This would be a good anecdote to include in a presentation that focuses on encouraging more critically thinking about climate change (and energy and emissions). Any update on publishing a ready-made guide on this, for putting on presentations to smaller or specific groups and thus extend the reach of this messaging? Apologize if I messed something that’s already been done!
I do not understand how wind speed can be used to characterize storm intensity without regard for duration and extent. Further, the statistical analysis of this paper (and many others) does not incorporate consideration of uncertainty and errors in the data which are certainly present. Don't statisticians tell us that regression only produces worthwhile results on data that has been verified?
Terrific analysis Roger. Reading this post I kept wondering why this happens - why faulty claims in published research become so amplified. And of course its because the data and results don't support the claims people have set out to make in the first place, and so they compensate. Consequently, the ethics of science and antics of scientists come under scrutiny creating the much bigger mess that you and others have been unpacking for years.
It is hard to impossible for news media to detect and mitigate abuses. Social media just makes it worse by providing an enabling platform for amplification. The researchers who lean into this style of 'performance science' have no motivation to change (worse, some have even sued others who have tried to correct the record). Motivation to increase transparency is, in this arena of climate and weather attribution, pretty low because so many interests benefit from the amplification.
How to fix this?
1- require reviews of university press releases. In this age of unhappiness with higher ed you'd think this would already be happening but I know (you know) that it isn't. Researchers want to promote their research and universities want to promote their researchers. There is no correcting mechanism. Government and private labs are just as bad.
2 - every major media outlet including weather services, some of the worst abusers, should have ombudsmen to police coverage.
3 - all of us who are privileged to have college degrees, especially in science, are duty bound to read between the lines and do the critical thinking. The entire climate politics mess lies at our feet - we abdicated our responsibility and worse, some of us are in the percentage who succumb to political agendas. In the end as usual we get what we deserve.
Roger, we worked extensively with the ERA group, particularly when they first started. The are excellent to work with and I would encourage you to contact them to discuss the discontinuities in the ERA due to observing system changes. I recall they wrote a number of Tech papers internally documenting much of this.
As I recall, major observing system changes that cause discontinuities in the time series include:
1. No upper air data prior to early 1950s (after the 'jet stream' was discovered by bombing of Japan in WW2)
2. 1978 - start of global infrared satellite sounder data (HIRS) - although a major advance, IR data is not able to observe below cloud top and so limited in storm systems
3. mid-1980s - some microwave imager data (but not sounder data) to help identify storm centers over the oceans
4. 1998 - start of microwave sounder data (AMSU). A major advance in sounding in cloudy atmospheres and a big advance in better tracking of storm centers.
I'm just back from travel and will need to more thoroughly read you paper a bit later. Will send additional feedback then. Regards
This is a wonderful snapshot on the biggest problem of all in these debates - the narrow, puny really, time window in which we have "modern" data and, even more, understandings! The jet stream always existed, I would put "discovered" in quotes. That is the anthropomorphic effect in all of this - our limited human lens.
Let me do a little further digging into this. The impacts of the observing system changes on ERA5 should be easily seen on the 500 mb height skill scores. ERA usually produces these times series and increasing skill over time is usual, and often shows clear jumps with the introduction of the new observing system. I'll see if I can find that or other similar metric.
A similar, but more dramatic problem occurred when deriving a trend for tropical cyclones (TCs) from data that is not homogenous. The strength of hurricanes on the Simpson-Saffer (Category 1-5) is assessed from pictures from space using the "Dvorak method". (IIRC, a clear eye in the center of a TC becomes readily apparent once a storm reaches Category 3. In the infrared, cloud top temperature get colder as hurricanes get stronger.) Planes regularly fly into TCs to measure wind strength only in the North Atlantic. Unfortunately, the resolution and quality of images of TCs from space have improved over the decades. A group of IPCC authors lead by Kossin degraded all satellite images to the resolution available at the start date of 1979, assigned Categories to those images using the same automated Dvorak software and looked at the increase in all TCs and all major TCs (Category 3 and above) between an early period and a late period. However, they didn't say much about about how homogenization changed the data. Being suspicious about everything climate scientists do, I calculated the changes produced by homogenization for myself: 14% fewer major TCs in the early period and 31% fewer in the late period. For all TCs, there were 27% fewer detected in the early period and 36% fewer detected in the later period.
If the view shared by both IPCC and leading climate realists that the from background noise emergence of many exteme weather phenomena will not occur this century is correct, this type of analysis also has limited value.
We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing.
Kim wrote: "We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing."
