Are severe convective storms in the US now a ‘primary’ peril for (re)insurers?

In November 2025, Dr. Scott St. George (WTW’s Head of Weather and Climate Risk) hosted a discussion of the shifting landscape for severe convective storms. His guests included Professor Kelsey Malloy (Assistant Professor of Climatology in the Department of Geography and Spatial Sciences at the University of Delaware) and Dr. Cameron Rye (Director of Natural Catastrophe Analytics at Willis Re).

Scott St. George: To start, can you give us a bit of “Climatology 101” about severe convective storms. That is a term that the insurance industry uses a lot to describe several different aspects of severe weather. What is a severe convective storm, and what makes that kind of storm different from a regular storm?

Kelsey Malloy: A severe convective storm has a specific definition from the Storm Prediction Center as having one or more of the following: a storm that produces a tornado, hail that is greater to or equal to one inch, so about quarter size, or straight-line damaging winds of about 58 miles per hour. It is defined by the strength of those features and whether they show up with the storm. It typically does not include things like flash flooding or lightning, which are also associated with severe convective storms but are not part of the formal definition.

St. George: Why are there some places that are so prone to these storms? If we look at a global map of annual hail probability, the central United States and central North America glow bright red and yellow. Is there something about that particular part of the world that lends itself to creating those perils?

Malloy: We see the greatest impacts from severe convective storms in places where we have these long-lived thunderstorms called supercells. These are the strongest types of thunderstorms that can produce tornadoes and hail specifically. When we think about the life cycle of a thunderstorm, we have updrafts, rising air that supplies moisture and warmth, basically the fuel for the storm to keep growing. Once it gets to a point where we have precipitation, we have a downdraft, where precipitation drags the air back down. For a short-lived thunderstorm, the downdraft cuts off the updraft, and the storm dissipates because it is no longer getting that fuel.

However, if we have an ingredient like vertical wind shear, which means winds are changing with height, we can separate the downdraft from the updraft. Therefore, we can still get that continuous fueling of a thunderstorm to produce these really strong supercell events. When we look at the big hotspot for hailstorms in the central US, we are seeing favorable environments for this setup. We have the Gulf of Mexico providing a southerly flow that brings unstable, warm, moist air. Then, because of the Rocky Mountains, we get setups of the jet stream that provide that necessary wind shear. It is a really conducive environment for producing supercells.

St. George: In 2024, you and your colleague Dr. Michael Tippett authored an WTW Insight piece titled “Uncovering Trends in US Tornado Outbreaks.” In that article, you wrote that, while we are interested in whether things like tornadoes or hailstorms are becoming more common, it remains a difficult question to answer. Why is it so hard to pin down the true risk posed by convective storms?

Malloy: I think there are three challenges to uncovering any kind of climate related signals in severe convective storms. The first is that our best set of observations are human reports compiled by the Storm Prediction Center, and those reports can have limitations and non-physical artifacts. Second, severe convective storms and their hazards are very rare and sporadic, random events. That means we need many, many samples to resolve the statistics related to climate. Finally, climate scientists typically move to models to find these signals, but severe thunderstorms, especially tornadoes and hail, are not even resolved in standard climate models.

St. George: You talked a little bit in your article about the confounding influence of luck. How does luck tie into the way that we imagine risk or the way that we interpret risk based on some of these products like first-person reports.

Malloy: Luck becomes a big factor when we’re looking at short data sets. In our current report data, we can have two adjacent locations that have experienced a totally different number of hazards, tornado and hail for example. But that’s not necessarily reflective of their true tornado or hail risk. Rather, we should think of it as the observational record is likely too short to account for the fact why one location might have so many more reports than another location. Instead, we would need thousands of years of tornado and hail events to resolve or converge at the true risk.

St. George: The work you led in partnership with WTW took a different approach by looking at the broader environmental factors. Can you explain how you used an “ingredients-based” approach to get around the limitations of first-person reports?

Malloy: Our idea is that we want to look at what fuels supercells. We need convective available potential energy, which tells us about the instability of the atmosphere, and we need wind shear to separate the updraft from the downdraft. Environmental data does not have the same limitations as reports; it does not have the same non-physical fingerprints or changing practices in human reporting. By modeling tornado activity with the environments as predictors, we can account for the total range of outcomes.

When we reconstructed a dataset of tornado outbreak activity using our model, creating a “Tornado Outbreak Index”, we could see how the risk has changed over time (Figure 1). If you look at the reports, areas in the Tennessee River Valley and the southeast US show that outbreak conditions have become two to three times more likely relative to the normal rate from the late 1970s. Our index matches this trend. Most of the eastern US is experiencing an upward trend in tornado outbreak potential according to our index. We found that all three key ingredients; convective precipitation, instability, and storm-relative helicity (rotating updraft potential), experienced an upward trend in the Tennessee River Valley region, which helps explain physically why we are seeing these changes.

St. George: Does this mean climate change is driving these trends?

