The effects of alternative automatic calibration techniques on parameter estimation for HEC-1

Abstract

Four automatic calibration techniques were used to estimate parameters in the U.S. Army Corps of Engineers HEC rainfall-runoff simulation HEC-1. The techniques were tested with three strategies in synthetic data studies and verification studies with a small, semi-arid watershed. The automatic calibrations employed two versions of Newton-Raphson optimization algorithms and weighted least squares (WLS) and maximum-likelihood based estimation criteria. Strategies were determined by the U.S. Army Corps of Engineers (the HEC strategy), and designed to avoid parameter interaction (Methods A and B). Overall, automatic calibration using a multivariate Newton-Raphson optimization based on explicit derivatives and the HMLE with Method B performed best in the synthetic data study in terms of parameter recovery. Verification studies suggest the model may need a soil moisture accounting mechanism to make application to the small, semi-arid watershed under study practical.