I disagree. In its 2012 Special Report on Extreme Events and Disasters (SREX), the IPCC concluded through statistical analysis of observations that at least three forms of extreme weather had increased or decreased significantly since 1950:
"It is very likely that there has been an overall decrease in the number of cold days and nights, and an overall increase in the number of warm days and nights, at the global scale, that is, for most land areas with sufficient data."
"There have been statistically significant trends in the number of heavy precipitation events in some regions. It is LIKELY that more of these regions have experienced increases than decreases, although there are strong regional and subregional variations in these trends."
"Likely" is not a statistically significant change, but in AR6, Figure 11.13 shows that the observed trends in the percentage of stations showing a statistically significant increase in annual maximum daily precipitation (RX1day) is statistically significant.
I mentioned Guadalupe River flash flood mostly only to illustrate the type of short intense precipitation that is increasing. In his linked article on that flood, Roger cites different statistics: Little increase since 1981 (not 1950) in a different metric R95p (days precipitation is above 95th percentile), flooding in general, and falling deaths from flooding. First of all, most "flooding" is mostly caused by "cut off lows" that cause large areas of rain to remain stationary for multiple days and that type of flooding is not being increased by warming associated with rising GHGs. Furthermore, the rivers that generalized flooding occurs on have been highly modified by levees and dams. Trends in this type of flooding aren't claimed to be caused by rising GHGs. Only the type of short intense rainfall that produces flash flooding is predicted to increase with rising GHGs/temperature and has been observed to increase.
Short intense precipitation is increasing in most places, but it only produces highly dangerous flash flooding where the local geography amplifies the danger, such as the Texas Hill Country. We apparently can observe a statistically significant increase in short intense rainfall in general, but do not have enough events in locations where it causes flash flooding to detect a statistically significant increase. The lack of warning systems on the Guadalope River was tragic negligence on the part of the local government. As best I can tell, my info on increasing short intense precipitation isn't contradicting anything Roger has written, (but that information is somewhat less compelling than I remembered from reading the IPCC's Special Report on Extreme Weather.)
For Americans who live in urban areas, the danger from increasing short intense precipitation is combined with the increase in runoff from increasing impenetrable surfaces into local streams. Localized flash flooding can carry away children and cars.
Respectfully Roger and Kim: There are a number of problems with suggesting that the background noise in the emergence of extreme weather phenomena means policy choices won't make any significant difference. I'm commenting mostly on the graphs in the "jaws of the snake" post you liked above.
The most obvious problem is ignoring the "body of the snake" (warming until present) and just focusing on the width of the jaws (future uncertainty). The simplest example is the change in short, intense, localized downpours that produce flash foods in vulnerable areas. The flash flood on the Guadalupe in Texas last summer that killed at least 135 people (including 20 children at a summer camp) is the kind of extreme flooding that already has been demonstrated to have significantly increased because of global warming/rising GHGs.
Future policy choices can't change what has already happened (the "body of the snake", for example 1+ degC of warming since 1970), however we attribute that warming to rising GHGs.
The overlap or non-overlap of confidence intervals associated with two lines technically is not a valid way to determine if there is a significant difference (usually in temperature) between the outcomes from two scenarios. The chances of two 95% confidence intervals just barely touching is 2.5%*2.5% = 0.065%, less than one in 1,000. Technically, you should probably use the formula for the statistical significance of the difference between two means. IIRC, in the simplest situation, the standard deviation for the difference of two means is bigger by a factor of SQRT(2), not two-fold bigger, than the standard deviation of each mean.
Even worse, the errors in each scenario are not randomly distributed, a requirement for the statistical significance of a difference to be meaningful. For example, if climate sensitivity proves to be lower than expected, the temperature in both scenarios will be lower.
In short, he statistical significance of the difference in temperature or other outcome from two policy choices or scenarios probably should not be deduced from whether the "jaws of the snake" are open or closed.
I'm not really sure what you mean by "policy choices" however, with respect to disaster planning, such as releated to the Texas floods, I discuss that here:
Roger: Kim wrote comment below, which you seemed to endorse by pointing to your "jaws of the snake" post
"If the view shared by both IPCC and leading climate realists that the from background noise emergence of many exteme weather phenomena will not occur this century is correct, this type of analysis also has limited value."
"We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing."
I replied above that we already have observed an increase in SOME forms of extreme weather, but not others such as hurricanes. In this comment, I'm trying to clarify my remarks about confidence intervals and their possible misuse in your "jaws of the snake" post.
