Effectiveness and reliability of emergency measures for flood prevention

Abstract

Floods in the summer of 2013 in Central Europe demonstrated once again that floods account for a large part of damage and loss of life caused by natural disasters. During flood threats emergency measures, such as sand bags and big bags, are often applied to strengthen the flood defences and attempt to prevent breaches. Although these measures are often used there is limited insight in the actual reliability of the measures and their effectiveness in increasing the safety of the flood defences. The objective of this research is to develop methods to analyse the reliability and effectiveness of emergency measures for flood defences. Attention will be paid to the quantification of the reliability of emergency measures through an extensive risk analysis. When including emergency measures in the reliability analysis of flood defences failure is defined as failure of both the emergency measure as well as the flood defence. To determine the failure probability of flood defences with emergency measures two assessments are made: 1) First the probability of failure of the emergency measure is determined and 2) second the effect of the emergency measures on the failure probability of the dike section. So even when emergency measures are successfully applied the dike could still fail (!). The reliability of emergency measures is determined with event and fault tree analyses. Ad 1) The probability of a correct functioning control and/or emergency measure depends on the completion of three phases: Detection – Placement – Construction. The system is modelled in an event tree: it forms a series system which functions when each event is completed on time and correctly. Ad 2) During the operational phase, when emergency measures are placed correctly, these will reduce the failure probability of the dike section. This reduction is determined with sensitivity analyses together with project VNK. For piping the effect of reducing the hydraulic head over the flood defence is calculated in steps of 0.5 meter. For overtopping the effect of filling up local ‘dents’ (i.e. spots with less elevation than the surrounding flood defence) in the flood defence height is determined. Overtopping measures only effectively reduce the failure probability of the dike section for water levels close to the crest while piping measures could potentially reduce the failure probability at lower levels compared to the crest height. An important aspect in the reliability assessment is the length effect; the longer the flood defence the higher the probability of it having a weak spot. In this report two types of length effect are treated: (1) The length effect of the flood defence (failure mechanism) and (2) the length effect of the emergency measure. With increasing amounts of weak spots along a flood defence the contribution of a system of ‘control’ and/or emergency measures to the reliability will then decrease. The length effect determines to a large extent the feasibility and type of emergency measure. Results case study dike ring 53: ‘Salland’ The framework developed is applied to a case study at the Dutch water board Groot Salland, for dike ring 53. According to VNK this dike ring has a high probability of flooding (>1/100 per year) as a result of a high vulnerability for piping (Piping probability of 1/63 per year) (Dijk & Plicht, 2013). The water board acknowledges the problems with piping as it is known that along several parts of the dike sand boils occur during high water on the river. Sometimes even boils occur at locations not known beforehand. The data sheet is used to determine the failure probability of such a system of ‘control’ or emergency measures. The failure probability for piping measures in dike ring 53 is estimated at 1/3 per event. Taking the effectiveness of the measures in to account this resulted in a decrease of the failure probability of the section with a factor 1.2 to 2.7. At dike ring level the failure probability is reduced to 1/120 per year, a factor 1.9. This validates the statement made that with increasing length (number of weak spots) the contribution of a system of emergency measures to the reliability of the flood defence decreases. The failure probability of the system depends largely on the probability of detecting weak spots in the dike. The reliability of the detection phase is influenced by the knowledge and experience of the detection personnel, but also by the weather conditions and visibility. The overtopping failure probability of the dike ring is estimated by VNK at 1/610 per year (Dijk & Plicht, 2013). The contribution of increasing local ‘dents’ in the dike is also determined. For these sections a failure probability is found of 1/9 per event. Together with the effectiveness this resulted in a reduction of the failure probabilities of the dike sections with a factor 2 to 6. This resulted in a failure probability of the dike ring with emergency measures of 1/3000 per year, a reduction with a factor 3.6. The failure probability of measures against overtopping is determined largely by the probability of detection of weak spots and the probability of correct placement of the emergency measure (sand bags). Both analyses show that overtopping measures are more reliable than piping measures, which is explained by the fact that it is easier