Part I
Robert Vislocky, Ph.D.
Over the past few years I’ve been perusing around various climate change debate threads on social media. Unfortunately, all too frequently I see the same flawed (but popular) metrics used to support the narrative that certain forms of severe weather are becoming significantly worse due to global warming. In this series, I’ll review some of these misleading or deceptive metrics and offer a critique on why they are flawed.
With hurricane season officially underway, I’ll kick this off this series by dissecting a measure known as the proportion of major tropical cyclones (TCs) to the total number of TCs (hereafter referred to as “the Ratio”). Several authors have used similar metrics in the past to assess the trend in the proportion of intense to total TCs. For example, see the following article links for Webster, Holland, Curry, and Chang (2005), and Klotzbach and Landsea (2015):
https://pubmed.ncbi.nlm.nih.gov/16166514/
https://journals.ametsoc.org/view/journals/clim/28/19/jcli-d-15-0188.1.xml
However the one most cited lately as evidence of an anthropogenic signal in hurricane intensity is a 2020 PNAS article by Kossin et al. titled “Global increase in major tropical cyclone exceedance probability over the past four decades”. Thus it is this article which will be the main focus.
https://www.pnas.org/doi/full/10.1073/pnas.1920849117
In the 2020 article the authors showed that from 1980 to 2017 the Ratio increased by a statistically significant amount which the authors later conflate to mean that there is an increasing trend in TC intensity. Figure 1 below is a graph of the Ratio out to 2025 that I produced using TC summary data from Colorado State University’s real-time tropical meteorology project. Although the authors in the PNAS article used a different dataset (i.e., 6-hour wind speed fixes inferred from a special satellite dataset), the percent changes in the Ratio from 1980 to 2017 are similar between the PNAS study and Figure 1 below.
https://tropical.atmos.colostate.edu/Realtime/index.php?arch&loc=global
Of course, upon publication of the PNAS study the media as always jumps on the climate bandwagon. For example, the New York Times story below says that “Climate Change is Making Hurricanes Stronger” and cites the details from the PNAS article:
https://www.nytimes.com/2020/05/18/climate/climate-changes-hurricane-intensity.html
NOAA also claimed a “significant global increase in hurricane intensity over a four decade period” as a result of the PNAS study:
https://www.ncei.noaa.gov/news/link-between-heat-and-hurricanes
So, what’s the problem? Simply put, the proportion of major TCs to total TCs is NOT a valid measure of absolute hurricane intensity. It is a relative measure with a moving denominator. It would be like trying to equate a high relative humidity with a high dew point, the two don’t necessarily go hand-in-hand. To show this flaw using an extremely simplified example, consider a season in which there was only one major hurricane and only one total hurricane. The Ratio metric would yield a value of 1, which is the highest value possible supposedly indicating a very intense hurricane season. Yet, by any reasonable measure a season with only one hurricane would be an extremely benign hurricane season.
This is not just a hypothetical example either as glaring discrepancies can be shown over an entire year at the global level using actual data (see table below). As one example, 2018 had 43% more TCs and 43% more major TCs than 2024. It also had a 75% higher ACE value. Yet according to the flawed Ratio statistic, both years were identical!! As another example, 1990 had 53% more total TCs and the same number of major TCs than 2009. Plus 1990 had a 53% higher ACE value than 2009. Yet the Ratio in 2009 was 53% higher than 1990 which is an absolute joke. There are numerous other discrepancies like those apparent in the annual TC data. The bottom line is the Ratio is not a suitable metric for measuring absolute TC intensity or computing the trends thereof, as it can increase with decreasing numbers of total TCs. Despite its flaw the Ratio is frequently referenced whenever a climate scientist or activist needs to cite a trend that shows hurricanes are becoming more intense (mostly because other indicators like global ACE in the post satellite era, or US landfalling hurricanes over the past century, show no significant trend).