Let's consider the temperature projections from two scenarios: RCP4.5 and RCP6.0. Trying decarbonize enough to change from a RCP6.0 to RCP4.5 is a policy choice. When the confidence intervals on those projections overlap (the jaws of the snake are closed), it appears as there will be no statistically significant benefit (lower temperature) from decarbonization. That is wrong. Uncertainty about our planet's climate sensitivity to rising GHGs contributes the confidence interval around each projection. However, our planet does have a fixed climate sensitivity, be it 2, 3, or 4 degC per doubling of CO2 (or the equivalent in other forcing). If climate sensitivity turns out to be 2, both projections will show lower temperatures and if climate sensitivity turns out to be 4, both projections will be higher. In either case there will still be a statistically significant DIFFERENCE in the outcome despite overlapping confidence intervals.
Confidence intervals are intends to show the effect of random noise in data. The possibility that climate sensitivity could be 2,3 or 4 is random noise in projections of RCP4.5 and RCP6.0. However climate sensitivity isn't random noise in projections of the DIFFERENCE BETWEEN RCP4.5 and RCP6.0. Whatever the correct value for our planet's climate sensitivity, it will have the same value in an RCP4.5 future and an RCP6.0 future.
There are other mathematical problems that come from try to interpret the apparent statistical significant of the overlap in confidence intervals that comes from standard deviations adding "in quadrature" (the square root of the sum of the squares), not arithmetically, but let's avoid confusion and focus on the role uncertainty in climate sensitivity plays in the DIFFERENCE between to projections.
The world is presently undershooting RCP4.5, our 2021 paper suggested a trajectory most consistent with SSP2-3.4. For argument purposes, let's just take those two. In terms of global average temperature, it is unlike that we could detect a change from 4.5 to 3.4 this century. I am already on board with the goal of accelrating the rate of decarbonization. However, basing such a goal on promises of detectable changes in climate experienced by people alive today does not seem well supported by climate science.
@Frank I just want to point out that the comment I made shouldn't be interpreted as saying anything about policy decisions. If we know why some potential material effect can't be observed, we have to act on what we can model / predict.
In the field of economics where I've been involved with creating attribution models, we basically don't have any observations we can ever make to validate what we predict. We simply have to make the best models we can and trust them.
Recognizing that emergence takes one hundred years (as an example) isn't a reason not to act in my view. It's a reason TO act - assuming we have other reasons beyond observed data to do so.
I'm sorry I can't reply more intelligently but I'm not quite able to follow the jaws of the snake argument.
Thank you for your great dissection of this paper. Truncating data and failing to address contrary data seems to be typical for the leading climate catastrophe cheerleaders. Their findings of increased economic destruction seems to have missed your outstanding research in why recent storms etc. are more costly.... more stuff to break. Thank you for keeping your writing understandable to those of us who laymen.
Great work as always Roger. I really appreciate how you assist the non scientifically trained ( I'm from a farming background) to better understand how to interpret the deluge of climate papers "proving that the science is settled " ....you can insert your headline of choice...... Thanks
Chris, I am even more lacking in expertise coming from a life in the financial world in the UK.
I subscribe to get full access to the voice of an expert sceptic. Not a denier but a sceptic. I am happy to accept his opinion on these winds, but I then came across a paper published in Weather and Climate Extremes which found that extreme heat in Central and Southern Europe has increased tenfold between 2010 and 2024 compared to the 1961-1990 period.
This backs up findings from the European State of the Climate. 2024 and 2024 Lancet Countdown Europe.
I find it very hard, as a layman, to evaluate the competing claims.
You write, "Independent validation studies confirm that ERA5 systematically underestimates wind speeds within extratropical cyclone centers, with elevated errors near the storm core and in the Gulf Stream region where nor’easters intensify most explosively."
Is it possible to use ERA5 methods to extrapolate from modern land-based observations (using a subset similar to the observations of the pre-satellite era), to create a more uniform time series?
Well done, excellent post. Have you considered how to broaden the reach of your findings beyond us Substack THB nerds? Maybe you can do a joint piece with Jonah Goldberg at The Dispatch.
I recall that Ben Zycher asserted in his recent THB post that as a general rule Climate Scientists are not very good statisticians. That seems to be born out by the sloppy work of the UPenn group.
On your use of AI, it would seem that a full professor at a research university would be using grad students to do the grunt work on data analysis and coding that you appear to have subcontracted to Claude. True or false?
Roger, some questions. I am interested in an insider's understanding of how climate science reporting works. Would I be correct that your reanalysis is a type of replication that we have so often been taught is an important part of scientific method, but is not highly rewarded by academia, and so is seldom done?
