Andrew, your call to scrutinise scientific arguments rather than attack scientists is reasonable. However, the article’s weakness is that its explanation of critics’ motives receives more development than its evaluation of their methodological objections.
Financial interests and political affiliations can warrant scrutiny, but they do not determine whether an objection is scientifically correct. The same standard should apply to attribution researchers and their critics.
You do quote Hossenfelder’s objection about event definitions and counterfactual probabilities. However, describing it as “vibes-based science” does little to explain how attribution methods address the concern. The useful questions are whether reasonable alternative definitions change the findings, whether the counterfactual is adequately justified, and whether the conclusions remain robust.
Requesting a specific example is fair. Requiring an example whose definition guarantees a particular result is unnecessarily demanding. A critic could make a valuable contribution by demonstrating sensitivity, bias, inadequate validation or an interpretation that exceeds the evidence.
A comparison with medical and laboratory science offers a constructive benchmark. We can examine causal claims through six overlapping approaches:
Controlled intervention: change a proposed cause and examine the response, recognising the different evidential limits of model experiments and clinical trials.
Observational comparison: compare exposures or conditions while examining confounding.
Counterfactual probability: estimate outcomes with and without the proposed cause.
Conditional mechanism analysis: investigate processes under specified conditions.
Fingerprint or signature evidence: test whether observed patterns distinguish the proposed cause from alternatives.
Individual-case causal assessment: examine the circumstances and competing explanations for a particular case.
These are neither six independent tests nor a validated checklist. Their value depends on their applicability, validation and different sources of error.
Uncertainty is present even when methods agree. In laboratory science, agreement between assays is valuable, but agreement alone does not establish accuracy if they share an interference or calibration error. Attribution analyses can likewise share data, assumptions or model limitations.
The benchmark should therefore be:
Causal confidence increases when applicable, independently validated methods converge on the same defined conclusion, competing explanations are adequately tested, and remaining uncertainty is reported.
The conclusion must also remain precise. Attribution of increased rainfall intensity does not automatically establish the same attribution for flood probability or resulting damage. Each requires additional evidence.
Critics contribute when they expose weaknesses in these connections, even if their broader position is mistaken. Their contribution should be assessed rather than presumed—and their objections should be tested rather than dismissed through an account of their motives.
Your article would be stronger if it showed which objections fail, which identify genuine limitations, and how those limitations affect what the public can reasonably conclude.
Walter, Andrew us super cool. And no one is perfect. Am sure he'd post a tempest more of the yins and yangs of climate attribution ( the yin or yang is the attribution rebuttal by case )
That illustration at the top... AI slop, my eyes are rotting. The only thing that would be worse would be if some human did it, but it's hard to imagine that anyone capable of mastering Photoshop would commit such an abomination.
Hossenfelder: "But making definite statements about the counterfactual probability of a particular event not occurring is, in my opinion, ill-defined mathematical nonsense. There is no amount of equations that will make this fundamental problem go away."
Dessler: 'Wait, what? To me, this argument amounts to "it doesn’t sound right, so it’s wrong.” I call this vibes-based science.'
Hossenfelder's confident claim extreme event attribution is "mathematical nonsense" is the argument from personal incredulity (https://en.wikipedia.org/wiki/Argument_from_incredulity). I, for one, don't think attribution science is nonsense, but rather a first-order estimate from physical models, subject to further refinement. This sounds like the Dunning-Kruger effect at work in a narrowly-credentialed expert, who evinces suspicion of global scientific culture bordering on paranoia.
Did Hossenfelder actually read the NAS report? It stops short of endorsing every fractional attribution of a weather event to climate change, but says models are improving and some attributions, e.g. heat waves, are well-supported. In formal scientific venues, of course, the NAS's implicity authority isn't probative, but it is in US courts. From the Politico article:
'“But for carbon atmospheric pollution, the 2021 Pacific Northwest heat dome would have not occurred,” the [County of Multnomah v. Exxon Mobil Corp.] lawsuit said.
'Those declarative statements, when bolstered by reputable scientists and studies, might move juries, legal experts say. But only if plaintiffs’ lawyers are allowed to make them in court, where judges are empowered to act as scientific gatekeepers. Under the so-called Daubert Standard, judges consider factors like whether a scientific theory has achieved widespread acceptance before allowing it to be used in a trial.
