Managing severe thunderstorm risk: Looking beyond the weather forecast

As we reach the middle of Spring, the U.S. hail and tornado season is in full swing.

March is a changeable month, in weather-terms. There’s an old saying that goes something like this: “March comes in like a lion, goes out like a lamb”. This mainly refers to temperatures, denoting the end of the winter’s cold weather at the start the month (the lion), followed by a warmer and fine end for many parts (the lamb).

However, this year there have been some extremes making this old saying not quite so apt. It’s been a busy month from start to finish when thinking about thunderstorms. Widespread damaging hail and tornadoes, especially in the last week, mean that March 2017 has come in like a lion and gone out like a pride of lions!

Thundery March

At the end of February and into the start of March there was a tornado outbreak in the Midwest. In fact there were tornadoes reported from Missouri, through Illinois and into Indiana: very rare so far north for the time of year.

In fact, the yearly tornado count was already ahead of average by the start of March, and continues to be above average, almost doubling the normal number of tornadoes (See chart below).

In the second warmest February on record, wildfires also spread across the Southern Plains.

At the end of the month, we saw a roar of severe weather lasting several days. Over the course of a week, there were multiple tornado outbreaks and final rash of tornado activity, and very large hail ran eastwards across the south of the U.S. As estimates of damage start to arrive, it seems clear that March will contribute a significant portion of the convective storm damage total this year.

Compared to last year

There were four occasions when individual storms caused greater than $1 billion in damage

Last year’s losses from convective storms were also high in the U.S., according to the Sigma 2/2017 Swiss re report. There were four occasions when individual storms caused greater than $1 billion in damage, with a total of $15 billion insured losses from tornado outbreaks and thunderstorms for 2016 (20% higher than the 10 year average), compared to $9.7 billion in 2015.

For a detailed run down of natural catastrophe events in 2016, including severe convective storms the “Willis Re Summary of Natural Cat Events 2016” may also be useful.

Comparing the average annual loss (AAL) of hurricane losses to severe convective storm we see virtual parity. Based on Verisk Analytics Property Claims Services (PCS) statistics, our recent Managing extreme thunderstorm risk report shows AAL between 2003 and 2015 for hurricanes is around $11.28 billion, while for severe convective storms it stands at $11.23 billion. The report goes on to analyse the influence of the El Niño-Southern Oscillation (ENSO) is on severe convective storm activity.

Using forecasts in insurance

The industry is used to considering long-range seasonal forecasts for hurricane activity. These forecasts look at the broad climate situation and use statistical or physics-based models to predict likelihoods of more, or less, active seasons.

Much of the predictability comes from the influence of the El Niño-Southern Oscillation, which plays a key part in any hurricane forecast. ENSO also influences severe convective storms and therefore can encourage or discourage the associated tornadoes or damaging hail activity.

Meteorological indices

Willis Research Network partners at Columbia University have been working on a new approach to using climate model outputs to predict the environmental conditions associated with extreme hail or tornado activity. These are physics-based forecasts of meteorological indices that represent winds, temperatures and moisture interactions as described in this paper by Michael Tippet et al. from 2012.

The team at Columbia has developed indices that give an indication whether hail or tornadoes are more likely called the Hail Environmental Index (HEI) and the Tornado Environmental Index (TEI). The Willis Re report mentioned above highlights how ENSO influences these indices.

Application of Science

We can view the forecasts for the next month to aid planning for claims handling or loss adjusting

So how might we use the links between ENSO and severe convective storms? Firstly, we can view the forecasts for the next month to aid planning for claims handling or loss adjusting, and potential for buying reinstatements if there is an increased risk likely to affect exposed parts of a portfolio.

The link to ENSO can also be used to provide scenarios, given that, to a degree, ENSO is predictable months in advance. This knowledge can aid realistic scenario testing given a higher or lower likelihood of an active convective storm season.

April Forecast

The latest view for April (typically a very active month) shows an increased risk from Texas up through “Tornado Alley” and across most of the east of the U.S.

The season has already produced roughly double the average of tornados and hail reports for February and March, with some outbreaks unusually far north, perhaps influencing the northward extension of the forecast TEI and HEI in the top two panels below.

The bottom two panels, show the raw observed observations for April based on 1990-2015). To find out more about these forecasts or the Willis Re report mentioned above, feel free to get in touch.

Source: Columbia University


This post was written with Prasad Gunturi. Prasad is Senior Vice President of Willis Re Analytics, where he leads the North American catastrophe modeling research and evaluation team. Prasad manages and leads specialized technical projects, including understanding the changes in the catastrophe models, technical evaluation of commercial catastrophe models, developing portfolio specific alternative views of risk and proprietary model development projects.

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 | Tags: ,

Leave a Reply

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