Improved parameter estimation for hydrological models using weighted object functions

This paper discusses the sensitivity of calibration of hydrological model parameters to different objective functions. Several functions are defined with weights depending upon the hydrological background. These are compared with an objective function based upon kriging, Calibration is applied to piezometric readings from the Isle of Goeree in the Netherlands. For a study on the permeability of the first aquifer, the kriging predictor yields weights that differ from using prior knowledge, and emphasizes more strongly spatially isolated points than commonly applied objective functions. It reduces the range of differences between measurements and model simulations, but the mean absolute error increases. For a study on the resistance of the top layer and of the aquitard, use of prior information in the objective functions leads to a reduction in standard deviations of the differences between measured and calculated values by 40-80%, Copyright (C) 1999 John Wiley & Sons, Ltd.