How might the IPCC's future assessment treat both the paper and and your analysis? Will either or both be cited?
Do you expect science news websites like "Live Science" or "Inside Climate News" to note your analysis? Are climate news websites, or general news outlets like NYT or The Guardian, responsible enough to acknowledge that some specific climate reporting needs to be re-evaluated or retracted on the basis of new analysis?
Does PNAS publish the sort of rebuttal you have made here in your sub stack or otherwise address it?
Modern science journalism is indeed a highly competitive business and I believe this drives a search for more sensational and catastrophic headlines. There are a lot of people who dwell on the negative and on perceived threats because they feel emotions that everyday life denies them.
Just as the regime media were competing against each other to see who could do the most to get Trump removed from office, science media competes to spread the alarmist narrative. The modern journalism space is really no better than the yellow journalism of the Gilded age or the blindly and nasty partisanship of the early 19th century. We forget that newspapers in the 1850's for example regularly portrayed Lincoln as a monkey.
Would I be correct that your reanalysis is a type of replication that we have so often been taught is an important part of scientific method, but is not highly rewarded by academia, and so is seldom done?
RP: Yes, yes, and yes
How might the IPCC's future assessment treat both the paper and and your analysis? Will either or both be cited?
RP: We shall see. My analysis will not be cited at it is not in the peer reviewed literature. See my experience with PNAS on the Grinsted paper. Not worth my efforts to publish.
Do you expect science news websites like "Live Science" or "Inside Climate News" to note your analysis?
RP: Not a chance.
Are climate news websites, or general news outlets like NYT or The Guardian, responsible enough to acknowledge that some specific climate reporting needs to be re-evaluated or retracted on the basis of new analysis?
RP: I’ve seen no evidence of this.
Does PNAS publish the sort of rebuttal you have made here in your sub stack or otherwise address it?
RP: PNAS publishes 500 word comments on published papers. I tried this on Grinsted and was stonewalled. Give who edited this paper not worth my time.
“The over-hyping of results and shedding of uncertainty and complexity is a pattern we often see in high-profile climate research, and chronicled in detail over the years here at THB. Loose standards of scientific quality may help to generate breathless headlines, but does not serve improved scientific understandings or public confidence in climate science.”
Strong, long recognized incentives in academia for this phenomenon wouldn’t you agree Roger? Do you think there has been any improvement? Perhaps counter-intuitively I would expect this sort of thing to worsen under a regime that cuts science funding and tries to undermine the standing of scientists in society. Potentially a bit of a vicious circle.
In climate science this phenomenon is indeed incentivized by academic reward structure. Note that Mann's paper was edited by his colleague Jim Kossin whose own work (on TCs) is highly dubious. I don't think the political regime makes much difference to these incentives. Though, I do think public health is different .. Stay tuned on that!
Was Claude a miner in West Virginia before retraining as a coder? Thanks for finding the time and knowledge to analyze the Chen paper.
Any paper with Michael Mann's imprimatur is and should be highly suspect. His viscous political science and legal reputation is well earned, as is his poor scientific work on the "hockey stick". I have great difficulty not just rejecting his work out of hand.
Mann's track record means he no longer gets the benefit of the doubt. The one minor saving grace here is that at least the Supplementary Analysis seemed more honest.
This would be a good anecdote to include in a presentation that focuses on encouraging more critically thinking about climate change (and energy and emissions). Any update on publishing a ready-made guide on this, for putting on presentations to smaller or specific groups and thus extend the reach of this messaging? Apologize if I messed something that’s already been done!
Somehow I had missed the news that Michael Mann moved from Penn State to U Penn. Why would a university hire a guy who is well known to be a liar (and, from a distance, seems also to be an a-hole)? See: https://www.thedp.com/article/2025/03/penn-michael-mann-sanctioned-judge-defamation-trial
I do not understand how wind speed can be used to characterize storm intensity without regard for duration and extent. Further, the statistical analysis of this paper (and many others) does not incorporate consideration of uncertainty and errors in the data which are certainly present. Don't statisticians tell us that regression only produces worthwhile results on data that has been verified?
Amen
Terrific analysis Roger. Reading this post I kept wondering why this happens - why faulty claims in published research become so amplified. And of course its because the data and results don't support the claims people have set out to make in the first place, and so they compensate. Consequently, the ethics of science and antics of scientists come under scrutiny creating the much bigger mess that you and others have been unpacking for years.
