Over one billion people

are protected against floods, or should be. Flood protection measures consist primarily of levees which, obviously, have to be strong enough to do their job. Yet, levees are different from other large structures like bridges or skyscrapers. For these, strength parameters are well established and can be more easily tested and extrapolated for use in real buildings. To build effective levees, we must learn from levees that have actually failed during real flood events.

The International Levee Performance Database (ILPD) is an online database that, after a four-year development period, contains data from 1529 failure cases in 20 countries. The ILPD is ready to be expanded and applied to support levee research and design around the world. Delft University asked Flows Productions to explain the ILPD to levee managers, assessors, engineers, crisis managers, insurance statisticians, levee data scientists, and all who contribute to the field of flood protection.

What goes into the database ?

The ILPD is not only a platform to share documents about levee failure events, it also has a “big data” structure connecting database entries with an overarching levee failure model. This means that anyone contributing to the ILPD can add documents and videos and complement these with failure-related data (geometric data and failure data, see the figures below) as JSON (excel) files. There are three levels of data and document depth: from basic, mostly geometric, metadata (1), to deeper intermediate “forensic” analyses (2), to highly controlled detailed failure experiment data (3).

For new tabular entries to be included in the database, some pre-processing by an ILPD junior (like a student-assistant or PhD student) is necessary to ensure conformity with ILPD terminology and structure. This ensures that diverse cases are decipherable by all users and, when the library of standardised data grows, allows for macro-level (meta) analyses. The more detailed the data and the more insightful the case, the more interaction with the ILPD team of administrators is expected before the data is made available to users. The overarching levee failure model complies with the scientific fundamentals developed at Delft University of Technology and described in the book Fundamentals of Flood Protection.

To be sure, levee failure understanding will benefit from an expansion of the database with input from all three depth-levels.

The interactive graphic below illustrates standardised ILPD terminology for failure and levee geometry – click on failure mechanisms for an illustration and to highlight contributing factors.

failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion internal erosion other failure mechanisms instability overflow waterside toe level return period water level waterside slope landside slope waterside berm width landside berm width levee core foreshore levee base crest width landside berm level landside toe level crest level instability + horizontal Sliding+ settlement+ micro-instability+ slope-instability failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion internal erosion other failure mechanisms overflow + collision+ earthquake+ drifting ice failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion internal erosion other failure mechanisms instability overflow failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion internal erosion other failure mechanisms instability overflow failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion internal erosion other failure mechanisms instability overflow external erosion + waterside slope+ landside slope failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping internal erosion other failure mechanisms instability overflow internal erosion + piping (backward interior erosion) failure geotechnical hydraulic + weight (loading / unloading)+ material properties + woody vegetation+ animal & human activity+ burrows + river / sea morphology (e.g. lateral movement of the channel) + hydrodynamic forces + changes in environment+ operational failures+ poor design + waves+ tides+ wind+ currents waveovertopping external erosion other failure mechanisms instability overflow

Standardised ILPD failure tree and dike geometry terminology. Click on a failure mechanism to highlight contributing factors and mutual relationships. 

How does the ILPD relate to similar databases ?

In their 2019 study, Özer, van Damme and Jonkman reviewed six levee performance databases, ten dam performance databases, and two flood event databases, six of which are still active. They concluded that currently no large scale, open access and systematically gathered datasets are available for detailed analysis of levee failure processes, development of empirical levee breach models, or validation of process-based models. 

The 2018 collaborative inventory produced by the EUCOLD Working Group on Levees and Flood Defences also stresses this lack of a central levee performance database and the sporadic availability of open-source data at the international level. Where national or regional databases are maintained, most do not focus on failure. Those that include tabulated historical failure information largely lack the depth of geotechnical and hydraulic information necessary for detailed analysis. 

Global databases for dam performance and disasters have been maintained and expanded for decades, serving the international community and illustrating the potential for databases to drive collaboration. With some effort, the ILPD could also play a similar role for levees and levee managers.

DatabaseLeading organizationGlobal coverageOpen AccessActive sinceRelationship to ILPD
China Levee Project Information Management System (CLPIMS)Chinese Ministry of Water ResourcesNoNo2018National database; focused on real-time monitoring, online analysis and early warning
International Disaster Database (EM–DAT)Centre for Research on the Epidemiology of Disasters (CRED)YesYes1988Contains flood disaster data, not specific to levee or dam failure
Italian Levee Database (INLED)Research Institute for Geo-Hydrological Protection (IRPI)NoNo2014National scale; closed access
National Performance of Dams Program (NPDP) DatabaseStanford UniversityLimitedYes1994Focus on in-service performance of dams, failures and other incidents
National Levee Database (NLD)U.S. Army Corps of EngineersNoYes2018National scale; no emphasis on failures
WaterveiligheidsportaalRijkswaterstaatNoNo 2017National scale; focus on monitoring, maintenance
World Register of Dams (WRD)International Commission on Large Dams (ICOLD)YesNo1958Longstanding database with statistics for large dams

Active databases comparable to the ILPD.  Click on database titles to learn more. Adapted from Özer et., al, 2019.