'“The National Academies of Sciences is obviously a very respected institution, so their report would carry sway among judges and others” if it draws stronger links between emissions and some types of natural disasters, said Hill, the former judge who is now an energy fellow at the Council on Foreign Relations think tank.'
So it's Hossenfelder, Koonin, and RPJr. against the NAS. Who would y'all bet on?
We need to capitalize on the Blue Wave and reinvigorate the passion for: (a) understanding that atmospheric carbon causes Climate Change, (b) understanding the humongous “disinformation campaigns” that conservative think tanks and private industry has created to promote DOUBT. Learn more from the 2010 book, “Merchants of Doubt,” by Naomi Oreskes and Erik M. Conway (or catch the 2014 documentary film if pressed for time).
Very nice article. I am disappointed to hear about Hossenfelder, though her ideas often cross into other scientists being unscrupulous and targeting her.
Some reactions.
Some years ago, and perhaps they still do this, the Heritage Foundation had lunchtime seminars that one could attend remotely. I attended a few, and even suggested my students to attend. At least as early as 2020, they were laying out the narrative than global warming is real, but there are still doubts as to the cause. There were two places that they were going to focus on in the future to mine the uncertainties - attribution science and the social cost of carbon.
Of course, one way to discredit the field is to question whether scientists are acting on truly scientific motivations and curiosity. It is an easily exploited myth and stereotype that science is some how pure and its purity is suspect if there is contact with the interests of the non-scientific world. Contact with lawyers is one way to do this. The idea that scientists having contact with lawyers is a sign of corruption, of course, plays on the stereotype that lawyers are intrinsically up to no good. The whole field of forensic meteorology and climatology is a formal interface to the legal community. If climate change is to be addressed then providing climate knowledge to lawyers, politicians, businesses, etc., etc. is necessary.
But ... the uncertainties in the field of event attribution are high. Verification is difficult. There is a demand for attribution products in law, insurance, risk management ... There is motivation to investigate attribution as scientific questions, as well as to provide numbers to customers. It is field where we need extra effort on evaluation, uncertainty descriptions, and development of profession standards (ethics) of practice. There needs to be transparency, boundaries, and checks and balances.
How much confidence would you put in a measuring instrument whose accuracy had not been adequately checked?
Keep that question in mind while reading the National Academies’ report on attributing extreme weather and its damages to climate change. My objection concerns the numbers. There’s quite a distance between “warming affected this event” and “warming made it twenty times more likely and caused $30 billion of the damage.”
That distance needs to be measured, not crossed with a running leap.
The report makes a revealing admission on page 96:
> “Observations are typically insufficient to validate model-based estimates of event probabilities or derived quantities like FAR [fraction of attributable risk] and RR [relative risk], meaning agreement among models is necessary but not sufficient for confidence.”
There’s the first problem, straight from the authors. The quantities they are discussing include the risk ratio: how likely an event is in the actual climate divided by how likely it would be in a hypothetical climate without human influence.
Suppose a model does a respectable job reproducing today’s weather statistics. Fine. That’s a necessary test. But how do we know it correctly calculates the DIFFERENCE between today’s climate and that hypothetical world?
Getting one world approximately right doesn’t establish that you’ve correctly calculated the difference between two worlds.
And running the model another thousand times won’t settle that question. It may tell you much more precisely what the model thinks. Whether the model is right remains a separate inquiry.
This gets troublesome out in the far tails of the distributions—the territory occupied by exceptionally rare events. The report acknowledges widely differing attribution estimates for the 2021 Pacific Northwest heatwave. When the estimated counterfactual probability is tiny, small changes in assumptions can produce enormous changes in the risk ratio.
For a simple illustration, suppose the actual-world probability stays fixed. Change the hypothetical-world probability from one in ten thousand to one in a million, and the risk ratio increases a hundredfold.
Same actual-world probability. Hundredfold difference in the headline multiplier.
I’d want a very good reason for choosing that denominator.
Then there’s the weather itself. The report acknowledges uncertainty about how human influence changes circulation, blocking patterns, and other dynamical processes. Some attribution experiments work around that difficulty by holding circulation approximately fixed and changing the background climate.
That can be a useful experiment. It asks, roughly, “With this weather pattern in place, how much difference did the warming make?”
But what happened to the probability of getting that weather pattern in the first place?
If we haven’t answered that, we haven’t calculated the whole change in event probability. We’ve calculated a conditional effect. Nothing wrong with that—provided the condition doesn’t disappear when the result reaches the headline.
