US tornado trends

The vast majority of tornadoes occur in the United States. Zhang et al. have a paper out this month that looks at trends in losses from 1958 to 2018. Here is a graph of the annual normalized losses:

Screenshot 2023-09-27 at 3.51.41 PM

The authors write that “both the severity of damage from individual events and the total annual losses from tornadoes are seen to have reduced over time.”

Time trends in losses from major tornadoes in the United States


1 Like

You can hide a lot in the word “normalized”. Normalization is not necessarily wrong but different normalizations give answers to different questions. It seems they normalize by three things, inflation, national wealth, and population. So the graph is the inflation adjusted loss per person as a fraction of national wealth. This is a far cry from saying tornadoes are decreasing.

Lot’s of other normalizations are possible and answer different questions. No one normalization is “right”. Adjusting for inflation is probably always a good thing. I wonder what the inflation adjusted loss looks like, without the other two normalizations? I wonder what it looks like if you adjust based on median wealth instead of national wealth? What if you adjust for loss as a percentage of wealth of those experiencing the loss? Trailer park residents who lose everything are not comforted by normalizing by national wealth when their personal wealth is destroyed.


Normalization is definitely necessary if you don’t want to fool yourself. For example, the New Madrid earthquakes, up to magnitude 8.2, occurred in 1811. The death toll and economic damage was quite different than if it had happened in 2011 – more people, more buildings, more assets in general.

Zhang writes about normalization:

“Exposure to the tornado damage can increase over time due to population change and economic growth. Meaningful statistical inference requires the use of normalization methods that transform losses so that they have the same exposure. A recent study by Simmons et al. (2013) examine the use of various normalization methods for U.S. tornado damage based on the SPC database. They find that different normalization methods lead to quite different results, but normalized tornado losses generally experience a downward trend.”

So, what does Simmons have to say?
A normalization analysis begins by asking a deceptively simple question: what damage would result if extreme events of the past occurred under the societal conditions of 2011? To answer this question, we employ a methodology of loss normalization that has been well developed over the past 15 years and applied to economic losses for events as diverse as US hurricanes and earthquakes, European floods and windstorms, tropical cyclones in India and China, and Australian extreme weather events…A first adjustment is for changes in inflation…”

OK. Here’s their Figure 1, inflation adjustment only:

Already we can see there is no upward trend since 1964. They go on to write:

“The basic form of a normalization uses base year economic damage (i.e. the index that was presented in the previous section), and applies as relevant, inflation, population, wealth (or housing unit) multipliers…”

Since tornadoes are smaller than hurricanes, they go on to use county data rather than national. And the result?

"The results are quite similar across the three methods. A linear trend fit to each of the three series indicates that normalized damage decreased by 35–63% from 1950 to 2011, depending on the series.

Normalized tornado damage in the United States: 1950–2011



Normalized is nonsense!

Fool who?


Did the study examine or take into account the affects of changes in building codes over time?

I can envision that stricter building codes could reduce losses over time.


1 Like

Also improved detection and warning systems. In general, more developed countries are better off in, say, an earthquake. Chile is likely to come through better than Haiti.


1 Like

Hardly. You do it every time you take inflation into account. Think back to buying a car decades ago. Was it cheaper?

And then, after you adjust for inflation you need to consider features. Did cars have seat belts, much less air bags? What about an AM radio versus satellite radio?

But back to your graph. The discussion has been about losses from tornadoes; your graph is of the number of tornadoes per year during January, February and March.

Detection systems have improved considerably over the last 70 years. The increase in tornado incidence has entirely been in the F1 category (F1 is the weakest, F5 the strongest) because of better detection and reporting.

Figure 1 in the Zhang paper shows the number of tornadoes per year by F-scale categories.

Only the weakest category has increased. F2 through F5 tornadoes have decreased over the decades. Those categories are responsible for about 90% of tornado damage (no surprise).


1 Like

:rofl: :rofl: :rofl:

I get it. We have a richer nation supposedly. But 50% of white males over 50 years old as of 2019 can not afford to retire.

Meanwhile, the costs change in relative terms when people are not insured. That is coming to pass.

You are burying that the storms are getting more frequent and stronger. Your version is a nonstarter.

Not at all. Look again at Zhang’s Figure 1. To quote from the paper:

"The graphs illustrate that the frequency of tornadoes with a rating of F1 on the F-scale has been increasing throughout our sample period, while the frequency of tornadoes with intensity ratings of F2, F3, F4, and F5 has been decreasing.



That certainly is not good enough for me. I have no idea on the veracity of your source.

A simple Google result

One study found the most damaging U.S. hurricanes are three times more frequent than 100 years ago. Why it matters: More severe storms, coupled with more development on risky coastlines, translate to more destruction. Hurricanes that caused more than $1 billion in damage have roughly doubled since the 1980s.Jul 24, 2023

1 Like

Chacun à son goût.

Can we stick to tornadoes in this thread?



The Zhang results successfully replicate the work published by Simmons (cited above). From the Simmons paper:

"We normalize for changes in inflation and wealth at the national level and changes in population, income and housing units at the county level. Under several methods, there has been a sharp decline in tornado damage. This decline corresponds with a decline in the reported frequency of the most intense (and thus most damaging) tornadoes since 1950.


The info Bob posts here about tornadoes is not especially new apart from perhaps the normalization used in the statistical analysis. Here is what the IPCC AR6 chapter on weather extremes (Ch. 11) says about tornadoes

Severe convective storms (tornadoes, hail, rainfall, wind, lightning)

Observed/Detected Trends Since 1950: Low confidence in past trends in hail and winds and tornado activity due to short length of high-quality data records.

Human Contribution to the Observed Trends Since 1950: Low confidence

They also talk about the inability of climate models to capture tornadoes

Climate models consistently project environmental changes that would support an increase in the frequency and intensity of severe thunderstorms that combine tornadoes, hail, and winds (high confidence ), but there is low confidence in the details of the projected increase.


Even in finer-resolution convection-permitting models, it is difficult to directly simulate tornadoes, hail storms, and lightning, so modelling studies of these changes are limited.

1 Like

National normalization trends do not apply symmetrically to regions where tornadoes occur without highly granular modeling of economic changes over the last several decades.

Inflation notwithstanding, rural development (where many of these storms occur) has been quite limited and, in some cases has gone the opposite direction for small towns vs larger coastal urban centers.

As we look into economic impacts, while inflation supports a higher unit replacement cost, the value of units and the density of same for affected areas would show a net negative bias for storm impact damage in these areas.

Even storms hitting the major cities (Rare due to the size, discrete length and sporadicity of tornadic generating storms), with higher unit impacts are likely to hit fewer newer structures due to demographic and socioeconomic trends for the midwest.

Example city growth: Kansas City, Missouri

Of Missouri’s 114 counties, 78 saw population decline reported in the 2020 Census. Ripley County in southeastern Missouri saw the biggest population decline, losing 24% over the past decade.
Housing quality in rural areas

In rural communities, more than half of all households — an estimated 9.4 million households — live in units that were built prior to 1980.

Every year, the homes become less valuable with more and more being abandoned or in dilapidated states of disrepair as compared to similar homes in urban areas where population pressure and economic circumstances foster upkeep and remodeling.

Quite simply, normalization using national trends will obscure some very important factors for tornado alley storm impacts over time.


Sorry about that. I do not pay attention to everything.