Comparison of different approaches to quantify the reliability of hydrological simulations

Abstract

The focus of this study was to compare different uncertainty estimation approaches to evaluate their ability to predict the total amount of uncertainty in hydrological model predictions. Three different approaches have been compared. Two of them were based on Monte-Carlo sampling and the third approach was based on fitting a probability model to the error series of an optimized simulation. These approaches have been applied to a lumped and a semi-distributed model variant, to investigate the effects of changes in the model structure on the uncertainty assessment. The probability model was not able to predict the total amount of uncertainty when compared with the Monte-Carlo based approaches. The uncertainty related to the simulation of flood events was systematically underestimated.