Year # Total TCs # Major TCs ACE Ratio
2018 60 33 1105 0.55
2024 42 23 630 0.55
Year # Total TCs # Major TCs ACE Ratio
1990 58 21 930 0.36
2009 38 21 607 0.55
Data courtesy of Colorado State University Department of Atmospheric Sciences.
https://tropical.atmos.colostate.edu/Realtime/index.php?arch&loc=global
Another interesting observation is that although the authors of the PNAS article were able to show a statistically significant trend in the Ratio statistic from 1980 to 2017, there has been no significant trend over the last 31 years from 1994 to 2025 (see Figure 2 below).
https://tropical.atmos.colostate.edu/Realtime/index.php?arch&loc=global
H/T to Charles Rotter
[editor’s note. I identified an error in the original submission~cr]
I can see a blatantly obvious flaw even as a geologist, not even close to a climatologist; it comes from the nature of orebodies.
If there is a relatively constant supply of tropical cyclone-generating energy available in any given year due to planetary hysteresis effects, any year in which cyclones develop the fewer the number, the higher the average dissipated energy per cyclone has to be.
I’m not a climatologist, but does that not mean that when there are fewer cyclones, that ratio will be higher?
As the phenomenon of rotating areas of atmosphere at the planetary surface is the topic, is not the the energy contained in the aggregate simply the energy being transferred from the warmer parts of the globe to the parts of the globe shedding the energy into outer space. This will vary with the variation of the total insolation and the lag time of the energy transport system. If there is a maths model for this process, expect it to include predominantly thermodynamic processes as well as rotational inertial properties.
Each individual whirlwind will exist having been commenced by random chance at that location at that particular time. Hint: Chaos theory and butterfly wings.
I only mention this as I cannot see how ore bodies have been collected by anything like the same processes. I am happy to be enlightened otherwise.
‘Well would you believe this?’– for a given source-rock quantity of gold being mobilized by a hydrothermal system (although I guess that does involve thermodynamics) if it precipitates in fewer higher-grade veins than in many lower-grade ones, the ratio of high grade veins to low grade ones will be higher. It was intended as a simple concept, somewhat comparable to a reduction in extreme lows, maybe due to increasing night-time cloud cover, making average temperatures seem to rise although highs do not. But many interpret the increasing average as an increase in atmospheric heat.
(Please understand that intro was a reference to Maxwell Smart.)
Yes, the number of smaller hurricanes has dropped a bit since about 1998.
So yes, as Charles says, taking a ratio of “small” to “big” is meaningless.
Given your stated assumptions, yes the energy per cyclone would be higher and given the math of this bogus statistical attribution study, the ration would be skewed as you suggested.
“Figure 1 below is a graph of the Ratio”
But it is not the same Ratio. The PNAS article averages global hurricanes over three years. That matters, because the quibble here is that you might get strange results if there were only one hurricane, say. Well, you might, but globally over three years the number of hurricanes is always large.
“there has been no significant trend over the last 31 years from 1994 to 2025″
Yes. The authors noted that. That is why they made the effort to extend the homogeneous series to a period where a significant result might be obtained. And it was.
30 years is “climate”.
Hurricanes data is not doing anything untoward.
No surprises with your ‘discovery’ here, Nick.
In the climate catastrophe tend-show, everyone who pays the entry ticket price to the barkers (media), or knows a guy who knows a guy, gets to jump up on stage with a graph of their take on how doomsday is progressing. Or not.
The main problem with the tent-show’s production is that any & every performer can draw up their own props using bits & pieces they can find in the props storage shed.
It makes for a piss-poor show, and that’s why increasing numbers of punters are now demanding their ticket money back.
(bonus thought – maybe if the tent-show hadn’t been promoted as something resembling “a tragedy in infinite acts”, it wouldn’t now be cratering at the science box office?)
A bad statistic averaged over multiple years, is still a bad statistic.
It is a different statistic, and not bad.
It is data mining for a predetermined result. Do we know any of these values accurately in the 1800’s, 1700’s, or 1600’s? Geez, how many hurricanes that never made landfall in the 1900’s are included here?
The first thing researchers should do is to check to see if the data is sufficient to make reasonable conclusions. It would appear here that sufficient data is not available to gauge human factors at all let alone natural climate changes. 100 years from now, there may be enough data to make a reasonable determination, but not now.
It started off bad, and averaging a bunch of bad numbers together doesn’t create a good number.
Nick,
They extended the series as far as satellite data was available. You might also ask yourself how many satellites were providing data in 1980 versus 2020.
Grouping it into 3 years doesn’t matter. There are still glaring discrepancies.
Example #1, 1990-1992 vs 2008-2010.
