To construct a digital simulation of a plume of brackish water in the surficial Biscayne aquifer of central Dade County, Florida, that originated from a flowing artesian well, it was necessary to quantify the rate of spillage and the consequent point-source loading of the aquifer. However, a flow-rate measurement (2,350 gallons per minute) made 2 months after drilling of the well in 1944 was inconsistent with later measurements (1,170 gallons per minute) in 1964, 1965, and 1969. Possible explanations were the: (1) drawdown of the aquifer over time; (2) raising of the altitude at which the water was discharged; (3) installation of 80 feet of 8-inch liner; (4) an increase in the density of the flowing water; and (5) gradual deterioration of the well casing. The first approach to reconciling the measured flow rates was to apply a form of the equation for constant-drawdown analysis often used to estimate aquifer transmissivity. Next, a numerical simulation analysis was made that pro- vided the means to account for friction loss in the well and recharge across vertically adjacent con- fining layers and from lateral boundaries. The numerical analysis required the construction of a generalized model of the subsurface from the surficial Biscayne aquifer to the cavernous, dolomitic Boulder Zone at a depth of 3,000 feet. Calibration of the generalized flow model required that the moddle confining unit of the Floridan aquifer system separating the artesian flow zone in the Upper Floridan aquifer from the Lower Floridan aquifer (the Boulder Zone) have a vertical hydraulic conductivity of at least 1 foot per day. The intermediate confining unit separating the flow zone from the surficial Biscayne aquifer was assigned a much lower hydraulic conductivity (0.01 foot per day or less). The model indicated that the observed mounding of Upper Floridan aquifer heads along the axis of the Florida Peninsula was related to the variable depth of the freshwater and brackish-water zone overlying deeper saline water. The analyses only partly reconciled the two rates. The second rate was accepted as representative of the con- ditions prevailing at the time of its measurement. On the basis of flowmeter logging, it was assumed that an additional 230 gallons per minute escaped through the corroded casing at that time. Factors not amenable to analysis, such as the inherent inaccuracy of the method of estimating flow from the well and possible error in estimating losses through the casing, could easily account for the remainder of the difference between the two measured rates.