The Role of Climate Change in Rising Severe Storm Costs
The costs of severe storms in the U.S. are rising, but is this due to climate change or something else?
Springtime means severe thunderstorm season—tornadoes, derechos, and hail, among other hazards—for much of the eastern two-thirds of the US.
Damage caused by severe thunderstorms is immense and represents the second most frequent type of billion-dollar disaster in NOAA’s disaster database for the United States. Severe thunderstorms are also the weather category responsible for the second most insured losses globally, according to SwissRe (in both datasets, tropical cyclones or hurricanes are responsible for the most damage). Both U.S. billion-dollar disasters and global insured disaster losses are increasing, and a large fraction of the overall increase seems to be driven by increases in losses from severe thunderstorms.
For many, especially in the media, the natural inclination is to see these rising trends in disaster losses from severe thunderstorms and reflexively attribute them to climate change.
Take, for example, The New York Times’ The Daily podcast (consistently a top 5 podcast) from May 15th, 2024, titled: “The Possible Collapse of the U.S. Home Insurance System.” The episode is about how these increases in insured losses are straining the books of insurers. The reporter, Christopher Flavelle, had this to say:
One of the really striking things about this data was it showed the contagion [climate change-driven disaster losses] had spread to places that I wouldn't have thought of as especially prone to climate shocks. For example, a lot of the Midwest, a lot of the Southeast…
Formerly unimportant weather events like hailstorms or windstorms didn't used to be the kind of thing that would scare insurance companies…But those are becoming so frequent and so much more intense that they can cause existential threats for insurance companies…
This market is starting to buckle under the cost of climate change, and this is all happening really fast.
However, as I emphasized in the previous series on floods, disaster impacts result from the combination of natural hazards (e.g., a physical event like hail or a tornado), exposure (the presence of people, resources, or infrastructure in harm’s way), and vulnerability (the propensity of people, resources, or infrastructure to be adversely affected).
Economic loss trends can be driven by changes in exposure and vulnerability even if the hazard is not changing. That’s why Roger Pielke often says, “Climate data should be the basis for claims of detection and attribution of changes in climate variables, not economic loss data.”
So, what does climate science say about historical and expected changes in severe thunderstorms and their associated hazards of tornadoes and hail?
As I have mentioned in previous posts, it is clarifying to divide scientific evidence into: (1) Historical trends; (2) Fundamental theory; and (3) Mathematical modeling.
Historical trends
Since thunderstorms are on a continuum of severity, it is not necessarily clear how to track changes in severe thunderstorms over time, and there are many different variables you could look at. An obvious place to start would be to just look at the annual number of tornadoes per year and how they are changing, but it turns out this is not as easy as it sounds.
You might think we have good data on phenomena as conspicuous as tornadoes, but unfortunately, not all tornadoes are reported, and there is a strong bias towards increased reporting in recent years due to increases in population and infrastructure that can be damaged by tornadoes (since damage itself is often used to report a tornado especially if it occurred at night). Thus, in order to study long-term trends in tornadoes, researchers need to account for these reporting biases. When studies have done this, they get results like those below.
The red line in the top figure shows no long-term trend in the number of US tornadoes, and the study concludes that “long-term climate change has not substantially affected the domain-wide tornado frequency during the 1975–2018 analysis period.” Interestingly, the study found a statistically significant decreasing trend in the strongest tornadoes (bottom right panel).
Other studies have confirmed no long-term trends in tornado occurrence but have noted changes in other statistics, such as a decrease in the number of days per year with tornadoes combined with an increase in the number of tornadoes on those days. Other studies have shown shifts in geography and seasonality of where tornadoes occur, but it is unclear how much of this might be due to natural decadal variability as opposed to being driven by long-term warming.
Hail is another major component of severe convective storms and is responsible for a great deal of economic losses from both property and crop damage. These losses are roughly $10 billion per year in the US, and some single events reach multiple billions, like a 2012 hailstorm in Phoenix, Arizona, which caused $4 billion in damage.
