numerical model was used simulate the regional ground-water-flow system in the Dayton area in southwestern Ohio. Ground water is the primary source of drinking water for the Dayton area. The aquifer consists of glacial sands and gravels in a buried bedrock valley. The shale bed rock in the area is poorly permeable, but the glacial deposits can yield up to 2,000 gallons per minute to wells. Interaction with surface water is an important component of the ground-water-flow system. A steady-state, three dimensional, three-layer MODFLOW model of the glacial deposits was constructed to simulate the ground-water-flow system. The modeled area encompasses about 241 mi2 in Montgomery, Greene, and Clark Counties. The model simulated steady-state conditions of September 1993 and included 187 pumped wells. Hydraulic conductivities in the model ranged from less than 1 foot per day to 450 feet per day. Simulated recharge rates ranged from 6 inches per year to 12.2 inches per year. Recharge was used in select areas to simulate inflow from the bed rock-valley walls. Measured water levels from 579 wells and streamflow gain-loss data from six river reaches were used to evaluate the model. Ninety-one percent of simulated heads were within 15 feet of the measured heads. The root-mean-square error and mean absolute difference between measured and simulated heads were 7.3 feet and 4.5 feet respectively for layer 1, 10.1 feet and 6.5 feet for layer 2, and 8.8 feet and 6.8 feet for layer 3. Recharge and river leakage accounts for 81 percent of the water entering the model; pumped wells and river leakage accounts for almost 91 percent of the ground water leaving the model. Interaction of the ground-water system and the major rivers, which include the Great Miami, Mad, Stillwater, and Little Miami Rivers, is known from previous investigations in the area; however, the model simulation indicates that the smaller streams also may have a significant local influence. The vertical hydraulic conductivity of the glacial deposits appears to have more effect on ground-water flow in some areas near the bed rock-valley walls than in the central areas of the valley. At a local scale, simulated heads in the central areas of the valley were generally insensitive to changes in aquifer parameters. The sensitivity of the model to changes in simulated hydraulic properties of the aquifer was assessed by systematically changing model parameters in four subareas of the model. All areas of the model were sensitive to changes in recharge. Changes in other parameters, such as hydraulic conductivity or riverbed conductance, had variable effects. The sensitivity of the model can be used to indicate the types of additional hydrogeologic data that would be most useful to future investigations.