It is hard to impossible for news media to detect and mitigate abuses. Social media just makes it worse by providing an enabling platform for amplification. The researchers who lean into this style of 'performance science' have no motivation to change (worse, some have even sued others who have tried to correct the record). Motivation to increase transparency is, in this arena of climate and weather attribution, pretty low because so many interests benefit from the amplification.
How to fix this?
1- require reviews of university press releases. In this age of unhappiness with higher ed you'd think this would already be happening but I know (you know) that it isn't. Researchers want to promote their research and universities want to promote their researchers. There is no correcting mechanism. Government and private labs are just as bad.
2 - every major media outlet including weather services, some of the worst abusers, should have ombudsmen to police coverage.
3 - all of us who are privileged to have college degrees, especially in science, are duty bound to read between the lines and do the critical thinking. The entire climate politics mess lies at our feet - we abdicated our responsibility and worse, some of us are in the percentage who succumb to political agendas. In the end as usual we get what we deserve.
Roger, we worked extensively with the ERA group, particularly when they first started. The are excellent to work with and I would encourage you to contact them to discuss the discontinuities in the ERA due to observing system changes. I recall they wrote a number of Tech papers internally documenting much of this.
As I recall, major observing system changes that cause discontinuities in the time series include:
1. No upper air data prior to early 1950s (after the 'jet stream' was discovered by bombing of Japan in WW2)
2. 1978 - start of global infrared satellite sounder data (HIRS) - although a major advance, IR data is not able to observe below cloud top and so limited in storm systems
3. mid-1980s - some microwave imager data (but not sounder data) to help identify storm centers over the oceans
4. 1998 - start of microwave sounder data (AMSU). A major advance in sounding in cloudy atmospheres and a big advance in better tracking of storm centers.
I'm just back from travel and will need to more thoroughly read you paper a bit later. Will send additional feedback then. Regards
This is a wonderful snapshot on the biggest problem of all in these debates - the narrow, puny really, time window in which we have "modern" data and, even more, understandings! The jet stream always existed, I would put "discovered" in quotes. That is the anthropomorphic effect in all of this - our limited human lens.
Thank you!
Let me do a little further digging into this. The impacts of the observing system changes on ERA5 should be easily seen on the 500 mb height skill scores. ERA usually produces these times series and increasing skill over time is usual, and often shows clear jumps with the introduction of the new observing system. I'll see if I can find that or other similar metric.
A similar, but more dramatic problem occurred when deriving a trend for tropical cyclones (TCs) from data that is not homogenous. The strength of hurricanes on the Simpson-Saffer (Category 1-5) is assessed from pictures from space using the "Dvorak method". (IIRC, a clear eye in the center of a TC becomes readily apparent once a storm reaches Category 3. In the infrared, cloud top temperature get colder as hurricanes get stronger.) Planes regularly fly into TCs to measure wind strength only in the North Atlantic. Unfortunately, the resolution and quality of images of TCs from space have improved over the decades. A group of IPCC authors lead by Kossin degraded all satellite images to the resolution available at the start date of 1979, assigned Categories to those images using the same automated Dvorak software and looked at the increase in all TCs and all major TCs (Category 3 and above) between an early period and a late period. However, they didn't say much about about how homogenization changed the data. Being suspicious about everything climate scientists do, I calculated the changes produced by homogenization for myself: 14% fewer major TCs in the early period and 31% fewer in the late period. For all TCs, there were 27% fewer detected in the early period and 36% fewer detected in the later period.
https://www.pnas.org/doi/pdf/10.1073/pnas.1920849117 (and correction)
If the view shared by both IPCC and leading climate realists that the from background noise emergence of many exteme weather phenomena will not occur this century is correct, this type of analysis also has limited value.
We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing.
Kim wrote: "We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing."
I disagree. In its 2012 Special Report on Extreme Events and Disasters (SREX), the IPCC concluded through statistical analysis of observations that at least three forms of extreme weather had increased or decreased significantly since 1950:
"It is very likely that there has been an overall decrease in the number of cold days and nights, and an overall increase in the number of warm days and nights, at the global scale, that is, for most land areas with sufficient data."
"There have been statistically significant trends in the number of heavy precipitation events in some regions. It is LIKELY that more of these regions have experienced increases than decreases, although there are strong regional and subregional variations in these trends."
https://www.ipcc.ch/report/managing-the-risks-of-extreme-events-and-disasters-to-advance-climate-change-adaptation/
"Likely" is not a statistically significant change, but in AR6, Figure 11.13 shows that the observed trends in the percentage of stations showing a statistically significant increase in annual maximum daily precipitation (RX1day) is statistically significant.
