Ground-water flow and stream-aquifer relations were simulated for seven aquifers in Coastal Plain sediments in the vicinity of the U.S. Department of Energy, Savannah River Site (SRS), in Georgia and South Carolina to evaluate the potential for ground water containing hazardous materials to migrate from the SRS into Georgia through aquifers underlying the Savannah River (trans-river flow). The work was completed as part of a cooperative study between the U.S. Geological Survey, the U.S. Department of Energy, and Georgia Department of Natural Resources. The U.S. Geological Survey three-dimensional finite-difference ground-water flow model, MODFLOW, was used to simulate ground-water flow in three aquifer systems containing seven discrete aquifers: (1) the Floridan aquifer system, consisting of the Upper Three Runs and Gordon aquifers in sediments of Eocene age; (2) the Dublin aquifer system, consisting of the Millers Pond, and upper and lower Dublin aquifers in sediments of Paleocene and Late Cretaceous age; and (3) the Midville aquifer system, consisting of the upper and lower Midville aquifers of sediments in Late Cretaceous age. Ground-water flow was simulated using a series of steady-state simulations of predevelopment (pre-1953) conditions and six pumping periods--1953-60, 1961-70, 1971-75, 1976-80, 1981-86, and 1987-92--results are presented for predevelopment (prior to 1953) and modern-day (1987-92) conditions. Total simulated predevelopment inflow is 1,023 million gallons per day (Mgal/d), of which 76 percent is contributed by leakage from the Upper Three Runs aquifer. Over most of the study area, pumpage induced changes in ground-water levels, ground-water discharge to streams, and water-budget components were small during 1953-92, and changes in aquifer storage were insignificant. Simulated drawdown between predevelopment and modern-day conditions is small (less than 7 feet) and of limited areal extent--the largest simulated declines occur in the upper and lower Dublin aquifers in the vicinity of the Sandoz plant site in South Carolina. These declines extend beneath the Savannah River and change the configuration of the simulated potentiometric surface and flow paths near the river. Predevelopment and modern-day flowpaths were simulated near the Savannah River by using the U.S. Geological Survey particle-tracking code MODPATH. Eastward and westward zones of trans-river flow were identified in three principal areas as follows: --zone 1-from the Fall Line southward to the confluence of Hollow Creek and the Savannah River; --zone 2-from the zone 1 boundary southward to the southern border of the SRS (not including the Lower Three Runs Creek section); and --zone 3-from the zone 2 boundary, southward into the northern part of Screven County, Ga. All zones for all model layers were located within or immediately adjacent to the Savannah River alluvial valley and most were located in the immediate vicinity of the Savannah River. Recharge areas for each of the zones of trans-river flow generally are in the vicinity of major interstream drainage divides. Mean time-of-travel simulated for predevelopment conditions ranges from 300 to 24,000 years for westward trans-river flow zones; and from 550 to 41,000 years for eastward zones. Corresponding travel times under modern-day conditions range from 300 to 34,000 years for westward zones and from 580 to 31,000 years for eastward zones. Differences in travel times between predevelopment and modern-day simulations result from changes in hydraulic gradients due to ground-water pumpage that alter flow paths in the vicinity of the river. Recharge to Georgia trans-river flow zones originating on the SRS was simulated for the Gordon and upper Dublin aquifers during predevelopment, and in the Gordon aquifer during 1987-92. During 1987-92, SRS recharge was simulated in 6 model cells covering a 2-square mile area, located away from areas of ground-water contamination. Si