Y. Moustakis, Pseudo-continuous stochastic simulation framework for flood flows estimation, Diploma thesis, 215 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, July 2017.
Typically, flood modelling in the context of everyday engineering practices is addressed through event-based deterministic tools, e.g., the well-known SCS-CN method. A major shortcoming of such approaches is the ignorance of uncertainty, which is associated with the variability of soil moisture conditions and the variability of rainfall during the storm event. In event-based modelling, the sole expression of uncertainty is the return period of the design storm, which is assumed to represent the acceptable risk of all output quantities (flood volume, peak discharge, etc.). In the meantime, the varying antecedent soil moisture conditions across the basin are represented by means of scenarios (e.g., the three AMC types by SCS), while the temporal distribution of rainfall is represented through standard deterministic patterns (e.g., the alternating blocks method). Furthermore, time of concentration is considered as a constant characteristic feature of a basin, which has actually been proved to be an invalid assumption. In order to address these major inconsistencies, while simultaneously preserving the simplicity and parsimony of the SCS-CN method, we have developed a quasi-continuous stochastic simulation approach, suitable for ungauged basins, comprising the following steps: (1) generation of synthetic daily rainfall time series; (2) update of potential maximum soil retention, on the basis of accumulated five-day Antecedent Precipitation; (3) estimation of daily runoff through the SCS-CN formula, using as inputs the daily rainfall and the updated value of maximum soil retention;(4) daily update of the value of time of concentration according to the runoff generated; (5) selection of extreme events and application of the standard SCS-CN procedure for each specific event. This scheme requires the use of two stochastic modelling components, namely the CastaliaR model, for the generation of synthetic daily data, and the HyetosMinute model, for the stochastic disaggregation of daily rainfall to finer temporal scales. Outcomes of this approach are a large number of synthetic flood events, allowing for expressing the design variables in statistical terms and thus properly evaluating flood risk. The proposed quasi-continuous stochastic simulation framework, along with a series of model variations is thoroughly investigated, in order to examine its response, prove its consistency and suggest further improvements and topics for future work.
Full text (13373 KB)
This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.