I mentioned Guadalupe River flash flood mostly only to illustrate the type of short intense precipitation that is increasing. In his linked article on that flood, Roger cites different statistics: Little increase since 1981 (not 1950) in a different metric R95p (days precipitation is above 95th percentile), flooding in general, and falling deaths from flooding. First of all, most "flooding" is mostly caused by "cut off lows" that cause large areas of rain to remain stationary for multiple days and that type of flooding is not being increased by warming associated with rising GHGs. Furthermore, the rivers that generalized flooding occurs on have been highly modified by levees and dams. Trends in this type of flooding aren't claimed to be caused by rising GHGs. Only the type of short intense rainfall that produces flash flooding is predicted to increase with rising GHGs/temperature and has been observed to increase.
Short intense precipitation is increasing in most places, but it only produces highly dangerous flash flooding where the local geography amplifies the danger, such as the Texas Hill Country. We apparently can observe a statistically significant increase in short intense rainfall in general, but do not have enough events in locations where it causes flash flooding to detect a statistically significant increase. The lack of warning systems on the Guadalope River was tragic negligence on the part of the local government. As best I can tell, my info on increasing short intense precipitation isn't contradicting anything Roger has written, (but that information is somewhat less compelling than I remembered from reading the IPCC's Special Report on Extreme Weather.)
For Americans who live in urban areas, the danger from increasing short intense precipitation is combined with the increase in runoff from increasing impenetrable surfaces into local streams. Localized flash flooding can carry away children and cars.
Exactly — This is discussed as well here: https://rogerpielkejr.substack.com/p/jaws-of-the-snake
Respectfully Roger and Kim: There are a number of problems with suggesting that the background noise in the emergence of extreme weather phenomena means policy choices won't make any significant difference. I'm commenting mostly on the graphs in the "jaws of the snake" post you liked above.
https://rogerpielkejr.substack.com/p/jaws-of-the-snake
The most obvious problem is ignoring the "body of the snake" (warming until present) and just focusing on the width of the jaws (future uncertainty). The simplest example is the change in short, intense, localized downpours that produce flash foods in vulnerable areas. The flash flood on the Guadalupe in Texas last summer that killed at least 135 people (including 20 children at a summer camp) is the kind of extreme flooding that already has been demonstrated to have significantly increased because of global warming/rising GHGs.
Future policy choices can't change what has already happened (the "body of the snake", for example 1+ degC of warming since 1970), however we attribute that warming to rising GHGs.
The overlap or non-overlap of confidence intervals associated with two lines technically is not a valid way to determine if there is a significant difference (usually in temperature) between the outcomes from two scenarios. The chances of two 95% confidence intervals just barely touching is 2.5%*2.5% = 0.065%, less than one in 1,000. Technically, you should probably use the formula for the statistical significance of the difference between two means. IIRC, in the simplest situation, the standard deviation for the difference of two means is bigger by a factor of SQRT(2), not two-fold bigger, than the standard deviation of each mean.
Even worse, the errors in each scenario are not randomly distributed, a requirement for the statistical significance of a difference to be meaningful. For example, if climate sensitivity proves to be lower than expected, the temperature in both scenarios will be lower.
In short, he statistical significance of the difference in temperature or other outcome from two policy choices or scenarios probably should not be deduced from whether the "jaws of the snake" are open or closed.
Hi Frank,
I'm not really sure what you mean by "policy choices" however, with respect to disaster planning, such as releated to the Texas floods, I discuss that here:
https://rogerpielkejr.substack.com/p/the-texas-flash-floods
I don't think any argument over attribution is particularly relevant to saving future lives in flash flood alley.
More generally, the relative role of adaptation to climate vs adaptation to climate change is the subject of our new paper: https://rogerpielkejr.substack.com/p/the-economics-of-climate-adaptation
Happy to engage further!
Roger: Kim wrote comment below, which you seemed to endorse by pointing to your "jaws of the snake" post
"If the view shared by both IPCC and leading climate realists that the from background noise emergence of many exteme weather phenomena will not occur this century is correct, this type of analysis also has limited value."
"We can't - through observations - conclude that intensity and frequency is definitively increasing. And we also can't say if it is decreasing."
I replied above that we already have observed an increase in SOME forms of extreme weather, but not others such as hurricanes. In this comment, I'm trying to clarify my remarks about confidence intervals and their possible misuse in your "jaws of the snake" post.
