The groundwater model documented in this report simulates the transient evolution of water levels in the aquifers and confining units of the Virginia Coastal Plain and adjacent portions of Maryland and North Carolina since 1890. Groundwater withdrawals have lowered water levels in Virginia Coastal Plain aquifers and have resulted in drawdown in the Potomac aquifer exceeding 200 feet in some areas. The discovery of the Chesapeake Bay impact crater and a revised conceptualization of the Potomac aquifer are two major changes to the hydrogeologic framework that have been incorporated into the groundwater model. The spatial scale of the model was selected on the basis of the primary function of the model of assessing the regional water-level responses of the confined aquifers beneath the Coastal Plain. The local horizontal groundwater flow through the surficial aquifer is not intended to be accurately simulated. Representation of recharge, evapotranspiration, and interaction with surface-water features, such as major rivers, lakes, the Chesapeake Bay, and the Atlantic Ocean, enable simulation of shallow flow-system details that influence locations of recharge to and discharge from the deeper confined flow system. The increased density of groundwater associated with the transition from fresh to salty groundwater near the Atlantic Ocean affects regional groundwater flow and was simulated with the Variable Density Flow Process of SEAWAT (a U.S. Geological Survey program for simulation of three-dimensional variable-density groundwater flow and transport). The groundwater density distribution was generated by a separate 108,000-year simulation of Pleistocene freshwater flushing around the Chesapeake Bay impact crater during transient sea-level changes. Specified-flux boundaries simulate increasing groundwater underflow out of the model domain into Maryland and minor underflow from the Piedmont Province into the model domain. Reported withdrawals accounted for approximately 75 percent of the total groundwater withdrawn from Coastal Plain aquifers during the year 2000. Unreported self-supplied withdrawals were simulated in the groundwater model by specifying their probable locations, magnitudes, and aquifer assignments on the basis of a separate study of domestic-well characteristics in Virginia. The groundwater flow model was calibrated to 7,183 historic water-level observations from 497 observation wells with the parameter-estimation codes UCODE-2005 and PEST. Most water-level observations were from the Potomac aquifer system, which permitted a more complex spatial distribution of simulated hydraulic conductivity within the Potomac aquifer than was possible for other aquifers. Zone, function, and pilot-point approaches were used to distribute assigned hydraulic properties within the aquifer system. The good fit (root mean square error = 3.6 feet) of simulated to observed water levels and reasonableness of the estimated parameter values indicate the model is a good representation of the physical groundwater flow system. The magnitudes and temporal and spatial distributions of residuals indicate no appreciable model bias. The model is intended to be useful for predicting changes in regional groundwater levels in the confined aquifer system in response to future pumping. Because the transient release of water stored in low-permeability confining units is simulated, drawdowns resulting from simulated pumping stresses may change substantially through time before reaching steady state. Consequently, transient simulations of water levels at different future times will be more accurate than a steady-state simulation for evaluating probable future aquifer-system responses to proposed pumping.