University of Twente Student Theses


Estimating and updating uncertainty with the GLUE methodology : Research report for flood forecasting procedures with an application to the Ve river

Putten, D.R. van (2009) Estimating and updating uncertainty with the GLUE methodology : Research report for flood forecasting procedures with an application to the Ve river.

[img] PDF
Abstract:This research was done within the framework of the Bachelor Thesis. The assignment of this research was to establish procedures to estimate and update uncertainty for flood forecasting using the WetSpa model. These procedures had to be made within the GLUE methodology, which takes into account the uncertainty of inputs and parameters of the WetSpa model. This model is a physicallybased distributed hydrological model. The study area for this research is the Ve river basin located in central Vietnam (Quang Ngai province). The uncertainty analysis method used is the ‘GLUE’ methodology, which is an abbreviation for Generalised Likelihood Uncertainty Estimation. The basis of the GLUE methodology, proposed by Beven and Binley (1992), is the premise that all model structures must, to some extent, be in error, and all observations and model calibration must also be subject to error. So there is no reason to expect that any one set of parameter values within a model will represent the true parameter set. When applying the GLUE-method one does not look for the optimum parameter set, but one makes an assessment of the likelihood of many parameter sets in a Monte Carlo analysis. This requires a goodness-of-fit index that must be chosen by the user. The calculated likelihoods are used in a GLUE-procedure to determine the uncertainty. The GLUE methodology also provides the possibility to update the likelihoods when new data become available and to evaluate these new data. The results of this research are two procedures, one for estimating uncertainty and one for updating uncertainty. To this aim six Matlab scripts were designed. Three of these scripts have been designed to calculate the likelihood of simulations in different ways, by Nash-Sutcliffe, Model Efficiency and Error Variance. The advantage of multiple Matlab scripts for a procedure instead of one script for a procedure is that adjustments can be made more easily. So more likelihood measures can be incorporated, the procedures can also be used for other models and study areas, and one can switch easily between simulation mode and forecasting mode. The procedures have been applied to the Ve river basin. Three data sets were available, dealing with three different floods. It was decided to use two data sets in simulation mode to test the estimating uncertainty procedure and the updating uncertainty procedure. The third data set has been used in forecasting mode. There are two main conclusions on the result for simulation mode. First, the hydrological responses of the two floods were not the same, due to parameter Ki. Secondly, the uncertainty bounds calculated by Nash-Sutcliffe were the most appropriate as compared to Model Efficiency and Error Variance. The result in forecasting mode for the third data set was poor, but this could be expected according to Doldersum (2009). He argued that this was due to the semi-open basin, a characteristic of the study area which is not incorporated into the model. Although the result was poor, it proved that the procedures work correctly in forecasting mode.
Item Type:Essay (Bachelor)
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering BSc (56952)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page