Let's consider the temperature projections from two scenarios: RCP4.5 and RCP6.0. Trying decarbonize enough to change from a RCP6.0 to RCP4.5 is a policy choice. When the confidence intervals on those projections overlap (the jaws of the snake are closed), it appears as there will be no statistically significant benefit (lower temperature) from decarbonization. That is wrong. Uncertainty about our planet's climate sensitivity to rising GHGs contributes the confidence interval around each projection. However, our planet does have a fixed climate sensitivity, be it 2, 3, or 4 degC per doubling of CO2 (or the equivalent in other forcing). If climate sensitivity turns out to be 2, both projections will show lower temperatures and if climate sensitivity turns out to be 4, both projections will be higher. In either case there will still be a statistically significant DIFFERENCE in the outcome despite overlapping confidence intervals.
Confidence intervals are intends to show the effect of random noise in data. The possibility that climate sensitivity could be 2,3 or 4 is random noise in projections of RCP4.5 and RCP6.0. However climate sensitivity isn't random noise in projections of the DIFFERENCE BETWEEN RCP4.5 and RCP6.0. Whatever the correct value for our planet's climate sensitivity, it will have the same value in an RCP4.5 future and an RCP6.0 future.
There are other mathematical problems that come from try to interpret the apparent statistical significant of the overlap in confidence intervals that comes from standard deviations adding "in quadrature" (the square root of the sum of the squares), not arithmetically, but let's avoid confusion and focus on the role uncertainty in climate sensitivity plays in the DIFFERENCE between to projections.
The world is presently undershooting RCP4.5, our 2021 paper suggested a trajectory most consistent with SSP2-3.4. For argument purposes, let's just take those two. In terms of global average temperature, it is unlike that we could detect a change from 4.5 to 3.4 this century. I am already on board with the goal of accelrating the rate of decarbonization. However, basing such a goal on promises of detectable changes in climate experienced by people alive today does not seem well supported by climate science.
@Frank I just want to point out that the comment I made shouldn't be interpreted as saying anything about policy decisions. If we know why some potential material effect can't be observed, we have to act on what we can model / predict.
In the field of economics where I've been involved with creating attribution models, we basically don't have any observations we can ever make to validate what we predict. We simply have to make the best models we can and trust them.
Recognizing that emergence takes one hundred years (as an example) isn't a reason not to act in my view. It's a reason TO act - assuming we have other reasons beyond observed data to do so.
I'm sorry I can't reply more intelligently but I'm not quite able to follow the jaws of the snake argument.
But I suspect we're saying the same thing.
Sort of.
Thank you for your great dissection of this paper. Truncating data and failing to address contrary data seems to be typical for the leading climate catastrophe cheerleaders. Their findings of increased economic destruction seems to have missed your outstanding research in why recent storms etc. are more costly.... more stuff to break. Thank you for keeping your writing understandable to those of us who laymen.
Great work as always Roger. I really appreciate how you assist the non scientifically trained ( I'm from a farming background) to better understand how to interpret the deluge of climate papers "proving that the science is settled " ....you can insert your headline of choice...... Thanks
Chris, I am even more lacking in expertise coming from a life in the financial world in the UK.
I subscribe to get full access to the voice of an expert sceptic. Not a denier but a sceptic. I am happy to accept his opinion on these winds, but I then came across a paper published in Weather and Climate Extremes which found that extreme heat in Central and Southern Europe has increased tenfold between 2010 and 2024 compared to the 1961-1990 period.
This backs up findings from the European State of the Climate. 2024 and 2024 Lancet Countdown Europe.
I find it very hard, as a layman, to evaluate the competing claims.
Thanks!
Decision based evidence making strikes once again.
Wouldn’t expect anything else from Piltdown Mann
You write, "Independent validation studies confirm that ERA5 systematically underestimates wind speeds within extratropical cyclone centers, with elevated errors near the storm core and in the Gulf Stream region where nor’easters intensify most explosively."
Is it possible to use ERA5 methods to extrapolate from modern land-based observations (using a subset similar to the observations of the pre-satellite era), to create a more uniform time series?
Great Q - Outside my wheelhouse. A good one for Ryan Maue.
Well done, excellent post. Have you considered how to broaden the reach of your findings beyond us Substack THB nerds? Maybe you can do a joint piece with Jonah Goldberg at The Dispatch.
I recall that Ben Zycher asserted in his recent THB post that as a general rule Climate Scientists are not very good statisticians. That seems to be born out by the sloppy work of the UPenn group.
