Modelling flood defence failure: a UK case study

One in four properties are at risk of flooding in the UK, according to JBA’s newest UK Flood Model. Flood defences play a significant role in managing this risk to properties and, while rare, defence failure can significantly increase flood risk when it occurs during a flood event.

JBA’s UK Flood Model now enables you to model this defence failure, alongside other defence analysis options. We explore this functionality, examining defence failure at national scale, as well as the localised impacts it can have through a case study of Wainfleet.

Flood defence analysis options

There are three defence scenarios under which we can use the UK Flood Model to assess flood risk:

  • Scenario 1 – Undefended: Properties are treated as being undefended and flood waters are allowed to expand out over the floodplain unimpeded by any defences that might obstruct the flow of water during a real event, providing a worst-case scenario.
  • Scenario 2 – Defence overtopping: Properties are subject to the overtopping of flood defences, which allows some water to flow over the top of the defence and cause damage to properties that are protected by the flood defence, although to a much lower level than for the undefended scenario. This is likely to be the most realistic view of risk during a real event.
  • Scenario 3 – Overtopping and defence failure: Properties are affected by overtopping as well as intermittent flood defence breaches (wherein the defence structurally fails).

While Scenario 3 is rare in the UK, due to flood defences being generally well maintained, it is important to consider the impact this sort of failure may have on properties when it occurs.

flood defence overtopping: impacts on uk losses

Each flood defence within the model is defined by a level of water that the defence can hold back during times of flood (standard of protection). Overtopping occurs when the water level during an event exceeds the amount prescribed by the flood defence standard of protection and water flows over the top, potentially causing damage to any properties within the vicinity of the overtopping event.

The UK Flood Model takes this excess of water into account and produces losses in line with the amount of overtopping, and subsequent damage, that occurs.

Figure 1: The overtopping of flood defences as river waters expand over a levee. (JBA archive)

Whilst the overtopping of defences generally produces a substantially lower volume of water on the floodplain than in an undefended scenario (due to flood defences generally being constructed to a high standard in the UK) the extent of flooding can still be significant, as seen in Figure 1.

Figure 2 shows loss estimates for UK properties when the mitigating impact of flood defences is considered, versus an undefended view.

Figure 2: The difference in losses produced by the UK Flood Model v3 for two scenarios: portfolio that is left undefended and a portfolio that experiences loss due to the overtopping of flood defences.

Losses when properties are modelled as undefended are 195% higher than estimated losses when defences are included in the overtopping scenario, highlighting the important role that flood defences play in keeping our homes and property safe.

flood defence failure: impacts on uk losses

Flood defence failures, or breaches, occur when the structure of the flood defence fails and water flows through the defence as if the defence were not there.

Figure 3: An example of a defence that has been breached. (JBA archive)

Figure 3 shows a defence breach that has occurred, allowing water to cross the threshold of the original embankment. This type of breach is often caused by erosion of the embankment, allowing weaknesses to form which eventually lead to the failure of the structure.

The UK Flood Model calculates the probability of failure occurring by considering the following factors:

  • The standard of protection assigned to individual defences within the area
  • The different types of defences that are included within the model e.g. embankment, flood wall
  • The existing material condition of the defences drawn from external information and research

Each of these static factors are combined with the associated water depths present during each flood event to produce a probability of failure. When a breach occurs, the model removes the flood defence from the calculation and allows water to flow unimpeded through the previously defended area.

A comparison between the national-level UK losses for defence breach and defence overtopping is shown in Figure 4. Losses are only very slightly higher when accounting for breaching than when only accounting for overtopping.

Figure 4: The difference between defence overtopping and breach combined, and defence overtopping on its own.

defence failure losses: counter-intuitive?

The fact that full defence failure losses are only very slightly higher than just defence overtopping losses may seem counter-intuitive, if the assumption is that the inclusion of a breach mechanism should lead to significantly higher losses.