Hail data suffers from the same issues as tornado data in that it is heavily influenced by reporting biases towards areas with higher population density. However, since at least around 2004, we have radar estimates of hail that don't suffer from these reporting bias issues (but come with their own laundry list of uncertainties and caveats). Nevertheless, this data paints a complicated picture of change that depends on location and season. For the spring, we see increases in severe hail that cause the most damage (greater than 2 inches in diameter) over Texas and Oklahoma but decreases over most of the rest of the country, and for the summer, we see increases in severe hail over the northern high plains and Minnesota but decreases over much of the rest of the country.
The authors note that natural variability (associated with phenomena like El Nino) is responsible for the bulk of this pattern and that it is unclear what, if any, effect long-term warming has had thus far. A recent review titled, “The effects of climate change on hailstorms,” concluded that, “Overall, no clear overarching national climatological hail trend has been found for the U.S.A.”
We see large increases in the economic losses associated with severe thunderstorms, but there are no well-documented increases in the physical hazards responsible for those losses. When reporting biases are accounted for, we see no increase in total tornado counts and perhaps a decrease in the occurrence of the strongest tornadoes. For hail, there is documented natural variability but no established long-term trend associated with climate change.
Fundamental Theory and Mathematical Models
Just because we haven't seen clear historical trends in tornados or hailstorms doesn't mean that warming is having no impact. It just means that so far, the effect is small enough to be buried in the noise of natural variability. So, to assess the effect of warming on the direction and magnitude of change in these hazards, researchers rely on fundamental theory and mathematical models.
Studying severe thunderstorms using climate models is particularly challenging because they are actually smaller than the resolution of the models and, thus, are not directly simulated. Because of this limitation, studies often rely on more indirect evidence. Specifically, research will often use fundamental theory on the ingredients for severe thunderstorms and look at how these ingredients have changed historically or how they change in climate models driven by increases in greenhouse gasses.
Fundamental theory tells us that severe thunderstorms require many ingredients, but at least two are particularly important. One is something called “Convective Available Potential Energy,” or CAPE, and the other is called “wind shear.” Convective Available Potential Energy is, just as it sounds, a measure of how much energy is available in the atmosphere for a storm to tap into. Wind shear is the change in wind speed or direction as you move up in the atmosphere, and it is an important ingredient for storms to develop rotation. You need a lot of both ingredients to get the “supercells” that are responsible for the longest-lived, and most severe tornadoes and hailstorms.
When we look at the historical changes in these ingredients, we don't see much change, which is consistent with the lack of change in reporting bias-adjusted tornado and hail reports.
However, models consistently predict that Convective Available Potential Energy will increase with warming. Wind shear is more uncertain, but the general expectation is that it will stay the same or potentially decrease.
Some studies have run climate change experiments using mathematical models with high enough resolution to explicitly simulate changes in supercells. The figure below shows the simulated difference in supercell counts per year (represented by a measure called UH) between the historical time period and 2100 (under around 3°C of global warming).
They show decreases over the Great Plains and upper Midwest and increases in the South. Overall, they simulate a 7 percent mean increase in supercell counts and a 5 percent mean increase in supercell hours in 2100, but variability is very large, as seen by the large confidence intervals (which represent the range across different years in the simulation).
Given the relatively small impact of long-term warming relative to natural variability (small change in the mean relative to the width of the confidence intervals), it is not surprising that we haven't seen the signal emerge from the noise in historical trends so far. Nevertheless, the balance of evidence seems to suggest that a warming climate will be more conducive to severe thunderstorms.
This wouldn't necessarily translate into an increase in hail, however, because hail melts in warmer temperatures. Essentially, the aforementioned increase in Convective Available Potential Energy should increase the updrafts that are necessary for hail formation, so we should expect to get more hail formed in clouds. However, that hail will fall through a warmer atmosphere and be more likely to melt into rain on the way down. So, which effect wins out? It is uncertain and depends on location and season. There are definitely some locations and seasons where the latter effect should win out. For example, one study showed that hail events mostly become nonexistent in future summers in Colorado.
A variety of studies have indicated that drier and cooler regions in North America will experience the largest increases in hail threat, while warmer and more humid regions will experience a reduced threat.
Exposure and vulnerability
So, changes in the hazard of severe thunderstorms and their sub-hazards are small to date and, for most measures, are expected to be on the order of 10 percent or less this century. What, then, explains the very large changes in economic losses from these events?