On your use of AI, it would seem that a full professor at a research university would be using grad students to do the grunt work on data analysis and coding that you appear to have subcontracted to Claude. True or false?
Sabine Hossenfelder touches on the grad/ postdoc students (cheap labour) here:
https://www.youtube.com/watch?v=JvgaZ_myFE4
Thanks
THB seems to get around, way beyond my wildest expectations. Still growing.
In general, sure. But there are also so outstanding statisticians in climate science.
Yes, in the past. I’d guess that is now changing with AI. Maybe more so for folks younger than me and less so for those older!
Michael Mann is the poster -child for sloppy, ignorant and over-confidence in his novel uses of statistics!
Not to mention some pathological personality issues. The man is a real disgrace to science.
Roger, some questions. I am interested in an insider's understanding of how climate science reporting works. Would I be correct that your reanalysis is a type of replication that we have so often been taught is an important part of scientific method, but is not highly rewarded by academia, and so is seldom done?
How might the IPCC's future assessment treat both the paper and and your analysis? Will either or both be cited?
Do you expect science news websites like "Live Science" or "Inside Climate News" to note your analysis? Are climate news websites, or general news outlets like NYT or The Guardian, responsible enough to acknowledge that some specific climate reporting needs to be re-evaluated or retracted on the basis of new analysis?
Does PNAS publish the sort of rebuttal you have made here in your sub stack or otherwise address it?
Modern science journalism is indeed a highly competitive business and I believe this drives a search for more sensational and catastrophic headlines. There are a lot of people who dwell on the negative and on perceived threats because they feel emotions that everyday life denies them.
Just as the regime media were competing against each other to see who could do the most to get Trump removed from office, science media competes to spread the alarmist narrative. The modern journalism space is really no better than the yellow journalism of the Gilded age or the blindly and nasty partisanship of the early 19th century. We forget that newspapers in the 1850's for example regularly portrayed Lincoln as a monkey.
Some replies …
Would I be correct that your reanalysis is a type of replication that we have so often been taught is an important part of scientific method, but is not highly rewarded by academia, and so is seldom done?
RP: Yes, yes, and yes
How might the IPCC's future assessment treat both the paper and and your analysis? Will either or both be cited?
RP: We shall see. My analysis will not be cited at it is not in the peer reviewed literature. See my experience with PNAS on the Grinsted paper. Not worth my efforts to publish.
Do you expect science news websites like "Live Science" or "Inside Climate News" to note your analysis?
RP: Not a chance.
Are climate news websites, or general news outlets like NYT or The Guardian, responsible enough to acknowledge that some specific climate reporting needs to be re-evaluated or retracted on the basis of new analysis?
RP: I’ve seen no evidence of this.
Does PNAS publish the sort of rebuttal you have made here in your sub stack or otherwise address it?
RP: PNAS publishes 500 word comments on published papers. I tried this on Grinsted and was stonewalled. Give who edited this paper not worth my time.
Maybe Andy Revkin can chime in on this.
I would love to hear Andy Revkin's take on this as well.
Enough about Mann on this thread. Let's focus on the science. Many thanks!
“The over-hyping of results and shedding of uncertainty and complexity is a pattern we often see in high-profile climate research, and chronicled in detail over the years here at THB. Loose standards of scientific quality may help to generate breathless headlines, but does not serve improved scientific understandings or public confidence in climate science.”
Strong, long recognized incentives in academia for this phenomenon wouldn’t you agree Roger? Do you think there has been any improvement? Perhaps counter-intuitively I would expect this sort of thing to worsen under a regime that cuts science funding and tries to undermine the standing of scientists in society. Potentially a bit of a vicious circle.
In climate science this phenomenon is indeed incentivized by academic reward structure. Note that Mann's paper was edited by his colleague Jim Kossin whose own work (on TCs) is highly dubious. I don't think the political regime makes much difference to these incentives. Though, I do think public health is different .. Stay tuned on that!
Too bad that the examined data ends 2024
If 2026 were included, it would be a different story!
Was Claude a miner in West Virginia before retraining as a coder? Thanks for finding the time and knowledge to analyze the Chen paper.
Any paper with Michael Mann's imprimatur is and should be highly suspect. His viscous political science and legal reputation is well earned, as is his poor scientific work on the "hockey stick". I have great difficulty not just rejecting his work out of hand.
Mann's track record means he no longer gets the benefit of the doubt. The one minor saving grace here is that at least the Supplementary Analysis seemed more honest.
I strongly suggest that we no longer call MM a “climate scientist.” “Climate activist” is more accurate and less demeaning to real scientists.