Through supervision of PhD and MSc students, the output of this project is better methods to assess the strength and performance of dikes using data from past events and experiments for optimising the design of flood defences. With the improved methods, we try to focus more on what is in the subsoil and how this affects dike performance. Overall, we concluded that it is important to look beyond models. The Dutch system is very focused on models, less on a wider view of what makes the dike perform. Dikes will not necessarily fail if budgets are not allocated perfectly but if subsoil properties, inspection and maintenance are not carried out properly.
Figure 1 - Top and bottom-left: Floodwall failure on 17th Street Canal from Hurricane Katrina in New Orleans (source: Reslo et al. 2009 and IPET, 2005). Bottom-right: Rebuild New Orleans floodwall in 2013 (Photo by Bianca Hardeman).
Motivation and practical challenge
Seeing the aftermath of the New Orleans flood in 2005 motivated me to work on dike reliability modelling (Figure 1 bottom-right in 2013). The flood consequences were grave and very impressive (Figure 1 top and bottom-left photo in 2005). Also, for dike reliability modelling here in the Netherlands and abroad, that event made me realise first-hand the difficulties in predicting dike failures and the need to reduce uncertainties.
On the one hand, it showed me failure mechanisms that are rarely observed outside books and laboratories. On the other hand, it showed that modelling these failures involves much more than applying well-known failure models. Very uncertain soil conditions determine the strength of the levee. For example, very small weak zones in the soil proved critical for slope stability. Hence, modellers and designers of dikes should better account for the uncertain factors influencing the dike strength as much as possible.
Figure 2. Components include data from top: the levee failure in Breitenhagen, Germany (source: Weichel (2013)); middle: the Flood Proof Holland backward erosion piping experiment (source: Pol 2018), and bottom: reconstructions around Kinderdjik in South Holland (source: SAFELevee). The piping and reinforcement schemes were adapted from van Beek (2015).
Our research helps in better understanding failure mechanisms for optimising the design of flood defences to better comply with the new flood protection standard. Some of the unique topics that I’m working on as a daily supervisor of the following PhD researchers are (see related projects):
- The temporal development of failure mechanisms. Together with Joost Pol (project D2), we look at the progression rate of piping using full scale and small scale experiments. This temporal development shows how long piping needs to occur to result in flooding along the coast and on riverine areas.
- Method to derive the most likely causes of failure of past breach events. Together with Job Kool (SAFELevee project), we improve the modelling of failure mechanisms via the structured deduction of failure scenarios from before and after data. We tested this approach to find the most likely cause of the failure of the Breitenhagen levee failure in Germany. The method is generically applicable to other locations.
- Optimisation of dike reinforcements. With Wouter Jan Klerk (project A2), we look at various measures to, for example, reduce uncertainties on the soil parameters and implement reinforcement techniques.
Relevant for whom and where?
Other researchers interested in probabilistic analysis and failure mechanisms modelling. Organisations planning the reinforcement of dikes and authorities setting the design requirements.
The research includes key locations in the Netherlands and abroad to use data from past events and experiments in the optimisation of flood defences.
Progress and practical application
For a detailed description of each finding, see the related outputs. The analysis of the 2013 failure on the Breitenhagen dike in Germany shows that the slope instability most likely occurred as the result of an old breach. This old breach probably eroded the soil in front of the reconstructed dike, creating a direct connection between river and aquifer, thereby increasing pore water pressures.
By including temporal progression rates in the failure probability assessment due to piping, the improvements on the dike safety are small for riverine cases, which have long-lasting flood levels. However, the improvements are much larger for the coastal cases, which have short-lasting flood levels resulting in insufficient time for piping to develop fully. There is still a considerable delay in the expected time of piping of several days in the riverine cases, which is beneficial for emergency response. Instead, for coastal cases, piping is less likely to occur with low duration floods.
Finally, our application example for five dike sections along the river Lek in The Netherlands shows that additional monitoring information is only valuable if the expected reinforcement decision is likely to be different.
Recommendations for practice
- Look beyond the models and into the dike, subsoil, history and more for how the dike may feel.
- Old dike breaches and former river meanders are the most critical dike sections.
- Take inspections and animal burrows more seriously.
- Put more effort into understanding piping.
- Case studies should be more central in dike assessment tools.
Last modified: 23/11/2021
Delft University of Technology
Bayesian inference of piping model uncertainties based on field observations
Method to estimate model uncertainty based on failure and survival piping observations.
11/12/2019 by Wim Kanning et al.
Bevat: Conference proceedings
Reliability-based partial factors for flood defences
Calibration procedure of the safety factors to ensure consistency between probabilistic and semi-probabilistic assessments at a cross-section and system level.
20/07/2019 by Wim Kanning et al.
Bevat: Publication upon journal access
A Bayesian hindcasting method of levee failures: The Breitenhagen case
Hindcasting of prior and posterior probabilities of failure using limited observations and photos of the geometry of the slope, with the goal of finding the most likely cause of failure.
03/09/2020 by Wim Kanning et al.
Bevat: Publication open access journal
Forensic analysis of levee failures: The Breitenhagen case
Generic steps to derive most likely failure scenarios from data prior, during and after a levee breach in Germany in 2013.
25/11/2019 by Wim Kanning et al.
Bevat: Publication open access journal
Also applicable to this project
Optimal planning of flood defence system reinforcements using a greedy search algorithm
We develop and validate greedy search algorithm that can find (near-)optimal combinations of reinforcement measures for dike segments. The approach was applied to an ongoing reinforcement project, and it was found to reduce investment costs by ~40% compared to the conventional approach.
Contains: Publication open access journal
Temporal Development of Backward Erosion Piping in a Large-Scale Experiment
Time-dependent reliability in flood protection decision making in the Netherlands
Exploration of the failure probability definition and the influence of including temporal correlation on failure probability estimates.
17/06/2018 by Wouter Jan Klerk et al.//=get_field('category', $output->ID)->name?>
Contains: Conference proceedings
Influence of monitoring on investment planning of flood defence systems
Exploring the value of monitoring for the reinforcement of a dike segment.
11/12/2019 by Wouter Jan Klerk et al.//=get_field('category', $output->ID)->name?>
Contains: Conference proceedings