to detect overtopping than piping. Comparison of strategies In the Netherlands about one thirds (1225km of total 3780km) of the flood defences currently do not meet the safety standards required for flooding. Besides reinforcements other options could be considered to improve the safety of the flood defence, each with their own effect on safety and costs. The question is what effect a system of emergency (or control) measures could have on the total cost, which consists of investments, operational cost and risk. On dike ring level dike reinforcements reduce the failure probability with a factor 10, compared to the factor 1.5 ~ 2 of emergency measures. Which strategy is preferred depends on the specifications of the dike ring. For typical dike rings along the Dutch rivers, with initial failure probabilities of 1/100, the increase in safety of a system of emergency measures (factor 2) is insufficient to be an alternative for dike reinforcements (factor 10), because the failure probability is limited to 1/1,250 by law. Dike reinforcements are more cost effective than a system of emergency measures. But, a system of emergency measures could be an interesting interim solution if investments in dike reinforcements take years (or decades). The total cost of all strategies depends largely on the initial failure probability (or annual risk) of the dike ring. Dike reinforcement is the best option for initial failure probabilities of 1/100 ~ 1/500, corresponding with an annual risk of flooding of 20 million euro (with an average damage cost during a flood of 2~10 billion euro), see Figure 4. For initial failure probabilities below 1/500 a system of emergency measures becomes more cost effective. It is expected this is more or less the optimal safety level for flood defences in this type of dike ring, which can be investigated with (Brekelmans, Hertog, Roos, & Eijgenraam, 2012). Conclusions and recommendations A comparison of emergency measures and dike reinforcements showed that both strategies contribute to a reduction of the probability of flooding. Emergency measures could reduce the failure probability of a dike with a factor 2 ~ 5, depending on the failure mechanism, organizational reliability and the length effect of the emergency measure. Dike reinforcements could achieve higher reductions of the failure probability. Looking at the stringent safety standards for flood defenses it is concluded that dike reinforcements are the only option to achieve the required safety levels (higher than 1/1,000 per year). If emergency measures are included in the assessment of flood defenses safety standards are required for their reliability. In other areas where temporary/moveable defenses are applied, for example in hydraulic structures, the probability of non-closure may not exceed 10% of the safety standard. For Dutch rivers, with a safety standard of 1/1,250 per year, this corresponds with a probability of 1/12,500 per year. Human failure is included in these methods. Taking the results of this research in to account it seems similar criteria for emergency measures are not feasible. The reliability of a system of emergency measures depends to a large extent on human performance during the detection and placement phase. For piping specifically investments in the personnel responsible for finding sand boils, are very effective as the failure probability of the emergency measures for piping depends largely on the probability of finding sand boils. Increasing the reliability of the organization is only effective up to a certain level, when other factors such as the reliability in time and effectiveness become dominant. Reductions up to a failure probability of 1/100 are effective, which corresponds with the level at which districts operate. Further reduction can be achieved by investing in logistics (placement speed). The feasibility in time has failure probabilities of one order lower than the organizational failure probabilities. It becomes dominant when the available time is around 24 hours. River systems have prediction times of 2 to 4 days, but coastal systems have much shorter available time (order 12 hours). It is expected that a system of emergency measures will have little effect on the reliability of a dike ring in a coastal system. The emergency measures treated (dikes of sand bags, sand boil containments and piping berms) proved to have technical failure probabilities (order 10-5 per demand) which are negligible compared to the failure probabilities of humans and/or the feasibility in time. The reliability of the emergency measures depends largely on the reliability of human actions. The assignment of error rates to the different employees of the water boards is based on expert judgement of the author, which was quite accurate when compared to observations in the field. However, further investigation (possibly with Bayesian networking, (Jager, 2013)) could provide more insights in human performance during floods. Research in the use of alternative (innovative) emergency measures is recommended, as a lot of products are currently being developed for flood fighting. The main disadvantage of sand bags is the required time for placement, which is rather high. Several new products are being tested which could be an alternative for the classical sand bag, yet these products have technical reliabilities which are lower than sand bags.