As part of the Place-Based Studies Program, the U.S. Geological Survey initiated a project in 1996, in cooperation with the U.S. Army Corps of Engineers, to quantify the rates and patterns of submarine ground-water discharge to Biscayne Bay. Project objectives were achieved through field investigations at three sites (Coconut Grove, Deering Estate, and Mowry Canal) along the coastline of Biscayne Bay and through the development and calibration of variable-density, ground-water flow models. Two-dimensional, vertical cross-sectional models were developed for steady-state conditions for the Coconut Grove and Deering Estate transects to quantify local-scale ground-water discharge patterns to Biscayne Bay. A larger regional-scale model was developed in three dimensions to simulate submarine ground-water discharge to the entire bay. The SEAWAT code, which is a combined version of MODFLOW and MT3D, was used to simulate the complex variable-density flow patterns. Field data suggest that ground-water discharge to Biscayne Bay relative to the shoreline is restricted to within 300 meters at Coconut Grove, 600 to 1,000 meters at Deering Estate, and 100 meters at Mowry Canal. The vertical cross-sectional models, which were calibrated to the field data using the assumption of steady state, tend to focus ground-water discharge to within 50 to 200 meters of the shoreline. With homogeneous distributions for aquifer parameters and a constant-concentration boundary for Biscayne Bay, the numerical models could not reproduce the lower ground-water salinities observed beneath the bay, which suggests that further research may be necessary to improve the accuracy of the numerical simulations. Results from the cross-sectional models, which were able to simulate the approximate position of the saltwater interface, suggest that longitudinal dispersivity ranges between 1 and 10 meters, and transverse dispersivity ranges from 0.1 to 1 meter for the Biscayne aquifer. The three-dimensional, regional-scale model was calibrated to ground-water heads, canal baseflow, and the general position of the saltwater interface for nearly a 10-year period from 1989 to 1998. The mean absolute error between observed and simulated head values is 0.15 meter. The mean absolute error between observed and simulated baseflow is 3 x 105 cubic meters per day. The position of the simulated saltwater interface generally matches the position observed in the field, except for areas north of the Miami Canal where the simulated saltwater interface is located about 5 kilometers inland of the observed saltwater interface. Results from the regional-scale model suggest that the average rate of fresh ground-water discharge to Biscayne Bay for the 10-year period (1989-98) is about 2 x 105 cubic meters per day for 100 kilometers of coastline. This simulated discharge rate is about 6 percent of the measured surface-water discharge to Biscayne Bay for the same period. The model also suggests that nearly 100 percent of the fresh ground-water discharge is to the northern half of Biscayne Bay, north of the Cutler Drain Canal. South of the Cutler Drain Canal, coastal lowlands prevent the water table from rising high enough to drive measurable quantities of ground water to Biscayne Bay. Annual variations in sea-level elevation, which can be as large as 0.3 meter, have a substantial effect on rates of ground-water discharge. During 1989-98, simulated rates of ground-water discharge to Biscayne Bay generally are highest when sea level is relatively low.