Malloy: We are not necessarily saying that all these environments have a climate change signal. We are just noticing that over this period, all three ingredients have noticed an upward trend. The evidence suggests that convective ingredients like atmospheric instability should produce an upward trend consistent with global warming, because a warmer atmosphere can hold more water vapor. But we are unsure about how trends in wind shear relate to global warming. There is no good theoretical understanding of whether trends in wind shear are a response to warming or just reflective of inter-annual variability, such as the El Niño Southern Oscillation. So, the near-term projection for 2030 or 2040 is very murky.

St. George: Cameron, you and I were talking earlier about the impact that severe convective storms have had on the insurance industry. There have been very high losses in the last couple of years.

Cameron Rye: Yes, exactly. So severe convective storm losses for the insurance industry have been very high recently. If we look at global insured losses from severe convective storms, in 2023 alone they exceeded $60 billion. And importantly for our audience, the vast majority of that, so over $50 billion, was from the US. That is the first time that severe convective storm losses have topped $50 billion in a single year for the industry.

St. George: To put that $50 billion number in context, does that make severe convective storms a primary peril?

Rye: Historically, we’ve often thought about perils in two categories: primary and secondary. Primary perils in the case of the US are typically hurricanes and earthquakes, and these are low frequency but high severity events. A nice recent example is Hurricane Ian in 2022, which caused around $50 billion of insured losses when it made landfall in Florida. These types of events typically drive our reinsurance programs and for the larger events, they have impact on capital requirements.

But secondary perils, and this is the group that severe convective storms fall into along with other perils like wildfires and floods, have traditionally been viewed as more manageable or attritional losses. But I think the shift that we’ve seen in the last few years illustrates why there’s been a bit of a surprise. All of a sudden, we’ve had several years starting around 2020 where we had $42 billion, and then in 2023 we had nearly a $60 billion loss. 2024 is another big loss over $50 billion (Figure 2).

St. George:As you noted, over the last 10 or 15 years, the insured losses from SCS have been gradually going up and then suddenly 2023, 2024 and again this year, the losses are very high. Kelsey just explained how explain how the environmental factors that produce tornadoes and severe weather have themselves become more likely over roughly the same period. Is it as simple as one causing the other?

Rye: In part. If you have more favorable environments, you will have more events and insurers will pick up more losses. But if you look at the claims data, it’s just that the change has been quite large and can’t be completely explained by changes in hazard alone.

St. George: If it is not just the hazard pushing it up so much year upon year, what are the other factors the industry is looking at to explain that change?

Rye: There are two main non-hazard factors. The first is exposure growth. We have seen a continuous expansion of urban areas and exposure growth in high-risk areas for convective storms. Since the turn of the century, there have been around 100,000 new properties built every year in areas at risk from convective storms. This essentially just means we have more assets in the way of storms that can get damaged.

The other key factor is inflation. The cost to replace a property when it has been damaged has risen. For hail specifically, the roof is often the biggest contributor to insured losses. We have seen a huge increase in the cost of building materials and labor. For example, the cost of asphalt roofs, one of the most common types in the US, has risen by about 250% since the year 2000. Post-COVID inflation alone drove a 45% increase in the last five years just to replace these types of roofs.

St. George: What other things are happening within the reinsurance market that might be relevant to the way that the industry deals with severe convective storms?

Rye: On the reinsurance side, in the last few years when we’ve been in a bit of a hard market, one of the things we’ve seen is that reinsurers have been trying to reduce or minimize their exposure to convective storms and other secondary perils. One of the ways in which they’re doing this is to write programs further up the risk curve. So essentially at higher return periods, they are retaining that exposure to say hurricanes and earthquakes higher up but trying to minimize the convective storm losses.

St. George: This is adding up to a tricky problem for the insurance industry. What do you think insurers and reinsurers ought to be doing to get a handle on this changing hazard and exposure?

Rye: One of the most important things for me is modeling and ensuring that your models that you’re using and your views of risk are up to date. Recently, we’ve seen the main catastrophe model vendors all update their US severe convective storm models. So, I think users of these models should now be thinking about evaluating these new versions, comparing them to recent loss experience, and comparing them to the old versions of the models.

Historically, convective storms have been considered as a secondary peril, and this has meant that the models haven’t maybe been updated as frequently as they should have been. Now that we’re seeing these updates come through, it’s a really good opportunity for insurers to re-evaluate their view of risk.

St. George: Do you think the last three years have shown us a worst-case scenario, or is this the new normal?

Rye: I do not think we have really seen the upper limit of what we could get from a really bad year for convective storms. The consensus within the market is that what we have seen in the last five years is perhaps what we should now expect for an average year. We are currently at about $45 to $50 billion in insured losses for this year, so we have had another year getting close to that mark. We should be expecting this now on average. If we have a truly extreme year, losses could be significantly higher.

Cameron Rye, Director Natural Catastrophe Analytics, Willis Re, cameron.rye@willisre.com

Scott St. George, Head of Weather & Climate Risk WTW, Scott.stgeorge@wtwco.com