A modified version of the Georgia Tech Watershed Model was applied for the purpose of flow simulation in three large river basins of west-central Florida. The model was calibrated for six streamflow stations located in these basins using 4 years of historical and current rainfall, runoff, and estimated evapotranspiration data. Watersheds modeled range in size from about 70 to 650 square miles.

Calibrations were evaluated by comparing the following synthesized and observed data: annual hydrographs for the 1959, 1960, 1973 and 1974 water years, flood hydrographs (maximum daily discharge and flood volume), and long-term annual flood-peak discharges (1950-72).

Annual hydrographs, excluding the 1973 water year, were compared using average absolute error in annual runoff and daily flows and correlation coefficients of monthly and daily flows. For stations used in the study, average absolute errors in simulated runoff range from 9 to 21 percent and errors in daily flows range from 48 to 71 percent. Correlation coefficients for monthly flows range from 0.81 to 0.95 and correlation coefficients for daily flows range from 0.68 to 0.87.

Correlation coefficients for simulated and observed maximum daily discharges and flood volumes used for calibration range from 0.91 to 0.98 and average standard errors of estimate range from 18 to 45 percent. Cor-relation coefficients for simulated and observed annual flood-peak dis-charges range from 0.60 to 0.74 and average standard errors of estimate range from 33 to 44 percent. The number of flood events used for calibration varies for each streamflow station, but range from 6 to 18. The number of annual flood-peak discharges used also vary but average about 20 for each station.

On the basis of these results, it is concluded that flood calibrations have been achieved for stations used in this study; however, because of data limitations, calibrations and prediction errors cannot be completely verified until additional rainfall, runoff, and evapotranspiration data become available.