Flood and low flow risk

A statistical analysis on the rainfall and the use of derived synthetic rainfall input will not lead to an accurate probability estimation of the hydraulic parameters in case of non-linear systems (i.e. river catchments). The reason is that the frequency of the effect (i.e. river state) is not equal to the frequency of the input (i.e. rainfall). Therefore, continuous simulations are necessary for the hydrological part of a river model. The rainfall runoff discharges can then be processed to determine representative hydrographs for the river routing (for a certain return period). This processing must be based on Quantity/Duration/Frequency-relationships (QDF) in order to take into account the downstream system behaviour (i.e. river routing). Based on these QDF-relationships, a series of representative hydrographs can be selected or composite hydrographs can be created. The processing of the selected hydrographs leads to large practical problems, while the use of composite hydrographs is an efficient and accurate alternative. By means of these composite hydrographs, river states with the same return period at all places can be calculated by one simulation.

The climate change impact on the risk of hydrological extremes along rivers and urban drainage systems is studied for the local hydro-climatologic conditions in Belgium in an on-going research project (CCI-project) for the Belgian Science Policy authority / Federaal Wetenschapsbeleid). For rivers, both floods and low flows are considered, while for urban drainage systems only flood extremes are of relevance. The research includes the study of climate change scenarios, the statistical analysis of trends and cycles in long-term series of historical rainfall, evapo(transpi)ration and river flow, and verification of the consistency of the climate change scenarios with the present and past climate, and the impact modelling towards flood risk and low flow risk along rivers and urban drainage systems.

The first version of climate change scenario's derived in the CCI-HYDR project have been applied in a parallel research project for the Flanders Hydraulics Research (WL Borgerhout) water administration of the Flemish Government. The hydrological impact has been investigated for three subbasins in the river Dender basin in Belgium.

More information: Website CCI-project on 'Climate change impact on hydrological extremes along rivers and urban drainage systems'.

<top>

Using a quasi 2D hydraulic floodplain model, in combination with lumped conceptual models for the different subcatchments in the river basin and a hydrodynamic model for the river, historical flood events can be simulated to validate the model. Using composite hydrographs, also representative flood events for various return periods can be simulated. Based on a digital elevation model (DEM) and a GIS system, the spatial extent of these flood events can be visualised.

During the quasi 2D flood model implementation, the GIS system is used to draw geometrical data from the DEM. It is also used to demarcate the potential flooding areas (areas to be modelled in a quasi two-dimensional way) and to visualize the spatial extent of the modelled floods. In this way, the GIS system is applied both as a pre-processing tool and a post-processing tool. The communication between the hydrodynamic model and the GIS system thus acts in two directions. An evaluation of this approach of river flood mapping has been made by the practical application of the river modelling package MIKE11 (Danish Hydraulic Institute), in combination with ArcView / MIKE-GIS, to the river Dender in Belgium. This work is done with support of the Flanders Hydraulics Research (WL Borgerhout) water administration of the Flemish Government. The GIS tasks are supported by a collaboration with Spatial Applications Division Leuven (SADL).

Because the results are succesful, the methodology is selected by the WL administration as standard approach for flood probability mapping and on the basis of flood risk calculations.

For calibration of the floodplain model and the flood mapping results, use is made of historical flood information. In the past, only information was available about the maximum spatial extent of the historical floods from the last 12 years. This information was collected by SADL within the framework of a project for the Ministry of the Flemish Community. Based on this flood information, the 'map of recent floods' was derived.

In the FAME project for the European Space Agency (ESA), additional use was made of satellite derived flood maps. That project demonstrated improvements in flood modelling performance by use of earth observation products (both radar-based ERS SAR and ENVISAT ASAR images for flood mapping, and Landsat ETM+ and IKONOS imagery for land use mapping and damage assessment).