Finally, we arrive at the damage bill.
The report deserves credit for rejecting a tempting shortcut: multiplying an individual event’s losses by its fraction of attributable risk. Making a rainfall threshold more likely does not tell us what fraction of the resulting flood damage warming caused.
To calculate that, we need to follow the water. Where did it fall? How did it drain? Which buildings flooded? How deep was the water? What would have happened with somewhat less rainfall?
Imagine a levee. A small increase in flood height might make little difference while the water remains below the top. Another inch could send it over.
Good luck calculating that damage from the rainfall probability ratio alone.
More elaborate impact models attempt to follow those steps, but the report acknowledges weaknesses in response functions, local data, and the tracking of uncertainty through the calculations. A credible estimate of additional rainfall cannot magically validate the subsequent flood model, damage model, or mortality estimate.
Each needs its own checking.
That is where I think the report falls short. It identifies many of these difficulties, sometimes quite plainly, then discusses increasingly consequential uses without consistently establishing which numerical estimates are accurate enough for which purposes.
I’m not arguing that warming cannot affect extreme weather. Nor am I claiming every attribution estimate is wrong.
I’m asking for VALIDATION before attaching that much confidence—and that many dollars—to the answer.
A measuring instrument doesn’t become accurate because we’ve found an urgent use for it.
And after complaining that opponents are engaging in "personal attacks on scientists" you are doing the exact same thing when you say
"Roger Pielke Jr. is affiliated with the American Enterprise Institute, another right-wing Washington think tank."
SO WHAT! Either Pielke is right or wrong, regardless of who he is "affiliated with". That's just a nasty personal attack, the very thing you accuse others of.
Andrews article and comments here are so similar to the establishment corporate democrats against the New blue wave progressives that just got elected. Scientists deserve to be respected but not at the expense of an educated public that has certainly done plenty of research that establishment scientists refuse to look more closely at.
There are extreme events that began in the late 1950s and still to this date continue.
There is an untold story of how an engineered water cycle
has hot-wired the Polar Climate System and led to Pan Arctic regions collapse and global climate warming. Data sets that NoAA has been using to determine thermal changes in the arctic regions is not taking into consideration of what has happened on the ground there over the past 75 years or so
Again I say it is time to look into existing research, beyond just the establishment confirmed scientific journals. On the ground, the water cycle in the northern Latitudes of the Northern Hemisphere ,has been seriously stripped away. There are approx. 20 major former rivers , some of the largest of the worlds systems, that stretch thru Siberia all the way to Labrador, These rivers no longer offer the traditional historical flow patterns, or historical thermal profiles, and even traditional flow volumes that the Arctic region has had for centuries before the 1950s.
In addition, Internationally most of the major worlds rivers are now disrupted, dammed ,and fragmented again altering their historical flow patterns, thermal Profiles, and volumes that reach into nearest oceans.
We again might expect our author to discount and criticize my comments and why is it so difficult for these establishment Folks to do the deeper dive required? And continuing to take the easiest way out, with information that is no longer relevant to the speed at which our climate is changing.
If you care to understand all the elements that have been left out of the full story of why the planet is heating so rapidly. ....happy to share our inexperience with anyone here. or I can do it privately with detailed graphs and errors that Noaa has made
The fossil fuel industry pumps roughly 37 billion metric tons of industrial carbon into the shared atmosphere every year, trapping extra heat equivalent to hundreds of thousands of gigawatts of continuous energy. This is not a political abstraction. It is a brute-force thermodynamic engine operated by corporate monopolies that privatize financial gains while forcing the public commons to absorb the physical shock.
Extreme event attribution science strips away their plausible deniability. By running open-source climate models that map the exact thermal fingerprint of human emissions against localized rainfall anomalies and heat domes, physicists are finally quantifying the precise debt these companies owe. When commentators abandon physical mechanics to complain about vibes and lawyer affiliations, they provide free cover fire for a business model that takes more than it gives.
The victims of these shifted probabilities are real bodies experiencing measurable trauma. They are human communities wiped out by sudden atmospheric rivers, and they are millions of wild animals that feel pain and fear, suffocating in 100,000-acre mega-fires because the air has been engineered to hold more moisture and heat.
We cannot allow PR campaigns to obscure basic physics. Defending the hard, verifiable math of extreme weather attribution is the necessary first step toward reclaiming our planetary boundaries from the executives who profit from their destabilization. The science is sound; now we must use it to hold the owners of the carbon infrastructure accountable.
