Get an overview about the project outputs and related knowledge

D3) Incorporating past performance

Start: 09/2017
End: 09/2021
Status: Active

Contact details

Mark van der Krogt

Delft University of Technology


This project developed methods to improve the reliability estimates for dike slope stability by considering performance information. By incorporating the information of survived (proof) loads and monitoring during the construction of dikes, reliability estimates become more credible, safety assessments improve, and the design of dike reinforcements can be made more efficient. Even when it takes money or risk to obtain the performance information, a strategy with obtaining performance information can be cost-effective, improving the efficiency of flood risk management.

Figure 1 - Top: Examples of slope instability in the landward side with and without flooding due to dike breach (source: Jüpner, 2013 and Landesbetrieb für Hochwasserschutz und Wasserwirtschaft Sachsen-Anhalt). Bottom: Construction of a dike (photo by Pascal Ogink, Combinatie dijkverbetering HOP, p103).

Motivation and practical challenge

Flood safety in the Netherlands is expressed in terms of the probability of typical failure mechanisms as of 2017. Dikes are assessed for safety by (semi-)probabilistic analyses methods. However, these failure probability estimates are dominated (among other factors) by the knowledge about the ground, which is often limited and uncertain. This limited knowledge is particularly the case for slope instability failures at the landward side, which may or may not lead to flooding due to a dike breach (Figure 1, top photos).

Measuring, monitoring and adding information about the past performance of dikes can reduce uncertainty and thus lead to better failure probability estimates. Some of this performance information is available when measuring the soil properties or monitoring existing dikes. Performance information is also gathered during dike reinforcement projects (Figure 1, bottom photo), such as pore water pressure monitoring and measurement of settlement during the dike construction. Unfortunately, not all information is used to assess dikes’ safety or improve estimates of the expected lifetime.

Research challenge

This research aims to develop updating methods to efficiently combine different performance information sources into safety assessments. I focus on dikes with high-reliability requirements, which are designed with a low probability of failure due to the high potential flood damage of flood defences in the Netherlands.

Innovative components

I explore several ways to improve failure probability estimates by incorporating different types of past performance information:

  • Identification of how error sources combine into the total uncertainty in the spatial average of soil properties measured using cone penetration tests (CPTs on figure 2.1).
  • The value of information for proof loading (figure 2.2) and pore pressure monitoring (figure 2.3) by using a decision tree, using a decision tree (joint component with related project A1).
  • How loads during the construction stage (figure 2.4), such as the weight of the soil used to reinforce dikes, can improve reliability estimates for a dike in flood conditions.
  • Further reducing uncertainty by combining observations during the construction (figure 2.4) such as survival, settlement measurement, and pore water pressure monitoring.
  • Development of event trees to estimate the probability of flooding based on several (figure 2.5) successive conditional instabilities.

Figures 2.1 to 2.5 with the type of performance information considered in this project.

Relevant for whom and where?

Researchers interested in probabilistic analysis and organisations wanting more targeted and cost-effective flood protection.

The components of this research are developed for a range of typical dike sections for the Dutch riverine area, and for a case study of a full-scale test embankment in Eemdijk.

Progress and practical application

The research demonstrated that characteristic values of soil properties determined with CPTs could be considerably higher if averaging of random errors is considered. Additionally, we can reduce uncertainty in site-specific transformation models by minimising the distance between boreholes and soil investigation tests (CPTs) and reducing bias in the measurements.

The construction of dikes is a large load on the soft subsoil. Using Bayesian updating, the information of construction survival can be incorporated in the dike reliability assessment. This reduces uncertainty and improves the estimates of the probability of failure. Depending on the situation, the probability of failure can reduce by more than a factor of 10. This reduction is especially significant when the load effects during construction are very similar to the future flood situation that is being assessed, such as for dikes on soft subsoils.

Uncertainty reduction measures such as proof loading and pore water pressure monitoring applied to typical dike sections in the Netherlands have a positive Value of Information, thus are often worth investing. For details about findings, see the related outputs for a detailed description of each finding.

Recommendations for practice

  • For reducing uncertainty in site-specific transformation models for statistically homogeneous geological deposits (on a regional scale), it is better to use multiple boreholes sufficiently distanced rather than multiple measurements within one borehole.
  • Performance information during the construction of dike reinforcements (such as the survival of the construction phase and measurements of settlement after raising dikes) is important information for reducing uncertainties. It should be collected and used for optimising designs during the dike reinforcement.
  • Consider performance information in future dike safety assessments as it may extend the dike reinforcement lifetime, which is particularly interesting in light of accelerated climate change.
  • Deliberately taking a small risk, for example, by adopting a more critical staged loading scheme or proof loading, can obtain a larger reliability update and thus a less costly design. This reliability update is especially important at locations where otherwise expensive (structural) solutions would be required.

Last modified: 12/01/2022

Contributing researchers

Mark van der Krogt

Delft University of Technology

Supervisory team

Dr. Timo Schweckendiek

Delft University of Technology Matthijs Kok

Delft University of Technology

Contributing partners

Proof loading and monitoring to optimize flood defence asset management

We calculate the expected value of extra information obtained from proof loading and monitoring. The additional information improves failure probability estimates for slope stability.

Mark van der Krogt

Delft University of Technology

Wouter Jan Klerk

Delft University of Technology

View storyline

View storyline

Project outputs

Do all dike instabilities cause flooding?

The case example shows that the current practice of assuming that the probability of flooding is equal to the probability of instability is very conservative especially at relatively low water levels.


View publication

Bevat: Conference proceedings

Value of information (VoI) of combinations of proof loading and pore pressure monitoring for flood defences

We calculate the expected value of extra information obtained from proof loading and monitoring. The additional information improves failure probability estimates for slope stability.


View publication

Bevat: Publication open access journal

Improving dike reliability estimates by incorporating construction survival

The approach is exemplified for a range of typical dikes and for a case study of a full-scale test embankment. The main result is that the reliability can increase significantly, especially for dikes on soft soil blankets.


View publication

Bevat: Publication open access journal

Uncertainty in spatial average undrained shear strength with a site-specific transformation model

We estimate the uncertainty of a transformation model to infer geotechnical properties from indirect measurements. When the measurement and transformation errors are random or spatially variable the uncertainty is lower and the inferred properties higher.


View publication

Bevat: Publication open access journal



Reflection: Data-driven dike enforcements - Constructive feedback from new and historical sources

The technical assessment of dike reinforcement projects in the Netherlands is flooded with data. Before, during and after dike reinforcements, a lot of data is collected and stored. This webinar attempts to bridge the gap between conventional and new data sources by presenting new methods for improving dike safety through data and discussing the general use of different data sources.

View event

All events