In the long term, the expanding bullseye effect that I emphasized in the flood series is a big part of the story.
Stephen Strader provides great examples of this. Below is a recent risk map of severe weather from this spring, in which 775,000 housing units in Oklahoma and Kansas were exposed to a “high risk” of a tornado (right side). But, the exact same risk map would have only included 321,000 housing units (41 percent) had it been issued in 1970.
That’s a specific example, but it generalizes. Systematic studies that encompass larger regions of the country and longer time periods reveal the same phenomena. Below is the number of structures per square km in tornado hotspots, showing an increase from about 2 structures per square km in 1940 to about 10 structures per square km today—a fourfold increase! These increases in exposure to severe weather hazards are much larger than any change in severe weather hazards themselves and thus are the dominant driver of trends in economic losses.
The insurance industry confirms as much. In its recent report on the matter, for example, SwissRe says that the main drivers of increased natural catastrophe losses are increases in exposure due to economic and population growth and increases in the cost of construction (independent of general inflation). They say, "So far, the impact of changing climates has been small."
An interesting finding from the SwissRe report is that hail is responsible for most (50-80 percent) of the insured losses from severe convective storms, and a key driver of increased losses from hail is increased vulnerability (in terms of the potential to incur monetary damage) in the form of increased solar installations. Again, it is not at all clear that the hail hazard has or is expected to become worse with warming. So what is in the way of hail and how much it costs to replace it, makes a big difference.
Similarly, the insurance company Aon had a recent report that breaks down changes in exposure into gross domestic product (GDP), fixed reproducible tangible wealth, property cost inflation, and population distribution in high-hazard zones.
These four components grew at rates of 2.3, 2.1, 2.8, and 1.1 percent per year, which would absolutely dwarf any type of climate change signal in tornadoes or hail. Overall, they found that changes in value exposed account for 93 percent of the overall change in economic losses from severe convective storms (8.3 percent per year compared to the loss trend of 8.9 percent per year), saying that the rest is unexplained and that climate change could be part of it.
It is worth noting that in terms of monetary losses, a great deal of the effect is due to the value exposed and the cost of reconstruction rather than the population exposed. This is part of the reason why fatalities from tornadoes for example, have seen a long-term decline.
Essentially, there are many more people in harm’s way than ever before, but at the same time, we are better protected, warned earlier, and better prepared for severe weather than we ever have been, which has been driving down deaths.
To conclude, we see large increases in economic losses from severe storms, but economic loss data cannot be used to infer changes in severe storms themselves. This is because the economic loss data is heavily influenced by changes in the value exposed and vulnerability of assets. When examining historical trends in tornadoes and hail over the US, we observe little change. However, looking to the future, we expect environments conducive to severe weather to become more common, but these changes are relatively modest, on the order of 10%.
In contrast, the changes in exposure—often referred to as the "expanding bullseye effect"—are an order of magnitude higher or more. This significant increase in exposure underscores that the rising economic losses are more a function of where and how we build rather than the frequency or intensity of the storms themselves. Therefore, the tendency of certain media outlets to attribute these increased losses primarily to climate change is misleading.
The takeaway is that while climate change does play a role, emissions reductions will be an incredibly inefficient lever in the fight against rising economic losses from severe storms. Instead, we should focus on managing exposure and reducing vulnerability in high-risk areas. This involves better planning, stronger building codes, and more effective disaster preparedness.
Patrick T. Brown is a Ph.D. climate scientist, co-director of the Climate and Energy Team at The Breakthrough Institute, and adjunct faculty at Johns Hopkins University.
I think this post is the definition of "erudite" - I feel smarter after reading this!
It also means getting real about policies that include stable, not unlimited growth in population to drive down the asset and personal risk exposure. And one easy way to decrease population growth is cut immigration numbers. If our goal has been to save the whole world by bring them here, that has already failed. So let’s get realistic about who we need to prioritize and that not only isn’t millions and millions of new consumers, nor is it lining the pockets of developers in the name of “smart” growth that as we can see here isn’t improving anything faster than the rate at which it is harmful.