However, this assumption may be misleading at a national scale.

Research funded by the JBA Trust suggests that widespread failure of flood defences is unlikely to occur during large flood events. Others have also commented on the overestimation of breach probability, for example Simm et al 2018 note that the Environment Agency fragility curves may overestimate the probability of breaches.

Further evidence for this view comes from a review of defence performance over the 10 years between 2007 and 2017, which highlights the rarity of river breaches (Simm et al 2017). In the summer 2007 flooding, a total length of approximately 1,000km of defences were ‘tested’ across the event, with only four cases of breach recorded. This represents only 0.01% of defence length or 0.2% by number of defence assets.

The 2015/2016 winter flooding arising from storms Desmond, Eva, and Frank provides further evidence of low probability of failures, with a total of seven breach locations recorded. As with the 2007 event, these breaches were in low consequence areas which did not contribute significantly to the overall loss experienced (Simm et al 2017).

Consequently, JBA has adjusted the Environment Agency fragility curves which are used as a baseline dataset to reflect a more plausible probability of breach occurrence. This view of risk has been used to investigate the impacts of spending on defence maintenance in a joint project with JBA Risk Management, Flood Re, and the ABI (Garrett et al 2021), which highlighted the benefit of defence maintenance in reducing overall losses.

However, it’s important to note that while this is the case at a national scale, there are nuances to the picture in specific scenarios – as can be seen in the case study below.

wainfleet: A CASE STUDY

Due to an interplay of defence type and condition, the complex realities of a flood event, and other factors, defence breach can cause significant losses in more localised conditions.

The Garrett et al (2021) report also highlights that it is likely that breaches may occur regularly in areas where damage is currently underreported.

To investigate this, we look at the village of Wainfleet, Lincolnshire, where a defence breach occurred in June 2019 leading to 88 properties being flooded and 1,000 people being evacuated from 580 homes.

Wainfleet sits just to the south-west of Skegness on the east coast of England. The River Steeping passes through the centre of Wainfleet, with properties situated on both its northern and southern sides.

During the preceding month, rainfall totals were 185% of the long-term average (Environment Agency, 2019), saturating the ground and increasing river flows in and around the northern part of the Anglian river basin. This excess rainfall led to the defences being breached on Wednesday 12 June 2019. Emergency response included the coordination of an RAF helicopter, brought in to temporarily fill the breach with sandbags and stem the flowing water.

Properties affected at a local level

In order to investigate the UK Flood Model’s behaviour on a smaller scale, we take a closer look at the losses produced by the model in and around the village of Wainfleet, repeating each of the three scenarios.

Figure 5: The centre of Wainfleet. Rivers are shown by solid blue lines with the 1,000-year return period flood extents shown by light blue shading. Overlaid are: panel a) the locations of around 1,000 exposure points, b) the locations of exposures that experience loss in an undefended scenario, c) the locations that experience loss in an overtopping and breach scenario, d) the locations that experience loss when only overtopping algorithms are applied.

Figure 5a shows the centre of Wainfleet. 1,185 individual properties were placed into the PE24 4 postcode sector using the JBA disaggregation scheme, which matches concentrations of exposed properties to areas within building footprints, with larger concentrations of disaggregated properties found in regions where the concentration of buildings in the area is at its highest. This gives us a realistic distribution of exposure within the vicinity of the village centre.

Many of these properties, however, are unaffected by the flooding. As a result, those properties which are modelled to be at risk of flooding, producing losses when the undefended option is selected in the model, are plotted in Figure 5b.

Figure 5c shows the number of properties affected by flooding when the overtopping and breach option is selected, and Figure 5d shows the number of properties affected by flooding when just the overtopping routine is engaged.

Figure 6: The number of properties affected by flood waters during a 10,000 year simulation for each scenario described within the text.

Figure 6 demonstrates a significant increase (of around 75%) in the number of properties affected by the overtopping and breach scenario when compared to a model run containing just the overtopping scenario.

