Research header

R.J.L. Groenewege (2022)

R.J.L. Groenewege (2022)

Multi-model simulation of wadi flash floods and disaster risk assessment, Utrecht University and Deltares


Download report

Download models:

A wadi (from the Arabic wādī or wād) is a valley or bed, typically of an ephemeral river that forms as the result of heavy rain during a short period in an otherwise year-round (semi-)arid region. Wadi flash floods are a particular type of desecraters regularly and increasingly occurring in the Middle East and North Africa (MENA) region and some countries in the Sub- Saharan region. Their impact can be devastating to the social environment, e.g. due to loss of lives and displacement, and to buildings and infrastructure. Other common problems include siltation of dam reservoirs due to high sediment yields as well as contamination of bays and rivers.

This research internship concerned the two-dimensional modelling of wadi flash floods as well as the assessment and mitigation of disaster risk. The primary objective was to validate and assess the performance of Deltares’ own morphodynamic solver Delft3D Flexible Mesh (FM) for wadi flash floods. For Delft3D FM, Deltares has released a rainfall module and an infiltration module, among many other available options: online bed update, graded sediment (sand + mud), and rocky places representation. This makes the software have many capabilities to capture the highly dynamic flash event (with the unsteady solver) and morphological change at a detailed level. However, the software had yet to be validated in pilot studies. The software also needed to be compared with other similar software that had already been validated in wadi studies. Moreover, many wadi basins are ungauged and lack proper monitoring, so using global data as input and automation of the model setup were evaluated.

Two case studies of wadi systems were selected, for which typical flash floods were modelled as they propagate through the catchment over an initially dry bed. The prime focus of the modelling was schematisation and investigation of the following physical wadi flash flood processes: rainfall-runoff, infiltration, soil erosion, and sediment transport under overland- and stream-flow conditions. This was primarily done in Deltares’ own software, i.e. Delft3D FM and wflow. Modelling results were compared between this software and a reference software suite. HEC-RAS (2D) was chosen as a reference because it has been used successfully for rain-on- grid simulations and cases like wadi flash floods.

It was shown that Delft3D FM and HEC-RAS can be successfully employed to model the hydrology and hydraulics of wadi flash floods in detail. From a hydrological perspective, the most important processes can be included and modelled, i.e. precipitation, infiltration, evaporation, runoff, and flow resistance. Delft3D FM and HEC-RAS may also be used for sediment transport and erosion modelling, but that is less straightforward and requires more calibration, tuning, and stabilisation. Global data that have shown excellent usability, at least for large wadi catchments (>100 km 2 ), are the Copernicus DEM for terrain and wadi basin delineation, ESA WorldCover for land cover and associated terrain roughness, and SoilGrids soil texture for modelling diffuse infiltration. However, the quality and resolution of the global rainfall data were found to be lacking and the main limitation to the exclusive use of global data. In short, two-dimensional wadi flash flood modelling is generally promising, but there are still few straightforward and well-established methods. An important obstacle experienced in this project was that established knowledge and “best practices” of wadi flash flood modelling are relatively limited, which was found to be consistent with the literature, where pilot and validation/verification studies are still prevalent. It is therefore recommended that further work is focused on increasing the reliability and scalability of Delft3D FM for two-dimensional wadi flash flood modelling. This can be achieved by solving of the numerical issues, more validation and calibration studies, and further development of the model setup automation.

It is shown that when the nourishment volume is increased by a certain factor, the effect of the nourishment is increased by a larger factor. Both the maximum increase in bed level and the length over which the nourishment increases the bed level are increased. Nourishing a larger volume of sediment at once is therefore found to be beneficial. When the larger nourishment is distributed into several smaller parts, the length over which the nourishment reduces erosion is increased even further. When the individual parts are of a sufficiently large volume, the maximum reduction of erosion does not decrease.

Finally, it is shown that discharge variability due to uncertainty in the discharge time series does not cause a shift in trend. However, it is able to cause local and temporal changes in bed level of up to 70 cm, which is larger than the nourishments investigated here. That means that nourishments should be evaluated under various discharge conditions.

In practice, it is likely that the characteristics of a sediment nourishment cannot be freely chosen. The knowledge on sediment nourishments gained in this study can be used as a basis to define a nourishment strategy based on the practical limitations and possibilities.