2. Load Interdependencies in The Netherlands
Since the 16th century, the compartmentalisation of floodplain areas in the Netherlands has developed primary defences that protect polders from the rivers and sea, and secondary defences that divide the polders into smaller regions [31
]. The primary defences consist of multiple dike rings that have been assigned higher and higher protection standards over the years, due to the occurrence of extreme events and increased economic and societal exposure [32
]. New protection standards to which the primary defences must conform have recently been imposed under Dutch law [33
]; however, it is known that most of the defences do not currently adhere to these standards.
The most recent estimate of current protection levels are given by the VNK2 project [34
], however, the calculated failure probabilities are generally considered to be conservative. Within the study, the probability of breaches along defined sections of dike was used to estimate the overall failure probability of ‘trajects’ or segments of the dike ring defences. The VNK2 project only accounted for load interdependency effects between dike rings 14, 15 and 44 (see Figure 1
), in effect treating this highly developed area as a single dike ring [35
]. However, investigations into load interdependencies by Delft Hydraulics [36
] suggest that negative effects are likely to be significant in a number of regions not addressed in the study, three of which are highlighted in the present study.
The first location for which spatial aspects of load interdependencies may be relevant is in dike ring 43 (the ‘Betuwe’, see Figure 1
, location A). The region is shown to be vulnerable to floods both economically [37
] and with respect to loss of life [38
]. Simulations of breach flows into this dike ring demonstrate how secondary defences delay and compartmentalise the flood waters, causing high water depths upstream of these defences. Studies suggest this effect reduces overall economic risk and further ‘compartmentalisation’ of the region would likely further reduce risk [39
]. Flood water can overflow back into the river system downstream in the dike ring, however failure of this system would likely result in a cascading or domino effect of flows into dike ring 16 (‘Alblasserwaard’) [40
Another potential floodplain shortcut highlighted by the authors of [36
] is dike ring 41, or ‘Land van Maas en Waal’ (Figure 1
, location B). As the name suggests, this region sits between the Waal and Meuse rivers, which converge to within 1 km of each other at the Western end of the dike ring. The dikes on the Meuse are generally lower than those of the Waal in the regions of dike rings 41 and 40 (see Figure 1
below). This, together with the Meuse’s smaller capacity means that large breach flows originating from the Waal could increase flood risk downstream on the Meuse. A probabilistic computational framework for system behaviour analysis of this area was described by Courage et al. [41
]. The authors concluded that load interdependencies are highly significant in the area, and that a framework encompassing the entire system was required for further analysis.
The potential for cascading and shortcutting of breach flows originating on the right bank of the German Rhine and propagating down the IJssel (through dike rings 48–53, see Figure 1
, location C) has been addressed in studies by Klerk [1
] and Bomers et al. [42
]. In both studies, a large redistribution of risk was observed. During high flows on the Rhine, bifurcation control structures convey only ~1/9 of the Rhine flow towards the IJssel, due to the limited capacity of this branch. However, breaches on the right bank of the Rhine, upstream of the bifurcations, have the potential to increase this flow beyond the capacity of the river, should the flows rejoin the system.
This paper analyses load interdependencies in the Dutch river system that include the potential for spatial inundation effects such as cascading and shortcutting. Scenarios are evaluated to model the effects of polder-side and regional defence failures, and the results from the three locations described above (dike ring 41, dike ring 43 and the IJssel valley) are analysed in detail. While load interdependency analysis that include the potential for negative interdependencies have been applied to isolated areas in the Netherlands, it was not previously possible to quantify the effect on overall hazard in the system. Furthermore, by developing scenarios that allow for polder-side and regional defence failures, the effects of a nonstatic floodplain domain can be assessed.
For many protected river systems, the overall and local flood risks computed from large-scale analyses have been shown to be dependent on the ‘load interdependency’ effects that are observed when dike breaching is accounted for. The probabilistic computation required for such studies can preclude the use of sophisticated 2D models, often resulting in certain river–dike–floodplain interactions to be overlooked in the analysis. Even with 2D models, static topographies do not allow for floodplain interactions such as regional and polder-side breaching.
As the floodplains of many large protected lowland river systems are compartmentalised, fast quasi-2d hydrodynamic models are often a reasonable alternative to estimate breach flow inundations. Furthermore, these models can include breaching of secondary defences that delineate the compartments. This paper conducted a load interdependency analysis of the Dutch river system using a fast quasi-2D model that allowed for system-wide and localised risk estimates to be generated and ‘negative load interdependencies’ to be demonstrated. The 2D domain was calibrated on existing breach simulations, and four scenarios were investigated to evaluate the effects of river-side, regional and polder-side breaching.
The case-study results showed that load interdependencies can have a significant impact on hydraulic load distributions and thus failure probabilities in the system, which is in agreement with results from the works of the authors of [1
]. The addition of a fast quasi-2D model to the 1D domain allowed for useful statistics (such as expected inundation volumes), to be estimated at each region, and highlighted areas in which inundation would cause significant pressures on secondary/regional defences or on the polder-side of primary defences. Failure of these defences was explored in two scenarios, and in many regions these failures have a large impact on the volume, level and timing of flood hazards. Specifically, the model highlighted the possibility of: ‘shortcutting’ and ‘cascading’ of flows in the regions of; the German Rhine to the IJssel, dike ring 41 and dike ring 43.
The study is limited in a number of areas. While fast, the quasi-2D approach does not allow for certain hazard variables to be calculated that may contribute to risk, such as flow velocities. The coincidence of river discharges and storm surges and tidal influences is also not currently included. Various assumptions were required for the analysis, such as the simplified thresholds for failure of regional and polder-side defences. The failure probability of the dike sections assumed in the paper (from the VNK2 study, [34
]) are considered to be too high, and many of these sections are currently being reassessed in light of the new safety standards.
These new standards are optimised based on (among other variables) expected economic damage from breaching, but do not include many of the interactions explored in this study. Therefore, calculating the expected economic damage and loss of life resulting from a system behaviour analysis under the new safety standards could be a worthwhile future research for the case study. In general, the approach developed can be used to evaluate risk and develop scenarios for many protected floodplain systems such as the Elbe, Po, Mekong and Mississippi rivers. The results could be of use for decision-making in many large-scale flood risk management applications, for example, when evaluating compartmentalisation or detention area mitigation strategies.