Parameter optimization for stimulating semi-arid watershed hydrology.

Abstract

The study was designed to explore the potential of numerical optimization techniques for obtaining unique optimum watershed model parameters. Successful parameter optimization would be valuable for correlations with watershed treatments or for regionalization for ungauged watersheds. A digital watershed model capable of simulating surface runoff hydrographs from small desert watersheds was developed from previous modeling studies. Six parameters were selected for optimization. All other parameters were estimated from previous studies or were measured on the watershed. The model was calibrated with data from a 14 acre watershed in the Tucson Basin. Independent data from the same watershed and data from a similar watershed were used to validate the calibrated parameter values. Calibration was carried out by an optimization strategy of response surface identification, followed by extensive surface climbing and concluding with a quadratic convergent fitting. Optimization results discovered several fitting problems presented by the model but the final results for the calibrated events appeared to be good. In comparison, the validation results were quite poor and indicated a large degree of data dependency in the optimization. It appeared doubtful that the model calibration, even with extensive data, would be able to obtain parameter values capable of correlation with watershed treatments. With model improvements, the approach may provide parameters adequate for regionalization for ungauged watersheds.