Load interdependencies of flood defences

Abstract

In the Netherlands flood risks are often assessed by using a dike ring approach. This approach has, aside from many advantages, also some disadvantages. The main disadvantage is that, when zooming in on a small area, the effect of the system on that small area and vice versa, is or cannot be accounted for properly. Research has shown that for a set of cases in the Netherlands these effects, called ‘load interdependencies’, have significant influence on flood risk estimates. Load interdependencies of flood defences are the effects of failures of dikes on load conditions of other dikes in the system. These can be both positive and negative; a failure causing a lower load at another locations is called positive, a failure causing a higher load at another location is called negative. The positive effects are caused by a decrease in discharge due to outflow into polders, the negative effects are generally caused by overland flow to another river, called shortcutting. In those cases flow and thus water levels increase, causing higher than expected water levels. To be able to assess these load interdependencies in flood risk analysis a different method is needed, as previous research shows and concludes that these effects cannot be ignored. Therefore the objective of this thesis was to develop a new method capable of dealing with these effects for cases at lowland rivers. Based upon previous research a new method was defined based upon a fast inundation model consisting of 1D branches and storage reservoirs. Without losing too much accuracy, this model was 150 times faster than a similar 2D model. Due to the fast nature of this model it was possible to execute a Monte Carlo analysis for the case of the Bovenrijn/IJssel, one of the cases in the Netherlands where load interdependencies have a significant influence on the outcome of risk analysis. By applying Importance Sampling techniques it was possible to execute the Monte Carlo analysis with a reduced number of runs. For the case of the Bovenrijn/IJssel, a schematization with 22 breach locations was used. By using the new method, it appeared that negative load interdependencies can indeed increase water levels at the IJssel, and it is thus necessary to account for these effects in further flood risk analysis. However, not at all locations at the IJssel negative effects are dominating; breaches along the IJssel compensate the negative effect. Along the Bovenrijn it was found that the effects are mainly positive. From the risk calculations it was shown that the risk at the left bank and the dike ring areas further downstream is mainly caused by negative load interdependencies. Other observations were that, although the flood waves are lower, they are also much longer, causing higher load durations. The water levels observed for the different discharge exceedence probabilities are generally much lower, which means that the actual safety of the dikes in the area is also generally higher. They also show a spreading for each discharge, meaning that the water levels are influenced by the uncertainty in strength of other dikes. From the case study it was concluded that the methodology performed quite well and gave interesting insights in the effects of load interdependencies in the area. However, this knowledge should not be used in all cases. For dike design for instance, it is not feasible to account for positive effects, as one dike reinforcement project would change the safety level of all other dikes. However, for calculations of the Maximum Probable Damage and for prioritization of dike reinforcement projects, accounting for load interdependencies can give a huge advantage.