It is a fact that World Weather Attribution was created to provide real time material for climate reporters and activists and evidence for litigation purposes. They have also been criticized quite reasonably for juicing their headline findings by using thresholds and relative risk as the basis for their conclusions. I think Otto was being transparent here: “Otto and others disagree with Trenberth; the statistics of likelihood, they argue, are *what the public wants to hear*.” (my emphasis)
Thank you for the article. I agree that talking to lawyers does not invalidate a scientist’s work, and that criticism should address methods and evidence. But the event-definition objection you quote is a methodological argument, and dismissing it as “vibes” leaves the central question unanswered.
The National Academies itself discusses this issue: robust attribution should examine whether conclusions remain qualitatively similar across reasonable event definitions, while recognizing that numerical estimates will differ. It also identifies event-selection bias as a concern. These are questions within attribution science, not reasons to reject the field.
Hossenfelder’s conclusion that attribution is inherently “ill-defined mathematical nonsense” goes too far. A probability can be well defined for a specified class of events—for example, a regional three-day temperature average exceeding a stated threshold. That differs from assigning a probability to every detail of one unique weather sequence.
But asking critics to demonstrate a definition that *guarantees* a particular result sets the wrong standard. A result can warrant criticism because it is sensitive to reasonable definitions, relies on inadequately validated models, or understates uncertainty. None of those requires proof of deliberate manipulation.
Your case would be stronger with a worked example showing how an attribution result responds to alternative event definitions, why its counterfactual is defensible, and how its uncertainty is assessed. That would answer the objection directly.
Finally, the instruction to stop judging people by their affiliations should apply consistently. Critics’ employers and scientists’ relationships with lawyers may warrant disclosure, but neither settles the science.
Attribution is a legitimate research framework. Its credibility is strengthened by explaining its assumptions and limits—and distinguishing criticism of a particular estimate from rejection of the entire field.
Hossenfelder might have a legitimate argument. I just wish she'd spent less time on conspiracy theories and more time fleshing the argument out. And then we can hash it out. That's how science works.
Pielke has made some detailed criticisms of this line of research. I know he has a lot of baggage but I think it would be good for someone to directly engage with his criticisms rather than just dismissing him because he’s associated with a right wing think tank. Because, as you say:
Thanks, Andrew. I agree that speculation about motives doesn’t resolve the science. My point was that the event-definition concern deserves a substantive answer, even if her dismissal of the entire field goes too far.
A worked example showing how an attribution estimate changes across reasonable choices of area, duration, and threshold—and which conclusions remain robust—would be a useful follow-up. I’d be interested in reading that. Thanks!
OP: "Roger Pielke Jr. is affiliated with the American Enterprise Institute, another right-wing Washington think tank."
Yep! The Politico item discusses RPJr in more detail:
'The obscure [NAS] panel has faced online criticism from the American Enterprise Institute, a right-leaning think tank, and other groups bankrolled by the fossil fuel industry. On the day of its first meeting in 2024, AEI scholar Roger Pielke Jr. published a Substack post accusing the National Academies of engaging in “stealth advocacy in support of climate litigation.”
'AEI spokesperson Brady Africk said its scholars’ views “are drawn from their own expertise and research, and we protect their ability to work independently.” Pielke told POLITICO’s E&E News that he was calling “things like I see them.”'
"Honest broker", my ass. It's clear he didn't help the AEI-inspired Republican Party much on Tuesday. Into the dustbin of history with him!
Occam had a duck named Razor. When it wandered off and someone called him to say they found it he asked the caller, "Does it waddle? Does it quack? Does it look like my little Razor?". When the caller said, "yes" Occam got his duck back.
If everything climate is on steroids you know it's climate change because the climate has never behaved like this everywhere, everything, all at once before.
I have my finger on the block button for anyone to misrepresents what I said. Read my post carefully and think before posting.
Hossenfelder's brand is contrarianism and negativity. Sometimes that's useful in science. Often it's not. But it's usually good for clicks.
Andrew, your call to scrutinise scientific arguments rather than attack scientists is reasonable. However, the article’s weakness is that its explanation of critics’ motives receives more development than its evaluation of their methodological objections.
Financial interests and political affiliations can warrant scrutiny, but they do not determine whether an objection is scientifically correct. The same standard should apply to attribution researchers and their critics.
