Enhancement of design flood estimation using continuous simulation

Publication Type:
Thesis
Issue Date:
2016
Full metadata record
Files in This Item:
Filename Description Size
01front.pdf431.12 kB
Adobe PDF
02whole.pdf9.94 MB
Adobe PDF
Estimation of design flood flow has been a concern for many years in both hydrologic research and in hydrologic practice. Design floods for a given location can be estimated by a number of approaches including analysis of past flood statistics or the use of catchment modelling approaches like design storm methods or continuous simulation. The thesis developed the use of a catchment modelling system for flood estimation conducted for a monsoonal catchment. How feasible and robust it is will depend on the availability of input data, the complexity of the catchment modelling system and the measure of fit used for searching model parameters in the calibration and validation process. The study improved the limitation of rainfall scale by application of disaggregation method to generate sequence of rainfall at an hourly step. Method of Fragment is used to test the hypothesis. The method was validated against 3 temporal rainfall scales namely 1 – hour, 3- hours and 5 hours and quantile flood flow. It was found that using disaggregation rainfall it was possible to improve the catchment modelling robustness for flood quantile estimation. The study introduced time series (AMS) fitting method as a measure of fit. Instead of reproduction of individual hydrographs, the focus of this approach was on reproducing the observed frequency curve. The annual maximum series was used as a model performance in the calibration process. Also, the study used simplified Bayesian framework to estimate errors associated with catchment response and model parameter values in simulation of the AMS. Bayesian technique was applied for the uncertainty of frequency curve and GLUE approach was used for selecting model parameter values. The physical semi-distribution HEC-HMS was applied to join the model inputs to the outputs through these model parameters. As a result the acceptable ranges of model parameter values were identified and fitted by a distribution (mostly normal distribution). It was found that the method is suitable for design flood estimation. The study also highlighted the differences in design flood magnitudes using hydrograph measure of fit compared with annual maximum series measure of fit. For an upstream gauge, it was found that there was an underestimation in design quantile estimations in the meantime this showed high uncertainty for the flow at a downstream catchment gauge. Complexity of catchment modelling system resulted in production of a large number of model parameters that need to be considered. The feasibility and applicability of model calibration for selection of spatially variable control parameters were improved by the introduction of mean and bias coefficients for each parameter. The use of these coefficients reduced a number of control parameters to be considered and, hence, reduced considerable time and effort for parameter values. In conclusion, the application of annual maxima series for design flood estimation is a new approach for catchment modelling parameter estimation. This can produce the flood quantiles without the assumption of neutrality in exceedance probability or of catchment initial moisture condition as can be seen in the traditional method. Furthermore, the uncertainty of modelling system errors is reduced due to the fact that the focus is directly on frequency curves. However, the limitation of the method is that it is suitable only for a continuous simulation approach which needs sufficient of observation data to reproduce flow sequences. The test had been conducted for a monsoonal catchment in Vietnam, further investigation needs to be considered.
Please use this identifier to cite or link to this item: