Parameter identifiability for catchment-scale erosion modelling: a comparison of optimization algorithms

The Simplex and the Shuffled Complex Evolution-University of Arizona (SCE-UA) optimization algorithms and the simple least-squares (SLS) and the heteroscedastic maximum likelihood estimator (HMLE) were applied to identify erosion parameters for a process-based model of catchment runoff and sediment yield. Four parameter identification problems were posed with the objective of studying the behaviour of the model under four optimization procedures. Error-free data as well as data with correlated error provided the means for determining the effectiveness of the optimization procedures studied. Results showed that the Simplex algorithm and the SLS objective function provided the best parameter estimates at lower levels of error, while the Simplex algorithm and HMLE objective function provided the best parameter estimates at higher levels of error. This was contrary to the expected theoretical outcome since the SCE-UA algorithm was more successful than the simplex algorithm in locating the global minimum on the response surface. (C) 1998 Elsevier Science B.V. All rights reserved.