Methodologies for flood risk assessment are worked out and/or compared in collaboration with Probabilitas Statistics and Risk Analysis Consultants (Prof. J. Van Dyck). A probabilistic approach will be used to combine the information on (1) the probability of flooding, (2) the spatial extent of floods for different severity levels (or return periods), and (3) the consequences of these floods (e.g. damage assessment). The research will make maximal use of the procedures worked out before by Probabilitas and the Hydraulics Laboratory for the insurance industry and the Flanders Hydraulics Research water administration. In the probabilistic approach, the most dominant uncertainty sources will be considered. Some uncertainty sources (e.g. intrinsic uncertainties increasing the temporal variability of the physical processes considered) will increase the flood risk, while other sources (e.g. statistical uncertainties on the model parameters) explain the statistical uncertainty on the flood risk. By considering the most dominant statistical uncertainty sources, an estimate is made of the order of magnitude of the uncertainty on the predicted flood risk.

The uncertainty considerations are also useful for the decision maker. The probabilistic results will allow a final decision to be developed based on the level of risk the decision-maker is willing to assume.

Flood forecasting models are developed for operational use at the Flanders Hydraulics Research (WL Borgerhout) water administration of the Flemish Government. These models predict in real time discharges and water levels at selected locations along the river, based on historical and forecasted rainfall and hydrological and hydraulic simulation models. Instantaneous model errors are forecasting for future time steps through a data assimilation procedure (model updating). The real-time forecasting models are implemented in the FloodWatch software of DHI Water & Environment. The real-time predictions are used by the WLB on the basis of the real-time managements of floods and low flows.
The research project includes evaluation of the accuracy of the real-time forecasting results for different time horizons through statistical performance indices. As part of this evaluation, the error structure (e.g. time correlation of the model residuals) is analysed, and the updating scheme of the model evaluated. Different types of updating schemes can be used, but their efficiency strongly depends on the model error structure, which can vary from river to river.

Techniques for Model Predictive Control (MPC) are implemented and tested for the real-time control of reservoirs (flood control applications). The reservoirs "Schulensmeer" and "Webbekom" along the river Demer upstream of the city of Diest in Belgium are used as test case. For this case, a real-time flood forecasting model is available: the so-called OBM Demer model. It is implemented in the FloodWorks software. The project is running through a doctoral scolarship funded by the Division Water (Afdeling Water) of the Flemish Environment Agency VMM. It is carried out in cooperation with the research unit ESAT-SISTA of K.U.Leuven.

The FRIEND/Nile project is one of the FRIEND projects of UNESCO, focusing on the hydrology in the Nile basin. The Flemish Government is donor in the project through the Flanders in Trust fund. The FRIEND/Nile project has components on Flood Frequency Analysis, Drought and Low Flow Analysis, Rainfall-runoff Modelling, Sediment Modelling, Eco-hydrology, and Integrated River Basin Management. P.Willems is the Flemish counterpart for the local research done in the Nile basin for the project components on Flood Frequency Analysis and Drought and Low Flow Frequency Analysis. In these two components, regional curves are developed for flood and drought or low flow frequency estimation in the river Nile basin. A first version of regional curves has been set up in the first phase of the project (2001-2005) and supported by M.Sc. theses at K.U.Leuven. They are based on at-site calibrations of flood and low flow frequency distributions and low flow QDF relationships for a large number of stations along the basin. To enable regionalisation analysis to be carried out, correlations have been investigated between the distribution's parameters and catchment characteristics such as the area of the catchment, topographical parameters, land use properties, etc. Based on the regional curves, predictions can be made of the high and low flow levels for given time scales and given return periods at ungauged sites in the basin. They can be used to estimate the return period of floods and water shortage problems for different types of water uses and water management applications, such as drinking water supply, agricultural applications, hydropower production, etc. The curves were, however, based on the assumptions that the data series are stationary without trends (e.g. no influence of climate change). In the second phase of the study (2006-2009), these trends will be investigated and the regional flood and low flow frequency curves updated accordingly.