Author(s): M. Werner; S. T. Khu
Keywords: Uncertainty; Monte carlo simulation; Muskingum routinggenetic algorithms
Abstract: Monte Carlo Simulation methods are widely applied for estimating predictive uncertainties in hydrological models due to model parameterisation. One such method is the Generalised Likelihood Uncertainty Estimator (GLUE) approach, where a likelihood value is assigned to a parameter set depending on how well the model can be considered behavioural. A disadvantage of the Monte-Carlo simulation approach is that it is requires multiple model evaluations for determining a reliable estimate of the uncertainty. Typically a large number of the model runs show poor results and can be disregarded in further analysis. In this paper a method is presented using a multi modal genetic algorithm to distinguish interesting parameter subspaces prior to application of the GLUE approach for estimating the uncertainties. The method potentially saves considerable computational effort and time, and is shown to determine similar uncertainty bounds. The method is first discussed and then demonstrated using a simple Muskingum routing model.