You do quote Hossenfelder’s objection about event definitions and counterfactual probabilities. However, describing it as “vibes-based science” does little to explain how attribution methods address the concern. The useful questions are whether reasonable alternative definitions change the findings, whether the counterfactual is adequately justified, and whether the conclusions remain robust.
Requesting a specific example is fair. Requiring an example whose definition guarantees a particular result is unnecessarily demanding. A critic could make a valuable contribution by demonstrating sensitivity, bias, inadequate validation or an interpretation that exceeds the evidence.
A comparison with medical and laboratory science offers a constructive benchmark. We can examine causal claims through six overlapping approaches:
Controlled intervention: change a proposed cause and examine the response, recognising the different evidential limits of model experiments and clinical trials.
Observational comparison: compare exposures or conditions while examining confounding.
Counterfactual probability: estimate outcomes with and without the proposed cause.
Conditional mechanism analysis: investigate processes under specified conditions.
Fingerprint or signature evidence: test whether observed patterns distinguish the proposed cause from alternatives.
Individual-case causal assessment: examine the circumstances and competing explanations for a particular case.
These are neither six independent tests nor a validated checklist. Their value depends on their applicability, validation and different sources of error.
Uncertainty is present even when methods agree. In laboratory science, agreement between assays is valuable, but agreement alone does not establish accuracy if they share an interference or calibration error. Attribution analyses can likewise share data, assumptions or model limitations.
The benchmark should therefore be:
Causal confidence increases when applicable, independently validated methods converge on the same defined conclusion, competing explanations are adequately tested, and remaining uncertainty is reported.
The conclusion must also remain precise. Attribution of increased rainfall intensity does not automatically establish the same attribution for flood probability or resulting damage. Each requires additional evidence.
Critics contribute when they expose weaknesses in these connections, even if their broader position is mistaken. Their contribution should be assessed rather than presumed—and their objections should be tested rather than dismissed through an account of their motives.
Your article would be stronger if it showed which objections fail, which identify genuine limitations, and how those limitations affect what the public can reasonably conclude.
Walter, Andrew us super cool. And no one is perfect. Am sure he'd post a tempest more of the yins and yangs of climate attribution ( the yin or yang is the attribution rebuttal by case )
That illustration at the top... AI slop, my eyes are rotting. The only thing that would be worse would be if some human did it, but it's hard to imagine that anyone capable of mastering Photoshop would commit such an abomination.
Hossenfelder: "But making definite statements about the counterfactual probability of a particular event not occurring is, in my opinion, ill-defined mathematical nonsense. There is no amount of equations that will make this fundamental problem go away."
Dessler: 'Wait, what? To me, this argument amounts to "it doesn’t sound right, so it’s wrong.” I call this vibes-based science.'
Hossenfelder's confident claim extreme event attribution is "mathematical nonsense" is the argument from personal incredulity (https://en.wikipedia.org/wiki/Argument_from_incredulity). I, for one, don't think attribution science is nonsense, but rather a first-order estimate from physical models, subject to further refinement. This sounds like the Dunning-Kruger effect at work in a narrowly-credentialed expert, who evinces suspicion of global scientific culture bordering on paranoia.
Did Hossenfelder actually read the NAS report? It stops short of endorsing every fractional attribution of a weather event to climate change, but says models are improving and some attributions, e.g. heat waves, are well-supported. In formal scientific venues, of course, the NAS's implicity authority isn't probative, but it is in US courts. From the Politico article:
'“But for carbon atmospheric pollution, the 2021 Pacific Northwest heat dome would have not occurred,” the [County of Multnomah v. Exxon Mobil Corp.] lawsuit said.
'Those declarative statements, when bolstered by reputable scientists and studies, might move juries, legal experts say. But only if plaintiffs’ lawyers are allowed to make them in court, where judges are empowered to act as scientific gatekeepers. Under the so-called Daubert Standard, judges consider factors like whether a scientific theory has achieved widespread acceptance before allowing it to be used in a trial.
'“The National Academies of Sciences is obviously a very respected institution, so their report would carry sway among judges and others” if it draws stronger links between emissions and some types of natural disasters, said Hill, the former judge who is now an energy fellow at the Council on Foreign Relations think tank.'
So it's Hossenfelder, Koonin, and RPJr. against the NAS. Who would y'all bet on?
