Parameter Uncertainty of a Watershed Model

Abstract

Parameter uncertainty of the Hydrologic Simulation Program Fortran (HSPF) was the focus of this work. The model was calibrated using data from the South Nation watershed located in eastern Ontario. Up to 18 parameters were used in independent model calibrations. Three uncertainty methods were used: the method of moments, Monte Carlo (MC) with Latin hypercube sampling (LHS) and induced correlation, and response surface methods. It was observed that as more parameters are simultaneously optimized higher uncertainty is typically observed. The method of moments shows the most conservative uncertainty. The largest uncertainty typically was estimated by the MC method. The 95% confidence intervals of parameter uncertainty correspond to up to 10% variations in spring maximum flows. © 2008, Taylor & Francis Group, LLC.