Evaluation of a watershed model to simulate sediment transport in a small agricultural watershed in Indiana

The streamflow and sediment components of the watershed model, ' Hydrologic Simulation Program-Fortran ', were evaluated using 2 years and 9 months of data from a 2.7-sq mi agricultural watershed in Indiana. Hydrographs of simulated sediment concentration generally have smaller peaks and longer recessions than do hydrographs of measured data. The attenuation of simulated peaks was partly caused by difficulties in designing a representative channel system. Simulated surface runoff during low-intensity rainfall was greater than measured surface runoff; this caused an overestimation of sediment concentration and discharge. Rainfalls of low intensity were mostly absorbed by the soil, but the model predicted that the amount of surface runoff from such storms would be large. The overly large volume of simulated surface runoff transported more sediment than actually occurred. The mean absolute error of mean sediment concentrations during storms is 45% of the measured average concentration. The error resulted from a combination of error in recording streamflow and rainfall, in simulating streamflows, and in simulating sedimentation processes. The largest percent errors are associated with simulated maximum streamflows and average sediment discharges. The root-mean-square errors are 93 and 102% of the measured maximum streamflow and average sediment discharges. (USGS)