How to model a windstorm

Storm Angus hit the U.K. earlier this month (also known as Nannette as designated by the Free University of Berlin). Although insurance industry losses are not expected have much impact on the reinsurance industry, the storm did produce high winds and heavy rain over the U.K., especially for Wales and Southwest England, and into parts of France. The quick assessment of potential losses from such storms, before, during and after the event, has become increasingly analytical over the last few decades, achieved through the development of a variety of catastrophe modelling approaches.

Catastrophe analytics and windstorms

A key component of the catastrophe modelling process for European windstorms is the hazard event set. To build this event set, there are two main approaches that are widely used in the industry:

  • One approach is to construct wind footprints from historical events, using observations and statistical resampling to produce an event set that represents the wide range of possible financial losses from extreme ‘tail’ events.
  • An alternative approach is to generate synthetic events that use the statistical relationships within the observations to characterise events.
Adjusting catastrophe model output can lead to a better representation of annual average losses

Generating a synthetic footprint is not a straightforward process, though, as there are a variety of approaches used to create the range of statistically possible events. The Willis Research Network’s longstanding partnership with the University of Exeter has recently focused its attention on new approaches to make this process more efficient and to give an alternative view of windstorm risk. This builds on our ongoing collaboration with Professor David Stephenson and his team of research staff, which has helped to bring greater understanding of windstorm clustering to the insurance industry over the years, which can be used to adjust catastrophe model output to better represent annual average losses.

New award-winning research

Dr Ben Youngman, lead author of this year’s winning entry into the Lloyd’s Science of Risk Prize, in the “Natural hazards” category, has been working to develop a new and computationally efficient approach to generating stochastic event sets for natural hazards. It is a novel approach that utilises the power of geostatistics and extreme value theory to create a method to give spatial dependence to values across a geographic region, while also statistically representing extreme values of the modelled hazard.

Here comes the science

Models are used to approximate generalized Pareto distributions to capture the intensity of the windstorms

The winning academic paper [preprint]uses statistical modelling of wind observations to capture European windstorm extremes.

For the statisticians out there, the paper describes how robust generalized additive models are used to approximate generalized Pareto distributions to capture the intensity of the windstorms, while a student’s t-process captures the spatial dependence to produce a continuous-space framework to footprint simulation. This process enables many years of simulations to be performed relatively quickly, making it suitable for identifying geospatially consistent extreme event footprints of maximum wind gust. These simulated events validate well against the observed wind gusts, and have realistic extreme value and spatial properties that can be applied to catastrophe model validation.

Appliance of science

It is planned that these stochastically generated event sets will provide a new tool for evaluating industry catastrophe model hazard event sets, starting with European windstorms, but potentially other hazards and other territories. There are potentially many other applications to this research.

Impact of industry / academia partnership

Dr Youngman is a Willis Research Network Fellow, working with the Willis Re Catastrophe Analytics teams to develop his approach to generating stochastic event sets for natural hazards. The recognition gained by Dr Ben Youngman and Professor David Stephenson’s paper by winning the Lloyd’s Science of Risk Prize, shows how successful co-production of academic science can be an important part of industry innovation.

As the world becomes more interconnected, and risks become more globally linked, the insurance sector must rise to the challenge of developing new and improved methods to assess risks across society for a sustainable future. Challenges, such as climate change or pandemic, are complex and affect many parts of society at once, so through multi-sectoral partnerships we can begin to understand how these risks can be quantified and managed for all interests.

About Geoffrey Saville

Geoffrey Saville is a member of Willis Towers Watson's Analytics Technology Team, having joined the company in 2013…
Categories: Natural Catastrophe, Reinsurance, Weather risk | Tags: , ,

Leave a Reply

Your email address will not be published. Required fields are marked *