The Impact of a New Flood Protection Approach on the Failure Probability of the Primary Flood Defence System Along the Waal

Abstract

The current flood protection strategy in the Netherlands, known as the protect open approach, is increasingly challenged by climate change and sea level rise. Within this approach, river dikes are heightened while maintaining an open connection with the sea. The primary flood defence system along the Waal has been strengthened until 2080 using the current flood protection approach. However, under future extreme climate scenarios, higher river discharges and compound river-coastal flooding are expected to place growing pressure on the primary flood defence system. This research investigates an alternative strategy for the long term: the advance-closed approach, in which the Delta21 concept is implemented. The advance-closed approach combines parts of the advance and protect-closed approaches as defined by the the sea level rise knowledge program. At the Haringvliet outlet there is an advancement towards sea by building Delta21 and the new waterway is closed off by a sluice following the protect-closed approach.

In the Delta21 concept an energy lake in build seaward of the Haringvliet outlet. During extreme river dischargers or sea conditions the Delta21 storm surge barrier can be closed off and water can be pumped out of the river system into the energy lake and subsequently into the sea.

The main objective of this study is to assess to what extent the advance-closed approach can reduce the probability of failure of the primary flood defence system along the Waal River, from Tiel to the Haringvliet outlet, compared to the current protect-open approach. The analysis focuses on extreme conditions under the KNMI’23 high emission, wet climate scenario for the year 2150.

The hydraulic response of both approaches is simulated using the one-dimensional hydrodynamic model SOBEK. The results of these simulations are translated into probabilistic water level distributions using a discrete Bayesian Network. Discrete Bayesian networks are probabilistic graphical models representing the joint distribution of a set of variables. The Bayesian Network first calculates the load on the primary flood defence system by calculating the water level statistics using the conditional probabilities to the boundary conditions of the hydrodynamic model. By applying the Markov blanket principle, the Bayesian Network remains transparent and computationally efficient while capturing the dominant dependencies in the system.

This probabilistic framework allows hydraulic loads and dike strength to be combined explicitly. The dike strength is defined by dike failure mechanisms such as backward erosion piping, macro-instability, and overflow which are incorporated through fragility curves derived from OKADER.

The results indicate that the advance-closed approach leads to lower water level exceedance prob- abilities and a significant reduction in the probability of dike failure along the Waal from Tiel to the Haringvliet outlet. Even in the far upstream, discharge dominated region, does the advanced-closed approach reduce the water levels compared tot the protect-open approach. For validated dike rings, the failure probability is reduced by factors ranging between 102 and 104 compared to the protect-open approach. This reduction is primarily attributed to the additional storage and pumping capacity provided by the Delta21 system during extreme river discharges and storm conditions. Although failure of the spillway and pump system of Delta21 results in high water levels, these scenarios have very low probabilities and therefore contribute only marginally to the overall failure probability.

In addition to quantifying failure probabilities, the Bayesian Network enables back-tracing of dike failure to specific hydraulic conditions and failure of individual flood barriers. This makes the approach a valuable tool for identifying vulnerabilities within complex flood-protection systems. At the same time, uncertainties remain, particularly in the representation of dike strength, assumptions of conditional independence, discretisation choices, and the omission of time as an explicit state variable which is especially relevant for pump reliability.

Despite these limitations, the study demonstrates that the advance-closed approach is a promising long-term strategy for reducing flood risk along the Waal under future extreme climate scenarios. Further refinement of hydrodynamic modelling, improved definition of failure mechanism probabilities, and explicit inclusion of time dependent processes are recommended to strengthen the robustness of the conclusions.