We need to capitalize on the Blue Wave and reinvigorate the passion for: (a) understanding that atmospheric carbon causes Climate Change, (b) understanding the humongous “disinformation campaigns” that conservative think tanks and private industry has created to promote DOUBT. Learn more from the 2010 book, “Merchants of Doubt,” by Naomi Oreskes and Erik M. Conway (or catch the 2014 documentary film if pressed for time).
Very nice article. I am disappointed to hear about Hossenfelder, though her ideas often cross into other scientists being unscrupulous and targeting her.
Some reactions.
Some years ago, and perhaps they still do this, the Heritage Foundation had lunchtime seminars that one could attend remotely. I attended a few, and even suggested my students to attend. At least as early as 2020, they were laying out the narrative than global warming is real, but there are still doubts as to the cause. There were two places that they were going to focus on in the future to mine the uncertainties - attribution science and the social cost of carbon.
Of course, one way to discredit the field is to question whether scientists are acting on truly scientific motivations and curiosity. It is an easily exploited myth and stereotype that science is some how pure and its purity is suspect if there is contact with the interests of the non-scientific world. Contact with lawyers is one way to do this. The idea that scientists having contact with lawyers is a sign of corruption, of course, plays on the stereotype that lawyers are intrinsically up to no good. The whole field of forensic meteorology and climatology is a formal interface to the legal community. If climate change is to be addressed then providing climate knowledge to lawyers, politicians, businesses, etc., etc. is necessary.
But ... the uncertainties in the field of event attribution are high. Verification is difficult. There is a demand for attribution products in law, insurance, risk management ... There is motivation to investigate attribution as scientific questions, as well as to provide numbers to customers. It is field where we need extra effort on evaluation, uncertainty descriptions, and development of profession standards (ethics) of practice. There needs to be transparency, boundaries, and checks and balances.
Good article. The only reason for quoting randos on Reddit should be to debunk them.
Here's my analysis of the NASEM report …
How much confidence would you put in a measuring instrument whose accuracy had not been adequately checked?
Keep that question in mind while reading the National Academies’ report on attributing extreme weather and its damages to climate change. My objection concerns the numbers. There’s quite a distance between “warming affected this event” and “warming made it twenty times more likely and caused $30 billion of the damage.”
That distance needs to be measured, not crossed with a running leap.
The report makes a revealing admission on page 96:
> “Observations are typically insufficient to validate model-based estimates of event probabilities or derived quantities like FAR [fraction of attributable risk] and RR [relative risk], meaning agreement among models is necessary but not sufficient for confidence.”
There’s the first problem, straight from the authors. The quantities they are discussing include the risk ratio: how likely an event is in the actual climate divided by how likely it would be in a hypothetical climate without human influence.
Suppose a model does a respectable job reproducing today’s weather statistics. Fine. That’s a necessary test. But how do we know it correctly calculates the DIFFERENCE between today’s climate and that hypothetical world?
Getting one world approximately right doesn’t establish that you’ve correctly calculated the difference between two worlds.
And running the model another thousand times won’t settle that question. It may tell you much more precisely what the model thinks. Whether the model is right remains a separate inquiry.
This gets troublesome out in the far tails of the distributions—the territory occupied by exceptionally rare events. The report acknowledges widely differing attribution estimates for the 2021 Pacific Northwest heatwave. When the estimated counterfactual probability is tiny, small changes in assumptions can produce enormous changes in the risk ratio.
For a simple illustration, suppose the actual-world probability stays fixed. Change the hypothetical-world probability from one in ten thousand to one in a million, and the risk ratio increases a hundredfold.
Same actual-world probability. Hundredfold difference in the headline multiplier.
I’d want a very good reason for choosing that denominator.
Then there’s the weather itself. The report acknowledges uncertainty about how human influence changes circulation, blocking patterns, and other dynamical processes. Some attribution experiments work around that difficulty by holding circulation approximately fixed and changing the background climate.
That can be a useful experiment. It asks, roughly, “With this weather pattern in place, how much difference did the warming make?”
But what happened to the probability of getting that weather pattern in the first place?
If we haven’t answered that, we haven’t calculated the whole change in event probability. We’ve calculated a conditional effect. Nothing wrong with that—provided the condition doesn’t disappear when the result reaches the headline.
Finally, we arrive at the damage bill.
The report deserves credit for rejecting a tempting shortcut: multiplying an individual event’s losses by its fraction of attributable risk. Making a rainfall threshold more likely does not tell us what fraction of the resulting flood damage warming caused.
