One Year Since Dudley, Eunice, and Franklin: Understanding Windstorm Clustering in Europe
Michèle LaiFebruary 21, 2023
In February 2022, a sequence of powerful windstorms known as Dudley, Eunice, and Franklin (also known as Ylenia, Zeynep, and Antonia respectively) caused billions of euros of insured loss in Europe. With the first anniversary of these storms, we are reminded of the need to consider clustering when modeling windstorms in Europe.
The 2022 windstorms happened as a large temperature gradient over the North Atlantic led to the formation of a series of low-pressure systems, which were then associated with a strong jet stream and deepened into three strong windstorms.
Between February 16-21, 2022, these events caused widespread damage across northern Germany, the Netherlands, and the U.K., as well as affecting Ireland, France, Belgium, Denmark, Switzerland, Austria, the Czech Republic, Poland, and Slovakia.
Within the three storms, Moody’s RMS analysis suggested that Eunice/Zeynep was the main storm and was expected to contribute between 2.5 and 3.5 billion euros (US$2.8–US$4.0 billion) to the overall insured loss total.
This ranks Eunice as the most damaging European windstorm event since Kyrill in 2007 (7.6 billion euros indexed to today’s value).
During this storm sequence, the Moody’s RMS Event Response team provided near real-time products to help clients understand likely losses from these storms.
Our loss estimate, issued just days after the storms, showed that insured losses from Dudley and Eunice (Ylenia and Zeynep), would likely fall between 3.0 and 4.5 billion euros (US$3.4–US$5.1 billion), with Franklin (Antonia) delivering hurricane force winds but being less destructive.
These loss estimates were later confirmed as other market estimates emerged for this storm sequence over the course of the following six months.
Looking at the geographic split of the losses from the three storms, losses in Germany would likely account for around 40 percent of the total loss, followed by the Netherlands at around 20 percent, and the U.K. at around 15 percent.
Storm Clustering is Not Uncommon
Storm clustering is driven by physical processes and is a fundamental part of windstorm behavior in Europe. This phenomenon refers to several types of events associated with different atmospheric dynamics, such as large-scale patterns (typically, the North Atlantic Oscillation is a key influencer for storm clustering in Europe) or the formation of secondary lows along primary cyclones (secondary cyclogenesis). These mechanisms are very well described by Dacre and Pinto in 2020.
For the (re)insurance industry, storm clustering simply describes the occurrence of multiple storms over a certain location such as Europe, within a certain time period (e.g. a week or a month).
Analysis of historical observations shows that annual storm occurrence tends to be over-dispersive, which indicates that storm occurrences are not independent, i.e., if one occurs, then it is more likely that others occur. It is this over-dispersive behavior that clustering models, as part of natural catastrophe models, aim to reflect.
Benchmark examples of storm clusters include Daria and Herta in 1990 and Lothar and Martin in 1999, but clusters of smaller storms are not uncommon. During the 2021-22 windstorm season alone, there were three clusters with Arwen/Barra in late 2021, Malik/Corrie in late January 2022, and Dudley/Eunice/Franklin in February 2022.
Storm clustering can increase the volatility in losses and thicken the tail of the aggregate exceedance probability distribution. But when storm clustering occurs, it remains challenging for the insurance industry to accurately assess the damage caused by each individual storm.
What are the key issues for (re)insurers around storm clustering, and how are Moody’s RMS and the new high-definition (HD) risk models looking to help?
A New Reference View of Risk
Moody’s RMS has modeled storm clustering since 2008 and released a major update in 2016. This update has withstood significant scrutiny from the insurance modeling community over the past five years and is the most robust approach available to model clustering.
It provides a more complete picture of clustering gained from historical data and new analytical techniques, helping to reduce uncertainty. Leveraging the HD modeling framework, the new Moody’s RMS Europe Windstorm HD Models™ enable clustering to be modeled natively and seamlessly to reflect the over-dispersive nature of this phenomenon.
Accurately Capture Hours Clauses
Temporal simulation, a key feature of HD models, enables the application of widely-used hours clauses, where a reinsurance or insurance contract covers losses occurring during a set time period – a typical period could be 72 hours.
Hours clauses can act to aggregate losses from multiple storms into a single loss, and clustering increases the probability of multiple storms falling within standard hours clause time windows.
Better Understand Windstorm Losses
When several storms occur close in space and time, it is often difficult to separate losses from each event. The February 2022 windstorm series is a good example of this. The losses for the three storms are usually reported together within the (re)insurance industry, making it difficult to assign a return period for each storm.
In addition, reading the combined losses on a Poisson-based Occurrence Exceedance Probability (OEP) curve would provide unrealistic numbers. For example, if you model windstorms Lothar and Martin (1999) as independent storms, this would give a return period of about 1,000 years, while a clustered OEP would estimate these storms as a 200-300-year event, which is more realistic.
Modeling clustering, therefore, has an impact on the Aggregate Exceedance Probability (AEP) and the OEP, when compared to the Poisson view. Clustering essentially increases the number of years when very few damaging storms occur, in between years where storms are clustered.
Effectively, clustering generally:
Increases the AEP at longer return periods and decreases at the shortest return period
Decreases the OEP, particularly at short and mid-return periods.
Modeling Seasonality and Multi-Year Contracts
The most severe European windstorms usually occur during the winter months from October to March, meaning that the calendar year – typically used for reinsurance contracts – contains half of two successive storm seasons.
To overcome this, the temporal simulation used in Moody’s RMS Europe Windstorm HD Models provides six-year periods that reflect the observed relationships in storminess between successive seasons due to multi-decadal variability of activity.
This allows clients to switch to modeling contracts starting on dates other than 1/1 and therefore capture a single storm season within the contract time period to better model the seasonality and deliver more realistic windstorm losses.
Windstorms and High-Definition Modeling
A key objective in the development of Moody’s RMS Europe Windstorm HD Models was to accommodate clustering windstorms using a time-based approach. The HD framework uses a full simulation engine allowing for a more realistic representation of seasonality and the native implementation of windstorm clustering.
This delivers improved loss distribution for better pricing and portfolio management and helps with the application of terms such as hours clauses.
As windstorm clusters are a regular occurrence in Europe and are typically responsible for the largest losses, it pays to analyze these events to understand potential losses across a range of scenarios including exploring tail-risk events.
Click here to find out more about how Moody’s RMS Europe Windstorm HD Models provide a more realistic representation of wind, storm surge, and climate change risk across 17 European countries.
Chloe is a senior product marketing manager at Moody's RMS, where she helps customers develop data-driven strategies to better understand their risk. She has more than a decade of experience in catastrophe modeling for the (re)insurance industry.
Previously, Chloe was a senior account director for CoreLogic, managing catastrophe modeling clients across Europe, Asia, and the U.S. Prior to CoreLogic, Chloe worked in various catastrophe modeling roles within a Lloyd’s Syndicate, AIG, and Aspen Re.
Chloe has a bachelor’s degree in Geography with Anthropology from Oxford Brookes University.
Michèle joined RMS in 2013, and is based at the Moody's RMS Zurich office as part of the Global Climate Product Management team, focusing on European climatic hazard models. As part of her role, Michèle is the senior product manager for the new Moody's RMS Europe Windstorm HD Models and the Moody's RMS Europe Severe Convective Storm HD Models.
She holds a master’s degree in Atmospheric and Climate Science from ETH Zurich.