Historically the Netherlands have always had to deal with the threat of flooding, both from the rivers and the sea as well as from heavy rainfall. The country consists of a large amount of polders, which are low lying areas of land protected from flooding by embankments. These polders require an extensive water storage and drainage system to discharge excess water to the surrounding ‘outside water’. Through a large system of ditches water is pumped onto large storage canals which in turn drain in to the ‘outside water’: in the sea or in rivers. In Dutch these drainage canals are called ‘Boezems’. The embankments which enclose the storage canals inside the polders are called Regional Flood Defences. The objective is to determine whether or not the flood risk approach can be applied to a system of regional flood defence systems, given the existing data and ‘state of the art’ methods. The proposed methodology will provide a basis for more thorough assessment of the regional flood defences, not only based on the safety standards, but including the failure probabilities of all relevant geotechnical failure mechanisms and corresponding consequences of flooding. The project focuses on ‘Boezemkaden’, which will be addressed as regional flood defences in the remainder of this document. These flood defences typically retain lower hydraulic heads than primary flood defences. Risk assessment methodology Flood risk is assessed by the annual expected damage due to flooding, which is estimated by multiplying the probability of flooding with the consequences of flooding. The largest knowledge gaps exist in the calculation of probabilities, because several ‘state of the art’ models are available to determine the flood consequences due to breaches in a flood defence system. Probability of flooding The probability of failure of a system of flood defences is determined based on a schematization of the system, which divides it in sections with similar strength properties. Probabilistic methods are used to determine the probability of failure of one section. Flood scenarios are defined as groups of sections which have similar consequences during a flood: these are chosen such that every breach in this group, regardless of its location, will lead to the same flood consequences. To obtain the probability of one flood scenario, the failure probabilities of individual dike sections are combined. The probability of flooding of the whole system can be determined by combining the scenario probabilities. The manner in which the failure probabilities are combined depends on the occurrence of relief in the system: • Relief: if failure of one section results in lower loads on the other sections relief is taken in to account. This can be done in two ways: by assuming that the weakest section fails first or by assuming that the first loaded section fails first; • No relief: if failure of one section does not have any effect on the loads of other sections, no relief is taken in to account. The occurrence of relief strongly depends on the volume of water inside the canal compared to the size of the inundated area after a breach. The extent of relief in canal systems requires careful investigation for every case study. With relief, we can assume only one breach is possible within one canal system. However, when there is no relief multiple breaches are possible. This is an important part of the schematization of the system, as the occurrence of relief has large effect on flood risk. Flood consequences It is assumed that floods resulting from a breach in a regional flood defence system only have inundation depths in the order of decimetres, except for the occasional deeper polders. Due to the low expected inundation depths no loss of life is taken in to account. The economic consequences are determined with HIS SSM and WSS. The HIS SSM model provides an estimate of the economic consequences and loss of life for large floods, with inundation depths of several meters. A disadvantage of the model is that it is inaccurate for low inundation depths. The ‘Water Schade Schatter’ is developed to determine the consequences of small floods due to heavy rainfall in polders. This model uses more accurate consequence functions for low inundation depths of several decimetres, providing more accurate estimations for small floods. A major disadvantage is that the consequence functions are limited to inundation depths below 0.3 meter. To account for larger inundation depths, the consequence functions for buildings specifically have to be changed. Both methods are used and compared in this report; the difference between both estimators can reach up to 20%. No clear distinction can be made of which estimator provides an upper or lower limit. Uncertainties in loads on regional flood defences To determine failure probabilities, insight is required in the statistical distribution of the governing loads, which for regional flood defences are: 1) Hydraulic loads: water levels inside the canals and resulting groundwater level; 2) Traffic loads: vertical loads on top of the flood defence; Waves in these canals are neglected. Currently a research program is undergoing on the stability of peat dikes during droughts. This load is not taken in to account at this stage, because the results of this research are expected to largely influence the assessment of regional flood defences. Furthermore, a case study is chosen for a region where earthquake loading is not present. Hydraulic loads We consider the volume of water in the canals being governed by the inflow from the polder drainage stations and the outflow to the outside water; neglecting rainfall, seepage and local wind set up. These water levels are regulated by the water boards. During extreme rainfall events, the canals have a certain storage volume available for storage of excess water out of the polders pumped on to the canals, which is determined by the difference between the target water level and the ‘drain stop level’; the maximum allowed water level on the canals. Once the ‘drain stop level’ is reached, the polder drainage stations are not allowed to keep pumping water on to the canals. This event may only occur with a probability of 1/100 per year. Whether or not the drain stop is successful depends on the way these are managed. During