To calculate that, we need to follow the water. Where did it fall? How did it drain? Which buildings flooded? How deep was the water? What would have happened with somewhat less rainfall?
Imagine a levee. A small increase in flood height might make little difference while the water remains below the top. Another inch could send it over.
Good luck calculating that damage from the rainfall probability ratio alone.
More elaborate impact models attempt to follow those steps, but the report acknowledges weaknesses in response functions, local data, and the tracking of uncertainty through the calculations. A credible estimate of additional rainfall cannot magically validate the subsequent flood model, damage model, or mortality estimate.
Each needs its own checking.
That is where I think the report falls short. It identifies many of these difficulties, sometimes quite plainly, then discusses increasingly consequential uses without consistently establishing which numerical estimates are accurate enough for which purposes.
I’m not arguing that warming cannot affect extreme weather. Nor am I claiming every attribution estimate is wrong.
I’m asking for VALIDATION before attaching that much confidence—and that many dollars—to the answer.
A measuring instrument doesn’t become accurate because we’ve found an urgent use for it.
Regards to all,
w.
Here is a scientific analysis of the NASEM study that, curiously, you seem to have overlooked.
https://rogerpielkejr.substack.com/p/climate-change-is-making-us-stupid
Here is another, again strangely missing.
https://www.realclearscience.com/articles/2026/10/10/will_the_supreme_court_climate_case_expose_the_ethical_shortfalls_at_the_national_academies_1210871.html
And after complaining that opponents are engaging in "personal attacks on scientists" you are doing the exact same thing when you say
"Roger Pielke Jr. is affiliated with the American Enterprise Institute, another right-wing Washington think tank."
SO WHAT! Either Pielke is right or wrong, regardless of who he is "affiliated with". That's just a nasty personal attack, the very thing you accuse others of.
w.
Andrews article and comments here are so similar to the establishment corporate democrats against the New blue wave progressives that just got elected. Scientists deserve to be respected but not at the expense of an educated public that has certainly done plenty of research that establishment scientists refuse to look more closely at.
There are extreme events that began in the late 1950s and still to this date continue.
There is an untold story of how an engineered water cycle
has hot-wired the Polar Climate System and led to Pan Arctic regions collapse and global climate warming. Data sets that NoAA has been using to determine thermal changes in the arctic regions is not taking into consideration of what has happened on the ground there over the past 75 years or so
Again I say it is time to look into existing research, beyond just the establishment confirmed scientific journals. On the ground, the water cycle in the northern Latitudes of the Northern Hemisphere ,has been seriously stripped away. There are approx. 20 major former rivers , some of the largest of the worlds systems, that stretch thru Siberia all the way to Labrador, These rivers no longer offer the traditional historical flow patterns, or historical thermal profiles, and even traditional flow volumes that the Arctic region has had for centuries before the 1950s.
In addition, Internationally most of the major worlds rivers are now disrupted, dammed ,and fragmented again altering their historical flow patterns, thermal Profiles, and volumes that reach into nearest oceans.
We again might expect our author to discount and criticize my comments and why is it so difficult for these establishment Folks to do the deeper dive required? And continuing to take the easiest way out, with information that is no longer relevant to the speed at which our climate is changing.
If you care to understand all the elements that have been left out of the full story of why the planet is heating so rapidly. ....happy to share our inexperience with anyone here. or I can do it privately with detailed graphs and errors that Noaa has made
Are you you the climate messiah who yields knowledge of the otherworld and gonna save us?
The fossil fuel industry pumps roughly 37 billion metric tons of industrial carbon into the shared atmosphere every year, trapping extra heat equivalent to hundreds of thousands of gigawatts of continuous energy. This is not a political abstraction. It is a brute-force thermodynamic engine operated by corporate monopolies that privatize financial gains while forcing the public commons to absorb the physical shock.
Extreme event attribution science strips away their plausible deniability. By running open-source climate models that map the exact thermal fingerprint of human emissions against localized rainfall anomalies and heat domes, physicists are finally quantifying the precise debt these companies owe. When commentators abandon physical mechanics to complain about vibes and lawyer affiliations, they provide free cover fire for a business model that takes more than it gives.
The victims of these shifted probabilities are real bodies experiencing measurable trauma. They are human communities wiped out by sudden atmospheric rivers, and they are millions of wild animals that feel pain and fear, suffocating in 100,000-acre mega-fires because the air has been engineered to hold more moisture and heat.
