Computing flood risk in areas protected by flood defences

Abstract

Risk analysis of areas protected by flood defence systems involves probabilistic analysis of a large number of scenarios in which one or more of the defence sections that make up the system has failed. In systems with large numbers of defence sections, the computational expense of this calculation can be prohibitive. When the probability of failure of each defence section is not negligibly small, sampling approaches that are now in widespread use may not converge on a stable risk estimate in reasonable computational time. To overcome this worrying limitation, this paper reformulates the flood risk calculation in terms of the cumulative distribution function of the volume of floodwater entering a floodplain. An algorithm is presented whose computational expense scales linearly with the number of sections in the flood defence system. The approach is applied to flood risk analysis in areas protected by extensive systems of flood defences in the Thames estuary, revealing how flood risk varies depending on the characteristics of the flood defence system and floodplain topography. It opens up the possibility of more exhaustive risk-based appraisal and uncertainty analysis of flood risk management options than have hitherto been feasible.

Probabilistic Assessment of Flood Hazard due to Levee Breaches Using Fragility Functions

“…In view of the present and previous assumption (5), breach scenarios characterized by different breach locations and failure times are considered as mutually exclusive. On the other hand, multiple failure states are usually treated as independent in literature (Dawson et al, 2005;Gouldby et al, 2008;Hall et al, 2003;Harvey et al, 2014), which means that the hydraulic interactions between breach events occurring during the same flood event are neglected. Including multiple breaches in the analysis can imply that a large number of combinations of breach locations and possibly times should be considered, especially for long levee reaches described with many cross-sections, thus requiring a prohibitive computational effort (Dawson et al, 2005).…”

“…Uncertainty in the loading conditions on a levee system has been extensively investigated in literature (Apel et al, 2004;Dawson et al, 2005;Dawson & Hall, 2006;Hall & Solomatine, 2008;Lendering et al, 2018;Winter et al, 2018). However, the temporal component of load sequencing is usually neglected, that is, all sections of the defense system are assumed to be subject to the same peak hydraulic load at almost the same time (Dawson et al, 2005;Hall et al, 2003;Harvey et al, 2014). 6.…”

“…Including multiple breaches in the analysis can imply that a large number of combinations of breach locations and possibly times should be considered, especially for long levee reaches described with many cross-sections, thus requiring a prohibitive computational effort (Dawson et al, 2005). Dawson and Hall (2006) and Harvey et al (2014) developed efficient flood risk methodologies to reduce this problem.…”

“…Flood risk assessment essentially consists in the estimate of the likelihood and severity of the adverse consequences of a flood event. It rationally supports a risk-based decision-making process aimed at developing effective risk communication and flood management policy, as well as at orienting judicious land use planning and economic resources allocation (e.g., Apel et al, 2004;Dawson et al, 2005;Gouldby et al, 2008;Hall et al, 2003;Harvey et al, 2014;Sayers et al, 2002;Sayers et al, 2011;Wagenaar et al, 2016). The flood hazard assessment, that is the quantification of the frequency and severity of a flooding event, is a fundamental phase of the overall flood risk analysis (USACE, 2017).…”

Editorial

“…They present a case study for Glasgow where they adopt a hierarchical approach: a coarse topographical model is used to identify hotspots for more detailed investigation. Harvey et al (2014) take a fundamental look at how we calculate risk at a large scale where there are many defence sections that may fail. Until now the sampling methods in use require prohibitive computer times.…”