The difference between the undefended properties scenario and the overtopping and breach scenario is less stark, with the undefended scenario sitting just 8% above the overtopping and breach scenario.

The incident that occurred in Wainfleet appears to contradict the national level results, which show similar levels of risk between the breach and the overtopping and breach scenarios. This highlights the fact that although defence breaches are infrequent occurrences, when they do occur there can be a significant increase in the number of properties affected at a local level. It’s vital that risk managers are able to assess these scenarios with the data available to them.

Annual Average Losses at a local level

When we look at the difference in average annual losses, however, a picture emerges which is very different. Figure 7 shows the average annual loss associated with each scenario.
As is the case for the national level losses, the undefended scenario produces large losses when compared to both the overtopping scenario and the overtopping and breach scenario. This is equivalent to an increase of 1,645% when using the overtopping and breach scenario as a benchmark against which to compare.

Figure 7: The average annual loss produced by each scenario for Wainfleet in Lincolnshire.

Figure 8: The average annual loss produced by the overtopping scenario and the overtopping and breach scenario for Wainfleet in Lincolnshire.

Figure 8 shows a small increase in losses when defence breach is incorporated in the modelling for Wainfleet.

The increase in losses between overtopping scenarios and overtopping and breach scenarios (18%) is much smaller than the increase in the number of properties affected. So even in a village like Wainfleet where the defences are vulnerable to breach (as demonstrated in 2019), the economic impact of a defence breach occurring is much smaller than might be implied by the increase in the number of properties that are affected.

However, the fact that Figure 8 shows a noticeable difference in losses between the two scenarios, whereas the difference in Figure 4 shows that at a national level the losses are fairly similar, suggests that the probability of increased losses due to a defence failing in and around Wainfleet is high relative to the probability of increased losses due to defence failures on a national level.

modelling defence failure: data available

This simple exercise has demonstrated that in certain locations around the UK, losses can be significantly increased (18% in Figure 8) by the occurrence of a defence breach. However, on a national level the difference in losses from defence breaches occurring is negligible.

The difference between the national and regional results highlights the complex interplay between exposure, event frequency and the severity of flooding in local areas, which can be captured in the new UK Flood Model.

The implementation of both overtopping and breach mechanisms within the UK Flood Model provides users with a more complete description of the physical processes that contribute to whether or not a property is at risk of flooding. This allows re/insurers to identify areas within the UK where their portfolios are at increased risk of flooding due to the possibility of a defence breach occurring during the underwriting, portfolio management, accumulation control and reinsurance pricing processes.

If you'd like more information about the UK Flood Model or would like to discuss a test and evaluation free trial, get in touch with the team.

references

Environment Agency, 2019. Rainfall and river flow 12 to 18 June 2019. [Online] Available at: https://webarchive.nationalarchives.gov.uk/ukgwa/20190703063215/https://www.gov.uk/government/publications/weekly-rainfall-and-river-flow-reports-for-england [Accessed 25 March 2022]

Garrett, J. Pettit, A. Waller, S. Millinship, I. Power, R. 2021. Modelling the impact of spending on defence maintenance on flood losses. JBA Risk Management, Flood Re, Association of British Insurers [Online] Available at: https://www.abi.org.uk/globalassets/files/publications/public/flooding/modelling-the-impact-of-spending-on-defence-maintenance.pdf [Accessed 04 April 2022]

Simm, J. and Tarrant, O., 2018, July. Development of fragility curves to describe the performance of UK levee systems. In Proceedings of the Twenty-Sixth International Congress on Large Dams, Vienna, Austria (pp. 1-7).

Simm, J., Flikweert, J., Hollingsworth, C. and Tarrant, O., 2017, July. Ten years of lessons learned from English levee performance during severe flood events. In 5th Annual Meeting of International Comminssion on Large Dams. Prague, Czech Republic: ICOLD.