heavy rainfall events, water boards may have to choose between having to exceed the ‘drain stop level’ on the canals to keep polders dry, or vice versa. The drain stop may fail due to factors which cannot be influenced by the waterboard, or because a certain part of a polder needs to remain dry. The event where the water levels exceed the drain stop level is defined as ‘failure of the drain stop’. Water level observations are used to determine the water level statistic of the regulated system. A Generalized Pareto Distribution is fitted through independent peaks of water levels in the canals. This distribution is modified, to account for the regulation of water levels in the system, by making a distinction successful and unsuccessful drain stop, see Figure 21. Traffic loads The combination of hydraulic loads (high water levels) and traffic loads is governing for the stability of the flood defence. Expert ellicitation was used to determine the statistical distribution of traffic loads. Water board employees responsible for the assessment of the regional flood defences were asked to provide estimates of the 5th, 50th and 95th quantiles of the statistical distribution of the traffic loads. Furthermore, they were asked to provide an estimate of the correlation between the traffic load and water level. The resulting traffic load distribution is shown in Figure 25, for green and grey flood defences. No correlations between the traffic load and water levels was expected with average water levels; the experts all agreed on this point. However, they did not agree on the correlation between the traffic load and the extreme water levels, which was either positive or negative. Different combinations of hydraulic and traffic loads are possible, which depend on the management of the water board. Specifically the policy regarding traffic loads on a regional flood defence determines which combinations of loads are most likely to occur. We determined the probability of failure of the regional flood defences with and without traffic loads. Using this methodology, we showed the influence of the traffic loads on the failure probability and risk of regional flood defences, which is significant. Uncertainties in strength of regional flood defences The following failure mechanisms are governing for regional flood defences: Overflow, Piping and Instability of the inner slope. FORM reliability calculations are used to determine the probability of failure for each mechanism. Overflow will occur when the water levels in the canal exceed the retaining height of the surrounding flood defence. The limit state function of overflow is based on a critical overflow amount, which can lead to erosion of the inner slope and breaching. The stability for piping is calculated with the updated Sellmeijer formula. ‘Hydraulic short circuiting’ is required for piping to develop under regional flood defences. Recent research has shown that hydraulic short circuiting is likely to occur when there is an aquifer below the canals. The response of the water pressure behind the dike to intrusion of water from the canal in the aquifer depends on the thickness of the aquifer. Field tests are required to determine the reduced hydraulic head over the flood defence, due to reduced infiltration of water from the canal to the aquifer. D-Geo Stability is used to determine the probability of failure for inner slope instability. The failure probability of critical slip circles is calculated with Bishop, for several combinations of water levels, piezo metric lines and traffic loads. Only slip circles which will lead to breaching of the flood defence are taken in to account (i.e. slip circles which protrude the crest of the flood defence). Finally, these are combined to obtain the failure probability of instability. Due to the absence of data on probabilities of phreatic lines, we assumed a distribution. We recommend to perform field tests to determine the actual distribution of the phreatic line. Proven strength Proven strength has high potential for updating failure probabilities of regional flood defences. In these canal systems, the difference between average and maximum water levels is very low which results in high potential for proven strength. Especially for overflow and piping the potential is great; the main uncertainty for these failure mechanisms lies in the water levels. The potential of proven strength for the instability failure mechanism is much lower, because not only the water level load determines the stability, but also the phreatic line and traffic loads; these are not always known for the survived load cases. Case study: HHNK Heerhugowaard The ‘Heerhugowaard’ polder is considered in our case study. It is surrounded by two canal systems: the Schermerboezem and the VRNK-boezem. The city of Heerhugowaard lies within the polder, on the Western side. The flood defence system was divided in 17 sections, based on strength properties. The water board made simulations of flooding for a number of breach locations along the flood defence system with Sobek. These were used to schematise the flood defence system in six flood scenarios, each consisting of a group of dike sections (Figure 42). Failure probabilities Overflow: The probability of overflow in this flood defence system is negligible; the retaining height of the flood defences is well above the water levels in the canals, which correspond with very low return periods (< 10-6 per year). Piping: The hydraulic heads over the considered regional flood defences result in rather high failure probabilities. However, there is no direct contact between the water in the canals and the aquifer, due to impermeable layers on the bottom of the canal. Therefore a reduced hydraulic head is taken in to account, which is based on field tests of the intrusion resistance of these layers. When taking the reduced hydraulic head in to account, more accurate failure probabilities are found (considering these flood defences have not failed in the last decennia). Note that these flood defences may still be at risk for piping if the geological profile is changed, for example due to dredging works or erosion of the bottom of the canals. This may expose the flood defence to the maximum hydraulic head, due to intrusion of the canal water in to the