A. Petsiou, Analysis of uncertainty of flood routing parameters - Application to the Pinios river , Diploma thesis, 111 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, Athens, 2017.
Flood forecasting and impact assessment is a multidisciplinary topic of high importance, spanning over meteorological, hydrological, hydraulic and environmental sciences and associated technologies. The EU Directive 2007/60 on the assessment and management of flood risk requires establishing flood hazard and flood risk maps showing the potential adverse consequences to human health, the environment, cultural heritage and economic activities. These should be accompanied by measures to manage floods and mitigate the frequency of potential flood damages. These should be accompanied by measures for flood management and mitigation of the frequency of the potential losses. However, floodplains dynamics is affected by various uncertainties such as climate variability, the hydro-morphological conditions, the land use and the demographic changes. It includes, namely, large uncertainty, derived from modelling assumptions and simplifications. Usually, flood modelling comprises two components, the hydrological simulation, which quantifies the size, duration and probability of the flood event and the hydraulic simulation using the dissemination of flood wave in the river channel and mapping of inundated areas. Inherent uncertainties are present in multiple aspects of the above approaches, involving the model structure, model parameters, boundary conditions and input data. However, most of such uncertainties, which may be surprisingly large, even in small basins, are usually neglected. In fact, in most applications, particularly of the everyday engineering practice, flood models are considered as fully deterministic tools, in which the unique expression of uncertainty is the return period of rainfall. Consequently, the hazard maps exported by hydraulic simulation models contain many uncertainties in the structure, data and parameters. In this thesis, we focus on the uncertainty derived from the river parameters, i.e. the channel’s geometry and the roughness coefficient that in most simulations are considered as constants or as uniformly distributed. The latter assumptions often do not correspond to reality. Therefore, by using a Monte Carlo approach, we perform an extensive analysis by simultaneous changing the river’s depth, width and roughness coefficient. Through literature’s review carried out, we can conclude that the uncertainty derived from sources associated with the river is quite large, especially in cases of limited measurements or insufficient information on the geometry of the river and the surface material. It is worth noting that the uncertainty analysis according to the references showed that the discharge (inflow) appears to have a long-term memory. In this way we aim at developing a generalized stochastic framework for estimating uncertainty arised by hydraulic simulation models. Such an integrated approach would ensure more consistent estimations of the total flood risk and the associated range of variability of flood inundation maps, especially for areas that they lack of data or they contain data with large uncertainty (topographical measurements) but will also enable a more accurate estimation of the potential financial losses related to such floods.