1990-92 had 78 major TC, 164 total TC and an ACE of 2954. 2008-10 had only 64 major TC, 117 total TC and an ACE of 1744. By all legit measures, 1990-92 was more active. But the Ratio says the latter period was more active (0.55 vs 0.48).
Example #2, 1996-1998 vs 2020-2022. 1996-98 had 76 major TC, 166 total TC, and an ACE of 2833. 2020-22 had 57 TC, 127 total TC, and an ACE of 1814. By all legit measures 1996-98 was way more active. Yet the Ratio says they were about the same (0.45 vs 0.46).
These are just two out of several more that can be found even when grouping into 3-year chunks. And note that these discrepancies always seem to boost the ratio in the later years causing a false upward trend.
And no the authors didn’t note there was no trend in the Ratio from 1994-2025 because their analysis only went to 2017.
“Grouping it into 3 years doesn’t matter.”
It’s what they did. And if you don’t do that, and find a fault, then the fault is in your statistics, not theirs.
“But the Ratio says the latter period was more active”
No, it doesn’t. That is your fallacy here. The Ratio says what it says. A greater fraction of the hurricanes were major. It does not claim to be a measure of ACE activity.
But they do claim that the proportion is a measure of TC intensity (as does the media). It is not a valid measure of intensity.
And in my last reply I did re-run the numbers and group into 3 year chunks. As shown above the obvious discrepancies are still there, unless you think a 3 year stretch with 64 major TCs and 117 total TCs is 15% more intense than a 3 year stretch of 78 major TCs and 164 total TCs.
Intensity is not the same as activity, and is not measured by ACE.
Consider two periods A and B with the same number of major hurricanes, but B has more non-major hurricanes. It could happen. A would have a higher Ratio (intensity), but B would probably have a higher ACE. There is no contradiction there.
B is more intense. Intensity in science is the “amount of energy passing through a given area over a set period of time”. That sounds more like ACE to me than whatever the Ratio purports to measure.
‘ “there has been no significant trend over the last 31 years from 1994 to 2025″
Yes. The authors noted that. That is why they made the effort to extend the homogeneous series to a period where a significant result might be obtained. And it was.’
Search and arrange the data to find the result you want. That isn’t science, that’s advocacy.
No, it’s proper science. We see a trend, but there isn’t enough data so say that it is significant. So get more data.
No. they worked until they could FIND a trend they wanted.
as you noted in your own words :
“That is why they made the effort to extend the homogeneous series to a period where a significant result might be obtained. And it was.”
Choosing data to fit your need is not science.
You use all the data you can get.
Here, with effort, they were able to extend the series.
“You use all the data you can get.” or in this case…make.
No new data was collected. Old information was rehashed until it showed what the was expected.
FTA:
“Based on physical understanding and robust support from numerical simulations, an increase in environmental potential intensity is expected to manifest as a shift in the TC intensity distribution toward greater intensity and an increase in mean intensity”
not based on observations or data….based on models and there ‘understanding”.
Also, per the article:
“The ADT-HURSAT should be considered a record that sacrifices some measure of absolute accuracy for homogeneity, and which allows more robust trend analysis.”
So the authors already know that they are on shaky ground, and giving up accuracy for ease of calculation.
Congratulations on robustly calculating the number of angels that can dance on the head of a pin.
Advocacy, not science.
What you just described is called P-hacking.
From Fig 2..compared to Figure 1…
…. it is obvious that the increased CO2 in the atmosphere this century has stabilised the hurricane ratio…. . 😉
The numbers of hurricanes or cyclones has dropped a lot since 1900.
Facts and data doesn’t matter to the climastistas, hot earthers, climate charletins, algores, anti-capitalists, communists, population control-ists, and the like. They can make shrill exclamations of the end of the earth, humanity, and the lap dog media laps it up.
Emotionalize an issue to sway public opinion.
Nothing to do with science, obviously, and everything to do with social neurosis.
I’ve seen this kind of manipulation many times.
A male weight lifter of 200 lb body mass can press 200 lb.
A female weight lifter of 160 lb body mass can press 160 lb.
Therefore there is no male advantage as both can press 1 lb. per lb. of body mass.
Statistics don’t lie, but liars use statistics.
Bingo! I usually see this stated as “Figures may not lie, but liars sure can figure.” 😉
Your way is more elegant as befits discussion on a higher plane. 🙂