A model parameter set for use with the Hydrologic Simulation Program FORTRAN watershed model was developed to simulate storm peaks and storm volumes for the 28 subbasins of the West Fork Trinity River Basin upstream from Lake Worth, northwest of Fort Worth, Texas, from the calibration and testing of 5 gaged subbasins. These parameters can be transferred to the 23 ungaged subbasins. The model simulates storm runoff for a channel-routing model that can be used to improve reservoir operation during floods in the basin.

Rainfall and runoff data were collected from October 1, 1992, to September 30, 1994. A total of 55 storms were recorded at the 5 streamgage stations during the 24 months. Twelve different pervious land segments were defined based on types of soil, land cover, and watershed slope. A total of 20 process-related parameters were defined for each land segment, and 6 basin-related parameters were defined for each stream reach.

The mean absolute errors for the 5 subbasins for simulation of storm peaks range from 48.0 to 470 percent and for simulation of storm volumes range from 34.4 to 416 percent. A sensitivity analysis was done to determine what a change in a parameter value has on the largest storm peak and on the total storm volume. The model then was recalibrated and tested on the basis of the analysis of the sensitivity of parameters and on the analysis of the errors from the initial model calibration and testing. The mean absolute errors for the 5 subbasins using the recalibrated parameters for simulation of storm peaks range from 47.1 to 297 percent, and for simulation of storm volumes range from 27.6 to 193 percent.

The model produced better results for simulation of the larger storm peaks and storm volumes than for simulation of the smaller storm peaks and storm volumes, especially after an extended period of no runoff. The same range in errors can be expected when transferring the parameters to the 23 ungaged subbasins. Additional data collection and model refinement could decrease the range of expected model errors. More storm data and improved discharge rating curves could result in model parameters that account for the wide seasonal variations in runoff in the study area.