What is the business case for the ILPD ?

A properly functioning and growing ILPD can be maintained and improved for less than 100.000 euro per year – the costs of one senior researcher and a few student-assistants. Ideally, research and educational schemes will be tied into the ILPD, potentially contributing to the total costs.

Constructing, reinforcing or upgrading levees costs between 2 and 20 million euro per kilometer. If the data used and insights gained from the ILPD would save a million euro on a single levee project, assuming the ILPD would cost 100.000 euros a year, the ILPD could be maintained for ten years and have a benefit-cost ratio of 1:1 (without future discounting). Along a similar line, if the ILPD prevented the total flood damages of hurricanes Katrina, Harvey, and Maria (see the table below), it could be maintained for 3.400.000 years!

Dike failure simulations and breach experiments are costly – the recent Dutch IJkdijk test site in the Netherlands for example, cost more than one million Euro. By improving the accessibility of experimental and real failure data, the ILPD cuts two ways: an experiment can build on existing failure data and more organisations can benefit from the findings of the experiment.

Of course, the organisations investing in and benefitting from are not the exact same ones – dealing with this issue may be the ILPD’s biggest challenge.

Ideally, high-income countries with high potential damage but low failure probabilities will be the primary investors. These countries have a great deal to learn from failures elsewhere, since their levees hardly ever fail. All countries benefit, especially those at greatest risk and with a willingness to reduce risks.

Flood risk and defense by the numbers.  Click individual items to view sources.

The Dutch IJkdijk, contributing ILPD data on level 3 (see FAQ 1). The ILPD supports these kinds of tests with background data and with visibility, and may on the long run reduce their need. See dijkmonitoring.nl for more information.

How has the ILPD been used so far ?

The ILPD has supported detailed studies of individual failures and studies of the course of entire floods (with multiple failures).

For example, Kool, Kanning, Heyer, Jommi and Jonkman nvestigated the Breitenhagen levee breach in 2013, supported by insights from other individual levee failures using the ILPD. The Breitenhagen breach was likely caused by unexpected high water pressures due to a connection between a pond and the aquifer, combined with unexpected saturation of the levee and locally weak soil conditions, most probably due to a historical breach at the same location. This teaches the important lesson that locations in levee stretches where breaches have historically occurred require special attention. Possibly, at some point, levee designers find this kind of historical information on the ILPD(!).  

When Özer, van Damme and Jonkman studied the 2002 and 2013 floods in Germany, they found that of all failure data entries, 13% were partial failures (breach height smaller than the levee height), 34% were total failures (breach as high as the levee), and 53% were total failures with scour (breach deeper than the levee). Breach size is often overlooked but crucial, as it largely determines the severity of the flood consequences. In 2002, the main failure mechanism was external erosion due to overtopping and overflow; in 2013, internal erosion and instability of the landside slope – which shows, once more, that levee height (overtopping, overflow) as well as strength (erosion and stability) both deserve equal attention in dike design. 

The team also studied the links between initial failure mechanisms, main failure mechanisms, and the eventual breach characteristics. They concluded that the initial failure mechanism defines the final breach. Furthermore, failures due to macro-instability and internal erosion are less frequent but lead to larger breaches (and thus more severe consequences). In addition, the research team identified that the higher the hydraulic load return period, the higher the breach density (number and length of breaches per kilometer) and made recommendations to discover patterns in breach distributions.

Moreover, the research team identified levee failure rates based on historical floods. This will help to inform future risk assessments, also in data-poor environments.

Findings and recommendations like these will be deepened and complemented as the ILPD grows. Macro-analyses on breach events from different years and different regions, perhaps continents, is one of the main objectives for a potential ILPD research follow-up.

0 total breach width per failure mechanism (m) average breach width per failure mechanism (m) instability overflow / overtopping+ Instability internal erosion+ instability external erosion+ instability overflow / overtopping+ internal Erosion overflow / overtopping+ external erosion internal erosion 0 40 80 120 160 200 2500 2000 1500 1000 500
Results from metadata analysis of the 2002 and 2013 flood events in the Elbe region of Germany showing average (green) and total breach width (grey) for different failure mechanisms (Özer et al. 2019).

How can low-income countries with a lack of information participate or benefit ?

Wealthier countries, like the Netherlands, can afford to build strong levees that will probably never fail. Yet, everyone benefits when data is made more widely accessible. As less wealthy economies grow, it is important that the flood defences that protect their people and prosperity grow with them. 

The Netherlands Environmental Assessment Agency (PBL) forecasts that without additional flood protection, economic damage due to flooding in Sub Saharan Africa, South Asia, and Southeast Asia will increase twentyfold by 2050 (compared to, at most, a fivefold increase in the rest of the world). The ILPD can contribute to the planning and design of smart, resilient flood defences that will be needed to ensure that growth in these regions is sustained.

In areas where monitoring, simulation, and forensic analysis of hydraulic and geotechnical performance is unaffordable, open-source data allows for comparisons between regions with similar climates and landscapes. Broader access to locally appropriate data will be a boon to both local experts and foreign consultants – information on local failures and levee conditions is extremely important, for example, when foreign consultants are asked to collaborate outside of their region of expertise.

