How insurance response models can improve flood predictions

Scroll to continue

MS Amlin is exploring how to improve flood predictions and help people be prepared. Research Manager in Catastrophe Modelling, Tina Thomson, talks about the collaboration with Imperial College, London which is using data from the Carlisle flood of 2015 to develop a localised flood response model.

As Hurricane Harvey made its horrific landfall last week flooding parts of Texas, USA, few people in the north of England can forget the devastation caused by Winterstorm Desmond.

After the inundations of 2005, Carlisle’s flood defences were upgraded and the people of this historic border town thought their town was ready should the rains come again. Sitting on the border between Scotland and England, the town grew up around the last bridge over the river Eden which flows into the sea nearby.

All insurers – including MS Amlin – rely on computer models to calculate risk and insurance premiums. But in the case of big, weather-related events these models are also used to anticipate where the damage is likely to happen so they can help warn home and business owners to take extra precautions.

But in 2015 torrential rain fell on already saturated land. Initially the new defences coped well, saving a significant number of buildings from flooding. But the Eden isn’t the only river in the Carlisle area and when the rain caused record rises in water levels in two other nearby rivers the town was trouble. Carlisle the unprecedented water levels caused left a trail of devastation affecting more than 2,000 properties and claimed the life of a 70-year-old man. The shocking pictures of the resulting devastation still linger in the memory.

The best laid plans…
While we’d like to think people will always follow advice this simply isn’t the case. Many factors influence whether people take action in response to flood warnings: from whether they’ve been affected by flooding before and how recently it happened to the reliability of forecasting and flood warnings.

We do know that timely and accurate warnings to specific policy-holders will increase their tendency to take action in advance and prepare their properties and businesses. This can significantly reduce the damage caused by extreme weather.  

At MS Amlin we are collaborating with Imperial College to develop models that can do just that.

Very local, highly unpredictable
Flood is such a localised peril it is notoriously difficult to model. Storms, Desmond, Eva and Frank affected 16,000 homes across the north of England and Scotland during December 2015. They caused £2bn of damage, of which only £1.3bn was insured with the data showing that a significant number of claims came from areas previously modelled to be low risk.

It’s generally agreed that in the Carlisle floods of 2015 the models struggled because of the unusually wet winter. One storm after another saturated the ground; unprecedented levels of rainfall continued to fall. With nowhere to go, three rivers burst their banks and flooded areas which the model failed to predict.
“Flood is such a localised peril it is notoriously difficult to model”

A local solution
Through our work with Imperial College we are exploring the potential for real-time hydrodynamic modelling solutions to improve the accuracy of our flood models.

Hydrodynamics’ is the study of the forces exerted by fluids, and how they react with solid materials around them. We believe there is potential to use scientific understanding from this field to create more accurate models that will predict in real-time how water flows are likely to impact on buildings. This could become a powerful tool to help people prepare for flood events and therefore mitigate losses and disruption. Research shows that flood warnings, flood emergency response plans, and setting standards on installations would make a positive difference to damage-saving efforts.

Of great value to affected households and businesses, these models may also speed up the payment of claims. If we know where the locations which will experience the greatest damage we can make sure funds are available quickly for repairs. Helping people get back on their feet quickly and continue with the lives and businesses is at the heart of the MS Amlin brand.

Developing the models
Plenty of data is available to forecasters and modellers, particularly in the UK. The Environment Agency has a real-time tool to measure flows in rivers across the UK using river gauges. They and the Met Office are also good sources of detailed real-time information on rainfall. There is also plenty of information about soil-types and digital terrain maps to help understand how heavy rainfall is likely to impact on the surrounding land.

We can now use crowdsourcing information from social media to help us understand quickly what is happening in specific locations and how that matches other data. While it’s difficult to model, it does provide aluable additional information to add to the information we get from our models.

More localised, more accurate, more supportive
Perhaps the biggest challenge of all is to have localised and very specific predictions of where the flood impact will be greatest.

The next generation of probabilistic flood models will allow for more flexibility in modelling local event information, such as temporary local defences, water management systems and different event durations as defined in the catalogue of flood events.  At the heart of all this is a desire to lessen the impact – both emotionally and financially for those affected.

Our collaborator at Imperial College, Jordan Seyedi, will be talking about his work at the Remote Sensing and Photogrammetry (RSPsoc) Annual conference in London, from September 5-8.  

Are you sure you want to remove this article from your library?