Evaluating three fitting criteria for the calibration of the U.S. Geological Survey precipitation runoff modeling system (PRMS)

Abstract

In automatic calibration, a fitting criteria, which is some function of the difference between the observed and the model generated flows, is optimized to get the best parameter set. The purpose of this investigation was to calibrate the U. S. Geological Survey Precipitation Runoff Modeling System (PRMS) model using three different fitting criteria; ordinary least squares (OLS), Ln transformation of the discharges using the OLS on the transformed flows (LOG), and maximum likelihood estimator for the heteroscedastic errors (HMLE). The performance of each criteria in terms of their ability to produce reliable forecasts was examined. The results of the research showed that the winter storms were reproduced best by the parameter sets chosen by the OLS fitting criteria and the summer storms were reproduced best by the HMLE parameter sets. However, the performance in terms of percent bias in different flow groups suggests that HMLE estimator is superior.