Will the ILPD support nature-based levees ?

The idea of nature-based levees (NBL) is that levee engineering involves natural processes and should for ecosystem benefits while contributing to a lower levee failure probability – see the table below for currently available “NBL-technologies”.

Expectations for nature-based levees are high, but they have to overcome a few difficulties. One of these is that the failure probability reduction (and the ecosystem benefits) face serious scientific uncertainties: not only in quantifying the presumed effects once the NBL solution looks exactly as it did on the drawing table, but also in how the nature-based contribution changes over time, in particular during the time it takes until the NBL has achieved its full strength (sediment accretion and growing trees, for example). With systematic performance and failure data, the ILPD can greatly help reduce these uncertainties.

Existing levee design guidelines rely on a body of past knowledge and experience. Nature-based levee technologies start anew, so speeding up the process of building a NBL-performance knowledge base will benefit even more from past and foreign cases than conventional levee practice. For example, Zhu and others discovered that vegetated foreshores greatly reduced breach sizes in the Dutch 1717 and 1953 floods. From the study of the wetlands to the northeast of New Orleans, Dijkman found that wetlands can lead to both a decrease and an increase of the surge level. With enough data, it would be interesting to retrieve hard numbers on how often falling trees have actually caused levee breaches – a common fear among levee managers around the world.

ILPD metadata entry, as described in previous FAQ, does not yet directly incorporate NBL technologies but it is a feasible addition if development on the database continues.

NBL technologyEcosystem benefitsEffect on failure probability
Eco-friendly breakwatersBreakwaters become an “artificial reef” for aquatic lifeA rougher breakwater may dissipate wave energy slightly better than conventional breakwaters
Eco-friendly levee claddingEco-friendly cladding provides shelter for aquatic lifeRevetment will provide roughness and limit wave runup
Vegetated leveesBiodiverse vegetation has a higher ecosystem value than plain grassVegetation may stabilize soil, but blown down trees may initiate breaches; rodents may dig burrows
Vegetated foreshoresVegetated accreting foreshores have a higher ecosystem value than non-vegetated eroding foreshores and much higher than stone cladding or asphaltHigher foreshores reduce wave height and dissipate wave energy; foreshores can be an alternative to revetment; high foreshores reduce breach size; sandy foreshores feed dune growth
Nourished (non-vegetated) sandy foreshoresNourishments momentarily disturb seabed ecosystems but for some time create a more diverse habitatAs above
Levee setbackThe previously embanked land becomes an ecologically valuable intertidal flat or floodplain; the levee location allows for more dynamic morphodynamics (migrating channels)The newly created foreshore may have the advantages and disadvantages above; a levee may be moved away from a potentially levee base-eroding channel (geul)
“Double dikes”The land between the two dikes becomes ecologically richer than beforeThe dikes can help with storm surge attenuation and wave dissipation; the land between the dikes will accrete; two dikes instead of one may be more expensive
Higher dikes for riverbed roughnessHigher allowed riverbed roughness makes room for more river nature dynamics (fluvial forests) More riverbed roughness slightly increases the load on river levees
Fish passages in leveesExpanding the options for fish who migrate between fresh and salt waterDiscontinuities in levees ask for special attention

Nature-based levees: best practices and associated impact.  Table by Ties Rijcken

How can the ILPD be realised ?

The ILPD has been developed until now (end of 2020) as part of the SafeLevee research project at TU Delft, funded by the Dutch Science Council. However, funding support has yet to be established for future development and maintenance. Ideally, there will be one sponsoring organisation to take the lead in expanding the ILPD into a mature platform, to which partner sponsors can relate. Partners of interest include international levee managers such as Rijkswaterstaat, USACE, EA as well as communities like the EUCOLD levee working group.

With the support of sponsors, Delft University of Technology would gladly take the lead in hosting and managing the ILPD. For 100.000 euro or more a year, a postdoc will supervise MSc and PhD students expanding the ILPD, manage promotion and visibility, communicate with sponsors and partners, coordinate research and education tied into the ILPD, and develop training for researchers and students at universities and institutes around the world. An online workshop of half a day should be enough to learn to use the ILPD; two days of training are required to learn to upload data and review uploads by others independently.

Expansion, usage, and scientific findings will be reported once or twice a year to the main sponsors and an independent ILPD board.

Film

Vincent de Gooijer, Ties Rijcken, Rex Steward, Robert Lanzafame and Bas Jonkman

Thanks

Meindert Van, Wouter ter Horst, Rémy Tourment, Martin van der Meer, Matthijs Kok, Richard Jorissen, Ece Özer and Alex Curran

 

PAPER

Ties Rijcken, Rex Steward, Robert Lanzafame and Bas Jonkman

PUBLISHER

Flows Productions, with TU Delft Hydraulic Structures and Flood Risk

INFOGRAPHICS

Eric van Bruggen

Final editing

Cees Oerlemans and Oscar Bradley

Sponsor

Dutch Science Council

Contact

leveefailures@tudelft.nl