We cannot allow PR campaigns to obscure basic physics. Defending the hard, verifiable math of extreme weather attribution is the necessary first step toward reclaiming our planetary boundaries from the executives who profit from their destabilization. The science is sound; now we must use it to hold the owners of the carbon infrastructure accountable.
It is a fact that World Weather Attribution was created to provide real time material for climate reporters and activists and evidence for litigation purposes. They have also been criticized quite reasonably for juicing their headline findings by using thresholds and relative risk as the basis for their conclusions. I think Otto was being transparent here: “Otto and others disagree with Trenberth; the statistics of likelihood, they argue, are *what the public wants to hear*.” (my emphasis)
https://e360.yale.edu/features/pinning-wild-weather-on-climate-change-scientists-are-upping-their-game
Some skepticism seems warranted and healthy.
Thank you for the article. I agree that talking to lawyers does not invalidate a scientist’s work, and that criticism should address methods and evidence. But the event-definition objection you quote is a methodological argument, and dismissing it as “vibes” leaves the central question unanswered.
The National Academies itself discusses this issue: robust attribution should examine whether conclusions remain qualitatively similar across reasonable event definitions, while recognizing that numerical estimates will differ. It also identifies event-selection bias as a concern. These are questions within attribution science, not reasons to reject the field.
https://www.nationalacademies.org/read/21852/chapter/4
Hossenfelder’s conclusion that attribution is inherently “ill-defined mathematical nonsense” goes too far. A probability can be well defined for a specified class of events—for example, a regional three-day temperature average exceeding a stated threshold. That differs from assigning a probability to every detail of one unique weather sequence.
But asking critics to demonstrate a definition that *guarantees* a particular result sets the wrong standard. A result can warrant criticism because it is sensitive to reasonable definitions, relies on inadequately validated models, or understates uncertainty. None of those requires proof of deliberate manipulation.
Your case would be stronger with a worked example showing how an attribution result responds to alternative event definitions, why its counterfactual is defensible, and how its uncertainty is assessed. That would answer the objection directly.
Finally, the instruction to stop judging people by their affiliations should apply consistently. Critics’ employers and scientists’ relationships with lawyers may warrant disclosure, but neither settles the science.
Attribution is a legitimate research framework. Its credibility is strengthened by explaining its assumptions and limits—and distinguishing criticism of a particular estimate from rejection of the entire field.
Thanks - Mark Morrissey
Hossenfelder might have a legitimate argument. I just wish she'd spent less time on conspiracy theories and more time fleshing the argument out. And then we can hash it out. That's how science works.
“
Pielke has made some detailed criticisms of this line of research. I know he has a lot of baggage but I think it would be good for someone to directly engage with his criticisms rather than just dismissing him because he’s associated with a right wing think tank. Because, as you say:
That's how science works.”
I have not seem them. Is there a particularly strong one you would recommend?
Here you go Andrew:
https://rogerpielkejr.substack.com/p/a-28-trillion-mistake
Have at it.
Thanks, Andrew. I agree that speculation about motives doesn’t resolve the science. My point was that the event-definition concern deserves a substantive answer, even if her dismissal of the entire field goes too far.
A worked example showing how an attribution estimate changes across reasonable choices of area, duration, and threshold—and which conclusions remain robust—would be a useful follow-up. I’d be interested in reading that. Thanks!
OP: "Roger Pielke Jr. is affiliated with the American Enterprise Institute, another right-wing Washington think tank."
Yep! The Politico item discusses RPJr in more detail:
'The obscure [NAS] panel has faced online criticism from the American Enterprise Institute, a right-leaning think tank, and other groups bankrolled by the fossil fuel industry. On the day of its first meeting in 2024, AEI scholar Roger Pielke Jr. published a Substack post accusing the National Academies of engaging in “stealth advocacy in support of climate litigation.”
'AEI spokesperson Brady Africk said its scholars’ views “are drawn from their own expertise and research, and we protect their ability to work independently.” Pielke told POLITICO’s E&E News that he was calling “things like I see them.”'
"Honest broker", my ass. It's clear he didn't help the AEI-inspired Republican Party much on Tuesday. Into the dustbin of history with him!
Occam had a duck named Razor. When it wandered off and someone called him to say they found it he asked the caller, "Does it waddle? Does it quack? Does it look like my little Razor?". When the caller said, "yes" Occam got his duck back.
If everything climate is on steroids you know it's climate change because the climate has never behaved like this everywhere, everything, all at once before.