aquifer below, which results in high probability of piping. In follow up research, we recommend including the effect of these events. The piping probability can be computed for scenarios with and without the reduced head, and then combined. Instability: The probability of failure for instability largely depends on the combination of the phreatic line and top loads. The influence of the outer water level for a given phreatic line is very low. Traffic loads reduce the reliability of the flood defence considerably. Due to the absence of data on probabilities of phreatic lines, we assumed a distribution based on expert judgement. We recommend performing field tests to determine actual distribution of the phreatic line in the defence more accurately. Several experts have stated that the current approach to for including traffic loads does not model the actual situation correct. We therefore recommend discussing the impact of having to include traffic loads on the strength of regional flood defences more thoroughly. The total failure probability of the regional flood defence system surrounding the Heerhugowaard polder is shown in Table 3. Proven strength We used First Order Survival Updating to update the failure probabilities found for piping. The probability of piping reduced to below 10-6, partly because the probability of water levels higher than the maximum survived water level is very small. However, if we analyse the equations used in this method we conclude that, for this specific case, the method is unrealistic, because the water level uncertainty has little influence on the failure probability (alpha values of 0,05). This results in an error in the formulas, with very low failure probabilities as a result. We therefore recommend to apply an exact method of Bayesian Updating for proven strength, which is described in (Schweckendiek, 2014). This will provide better estimates of the posterior failure probability. Flood risk HIS SSM and WSS were used to compute the consequences of flooding for each flood scenario. The calculated consequences both lie in the same order of magnitude (except for scenario 5, due to large difference in the damage to industry). We assume the WSS model to determine the flood damages for regional flood defences more accurately than HIS SSM, because, in general, these floods have lower inundation depths. The flood risk of each scenario is shown in Table 34. The largest flood risk is determined by scenario 3, or section 11, which has a large probability of flooding combined with high flood consequences. Moreover, the Schermer canal has higher flood risks than the VRNK canal. Cost effectiveness The results of a flood risk assessment for regional flood defences can be used to make cost benefit assessment for interventions in the system. Currently interventions in the system are based on the assessment of regional flood defences, wherein weakest sections are prioritized over stronger sections. However, the weakest sections within a system may very well not be the sections where interventions are most cost effective. Note that the results in this assessment are presented to illustrate the method. The expected total costs of several interventions aiming to reduce flood risk in a system of regional flood defences were compared: - Reducing the hydraulic loads on regional flood defences is not cost effective, as the influence on the flood risk of reducing the drain stop level is negligible; - Restricting traffic loads on regional flood defences can be cost effective, if instability is the governing failure mechanism for the considered section. - Compartmentalization of canals, to reduce the consequences after a flood, can be a cost effective intervention. - Reinforcements prove to be the most cost effective measure. Discussion and conclusions We conclude that the flood risk approach can be applied to regional flood defence systems using the data used in the safety assessment of regional flood defences. The approach not only provides insight in the failure probabilities of the flood defences, but also in the corresponding consequences of flooding and therefore the flood risk. The results can be used to compare the flood risk within the system and prioritize interventions based on the expected risk reduction and cost effectiveness. To obtain more accurate results we recommend investigating how the data obtained in the assessment can be used more effectively in the flood risk approach and/or proven strength assessments. For example, more insight in the relation between the outer water level and rainfall on the phreatic line may provide better estimates of the probability density function of the phreatic line. For these assessments, insights obtained from the water board dike supervisors can play a useful rule. According to the IPO safety standards, the probability of flooding for these flood defence system is required to be 20% of the probability of a drain stop, which is 1/500 per year. The probabilities found for overflow comply with these requirements. However, several sections do not comply with the safety standard for piping and instability, which was also concluded in the safety assessment of the flood defence system. Temporary measures to increase the strength of flood defences during calamities have not been considered in this report. We recommend investigating the potential effectiveness of these measures and comparing this with more traditional reinforcements of the flood defences and consequence reducing measures, such as compartmentalization of the canals. The results of a flood risk assessment of regional flood defence systems can be compared with the results of flood risk from primary flood defences. To do so, more research is required in system behaviour of several regional flood defence systems, as one primary flood defence system often surrounds several of these systems. This report focussed on the flood risk from regional flood defences loaded by canal systems; however, these flood defences are also used to protect polders from flooding from the larger lakes and several ‘regional rivers’ (e.g. the Dommel). The load uncertainty (i.e. water level difference) in these systems is larger, which can result in different conclusions than those obtained in this report. This also requires further research.