Evaluation and modification of five techniques for estimating stormwater runoff for watersheds in west-central Florida

Abstract

Several traditional techniques have been used for estimating stormwater runoff from ungaged watersheds. Applying these techniques to watersheds in west-central Florida requires that some of the empirical relations be extrapolated beyond tested ranges. As a result, there is uncertainty as to the accuracy of these estimates. Sixty-six storms occurring in 15 west-central Florida watersheds were initially modeled using the Rational Method, the U.S. Geological Survey Regional Regression Equations, the Natural Resources Conservation Service TR-20 model, the U.S. Army Corps of Engineers HEC-1 model, and the U.S. Environmental Protection Agency Storm Water Management Model. The techniques were applied according to the guidelines specified in the user manuals or standard engineering textbooks as though no field data were available and the selection of input parameters was not influenced by observed data. Computed estimates were compared with observed runoff to evaluate the accuracy of the techniques. One watershed was eliminated from further evaluation when it was determined that the area contributing runoff to the stream varies with the amount and intensity of rainfall. Therefore, further evaluation and modification of the input parameters were made for only 62 storms in 14 watersheds. Runoff ranged from 1.4 to 99.3 percent of rainfall. The average runoff for all watersheds included in this study was about 36 percent of rainfall. The average runoff for the urban, natural, and mixed land use watersheds was about 41, 27, and 29 percent of rainfall, respectively. Initial estimates of peak discharge using the Rational Method produced average watershed errors that ranged from an underestimation of 50.4 percent to an overestimation of 767 percent. The coefficient of runoff ranged from 0.20 to 0.60. Calibration of the technique produced average errors that ranged from an underestimation of 3.3 percent to an over estimation of 1.5 percent. The average calibrated coefficient of runoff for each watershed ranged from 0.02 to 0.72. The average values of the coefficient of runoff necessary to calibrate the urban, natural, and mixed land use watersheds were 0.39, 0.16, and 0.08, respectively. The U.S. Geological Survey regional regression equations for determining peak discharge produced errors that ranged from an underestimation of 87.3 percent to an overestimation of 1,140 percent. The regression equations for determining runoff volume produced errors that ranged from an underestimation of 95.6 percent to an overestimation of 324 percent. Regression equations developed from data used for this study produced errors that ranged between an underestimation of 82.8 percent and an overestimation of 328 percent for peak discharge and from an underestimation of 71.2 percent to an overestimation of 241 percent for runoff volume. Use of the equations developed for west-central Florida streams produced average errors for each type of watershed that were lower than errors associated with use of the U.S. Geological Survey regional equations. Initial estimates of peak discharges and runoff volumes using the Natural Resources Conservation Service TR-20 model, produced average errors of 44.6 and 42.7 percent, respectively, for all the watersheds. Curve numbers and times of concentration were adjusted to match estimated and observed peak discharges and runoff volumes. The average change in the curve number for all the watersheds was a decrease of 2.8 percent. The average change in the time of concentration was an increase of 59.2 percent. The shape of the input dimensionless unit hydrograph also had to be adjusted to match the shape and peak time of the estimated and observed flood hydrographs. Peak rate factors for the modified input dimensionless unit hydrographs ranged from 162 to 454. The mean errors for peak discharges and runoff volumes were reduced to 18.9 and 19.5 percent, respectively, using the average calibrated input parameters for each watershed. Initial estimates of peak discharges and runoff volumes using the U.S. Army Corp of Engineers Hydrologic Engineering Center-1 model, produced average errors of 105 and 26.8 percent respectively, for all the watersheds. Curve numbers and lag times were adjusted to match estimated and observed peak discharges and runoff volumes. The average change in the curve number for all the watersheds was a decrease of 2.5 percent. The average change in the lag time was an increase of 169 percent. The mean errors for peak discharges and runoff volumes were reduced to 5.8 and 1.4 percent, respectively, using the average calibrated input parameters for each watershed. The observed and estimated peak discharges and runoff volumes could be matched by adjusting curve numbers and lag time using the U.S. Army Corp of Engineers Hydrologic Engineering Center-1 model; however, the shape of the estimated flood hydrograph and timing of the peak could not be matched. The input dimensionless unit hydrograph must also be changed to increase the accuracy of the Hydrologic Engineering Center-1 model for watersheds in west-central Florida. The source code has to be modified and recompiled to enter different dimensionless unit hydrographs into the HEC-1 program. During application of the U.S. Environmental Protection Agency Storm Water Management Model, two separate infiltration methods were evaluated. Initial estimates of peak discharges and runoff volumes produced mean errors of 46.5 and 6.8 percent, respectively, for all watersheds using the Green-Ampt infiltration method, and 48.8 and 9.5 percent, respectively, using the Horton infiltration method. The mean errors were reduced to 18 and 0.3 percent for the Green-Ampt method and 20.9 and 7.2 percent for the Horton method using the average calibrated input parameters for each watershed. Estimates of peak discharges and runoff volumes were initially made for watersheds in west-central Florida using recommended procedures, then compared to observed peak discharges and runoff volumes. Subsequently, the procedures were modified to increase accuracy for this area. The same methods used during the study could be used in other parts of the world to evaluate the accuracy of standard methods for estimating stormwater runoff.