Ground-water flow under 2002 hydrologic conditions was evaluated in an eight-county area in Georgia and South Carolina near the Savannah River Site (SRS), by updating boundary conditions and pumping rates in an existing U.S. Geological Survey (USGS) ground-water model. The original ground-water model, developed to simulate hydrologic conditions during 1987-92, used the quasi-three-dimensional approach by dividing the Floridan, Dublin, and Midville aquifer systems into seven aquifers. The hydrogeologic system was modeled using six active layers (A2-A7) that were separated by confining units with an overlying source-sink layer to simulate the unconfined Upper Three Runs aquifer (layer A1). Potentiometric- surface maps depicting September 2002 for major aquifers were used to update, evaluate, and modify boundary conditions used by the earlier ground-water flow model. The model was updated using the USGS finite-difference code MODFLOW-2000 for mean-annual conditions during 1987-92 and 2002. The specified heads in the source-sink layer A1 were lowered to reflect observed water-level declines during the 1998-2002 drought. These declines resulted in a decrease of 12.1 million gallons per day (Mgal/d) in simulated recharge or vertical inflow to the uppermost confined aquifer (Gordon, layer A2). Although ground-water pumpage in the study area has increased by 32 Mgal/d since 1995, most of this increase (17.5 Mgal/d) was from the unconfined Upper Three Runs aquifer (source-sink layer A1) with the remaining 14.5 Mgal/d assigned to the active layers within the model (A2-A7). The simulated water budget for 2002 shows a decrease from the 1987-92 model from 1,040 Mgal/d to 1,035 Mgal/d. The decreased ground-water inflows and increased ground-water withdrawal rates reduced the simulated ground-water outflow to river cells in the active layers of the model by 43 Mgal/d. The calibration statistics for all layers of the 2002 simulation resulted in a decrease in the root mean square (RMS) of the residuals from 10.6 to 8.0 feet (ft). The residuals indicate 83.3 percent of the values for the 2002 simulation met the calibration error criteria established in the original model, whereas 88.8 percent was within the specified range for the 1987-92 simulation. Simulated ground-water outflow to the Savannah River and its tributaries during water year 2002 was 560 cubic feet per second (ft3/s), or 86 percent of the observed gain in mean-annual streamflow between streamflow gaging stations at the Millhaven, Ga., and Augusta, Ga. At Upper Three Runs Creek, simulated ground-water discharge during 2002 was 110 ft3/s, or 83 percent of the observed streamflow at two streamflow gaging stations near the SRS. These results indicate that the constructed model calibrated to 1987-92 conditions and modified for 2002 dry conditions is still representative of the hydrologic system. The USGS particle-tracking code MODPATH was used to generate advective water-particle pathlines and their associated time-of-travel based on MODFLOW simulations for 1987-92, 2002, and each of four hypothetical ground-water management scenarios. The four hypothetical ground-water management scenarios represent hydrologic conditions for (1) reported pumping for 2002 and boundary conditions for an average year; (2) reported pumping for 2002 with SRS pumping discontinued and boundary conditions for an average year; (3) projected 2020 pumping and boundary conditions for an average year; and (4) projected 2020 pumping and boundary conditions for a dry year. The MODPATH code was used in forward-tracking mode to evaluate flowpaths from areas on the SRS and in backtracking mode to evaluate further areas of previously documented trans-river flow on the Georgia side of the Savannah River. Trans-river flow is a condition in which the local head gradients might allow migration of contaminants from the SRS into the underlying aquifers and beneath the Savannah River into Georgia. More...